Proceedings of the meeting - Department of Physics - University of ...

ph.surrey.ac.uk

Proceedings of the meeting - Department of Physics - University of ...

_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR=^ìÖìëíI=OMMSïïïKéÜKëìêêÉóK~ÅKìâLÄÅ|áëãêãMS


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMS^=åçíÉ=Ñêçã=íÜÉ=`Ü~áê=çÑ=íÜÉ=içÅ~ä=lêÖ~åáëáåÖ=`çããáííÉÉqÜáë=ÄççâäÉí=Åçåí~áåë=íÜÉ=éìÄäáëÜÉÇ=éêçÅÉÉÇáåÖë=çÑ=íÜÉ=_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjoj=~ååì~ä=ãÉÉíáåÖ=ÜÉäÇ=É~êäáÉê=íÜáë=óÉ~ê=áå=^ìÖìëíKaÉëéáíÉ=íÜÉ=ãçîÉ=íç=~=ä~íÉ=^ìÖìëí=Ç~íÉ=ã~ÇÉ=åÉÅÉëë~êó=Äó=íÜÉ=råáîÉêëáíó=çÑ=pìêêÉóÛë=ìåìëì~ä=ëÉãÉëíÉê=ëóëíÉãI=f=~ã=ÇÉäáÖÜíÉÇ=íç=ë~ó=ïÉ=Ü~Ç=~å=ÉñÅÉääÉåí=ãÉÉíáåÖI=ïÜáÅÜ=ï~ë=ÉñíêÉãÉäó=ïÉää=ëìééçêíÉÇ=Äó=íÜÉ=rh=jo=ÅçããìåáíóK=få=íÜÉ=ÉåÇI=íÜÉêÉ=ïÉêÉ==RV=ëìÄãáííÉÇ=~Äëíê~Åíë=áå=~ÇÇáíáçå=íç=íÜÉ=U=áåîáíÉÇ=äÉÅíìêÉëKqÜÉ=ãÉÉíáåÖ=ï~ë= ÖÉåÉêçìëäó=ëìééçêíÉÇ=Äó=çìê=áåÇìëíêá~ä=ëéçåëçêëI=íç=ïÜçã=f=~ã=áããÉåëÉäó=Öê~íÉÑìäK=qÜÉáê=Ççå~íáçåë=~ääçïÉÇ=ìë=íç=ÅçåíáåìÉ=íÜÉ=ëìÅÅÉëëÑìä=ëíìÇÉåí=ëíáéÉåÇ=éêçÖê~ããÉ=~ë=ïÉää=~ë=ÇÉÑê~ó=íÜÉ=ÉñéÉåëÉë=çÑ=çíÜÉê=é~êíë=çÑ=íÜÉ=ãÉÉíáåÖK=qÜÉ=ÅçåÑÉêÉåÅÉ=ÇáååÉê=~í aÉåÄáÉë táåÉ=bëí~íÉ=áå=açêâáåÖ=ï~ë=ãÉãçê~ÄäÉW=Ñçê=íÜÉ=äçÅ~íáçåI=íÜÉ=íçìêë=~åÇ=íÜÉ=ïáåÉ=í~ëíáåÖI=íÜÉ=Ñ~Äìäçìë=ÄáÖ=Ä~åÇ=Á=~åÇ=åçí=äÉ~ëí=Ñçê=íÜÉ=~ÇÇÉÇ=E~åÇ=ìåéä~ååÉÇF=ÉåíÉêí~áåãÉåí=çÑ=ï~íÅÜáåÖ=íÜêÉÉ=í~ää=Åç~ÅÜÉë=ëèìÉÉòáåÖ=ìåÇÉê=~=îÉêó=äçï=ê~áäï~ó=ÄêáÇÖÉ>qÜáë=óÉ~ê=ïÉ=ÅçåíáåìÉÇ=íÜÉ=éçéìä~ê=íê~Çáíáçå=çÑ=ÜçäÇáåÖ=~=ëáåÖäÉJëÉëëáçå=ãÉÉíáåÖI=ïáíÜ=~=ë~íÉääáíÉ=ïçêâëÜçé=çå=píÉã=`Éää=fã~ÖáåÖI=îÉêó=Å~é~Ääó=çêÖ~åáëÉÇ=Äó=jáâÉ=jçÇçI=ïÜáÅÜ=êÉéêÉëÉåíÉÇ=~=íáãÉäó=ÅçåíêáÄìíáçå=íç=~å=ÉãÉêÖáåÖ=ÑáÉäÇKqÜÉ=éêçéçë~ä=íÜ~í=qÜáë=eçìëÉ=ÄÉäáÉîÉë=íÜ~í=jo=ëÜçìäÇ=ëíçé=íêóáåÖ=íç=ÄÉ=mbqÒ=ÖÉåÉê~íÉÇ=ÄçíÜ=áåÑçêãÉÇ=~åÇ=îÉêó=äáîÉäó=ÇáëÅìëëáçåI=ïáíÜ=~=é~êíáÅìä~êäó=ÉåíÉêí~áåáåÖ=ÅçåíêáÄìíáçå=Ñêçã=j~êâ=ióíÜÖçÉI=ëéÉ~âáåÖ=íç=ìë=Ñêçã=íÜÉ=Ç~êâ=ëáÇÉÒK^åçíÜÉê=ÜáÖÜäáÖÜí=ï~ë=íÜÉ=ëéÉÅá~ä=áåîáíÉÇ=äÉÅíìêÉI=ÇÉäáîÉêÉÇ=Äó=píÉéÜÉå hÉÉîáäI=çå=~å=br=ÇáêÉÅíáîÉ=íÜ~íI=áÑ=áãéäÉãÉåíÉÇI=ïçìäÇ=ÉÑÑÉÅíáîÉäó=Ä~å=áåíÉêîÉåíáçå~ä=jof=éêçÅÉÇìêÉë=áå=íÜÉ=rhK=qÜáë=éêçîÉÇ=~=Ñ~ëÅáå~íáåÖ=áåëáÖÜí=áåíç=íÜÉ=~ÅíáîáíáÉë=çÑ=çìê äÉÖáëä~íáîÉ=~åÇ=êÉÖìä~íçêó=ëóëíÉãK==qÜáë=~Äëíê~Åí=Äççâ=Åçåí~áåë=íÜÉ=ã~àçêáíó=çÑ=íÜÉ=~Äëíê~Åíë=éêÉëÉåíÉÇK=få=~=ÑÉï=Å~ëÉëI=~ìíÜçêë=Ñçê=î~êáçìë=êÉ~ëçåë=ÉäÉÅíÉÇ=åçí=íç=ã~âÉ=íÜÉáê=ïçêâ=~î~áä~ÄäÉ=íç=íÜÉ=ÖÉåÉê~ä=éìÄäáÅK=_ó=íÜÉ=ë~ãÉ=íçâÉåI=óçì=ïáää=ÑáåÇ=çå=íÜÉ=_êáíáëÜ=`Ü~éíÉê=ïÉÄëáíÉ=íÜÉ=ã~àçêáíó=çÑ=íÜÉ=í~äâë=éêÉëÉåíÉÇ=~í=íÜÉ=ãÉÉíáåÖI=ã~áåäó=áå=íÜÉ=Ñçêã=çÑ=jáÅêçëçÑí=mçïÉêéçáåí=ÑáäÉëK=^ää=íÜÉ=ã~íÉêá~ä=ã~ó=ÄÉ=ÑêÉÉäó=Ççïåäç~ÇÉÇ=Ñçê=íÜÉ=íÜÉ=éìêéçëÉë=çÑ=éÉêëçå~ä=ëíìÇóK eçïÉîÉêI=éäÉ~ëÉ=åçíÉ=íÜ~í=íÜÉ=ÅçéóêáÖÜí=çÑ=~ää=áíÉãë=êÉã~áåë=ïáíÜ=íÜÉ=êÉëéÉÅíáîÉ=~ìíÜçêë=çÑ=íÜÉ=î~êáçìë=ÅçåíêáÄìíáçåëK=vçì=~êÉ=íÜÉêÉÑçêÉ=âáåÇäó=êÉèìÉëíÉÇ=åçí=íç=êÉJìëÉ=çê=êÉíê~åëãáí=~åó=çÑ=íÜÉ=ã~íÉêá~ä=ïáíÜçìí=íÜÉáê=éêáçê=éÉêãáëëáçåKpáãçå=g=açê~åïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMSpíìÇÉåí=ëíáéÉåÇ=êÉÅáéáÉåíëeÉáÇá=_çååáÅáf~å=`~îáåbäÉ~åçê=`çñj~åháå=`ÜçóhÉååÉíÜ=`ÜÉìåÖp~ääó=bäÇÉÖÜ~áÇó`êáëíáå~=d~ÄÉääáÉêám~å~Öáçíáë=hóêí~íçëf~å=j~äçåÉh~êÉå=jìääáåÖÉêsáêÖáåá~=kÉïÅçãÄÉgç~ç=mÉêÉáê~jáÅÜ~Éä=mççäÉoçë~=p~åÅÜÉòj~êáÉ=pÅÜêçÉÇÉêa~åáÉä=píìÅâäÉóvìÉÜìá=q~çoçã~å=tÉëçäçïëâámÉíÉê=têáÖÜíråáîÉêëáíó=çÑ=bÇáåÄìêÖÜråáîÉêëáíó=çÑ=kçííáåÖÜ~ãråáîÉêëáíó=çÑ=kçííáåÖÜ~ãfåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜråáîÉêëáíó=`çääÉÖÉ=içåÇçåråáîÉêëáíó=çÑ=kçííáåÖÜ~ãfåëíáíìíÉ=çÑ=`~åÅÉê=oÉëÉ~êÅÜfåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜråáîÉêëáíó=çÑ=`~ãÄêáÇÖÉråáîÉêëáíó=çÑ=kçííáåÖÜ~ãråáîÉêëáíó=çÑ=`~ãÄêáÇÖÉråáîÉêëáíó=çÑ=`~ãÄêáÇÖÉråáîÉêëáíó=çÑ=kçííáåÖÜ~ãråáîÉêëáíó=çÑ=kçííáåÖÜ~ãråáîÉêëáíó=çÑ=lñÑçêÇråáîÉêëáíó=çÑ=lñÑçêÇråáîÉêëáíó=çÑ=bÇáåÄìêÖÜråáîÉêëáíó=çÑ=kçííáåÖÜ~ãråáîÉêëáíó=çÑ=kçííáåÖÜ~ãïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMSpÅáÉåíáÑáÅ=éêçÖê~ããÉ^ää=ÉîÉåíë=ÉñÅÉéí=íÜÉ=ÅçåÑÉêÉåÅÉ=ÇáååÉê=ïáää=í~âÉ=éä~ÅÉ=áå=íÜÉ=^ìëíáå=mÉ~êÅÉ=_ìáäÇáåÖ=çÑ=íÜÉ=råáîÉêëáíó=çÑ=pìêêÉóKtÉÇåÉëÇ~ó=OP=^ìÖìëítçêâëÜçéW=“píÉã=`Éää=fã~ÖáåÖÒI=çêÖ~åáëÉê=jáâÉ=jçÇçNKMM=Ó NKPM=éãNKPM=Ó OKMM=éãOKMM=Ó OKQM=éãOKQM=Ó PKOM=éãPKOM=Ó PKQM=éãPKQM=Ó QKOM=éãQKOM=Ó RKMM=éãRKMM=Ó RKPM=éã`çÑÑÉÉ=~åÇ=êÉÖáëíê~íáçåqÜÉ=ÇáîÉêëáíó=çÑ=ëíÉã=ÅÉääëp~Ö~=gçÜ~åëëçåI=háåÖë=`çääÉÖÉI=içåÇçåtÜó=Çç=ïÉ=åÉÉÇ=ëíÉã=ÅÉää=áã~ÖáåÖ\=jáâÉ=jçÇçI=háåÖë=`çääÉÖÉI=içåÇçåaÉëáÖåáåÖ=Åçåíê~ëí=~ÖÉåíë=Ñçê=ëíÉã=ÅÉää=áã~ÖáåÖháëÜçêÉ _~ÜâççI=fãéÉêá~ä=`çääÉÖÉqÉ~fëëìÉë=~åÇ=éêçÄäÉãë=ëìêêçìåÇáåÖ=íÜÉ=ä~ÄÉääáåÖ=çÑ=ëíÉã=ÅÉääë=~åÇ=íÜÉáê=íÜÉê~éÉìíáÅ=éçíÉåíá~äj~êâ=ióíÜÖçÉI=råáîÉêëáíó=`çääÉÖÉ=içåÇçåfã~ÖáåÖ=çÑ=ÖÉåÉ=ÉñéêÉëëáçåmçJt~Ü pçI=fãéÉêá~ä=`çääÉÖÉ`äçëáåÖ=êÉã~êâëSKPM=Ó VKMM=éãaêáåâë=êÉÅÉéíáçåI=Üçí=ÄìÑÑÉí=~åÇ=éçëíÉêëïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


qÜìêëÇ~ó=OQ=^ìÖìëíUKMM=Ó VKOR=~ãVKOR=Ó VKPM=~ãVKPM=Ó NMKNR=~ã`çÑÑÉÉI=é~ëíêáÉë=~åÇ=êÉÖáëíê~íáçåtÉäÅçãÉfNW=_áää=jççêÉ=jÉãçêá~ä=iÉÅíìêÉeìã~å=_ê~áå=jof=~í=QKT=qÉëä~oçÖÉê=lêÇáÇÖÉI=råáîÉêëáíó=`çääÉÖÉ=içåÇçåNMKNR=Ó NNKNR=~ãNMKNRNMKOTNMKPVNMKRNNNKMPpÅáÉåíáÑáÅ=pÉëëáçå=fW=qÉÅÜåáÅ~ä=ÇÉîÉäçéãÉåíë=áå=áã~ÖáåÖ`Ü~áêëW=oçÖÉê=lêÇáÇÖÉ ~åÇ=d~êÉíÜ=_~êâÉêlNW=c~ëíÉê=aáñçå=c~íJt~íÉê=fã~ÖáåÖ=táíÜ=päáÅÉ=jìäíáéäÉñÉÇ=mìäëÉëhì~å g=iÉÉ^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇI=pÜÉÑÑáÉäÇlOW=^å=fåëÉêí~ÄäÉI=pÜçìäÇÉêJpäçííÉÇ=dê~ÇáÉåí=~åÇ=pÜáã=pÉí=Ñçê=aóå~ãáÅ=pÜáããáåÖjáÅÜ~Éä=mççäÉI=oáÅÜ~êÇ=_çïíÉää=páê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãlPW=máíÑ~äë áå=íÜÉ=ÇÉîÉäçéãÉåí=çÑ=räíê~=pÜçêí=bÅÜç=ErqbF=áã~ÖáåÖ=~í=ÜáÖÜ=ÑáÉäÇ=ëíêÉåÖíÜ^äÉñ~åÇê hÜê~éáÅÜÉî N I=káÅçä~=páÄëçå N I=j~ííÜÉï=oçÄëçå O I=^åÇêÉï=jK=_ä~ãáêÉ PN aÉé~êíãÉåí=çÑ=mÜóëáçäçÖóI== O lñÑçêÇ=`ÉåíêÉ=Ñçê=`äáåáÅ~ä=j~ÖåÉíáÅ=oÉëçå~åÅÉI=råáîÉêëáíó=çÑ=lñÑçêÇX= P kÉïÅ~ëíäÉ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=råáîÉêëáíó=çÑ=kÉïÅ~ëíäÉlQW=^å áåëÉêí=ëóëíÉã=Ñçê=~Çî~åÅÉÇ=áã~ÖáåÖ=êÉëÉ~êÅÜ=çå=ÅäáåáÅ~ä=jo=áã~ÖÉêëj~êíóå m~äÉó N I=bìÖÉåó hêàìâçî N I=g~ãÉë=táäÇ N I=jáÅÜ~Éä=i~ãéÉêíÜ O I==f~å=vçìåÖ ON^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇX== O jÉÅÜ~åáÅ~ä=båÖáåÉÉêáåÖI=fãéÉêá~ä=`çääÉÖÉ=içåÇçålRW=léíáãáë~íáçå çÑ=píÉ~Çó=pí~íÉ=cêÉÉ=mêÉÅÉëëáçå=pÉèìÉåÅÉë=Ñçê=eóéÉêéçä~êáëÉÇ=P eÉ=jofgK=jK=táäÇ N I=hK=qÉÜ N I=kK=tççÇÜçìëÉ N I=jKkKgK=m~äÉó N I=kK=ÇÉ=w~åÅÜÉ O I=iK=h~ëìÄçëâá PN^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇI=pÜÉÑÑáÉäÇI=råáíÉÇ=háåÖÇçãI= O fåëíáíìíÉ=çÑ=_áçãÉÇáÅ~ä=båÖáåÉÉêáåÖI=wìêáÅÜI=pïáíòÉêä~åÇI= P mÜáäáéë=jÉÇáÅ~ä=póëíÉãëI=`äÉîÉä~åÇI=leI=råáíÉÇ=pí~íÉëNNKNR=Ó NNKQR=éã`çÑÑÉÉNNKQR=Ó NOKNR=éãNNKQRpÅáÉåíáÑáÅ=pÉëëáçå=ffW==_ê~áå=ëíêìÅíìêÉ=~åÇ=ÅçÖåáíáîÉ=ÑìåÅíáçåEíê~ÅíçÖê~éÜóI=ÑjofI=ÉíÅKF`Ü~áêëW=píÉÑ~å=aÉÄÉåÉê ~åÇ=^ååÉííÉ=píÉêêfåîáíÉÇ=äÉÅíìêÉ=fOW oÉÅÉåí=~Çî~åÅÉë=áå=bbdLÑjof áåíÉÖê~íáçåpíÉÑ~å=aÉÄÉåÉê NION jo`=fåëíáíìíÉ=çÑ=eÉ~êáåÖ=oÉëÉ~êÅÜI=oçó~ä=pçìíÜ=e~åíë=eçëéáí~äI=pçìíÜ~ãéíçå=plNQ=lvdX= O`äáåáÅ~ä=kÉìêçëÅáÉåÅÉëI=pÅÜççä=çÑ=jÉÇáÅáåÉI=råáîÉêëáíó=çÑ=pçìíÜ~ãéíçåI=pçìíÜ~ãéíçåI=rh


NOKNRNOKOTNOKPVlSW=fãéêçîÉÇ=^êíáÑ~Åí `çêêÉÅíáçå=Ñçê=`çãÄáåÉÇ=bbdLÑjofh~êÉå=gK=jìääáåÖÉê N I=m~ìä=pK=jçêÖ~å O I=oáÅÜ~êÇ=tK=_çïíÉää N I=N páê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëíáó çÑ=kçííáåÖÜ~ãX== O^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãlTW sçñÉäJ_~ëÉÇ=fåîÉëíáÖ~íáçå=lÑ=tÜáíÉ=j~ííÉê=`Ü~åÖÉë=få=kçêã~ä=^ÖÉáåÖ=rëáåÖ=aáÑÑìëáçå=qÉåëçê=fã~ÖáåÖ==^åÇ=qê~ÅíçÖê~éÜóqÜçã~ë=oK=_~êêáÅâI=oÉÄÉÅÅ~=^K=`Ü~êäíçåI=jáÅÜ~Éä=lÛpìääáî~åI=eìÖÜ=pK=j~êâìëI=`Üêáë=^K=`ä~êâ`ÉåíêÉ=Ñçê=`äáåáÅ~ä=kÉìêçëÅáÉåÅÉI=píK=dÉçêÖÉÛëI=råáîÉêëáíó=çÑ=içåÇçålOSW=`Ü~ê~ÅíÉêáë~íáçå=çÑ=Äê~áå=~åáëçíêçéó=ìëáåÖ=ÇáÑÑìëáçå=jofj=j `çêêÉá~I=d=_=táääá~ãëI=q=^=`~êéÉåíÉêtçäÑëçå _ê~áå=fã~ÖáåÖ=`ÉåíêÉI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉNOKQR=Ó OKMM=éãiìåÅÜI=éçëíÉêë=~åÇ=ÉñÜáÄáíáçåOKMM=Ó PKPM=éãOKMMOKPMOKQOOKRQPKNM Ó PKRM=éãPKRM Ó RKNR=éãcçê^Ö~áåëícçê^Ö~áåëípÅáÉåíáÑáÅ=pÉëëáçå=fffW=kÉìêçäçÖáÅ~ä=ÇáëÉ~ëÉë=~åÇ=ÇáëçêÇÉêëpÉëëáçå=`Ü~áêëW=dáååó=kÖ=~åÇ=`ä~ìÇá~=tÜÉÉäÉêJháåÖëÜçíífåîáíÉÇ=äÉÅíìêÉ=fPW=`Ü~ääÉåÖÉë=~åÇ=åÉÉÇë=Ô jo=íÉÅÜåáèìÉë=áå=íÜÉ=åÉìêçëÅáÉåÅÉëdáååó=kÖN`ÉåíêÉ=Ñçê=kÉìêçáã~ÖáåÖ pÅáÉåÅÉëI=j~ìÇëäÉó eçëéáí~äI=aÉåã~êâ=eáääI=içåÇçå=pbR=U^wX= O fåëíáíìíÉ=çÑ=mëóÅÜá~íêóI=aÉ=`êÉëéáÖåó m~êâI=içåÇçå=pbR=U^clUW=fåÅçêéçê~íáåÖ=açã~áå=håçïäÉÇÖÉ=áåíç=cìòòó=`çååÉÅíÉÇåÉëë=fã~ÖÉ=pÉÖãÉåí~íáçåW=^ééäáÅ~íáçå=íç=_ê~áå=iÉëáçå=sçäìãÉ=bëíáã~íáçå=áå=jìäíáéäÉ=pÅäÉêçëáëj~êâ=^K=eçêëÑáÉäÇ N I=oK=_~âëÜá O j~êÅç=oçî~êáë P I=j~ê~=^K=oçÅÅ~ P I=sÉåâ~í~ pKoK=a~åÇ~ãìÇá O I=m~çä~=s~äë~ëáå~ P I=bäÇ~ gìÇáÅ~ P I=cìäîáç iìÅÅÜáåá P I=`Ü~êäÉë=dìííã~åå O I=j~êá~=má~ pçêã~åá Q ~åÇ=j~ëëáãç=cáäáééá PN råáîÉêëáíó=çÑ=iÉáÅÉëíÉêI=iÉáÅÉëíÉê=ibN=RttX= O e~êî~êÇ=jÉÇáÅ~ä=pÅÜççäI=_çëíçå=j^X= P råáîÉêëáíó=çÑ=jáä~åI=fí~äóX= Q råáîÉêëáíó=çÑ=dÉåçî~I=fí~äólVW=qÉãéçê~ä ÅÜ~åÖÉë=áå=ÇáÑÑìëáçå=áå=ãçÇÉê~íÉ=íç=ëÉîÉêÉ=íê~ìã~íáÅ=Äê~áå=áåìêókÉïÅçãÄÉ scgI=kçêíàÉ gI=_ê~ÇäÉó=mdI=e~êÇáåÖ=pdI=`Ü~íÑáÉäÇ=a^I=lìííêáã gdI=`çäÉë=gmI=eìíÅÜáåëçå=mgI=máÅâ~êÇ=gaI=`~êéÉåíÉê=q^I=jÉåçå ahI=táääá~ãë=d_==tçäÑëçå _ê~áå=fã~ÖáåÖ=`ÉåíÉêI=^ÇÇÉåÄêççâÉÛë eçëéáí~äI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉlNMW=^Äåçêã~ä íÜ~ä~ãáÅ=ãÉí~Äçäáëã=çå=eJjop=áå=ÚéÉêáéÜÉê~äÛ=Çá~ÄÉíáÅ=åÉìêçé~íÜófa=táäâáåëçå N I=a=pÉäî~ê~à~Ü O I=o=d~åÇÜáI=k=tççÇÜçìëÉI=`=bãÉêóI=ma=dêáÑÑáíÜëI=qÉëÑ~óÉ pN^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇ=X= O aá~ÄÉíÉë=råáíI=oçó~ä=e~ää~ãëÜáêÉeçëéáí~äI=däçëëçé=oç~ÇI=pÜÉÑÑáÉäÇ=pNM=OgcqÉ~aÉÄ~íÉÒqÜáë=ÜçìëÉ=ÄÉäáÉîÉë=íÜ~í=jof=ëÜçìäÇ=ÖáîÉ=ìé=íêóáåÖ=íç=ÄÉ=mbqÒ`Ü~áêã~åW=páãçå=açê~åfåîáíÉÇ=ëéÉ~âÉêëW^Çêá~å=`~êéÉåíÉêI=tçäÑëçå _ê~áå=fã~ÖáåÖ=`ÉåíêÉI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉ^åÇó=gìê~ëòI=mçáåí=qÜÉê~éÉìíáÅë=fåÅKI=_çëíçåa~îáÇ=_êççâëI=jo`=`óÅäçíêçå=råáíj~êâ=ióíÜÖçÉI=fåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜ


SKMM=éã=Ó`çåÑÉêÉåÅÉ=ÇáååÉê=~í aÉåÄáÉë táåÉêóI=açêâáåÖI=båÖä~åÇÛë=ä~êÖÉëí=îáåÉó~êÇ^å=~ééêçñáã~íÉ=íáãÉí~ÄäÉ=áë=~ë=ÑçääçïëWRKPM=éêçãéíSKPM=Ó TKNR`ç~ÅÜÉë=äÉ~îÉ=Ñêçã=^ìëíáå=mÉ~êÅÉ=_ìáäÇáåÖlå=~êêáî~äI=íÜÉ=é~êíó=ïáää=ëéäáí=áåíç=íÜêÉÉK=lÑ=íÜçëÉ=ïÜç=ÄççâÉÇ íç=Öç=çå=~=íçìêI=çåÉ=Öêçìé=ïáää=í~âÉ=~å=ÉñíÉêå~ä=íçìê=çÑ=íÜÉ=ïáåÉêó=~åÇ=~=ëÉÅçåÇ=Öêçìé=ïáää=íçìê=íÜÉ=áåíÉêå~ä=é~êí=çÑ=íÜÉ=ïáåÉêóKtÉ=~éçäçÖáëÉ=íÜ~í=ÇìÉ=íç=äáãáí~íáçåë=áå=åìãÄÉêë=íÜ~í=Å~å=ÄÉ=~ÅÅçããçÇ~íÉÇ=çå=íÜÉ=íçìêëI=åçí=ÉîÉêóÄçÇó=ïÜç=ÅÜçëÉ=íç=Çç=íÜáë=ïáää=Ü~îÉ=íÜÉ=çééçêíìåáíóK=tÉ=ïáää=çéÉê~íÉ=íÜáë=çå=~=ÑáêëíJÅçãÉ=ÑáêëíJëÉêîÉÇ=Ä~ëáëI=ÇÉéÉåÇÉåí=çå=íÜÉ=Ç~íÉ=çÑ=êÉÖáëíê~íáçå=~åÇ=~=Ñìää=êÉÑìåÇ=çÑ=~åó=ÑÉÉë=é~áÇ=ïáää=ÄÉ=ÖáîÉåKqÜÉ=êÉã~áåáåÖ=Öêçìé=ïáää=ÄÉ=çÑÑÉêÉÇ=~=ïáåÉJí~ëíáåÖKTKQR=Ó NNKMM=NNKMM=Ó NNKPMaáååÉê=~åÇ=ÉåíÉêí~áåãÉåí`ç~ÅÜÉë=äÉ~îÉ=Ñçê=íÜÉ=råáîÉêëáíó


cêáÇ~ó=OR=^ìÖìëíUKMM=Ó UKPM=~ã`çÑÑÉÉI=é~ëíêáÉë=~åÇ=êÉÖáëíê~íáçåVKMM=Ó VKPM=~ãpéÉÅá~ä=áåîáíÉÇ=äÉÅíìêÉ=fUW=qÜÉ=br=mÜóëáÅ~ä=^ÖÉåíë=EbjcF=aáêÉÅíáîÉ=~åÇ=áíë=fãé~Åí=çå=jof=J mêìÇÉåí=mêÉÅ~ìíáçå=çê=píê~áÖÜí=_~å~å~ë\píÉéÜÉå hÉÉîáäháåÖÛë=`çääÉÖÉ=içåÇçåVKPM=Ó NMKPR=~ãVKPMNMKMMNMKNONMKOQNMKPSpÅáÉåíáÑáÅ=pÉëëáçå=fsW=qÜÉ=ÑìíìêÉ=çÑ=ÅäáåáÅ~ä=jof=~åÇ=áíë=ìäíáã~íÉ=äáãáíëpÉëëáçå=`Ü~áêëW=j~ííÜÉï `äÉãÉåÅÉ ~åÇ=a~îáÇ=`çääáåëfåîáíÉÇ=äÉÅíìêÉ=fQW=qÜÉ=jof=ëóëíÉã=çÑ=íÜÉ=ÑìíìêÉW=áååçî~íáçåI=ÇÉîÉäçéãÉåí=~åÇ=ÇáêÉÅíáçåë=Ó ~=éÉêëçå~ä=áåÇìëíêá~ä=éÉêëéÉÅíáîÉj~ííÜÉï `äÉãÉåÅÉ`äáåáÅ~ä=pÅáÉåÅÉ=pìééçêí=joI=mÜáäáéë=jÉÇáÅ~ä=póëíÉãë=rhlNNW=qOG=ïÉáÖÜíÉÇ=~åÇ=éÜ~ëÉ=áã~ÖáåÖ=~í=T=qÉëä~iÉá gá~åÖI=^åÇêÉ~ë=pÅÜ~ÉÑÑÉêI=mÉååó dçïä~åÇ ~åÇ=oáÅÜ~êÇ _çïíÉääpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kçííáåÖÜ~ãI=kdT=OoalNOW=j~ÖåÉíáÅ=cáÉäÇ=fåÇìÅÉÇ=sÉêíáÖçf~å `~îáåI tÉåiçåÖ=ná~åI=oáÅÜ~êÇ _çïíÉääI=mÉååó dçïä~åÇI=m~ìä=däçîÉêpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kdT=OoalNPW=qÜÉ=ëíìÇó=çÑ=ã~ÖåÉíáÅ å~åçé~êíáÅäÉë ~ë=~=Åçåíê~ëí=ÉåÜ~åÅÉê=Ñçê=jof=íç=íê~Åâ=Å~åÅÉê=ãÉí~ëí~ëáëi~ìê~=j=m~êâÉë N oáÅÜ~êÇ=eçÇÖëçå N I=f~å=oçÄáåëçå O I=a~îáÇ=cÉêåáÖ P I=kÖìóÉå=qh=qÜ~åÜ OIPN j~ÖåÉíáÅ=oÉëçå~åÅÉ=~åÇ=fã~ÖÉ=^å~äóëáë=oÉëÉ~êÅÜ=`ÉåíêÉ=Ej^of^o`FXOaÉé~êíãÉåí=çÑ=`ÜÉãáëíêóX= P`ÉåíêÉ=Ñçê k~åçëÅ~äÉ pÅáÉåÅÉI=pÅÜççä=çÑ=_áçäçÖáÅ~ä=pÅáÉåÅÉëI=qÜÉ=råáîÉêëáíó=çÑ=iáîÉêéççälNQW=`Ü~ê~ÅíÉêáò~íáçå=çÑ=pçäáÇ=pí~íÉ=aóå~ãáÅ=kìÅäÉ~ê=mçä~êáò~íáçå=Ñçê jÉí~ÄçäáÅ=fã~ÖáåÖjK=pÅÜêçÉÇÉêI=iK `çÅÜäáåI=aK=qóäÉêI=dK o~ÇÇ~I=hK=`ä~êâÉaÉé~êíãÉåí=çÑ=mÜóëáçäçÖó=^å~íçãó=~åÇ=dÉåÉíáÅëI pÜÉêêáåÖíçå _ìáäÇáåÖI=råáîÉêëáíó=çÑ=lñÑçêÇI=m~êâë=oÇI=luN=PmqNMKRM=Ó NNKNR=~ã`çÑÑÉÉNNKNR=Ó NOKPM=éãNNKNRpÅáÉåíáÑáÅ=pÉëëáçå=sW bñíê~Åê~åá~ä jofpÉëëáçå=`Ü~áêëW=a~îáÇ içã~ë ~åÇ==píÉéÜÉå hÉÉîáäfåîáíÉÇ=äÉÅíìêÉ=fRW=fåíÉê~ÅíáîÉ=C=oÉ~äJíáãÉ=_çÇó=jofa~îáÇ içã~ëaÉé~êíãÉåí=çÑ=o~ÇáçäçÖóI=pÅÜççä=çÑ=`äáåáÅ~ä=jÉÇáÅáåÉI=råáîÉêëáíó çÑ=`~ãÄêáÇÖÉ


NNKQRNNKRTNOKMVlNRW=jof=íê~ÅâáåÖ=íç=Éëí~ÄäáëÜ=íÜÉ=êÉä~íáçåëÜáé=ÄÉíïÉÉå=ãóçÅ~êÇá~ä=áåàìêó=~åÇ=ëíÉã=ÅÉää=ÜçãáåÖ`~êçäóå=`~êê N I=a~åáÉä=píìÅâÉó N I=içìáëÉ=q~ííçå O I=a~ãá~å=qóäÉê N IgìÉêÖÉåpÅÜåÉáÇÉê N Ipá~åe~êÇáåÖ P I=háÉê~å=`ä~êâÉ NN aÉé~êíãÉåí=çÑ=mÜóëáçäçÖóI=^å~íçãó=~åÇ=dÉåÉíáÅëI=råáîÉêëáíó=çÑ=lñÑçêÇX= O k~íáçå~ä=_äççÇ=pÉêîáÅÉI=gçÜå o~ÇÅäáÑÑÉ eçëéáí~äI=lñÑçêÇX= P k~íáçå~ä=eÉ~êí=~åÇ=iìåÖ=fåëíáíìíÉI=fãéÉêá~ä=`çääÉÖÉI=içåÇçålNSW=d~ëíêáÅ=~åÇ=ëã~ää=ÄçïÉä=êÉëéçåëÉ=íç=~äÖáå~íÉ=ÄÉ~ÇëbK=cK=`çñ N I=`K=iK=eç~Ç N I=mK=o~óãÉåí O I=oK=`K=péáääÉê P I=mK=gK=têáÖÜí N I=jK=_ìíäÉê O I=iK=j~êÅá~åá NIP I=mK=^K=dçïä~åÇ NN páê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãX== O`çêéçê~íÉ=oÉëÉ~êÅÜI råáäÉîÉê o=C=aI`çäïçêíÜI=_ÉÇëKX= P tçäÑëçå=aáÖÉëíáîÉ=aáëÉ~ëÉë=`ÉåíêÉI=råáîÉêëáíó=eçëéáí~äI=kçííáåÖÜ~ãlNTW=^ÅÅìê~íÉ=^ìíçã~íÉÇ=jÉ~ëìêÉãÉåí=çÑ=aóå~ãáÅ=^êíÉêá~ä=iìãÉå=^êÉ~=Äó `jo`ä~êÉ=bK=g~Åâëçå N I `ÜÉÉê~Ö `K=pÜáêçÇ~êá~ N I=gìëíáå=jK=pK=iÉÉ N I=oçÄáå=mK=`ÜçìÇÜìêó N I=hÉáíÜ=jK=`Ü~ååçå N I=píÉÑ~å=kÉìÄ~ìÉê N I=j~ííÜÉï=aK=oçÄëçå NN råáîÉêëáíó=çÑ=lñÑçêÇ=`ÉåíêÉ=Ñçê=`äáåáÅ~ä=j~ÖåÉíáÅ=oÉëçå~åÅÉ=oÉëÉ~êÅÜI=lñÑçêÇNOKOR=Ó NKPM=éãiìåÅÜI=éçëíÉêë=~åÇ=ÉñÜáÄáíáçåNKPM=Ó PKNR=éãNKPMOKMMOKNOOKOQOKPSpÅáÉåíáÑáÅ=pÉëëáçå=sfW=_áçãÉÇáÅ~ä=áã~ÖáåÖ=~åÇ=ëéÉÅíêçëÅçéó`Ü~áêëWkáâá=páÄëçåIcê~åâäóåeçïÉfåîáíÉÇ=äÉÅíìêÉ=fSW=fã~ÖáåÖ=`kp=áåÑä~ãã~íáçåW=ïÜ~í=jo=Å~å=~åÇ=Å~åÛí=íÉää=ìëkáâá=páÄëçåbñéÉêáãÉåí~ä kÉìêçáã~ÖáåÖ dêçìéI=råáîÉêëáíó=i~Äçê~íçêó=çÑ=mÜóëáçäçÖóI=lñÑçêÇlNUW=jo=ãÉ~ëìêÉãÉåíë=çÑ=ÇáÑÑìëáçåI=éÉêÑìëáçå=~åÇ=qO=áå=íÜÉ=ÜáééçÅ~ãéìë=ÑçääçïáåÖ=ëí~íìë ÉéáäÉéíáÅìë áå=íÜÉ=ê~í=äáíÜáìãJéáäçÅ~êéáåÉ ãçÇÉäj=`Üçó N I=oK`K=pÅçíí N I=aKiK=qÜçã~ë O I=aKdK=d~Çá~å N I=jKcK=ióíÜÖçÉ NN r`i=fåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜI=içåÇçåX= O aÉé~êíãÉåí=çÑ=jÉÇáÅ~ä=mÜóëáÅë=~åÇ=_áçÉåÖáåÉÉêáåÖI=råáîÉêëáíó=`çääÉÖÉ=içåÇçåI=içåÇçålNVW= PN m=jop=ëíìÇó=çå=áåíê~ÅÉääìä~ê=ée=~åÇ ÉåÉêÖÉíáÅë áå=Üìã~å=Äê~áå=ÇìêáåÖ=îáëì~ä=ëíáãìä~íáçå=ìåÇÉê Éìçñá~ ~åÇ=ãáäÇ=Üóéçñá~oáëíç ^K=h~ìéáååÉå N IoáëÜã~sáÇó~ë~Ö~ê NIO I ^äáãìä `ÜçïÇÜìêó OpÅÜççäë=çÑ= N péçêí=~åÇ=bñÉêÅáëÉ=pÅáÉåÅÉë=~åÇ= O mëóÅÜçäçÖóI=råáîÉêëáíó=çÑ=_áêãáåÖÜ~ãI bÇÖÄ~ëíçåI=_áêãáåÖÜ~ã=_NR=OqqlOMW=içåÖ=~åÇ=ëÜçêí=ÉÅÜç=íáãÉ=éêçíçå=ã~ÖåÉíáÅ=êÉëçå~åÅÉ=ëéÉÅíêçëÅçéó=çÑ=íÜÉ=ÜÉ~äíÜó=~ÖÉáåÖ=Äê~áåaçãáåáÅâ=gKlK=jÅfåíóêÉ N I=oÉÄÉÅÅ~=^K=`Ü~êäíçå O I=eìÖÜ=pK=j~êâìë O Icê~åâäóå^K=eçïÉ NN _~ëáÅ=jÉÇáÅ~ä=pÅáÉåÅÉëI=pí=dÉçêÖÉDëI=råáîÉêëáíó=çÑ=içåÇçåX= O`äáåáÅ~ä=kÉìêçëÅáÉåÅÉëI=pí=dÉçêÖÉDëI=råáîÉêëáíó=çÑ=içåÇçålONW=jof=áåÇáÅ~íÉë=íÜ~í=ÄçåÉ=ã~êêçï=ëíÉã=ÅÉääë=êÉã~áå=Öê~ÑíÉÇ=áå=íÜÉ áåÑ~êÅíÉÇ ê~í=ÜÉ~êí=Ñçê=ëáñíÉÉå=ïÉÉâëI=Äìí=Çç=åçí=áãéêçîÉ=Å~êÇá~Å=ÑìåÅíáçåpíìÅâÉó=ag N I=`~êê=`^NI=j~êíáåJoÉåÇçå b O I=qóäÉê=ag N I táääãçíí `OI=`~ëëáÇó=mg N I=e~äÉ=pgj O I=pÅÜåÉáÇÉê=gb N I q~ííçå i O I=e~êÇáåÖ=pb P I o~ÇÇ~ dh N I=t~íí=p O I=`ä~êâÉ=h NN`~êÇá~Å=jÉí~Äçäáëã=oÉëÉ~êÅÜ=dêçìéI=i~Äçê~íçêó=çÑ=mÜóëáçäçÖóI=^å~íçãó=~åÇ=dÉåÉíáÅëI=råáîÉêëáíó=çÑ=lñÑçêÇX= O píÉã=`Éää=oÉëÉ~êÅÜ=i~Äçê~íçêóI=k~íáçå~ä=_äççÇ=pÉêîáÅÉI=gçÜå o~ÇÅäáÑÑÉ eçëéáí~äK=lñÑçêÇX= P k~íáçå~ä=eÉ~êí=~åÇ=iìåÖ=fåëíáíìíÉI=fãéÉêá~ä=`çääÉÖÉ=pÅÜççä=çÑ=pÅáÉåÅÉI=qÉÅÜåçäçÖó=~åÇ=jÉÇáÅáåÉI=içåÇçå


OKQUlOOW=içåÖáíìÇáå~ä=pÜçêí=qb=mêçíçå=j~ÖåÉíáÅ=oÉëçå~åÅÉ=péÉÅíêçëÅçéó=~åÇaáëÉ~ëÉ=mêçÖêÉëëáçå=áå=fåÜÉêáíÉÇ mêáçå aáëÉ~ëÉeó~êÉ e NIO I páÇÇáèìÉ a OI=P I=tÉÄÄ=q OI=P ItêçÉp OI=P I `çääáåÖÉ g OI=P I=qÜçêåíçå=gp N Ivçìëêó q NN ióëÜçäã=aÉé~êíãÉåí=çÑ kÉìêçê~ÇáçäçÖóI=k~íáçå~ä=eçëéáí~ä=Ñçê=kÉìêçäçÖó=~åÇ=kÉìêçëìêÖÉêóI= O jo` mêáçå råáíI=aÉé~êíãÉåí=çÑ kÉìêçÇÉÖÉåÉê~íáîÉ aáëÉ~ëÉëI=fåëíáíìíÉ=çÑ=kÉìêçäçÖóI=r`iI= P k~íáçå~ä mêáçå `äáåáÅI=k~íáçå~ä=eçëéáí~ä=Ñçê=kÉìêçäçÖó=~åÇ=kÉìêçëìêÖÉêóI=nìÉÉå=pèì~êÉI=içåÇçåPKMM=Ó PKPM=éãqÉ~PKPM=Ó QKQM=éãPKPMQKMMQKNOQKOQpÅáÉåíáÑáÅ=pÉëëáçå=sffW=nì~åíáí~íáîÉ=jofI=Å~åÅÉê=~åÇ=Åçåíê~ëí=~ÖÉåíëpÉëëáçå=ÅÜ~áêëW=dÉçÑÑ=m~êâÉê=~åÇ=j~êíáå=iÉ~ÅÜfåîáíÉÇ=äÉÅíìêÉ=fTW=nì~åíáí~íáîÉ=jof=çÑ=qìãçìêë=ìëáåÖ=`çåíê~ëí=^ÖÉåíëdÉçÑÑ=m~êâÉêfã~ÖáåÖ=pÅáÉåÅÉ=~åÇ=_áçãÉÇáÅ~ä=båÖáåÉÉêáåÖI píçéÑçêÇ _ìáäÇáåÖI=råáîÉêëáíó=çÑ=j~åÅÜÉëíÉêI=lñÑçêÇ=oç~ÇI=j~åÅÜÉëíÉê=jNP=VmqlOPW=^=kÉï=cìåÅíáçå~ä=cçêã=Ñçê=^êíÉêá~ä=fåéìí=cìåÅíáçå=jçÇÉääáåÖj~ííÜÉï=loqlk N I=g~ãÉë=aÛ^êÅó N I=páãçå=t~äâÉêJp~ãìÉä N I=j~êíáå=iÉ~ÅÜ N I=a~îáÇ=e~ïâÉë O I=a~îáÇ=`çääáåë NN fåëíáíìíÉ=çÑ=`~åÅÉê=oÉëÉ~êÅÜI=pìííçåI=pìêêÉó=pjO=RkdX= O`ÉåíêÉ=Ñçê=jÉÇáÅ~ä=fã~ÖÉ=`çãéìíáåÖI=råáîÉêëáíó=`çääÉÖÉ=içåÇçålOQW=m~êíá~ä=îçäìãÉ=ÅçêêÉÅíáçå=Ñçê=ã~ÖåÉíáÅ=êÉëçå~åÅÉ êÉåçÖê~éÜó ëíìÇáÉëoçÇêáÖìÉò=dìíáÉêêÉòI=a N I=aá~ò jçåíÉëÇÉçÅ~I=l O I=tÉääëI=h N I jÉåÇáÅÜçîëòâóI=f P I=~åÇ=dçêÇçå=f PN`ÉåíêÉ=Ñçê=sáëáçåI=péÉÉÅÜ=~åÇ=páÖå~ä=mêçÅÉëëáåÖI=pÅÜççä=çÑ=bäÉÅíêçåáÅë=C=mÜóëáÅ~ä=pÅáÉåÅÉëI=råáîÉêëáíó=çÑ=pìêêÉóX= O brfqqI råáîÉêëáÇ~Ç ÇÉ=i~ë=m~äã~ë=ÇÉ dê~å=`~å~êá~I=pé~áåX= P o`p=råáí=çÑ=_áçéÜóëáÅëI=fåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜI=råáîÉêëáíó=`çääÉÖÉ=içåÇçålORW=fã~ÖáåÖ=áå=ÇêìÖ=ÇÉîÉäçéãÉåíW=äÉëëçåë=Ñêçã=íÜÉê~éÉìíáÅë=~åÇ=í~êÖÉíÉÇ=ãçäÉÅìä~ê=áã~ÖáåÖ=~ÖÉåíë^åÇó aòáâJgìê~ëòmçáåí=qÜÉê~éÉìíáÅëI=fåÅKI=_çëíçåI=j^QKQM=éã`äçëÉ=çÑ=ãÉÉíáåÖ


mçëíÉêëmNmOmPmQmRmSmTmUmV`ä~ëëáÑáÅ~íáçå=çÑ=^ìíçã~íÉÇJdóêáÑáÅ~íáçå fåÇÉñ=aÉëÅêáéíçêë=ìëáåÖ=pìéÉêîáëÉÇ=iÉ~êåáåÖ=jÉíÜçÇëW=^=cÉ~íìêÉ=_~ëÉÇ jçêéÜçãÉíêó píìÇóeÉáÇá=j _çååáÅáI=gçå~íÜ~å=j=e~êêáëI==^åÇêÉï=`=pí~åÑáÉäÇI=açãáåáÅ=gçÄI=d=h~íÜÉêáåÉ=p ióãÉêI=píÉéÜÉå=j i~ïêáÉI=q=táääá~ã=g=jççêÜÉ~Çaáîáëáçå=çÑ=mëóÅÜá~íêó=I=qÜÉ=råáîÉêëáíó=çÑ=bÇáåÄìêÖÜ`Ü~ê~ÅíÉêáë~íáçå=çÑ=~=`~êçíáÇ=fåàìêó=jçÇÉä=áå=o~í=áå=îáîç=rëáåÖ=jofhKhK=`ÜÉìåÖ NIO I=jKgK=o~ãáêÉò P I=mK=iÉÜíçä~áåÉå P I=aKdK=d~Çá~å O I=^KjK=q~óäçê Q I=oKgK=lêÇáÇÖÉ N I=jKcK=ióíÜÖçÉ ON aÉé~êíãÉåí=çÑ=jÉÇáÅ~ä=mÜóëáÅë=~åÇ=_áçÉåÖáåÉÉêáåÖI=råáîÉêëáíó=`çääÉÖÉ=içåÇçåI=içåÇçåX= O o`p=råáí=çÑ=_áçéÜóëáÅëI=r`i=fåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜI=içåÇçåX= P`ÉåíêÉ=Ñçê=`~êÇáçî~ëÅìä~ê=_áçäçÖó=~åÇ=jÉÇáÅáåÉI=aáîáëáçå=çÑ=jÉÇáÅáåÉI=råáîÉêëáíó=`çääÉÖÉ=içåÇçåX= Q`~êÇáçíÜçê~ÅáÅ=råáíI=dêÉ~í=lêãçåÇ=píêÉÉí=eçëéáí~ä=Ñçê=`ÜáäÇêÉåI=içåÇçå^=içÅ~ääó=^Ç~éíáîÉ=dê~ÇáÉåí=`çåíêçääÉÇ=pé~íá~ä=oÉÖìä~êáë~íáçå=m~êíá~ä=sçäìãÉ=`ä~ëëáÑáÉêgçÜå=mK `ÜáîÉêíçåI=hÉîáå=tÉääë`ÉåíêÉ=Ñçê=sáëáçåI=péÉÉÅÜ=~åÇ=páÖå~ä=mêçÅÉëëáåÖI=pÅÜççä=çÑ=bäÉÅíêçåáÅë=~åÇ=mÜóëáÅ~ä=pÅáÉåÅÉëI=råáîÉêëáíó=çÑ=pìêêÉóI=drO=TuepÜáÉäÇÉÇ Äáéä~å~ê Öê~ÇáÉåí=Åçáä=ÇÉëáÖå=Ñçê=~å=çéÉå=éÉêã~åÉåí=äçï=ÑáÉäÇ=jofiá pòÉ `ÜçïI hì~å gKiÉÉI=gáã=j=táäÇI j~êíóå kKgKm~äÉópÉÅíáçå=çÑ=^Å~ÇÉãáÅ=o~ÇáçäçÖóI råáîÉêáëíó çÑ=pÜÉÑÑáÉäÇI=oçó~ä e~ää~ãëÜáêÉ eçëéáí~äI=däçëëçé=oç~ÇI=pNM=OgcjÉ~ëìêáåÖ=q OìëáåÖ=q OmêÉé~êÉÇ=_~ä~åÅÉÇ=qìêÄç=cáÉäÇ=bÅÜç=fã~ÖáåÖ=~í=PKMq=~åÇ=TKMqbK=cK=`çñI=`K=iK eç~ÇI=mK=^K dçïä~åÇpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kçííáåÖÜ~ãkÉÖ~íáîÉ=_lia=~ÅÅçãé~åáÉë=ëÜ~êéÉåáåÖ=çÑ=ïÜáëâÉê=êÉéêÉëÉåí~íáçåë=áå=ê~í=Ä~êêÉä=ÅçêíÉñ=îáëì~äáòÉÇ=ïáíÜ=ÑìåÅíáçå~ä=ã~ÖåÉíáÅ=êÉëçå~åÅÉ=áã~ÖáåÖ_Éåáíç=ÇÉ `Éäáë ^äçåëç N I=^åÇêÉï=pK=içïÉ N I ^áëäáåÖ iK=aáñçå N I=gçÜå=mK=aÉ~ê O Ih~äçâ`K=iÉÉ P I=píÉîÉå=`K=oK=táääá~ãë N ~åÇ=dÉê~äÇ=qK=cáååÉêíó NN jo`=`ÉåíêÉ=Ñçê=kÉìêçÇÉÖÉåÉê~íáçåI=háåÖÛë=`çääÉÖÉ=içåÇçåI=aÉ=`êÉëéáÖåó=m~êâI=içåÇçå=pbR=U^cX= O aÉé~êíãÉåí=çÑ=jÉÅÜ~åáÅ~ä=båÖáåÉÉêáåÖI=fãéÉêá~ä=`çääÉÖÉ=içåÇçåI=içåÇçå=ptT=O^wX= P aáîáëáçå=çÑ=båÖáåÉÉêáåÖI=háåÖÛë=`çääÉÖÉ=içåÇçåI=píê~åÇI=içåÇçå=t`Oo=Oip^å=~ëóããÉíêáÅ=èì~Çê~íìêÉ=ÄáêÇÅ~ÖÉ=ÄçÇó=qJo=Åçáä=Ñçê=ÜóéÉêéçä~êáëÉÇ==PeÉ=äìåÖ=jofk=ÇÉ=w~åÅÜÉ P I=k=`Üáåå~ O I=h=qÉÜ N I=`=o~åÇÉää O I=gj=táäÇ NN^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇI=rhI= O mìäëÉíÉè=iíÇI=rhI=_áçãÉÇáÅ~ä=båÖáåÉÉêáåÖI= P råáîÉêëáíó=~åÇ=bqeI=wìêáÅÜI=pïáíòÉêä~åÇc~ëí=~åÇ=^ÅÅìê~íÉ=j~ééáåÖ=çÑ=íÜÉ=cäáé=^åÖäÉ=rëáåÖ=íÜÉ=NUM° kìääkáÅÜçä~ë=d=açïÉää=~åÇ=m~ìä=p=qçÑíëfåëíáíìíÉ=çÑ=kÉìêçäçÖóI=r`iI=nìÉÉå=pèì~êÉI=içåÇçå=t`Nk=P_dfåîÉëíáÖ~íÉ=íÜÉ=ÇÉéÉåÇÉåí=çÑ=mPMM=çå=q~êÖÉí=íç=q~êÖÉí=fåíÉêî~äp~ääó=bäÇÉÖÜ~áÇóI=h~ó=bK=eÉ~ÇI=mÉååó=dçïä~åÇI=pìÉ=cê~åÅáëpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=råáîÉêëáíó=m~êâI=kçííáåÖÜ~ã kdT=Ooa


mNMmNNmçëíÉê=ïáíÜÇê~ïåmNOmçëíÉê=ïáíÜÇê~ïåmNPmNQmNRmNSmNTmNUoÉëçäîáåÖ=ÅçåÑçêã~íáçå~ä=ÉñÅÜ~åÖÉ=áå éáãçåáÇ~òçäÉ Ó ~=Üóéçñá~=ã~êâÉê`êáëíáå~ d~ÄÉääáÉêáI=qÜçã~ë=oK bóâóåI=dÉçÑÑêÉó=pK=m~óåÉI=j~êíáå=lK=iÉ~ÅÜ`~åÅÉê=oÉëÉ~êÅÜ=rh=`äáåáÅ~ä=j~ÖåÉíáÅ=oÉëçå~åÅÉI=qÜÉ=fåëíáíìíÉ=çÑ `~åÅÉê=oÉëÉ~êÅÜI=oçó~ä j~êëÇÉå cçìåÇ~íáçå=kepI=pìííçåI=pìêêÉóqÉñíìêÉ=~å~äóëáë=Ñçê=ÇáÑÑÉêÉåíá~íáçå=ÄÉíïÉÉå=ÄÉåáÖå=~åÇ=ã~äáÖå~åí=äáîÉê=íìãçìêë=áå=jof^K dÜ~êÄ~äáI=aêK=oK^K iÉêëâáaÉé~êíãÉåí=çÑ=jÉÇáÅ~ä=mÜóëáÅëI káåÉïÉääë eçëéáí~ä=C=jÉÇáÅ~ä=pÅÜççäI=råáîÉêëáíó=çÑ=aìåÇÉÉ`Ü~ê~ÅíÉêáò~íáçå=çÑ=ëÅ~ååÉê=ëí~ÄáäáíóX=ãÉ~ëìêáåÖ=ëìÄíäÉ=ÅÜ~åÖÉë=áå=~éé~êÉåí=íê~åëãáííÉê=çìíéìí=~åÇ=êÉÅÉáîÉ=Ö~áåg=p=g~ÅâëçåI=a=g qçòÉêI=m=p=qçÑíëfåëíáíìíÉ=çÑ=kÉìêçäçÖóI=r`iI=nìÉÉå=pèì~êÉI=içåÇçåI=t`Nk=P_diçï=ÑáÉäÇI==äçï=Åçëí=ãìäíáJåìÅäÉ~ê=áã~ÖáåÖ=êÉëÉ~êÅÜ=ëóëíÉãëbìÖÉåó=hêàìâçîI=g~ãÉë=táäÇI j~êíóå=m~äÉó^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇI=oçó~ä e~ää~ãëÜáêÉ eçëéáí~äI=pÜÉÑÑáÉäÇcÉêìãçñáÇÉ ä~ÄÉääáåÖ=çÑ=`aNPPH=ÅÉääë=Ñçê=í~êÖÉíÉÇ=êÉJÉåÇçíÜÉäá~äáò~íáçåW=éçíÉåíá~ä=~åÇ=éáíÑ~ääëm~å~Öáçíáë hóêí~íçë N I m=iÉÜíçä~áåÉå O I=jg=o~ãáêÉòOI=ad=d~Çá~å N I=n=m~åâÜìêëí P I=jc=ióíÜÖçÉ NN r`i fåëáíìíÉ çÑ=`ÜáäÇ=eÉ~äíÜI=içåÇçåX== O r`i=aÉé~êíãÉåí=çÑ=jÉÇáÅáåÉI=içåÇçåXP içåÇçå=`ÉåíêÉ=Ñçê k~åçíÉÅÜåçäçÖóI=içåÇçårëáåÖ=íÜÉ=ÄäççÇ=çñóÖÉå=äÉîÉä=ÇÉéÉåÇ~åí=ã~ÖåÉíáÅ=êÉëçå~åÅÉ=ëáÖå~ä íç=çÄëÉêîÉ=íÜÉ=åÉìêçå~ä=ÉÑÑÉÅíë=çÑ=~ÅìíÉ=~åÇ=ÅÜêçåáÅ ÑäìçñÉíáåÉ ~Çãáåáëíê~íáçå=áå=íÜÉ=ã~ãã~äá~å=Äê~áåq~ãëáå i~åÖäÉó N I=píÉîÉå=`KoK=táääá~ãë N I=jáÅÜ~Éä=lÛkÉáää O ~åÇ=káÅÜçä~ë=gçåÉë NN fåëíáíìíÉ=çÑ=mëóÅÜá~íêóI=içåÇçåI=rKhK== O bäá=iáääó=C=`çK=pìêêÉóaÉîÉäçéãÉåí=çÑ=~=ÅçãÄáåÉÇ ãáÅêçmbq∆Jjo=ëóëíÉã^KgKiìÅ~ë NIO I=oK`Ke~ïâÉëNI=oKbK^åëçêÖÉ O I=dK_Ktáääá~ãë N I=oKbKkìííPI=gK`K`ä~êâ N I=qKaKcêóÉê N I=qK^K`~êéÉåíÉê NN tçäÑëçå=_ê~áå=fã~ÖáåÖ=`ÉåíêÉI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉI=_çñ=SRI ^ÇÇÉåÄêççâÉÛëeçëéáí~äI=`~ãÄêáÇÖÉ=`_O=OnnX== O`~îÉåÇáëÜ=i~Äçê~íçêóI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉI=gKgK=qÜçãéëçå=^îÉåìÉI=`~ãÄêáÇÖÉ=`_P=MebX= P páÉãÉåë=jçäÉÅìä~ê=fã~ÖáåÖmêÉÅäáåáÅ~ä pçäìíáçåëI=UNM=fååçî~íáçå=aêáîÉI=håçñîáääÉI=qk=PTVPOI=rp^qÜÉ=êÉä~íáçåëÜáé=ÄÉíïÉÉå=Äê~áå=ëíêìÅíìêÉ=~åÇ ëÅÜáòçíóé~ä ãÉ~ëìêÉë=áå=ëìÄàÉÅíë=~í=ÜáÖÜ=êáëâ=çÑ=ÇÉîÉäçéáåÖ=ëÅÜáòçéÜêÉåá~d=Kh~íÜÉêáåÉ=pK ióãÉêI=açãáåáÅ=bK=gçÄI=qK=táääá~ã=gK=jççêÜÉ~ÇI=^åÇêÉï=jK=jÅfåíçëÜI=a~îáÇ=dK=lïÉåëI=bîÉ=`K gçÜåëíçåÉ ~åÇ=píÉéÜÉå=jK i~ïêáÉaáîáëáçå=çÑ=mëóÅÜá~íêóI=oçó~ä=bÇáåÄìêÖÜ=eçëéáí~äI=jçêåáåÖëáÇÉ=m~êâI=bÇáåÄìêÖÜ=beNM=Recjof=_~ëÉÇ=^ííÉåì~íáçå=`çêêÉÅíáçå=Ñçê=`çãÄáåÉÇ=mbqLjof~å=_=j~äçåÉ O I=oáÅÜ~êÇ=b=^åëçêÖÉ NIO I=qáã=a=cêóÉê N I=dìó=_=táääá~ãë N I=q=^Çêá~å=`~êéÉåíÉê NN tçäÑëçå=_ê~áå=fã~ÖáåÖ=`ÉåíêÉI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉI=_çñ=SRI ^ÇÇÉåÄêççâÉÛëeçëéáí~äI=`~ãÄêáÇÖÉ=`_O=OnnX= O`~îÉåÇáëÜ=i~Äçê~íçêóI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉI=gKgKqÜçãëçå=^îÉåìÉI=`~ãÄêáÇÖÉ=`_P=Meb=


mNVmOMmONmOOmOPmOQmORmOSmOTmOUqN=ã~ééáåÖ=áå=ÅÜáäÇÜççÇ=~ÄÇçãáå~ä=íìãçìêë=Ó éêÉäáãáå~êó=ÉñéÉêáÉåÅÉfçëáÑ jÉåÇáÅÜçîëòâó N IÕóëíÉáåbK=läëÉå ON o`p=råáí=çÑ=_áçéÜóëáÅëI=r`i=fåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜI=içåÇçåX= O o~ÇáçäçÖó=aÉé~êíãÉåíI=dêÉ~í=lêãçåÇ=píêÉÉí=eçëéáí~ä=Ñçê=`ÜáäÇêÉå=kep=qêìëíI içåÇçå^å=áåÅêÉ~ëÉÇ=ã~íêáñ=ëáåÖäÉ=ÉÅÜç=~Åèìáëáíáçå=j^j_^=Oa=~êê~ój~êíóå=m~äÉóI=g~ãÉë=táäÇI=hì~å=iÉÉ^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=çÑ=pÜÉÑÑáÉäÇfãé~Åí=çÑ=áåÅçåëáëíÉåí=êÉëçäìíáçå=çå=s_j=ëíìÇáÉëW=~å=Éñ~ãéäÉ=ìëáåÖ=pÉã~åíáÅ=aÉãÉåíá~gç©ç=jK=pK=mÉêÉáê~ N I=mÉíÉê=gK=kÉëíçê O I=dìó=_K=táääá~ãëN tçäÑëçå=_ê~áå=fã~ÖáåÖ=`ÉåíêÉI=aÉéíK=çÑ=`äáåáÅ~ä=kÉìêçëÅáÉåÅÉëI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉX== O kÉìêçäçÖó=råáíI=aÉéíK=çÑ=`äáåáÅ~ä=kÉìêçëÅáÉåÅÉëI=råáîÉêëáíó=çÑ=`~ãÄêáÇÖÉräíê~JbÑÑáÅáÉåí=pÜáÉäÇÉÇ=açãÉ=dê~ÇáÉåí=`çáäëjáÅÜ~Éä=mççäÉ=~åÇ=oáÅÜ~êÇ _çïíÉääpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=råáîÉêëáíó=m~êâI=kçííáåÖÜ~ã kdT=Ooa^ééäáÅ~íáçå=çÑ=eáÖÜÉêJlêÇÉê=_çìåÇ~êó=bäÉãÉåí=jÉíÜçÇ=íç=dê~ÇáÉåí=`çáä=aÉëáÖå`K `çÄçë p~åÅÜÉò N IiK=j~êáå O I=oKtK=_çïíÉää N I=eK=mçïÉê O I=mKjK=däçîÉê N I=^K^K=_ÉÅâÉê O ~åÇ=fK^K=gçåÉë ON páê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=qÜÉ=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kçííáåÖÜ~ã=m~êâI==kçííáåÖÜ~ã=kdT=OoaX= O pÅÜççä=çÑ=jÉÅÜ~åáÅ~äI=j~íÉêá~äë=~åÇ=j~åìÑ~ÅíìêáåÖ=båÖáåÉÉêáåÖI=qÜÉ=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kçííáåÖÜ~ã=m~êâI=kçííáåÖÜ~ã=kdT=OoaaÉíÉÅíáçå=çÑ=íÜÉ píêá~ çÑ dÉåå~êá ìëáåÖ=qìêÄç=péáå=bÅÜç=fã~ÖáåÖ==~í=Pqoçë~=p~åÅÜÉòI=pìë~å=cê~åÅáëI=oáÅÜ~êÇ _çïíÉääpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=råáîÉêëáíó=m~êâI=kçííáåÖÜ~ãI=kdT=Ooa^ìíçã~íÉÇ=qÉÅÜåáèìÉë=Ñçê jìêáåÉ=mÜÉåçíóéáåÖ áå=eìåíáåÖíçåÛë=aáëÉ~ëÉpK=gK=p~ïá~â N I=dK=_K=táääá~ãë N I=kK=fK=tççÇ OIP I=^K=gK=jçêíçå P I=qK=^K=`~êéÉåíÉê NN tçäÑëçå=_ê~áå=fã~ÖáåÖ=`ÉåíêÉI=_çñ=SRI ^ÇÇÉåÄêççâÉÛë eçëéáí~äI=eáääë=oç~ÇI=`~ãÄêáÇÖÉ=`_O=OnnX= O`~ãÄêáÇÖÉ=`ÉåíêÉ=Ñçê=_ê~áå=oÉé~áêI=bK=aK=^Çêá~å=_ìáäÇáåÖIcçêîáÉ páíÉI=oçÄáåëçå=t~óI=`~ãÄêáÇÖÉ=`_O=OmvX= P aÉé~êíãÉåí=çÑ=mÜ~êã~ÅçäçÖóI=qÉååáë=`çìêí=oç~ÇI=`~ãÄêáÇÖÉ=`_O=NmadÉåÉê~äáëÉÇ=fåÇÉñ=cçêãìä~É=Ñçê mêçä~íÉI=lÄä~íÉ=~åÇ=péÜÉêáÅ~ä=pÜ~éÉ=aáÑÑìëáçå=qÉåëçêëvìàá=pÜÉå N I=oáëíç=h~ìééáåÉå N I=fÇ~=mì O I=`Üêáë=`ä~êâ PN pÅÜççä=çÑ=péçêí=~åÇ=bñÉêÅáëÉ=pÅáÉåÅÉëI=råáîÉêëáíó=çÑ=_áêãáåÖÜ~ãX O aÉé~êíãÉåí=çÑ=`çãéìíáåÖI=dçäÇëãáíÜë=`çääÉÖÉI=råáîÉêëáíó=çÑ=içåÇçåX= P fåëíáíìíÉ=çÑ=`ÜáäÇ=eÉ~äíÜI=råáîÉêëáíó=`çääÉÖÉ=içåÇçå=qÜÉ=áåÑäìÉåÅÉ=çÑ=í~ëâ=ÇÉã~åÇ=çå=íÜÉ=_lia=êÉëéçåëÉ=áå=~=ÅçãéäÉñ=éçïÉêJÖêáé=í~ëâ^ååÉííÉ=píÉêê N I=pÜ~å=pÜÉå N I=d~åÖ=d~ç N I=tÉåÖëÜÉå=eçì O I=^åÇêÉ=pò~ãÉáí~í NN aÉé~êíãÉåí=çÑ=mëóÅÜçäçÖóI=råáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇI=rhX= O aÉé~êíãÉåí=çÑ=_áçãÉÇáÅ~ä=pÅáÉåÅÉëI=råáîÉêëáíó=çÑ=`ÜçåÖèìáåÖI=`ÜçåÖèìáåÖI=`Üáå~^=Åçãé~êáëçå=çÑ=ãçÇáÑáÉÇ=âJí=_i^pq=~åÇ=~=âJí=î~êá~ÄäÉJÇÉåëáíó=ìåÇÉêJë~ãéäáåÖ=íÉÅÜåáèìÉ=Ñçê=êÉíêçëéÉÅíáîÉäó=Ö~íÉÇ=jo=Ñäçï=ãÉ~ëìêÉãÉåívìÉÜìá=q~çI=f~å=j~êëÜ~ääjÉÇáÅ~ä=mÜóëáÅë=~åÇ=pebc`=_êá~å=fã~ÖáåÖ=oÉëÉ~êÅÜ=`ÉåíêÉ=Ñçê=pÅçíä~åÇI=råáîÉêëáíó=çÑ=bÇáåÄìêÖÜI=bÇáåÄìêÖÜ


mOVmPMmPNmPOmPPmPQmPRmPSpäáÅÉ=éêçÑáäÉ=ÉÑÑÉÅíë=áå=î~êá~ÄäÉ=Ñäáé=~åÖäÉ=em= P eÉ=jofhK qÉÜI=hKgK=iÉÉI=gKjK=táäÇ^Å~ÇÉãáÅ=o~ÇáçäçÖóI=råáîÉêëáíó=lÑ=pÜÉÑÑáÉäÇI=pÜÉÑÑáÉäÇnì~åíáí~íáîÉ=j~ÖåÉíáë~íáçå=qê~åëÑÉê=çÑ=içï=dê~ÇÉ däáçã~a~åáÉä=gK qçòÉêI=`ÜêáëíçéÜÉê=bK=_ÉåíçåI=m~ìä=pK=qçÑíë=~åÇ=gÉêÉãó=oK=oÉÉëfåëíáíìíÉ=çÑ=kÉìêçäçÖóI=råáîÉêëáíó=`çääÉÖÉ=içåÇçåI=nìÉÉå=pèì~êÉI içåÇçåI=t`Nk=P_dbmfpq^o=mìäëÉÇ=^êíÉêá~ä=péáå=i~ÄÉääáåÖ=mÉêÑìëáçå=fã~ÖáåÖ=~í=T=qÉëä~^äÉñ=d~êÇÉåÉêI=oçã~å tÉëçäçïëâáI=mÉååó dçïä~åÇ ~åÇ=pìë~å=cê~åÅáëpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kçííáåÖÜ~ãI=kdT=OoamêÉäáãáå~êó=a~í~=Ñçê=nì~åíáí~íáîÉ=qïçJaáãÉåëáçå~ä=eoj^p=péÉÅíêçëÅçéó=çÑ=eìã~å=_ê~áå=qìãçìêë^ä~å=têáÖÜí N I háêëíáÉ léëí~Ç N I=gçÜå=dêáÑÑáíÜë N I=_K=^åíÜçåó=_Éää O ~åÇ cê~åâäóå eçïÉ N=N`orh=_áçãÉÇáÅ~ä=j~ÖåÉíáÅ=oÉëçå~åÅÉ=oÉëÉ~êÅÜ=dêçìéI=aáîáëáçå=çÑ=_~ëáÅ=jÉÇáÅ~ä=pÅáÉåÅÉë=~åÇ= O aÉé~êíãÉåí=çÑ=kÉìêçëìêÖÉêóI=pí=dÉçêÖÉÛë=EråáîÉêëáíó=çÑ=içåÇçåFI=ptNT=Mobq NjÉ~ëìêÉãÉåíë=Ñçê=`çêíáÅ~ä=dêÉó=j~ííÉêI=tÜáíÉ=j~ííÉê=~åÇ=pìÄJ`çêíáÅ~ä=dêÉó=j~ííÉê=~í=TqmK=gK=têáÖÜíI=^K=mÉíÉêëI=jK=_êççâÉëI=oK `çñçåI=mK=jçêêáëI=pK=cê~åÅáëI=oK _çïíÉääI=mKdçïä~åÇpáê=mÉíÉê=j~åëÑáÉäÇ=j~ÖåÉíáÅ=oÉëçå~åÅÉ=`ÉåíêÉI=pÅÜççä=çÑ=mÜóëáÅë ~åÇ=^ëíêçåçãóI=råáîÉêëáíó=çÑ=kçííáåÖÜ~ãI=kçííáåÖÜ~ã=kdT=OoakÉáÖÜÄçìêÜççÇ íê~ÅíçÖê~éÜóW=~=åÉï=~ééêç~ÅÜ=íç=ëÉÉÇ=éçáåí=éä~ÅÉãÉåí=Ñçê=Öêçìé=ÑáÄêÉ=íê~ÅâáåÖgçå~íÜ~å=aK=`i^vabk N I=j~êâ=bK=_~ëíáå O I=^ãçë=gK=píçêâÉó PN kÉìêçáåÑçêã~íáÅë=açÅíçê~ä=qê~áåáåÖ=`ÉåíêÉI=pÅÜççä=çÑ=fåÑçêã~íáÅëX= O jÉÇáÅ~ä=~åÇ=o~ÇáçäçÖáÅ~ä=pÅáÉåÅÉë=EjÉÇáÅ~ä=mÜóëáÅëFX= P fåëíáíìíÉ=Ñçê=^Ç~éíáîÉ=~åÇ=kÉìê~ä=`çãéìí~íáçåI=pÅÜççä=çÑ=fåÑçêã~íáÅëX=råáîÉêëáíó=çÑ=bÇáåÄìêÖÜI=bÇáåÄìêÖÜI=rhs~äáÇ~íáçå=çÑ=akmJjo=Eaóå~ãáÅ=kìÅäÉ~ê=mçä~êáë~íáçå=j~ÖåÉíáÅ=oÉëçå~åÅÉF=ãÉ~ëìêÉãÉåí=çÑ=ëìÄëíê~íÉ=ìéí~âÉ=~åÇ=ãÉí~Äçäáëã=áå=íÜÉ=ÜÉ~êíiçïêá=`çÅÜäáåI=j~êáÉ=pÅÜêçÉÇÉêI=a~ãá~å=qóäÉêI=dÉçêÖÉ o~ÇÇ~I=háÉê~å=`ä~êâÉ`~êÇá~Å=jÉí~Äçäáëã=oÉëÉ~êÅÜ=dêçìéI=aÉé~êíãÉåí=çÑ=mÜóëáçäçÖóI=^å~íçãó=~åÇ=dÉåÉíáÅëI pÜÉêêáåÖíçå _ìáäÇáåÖI=råáîÉêëáíó=çÑ=lñÑçêÇI=rhfãéêçîÉÇ=ëíìÇó=çÑ=ãóçÅ~êÇá~ä=áåÑ~êÅíáçå=ìëáåÖ=Åçåíê~ëí=ÉåÜ~åÅÉÇ=jofd~åÖ=d~ç NIO I=m~ìä=`çÅâëÜçíí N I_àçÉêådêçÉååáåÖ N I=^ååÉííÉ=píÉêê ON råáîÉêëáíó=çÑ=dä~ëÖçïI=rhX== O råáîÉêëáíó=çÑ=ëìêêÉóI=rh


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMS_áää=jççêÉ=jÉãçêá~ä=iÉÅíìêÉïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


I1Bill Moore Memorial Lecture: Human Brain MRI at 4.7 TeslaRoger Ordidge, David Thomas, David Carmichael, Enrico DeVita and Robert TurnerUniversity College London, United KingdomIntroductionAn aim of this research is to develop an approach for high resolution MRI of the humanbrain that could produce high quality images in a clinically feasible scan time using a 4.7Tesla magnet. Higher resolution images are desirable in order to visualise the smaller andmore intricate brain structures that are currently not observable in standard MR images. Amajor limitation to obtaining higher spatial resolution is the limited signal-to-noise ratio(SNR) of the MR data acquisition.MethodsIn order to achieve this goal, it was necessary to:(1) Increase the main magnetic field strength to 4.7 Tesla.(2) Perform pulse sequence optimisation. At higher field strengths, the RF powerrequired for a given excitation flip angle is proportionally larger. This means thatthe power deposition (SAR) limits are reached more quickly. Also, the transmitRF field is less uniform due to field focusing effects which can lead to nonuniformimages with variable contrast. This problem proliferates in sequenceswith multiple RF pulses, where the net effect becomes more pronouncedfollowing each RF pulse.(3) Use RF receiver coil arrays. Over recent years, array coils have been usedincreasingly because they enable higher sensitivity to be achieved over a largervolume.(4) Improve the point spread function (PSF). Novel signal acquisition strategiescombined with parallel imaging methodology can be used to mitigate the effectson the PSF of T2 relaxation during multiple-echo MR imaging. The deleteriouseffects of related imaging ghosting can also be minimised.Results and DiscussionIn this presentation, the configuration of our high field 4.7T MR scanner, customised forneuroimaging applications, will be described. High resolution T2-weighted human brainimages have been acquired using a modified fast spin echo (FSE) imaging sequence and aprototype 4-channel array coil. The array coil was designed to fit inside a standard designhead birdcage coil and to fit closely to the contours of the human head, enablingmaximum gradient performance, minimum SAR and optimum image SNR. Using thisapproach, we have obtained high quality images with an in-plane pixel resolution of352µm ×352µm (slice thickness = 2mm) in an acquisition time of 5mins 40s for a 17slice data set. For T1-weighted MRI, a modified MDEFT sequence was developed toovercome the effects of field-focussing.


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMSfåîáíÉÇ=iÉÅíìêÉëïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


Recent advances in EEG/fMRI integrationStefan Debener 1,21 MRC Institute of Hearing Research, Royal South Hants Hospital, Southampton SO14 OYG, UK2 Clinical Neurosciences, School of Medicine, University of Southampton, Southampton, UKI2IntroductionElectroencephalography (EEG) and functional magnetic resonanceimaging (fMRI) provide complimentary information regarding thetemporal and spatial resolution of brain activity. Accordingly, theintegration of both signals holds great promise for a betterunderstanding of how cognitive processes are implemented in distinctcortical areas. Following an overview about the virtues and pitfalls ofsimultaneous and separate recordings, it will be described howartifacts unique to inside scanner EEG recordings can be removed. Itwill be shown that, following scanner artifact correction, signals ofsufficient quality can be obtained from inside scanner EEG recordings(1). Importantly, this enables the application of advanced single-trialoriented EEG analysis techniques such as independent componentanalysis (ICA, e.g., 2). A recent EEG/fMRI study on cognitiveperformance monitoring will be presented showing that thecombination of simultaneous recordings and ICA enables to study thedirect trial-by-trial coupling between EEG and fMRI data (3). Thebenefits of this new integration approach for the study of cognitivefunction will be discussed.Mining event-related brain dynamicsUsually only the evoked fraction of the EEG signal, the time-domainaveraged event-related potential (ERP), is considered and all trial-bytrialfluctuations are regarded as noise. However, trial-by-trialfluctuations of event-related EEG signals are meaningful, since theyhave been shown to be specifically correlated with the BOLD signal(4). From an information-based point of view, it is reasonable toconsider event-related EEG signals within a multi-dimensional signalspace (5). It is proposed that the consideration of EEG signals beyondthe ERP can help to investigate the coupling betweenelectrophysiological and hemodynamic properties of brain activity.Separate versus simultaneous recordingsPrevious EEG/fMRI integration attempts often focused on separatelyrecorded signals, and combined the results of both measures ratherthan the signals directly. However, separate EEG/fMRI recordingprotocols are not well suited for the investigation of a possible directcoupling between hemodynamic and electrophysiological measures ofbrain activity. More recent advances in EEG hardware and softwaredevelopments have made it feasible to simultaneously record EEG andfMRI data. Only simultaneous recordings guarantee that the identicalsensory, mental, and behavioral conditions are being studied.Moreover, simultaneous recording protocols allow to jointly exploitthe trial-by-trial fluctuations of both measures, which can revealimportant insights into the dynamic properties of cognitive function.Quality of EEG signals recorded inside the MRIThe quality of EEG data recorded simultaneously with fMRI ismassively compromised by gradient artefacts and ballistocardiogramartefacts (Figure 1). Gradient artefacts can cause amplitudes beingorders of magnitudes larger than the EEG signal, but are relativelyinvariant over time, and therefore can be subtracted out (Figure 1 B).However, the major challenge is the ballistocardiogram (BCG)artefact, which originates from the pulsatile movement of blood andthe pulsatile movement of EEG electrodes placed adjacent to largerblood vessels (Figure 1 B-D). Both channel-wise and spatial filtertechniques have been proposed for BCG-correction. Results of asystematic comparison of three different BCG-correction approaches(1) will be presented. It will be shown that a multivariate combinedchannel-wise and spatial filter approach based on ICA is superior toalternative techniques. Following appropriate BCG-correction, EEGsignals of reasonable quality can be obtained.Direct coupling between event-related EEG andfMRI in a cognitive taskGoal-directed behavior requires the continuous monitoring and dynamicadjustment of ongoing actions. We simultaneously recorded 32channel EEG and fMRI data from subjects performing an Eriksenflanker task (3). By applying ICA to the EEG signals, we found thesingle-trial error-related negativity of the EEG to be systematicallyrelated to behavior in the subsequent trial. This result reflects theimmediate behavioral adjustment of a cognitive performance monitoringsystem. Moreover, the trial-by-trial EEG measure of performancemonitoring predicted the fMRI activity in the rostral cingulate zone, abrain region known to play a key role in processing of response errors.Importantly, this result was more specific when compared to a conventionalfMRI analysis, which supports the conclusion that temporalinformation, provided by the EEG, can improve the processing offMRI signals. Furthermore, the study shows that it is possible to investigatethe dynamic coupling of non-invasively recorded EEG andfMRI data. This new approach, EEG-informed fMRI analysis, providesa powerful way for the study of brain function related to cognitiveprocessing.Figure 1: Example 10-sec traces from inside scanner EEG (A-C) andEKG (D) recordings. EEG data from six representative channels areshown before (A) and after correction for MR gradient (B) andballistocardiogram (C) artefacts.References(1) Debener S et al. submitted (2006)(2) Debener S et al. Cog Brain Res. 22, 309-321 (2005)(3) Debener S et al. J Neurosci. 25, 11730-11737 (2005)(4) Debener S et al. Neuroimage 31, S31 (2006)(5) Makeig S et al. Trends in Cog Sci 8, 204-210 (2004)


Challenges and needs - MR techniques in the neurosciencesI3V Ng 1,21 Centre for Neuroimaging Sciences, Maudsley Hospital, Denmark Hill, London SE5 8AZ; 2 Institute of Psychiatry, De Crespigny Park,London SE5 8AFMR imaging has revolutionised the diagnostic capabilities within clinical neurosciences. Investigations of the brainand spine were previously done with invasive procedures, and the findings were inferred from either vesseldisplacement or outlining of cerebrospinal fluid lined spaces. Being able to interrogate the structural, physiologicaland biochemical aspects of the central nervous system, non-invasively and efficiently, has benefitted countlesspatients, but also given us a greater understanding of the machinations of the brain in particular. However, it is safeto say that there have been and will be many occasions, when the techniques are left wanting. This talk will useexamples of some commonly encountered issues in diagnostic neuroradiology, as a starting point in the discussionof the challenges that we need to overcome in order to make even more progress.


The MRI system of the future: innovation, development and directions – a personal industrial perspectiveMatthew ClemencePhilips Clinical Science Research Support, Philips Medical Systems UKI4MRI underwent rapid development and acceptance into the medical arena and predictions on future trends in thefield are notoriously difficult. Technological innovation remains a powerful driving force in the development ofsystems; but equally in today’s markets, proven clinical utility and cost effectiveness are equal partners in the designprocess. From early machines that offered novel tissue contrasts and a service providing a safer alternative to highdose CT, modern machines face competition from other modalities which can offer faster scan times, cheaper initialcosts and, in the case of CT, much lower doses than those in the past. MRI remains a complex modality to learn anduse effectively and much of the recent development has been driven by a need to address this issue. Drawing onsignificant innovations of the last 18 years, such as the dramatic increase in available computing power and whatopportunities that has opened up, this presentation will explore how these have moved the modality forwards and theinteraction between technical development for future systems, the roles of physicists within that process, and howthat balances with the interests from those who now hold the finances which purchase the majority of systems.


I5Interactive & Real-time Body MRIDavid LomasDepartment of Radiology, School of Clinical Medicine, University of CambridgeThis presentation will address the technical aspects involved in exploiting MRI for interactive and real-time use withexamples of relevant clinical applications. The potential advantages and roles for MR as against conventional X-rayfluoroscopy will be discussed.


I6Imaging CNS inflammation: what MR can and can’t tell usNiki SibsonExperimental Neuroimaging Group, University Laboratory of Physiology, OxfordDespite the extensive use of MR imaging and spectroscopy in both basic and clinical science, it is fair to say thatthere are still numerous changes in the MR signals that we don’t understand. Nevertheless, by taking wellcharacterised experimental models and combining MR and non-MR approaches, we hope to begin to shed light onsome of the outstanding questions.In Oxford we have, to a large extent, taken a reductionist view of the problem, and have spent much of our timefocusing on very simple experimental models with the aim of teasing apart the different components ofneuropathology. Expression of pro-inflammatory cytokines has been implicated in the pathogenesis of a broadspectrum of neurological diseases, and we have been investigating the effects of specific pro-inflammatorycytokines, in isolation, on CNS integrity. To date, we have determined that injection of TNF- into the brain resultsin an early and profound reduction in cerebral blood volume detected by MRI, which can be reversed by endothelinreceptor antagonists and is dependent on activation of the TNFR2 pathway. We have also established that TNF-induces delayed changes in brain water diffusion, which may reflect perturbations of metabolism. The TNF-inducedchanges in CBV led us to investigate the effect of ET-1 per se in the CNS, with the intriguing finding of anacute increase in T1 relaxation in the injected region, which we were also able to induce with a model of NMDAinducedexcitotoxicity. Similar changes have been reported in a very small number of studies of cerebral ischaemia,but the underlying causes have remained unclear. We now find that this acute T1 change appears to reflect earlyastrocyte activation.In addition to these simple models of cytokine-induced inflammation, we have also been studying more complexmodels of CNS neuropathology with the aim of identifying early MR-markers of disease activity. In particular, ourstudies with a well characterised model of MS in the rat brain have revealed that early perfusion changes appearbefore overt pathology is visible with conventional, clinically-used MRI methods. This finding has recently beenborne out in MS patients. Moreover, we have demonstrated that induction of a systemic inflammatory response willreactivate this same MS-like lesion in the brain, with implications for the management and treatment of MS patients.In the course of these studies, however, it has become clear that there are many MR-invisible, and indeed clinicallysilent, pathological events in the brain that it would be of enormous benefit to be able to detect. With this in mind,we have been working with chemists both in Mons (Belgium) and Oxford, to develop novel targeted MR contrastagents that can identify “invisible” lesions and early signs of neuropathology.Overall, these studies have revealed new and previously unrecognised aspects of inflammation biology in thecontext of the CNS. Yet, it is undoubtedly the case that the more we think we understand, the more we realise howmuch more is out there to discover.


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMSpéÉÅá~ä=fåîáíÉÇ=iÉÅíìêÉïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


The EU Physical Agents (EMF) Directive and its Impact on MRI:Prudent Precaution or Straight Bananas?I8Stephen KEEVILDepartment of Medical Physics, Guy’s and St Thomas’ NHS Foundation Trust, Guy’s Hospital, London, SE1 9RTIntroductionThe Physical Agents (Electromagnetic Fields) Directive (1) wasadopted by the European Union in 2004. Member states are obliged toincorporate it into law by April 2008, and in the UK this will take theform of the Control of EMF at Work Regulations 2008. The Directiverestricts occupational exposure to electromagnetic fields (EMF) withfrequencies up to 300 GHz. It contains ‘exposure limit values’,expressed in terms of induced current density in the head and trunk atlower frequencies and of specific absorption rate (SAR) at higherfrequencies. There are also supplementary ‘action values’ expressed inmore easily measurable terms to ensure compliance with the limits.These values are taken from guidelines issued some years ago by theInternational Commission on Non-Ionising Radiation Protection (2).This lecture will describe the potential impact of the Directive onMRI, the scientific basis for the limits, and the current state of playwith efforts by the MR community to have the Directive amended.Impact on MRIAll three of the EMF frequency ranges used in MRI are in principlewithin the scope of the Directive. The table summarises relevantaction values and exposure limits in each frequency range, togetherwith estimated maximum occupational exposures for MRI workers.Static magnetic field. The draft Directive contained a static magneticfield limit of 2 T, but this was removed during negotiation; the actionvalue of 200 mT is of little practical consequence. However,movement through the temorally static but spatially varying fieldexposes staff to a slowly time-varying field, and it seems likely thatthe induced current density limit will be exceeded by a considerablemargin (3, 4). These limits are absolute, with no scope for timeaveraging or exemption for brief exposure.Switched Gradients. The exposure limit in the 100s – 1000s Hz rangewill create an ‘exclusion zone’ around the bore opening while imagingis taking place. The extent of this zone will depend on magnet andgradient coil design and on the pulse sequences being used. Againthere is no scope for time averaging in this frequency range, so staffwill not be permitted to enter the zone during imaging.Radiofrequency. Whilst the exposure limits are low, there is allowancefor spatial and temporal averaging, and it seems unlikely that MRworkers will exceed them. However, the localised SAR limits for thelimbs may be problematic in some cases, such as interventional MRI.Taken together, these limits threaten current clinical use and futuredevelopment of MRI. It may be impossible to work close a scannerduring imaging to provide patient care or monitoring or to performresearch procedures. Interventional MRI may be prohibited essentiallyin its entirety. Issues with the static field may impact on installation,cleaning, maintenance and patient set-up, as well as situations inwhich staff need to work close to patients or experimental apparatus.Basis of Exposure LimitationThe Directive aims to protect workers from ‘known short-termadverse effects’ arising from exposure to EMF. In the frequency rangeup to 100 kHz, there is little scientific evidence for such effects. In theICNIRP paper (2), occurrence of harmless physiological effects suchas magnetophosphenes at tens of hertz has been taken as indicative ofpossible adverse health effects over the whole frequency range. It isassumed, with no evidence at all, that these hypothetical effects followthe same frequency response as peripheral nerve stimulation, but withan onset threshold about two orders of magnitude lower. ICNIRP hasmore recently described the guidelines as having been ‘written manyyears ago, and... now under review’ (5), while in the UK, NRPBdescribed the limits as ‘a cautious approach… to indicate thresholdsfor adverse health effects that are scientifically plausible’ (6). There isno basis for curtailing MRI on the strength of this evidence.Prospects for AmendmentProfessional bodies involved with MRI have been campaigning forseveral years for amendment of the Directive prior to implementation.Little progress was made until September 2005, when a press briefingwas held with the assistance of the charity Sense About Science. Thisattracted significant media coverage and led to questions being askedin the House of Lords. The relevant government minister becamepersonally involved, and a working party has been established withthe Health and Safety Executive to examine impact on MRI and howit might be mitigated. The European professional bodies andmanufacturers have also been active, and in March 2006representatives met with the EU Social Affairs Commissioner. Afurther working party has subsequently been established by theEuropean Commission, with a very similar remit. Lobbying isbeginning in other EU members states, and this is very important if weare to prevent implementation without amendment in 2008.In late 2005 the House of Commons Science and Technology SelectCommittee announced an inquiry into the issue as a case study ingovernment handling of scientific advice. The committee’s recentreport (7) is extremely critical of the Directive and of various nationaland international agencies for their handling of scientific evidence andof representations made by the MR community.References(1) Directive 2004/40/EC. Official Journal of the European Union L159 of 30 April 2004, and corrigenda L 184 of 24 May 2004.(2) ICNIRP Health Physics 74, 494-522 (1998).(3) Liu F et al J Magn Reson 161 99-107 (2003).(4) Crozier S and Liu F Prog Biophys Molec Biol 87 267-278 (2005).(5) ICNIRP Health Physics 87 197-216 (2004).(6) NRPB Doc of the NRPB 15 (3) (2004).(7) House of Commons Science and Technology Committee Watchingthe Directives: Scientific Advice on the EU Physical Agents(Electromagnetic Fields) Directive (2006).Static magnetic fieldFrequency Exposure limit Action value formagnetic fluxdensity0 Hz None 0.2 T 3 T (clinical)7 T (research)


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMSpìÄãáííÉÇ=lê~ämêÉëÉåí~íáçåëïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


Faster Dixon Fat-Water Imaging With Slice Multiplexed PulsesO1Kuan J LeeAcademic Radiology, University of Sheffield, Sheffield, UKIntroductionPhase sensitive imaging for fat-water separation was first proposed byDixon (1): for each slice, two images are acquired, one with fat andwater “in-phase”, and the other π “out-of-phase”. However, theoriginal method could fail if there were significant fieldinhomogeneities which introduced additional phase shifts. Toovercome this, Glover (2) suggested acquiring an additional 2π image.Unfortunately this leads to a threefold increase in minimum imagingtime.Recently we introduced a new method for multislice imaging usingthe “multiplex pulse” (3). The multiplex pulse combines severalcomponent pulses into a single pulse which has the duration of asingle component pulse. Each component pulse simultaneously selectsa different slice which is rephased with a different slice gradientrefocusing lobe, thereby avoiding aliasing. The aim of this work is toshow that multiplex pulses can be used to speed up Dixon fat-waterimaging.MethodsA 4 ms three-slice multiplex pulse (Fig 1), the slices of which arerephased by +0.15, -0.5 and -0.85 of the slice select lobe, wasdesigned as described in Ref (4). This was used in a spin-echosequence with a non-slice selective refocusing pulse; the schematic(not to scale) is shown in Fig 2. First, images were collected atvarying TE with fat and water phantoms in order to determine the inphaseecho time for each slice. Relative to the 0.5-rephased slice, the0.15 pulse’s slice was measured to be in phase 3.1 ms earlier and the0.85 pulse’s slice was in phase 1.9 ms later (data not shown). Thetiming of the readout gradients was adjusted to take this into accounti.e. all echoes to have the same relative fat-water shift.RF/SignalSSPEFEwas a non-minimum power delayed focus DBURP1 pulse). This delayis actually helpful because the time interval can be used to switch theread gradient. Interestingly, it indicates that the relative in-phase echotimes may be adjusted e.g. simply by shortening or lengthening theexisting pulse. This may be useful when adapting a sequence for useon scanners of different field strengths. Adjustment will also benecessary for other variants of Dixon imaging requiring different fatwatershifts e.g. the IDEAL method (-π/6, π/2, 7π/6) (5), or the twopoint POP (0, 135°) (6).The proof-of-principle sequence presented here uses one non-sliceselective refocusing pulse. The next step will be to implementmultiplex fat-water imaging with FSE sequences. The sequence willbe similar to GRASE (7), where extra gradient echoes are collected inbetween refocusing pulses. However, with multiplex pulses, the extragradient echoes correspond to extra slices, instead of extra k-spacelines within the same slice.It is hoped also to implement this method for fat suppression at lowfields (


An Insertable, Shoulder-Slotted Gradient and Shim Set for Dynamic ShimmingMichael POOLE, Richard BOWTELLSir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, University Park,Nottingham, England, NG7 2RD (correspondance to Michael Poole: ppxmp@nottingham.ac.uk)O2IntroductionMagnetic susceptibility induced field distortions may cause signal-lossand geometric distortions in MRI, particularly in EPI. Dynamicshimming [1] aims to ameliorate magnetic field distortions for eachslice of a multi-slice acquisition. The efficacy of field homogenisationis greater than conventional shimming of the whole volume, butimplementing dynamic shimming requires shim coils of high strengthand low inductance to achieve the necessary field switching times.Here we present an insertable set of gradient and shim coils with lowinductance and high efficiency designed for use in dynamic shimmingwith an arbitrary-surface, boundary element method (BEM) [2]. Thismethod of design was used because the coils were to have diametersfalling between 380 and 470 mm, and therefore required cut-outs toaccommodate subjects’ shoulders. Half size versions of the Z2 and ZXcoils have been built and tested and a full-size shim set is currentlyunder construction.MethodsA method where the conducting surface on which the coils are woundis discretised into triangular elements [2] was used to design coils withthe geometry shown in Fig. 1. Thsi geometry is a 800 mm longcylinder with a diameter that is different for each coil. Table 1 lists thecoils that make up the shim set in radial order. The order and degreeof the associated spherical harmonic is given alongside the radius thatthe coil is designed at and the cartesian form of the spherical harmonicused as the function for the target field. Z0 (P) and (S) are the primaryand shield coils of the Z0 shim respectively. The shoulder slots are220 mm long and are 150 mm wide. In Fig. 1 half of the discretisedsurface is shown in blue, and has a total of 5456 triangular elements.The region of uniformity is a 160 mm diameter spherical volumecontaining 93 target field points. The set consists of X, Y, Z gradientsand all 2nd order shim coils as well as a shielded Z0 coil. The Z0 coilwas designed on a 320 mm long cylinder with a method that expressesthe current density as a weighted set of sinusoidal harmonics [3]. TheBEM method allows the optimisation of the RMS field error,inductance and resistance whilst imposing torque-balancing. The fieldgenerated by the coil was modelled by applying the Biot-Savart law tothe wire-paths, and the inductances and resistances were estimatedusing FastHenry© [4], a multipole impedance calculation. Theelectromagnetic properties of the coils will be tested when theconstruction is complete.Table 1. A list of the coils that make up the shim set.Figure 2. Wire paths for half of the a) X gradient and b) Z2 shimcoils.Table 2. The efficiency, inductance, resistance, figure of merit, andthe minimum gap between wires for all the gradient and shim coils.Figure 1. The geometry of the coils.ResultsWire-paths of the X gradient and Z2 shim coils are shown in Fig. 2.The red wires on the X gradient indicate reversed current flow withrespect to the blue wires. The current direction for the Z2 shim isshown by arrows. The efficiency, inductance, resistance, and figure ofmerit (FOM) η 2 /L of all the coils are shown in Table 2. Inductancesand resistances were modelled with 3 mm diameter wire usingFastHenry©.ConclusionAn insertable gradient and shim set with shoulder slots containing allcoils up to 2nd order has been designed using a boundary elementmethod [2]. These coils have high efficiency, low inductance, lowresistance, and are torque-balanced. The full-size set is currently inproduction and our ultimate aim is to implement dynamic shimming[1] at 3T and 7T with it.References[1] A. M. Blamire, D. L. Rothman, T. Nixon. Magn. Reson. Med. 36,159-65 (1996). [2] R. A. Lemdiasov, R. Ludwig. Concepts Magn.Reson. B. 26B, 67-80 (2005). [3] M. Poole, R. Bowtell. BCISMRM,2005, p. 3. [4] M. Kamon, M. J. Tsuk, J. K. White. IEEE Trans. onMTT. 42, 1750-1758 (1994).AcknoledgementsWe wish to thank Magnex Scientific Ltd. (part of the Varian, Inc.group) for the support of a studentship


Pitfals in the development of Ultra Short Echo (UTE) imaging at high field strengthO3Alexandr Khrapichev 1 , Nicola Sibson 1 , Matthew Robson 2 , Andrew M. Blamire 3 ,1 Dept of Physiology, 2 Oxford Centre for Clinical Magnetic Resonance, University of Oxford;3 Newcastle Magnetic Resonance Centre, University of NewcastleIntroductionUltra short echo (UTE) imaging is the class of experiments whichattempt to acquire data with effective echo times ranging from 50-150µs (1). Such methods allow imaging of fast relaxing spins withtransverse relaxation times above 40µs. Recently UTE sequenceshave been demonstrated on clinical MRI systems and have revealedinteresting image contrast across a number of body areas (1, 2). Ofparticular interest to us is the suggestion that image contrast seenusing UTE in a patient with multiple sclerosis is reporting on thedemyelination process of this disease (2).In order to determine the fundamental nature of the UTE signal seenin vivo requires direct correlation of the UTE signal with quantifiedlevels of myelin in the tissue. This can be achieved by applying UTEsequences in small animal models with direct comparison tohistological specimens. We are therefore developing UTE for highfield, small animal work.To achieve such short echo times requires self-refocussed RF pulsesand radial k-space sampling (akin to back-projection techniques) toeliminate delays associated with slice refocussing, read-dephasing andphase encoding. Further, since significant relaxation may occurduring data sampling itself, limiting the achievable resolution, it isdesirable to work with short data sampling time (of the order of T 2 ).This in turn requires high readout gradient strength (ideally maximumgradient) and large sampling bandwidth. The manufacturingtolerances for small scale gradient sets are proportionally much largerthan in clinical systems, and the gradient fields produced areincreased, so gradient system performance (e.g. gradient cross terms,residual eddy current effects, etc) may have a more dominant effect onimage formation than in clinical systems. In this paper we report thetechnical challenges associated with implementing UTE at high fieldon a small bore system and discuss measurements and correctionschemes to achieve UTE data.MethodsThe UTE sequence (3) shown in figure 1(a) was implemented on a 7TVarian Inova spectrometer running vnmr6.1c software. The system isequipped with a Magnex gradient system capable of 175 mT/m in arise time of 80µs. Imaging experiments were performed using awater-filled Varian microimaging phantom (figure 2(a)) and abirdcage RF coil. Each radial FID consisted of 64 data points sampledat 300kHz. Gradient performance was assessed using the sequenceshown in figure 1 (b) (4) where gradient strength is determined fromthe phase accumulation during the delay t p . A water-filled 4mm i.d.NMR tube aligned coaxially with the magnet bore was used toapproximate a point-source phantom in the transverse plane. Datawere collected at isocentre to measure B o eddy current effects and atseveral location along the X and Y axes to measure spatiallydependent eddy current terms.Image processing was performed using regridding of the radial K-space data using standard algorithms with corrections for gradientperformance included.ResultsAlthough the system is capable of routinely obtaining high qualitydata using fast imaging sequences such as EPI, basic image qualityusing the UTE sequence without detailed characterisation of gradientperformance was poor (figure 2(b)).Characterisation of the gradient system using the sequence in figure1(b) demonstrated a significant B o eddy current term from pulsing theY gradient. This is shown in figure 2(c) for gradient pulses between10 and 90mT/m. This response was found to be well described by amathematical function where the initial slope is a function of therising field gradient, whereas the decay is described by an exponentwith a time constant (~25µs) characterising the physical property ofthe system. In the image, this B o term leads to the phase role across k-space, which produces the smearing artefact seen in figure 2(b).Phase correction of each FID to compensate for this unwanted phaseaccumulation results in the image in figure 2(d).The gradient waveforms on our system run at constant slew rate,which results in gradient rise time dependent on the required demand.This therefore varies with each radial projection and each individualk-space trajectory reaches a plateau at the different acquisition point,causing a disortion of k-space around the origin, as shown in figure2(e) and is the cause of the remaining artefacts seen in figure 2(d).When k-space trajectory is corrected the image of figure 2(f) isproduced.RFG zG rADCRFG rADCFigure 1 (a) UTE sequence, (b) Gradient mapping sequenceaceFigure 2: UTE imagesConclusionThe UTE sequence was implemented at 7T on a small bore MRIsystem. The effect of very short time constant B o eddy currents whichare mainly active during gradient switching was measured and foundto be the cause of major artefacts in the UTE data. Since the B o fieldaffects all spins equally, post-processing corrections can be simplyapplied to unwrap the unwanted phase accumulation. Correction fordistortion of each radial projection due to gradient rise time effects ledto a further improvement in image quality.References(1) Waldman et al. Neuroradiology 2003; 45(12) p887-892(2) Gatehouse PD, Bydder GM. Clin Radiol. 2003; 58(1) p1-19(3) Robson, M. D. et al. JCAT 2003; 27(6) p825-846.(4) Goodyear, D. J. et al. J Magn Reson 2003 ; 163, p1-7.t p


An insert system for advanced imaging research on clinical MR imagersO4Martyn PALEY 1 , Eugeny KRJUKOV 1 , James WILD 1 , Michael LAMPERTH 2, Ian YOUNG 21 Academic Radiology, University of Sheffield; 2 Mechanical Engineering, Imperial College London;IntroductionMany clinical magnetic resonance systems do not have access toadvanced techniques such as ultra-short echo time, multi-nuclear ormagic angle spectroscopy and imaging. These methods have thepotential to dramatically improve clinical diagnoses in certainapplications. The purpose of this project is to develop an insertupgrade for standard imaging systems which provides access to thesenew research applications without disturbing routine clinicaloperation. The initial targets are improved orthopaedic imaging with1-H and 31-P capability and the enabling of hyperpolarised 3-Heimaging for neonatal and paediatric lungs.MethodsTo provide minimum disturbance to the clinical scanner hardware, theonly original system component which is used is the magnet. A newspectrometer and software system based on National Instruments (NI)hardware and Labview software has been developed and integratedwith RF (Tomco, Au) and gradient amplifiers (Analogue Associates,UK) and an insert gradient and RF coil system to provide a completeindependent imaging system. The new data acquisition system isbased on a PXI chassis with an embedded Windows XP controllerwhich can acquire 32 channels simultaneously at 14 bit resolution atup to 2.5 MS/s. 8 channels of 16 bit D-A conversion provide thecontrol, RF and gradient waveforms at a maximum output rate of 1MS/s. Waveforms are generated using a standard NI analoguewaveform editor which provides a very powerful graphical sequencedevelopment tool. The set of control waveforms are read into theLabView MR Virtual Instrument (VI) and displayed in real time as thesequence progresses. Signals acquired from all 32 channels are alsodisplayed in real time. Limited 2DFT reconstruction is built into thesystem but raw data is automatically stored so that off-linereconstruction using e.g. Matlab is possible for more advancedtechniques. The RF spectrometer is based on modular RF componentsfrom Mini-Circuits and is broadband from 250 KHz up to 512 MHz.So far only four of the 32 receive demodulators have been built andonly a single quadrature channel used for MR data acquisition.The volume and surface gradients were initially tested on a 3T MRsystem (Philips Intera, NL) by phase mapping with and without aweak DC current applied. The gradient and RF leads were filtered andfed into the scanner room through a waveguide. The completedsurface gradient coil coil is shown in Fig. 3.Fig 3For initial testing, a single 100m loop TR RF surface coil wasemployed.ResultsFig 4 shows phase maps from the X volume gradient coil at iso-centreand at the edge of the predicted linear region with a DC current of 1Aapplied. The phase pattern is asymmetric as the test object was offsetto image the edge of the field in one direction and the coil was notquite horizontal in the magnet. The gradient field is linear as expectedfrom simulations.Fig 4Figure 5 shows a phase map from the Z surface gradient coil with400mA current applied and illustrating close agreement with thesimulation of figure 1b. The quasi-linear region at the center of thecoil is used in conjunction with a TR surface coil and will also be usedwith gradient warp correction.Fig 1A conventional 3 axis gradient coil set consisting of a Maxwell pairfor Z and shortened Golay saddle coils for X and Y axes was woundwith 10 and 5 turns respectively of 2mm diameter wire on a 315mmdiameter rigid former, large enough to accommodate the adult headand knee (Fig 1). The mean spacing of the Maxwell pair was 270mmand the spacing of the inner turns of the Golay coils was 122mm. Theresistance of the Maxwell pair was 0.13 Ohms and the Golay coils0.21 Ohms. The coil weight was 5Kg and was easily located on thepatient table where it was secured using Velcro straps. Surfacegradients provide flexibility to move e.g. a knee ligament to the magicangle within the magnet (1). A planar 3 axis surface gradient coil setwas designed and built using 2mm diameter wire. The coil windinglocations and current directions for the Z-gradient are shown in Fig 2aand the calculated fields lines in Fig 2b. The length of the conductorswas 200mm. The return loop design provides a certain amount ofshielding below the coils although this is not really necessary as thesmall physical size means there is little interaction with the magnetstructures. Similar planar designs were produced for the x and y axes.2a 2bFig 5Image data sets have now been acquired with a single channel of thedata acquisition system. The acoustic noise of the insert has not yetbeen quantified but subjectively is much quieter then the whole bodygradient system. No significant motion of the gradient coil sets wasobserved with a maximum current of 55A. Eddy current interactionwith the magnet has not yet been examined but is expected to beweak. Further development will include multi-channel RF coils forparallel imaging of 1-H, 31-P and 3-He nuclei.ConclusionInitial testing has shown that a fully independent MR insert systemcan provide access to research methods on a standard clinical imagerwith no disturbance of normal operation and rapid changeoverbetween research and clinical scan modes. Further work willimplement a full range of advanced methods for multi-nuclearimaging and spectroscopy. A LabView based MR compatible motionsystem is also in development which will allow limbs to be rapidlymoved to the magic angle under image guidance.References(1) Cho et al., J. Magn. Reson. 1991; 94: 471-485This project is funded by DH-NEAT Grant G019


Optimisation of Steady State Free Precession Sequences for Hyperpolarised 3 He MRIJ. M. Wild 1 , K. Teh 1 , N. Woodhouse 1 , M.N.J. Paley 1 , N. de Zanche 2 , L. Kasuboski 31 Academic Radiology, University of Sheffield, Sheffield, United Kingdom, 2 Institute of Biomedical Engineering, Zurich, Switzerland,3 Philips Medical Systems, Cleveland, OH, United States.O5Introduction In previous work the magnetizationresponse of hyperpolarised (HP) 3 He gas to a SSFPsequence was simulated using matrix operators [1]. Inthis work the theory is compared with NMR and MRIexperiments and the results used to optimize SSFP forin-vivo MRI.Methods3 He MRI was conducted on a 1.5 Tmagnet (Philips, Eclipse) with T-R capabilities at 48.5MHz. Gas was polarised to 30% with rubidium spinexchangeapparatus. A low pass quadrature TR birdcagecoil was used for phantom work, a flex twin saddle coilfor in-vivo work. A 2D SSFP sequence wasprogrammed with TE/TR = 5/10 ms and had an α/2 –TR/2 start-up pre-pulse. Gas phantom studies wereconducted with 1 l plastic bags filled with 100 ml 3 Heand 900 ml N 2 . In-vivo ventilation imaging wasperformed with a breath-hold of a gas mixture of 300 l3He/700 ml N 2 . The performance of SSFP as a functionof flip angle was investigated in phantoms and in-vivoand compared to the performance of optimised 2Dspoiled gradient echo [2].Results & Discussion Phantom experiments,confirm predicted theory of increased SNR with SSFPthrough use of higher flip angles (α) when compared tooptimized spoiled gradient echo (SPGR) - (Fig. 1-3).SSFP with an optimized α=20°, and 128 sequentialphase encodes, showed higher (1.7 x) SNR thanoptimized SPGR (α=8°) –also see Fig.4. Simulationsand experiments show some compromise to SNR andPSF at high α due to diffusion dephasing from thereadout gradient (Fig.3). In 3 He NMR experiments,diffusion dephasing can be mitigated, and the effectiveT2 is long (1 s). Under these circumstances SSFPbehaves like CPMG with sin(α/2) weighting of SNR(Fig. 2). Experiments and simulations were alsoperformed to characterize off-resonance behavior of theSSFP HP 3 He signal. Banding artifacts were observed insome in vivo SSFP images, close to the diaphragmwhere B 0 inhomogeneity is highest. Despite theseartifacts, higher (1.6 x) SNR was observed with SSFPin-vivo when compared to SPGR (Fig. 5). The predictedtheory of increasing SSFP SNR with increasing α wasobserved in the range α =10°-20° (Fig. 5c-d) withoutcompromise to image quality through blurring causedby excessive k-space filtering [3].Conclusion It has been demonstrated theoreticallyand experimentally that SSFP can provide high spatialresolution images of lung ventilation at breath-hold withHP 3 He at 1.5 T with the potential for higher SNR thancurrently used low flip angle SPGR methods. AlthoughHP gas M 0 is not renewable, under the right conditionsof long effective T2, a pseudo-steady state can beachieved within the time course of an MR imagingexperiment. When effective T2 is shortened bydiffusion dephasing, SSFP yields a decayingmagnetization response. Nevertheless this non-steadystate can still be used to good effect in-vivo as itprovides higher SNR for a given level of blurring in thePSF than an optimized SPGR sequence.SSFPSpoiledFig. 1 Simulations of SSFP & SPGR SNR as a function of αSignal intensity at n=640.140.120.10.080.060.040.02SSFP low b, in vivo, T2 eff 76 msSSFP low b, phantom T2 eff 68 msLow b spoiled00 10 20 30 40 50 60α°Fig. 2 SSFP NMR of HP 3 He - data shows a good fit to theorySSFP n= 64 signal intensity normalized by M 0 (a.u.)SSFP intensity at n=64 (a.u.)807060504030201025020015010050sin(α/2)00 20 40 60 80 100 120 140 160 180 200α°Fig. 3. SSFP phantom imaging performance with flip angleexperiment05 10 15 20 25 30 35 40α°Fig. 4. SSFP phantom imaging versusSPGR –note reduced dephasing due tosusceptibility gradientsabFig 5a coronal SSFP image acquired with α = 20°. Fig 5b is the SPGR imagefrom the same slice acquired with the same volume of gas. Note thecharacteristic banding in the SSFP image –arrow.Figs 5c and 5d are coronal SSFP images from the same slice in a secondsubject, acquired with α = 10° and α = 20° respectively. SNR increases witha without a noticeable change in the level of blurring.References. [1] Proc. Intl. Soc. Man. Reson. Med. 13(2005) 50. [2] Magn Reson Med. 2004;52(3):673-8 [3]Magn Res Med 2002;47:687-695AcknolwledgementsEPSRC, UK. Grant # GR/S81834/01(P). Spectra gases, GE Healthcaretheorycd


Improved Artifact Correction for Combined EEG/fMRIO6Karen J. Mullinger 1 , Paul S. Morgan 2 , Richard W. Bowtell 1 ,1 Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, Universtiy of Nottingham; Academic Radiology,University of Nottingham;IntroductionSimultaneous EEG and fMRI has been made possible by the developmentof EEG hardware and correction methods that allow artifactsgenerated by the scanner gradients and cardiac pulse to be characterisedand then subtracted [1]. Correction methods generally rely on thegeneration of reproducible gradient artifacts and accurate detection ofcardiac R peaks. Here, we describe the implementation of two methodologicaldevelopments aimed at improving the reliability of artifactcorrection: synchronization of EEG sampling to the MR scanner clockand use of the scanner’s physiological logging in identifying R-peaks.MethodsfMRI and EEG data were acquired simultaneously using a PhilipsIntera Achieva 3.0 T MR scanner and a BrainAmp MR EEG amplifier,Brain Vision Recorder software (Brainproducts, Munich) and theBrainCap MR electrode cap with 32 electrodes (5 kHz sampling rate).A standard EPI sequence was implemented (64×64×20 matrix,3.25×3.25 mm 2 in-plane resolution and 3mm slice thickness). Cardiacand respiratory cycles were simultaneously recorded using the scanner’sphysiological monitoring system (vector cardiogram (VCG) [2]and respiratory belt) whose output is sampled at 500 Hz. Triggersmarking the beginning of each volume acquisition were recorded onthe EEG system. Data was recorded for 6 minutes (180 volumes) for3 different situations: (i) the EEG sampling and scanner waveformswere synchronised by driving the BrainAmp clock using a 5 kHzsignal derived from the 10 MHz MR scanner clock (TR = 2 s); (ii) theEEG sampling was not synchronised to the scanner (TR = 2 s); (iii) theEEG and MR clocks were synchronised, but a TR of 2.0001 s which isnot a multiple of the scanner clock period, was employed. In eachexperiment the time between repetitions of scanner waveforms wasTR/20 ≈ 100 ms, so that gradient artifacts occurred at multiples of10Hz.Off-line EEG signal correction was based on averaging and then subtractinggradient and pulse artifacts, as implemented in Brain VisionAnalyzer [1]. Gradient artifact correction employed an average artifactwaveform of 2 s duration, formed from the average of the 180 TRperiods,using the scanner-generated markers. Pulse artifact correctionwas carried out based on R-peak markers derived from the ECGor VCG traces.RatioSynchronised, TR=2sSynchronised, TR=2.0001s0.50.40.30.20.100.050.04Not synchronised, TR=2s102030405060708090100110120130140150Frequency (Hz)imperative to employ a TR which is a multiple of the scanner clockperiod in order to achieve good correction. The ECG traces producedbefore and after gradient artifact correction are shown in Figure 2.Figure 3 shows a corresponding VCG trace. Figure 4 indicates thatusing the VCG rather than ECG trace to define R-peak markers providesa similar/slightly improved level of pulse artifact correction.Use of the VCG consequently offers an alternative approach wheredifficulties in obtaining an adequate quality ECG trace occur.Voltage (a.u.)Voltage (Micro Volts)Voltage (Micro Volts)20000.0015000.0010000.005000.000.00-5000.00-10000.00-15000.000 2.5 5 7.5 10 12.5 15 17.5-8000.00-8200.00-8400.00-8600.00-8800.00-9000.00-9200.00-9400.00-9600.00-9800.000.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5Time (s)Figure 2: Typical ECG trace before (top) and after (bottom)gradient correction.3500300025002000150010005000-500-1000-15000.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5Time (s)Figure 3: VCG trace obtained straight from the scannerwithout any correction.4Ratio0.030.020.010102030405060708090100Frequency (H z)Figure 1: Ratio of signal power at 10 Hz harmonics forcorrected/uncorrected data, averaged over all channels.Error bars show the standard deviation of signal over allchannels. B shows the same data as A, but at a differentscale.110120130140150Voltage (Micro Volts)3211 2 34Results and DiscussionFigure 1 shows the ratio (corrected/uncorrected) of EEG power atmultiples of 10 Hz, indicating that synchronisation of the EEG samplingto the MR scanner clock, leads to better correction of gradientartifacts. The reduced residual artifact at high frequencies will beparticularly advantageous for measuring gamma band activity in thescanner. Figure 1 also shows that even with synchronisation it isFrequency (Hz)Figure 4: EEG artifacts on channel TP10 due to theballistic effect before (black) and after correction usingVCG (red) and ECG (green).References1. Allen et al. Neuroimage 8:229-239,1998.2. Chia et al. JMRI, 12:678-688,2000.


Voxel-Based Investigation Of White Matter Changes In Normal Ageing Using Diffusion Tensor Imaging And TractographyThomas R. Barrick 1 , Rebecca A. Charlton 1 , Michael O’Sullivan 1 , Hugh S. Markus 1 , Chris A. Clark 11 Centre for Clinical Neuroscience, St. George’s, University of London, United Kingdom.IntroductionIn recent years the importance of changes in white matter integrity in normal ageing has been reported (1). Diffusion tensor imaging has enabled us to investigatewhite matter structural integrity in vivo, and region of interest approaches have lead to debate about the regional specificity of white matter decline in ageing, inparticular that the frontal lobes are preferentially affected (2, 3, 4). Voxel-based statistical analyses allow us to investigate anatomical distribution of white matterdamage across the ageing brain. Here we present data from a large sample of normal older adults using voxel-based statistical analysis. We analyse 3D meandiffusivity (MD) and fractional anisotropy (FA) maps, and investigate MD and FA perpendicular to the mid-sagittal plane (using 2D column maps, 5) and throughcoronal slices (1D coronal slice profiles, 5) in standard space. Post-hoc streamline tractography was performed through significant clusters, in order to investigatethe extent of white matter pathways affected in the ageing process.MethodsMRI data acquisition and preliminary analysis: 99 healthy adults between 50 and90 years were scanned on a 1.5T GE Signa MRI system (max. field gradientstrength 22 mTm -1 ). Diffusion tensor imaging (DTI) was achieved using a singleshot echo planar sequence with 12 diffusion sensitised directions as describedpreviously (6). Two interleaved acquisitions comprising 25 slices each providedwhole brain coverage (resolution: in plane 2.5mm; through plane 2.8mm). Eachsubject’s DTI was normalised to standard space by affine transformation (7) andresampled to contain 1.0 mm 3 voxels. MD and FA images were generated foreach subject DTI and voxels corresponding to CSF were removed prior to furtheranalysis.Generation of MD and FA (2D) column and (1D) coronal slice profile maps: Foreach subject mean MD and FA values were computed perpendicular to the midsagittalplane through white matter of both cerebral hemispheres in standardspace. This generated pixelated 2D column maps (5) of MD and FA through theentire brain of each subject. In addition, coronal slice profile maps (5) weregenerated by computing the mean values of MD and FA through white matter ofboth cerebral hemispheres for each coronal slice in standard space.Statistical analysis: 3D and 2D data were smoothed (3D 3mm FWHM; 2D 2mmFWHM) and statistical analysis was performed in SPM2 using a linear correlationmodel. MD and FA 3D, 2D column and 1D coronal slice profile maps werecorrelated with age. In all analyses, multiple comparisons correction was appliedusing a family wise error at p < 0.05 corrected. Significant positive and negativecorrelations are represented using hot and cold colours, respectively, in thesignificance maps.Post-hoc tractography: Subvoxel streamline tractography was performed asdescribed previously (6). Streamlines (vector step length 1.0mm, terminationcriteria FA < 0.08) were initiated from the centre of every voxel in each Figure 1individual normalised DTI dataset. Only streamlines passing through significant3D clusters determined from the 3D MD and FA statistical analyses were retained and wereconverted to binary images. These binary images were statistically analysed on a voxelwise basisusing the binomial test including a multiple comparisons correction (p < 0.05).(i)(ii)(iii)(i)t-statistic(a) Significant MD increase-120 -100 -80 -60 -40 -20 0 20 40 60 80z=0mm151050-5-10-15Talairach y-coordinate (mm)t-statistict-sttic1-1(a) Mean Diffusivity-120 -100 -80 -60 -40 -20 0 20 40 60 80t-statistic12.92-12.92(b) Significant FA decreasez=0mmO7151050-5-10-15Talairach y-coordinate (mm)(b) Fractional AnisotropyRight Left Right LeftResultsWhite matter coronal slice profile significance maps showed MD age to exhibit a significantpositive correlation with age, such that MD increased with age across the entire brain (Figure 1ai).White matter MD column maps were also significantly correlated with age across the brain (Figure1aii). Voxel-based 3D statistical analysis of the MD maps again revealed extensive, significantvoxel clusters representing positive correlations with age across the white matter and in thethalamus (Figure 1aiii, also see Figure 2ai for surface rendering of significant voxel clusters). Theselocal changes were shown to affect white matter pathways across the entire brain in the post-hoctractography statistical analysis (Figure 2aii).Coronal slice profile significance maps showed FA to have a significant negative correlation withage, such that FA decreased with age across the entire brain (Figure 1bi). White matter FA columnmaps also showed a significant negative correlation with age across the brain (Figure 1bii) withgreater significance found in the anterior frontal lobe. Voxel-based statistical analysis of the FAmaps revealed extensive negatively correlated peri-callosal clusters located predominately in thefrontal lobes of both hemispheres (Figure 1biii, and Figure 2bi). Although significant FA clusterswere small in comparison to MD clusters post-hoc tractography revealed large networks of frontofronto,fronto-parietal and fronto-occipital pathways were affected by age-related FA changes(Figure 2b).(i)z=0mm(ii)Figure 2z-statistic9.950.0Discussion These results suggest decline of white matter integrity across the whole brain, but with the anterior frontal lobes being particularly affected. This is inkeeping with results from other authors (8). In addition, tractography results suggest that localised peri-callosal white matter degeneration may affect pathwayconnectivity throughout the whole brain. Consequently, white matter pathway degeneration may explain the pattern of impaired and spared cognitive functions thatare apparent in normal ageing. In particular, abilities that activate diffuse networks could be more affected, compared to activities relying upon localised networkswhich remain relatively stable (9). It remains to be investigated as to whether voxel-based analyses of cognitive function will support this inference.References[1] Morrison, JH & Hof PR, 1997. Science, 278: 412-419. [2] Pfefferbaum A et al., 2000. Mag Res Med, 44: 259-268.[3] Sullivan EV et al., 2001. NeuroReport, 12: 99-104. [4] Charlton RA et al., 2006. Neurology, 66(2): 217-222.[5] Barrick TR et al., 2005. NeuroImage, 24(3): 678-691. [6] Barrick TR & Clark CA, 2004. NeuroImage, 22(2): 481-491.[7] Alexander DC et al., 2001. IEEE Trans Med Imaging, 20(11): 1131-1139. [8] Head D et al., 2004. Cerebral Cortex, 14: 410-423.[9] Mega MS & Cummings JL, 1994. J of Neuropsychiatry & Clin Neuro, 6: 358-370.


O8Incorporating Domain Knowledge into Fuzzy Connectedness Image Segmentation:Application to Brain Lesion Volume Estimation in Multiple SclerosisMark A. Horsfield 1 , R. Bakshi 2 Marco Rovaris 3 , Mara A. Rocca 3 , Venkata S.R. Dandamudi 2 , Paola Valsasina 3 , Elda Judica 3 ,Fulvio Lucchini 3 , Charles Guttmann 2 , Maria Pia Sormani 4 and Massimo Filippi 31 University of Leicester, Leicester LE1 5WW; 2 Harvard Medical School, Boston MA;3 University of Milan, Italy; 4 University of Genova, ItalyIntroductionMultiple sclerosis (MS) is a neurological disorder affecting the brainand spinal cord thought to be of autoimmune origin [1]. Focal areas oftissue damage (lesions) occur, and are visible as hyperintensities onT 2 -weighted MR images. Measuring the change in volume of theselesions (the ‘lesion load’) plays an import part in placebo-controlledclinical trials of new treatments for MS.However, quantitative assessment of lesion load is time-consuming,and the volumes obtained are operator-dependent and prone tooperator-induced errors [2]. MS pathology, whilst having focalabnormalities, also has a diffuse component, and the lesions seen onMRI have no clearly-defined borders, making the delineation of suchborders highly subjective. Several workers have addressed theproblem of improving the reproducibility of the measurement of MSlesion volumes using computer-assisted or fully automatedcomputerized methods.Fuzzy connectedness is a general image segmentation framework inwhich the object membership of pixels depends on the way they “hangtogether” spatially in spite of gradual variations in their intensity.Fuzzy connectedness has previously been applied to MS lesionsegmentation, as part of a complex image analysis procedure [3]. Inthis work, we wished to simplify the task of lesion segmentation, toreduce the operator workload as much as possible. To this end weshow how prior knowledge can be incorporated into the fuzzyconnectedness framework, to improve the segmentation for thisparticular task.MethodsFuzzy Connectedness: The fuzzy affinity between any two imageelements (pixels) depends on the degree of adjacency of the pixels, aswell as the similarity of their intensity values. The closer the pixelsare, and the more similar their (possibly multi-parametric) intensities,the greater should be the affinity between them. The strength ofconnectedness between any pair of pixels (c, d) is defined byconsidering all possible connecting paths of pixels between c and d,where such a path is a sequence of links between adjacent pixels alongthe path. The strength of any one such path is the strength of theweakest link in it. Finally, the strength of connectedness between cand d is the strength of the strongest of all possible paths between cand d. Based on a set of seed pixels, provided by the user, a fuzzyconnected objected is defined as the set of all pixels with a connectionstrength to the seeds greater than a pre-set threshold.Incorporating Domain Knowledge: We modify the affinity betweenpixels to include three extra components:• The first relates to the known intensity characteristics of the imagefeature – in this case MS lesions – relative to the surrounding nonfeaturepixels. The user can provide “intensity hints” to indicate thatthe feature is brighter or darker than the background in a particularimage. This skews the implicit distribution of intensities from whichthe affinity is calculated.• The second consists of a probabilistic model of the spatial variationin the size characteristics of the features. Feature size is modelled asa spatially-varying feature membership correlation.• Third, we incorporate a probabilistic model of the known spatialdistribution of the feature to provide a “prior affinity”.Total affinity is a weighted sum of the prior affinity and the imagederivedaffinity.The feature size and distribution models were derived from asample of 300 MRI scans from the MS patient population, with thelesions manually segmented by neurologists. To assess an individualpatient’s scan, a proton-density template is registered to it, and thesame transform applied to the probability & feature size images.Testing: The method was tested against the performance of anestablished semi-automated methods of MS lesion segmentation basedon edge detection and contour following [3]. Using both methods, twooperators independently segmented the lesions in 14 MS patients fromdual-echo brain MRI scans. Each patient was scanned twice on thesame day so that we could assess both the scan-rescan and interobservervariability.The different sources of variation for the lesion volumemeasurements were modeled with a random effect analysis of variance(ANOVA), with three random factors (subject, observer and scannumber), plus three interaction terms, and a residual.In addition, the concordance between measurements was evaluatedas:A ∩ Bconcordanc e = × 100%A ∪ Bwhere A is the set of pixels segmented on one occasion and B the setsegmented on a second occasion. This is a measure of the consistencyof which pixels are delineated as lesion.ResultsFigure 1 shows one slice from the templates derived from the 300manually-segmented MRI scans.Figure 1: Representative slice from the template images: protondensityweighted template (left) lesion probability (centre); and lesionfeature size (right).The operator time required to segment the MS lesion was reducedfrom an average of 111 minutes per patient for the contouring method,to 16 minutes per patient for the fuzzy connections method.As expected, the ANOVA showed that the scan subject was by farthe biggest contributor to the total variance. However, the scannumber (first or second scan) made a neglibible contribution tovariance. A reduced model was therefore used, having removed theeffect of the scan number. Table 1 shows the non-negligiblecontributions to the variance for the reduced model.Contributor Contouring Fuzzy ConnectionsObserver 18.9% 9.3%Subject×Observer 4.4% 2.2%Table 1: contributors to variance for the two volume measurementmethods (ANOVA).Table 2 shows the mean concordances achieved when the sameobserver measured the first and second scans (scan-rescan), when thesame observer measured the same scans twice (intra-observer) andwhen the two observers measured the same scan (inter-observer).Contouring Fuzzy ConnectionsScan-rescan 55.1% 60.6%Intra-observer 74.36% 80.8%Inter-observer 47.6% 59.9%Table 2: concordances for the two volume measurement methods.ConclusionMS lesion segmentation based on the Fuzzy Connections algorithmsignificantly reduces the operator time, since the task is reduced to oneof identifying lesions and marking them, rather than delineating theborders. The method also improves the reproducibility andconsistency of the segmented lesion volumes, compared to a widelyusedsemi-automated method. The incorporation of prior knowledgeabout the distribution, size and brightness characteristics into thealgorithm is essential to achieve the required reproducibility.References1. H Lassman. In: McAlpine's Multiple Sclerosis, ChurchillLivingstone, London (1998).2. M. Filippi et al. Brain 118, 1593-1600 (1995).3. J.K. Udupa, L. Wei, S. IEEE Transactions on Medical Imaging. 16,598-609 (1997).


Abnormal thalamic metabolism on H-MRS in ‘peripheral’ diabetic neuropathyO10ID Wilkinson 1 , D Selvarajah 2 , R Gandhi, N Woodhouse, C Emery, PD Griffiths, Tesfaye S1 Academic Radiology, University of Sheffield ; 2 Diabetes Unit, Royal Hallamshire Hospital, Glossop Road, Sheffield S10 2JFIntroductionDiabetic neuropathy (DN) is a common debilitatingcomplication of diabetes mellitus involving the extremities. Infact, diabetes is the leading cause of non-traumatic lower limbamputation in the western world. This particular complicationhas historically been classified as a disease of the peripheralnervous system and pathological abnormalities of theperipheral nerves in DN have been well quantified. Recentevidence from our group indicates spinal cord involvement inearly-stage DN (1). However, the extent of central nervoussystem involvement remains unresolved. A completeunderstanding of the full extent of nervous systeminvolvement is crucial to elucidating the pathogenesis of DNand may facilitate the development of novel, rationaletreatments. This study aims to extend our understanding ofCNS involvement by investigating the metabolic status of thethalamus (common junction for most of the ascending sensorypathways) in DN using proton magnetic resonancespectroscopy (H-MRS).MethodsSixteen diabetic patients and 8 matched healthy volunteershave been studied. The diabetic patients consisted 2 groups: 8normal diabetics with no DN and 8 with establishedneuropathy determined according to published criteria (2).Spectra were obtained at 1.5T (Eclipse, Philips MedicalSystems, Best, The Netherlands) from an 8ml cubic voxelplaced over the right posterior lateral thalamic nucleus Twospectra were acquired per subject: one proton densityweighted(STimulated Echo Acquisition Mode - STEAMtechnique: TE=20ms, TR=3s) and the other T2-weighted(Point RESolved Spectroscopy - PRESS technique:TE=135ms, TR=1600ms). Short TE results are expressed asareas under the mI, Cho, Cr and NA resonances relative to thatfrom unsuppressed water. Long TE results are expressed asthe ratios (NA/Cho, NA/Cr & Cho/Cr) of areas under the threemain peaks, Cho, Cr and NA.ResultsAt long TE, NA/Cho was lower in established DN comparedto diabetics without neuropathy (p=0.036) and healthycontrols (p=0.015) (fig 2). At short TE, no significant groupdifferences were apparent.Discussion & ConclusionsThe posterior lateral nucleus of the thalamus was chosen forspectroscopic examination because most ascending sensorypathways relay within this nucleus before projections are sentto higher cortical centres. If the observed variation resultsfrom reduced NA signal at long echo-time, this data suggestsaltered neuronal biochemistry in subjects with established DN.The lack of significant difference at short TE suggests that thismay be due to changes in metabolite T2. This data providesfurther evidence that the CNS is involved in ‘peripheral’diabetic neuropathy.Figure 2: NA/Cho ratio obtained at long TE for the 3 groups.Figure 1: Example (a) spectroscopic ROI and long TE (135ms)spectra from (b) a normal control and (c) a diabetic subject withestablished neuropathyReferences1. Eaton SEM, Harris ND, Rajbhandari SM, Greenwood P,Wilkinson ID, Ward JD, Griffiths PD, Tesfaye S. Spinalcord atrophy in diabetic peripheral neuropathy. Lancet2001; 358:35-362. P.J. Dyck, J.L. Davies, W.J. Litchy, P.C. O’Brien. Longitudinalassessment of diabetic polyneuropathy using acomposite score in the Rochester Diabetic NeuropathyStudy cohort. Neurology(1997), 49, 229-239.


T2* weighted and phase imaging at 7 TeslaO11Lei JIANG 1 , Andreas Schaeffer 1 , Penny Gowland 1 and Richard Bowtel 1 .1 Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Nottingham, NG7 2RD, UKIntroductionUltrahigh field offers improved signal to noise ratio and new sourcesof image contrast. In particular spoiled gradient echo sequences provideconsiderable contrast to vessels, and brain areas that have previouslybeen associated with high iron levels ,due to increased T2*decay rates associated with magnetic susceptibility induced fieldperturbations. Furthermore the phase images associated with thissequence provide excellent contrast between the grey and white matter;a similar effect has previously been reported at lower field (1).The aim of this study is to characterise this effect with the future intentionof optimizing the sequences for contrast to noise ratio per unittime in the magnitude and phase images and of investigating thesources of this contrast.Methods and AnalysisData were acquired on a 7 Tesla Philips Intera Achieva scanner. Threehealthy human adult volunteers were studied, who gave their informedconsent. 3D-spoiled gradient echo images were acquired with an echotime of 15 ms, the shortest accessible TR of 27 ms and a flip angle of11 o . The voxel size was 0.4 x 0.4 x 0.7 mm 3 , with a matrix of 512 x512 x 120. The total acquisition time was 11 min. Phase images wereunwrapped in MATLAB (MathWorks, MA) and fitted to a polynomialto remove ‘low spatial frequency’ variations due to inhomogeneitiesin the applied field.Figure 3: Modulus image (left) and corresponding phase map (right)acquired at 7T (TE=15 ms).Results.Figure 4: Maximum intensity projection through a 3D spoiled gradientecho image set, showing the veins (TE=20 ms)The frequency shift between the grey and white matter was found tobe approximately 3 Hz.Figure 1: T2* weighted images through the caudate head, at 7T (left)and 3T (right) (TE=21 ms). Considerably increased contrast can beobserved at 7T.ConclusionWe have demonstrated that gradient echo images provide novel contrastat 7T in both the magnitude and phase data, in regions with largevenous contributions, in deep grey nuclei and in some major whitematter tracts. Possible biophysical explanations for this contrast includedeoxygenated blood in the venules and veins, known differencesin iron content of different brain regions or differences in myelinationbetween brain regions. Experiments are underway to try to understandthis source of contrast, and to improve the processing of the phasedata.References(1) E. Mark Haacke et al. Magn. Reson. Med. 52, 612-618 (2004).(2) Sofic E et al. Jour. Neural Transmission 74 (3), 199-205 (1988).Figure 2: Zoomed version of fig. 1 clearly showing new structurevisible in the caudate head at 7T.Acknowledgements: This work was funded by the Wellcome Trust,MRC, EPSRC and HEFCE, with technical support from Philips MedicalSystems.


Magnetic Field Induced VertigoO12Ian Cavin 1 , WenLong Qian 1 , Richard Bowtell 1 , Penny Gowland 1 , Paul Glover 11 Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, NG7 2RD, UKIntroductionIt is well known that subjects and operators report feelings of apparentmotion in and around high field magnets. We have called these sensationsMagnetic Field Induced Vertigo (MFIV). In some subjectsnausea can result whilst others are unaffected. In this work we presenttheoretical predictions together with experimental evidence for andagainst for a number of candidate mechanisms. Movement of thehuman head is detected in the vestibular system, which consists ofthree approximately orthogonal toroids of conducting fluid, eachcontaining a pressure sensitive cupula. These organs detect angularaccelerations of the head. The maculae of the vestibular system are theorgans which detect lateral and vertical accelerations, and are composedof a gel matrix which supports otoliths (aragonite form ofcalcium carbonate)[1]. It is possible to postulate a number of mechanismsfor the induction of vertigo during exposure to large magneticfields. These include: magneto hydrodynamic effects within the vestibularfluids (MHD); galvanic vestibular stimulation of hair cells dueto induced currents (iGVS); forces due to magnetic susceptibilitydifferences between the otolithic membrane and/or cupula and surroundingfluid.HypothesesMHD: These effects have previously been proposed as the main causeof vertigo in large magnetic fields [2]. Careful examination of MHDtheory does not however support this explanation. The expressiongiven in Ref. 2 for the pressure changes induced in a toroid due tomotion in a strong magnetic field assumes angular rotation about thetoroid diameter rather than its axis. Schenck assumes that the radialforce is perfectly coupled to the cupula of an orthogonal canal, but inreality it is likely to be a small fraction. Calculation of the effect ofrotation in the plane of the toroid gives a much smaller pressurechange than that calculated in Ref. 2 which is significantly less thanthe 13 micro Pa threshold [3] for perception.iGVS: Hair cells form linear deflection transduction sensors and respondto the displacement of the cupulae and the maculae. The haircell has a static firing rate at zero displacement; deflection causes achange in firing rate which the brain interprets as movement [4].Changes in the electric field across the cell will also modulate thefiring rate, giving a false perception of movement. The current densitiesrequired to elicit an effect are therefore likely to be lower than theaccepted thresholds for peripheral nerve stimulation. Unfortunately, itis difficult to derive a direct relationship between the rate of change ofmagnetic field (dB/dt) experienced and the perceived accelerationbecause of the complicated nature of the signal transduction. In addition,the pattern of induced electric fields is likely to depend on thefine structure of the inner ear which it is difficult to simulate withcurrently available numerical modelling software. However, directcurrents applied via vestibular (10µA [5]) or mastoid electrodes (1mA[4]) have been shown to modulate the firing rate of the vestibular haircells and hence give a subject a perception of rotation. A simpleanalysis using modelling [6] of induced and directly applied currentswould suggest that a dB/dt value of 5Ts -1 could induce a current densityof 100mAm -2 i.e. of the order of those occurring when 1mA isapplied to the mastoid.Susceptibility: Comparing the magnetically induced force on anotolith, which depends on the susceptibility difference between otolithand fluid, ∆χ, and the product of the field and (axial) gradient B x G,together with the buoyancy force, yields a perceived axial accelerationof − ∆χBG/ µ0∆ρ, where ∆ρ is the difference in otolith and fluid density.Assuming a value for B x G of 46 T 2 m -1 (which occurs just 30cminside the bore of a 7 T magnet) and taking literature values of ∆χ and∆ρ [7], this mechanism would produce a perceived acceleration of0.01 g (~ 10 times the perception limit for lateral accelerations [1]). Asimilar calculation can be performed for the cupula. However, in thiscase there are no published values related to the magnetic susceptibilityof the cupula. The ∆χ is likely to be very small as it is a gelatinousmembrane. The threshold of perception for this mechanism may belower than 5 T 2 m -1 in some subjects.MethodsA number of experiments have been employed in order to determinethe exact mechanism of magnetic interaction with the vestibular system.Only three are reported here:A) A 26 cm inner diameter solenoid, capable of delivering single unipolarand multiple bi-polar dB/dt impulses at 2Ts -1 for 50ms to a staticsubject was constructed. Twelve subjects stood upright for 1 minuteon an Accusway forceplate which recorded their centre-of-pressuremovements whilst 8 repeats of the stimulus was presented.B) Subjects stood perfectly stationary next to the 7T magnet (BxG =1T 2 m -1 ). They closed their eyes for 20s and opened them for 20s over3 cycles. Their forward sway motion was tracked by video cameraand head markers. The subject’s response was recorded at low field.C) Ten subjects were placed on the bed and moved to iso-centre of the7T magnet. They then performed slow rotations (


O13The study of magnetic nanoparticles as a contrast enhancer for MRI to track cancer metastasisLaura M PARKES 1 Richard Hodgson 1 , Ian Robinson 2 , David Fernig 3 , Nguyen TK Thanh 2,3*1 Magnetic Resonance and Image Analysis Research Centre (MARIARC), 2 Department of Chemistry3 Centre for Nanoscale Science, School of Biological Sciences, The University of Liverpool.IntroductionMagnetic nanoparticles offer a unique opportunity for cell tracking invivo. By incorporating magnetic nanoparticles inside cells, their positionwithin the body can be tracked by MRI. In particular, they providethe means to solve an outstanding problem in cancer research: thetracking of metastatic cancer cells. It is difficult to image the ongoingmetastatic process inside the body, and consequently much of what isknown about metastasis has been deduced from static pictures arisingfrom the identification of tumour cells in pathology samples.Magnetic nanoparticles composed of iron oxides are currently used asimage enhancers in magnetic resonance imaging (1, 2). However ironoxides have a relatively low saturation magnetisation, requiring theuse of larger particles. Transition metal nanoparticles, e.g. those madefrom cobalt, have a much higher saturation magnetisation value, allowingthe use of smaller particles (< 8 nm), without compromisingsensitivity.AimThe first aim of this work was to compare the MR sensitivity of cobaltand iron oxide nanoparticles. The second aim was to demonstrate MRcontrast due to nanoparticles in a cell culture.MethodsA 3T Siemens Trio system was used for imaging. T 1 , T 2 and T 2 * relaxationtimes were measured for a range of concentrations of theparticles (both iron oxide and cobalt) dispersed in 2% agarose gel.T2 quantification:A multi-echo spin echo sequence was used with the following parameters:TE: 16 echo times from 15 to 240ms, TR=2.5s, resolution0.9x0.9x3mm.T1 quantification:An inversion recovery turbo spin echo sequence was used at a rangeof inversion times: inversion times:24,100,300,570,750,1000,1500,3000ms, TR=5.9s, TE=12ms, resolution0.9x0.9x3mm.T2* quantification:A 3D gradient echo sequence was used at a range of echo times: TE=10,20,30,40,60ms, TR=70ms, 1.5mm isotropic resolution.Matlab was used to fit the data on a voxel by voxel basis producingmaps of T1, T2 and T2*. A region of interest from the central portionof the tubes was chosen to compare relaxation parameters betweeniron oxide and cobalt.To demonstrate MR contrast from nanoparticles in cells we grewHaCaT (human keratinocyte cell line) cells in the cell culture mediumDMEM + 5% (v/v) fetal calf serum. 100 µL containing 1300, 430,130, 43, 13, 4.3 µg of iron oxide was injected onto the plate over anarea of 8 x 16mm. The cells were incubated with iron oxide nanoparticlesin DMEM alone for 3 hrs. Following this the medium waswashed and a T2*-weighted 3D gradient echo sequence at echo timeof 30ms was used for imaging.ResultsFigure 1 shows the MR relaxation times for a range of concentrationsof both iron oxide and cobalt dispersed in agarose gel. The cobalt andiron oxide particles had approximate sizes 7 ± 1 nm and 18 ± 4 nm(n=200) respectively. It can be seen that both particles show measurabledifferences in relaxation times at relatively low concentrations,with the effect on T 2 * being the greatest. The effect of the iron oxideparticles is greater, particularly for T 2 *, reflecting the larger particlesize of iron oxide, resulting in approximately 17 times the mass concentrationof iron oxide compared to cobalt. Further work will considerthe effect of particle size.Figure 1T1 (s)T2 (ms)T2* (ms)1.61.51.41.31.21.18070605040302010050403020100Figure 20 10 20 30 40particles/ml x 10 120 10 20 30 40particles/ml x 10 12Iron oxideCobalt0 10 20 30 40particles/ml x 10 12µg130043013043134.3000DetectionlimitFigure 2 shows clear MRcontrast due to thepresence of iron oxidenanoparticles in cells.The detection limit in thisexperiment is 13ug ofiron oxide.ConclusionCobalt nanoparticles show promise as a contrast agent. Theirrelatively small size makes them more suitable for cell tracking.References1. Pankhurst, Q.A., et al., Journal of Physics D-Applied Physics,2003. 36(13): p. R167-R181.2. Kim, D.K., et al., Scripta Materialia, 2001. 44(8-9): p. 1713-1717.


Characterization of Solid State Dynamic Nuclear Polarization for Metabolic ImagingO14M. Schroeder, L. Cochlin, D. Tyler, G. Radda, K. ClarkeDepartment of Physiology Anatomy and Genetics, Sherrington Building, University of Oxford, Parks Rd, OX1 3PTIntroductionWhilst MRI is widely used in healthcare to assess tissue structure andfunction, the application of MR in metabolic imaging has been limitedby intrinsically low sensitivity. This sensitivity originates from thelow magnetic energy of nuclear spins compared with their thermalenergy at room temperature. In proton MR, low polarization is compensatedby a high proton concentration in samples; unfortunately thisis not true for low natural abundance magnetic nuclei such as 13 C.Dynamic nuclear polarization (DNP) is an established method that canachieve virtually 100% polarization levels in solid-state,paramagnetic samples (1,2). Recently, Ardenkjær-Larsen et al (3)utilized DNP in a novel method, DNP-MR, for obtaining highlypolarized nuclear spins in solution. DNP-MR consists of DNP of 13 Cnuclei to a polarization level of approximately 50%, followed by rapiddissolution that results in liquid state polarization in excess of 20%(3). As a result, dynamic, in vivo visualization of a given metabolite,and its conversion to other species, has become a possibility.This study was carried out to characterize the solid state DNP mechanismspecific to the application of DNP-MR, and to optimize solidstate polarization with a view to maximize liquid state polarization.All equipment associated with sample polarization, in a hyperpolarizerdesigned specifically for DNP-MR, has been previously described (3).MethodsA 40 mg aliquot of 99% enriched 1- 13 C-pyruvic acid, doped with 15mM of paramagnetic radical, was placed in the hyperpolarizer (3).Frequency of incident microwaves were increased from 93.840 GHzto 93.940 GHz, in 2 MHz increments, at an incident power of 20.1mW. Polarization at each frequency step was measured after 300 s ofbuild-up, with a 90º RF pulse.At the optimum frequency for positive polarization enhancement,determined from the frequency sweep at 20.1 mW, 1- 13 C-pyruvic acidwas polarized for 9000 s. Polarization build-up was monitored withvery low flip angle RF pulses at 300 s intervals. DNP parametersinvestigated included sample sizes from 30 mg-50 mg at intervals of10 mg, and 14 logarithmically-spaced values of microwave powerspanning the output range and sensitivity of the microwave source.Hyperpolarizer function was carefully monitored, to ensure the samplewas located in a bath of liquid helium, at less than 1.2 K and a magneticfield of 3.354 T.Solid-state polarization build-up was recorded and modelled as a firstorderexponential function. Polarization time constant was calculatedfor each build-up curve. Frequency sweep datasets were examined forindications of the mechanism of DNP (4).ResultsThe frequency sweep enhancement function can be seen in Figure 1.Optimal polarization frequency with this setup was determined to be93.890 GHz for positive enhancement, and 93.950 GHz for negativeenhancement. The enhancement function was asymmetric. The enhancementpeaks displayed a 60 MHz frequency spread, as comparedto the 72 MHz spread (twice the Larmor frequency) one would expectfrom the solid effect, indicating that thermal mixing may be the dominantmechanism of DNP (4). At optimal frequency, and at the rat dosesample size of 40 mg, maximum polarization was reached at microwaveintensity of 33.1 mW with a time constant of 1575 s.Figure 2 depicts polarization build-up under these conditions. A directrelationship was observed between the natural logarithm of incidentmicrowave power, and build-up time constant. A similar relationshipwas observed between microwave power and polarization magnitude;however at high microwave intensity (33.1 mW and above) the relationshipreversed, due to sample heating. Polarization time seems tobe independent of sample size. High power levels had less of a heatingeffect on larger samples. Figure 3a shows the relation between microwavepower and build-up time constant at various sample sizes, andfigure 3b shows the relationship between microwave power andmeasured NMR signal level.Signal magnitude / a.u.Time constant / s200015001000Frequency sweep enhancement function5000-500-1000-15003000250020001500100050093.8 93.85 93.9 93.95 94 94.050Freqeuncy, MW source / GHzFigure 2Polarizationbuild-up over9000s at93.890 GHz,33.1 mW.(a) Build-up Time Constant-2.00 0.00 2.00 4.00 6.00Ln (microwave power / mW )30mg50mg40mgSignal magnitude / a.u.14000120001000080006000400020000(b) NMR signal magnitude50mg40mg30mg-2.00 0.00 2.00 4.00 6.00Ln (microwave power / mW)Figure 3 The effect of sample mass and the natural logarithm ofmicrowave power on polarization build-up. a) Polarization timeconstant at 30 mg, 40 mg, and 50 mg. b) Signal magnitude, inarbitrary units, at same mass points. At higher microwave powerlevels sample heating counteracts the higher polarization rate.ConclusionWhen performing in vivo metabolic imaging experiments, it isessential that contrast agent polarization levels are sufficiently highfor the visualization of both the contrast agent and its lowerconcentration metabolic products. This study has revealed severalfactors that must be considered to ensure efficient solid state DNP, afundamental precursor to liquid state contrast agent polarization. Wenow have an understanding of how user-controlled hyperpolarizerparameters, such as microwave power and frequency, and sample size,affect polarization levels. In the future we will operate thehyperpolarizer at the conditions determined by this study, and use thisdata to monitor polarization build-up as a part of both hyperpolarizermaintenance and in development of novel metabolic tracer molecules.Future work will involve collection of data at sample size data pointsranging from 20 mg-60 mg, at 5 mg intervals, to fully characterize therelationships explored in this study. Thermal equilibrium spectra willbe acquired for calculation of absolute polarization levels, so that therelationship between microwave power and polarization magnitudecan be quantitatively modelled. Further, frequency sweeps at thehighest and lowest power extremes of the microwave source will beperformed to compare the resultant enhancement function lineshapes.This may pinpoint the relative contributions of thermal mixing and thesolid effect to the DNP mechanism.References(1) Jeffries. Phys Rev. 106, 164-165 (1957).(2) de Boer and Niinikoski Nucl Inst Meth. 114, 495-498 (1974)(3) Ardenkjaer-Larsen et al. Proc Natl Acad Sci USA. 100, 10158-63(2003).(4) Wind et al. Prog NMR Spec. 17, 33-67 (1985).Acknowledgements:Figure 1 Frequencysweep results. This systemis optimized at amicrowave frequency of93.895 GHz for positiveenhancement, and93.950 GHz for negativeenhancement.TimeThis study was supported by the British Heart Foundation and GeneralElectric.


O15MRI tracking to establish the relationship between myocardial injury and stem cell homing.Carolyn Carr 1 , Daniel Stuckey 1 , Louise Tatton 2 , Damian Tyler 1 , Juergen Schneider 1 , Sian Harding 3 , Kieran Clarke 11 Department of Physiology, Anatomy and Genetics, University of Oxford; 2 National Blood Service, John Radcliffe Hospital, Oxford;3 National Heart and Lung Institute, Imperial College, LondonIntroductionStem cells offer a promising approach to the treatment of myocardialinfarction and prevention of heart failure. Clinical trials using mononuclearcells, bone marrow-derived stromal cells (BMSCs) and skeletalmyoblasts are completed or underway 1 . In these trials, cells havebeen administered either by direct injection into the myocardium or byinfusion into the coronary system. In order to optimise cell-basedtherapy, the fate of the administered stem cells should be known,particularly where they are administered by infusion. Homing of stemcells to the infarcted myocardium has been shown by histologicalmethods to be related to the extent of myocardial injury 2 . Previouslywe have used iron oxide-labelling of BMSCs to non-invasively trackdonor cells injected into the infarcted myocardium by MRI and toisolate the donor cells from the grafted hearts using the magneticproperties of the injected BMSCs 3 . Here we infused labelled cells intothe vasculature to investigate the relation between myocardial damageand cell homing.AEF 65%CBEF 45% 3000 BMSCsDMethodsRat BMSCs were isolated from the tibia and fibia of male Wistar ratsand adherent cells were cultured to passage two. Prior toadministration, cells were incubated with 0.9 µm iron oxide particles(Bangs Laboratories Inc) for 24 hours and Vybrant DiI cell trackerdye for 1 hour. Myocardial infarction was induced in female rats byligation of the coronary artery. One or three days after coronary arteryligation, 5 x 10 6 labelled BMSCs were infused via the tail vein. Heartfunction was determined using cine MRI at 2, 7 and 28 days afterinfarct surgery. Animals were sacrificed at 4 weeks and hearts takenfor ex vivo MR microscopy or cell isolation.Cine MR acquisition: cardiac cine-MRI was performed using an11.7 T MR system with a Bruker console and a 60 mm birdcage coilas described previously 4 . A stack of contiguous 1.5 mm true shortaxis ECG and respiration-gated cine images (FOV 51.2 mm 2 , matrixsize 256 x 256, TE/TR 1.43/4.6 ms, 17.5° pulse, 25-35 frames percardiac cycle) were acquired to cover the entire left ventricle.MR microscopy: hearts were excised, fixed in paraformaldehyde, andembedded in 1% agarose doped with Gadolinium DTPA for highresolution MRI (fast gradient echo, FOV, 20mm 3 ; matrix size,256×256×512; TR/TE, 1.8/30 ms).Cell re-isolation: hearts were excised and digested with collagenase.Total digest was visualised by microscopy. In some cases, ironoxide/DiI + cells were separated from digested heart tissue using amagnet.ResultsMRI: Infarct surgery resulted in hearts with a range of ejection fractions(EF; 35-75%) and scar sizes (2-35%). In some hearts, a cleararea of signal void could be seen in the in vivo cine MRI throughoutthe infarct scar (Figure A, B). This area of signal void was confirmedby ex vivo MR microscopy (Figure C). Hearts with low levels ofdamage did not contain these areas of signal void (Figure D).Cell re-isolation: Fluorescent DiI labelled cells could clearly be seenin the tissue digest. Hearts with a more severe infarct were found tocontain a greater number of BMSCs (Figure B). DiI+ BMSCs couldalso be seen in sections of undigested scar tissue. Low numbers ofcells, undetected by MRI, were observed in hearts with smaller infarcts(Figure D). On rare occasions, contracting DiI+ cardiomyocyteswere observed. Contraction analysis showed that these were of adultphenotype and therefore unlikely to have resulted from differentiationof the BMSCs. DiI is a cytoplasmic stain and thus can be transferredbetween cells during scavenging or cell fusion 5 .DiscussionIron oxide labelling can be used to track BMSCs successfully whenthe cells are injected directly into the myocardium 3 since this results ina large number of labelled cells located within a small area of tissue.Infusion of BMSCs into the vasculature is associated with high loss ofcells as BMSCs become trapped in other organs, such as the lungs,EF 67% 1200 BMSCsIn vivo cine MRI 4 weeks post infarct surgery (A, B, D) and ex vivoMR microscopy (C) of the heart shown in A. The number of DiI+BMSCs detected after cell isolation from hearts B and D is shown.liver and spleen 6 . Nevertheless, cells infused into the vasculature havebeen detected in the infarcted heart using histology by us and others 2 .In hearts with a moderate infarct, regions of signal void could bedetected through the scar by cine MRI (Figure A) and MR microscopy(Figure C). Visualisation of the iron oxide labelled BMSCs in severelyinfarcted hearts can be more challenging as the scar is very thin andsignal along the endocardial edge of the scar may be diminished dueto the presence of compact collagen fibres and fibroblasts. In heartswith low levels of damage (Figure D), signal voids were not observed.Isolation of cells from heart tissue by total heart digest enabled thenumber of BMSCs that had homed to the heart and remained there forfour weeks to be established. Previously 3 we have shown that retentionof BMSCs injected directly into the myocardium correlates withthe degree of damage to the heart. Where cells are infused, the largernumber of BMSCs in more infarcted hearts may be due to increasedhoming to the heart or to improved retention of cells by the damagedmyocardium. The low numbers of DiI-labelled cardiomyocytes observedand their adult phenotype indicates that myocyte formationfrom BMSCs did not occur in this study.ConclusionIron oxide labelling can be used to track BMSCs administered to theinfarcted heart by intravenous infusion and the magnetic labellingfacilitates separation of grafted cells from those of the hostmyocardium after digestion of the heart. However, due to the lownumbers of BMSCs remaining in the mildly infarcted heart and thethin scar of the severely infarcted heart, this technique is less robustwhen intravenous infusion is used rather than direct myocardialinjection. A labelling technique that gives positive contrast would bemore valuable for low cell numbers in scar tissue.References1) Cleland et al, 2006, Eur J Heart Fail, 8, 105-1102) Ciulla et al, 2004, Transfusion, 44, 239-443) Stuckey et al, 2006, Stem Cells, in press4) Tyler et al, 2006, J Cardiovas Magn Reson, 8, 327-335) Garbade et al, 2005, Eur J Cardiothorac Surg, 28, 685-6916) Aicher et al, 2003, Circulation, 107, 2134-9This project was funded by the British Heart Foundation


Gastric and small bowel response to alginate beads.O16E. F. Cox 1 , C. L. Hoad 1 , P. Rayment 2 , R. C. Spiller 3 , P. J. Wright 1 , M. Butler 2 , L. Marciani 1,3 , P. A. Gowland 1 .1 Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, United Kingdom;2 Corporate Research, Unilever R & D, Colworth, Beds., United Kingdom;3 Wolfson Digestive Diseases Centre, University Hospital, Nottingham, United KingdomIntroductionThe study of gastrointestinal (GI) diseases using MRI is wellestablished 1 , but the study of normal small bowel physiology usingserial MRI has not yet been realised. Small bowel disease is visualisedby distending the small bowel lumen with water by using preparationsthat prevent absorption. This approach is not acceptable whenstudying normal physiology as water is naturally secreted andabsorbed in the small bowel. MRI is potentially a powerful techniquefor studying gastrointestinal physiology, as it is relatively noninvasive,and because it can measure many different, related functionswithin an experiment. The aim of this study was to determine theeffect of alginate gel bead strength on the gastrointestinal response toa model meal containing solid beads, using MRI to visualise andquantify the beads in the stomach and small bowel.MethodsVolunteer SelectionTen healthy volunteers, with no history of gastro-intestinal disease,formed the study group. The study was approved by the Local EthicsCommittee and all volunteers gave written informed consent.Meal DescriptionsTwo different bead types were used as model ‘solids’ in the study;solid centre alginate beads (strongly gelled) and liquid centre (weaklygelled). These were 2-4 millimetres in diameter and made by dropping200 ml of 1.5% w/w solution of Manugel DMB (ISP, Köln, Germany)into 0.37% CaCl solution for different lengths of time 2 . The mealswere randomised in a double blind fashion. In addition to the beads,the volunteers consumed 500ml of distilled water to help the volunteerswallow the beads without chewing and to provide contrast to thebeads in vivo.Study ProtocolVolunteers were asked to attend at 7:45am having fasted overnightand having abstained from alcohol for 24 hours, and from caffeine andstrenuous exercise for 18 hours. Volunteers were scanned beforeconsumption of the test meals to provide a baseline set ofmeasurements for the study day. A fat pre-load meal of 50ml Calogen(SHS International Ltd., Liverpool, UK) was given to the volunteers15 minutes before the main bead meal to turn the GI tract into a fedstate. Volunteers consumed the bead meal over 15 minutes. Imageswere acquired on a 3.0T Philips Achieva MRI scanner. CoronalRARE (TE = 400ms) images of the small bowel and transverseHASTE (TE = 59ms) images of the stomach were acquired during 2breatholds for each image type. These acquisitions were repeated atapproximately 30 minute intervals over 4 hours. Five minutes beforethe 4 hour scan 200ml of water was given to the volunteer to aidvisualisation of any beads that were left in the stomach. A satietyquestionnaire 3 was completed by the volunteers before each imagingperiod.AnalysisGastric half emptying times and percentage of ingested volumeremaining at 60 mins were measured from gastric volumemeasurements made from the HASTE images. Both total mealvolumes and bead only volumes were measured. Bead visualisationwas scored by one operator using the RARE images in the wholeimage and four quadrants of the intestine (0 – no beads visible, 1 –few beads visible, 2 – many beads visible). (See figure 1) Thequadrants were defined in the coronal image, with the centre being atthe inter-vertebral disc between L2 and L3 of the spine. These scoreswere integrated over the whole time period to give a visualisationscore (max 16). The time for the meal to initially reach the cecumfrom the mouth, was also measured from the coronal images. Theareas under the satiety curves (AUC) were calculated.ResultsThe median half emptying times for the meals containing weak andstrong gelling beads were both 48 mins (interquartile range: weak 40-64 mins, strong 41-64 mins). The median percentage volumesFigure 1: Typical coronal RARE image of the small bowel. Arrowsshow visible beads in the small bowel. The four qaudrants used in theanalysis are indicated.remaining at 60 mins were 30% and 35% respectively for the weakand strong beads (p = 0.037, N = 10, Wilcoxon Signed Ranks Test).For bead volumes only, the median half emptying time (weak 39mins, strong 54 mins) and the median percentage volume remaining at60 mins (weak 42%, strong 63%) were found to be close to the chosenstatistical significance threshold of p


Accurate Automated Measurement of Dynamic Arterial Lumen Area by CMRO17Clare E. JACKSON 1 , Cheerag C. Shirodaria 1 , Justin M. S. Lee 1 , Robin P. Choudhury 1 , Keith M. Channon 1 , StefanNeubauer 1 , Matthew D. Robson 11 University of Oxford Centre for Clinical Magnetic Resonance Research, Oxford, United KingdomIntroductionMagnetic resonance imaging is uniquely suited to study thepathophysiology of arteriosclerosis. By studying the dynamic changesin the lumen and relating this to the observed pulse pressure, it ispossible to derive a measure of arterial compliance (a mechano-elasticproperty of the arterial wall that reflects stiffness and endothelialdysfunction and is an important indicator of vascular function). MRmeasurements of vessel dimensions generally use manual tracing ofthe vessel lumen. However, these studies tend to generate very largedatasets so such data post-processing is very time consuming and haslimited accuracy and objectivity.MethodsMethods: All studies were performed on a 1.5T clinical MR scanner(Siemens Sonata, Erlangen, Germany). Bright blood SSFP cine MRimages were acquired of 33 newly diagnosed coronary artery diseasepatients, who were each imaged on two separate occasions. Themethod was used to automatically characterise the ascending,proximal descending and distal descending aorta on each of the 66examinations. Left and right carotid artery lumen area was also foundat end diastole and end systole for each examination. An automatedmeasurement method was implemented under Matlab (Version 6.5,The Mathworks, Inc, Natick, MA). This method was based upon asemi-automated method for the measurement of vessel wall thicknessthat was validated in [1]. The data analysis method was based upon(1) vessel wall unwrapping, followed by (2) a gradient detectionalgorithm for MR data post-processing and (3) automated tracking ofthe maximum gradient. All the cine MR images were analysed usingthis method, with the only user input being to outline the vessel on thefirst image. Analysis by manual tracing was also performed for eachof these vessels and the results were compared with those from theautomated method. Plexiglass tubes of known varying diameters werealso imaged and analysed using both the automatic and manualmethods. The repeatability of the automated method was tested byrepeatedly (5 times) acquiring bright blood SSFP cine MR images ofthe proximal descending aorta of 6 normal volunteers. Each data setwas analysed using the automated method. One data set from eachvolunteer was also analysed 5 times.ResultsPlexiglass tubes: There was a close correlation between the results ofthe automated method and the true dimensions of the plexiglass tubes(r=0.9998, MR-derived tube area = 1.008 * true tube area +7.239mm 2 , p0.7)between the manual and automated methods was found for 184 of the198 lumen area curves (93%). As the individual measurementsthrough the cardiac cycle are independent, we have used a measure oftemporal smoothness to determine an upper limit on the measurementerrors of each approach. The average Coefficient of Variation in themanual method was found to be 1.21% and the error in the automatedmethod was 0.58% and corresponds to sub-pixel accuracy. This erroris small when compared to peak physiological fluctuations of area ofaround 15.7% and shows that the automatic method is morereproducible. A poor correlation was found for 14 of the 198 lumenarea curves. 6 of these cases had obvious artefacts in the images.Another 8 failed on the ascending aorta, where the slice plane chosenresulted in the movement of the aorta in and out of the image planewhich had a large effect on the observed vessel area.Carotid arteries: Inner areas of the carotid arteries were found at enddiastole and end systole using both manual tracing and the automatedmethod. A close correlation (r = 0.9535) was found between the twomethods.Reproducibility: Figure 2 shows example area/time curves for therepeated analysis using the automated method. Figure 2(a) showsrepeated analysis (5 times) of the same data set. Figure 2(b) showsanalysis of 5 sets of data from the same volunteer. Mean correlationcoefficients are given in Table 1 for each of the 6 volunteers. Themean correlation was high (r>0.99) for both the repeated analysis ofthe same data set and analysis of repeatedly acquired data from theeach volunteer. As could be expected, the correlation coefficientswere all slightly higher for the repeated analysis of the same data setthan for the analysis of the repeatedly acquired data.Figure 1: Comparison of results of automatic method with actualmanually traced vessel dimensions for (a) the ascending aorta, (b) theproximal descending aorta and (c) the distal descending aorta. (d)shows example area/time curves for one data set.Figure 2: Example curves obtained when (a) automated analysis wasrepeated 5 times for the same data set and (b) the automated analysiswas performed for 5 different data sets from the same volunteer.1 2 3 4 5 6 meansame acquisition 0.99884 0.9973 0.99845 0.99935 0.99936 0.99829 0.998598different acquisitions 0.98957 0.98896 0.98851 0.99631 0.99601 0.99081 0.991695Table 1 : R values for the correlation of the automated analysis of thesame data (5 times) and the automated analysis of 5 separate sets ofdata acquired of each of 6 volunteersConclusionAutomated dynamic lumen area (DLA) segmentation of bright bloodSSFP images yields reliable and accurate measurements. Further workon this project will involve integration pulse wave velocity (PWV)calculation into the same software package and investigating thefunctional form of both the PWV and the DLA.References(1) Qian, W., Robson M.D., et al., Accuracy of Quantitative MRVessel Wall Imaging Applying a Semi-Automated Gradient DetectionAlgorithm-A Validation Study. JCMR, 6, 895-907 (2004)This project funded by the EPSRC (EP/D060834/1)


MRI indicates that bone marrow stem cells remain grafted in the infarcted rat heart for sixteen weeks,but do not improve cardiac functionStuckey DJ 1 , Carr CA 1 , Martin-Rendon E 2 , Tyler DJ 1 , Willmott C 2 , Cassidy PJ 1 , Hale SJM 2 , Schneider JE 1 , Tatton L 2 ,Harding SE 3 , Radda GK 1 , Watt S 2 , Clarke K 11 Cardiac Metabolism Research Group, Laboratory of Physiology, Anatomy and Genetics, University of Oxford.2 Stem Cell Research Laboratory, National Blood Service, John Radcliffe Hospital. Oxford.3 National Heart and Lung Institute, Imperial College School of Science, Technology and Medicine, London.O21IntroductionA promising, novel approach to the treatment of myocardial infarction(MI) and prevention of heart failure is cell grafting in the damagedmyocardium 1 To optimise stem cell therapy a non-invasive methodthat can show the tissue distribution of the administered cells andmonitor cardiac function is required 2 . We have used iron-labelling ofbone marrow stromal cells (BMSCs) to non-invasively track celllocation in the infarcted rat heart over 16 weeks using cine-MRI andto isolate the donor cells from the grafted hearts using the magneticproperties of the injected BMSCs. Cardiac function was measuredfrom 1 to 16 weeks post MI and BMSC administration.MethodsBMSCs were isolated from rat bone marrow, characterised by flowcytometry, transduced with lentiviral vectors expressing GFP andlabelled with 0.9 µm iron particles (Bangs Laboratories Inc). Ratswere anaesthetized and MI was induced by ligation of the left anteriordescending coronary artery. Ten minutes post MI, either a total of 5 ×10 5 BMSCs (n = 8), or saline (n = 7), was injected into 4 separatelocations in the infarct periphery. Cardiac cine-MRI was performed at1, 4, 10 and 16 weeks post MI using an 11.7 T MR system with aBruker console and a 60 mm birdcage coil as described previously 3 .A stack of contiguous 1.5 mm true short axis ECG and respiratorygated cine images (FOV, 51.2 mm 2 ; matrix size 256×256; TE/TR,1.43/4.6 ms; 17.5° pulse; 25 to 35 frames per cardiac cycle) wereacquired to cover the entire left ventricle. Cell distribution andcardiac function were measured in each slice. At 16 weeks, heartswere removed, fixed and imaged ex vivo using high resolution 3D MRmicroscopy (fast gradient echo, FOV, 20 mm 3 ; matrix size,256×256×512; TR/TE, 1.8/30 ms), then sectioned for histology ordigested to single cells using collagenase for donor cell re-isolation.ResultsSignal voids in the cardiac MR images caused by the iron particles inthe BMSCs were detected in all rats at all times (Fig 1A). MRmicroscopy identified hypointense regions at the same position asthose identified in vivo (Fig 1B). In mildly infarcted hearts, thevolume of the signal void decreased over the 16 weeks (Fig 1C),whilst the signal void volume did not decrease significantly inseverely infarcted hearts. When non-labelled BMSCs or naked Bangsiron particles were injected into the heart, no signal voids wereidentified after the first week. Donor cells containing iron particlesand expressing GFP were identified in MR-targeted heart sections andafter magnetic cell separation from digested hearts. Approximately1.5% of administered BMSCs were identified at 16 weeks. Themajority of donor cells maintained their BMSC phenotype, but on rareoccasions, fluorescently labelled cells with a cardiomyocytephenotype were identified.No improvements in left ventricular ejection fraction (Fig 2), strokevolume, cardiac output or reduction in non-contractile scar regionwere found between the control and BMSC treated animals at any ofthe time points studies.Ejection Fraction (%)80706050400 5 10 15WeeksMI +BMSCsn = 8MIn = 7shamn = 6Figure 2: Left ventricular ejection fraction from rats that underwenteither MI, MI + BMSC injection, or sham operation.ConclusionMRI can be used to track cells labelled with iron particles in damagedtissue for at least 16 weeks after injection and to guide tissuesectioning by accurately identifying regions of cell engraftment. Themagnetic properties of the iron labelled donor cells can be used fortheir isolation from host tissue to enable future characterisation. Inthis study, although the grafted BMSCs remained present in theinfarcted hearts for 16 weeks, they were unable to improve cardiacfunction.References1 Mather & Martin, 2004, Lancet 364 p 183-922 Bartunek et al, 2006, Eur Heart J, 11, p1338-403Tyler et al, 2006, J Cardiovas Magn Reson, 8, 327-33This project was funded by the British Heart FoundationFigure 1: In vivo (A) and ex vivo (B) MR images of iron labelled BMSCs in the infarcted rat heart 16 weeks after administration. By tenweeks the signal void volume arising from the donor cells was significantly larger in the hearts with greater damage (C).


Longitudinal Short TE Proton Magnetic Resonance Spectroscopy and Disease Progression in Inherited Prion DiseaseHyare H 1,2 , Siddique D 2, 3 , Webb T 2, 3 , Wroe S 2, 3 , Collinge J 2, 3 , Thornton JS 1 , Yousry T 11 Lysholm Department of Neuroradiology, National Hospital for Neurology and Neurosurgery,2 MRC Prion Unit, Department of Neurodegenerative Diseases, Institute of Neurology, UCL,3National Prion Clinic, National Hospital for Neurology and Neurosurgery, Queen Square, London, UKO22IntroductionInherited prion disease is a progressive neurodegenerativedisease characterised by different mutations within theprion protein (PRNP) gene 1 . Conventional MR sequencesare often unremarkable but with a therapeutic trialunderway, non-invasive monitoring of disease progressionis vital. To date, proton magnetic resonance spectroscopy( 1 H-MRS) studies in inherited prion disease have beenlimited to case reports and small series 2-4 . The purpose ofthis study was to determine longitudinal changes in cerebral1 H-MRS metabolite-ratios in inherited prion disease andinvestigate their value as surrogate markers of diseaseprogression.MethodsFive symptomatic patients (3 male, 2 female, mean age 46years) with inherited prion disease, referred to the NationalPrion Clinic, National Hospital for Neurology andNeurosurgery, for treatment with quinacrine as part of theMRC Prion 1 Trial, were studied. Short echo time (TE),single voxel 1 H-MRS was performed using an automatedpoint-resolved spin-echo localisation (PRESS) technique,with TE 35ms, TR 3000, NEX 8 at 1.5T (GE MedicalSystems, Milwaukee, WI). Spectra were acquired fromtwo voxels (volume 3.3ml-4.4ml) centred on the right headof caudate (RHC) to include the anterior right putamen, andthe right thalamus (RTH) (Fig 1). Signal ratios for themetabolites N-acetylaspartate (NAA), choline containingcompounds (Cho) and myo-inositol (MI) relative to totalcreatine (Cr) were determined using LCModel software 5 .Spectra were obtained serially using the same protocol inall patients at 3 month intervals to a maximum of 9 monthsfollow up.ResultsClinically all patients demonstrated disease progressionwith a decrease in mini mental state examination (MMSE)and activities of daily living (ADL) clinical scores.Appearances on conventional MRI were unremarkableexcept for mild volume loss. In the RHC voxel, all patientsdemonstrated a decrease in NAA/Cr and an increase inMI/Cr with time (Figs. 2 and 3). Linear regression analysisdemonstrated a mean slope of -0.0520 (SE 0.013) permonth for NAA/Cr (p=0.018) and 0.0975 (SE 0.025) permonth for MI/Cr (p=0.018). No significant change wasseen for Cho/Cr in the RHC voxel, or any metabolite-ratioin the RTH voxel.ConclusionAnatomically specific changes in NAA/Cr and MI/Cr wereobserved concomitant with clinical deterioration.Spongiosis, gliosis and neuronal loss are histopathologicalfeatures of inherited prion disease and changes are oftenextensive in the caudate and putamen 6 . Elevated MI/Cr isthought to be associated with gliosis, and reduced NAA/Crwith neuronal loss in neurodegenerative conditions 7 .In the RHC voxel we observed a proportionally greaterchange in MI/Cr than NAA/Cr, suggesting that MI/Cr maybe a more sensitive index of pathological changes.As opposed to conventional MR imaging, longitudinalshort TE1 H-MRS may provide important surrogatemarkers of disease progression in patients with inheritedforms of prion disease.Fig 1: Axial T2 FSE images demonstrating positions of the RHC voxel(A) and RTH voxel (B).NAA/Cr21.81.61.41.210.80.60.40.20Right head of caudate and right putamen voxel0 1 2 3 4 5 6 7 8 9Time (months)Fig 2: NAA/Cr versus time in RHC voxelMI/Cr1.41.210.80.60.40.20ARight head of caudate and right putamen voxel0 1 2 3 4 5 6 7 8 9Time (months)Fig 3: MI/Cr versus time in the RHC voxelPatient 1Patient 2Patient 3Patient 4Patient 5Patient 1Patient 2Patient 3Patient 4Patient 5References(1) Malluci G et al. Curr Opin Neurol. 17(6), 641-647 (2004)(2) Shyu W-C et al.. J Neurol Sci. 138, 157-160 (1996)(3) Konaka K et al. Neuroradiology. 42, 662-665 (2000)(4) Waldmann AD et al. Neuroradiology. 48(6), 428-33 (2006)(5).Provencher SW et al. Magn Reson Med. 30, 672-679 (1993)(6) Nicholl D et al. J Neurol Neurosurg Pyschiatry. 58, 65-69 (1995)(7) Valenzuela et al. Neurology. 56, 592-598 (2001)This project was funded by the Medical Research Council.B


A New Functional Form for Arterial Input Function ModellingMatthew ORTON 1 , James D’Arcy 1 , Simon Walker-Samuel 1 , Martin Leach 1 , David Hawkes 2 , David Collins 11 Institute of Cancer Research, Sutton, Surrey, UK. SM2 5NG.2 Centre for Medical Image Computing, University College London, WC1E 6BT, UK.O23IntroductionThe use of dynamic MR methods for first-pass kinetic modelling of diffusibletracers has received much recent attention. The holy-grail is the accuratequantification of various pharmacokinetic parameters which have importantdiagnostic and prognostic capabilities. A crucial component of such analyses isthe Arterial Input Function (AIF), which is the concentration of tracer in theblood plasma as a function of time.Since the AIF is not fixed, either between patients, or for the same patient ondifferent examinations, it is of primary importance to use an AIF that isappropriate to the data being considered. Failure to do this will have an obviousimpact on the estimation of the pharmacokinetic parameters, and this has beenwidely reported by many researchers. Current solutions include directmeasurement in a blood vessel (1), reference region methods (2), or using afunctional form of some kind (3,4) – each of these methods have advantages anddisadvantages.The approach proposed here is the use of a particular functional form for theAIF, which is given in the next section. The ultimate aim of this work will be toestimate the parameters of the proposed AIF directly from the data within theROI, removing the need for data from either blood vessels or reference regions.This abstract presents the form for the AIF along with a theoretical justification,discusses its properties, and presents some preliminary results where the AIFparameters are estimated from the tissue uptake curves taken from a primarybreast carcinoma.MethodsFunctional form for AIF: The proposed AIF is given by2 −m1t−m2t−m1tcp( t)= a1m1t e + a2( e − e ).This function has some similarities with that described by Weinmann et al. (3), inthat it contains exponentially decaying terms with two rate constants, and twoamplitude parameters. However, figure 3 shows that it has a very different initialshape from the bi-exponential form of Weinmann. An important difference isthat c p (0) = 0, so that this function correctly models the fact that the tracerconcentration gradually reaches its maximum level, rather than instantaneously.An advantage of this form is that when combined with the Tofts model for tracerdiffusion from the plasma (5), the resulting convolution has an analyticalsolution. The inclusion of m 2 1 in the first term means that a 1 is the area under thatpart of the curve.Theoretical justification for AIF: The AIF can be split into several phases – theinitial bolus, possible recirculation peak(s), equilibration with the whole bodyextravascular space, and finally renal excretion. Renal excretion occurs overseveral hours, so for first-pass modelling this effect can be neglected. Separaterecirculation peaks are also neglected since even though they can sometimes beseen in the MR signals from blood vessels, they have a diminishing effect at thetissue level. By the time the bolus reaches the aorta, its initial rectangular shapewill have been dispersed. A gamma function of the form c b (t) = α 1 t n e -m t 1is knownto be a reasonable model for this process (6), and that is the form used here.Increasing n forces the function to be near-zero for a longer initial period, whileincreasing m 1 makes both the rise and the fall faster. In principle n could takeany positive value, but for simplicity it is fixed at n = 1, which still retains themost important features of the function. The results of Weinmann et al. indicatethat the equilibration between the blood plasma and the whole body extravascularspace takes of order 10-20 minutes, during which time the initial bolus willbecome well mixed with the blood plasma. A simple dual-compartment model istherefore adequate to model the equilibration process, and this is characterised bya residue function of the form r(t) = α 2 e -m t 2. The complete AIF is the sum of theinitial bolus and the equilibrium phase, and is given bycp( t)= cb(t)+ r(t)⊗ cb(t),where ⊗ represents convolution. Some simple calculus and algebra recovers theexplicit form given above, where the amplitude terms have been condensed intothe parameters a 1 and a 2 .ResultsInitial results were obtained by using the Bayesian Fitting algorithm reportedbriefly in (7), that uses a probabilistic model to jointly estimate the AIFparameters and the pharmacokinetic parameters for a given ROI. The outputfrom the algorithm is a large collection of random samples from the inferredposterior distribution that can be used to provide estimates for the parameters inmany different forms. In figure 1, images of posterior mean estimates of thepharmacokinetic parameters k ep and K trans are shown for a primary breastcarcinoma, overlaid on the anatomical image. Of more interest in this context arethe results for the AIF parameters, and this is presented in two forms. Firstly infigure 2 estimates of the posterior distributions for each parameter are shown,which were obtained by smoothing a histogram of the samples from the fittingalgorithm. The peak of this distribution can be used as a point estimate for the0.0 1.0 2.0kep (min -1 )Figure 1: Images of pharmacokinetic parameters overlaid on the anatomicalimage of a breast carcinoma.p(a 1| data)p(m 1| data)c p(t) (mM)6543210.8 0.9 1a 1(mM min)13 14 15m 1(min −1 )1 1.1 1.2a 2(mM)Figure 2: Smoothed Posterior histograms for the four AIF parameters.Figure 3: Estimated AIF drawn as a scatter plot to represent uncertainty.0.3 0.6K trans (min -1 )00 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6 1.8 2time (min)p(a 2| data)p(m 2| data)0.00.095 0.1 0.105 0.11m 2(min −1 )parameter, while the width gives a measure of the posterior uncertainty in theparameter. A more direct description is the AIF itself, and each sample from thefitting algorithm can be used to give an AIF curve that is a possible fit to the data.This is shown in figure 3 where each sample is used to generate a few pointsalong the AIF curve. The density of points in any region gives a measure of howlikely the curve is to pass through that region, given the uncertainty in theparameter estimates. This indicates that there is more uncertainty in the bolusthan the equilibration phase, which is not surprising since there is less data fromwhich to estimate the bolus parameters.ConclusionThe preliminary results presented here demonstrate that this functional form forthe AIF is reasonably realistic, and has some theoretical justification. When it isused in a fitting algorithm that estimates the parameters from tissue uptake data,the resulting curve is seen to be plausible, and commensurate with directlymeasured AIFs that have been previously reported (1,4). In particular, theestimate of m 2 is very similar to published values for the rate of equilibrationbetween the blood plasma and the whole body extravascular space (3).References(1) Fritz-Hansen T, et al. Mag. Res. Med. 3, 225-231 (1996)(2) Yang C, et al. Mag. Res. Med. 52, 1110-7 (2004)(3) Weinmann HJ, et al. Physiol. Chem. Phys. Med. NMR. 16, 167-172 (1984)(4) Parker GJ, et al. Proc. Intl. Soc. Mag. Res. Med. 13, 2100 (2005)(5) Tofts PS, et al. J. Magn. Reson. Imag. 10, 223-32 (1999)(6) Starmer CF, et al. J. Appl. Physiol. 28, 219-20 (1970)(7) Orton M, et al. Proc. Intl. Soc. Mag. Res. Med. 14, 3490 (2006)This project was funded by EPSRC grants GR/T20434/01 and GR/T20427/01(P),and CRUK grant C1060/A808.


Partial volume correction for magnetic resonance renography studiesO24Rodriguez Gutierrez, D 1 , Diaz Montesdeoca, O 2 , Wells, K 1 , Mendichovszky, I 3 , and Gordon I 3 ,1 Centre for Vision, Speech and Signal Processing, School of Electronics & Physical Sciences, University of Surrey, UK.2 EUITT, Universidad de Las Palmas de Gran Canaria, Spain.3 RCS Unit of Biophysics, UCL Institute of Child Health, London, UK.IntroductionMagnetic resonance renography (MRR) has been suggested as apossible alternative to nuclear radioisotope renography (NRR) forestimating differential renal function (1) (i.e. filtration). As in NRR, acompound with known renal handling (e.g. DTPA) is tagged with asuitable contrast agent, in this case gadolinium (Gd). Data are thenacquired capturing the distribution of the contrast agent within thekidney as a function of time. Some of the advantages of MRR are thelack of ionising radiation, increased spatial resolution, and volumetricdata that contains both tracer kinetics and anatomical information.To date, MRR and, even NRR, techniques have failed to robustlyestimate glomerular filtration rate (GFR) when compared to ‘goldstandard’ plasma-sampling methods. Prior work has identifiedmovement artefacts (2), non-linear relation between signal-intensityand Gd concentration (3), suitability of models used to estimate GFR(4), or the selection of a suitable region of interest (ROI) for theanalysis (5) as factors that need to be considered for accuratequantification. However, we propose that partial volume effect (PVE)must also be considered.In this work, we present the first results demonstrating partial volumecorrection for adjacent organs on time-intensity curves obtained fromrenal ROIs.MethodsThe proposed approach is based on estimation of the mixing betweendifferent tissues within every voxel, and the definition of observedtime-intensity curves as the sum of individual time-intensity curvesthat correspond to each tissue class.Assuming linear mixing, the problem of partial volumes (PVs) mightbe thought of as that of finding, for every voxel, a mixing vector, α:α = ( α 1, α2,...α n)where n is the number of tissue signals contributing to the voxelintensity. Thus, the mixing vector represents the percentage of eachtissue captured within that voxel.Our approach to generating the mixing vectors consists of acquiring ahigh resolution anatomical volume, prior to contrast injection that isthen segmented to produce noiseless high resolution binary tissuetemplates. Assuming linear mixing, these high resolution templatesare then individually filtered using the point spread function (PSF) ofthe sequence used during dynamic contrast-enhanced acquisition anddown-sampled to the same dimensions of the dynamic data. Thus, theindividual components (α k ) of the mixing vector for each voxel aregenerated (see Figure 1). The PSF can be obtained using methods suchas in (6).Having obtained the corresponding α-maps, the observed intensity I obsfor each voxel at time, t, within or surrounding the kidney can bedefined as:nobs( t)= ∑αj× Ij( t)j = 1IKnowing I obs and the α-maps, the above equation can be solved for thesignal intensity curves of individual tissue components, provided thereare at least n, or more, observations per ROI .To illustrate the application, several datasets from a healthy volunteerwith normal renal function were acquired on a 1.5T Siemens Avantoscanner. An anatomical high resolution scan was followed by adynamic acquisition using a SPGR 3D-FLASH pulse-sequence(VIBE): TE/TR = 0.53/1.63 ms, flip angle = 17º, acquisition matrix =128x104 and 400x325 (mm) field of view (FOV). The functionaldataset consisted of 3D volumes with 18 slices of 7.5mm thickness(no gap) and an in-plane resolution of 3.1x3.1 mm, acquired every 2.5s. The injected Gd-DTPA dose was 0.05 mmol/kg body weight. Datawere corrected using a movement correction technique similar to thatdescribed in (2), but extended to 3D.ResultsPreliminary results on a region close to the liver, as an example onhow to eliminate the contributions from the liver and fat surroundingthe right kidney from the renal signal-intensity curve are presented.Figure 1: High resolution image and segmentation of liver andright kidney (left) and α-map corresponding to the right kidney(right).Figure 2: Cortical ROI (green pixels) overlaid on to the dynamicdata (left) and corresponding time-intensity curves: The red plotcorresponds to the average intensity in the ROI (uncorrecteddata), the blue and green plots are the resulting kidney and livercomponents, using the method described in the text (right).Figure 2 shows a cortical ROI selected from voxels inside the kidneyaffected by liver PVs. The graph shows time-intensity curves for theROI both before and after PV correction. The corrected curvecorresponding to kidney-only clearly shows a blood perfusion peak,followed by an ascending intensity phase that corresponds to filtrationand a descending intensity phase as tracer begins to exit the kidney. Incontrast, the pre-correction curve is contaminated by the signal fromthe liver, and does not resemble the expected renal curve.ConclusionThe work presented here is, to our knowledge, the first attempt toquantify PV effects in MRR. As data within renal ROIs isquantitatively used for GFR estimation, non-correction of these PVEsmight be one of the factors preventing robust renal functionestimation. Further work is currently directed towards correction ofPV effects from intra-kidney tissues not involved in the filtrationphase.References(1) Huang AJ, et al. Radiol. Clin. North Am. 41, 1001-1017 (2003)(2) Giele ELW, et al. J Magn. Reson. Imaging 14, 741-749 (2001)(3) Pedersen M, et al. Magn. Reson. Med 51(3), 510-517 (2004)(4) Annet L, et al. J Magn. Reson. Imaging 20, 843-849 (2004)(5) de Priester JA, et al. J Magn. Reson. Imaging 14, 134-140 (2001)(6) Steckner MC et al. Med. Phys. 21 (3), 483-489 (1994)


O25Imaging in drug development: lessons from therapeutics and targeted molecularimaging agents.Andy Dzik-Jurasz, VP of Clinical Development, Point Therapeutics, Inc., Boston, MA,USA. adzikjurasz@pther.comPrecise quantitative knowledge of clinical drug pharmacology, in normal and diseasedstates is essential to the successful development of drugs. Several biomarker strategies,including imaging, are currently under development and have had success in supportingidentification of the mechanisms of action of drugs, and developing early read-outs ofresponse and prognosis for the purpose of patient stratification. Imaging has the greatbenefit over other biomarkers in providing functional and anatomical data. Usingexamples from the field of oncology the role of imaging in supporting the development ofoncology drugs in early clinical trials will be discussed. Many of these techniques are farfrom being accepted as outcome measures for regulatory submission of new chemicalentities (putative drugs that have not yet been approved for commercialization). On theother hand, conventional anatomical imaging represents the main body of imaging usedin late phase drug development (phase 3 and 4) and differs substantially in its role to thenewer functional techniques in development. The manner in which imaging is used forregulatory and exploratory research purposes will also be discussed.Many drugs currently under development are characterized by their unique moleculartargeting which is as a result of the increased understanding of key molecular targetscontrolling disease behavior i.e. apoptosis, angiogenesis. The importance of molecularevents in disease has led to an attempt to image those events in man in the hope ofaltering disease management to the patient’s advantage compared with the standard ofcare. The development of such agents is no less demanding than the development of atherapeutic and presents certain unique problems for those hoping to bring a targetedimaging agent to market. Not least of these is the regulatory hurdle to approval. Examplesof recent experience with targeted magnetic resonance agents will be used to illustrate theissues faced in bringing a targeted imaging agent to the clinic.


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR íÜ ^ìÖìëíI=OMMSpìÄãáííÉÇ=mçëíÉêmêÉëÉåí~íáçåëïïïKéÜKëìêêÉóK~ÅKìâLÅåêéLÄÅMS


Characterisation of a Carotid Injury Model in Rat in vivo Using MRIP2K.K. Cheung 1,2 , M.J. Ramirez 3 , P. Lehtolainen 3 , D.G. Gadian 2 , A.M. Taylor 4 , R.J. Ordidge 1 , M.F. Lythgoe 21. Department of Medical Physics and Bioengineering, University College London, London, UK – *k.cheung@ich.ucl.ac.uk2. RCS Unit of Biophysics, UCL Institute of Child Health, London, UK3. Centre for Cardiovascular Biology and Medicine, Division of Medicine, University College London, UK4. Cardiothoracic Unit, Great Ormond Street Hospital for Children, London, UKIntroductionPercutaneous transluminal angioplasty (PTA) is a revascularisationtechnique for ischaemic heart diseases (IHD) with low morbidity andhigh success rate. However post-angioplasty restenosis is commondue to neointimal hyperplasia following endothelial injury,necessitating repeat surgery to re-expand the target vessel 1 . Incardiovascular diseases, both endogenous and exogenous bloodderivedcells appear to have the capacity to differentiate and transdifferentiateinto endothelial and vascular smooth muscle cells, andcan participate in angiogenesis and neovascularisation 2 . Endothelialprogenitor cells (EPCs) are a type of stem cell that has beendemonstrated to reduce neointimal hyperplasia 3 . We have establishedan animal model to investigate the effect of EPCs on post-angioplastyinjury of the common carotid artery (CCA). This study aims tocharacterise this model by monitoring the development of injury invivousing MRI, which would provide the initial platform for alongitudinal study incorporating stem cell tracking.MethodsStudy Design: Nine male adult Sprague-Dawley rats (400g-570g)were divided into two groups (G1&G2). Both groups were scannedbefore undergoing balloon surgery, and were subsequently scanned onday 2, 7, 14, 21 and 28 after surgery; G2 was scanned at two furthertime points on day 48 and 52. Animals were sacrificed at the last timepoints for tissue extraction.Surgery Protocol: Animals were anaesthetised with ketamine andxylazine. The left CCA and external carotid artery (ECA) wereexposed through a midline incision in the neck. A 2F embolectomycatheter was inserted via the ECA into the CCA down to the aorticarch. The balloon was then inflated and withdrawn with rotation todenude the endothelium. The ECA was tied and the wound wasclosed. Animal were recovered and kept ad libitum post surgery.MRI Evaluation: Transverse images of the CCA were obtained usinga 2.35T horizontal bore SMIS system with a 15mm circular surfacecoil. The animals were anaesthetized with 2% isoflurane and 1L/minO 2 . A multislice spin echo (SE) 2DFT sequence was used, with thefirst slice positioned immediately proximal to the bifurcation of theCCA. (TR = 1000; TE = 30; FOV = 25mm; 9 slices; slice thickness =1.5mm; 256×128 pixels reconstructed to 256×256; apparent in-planeresolution = 98µm×98µm). Data acquisition time for each animal wasapproximately 43 minutes. For quantitative analysis, lumens of the leftand right CCA were outlined for every slice and then measured withImageJ software using a bimodal threshold segmentation technique 4 .Mean lumen areas (MLA) were then obtained by averaging the lumenareas from all nine image slices.Data Analysis and Statistics:The right CCA MLA was used as the control at each time point, andthe ratio between the left and right MLA was calculated. This indexwas chosen to account for lumen changes due to growth. Within eachanimal, the ratio for each time point is then normalised to its preinjuryL:R ratio.Comparisons were made between different rats at different timepoints. Each rat in G1 and G2 accounted for 5 and 7 timemeasurements respectively. Data were expressed as means +/-standard error (SE). L:R ratio at each time point was compared to theday-0 (pre-injury) ratio with paired T-test. A difference wasconsidered to be significant at p


A Locally Adaptive Gradient Controlled Spatial Regularisation Partial Volume ClassifierP3John P. Chiverton, Kevin Wells,1 Centre for Vision, Speech and Signal Processing, School of Electronics and Physical Sciences,University of Surrey, GU2 7XH, UKIntroductionMR imaging of the human brain is an important imaging modality inclinical and research applications of neurology. The accuratequantitation of the gross anatomical tissues, i.e. Grey Matter (GM)and White Matter (WM) is an important step in many diagnosticapplications. Unfortunately MR images suffer from imaging artefacts,notably the Partial Volume (PV) effect where individual voxels maybe composed of a mixture of these tissues. This work presents a novelmethodology, utilising a locally adaptive Gradient-controlled SpatialRegulariser (GSR) in a Bayesian formulation for the quantification ofthese tissues in the presence of the PV effect. This new formulationresults in competitive and in some instances superior PV classifierperformance in comparison with other existing classifiers, due to theadaptive spatial regularisation.MethodsThe PV effect is modelled here by utilising a Bayesian formulation.The per voxel tissue content is modelled as a mixture random vector,α = (α 1 α 2 ... α n ) T for n tissues. Each vector element, α i , isrepresentative of the amount of tissue in the voxel, where 0 representszero content and 1 represents full content. The probability density forthis random vector given a voxel intensity, g, gradient magnitude, z,and the mean mixture vector, α N , of the neighbouring voxels is givenby the posterior density via Bayes theorem:p( α | g, α N ,θ,z) ∝ p( g | α,θ) . p( α | α N ,z)(2)where the first term on the r.h.s. is the intensity likelihood and in thiscase a Gaussian distribution (given a particular mixture vector) is mostappropriate for neurological MR magnitude data, see [1]. Theparameters of the Gaussian distribution are described by θ containingmeans and standard deviations. The right most term is the locallydefined spatial mixing prior dependent on the neighbouring mixturevalues and the gradient magnitude. The form of this spatial mixingprior will now be discussed.A quadratic function is useful to measure the dissimilarity betweenthe voxel’s mixing vector, α, and the mean mixing vector ofneighbouring voxels, α N , the result of which is a Gaussian distributedrandom variable. This was previously found by Choi et al, 1993 [2].We however have also found a functional relationship exists betweenthe locally calculated image gradient magnitude, z, and the standarddeviation of this distribution, i.e. the amount of spatial regularisationrequired. This relationship allows the spatial regularisation of themixing prior to be regulated on a per voxel basis. Thus regions of highgradient magnitude require less spatial regularisation due to the likelydissimilarity of the neighbouring voxel tissue content. We found thefunctional relationship to be well modelled by a Beta density function.Parameter estimation was undertaken by simulating theconditional posterior densities with an Independent MetropolisHastings (IMH) Markov Chain Monte Carlo (MCMC) approach. Themean and variance prior distributions were given Gaussian andGamma PDFs respectively and were used as the proposal distributionsin the IMH algorithm to simulate the posterior distributions of themean and variance parameters. The aforementioned constraints on themixing parameter vector prevent the spatial prior mixing distributionfrom being the proposal distribution in the simulation of the mixingvectors. Therefore, a Dirichlet density was selected as the mixingvectors’ proposal distribution. This density fulfils the mixingconstraints whilst matching the approximate form of theoreticalmixing densities previously found in [3].Performance assessment was undertaken with publicly availabledata sets: (1) Simulated MR brain data from the Brain Imaging Center,Montreal Neurological Institute, Canada [4]; (2) 20 normal MR braindata sets and their manual segmentations from the Center forMorphometric Analysis at Massachusetts General Hospital, USA, [5],with variable image quality due to the PV effect and other imagingartefacts such as intensity inhomogeneities. The simulated MR braindata also provides PV ground truth values while the 20 normal MRIdata sets have expert discrete voxel manual segmentations, i.e. no PVground truth. Therefore the voxel RMS error metric was used on thesimulated MR brain data sets while the discrete metric known as theJaccard similarity measure was used for the 20 normal MRI braindata sets. The Jaccard metric measures the amount of ‘overlap’between the classifier output and the ground truth where 1.0 is perfectand 0.0 is not segmented at all.ResultsAfter applying the GSR classifier to the simulated MR brain data with3% noise, we found the classified voxel RMS values were 12% and9% for GM and WM respectively. This compares to RMS values in[6] using three fuzzy clustering techniques: 14%, 15%, 12% for GMvoxels and 13%, 11%, 11% for WM voxels. 5%, 7% and 9% noisedata sets were also classified with results given in table 1 togetherwith alternative PV classifiers from [7] which are Maximum aposteriori (SMAP) and Maximum Likelihood (SML) techniques thatmodel the PV voxels as coming from separate classification classes.The results in table 1 illustrate the superior performance (of GSR) forthe majority of the classified PV voxels. GSR classified GM and WMimage slices for the simulated brain data with the most noise (9%) canbe seen in fig. 1. Visual comparison with the ground truth (also infig.1) illustrates the similarity between the ground truth and the outputof GSR.Table 1: RMS PV classifier errors for the simulated MRI brain data sets withnoise of 3 – 9%. SMAP, SML are PV classifiers in [7].Figure 1: GSR classifier output (WM-a GM-b) and PV ground truth maps(WM-c,GM-d) for 9% noise simulated MRI brain data.The results of applying GSR on the 20 real MRI data sets alsoproduced promising results with mean Jaccard similarities: 0.57 forWM voxels and 0.58 for GM voxels. A Jaccard value of 1 is optimum,although human expert segmenters only achieve between 0.8 and 0.9,[5,8]. GSR does not include any modelling of the prominent intensityinhomogeneities present in most of the data sets and could beimproved significantly with additional model refinements. Thesevalues can be compared with a number of standard baseline classifierswhose Jaccard similarity results in [5,8] are: 0.57, 0.56, 0.57, 0.55,0.55, 0.57 for the WM voxels and 0.56, 0.56, 0.47, 0.55, 0.54, 0.48 forthe GM voxels.ConclusionThis short abstract has briefly described a novel per voxel gradientcontrolled and adaptive spatially regulated Bayesian formulation forPV classification of MR image data. Classifier performanceassessment has been undertaken and the results have been found to becompetitive and in some cases superior to other high performanceclassifiers in the medical image analysis literature. Further work willinclude modelling of intensity inhomogeneities and other imagingartefacts, to further improve the classification rate.Acknowledgements The authors gratefully acknowledge the Center forMorphometric Analysis at Massachusetts General Hospital and the BrainImaging Center, Montreal Neurological Institute for the availability of the datasets. This work was supported by UK EPSRC.References(1) Gudbjartsson H, Patz S. Magn. Reson. Med. 34, 910-914 (1995)(2) Choi, HS et al. IEEE Trans. Med. Imag. 10, 395-407 (1991)(3) Chiverton J, Wells K, IEEE Sig Proc Letters, 13, 369-372 (2006)(4) Kwan R et al, Lecture Notes in Comp. Sci. 1131, 135-140, (1996)(5) available online: http://www.cma.mgh.hardvard.edu/ibsr/(6) Pham DL, Prince J, In Proc. IEEE MMBIA, 170-177, (2000)(7) Shattuck DW et al Neuroimage 13, 856-876 (2001)(8) Rajapakse JC, Kruggel F, Image Vis. Comp. 16, 165-180, (1998)


Shielded biplanar gradient coil design for an open permanent low field MRIP4Li Sze CHOW 1 , Kuan J.Lee 1 , Jim M Wild 1 , Martyn N.J.Paley 11 Section of Academic Radiology, Univeristy of Sheffield, Royal Hallamshire Hospital, Glossop Road, S10 2JFIntroductionThis study presents the design of an optimized shielded biplanar gradient coil setused for an open permanent low field (0.2T) magnet, which will be used forlung imaging in the standing position with hyperpolarized (HP) gas (He-3).Many formulations have been proposed for the design of shielded gradients overrecent years [1, 2]. Our design was based on a modified version of the coil setproposed by Tomasi et al [3]. A half-size prototype of the longitudinal gradientcoil was built for testing and compared with the simulation.MethodsSimulation of the gradient field using the Biot-Savart law for both longitudinaland transverse gradient coils were performed with Matlab. The design used fourplanes perpendicular to the z axis where the primary and shielding planes wereplaced at z = ± a and z = ± 1.15a respectively for an efficient shielding effect.Since the shielding planes are restricted at ± 220mm to fit in the magnet face,then a ≈ ± 191.3mm.Longitudinal gradient coilThe Gz coil used 34 circular wires (with radius R p , p = 1,...,34) for the primaryplane and 58 circular wires (with radius R s , s = 1,...,58) for the shielding. Thewire distribution in both primary planes was identical to maximize the fielduniformity. The currents in each plane were opposite; the shielding coil currentwas a quarter of the primary coil current and in opposite direction to therespective primary coil current in each plane [3].Transverse gradient coilThe Gx and Gy coils used similar design of 2n straight wires of length 450mm at±X p and ±Y p (p=1,…,n where n=42) respectively on each primary plane, andanother 2n straight wires of similar length at ±X s and ±Y s (p=1,…,n) for theshielding planes. The primary coil currents flow in the same direction for allwires: along y for Gx coil, and along x for Gy coil, respectively. The shieldingcoil is used as a return path for the primary coil current, hence in the oppositedirection to the primary coil current. The gradient optimization used a simplexsearch method [4] to get the optimum radii (R p ,R s ) of each wires for the Gz coiland the optimum position (X p ,X s ,Y p ,Y s ) of the straight wires for the Gx and Gycoils.ResultsTable 1 records the optimized radii of the Gz coil and the positions of the Gxand Gy coils. Fig. 1 shows the simulated fields for each of the coils and thegradient along a middle profile of each of the respective planes. The maximumand minimum points of these fields were at z = ± a; the field drops to zerooutside the shielding plane. The simulations give a field strength of 0.47mT/m/Aand 0.27mT/m/A for the Gz and Gx or Gy coil respectively. The theoreticalvalues of inductance and resistance are 0.51mH and 1.19Ω for the Gz coil, and4.39mH and 0.44Ω for the Gx or Gy coil, with 2mm diameter copper wire.Table 1: Optimized radii and positions for the biplanar gradient coil.R P(mm)2 turnseachR s(mm)(N) indicate N turnswith same radius.X p or Y p(mm)3 wireseachX s or Y s(mm)3 wireseach5480150160166176178182198204206212226230240244252345464748494104114124134146 (2)156 (3)166 (3)176 (4)186 (4)196 (5)206 (4)216 (4)226 (4)236 (4)246 (4)256 (3)266 (3)276± 0± 18± 30± 50± 51± 52± 52± 53± 64± 66± 67± 70± 88± 180± 2± 11± 25± 46± 167± 168± 170± 175± 175± 177± 177± 177± 179± 185A half-size Gz prototype was built using 0.1mm diameter copper wire. Thetheoretical values of inductance and resistance are 0.25mH and 238Ω. Themeasured fields of the shielded and unshielded coils along the coil axis aredisplayed in Fig. 2. The ratio of V shielded /V unshielded is 0.28 at z = 15cm if bothpeaks are normalised to the same value, which is a suppression of 72% of the zcomponent of the magnetic field. Fig. 3 shows the phase images (sagittal plane)obtained on a 1.5T MR scanner where a bottle phantom was placed in the centreof the coil, and another phantom was placed outside one of the shields. When acurrent (28mA) was applied to the Gz coil (Fig. 3b), the phase image showed thecontours of the constant Bz field from the Gz coil but only a small effect on thephantom outside the shield. The linear Gz gradient was obtained from the linearphase ramp in Fig.3b, using function polyfit and polyval in Matlab and found tobe approximately 4.0mT/m/A along the coil axis of the half-size prototype. Thelinear gradient was then taken out to investigate any residual inhomogeneity dueto the Gz coil as shown in Fig.3c, which is fairly smooth in the target region.The largest phase shift in this region due to the inhomogeneity was 2.77 radiansat the bottom right corner pixel (see arrow), equivalent to ≈ 0.62µT.(b)(a)Fig. 1: Simulated fields (mT) for (a) Gz coil on x-z plane, (b) Gx coil on x-y plane, (c) Gycoil on x-y plane.arbitrary unit8006004002000-400-600-800(c)-25 -15 -5 -200 5 15 25Position (cm)ShieldedUnshieldedFig. 2: Measured fields of the half-size Gz coil using an oscilloscope from the signalinduced in a search coil when the Gz coil was driven with a low frequency AC current.(a)(c)Fig. 3: Phase images: (a) without Gz coil, (b) 28mA applied to the shielded Gz coil, (c)residual Gz coil’s inhomogeneity. Rectangle shows a target volume. (*pixel with largestphase shift due to inhomogeneity.)ConclusionAn optimized shielded biplanar coil was designed for vertical MR whichsuccessfully suppresses the magnetic field outside the coil.ReferencesBottle phantomspixel*primaryplane[1] Turner et al. Mag.Reson.Imag. 11:903-920 (1993)[2] Yoda et al. J.Appl.Phys. 67:4349-4353 (1990)[3] Tomasi et al. J.Mag.Reson 140, 325-339 (1999)[4] Lagarias et al. SIAM J.Optimization 9,112-147 (1998)Acknowledgement: We acknowledge DH-LINK (HTD-021) for funding this work.(b)shieldingplanephantomoutside shield


Measuring T 2 using T 2 Prepared Balanced Turbo Field Echo Imaging at 3.0T and 7.0TP5E. F. Cox 1 , C. L. Hoad 1 , P. A. Gowland 1 ,1 Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, Nottingham, UKIntroductionImaging using the balanced turbo field echo (bTFE or TrueFISP)produces rapid images of high SNR with minimal distortionsparticularly in the abdomen and at high field strengths. It isparticularly important to be able to measure relaxation parameters at7.0T to optimise image contrast at this field strength. It has previouslybeen shown that it is possible to measure T 1 using this sequence with asuitable preparation pulse 1 . T 2 preparation pulses can also be usedwith bTFE sequences 2 (T 2 -prep-bTFE). However quantifying T 2 fromthe bTFE data is not trivial as the final magnetisation at the centre ofk-space will depend not only on the preparation phase but also on thenumber and timing of the r.f. pulses of the acquisition. This work aimsto determine how accurately T 2 can be measured with bTFE usingboth simulations and data from phantom and in vivo experiments.MethodsSimulations: Signal intensities were simulated with a T 2 preparationphase of a 90º/180º/90º pulse combination (total time TE) followed bya half-fourier bTFE acquisition block consisting of a α/2 preparationpulse and successive alternating ± α pulses with a TR of 3ms. Therewas a time delay of 15s at the end of the bTFE acquisition to allow thesystem to fully relax to maximise SNR. The signal intensity at thecentre of k-space was simulated at several TEs between 50ms and1500ms for both long (T 2 =0.5s, T 1 =2s) and short (T 2 =0.05s, T 1 =0.5s)T 2 /T 1 combinations. Both the fitting and simulation program modelledthe signals by repeated application of rotation (1) and relaxation (2)functions:(1) (2)Mx= M xMx= Mxexp( − t T 2)My= Mycos( α ) + Mzsin( α ) My= Myexp( − t T 2)Mz= −Mysin( α ) + Mzcos( α ) Mz= Mzexp( − t T1 ) + M0( 1−exp( − t T1))For each multi-TE data set, simulated for different combinations of T 2 ,T 1 , M 0 and α, a Powell fitting algorithm 3 was used to fit combinationsof the parameters that would be unknown for real images, i.e. T 2 , T 1and M 0 . T 2 , T 1 and M 0 4 were fitted for estimates of α between 15º and80º with simulated α of 30º, 40º and 50º. Gaussian noise was added tothe simulated data before fitting and the effect of changing the noiselevel was investigated.Experimental: Abdominal imaging was carried out on a 3.0T PhilipsAchieva whole body MRI scanner using the SENSE Torso coil. Sixhealthy volunteers were scanned on a single occasion andmeasurements of T 2 in the liver, spleen, kidney and skeletal musclewere acquired, from 2 different slices through the abdomen using theT 2 -prep-bTFE sequence. Half Fourier acquisitions were used toincrease sensitivity to the preparation step. Decay curves weregenerated using 8 different T 2 preparation times (20 to 400 ms). BTFEdata was fitted by modelling the evolution of the magnetisation aftereach r.f. pulse in the sequence, and the Powell algorithm 3 minimisedthe difference between measured and modelled data to fit T 2 , T 1 andM 0 . Phantom and head imaging was carried out on a 7.0T PhilipsAchieva head MRI scanner using the T/R head coil. A gel phantomwith 4 quadrants of differing T 1 and T 2 was scanned using T 2 -prepbTFEand Spin-echo EPI imaging sequences to compare measured T 2values. Work is in progress to quantify T 2 in different regions of brainusing the T 2 -prep-bTFE sequence.ResultsSimulations: Figures 1 and 2 show sensitivity of fitted values of T 2and M 0 respectively to estimates of flip angle, when fitting for T 2 , T 1and M 0 , for T 2 /T 1 =0.5/2s. The fitted values of T 2 are extremelyinsensitive to the estimates of α (Figure 1), indicating that thesequence can be used in regions of B 1 inhomogeneity. Similarly thefitted values of T 1 were very insensitive to the estimated value of α.When α is underestimated, the fitting program compensates byincreasing M 0 (Figure 2). For T 2 /T 1 =0.05/0.5s the fitted T 2 valuesshowed no detectable dependence on the estimate of α. Table 1 showsthe effect of adding gaussian noise to simulated data.Experimental: Table 2 shows the mean data for the six healthyvolunteers. Table 3 compares the fitted T 2 values using the T 2 -prepbTFEsequence and the Spin-echo EPI sequence at 7T.ConclusionThese simulations show that T 2 can be measured accurately using abTFE sequence, which has particular advantages in the abdomen athigh fields, where EPI techniques are not applicable and HASTE orRARE techniques have a very high SAR. The fitted T 2 values becameless accurate with more spread in the data as the noise increased.Results from the six healthy volunteer were similar to previouslypublished data 5 for the liver and spleen, however measurements of thekidney and skeletal muscle differed more. Partial volume effects inthese structures may account for the differences as well as the smallsample size. The bTFE phantom T 2 results at 7T are in goodagreement with the EPI data, although all T 2 -prep-bTFE values wereslightly higher. This may have been due to the fitting program usinginstantaneous r.f. pulses, whereas the experimental data had finiteduration r.f. pulses or due to the assumption that the SNR isdetermined by signal at the centre of k-space. Future work will use theapproach described here to determine the most efficient protocol (interms of signal to noise per unit time, insensitivity to sequenceparameters and minimum total imaging time) with which to measureT 1 and T 2 simultaneously using TrueFISP 6 .References1. Scheffler K, et al. Mag. Res. Med. 45; 720-723(2001)2. Kaul MG, et al. Rofo. 176; 1560-1565 (2004)3. Press WH, et al. Numerical Recipes in C (2ndEd.); 4124. Hoad CL, et al. Proc. ISMRM P2506 (2006)5. Bazelaire CMJ, et al Radiol. 230; 652-659 (2004)6. Newbould R, et al. Proc. ISMRM P2191 (2005)Figure 1: Graph to show how fitted T 2 varies withestimates in flip angle. Simulated T 2 =0.5s.% NoiseT 2 =0.5s, T 1 =2sT 2 =0.05s, T 1 =0.5s0.050.5 ± 0.00390.05 ± 0.00080.10.5004 ± 0.00980.05 ± 0.00170.20.5043 ± 0.02140.0502 ± 0.00320.50.5076 ± 0.04090.0513 ± 0.00851.00.4973 ±0.06260.0549 ± 0.0353Table 1: Fitted T 2 results (in seconds) for data simulatedwith T 2 shown with added noise (mean ± std. dev)Figure 2: Graph to show how fitted M 0 varies withestimates in flip angle. Simulated M 0 =0.5s.NT 2Liver60.029 ± 0.007Spleen60.054 ± 0.013Kidney (Medulla)60.053 ± 0.009Skeletal Muscle30.079 ± 0.034Table 2: In vivo measurements (in seconds)of 6 healthy volunteers scanned on a singleoccasion (mean ± std. dev)RegionEPIbTFEA0.0370.039 ± 0.002B0.0800.084 ± 0.003C0.0690.074 ± 0.001D0.0770.081 ± 0.003Table 3: T 2 results of phantom experiments at 7T(in seconds), showing EPI and mean ± std. devaveraged over 6 different bTFE experiments.


An asymmetric quadrature birdcage body T-R coil for hyperpolarised 3 He lung MRIN de Zanche 3 , N Chinna 2 , K Teh 1 , C Randell 2 , JM Wild 11 Academic Radiology, University of Sheffield, UK, 2 Pulseteq Ltd, UK, Biomedical Engineering, 3 University and ETH, Zurich, SwitzerlandP6Introduction Birdcage resonators have a spatiallyhomogeneous B1 profile, which is important inhyperpolarised (HP) gas MRI given its inherentsensitivity to flip angle due to the non-renewablepolarisation. In this work an asymmetric quadraturebirdcage body coil is presented for HP 3He lung MRI.The coil performance is compared to that of a flexibletwin-saddle T-R coil currently in use in many sitesworldwide.Methods The positions of the coil’s 12elements and shield shape were determined usingconformal mapping methods [1] and performance wasmodelled using a method of moments package (NEC-2).Overall dimensions: length 600 mm, width 588 mm,height 450 mm. The coil is a ‘split design’ for ease ofpatient/volunteer access/positioning. The coil fits insidethe proton body transmit coil and consequently wasshielded to minimise coil coupling. Network analyzermeasurements were performed inside the magnet toascertain Q and port isolation for a range of subjectloading (51 kg, 80 kg, 110 kg).3 He MRI was conducted on a 1.5T magnet with T-Rcapabilities at 48.5 MHz. Gas was polarized to 30%with rubidium spin-exchange apparatus. A 2-D B1mapping sequence [2] was used to characterize the coilhomogeneity. A large phantom (100 l bag filled with100 ml 3He and 100 l N2) was used to map B1. The invivoperformance of the two coils was assesed in avolunteer study on the basis of B1 homogeneity andwith imaging comparisons using an optimised 3Dgradient echo sequence [3].Results & DiscussionIsolation between quadrature ports was measured asbetter than -15dB when loaded with 51, 80 and 110 kgsubjects, while Q unloaded = 182. The ratio of theQ unloaded /Q loaded > 2 for the 50 kg load. Fig.1 shows B1field simulation of the coil in the axial x-y plane. Fig.2shows a B1 homogeneity map acquired from the floodphantom –a good agreement with theory of Fig.1 isdemonstrated. Fig. 3 shows the coil in position on thescanner about to be placed inside the bore. Fig.4 showsin-vivo B1 homogeneity maps derived from a 2D B1mapping sequence [2] from the same volunteer with thetwo coils. The B1 mapping sequence was run with thesame B1 pulse voltage for a direct comparison of RFpower requirements. The birdcage shows a morehomogeneous flip angle across the whole lung structure(B1 inhomog. < 8%) when compared to the flex coil(B1 inhomog. >16%). The birdcage is also less powerdemanding, delivering almost twice the flip angle for agiven B1 voltage. Fig. 5 shows in-vivo images acquiredwith a 3D sequence [3] from the same volunteer withthe two coils using the same volume of gas. The SNR ison average 1.5 times higher with the birdcage coil whenaveraged over all slices in both lungs. Moreover theimage quality is superior as a result of the homogeneousflip angle introducing less blurring via broadening of thepoint spread function due to RF depletion of the finiteHP gas magnetisation [4].3Flex coil: mean flip = 5.5° +/- 0.9°Birdcage coil: mean flip = 9.6° +/- 0.7 °2Fig. 4 B1 maps of the two coils acquired from same slice in same volunteerFlex coilFig. 5 Ventilation images from the same slice in the same volunteer from thetwo coils. The bircage coil provides images with more spatial homogeneity andconsistently higher SNR over all slices.Conclusion A novel whole body asymmetricbirdcage design has been demonstrated for lung imagingof HP 3He. The improved B1 homogeneity of the coil ishighly desirable in HP gas MRI where optimum use ofpolarisation is a major challenge. The shieldingmitigates coupling interactions with the proton bodycoil and allows efficient quadrature performance. Whilethe coil has been demonstrated for use in lungs with3He, the design is transferable to other nuclei whichmay require a uniform B1 over the whole body e.g. HP129Xe and HP 13C. The compact body coil design mayprovide a solution to whole body imaging at higher B 0where RF power constraints are experienced.References. [1] Magn Reson Med;53(1):201-211 [2]Magn Reson Med. 2005 ;53(5):1055-1064. [3] MagnReson Med. 2004;52(3):673-8 [4] Magn Res Med2002;47:687-695AcknolwledgementsEPSRC, UK. Grant # GR/S81834/01(P).Spectra gases, GE Healthcare9876543Birdcage coilFig. 21412108642


Negative BOLD accompanies sharpening of whisker representations in rat barrel cortex visualized with functionalmagnetic resonance imagingBenito de Celis Alonso 1 , Andrew S. Lowe 1 , Aisling L. Dixon 1 , John P. Dear 2 , Kalok C. Lee 3 , Steven C. R. Williams 1 and Gerald T. Finnerty 11 MRC Centre for Neurodegeneration, King’s College London, De Crespigny Park, London SE5 8AF, UK. 2 Department of Mechanical Engineering, Imperial CollegeLondon, London SW7 2AZ UK. 3 Division of Engineering, King’s College London, Strand, London WC2R 2LS, UK.Email: spgtbca@iop.kcl.ac.uk (Benito de Celis) or spgtgtf@iop.kcl.ac.uk (Gerald T. Finnerty)P7Introduction:Representations of the sensory periphery are commonly arranged in neocortex astopographic maps 1 . These cortical maps are dynamic, often varying with the frequency ofafferent input. For example, rat whisker representations in cortex sharpen with repetitivedeflection at frequencies in the upper part of the physiological whisking range (4 – 12 Hz)as has been demonstrated with optical and Doppler imaging 2 . Here, we imaged sharpeningof rat whisker representations using blood oxygen-level dependent (BOLD) fMRI.Figure 2: Representation of C1 and C2 whiskers at 3, 7 and 10 HzAMethods:Animal Preparation.Anaesthesia of 11 male Sprague Dawley rats was performed with a bolus of α-chloralose(65 mg/kg) and continued with a 30 mg/kg/hour infusion of the same anaesthetic.Respiratory rate, cardiac rate and rectal temperature (37.0 ± 0.5 °C) were monitored duringscanning.Whisker stimulation.All whiskers except C1 and C2 were trimmed (Fig. 1).Whiskers were placed in the teeth of a comb thatmoved rostro-caudally (8 mm displacement) at 3 Hz,7 Hz or 10 Hz. ON and OFF blocks were randomized.Figure 1: Whiskertrim patternMRI and fMRI methods.Imaging was performed in a 9.4T magnet (Oxford Instruments, Oxford, UK) with a 25mm diameter transmit/receive surface coil (Varian, Palo Alto, CA, USA).Anatomical scans: Spin Echo sequences, TR/TE = 1000/20 ms ; 0.5 mm thickness; field ofview of 3.2 x 3.2 cm; matrix size of 192 x 192 averaged 4 times. fMRI data: Five-Echo-Gradient-Echo sequence. Flip angle, 31 degrees ; TR, 340 ms; TE = 4, 8, 12, 16, 20 ms;FOV, 32 x 32 mm; matrix, 96 x 96; brain volumes of 12 slices 0.5 mm thick employing32.95 seconds per block. Each fMRI session consisted of 120 blocks (60 ON and 60 OFF),lasting for 1 hour 5 minutes.BCData analysis: Data was analysed with SPM99. The contribution of large draining vesselswas reduced by compiling a coefficient of variance map 3 . Probabilistic independentcomponent analysis (PICA) was performed using MELODIC to reduce noise. The designmatrix included muscle signal as covariate of no-interest serving as an independentmeasure of whether global effects modified the BOLD signal during the imaging session.MarsBar tool (MARSeille Boîte À Région d'Intérêt) was used to measure the size ofclusters, the changes in BOLD signal intensity and P values for voxels in different ROIs.Results:3 Hz stimulation evoked a positive BOLD response (PBR) in contralateral neocortex in theSI region 2.5 – 3 mm caudal to bregma (Fig 2A). 7 Hz stimulation overlapped the 3 Hzactivation, but was larger and extended laterally towards SII (Fig. 2B). 10 Hz whiskerstimulation evoked two PBRs in contralateral neocortex in SI and SII. The PBRs wereseparated by a negative BOLD response (NBR). NBRs were also present ipsilateral to thewhisker stimulation homotopic to the PBRs in contralateral SI and SII (Fig 2C).ROIs and a line drawing from a rat atlas are superimposed on a brain slice. The colourcode (Red, PBR; blue, NBR) denotes the principal BOLD response in each ROI (Fig. 3A).Change in BOLD signal in SI (filled circles) and SII (open circles) with varying whiskerdeflection frequency (Fig. 3B). Relationship between maximum signal intensity in SI andSII. The dashed line is the unity line. The solid line is the linear regression fit to thevalues. The PBR in SI had a larger amplitude than the PBR in SII (n = 22, P < 0.001,signed rank test). The amplitudes of PBRs in SI and SII were positively correlated (r =0.60, P = 0.003, n = 22) (Fig. 3C). Relationship between the maximum PBR in SI and themaximum NBR in adjacent cortex (ROI M and ROI N). A linear regression has been fittedto the values (Fig. 3D).ACFigure 3: Single subjects data analysisBDConclusions:1. Deflection of two whiskers at 3–10 Hz evokes a positive BOLD response in SI andSII.2. The PBRs in SI and SII are correlated.3. The NBR that separates and partially surrounds the activations in SI and SII emergeswith whisker deflection frequencies associated with sharpening of whiskerrepresentations.4. Ipsilateral hemisphere negative BOLD response was correlated with the amplitude ofthe positive BOLD response in adjacent SI suggesting coupling of the PBR andadjacent NBR.References:1. Kaas, J.H., 1987. The organization of cortex in mammals: Implications fortheories of brain function. Annu. Rev. Psychol. 38, 124-151.2. Sheth, B.R., Moore, C.I., Sur, M., 1998. Temporal modulation of spatialborders in rat barrel cortex. J. Neurophysiol. 79, 464-470.3. Hlustik P., Noll D.C., Small S.L., 1998. NeuroImage 7, 224-231.All scans were performed at the MRC Biological Imaging Centre, Imperial CollegeLondon. Supported by the Wellcome Trust.


Fast and Accurate Mapping of the Flip Angle Using the 180° NullP8Nicholas G Dowell and Paul S Tofts1 Institute of Neurology, UCL, Queen Square, London WC1N 3BG; UKIntroductionUncertainty in excitation flip angles due to RF field B 1nonuniformity (NU) is the largest cause of error in quantitative MRimages and spectra.(1) A knowledge of the actual flip angle at everyregion in the volume will improve the accuracy and precision ofmany qMR parameters. Here, we introduce a technique that candeliver a flip angle scale factor map covering the whole brain in lessthan 4 minutes without confounding effects from T 1 , T 2 and protondensity (PD). This rapid acquisition is an advantage over thecommonly-used double angle method DAM for mapping B 1 ,(2)which needs a lengthy TR to remove any T 1 dependence.MethodThe flip angle scale factor, given by the Greek letter zeta ζ, relatesthe nominal (system calibrated) θ n and actual θ a flip angle, i.e. θ a =ζθ n . The NMR signal from a spoiled gradient echo sequence is givenby M xy = M 0 [1 – exp{–TR/T 1 }] sinθ a / [1 – cosθ a exp{–TR/T 1 }],where M 0 is the magnetisation at thermal equilibrium.(1) Figure 1shows a plot of the gradient echo signal versus flip angle. The signalintensity has a TR/T 1 dependence at θ ≠ 180°, but the NMR signal isnulled following a 180° pulse, irrespective of T 1 .(3) Therefore, ifθ n null (the nominal flip angle that gives a signal null) is known, thenthe value of ζ can be calculated, since ζ = 180°/θ n null . In the region ofthe null-point, the signal has an approximately linear dependence onflip angle θ a in the vicinity of the null-point. Consequently, it ispossible to use linear regression (3) to determine θ n null on a pixel-bypixelbasis from a series of images acquired with flip angles in thelinear region where θ a ≈ 180°.The method was experimentally validated by manually adjusting thetransmitter output amplitude by a factor of 1.122 and 0.891 (= ±10TG units on a GE scanner) and measuring the resulting change in ζfrom the flip angle scale factor maps. In addition, the accuracy ofthis technique was evaluated under off-resonance conditions byvarying the transmitter output frequency and measuring the changein ζ.System: GE Signa 1.5 T with a birdcage head coil. Sequence: Simplymodified 3D spoiled gradient echo (GRASS), without the need forrecoding the sequence. Imaging parameters: acquisition matrix =128x64; TR = 33 ms, TE = 6 ms; slices = 28; slice thickness = 9 mm.Flip angle scale factor maps were constructed from only three seriesof images. The set of flip angles (θ n = 145°, 180°, 215°) wasoptimised numerically for improved accuracy. Overall scan time:3x70 seconds. The technique was applied to water and oil phantomsand was also used in vivo. Off-resonance conditions were investigatedby changing the transmitter frequency and observing the effect on ζ.ResultsFigure 2a shows an axially reformatted flip angle scale factor map(using sagittal-plane images) of the brain. It demonstrates the methodcan report the flip angle scale factor in spite of variations in protondensity and T 1 across the brain, even from the CSF-filled ventricles.The profile (Fig. 2b), extracted from the region indicated by the line inFig. 2a, reveals that B 1 is enhanced at the centre of the head due to thepresence of RF standing waves.(4,5) The flip angle map of a waterphantom (Fig. 2c) shows a similar effect, although water has a highdielectric constant that results in a much larger RF field at the centreof the phantom. This is clearly demonstrated by the accompanyingprofile (Fig. 2d).Validation was completed when the flip angle maps accurately determined(to better than 0.1%) the change in ζ owing to the manual increasein transmitter output as described above.Figure 1Plot of simulated spoiled gradient echo signal intensity versus flip angleθ a for a range of T 1 .(1)Figure 2(a) Axially reformatted flip angle error map of the brain. The verticalline indicates the region from which the profile in (b) is extracted. (c)Slice from a flip angle error map of a 15-cm diameter spherical waterphantom and (d) the intensity profile extracted along the regionindicated with the horizontal line in (c).ConclusionsThis method can accurately and precisely determine a flip angle mapcovering the entire brain in less than 4 minutes using a pulse sequenceavailable on commercial scanners. Consequently, this technique canbe incorporated into MR protocols to accurately determinequantitative MR parameters in the presence of B 1 inhomogeneity.Reference List1. Tofts PS. The Measurement Process. In: Tofts PS, editor.Quantitative MRI of the brain. Chichester: Wiley, 2003: 17-54.2. Stollberger R, Wach P, McKinnon G, Justich E, Ebner F. In:Proceedings of the SMRM, 7th Annual Meeting, San Francisco, 1988;106.3. Venkatesan R, Lin WL, Haacke EM. Magnetic Resonance inMedicine 1998; 40(4):592-602.4. Barker GJ, Simmons A, Arridge SR, Tofts PS. British Journal ofRadiology 1998; 71(841):59-67.5. Tofts PS. Journal of Magnetic Resonance Series B 1994;104(2):143-147.


P9Investigate the dependent of P300 on Target to Target IntervalSally Eldeghaidy, Kay E. Head, Penny Gowland, Sue FrancisSir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham,University Park, Nottingham, England, NG7 2RDIntroduction:Simultaneous EEG-fMRI for evoked activity has a growing interest in understanding brain function due to the highspatial and temporal resolution. The event related P300 is widely believed to be a neural signature of both attentionand working memory of task-relevant stimuli [1]. Many factors affect P300’s amplitude and latency such asprobability, inter-stimulus-interval (ISI), and ordering of targets within the sequence [2] which all effect the target totarget interval (TTI) [3, 4]. This study aimed to investigate the TTI effect on P300 ERP and fMRI and to determine anycorrelation between amplitude of the BOLD signal and ERP P300 response to an auditory oddball task.Methods: The paradigm was initially tested outside the scanner on three subjects to investigate the effect of the TTIand subject response (count vs button), to optimise the P300. A pilot study of two volunteers with simultaneous EEGand fMRI was then performed. All subjects gave written consent and had no known neurological disorders.Paradigm: An auditory oddball paradigm with 94 % non-target stimuli (1979 Hz tones) and 6 % target stimuli (24tones, 2095Hz). Stimuli were generated using presentation (http://www.neurobehavioralsystems.com/software/presentation) and delivered to the subject via MR-compatible headphones in a pseudorandomised order, (TTIs) were8.8, 15.4, 19.8, 35.2 s (6 tones each) and ISI 2.2 s. All participants reported they could discriminate the stimuli frombackground scanner noise. Participants were asked to silently count target tones with their eyes closed. Beforebeginning the task, each participant performed a practice block of trials.EEG Recording and Analysis: EEG data were recorded on a MRI compatible 32 Ag/AgCl channel brain visionrecorder (Brain Products, Munich, Germany) placed on the scalp according to the 10-20 international system;electrode skin impedance was less than10 KOhms. Data were collected with a sampling rate of 5 KHz and processedusing Brain Vision Analyzer software (Brain Amp, Brain Products, Munich, Germany). Cardio-ballistic pulse artefactsin the EEG traces were corrected using the ECG electrode and the VCG from the MRI scanner, and MRI artefactswere detected and corrected with the EEG software.fMRI Acquisition and Analysis: Data was acquired on Philips 3 T MRI scanner using a SENSE head RF coil. 14transverse GE-EPI images [64 x 64 matrix size, echo time (TE) of 35 ms, and 3 x 3 x 6 mm voxel size] were acquiredevery 1 s throughout the fMRI paradigm. Functional MRI data were acquired in six sessions in a 3 hour period, eachsession took 17 minutes. fMRI data were analysed using SPM2. Data was corrected for slice timing, realigned,normalised to MNI co-ordinates, and spatially smoothed with a 5 mm Gaussian kernel. Statistical parametric mapswere then generated by modelling the targets at each of the four TTIs versus the standard non-target stimuli whichwas used as the baseline condition. Activation maps were thresholded at p


23.0 22.5ppmResolving conformational exchange in pimonidazole – a hypoxia markerCristina Gabellieri 1 , Thomas R. Eykyn 1 , Geoffrey S. Payne 1 , Martin O. Leach 11 Cancer Research UK Clinical Magnetic Resonance, The Institute of Cancer Research,Royal Marsden Foundation NHS, Sutton, Surrey, United KingdomP10IntroductionPimonidazole is widely used as a marker for qualitative and quantitativeassessment of tumour hypoxia (1). Under conditions of low oxygentension the bioreduction process is irreversible and the product isbound selectively under hypoxic conditions.Pimonidazole hydrochloride (Hypoxyprobe TM ) is highly water-solubleand is used to detect hypoxia and oxygen gradients at the cellular levelwithout tissue disruption, and is extensively used for immunohistochemicalvalidation of hypoxic conditions. Tumour hypoxia is associatedwith cancer aggressiveness and is recognised as a limiting factorin the successful treatment of solid tumours by conventional radiotherapyand chemotherapy. However, its unique occurrence in suchtumours is also viewed as offering the possibility of selective chemotherapy.Related compounds have been developed as MR hypoxiamarkers such as SR4554. The bound adduct is retained in hypoxicregions and may be observed by 19 F MRS. Thus the dynamic propertiesof these and other related compounds is of interest to better understandthe binding process.We present a study of the conformational exchange in pimonidazole.Exchange between different magnetic sites has been observed. At lowtemperatures, just above the freezing point, the rates of exchange aresmaller than the chemical shift difference between exchanging sitesand the spectra are well resolved. As the temperature is increased therates of exchange become appreciable to the difference in chemicalshifts and the NMR spectrum is drastically affected. Employing 13 CNMR the conformational exchange process is unambiguouslyattributed to ring flipping of the 6-membered ring.NMR MethodsNMR experiments were conducted on a Bruker Avance 500MHzspectrometer (Bruker Instruments, Germany). Pimonidazole wasdissolved in D 2 O (380 mM). 1 H and 13 C spectra were recorded as afunction of temperature. Experimental peaks have been assigned withthe help of ACD/HNMR and ACD/CNMR predictor sofwtare(ACD/Labs).Two-dimensional 1 H- 1 H EXSY (EXchange SpectroscopY) have beenrecorded to measure the rates of chemical exchange and thereby leadto a better understanding of the underlying processes. Positive offdiagonal peaks give information about chemical exchange, whilenegative off diagonal peaks show cross-relaxation processes.Exchange rates have been calculated with the programme ExsyCalc(MestreC) and assumes slow chemical exchange on the time scale ofthe chemical shift difference.ResultsFigure 1 shows 1 H spectra at287K and 337K. Two distinctgroups of protons may beobserved, those that arebroadened at high temperatureindicating exchange and thosethat remain narrow, indicatingthe absence of exchange. Figure2 shows13 C spectra ofpimonidazole at 280K and330K. The peaks undergoingexchange correspond to the fourlateral carbon nuclei in the 6-membered ring of the molecule,indicating a rotation of the ring about the axis of the nitrogen-carbonbond thus interchanging the two ortho positions as well as the twometa positions. At low temperature the peaks are separated indicatingslow chemical exchange and rates that are smaller than the chemicalshift difference (6.45 Hz for the peaks at 22 ppm). At highertemperature, the two peaks at 22 ppm collapse to a single averagechemical shift, while the two separated by 401.9 Hz (ortho position)coalesce indicating an intermediate exchange situation.T=330K4.5 4.0 3.5 3.0 2.5 2.0 1.5 ppm337KFigure 1: 1 H NMR spectra of pimonidazole in H 2 O at twodifferent temperatures.401.9 Hz70 65 60 55 50 45 40 35 30 25 20 ppmFigure 2: 13 C NMR spectra of pimonidazole in H 2 O at two differenttemperaturesThe exchange rates can be quantified plotting the intensity of the offdiagonalpeaks in a 2D 1 H- 1 H EXSY experiment as a function of themixing time. The fitting algorithm assumes a two-site exchangeprocess. The calculated rotation rate at T=285K is (4.4 ± 0.4) Hz,consistent with the observation in the carbon spectrum. Exchangerates have been measured as a function of temperature and allowed usto derive thermodynamic parameters for the exchange process.ConclusionWe have presented a study of conformational chemical exchange inpimonidazole. We are able to identify the dynamic process of rotationof the ring and quantify the exchange rates of this process. Knowledgeof these conformational exchange processes yields a betterunderstanding of the NMR response of this widely used hypoxiamarker.References(1) Arteel GE et. al. Br. J. Cancer 72, 889-895, (1995).(2) Bain AD Prog. Nucl. Magn. Reson. Spectr. 43, 63-103 (2003).This work was supported by Cancer Research UK [CUK] grant numberC1060/A808 and by the Council for Research Councils BasicTechnology Programme grant number GR/S23612/01.6.45 Hz


Low field, low cost multi-nuclear imaging research systemsEugeny Krjukov, James Wild, Martyn PaleyAcademic Radiology, University of Sheffield, Royal Hallamshire Hospital, Sheffield, UKP13IntroductionMagnetic resonance imaging (MRI) is a widely used and powerfulmethod for in-vivo diagnosis of disease. Unfortunately physicalrestrictions limit its application. At room temperature nuclearpolarisation is low and as a consequence MR signal amplitudes at lowfield strength are very weak. In order to increase image contrast tonoise ratio, high magnetic fields are usually applied. For instance 3Twhole body magnets are routinely used in medical practice andmagnets up to and beyond 7T are used in research. However, they areexpensive to manufacture and require cryostats and liquid helium as acoolant which increases cost dramatically. Another potential problemis nonuniform magnetic susceptibility of the object being imaged. Thisis especially significant for high magnetic fields and can reduce imagecontrast and produce distortion. Hyperpolarized contrast agents suchas 3-He or 129-Xe can be an effective solution for these problems forcertain applications where gases can be used.EquipmentMagnets: Use of parallel techniques with hyperpolarised gases shouldbe very useful as the number of RF pulses can be minimised ropreserve polarisation. Two low field resistive magnets were designedand built in order to apply the MAMBA and SPIRIT parallel imagingmethods respectively (1, 2). A corrected solenoidal magnet (S) withdiameter of 20cm and length of 100cm shown in Fig. 1a was designedfor industrial hyperpolarised gas flow chamber’ applications. A sixcoil open magnet (H) with a main coil diameter of 60 cm and anintercoil distance of 15cm, shown in Fig. 1b and 1c was designed forneonatal lung hyperpolarised 3-He imaging. Homogeneous magneticfields of 0.008T (S) and 0.015T (H) were obtained at the centre ofboth magnets respectively with stable temperatures below 60 o C (1) asreported previously (3, 4).Spectrometer: A four channel direct detection MR spectrometer wasbuilt from standard Mini-Circuit components linked to a NationalInstruments data acquisition and control system programmed inLabView software. Solenoidal RF coils and passive TR switches wereused for initial testing of the system.ResultsFigure 2a shows FIDs of 3-He and 1-H at 450 kHz in the H magnet,central and top curves respectively using the same RF coil andadjusting for gyromagnetic ratio by ramping the field. Flip angle wasclose to 90 o for 1-H and about 2 o for 3-He. 3-He nuclei were 30%optically prepolarized. The bottom curve shows the 1-H FID at 350kHz in the S magnet. Receiver amplification and spectrometer noiselevels were the same for all curves. In all cases an 18 mm diameter, 70mm long water phantom doped with CuSO4 was used with T1 ~100ms. Decay curves show good homogeneity over the phantomvolume. A strong FID could be acquired over a 300mm length of the Smagnet, as predicted, which should be sufficient for gas flow imagingapplications.Figure 2b-d show sagittal and coronal 1-H proton images and a 3-Hesagittal image. For the 1-H images, a letter F profile object withsection thickness 2mm was put inside the phantom in order todemonstrate image resolution. The 3-He image used a larger FOV asweaker gradient fields were required in order to prevent signal lossdue to diffusion. However, the spatial resolution was still similar tothat used for in vivo lung imaging on whole body systems.ConclusionWe have demonstrated operation of two simple MR systems with lowweight, low cost as well as low energy consumption which could beuseful in mobile MRI. They could be effectively used in industrialapplications (S) and in medical practice (H) where cost and/or safetyare the dominant factors. Imaging of neonatal lungs at the cotsidewhere low acoustic noise, low RF absorption and low static fieldprovide a major advantage when linked to high SNR hyperpolarised 3-He imaging should form an important future application.AcknowledgementWe acknowledge support from the Royal Society Paul InstrumentFund and DH-NEAT for this workFig. 1 a – Solenoidal magnet; 1b and 1c – Helmholtz magnet.FID (V)1050-5-101aHe 3 450kHzH 1 450kHzH 1 350kHz0 5 10 15 20 25time (ms)2cFig. 2 a – Fid’s from 3-He and 1-H at 450 kHz in the H magnet and 1-H at350 kHz in the S magnet;2b and 2c – sagittal and coronal 1-H phantomimages; 2d – sagittal 3-He image, all acquired in the H magnet.References2a1b1c(1) Paley M et al. Mag. Res. Med. 48, 1043-1050 (2002)(2) Paley M et al. Mag. Res. Imag. 24, 557-562 (2006)(3) Fichele S et al. Proc. BC ISMRM., 2005.(4) Paley M et al., Proc. BC ISMRM, 2005.2b2d


P15Using the blood oxygen level dependant magnetic resonance signal to observe the neuronal effects ofacute and chronic fluoxetine administration in the mammalian brainTamsin Langley 1 , Steven C.R. Williams 1 , Michael O’Neill 2 and Nicholas Jones 11 Institute of Psychiatry, London, U.K. 2 Eli Lilly & Co. Surrey, U.K.IntroductionPrevious research has indicated that chronic but not acute dosing ofSSRIs is required to confer a full clinical effect 1 . The initial inhibitionof the 5HT 1A autoreceptor following acute SSRI treatment and itssubsequent gradual disinhibition following chronic administration hasbeen proposed to be inherent in this slow therapeutic onset, known asthe ‘clinical lag phase’ 2 . Pre-clinical studies using various invasiveneuroimaging techniques such as microdialysis 3 , electrophysiology 2and autoradiography 4 have shown reduced neurotransmission withinthe serotonergic system, as well as reduced 5HT concentrations inspecific brain regions during this time period. In this study, we usedthe non-invasive in vivo imaging technique, pharmacological magneticresonance imaging 5 (phMRI) to examine the effects of acute andchronic dosing of fluoxetine on neuronal activation in the rat.MethodsThe acute study consisted of Sprague Dawley (S.D.) rats (n=9) beinganaesthetised and placed in a 4.7 T superconducting magnet beforereceiving an intraperitoneal (ip) dose of either vehicle or fluoxetine.The chronic study consisted of S.D. rats (n=9) receiving either vehicleor fluoxetine 10mg/kg orally (p.o.) for 20 days. On day 21 of chronicdosing, subjects were placed into the same magnet and the protocolfor the acute study was followed. For both studies subjects werescanned using a continuous, three echo, gradient echo sequence (TE=5,10,15 ms; TR =940 ms; acquisition matrix= 64x64x24; FOV=4cm 2 ;voxel resolution of 0.5x0.5x0.5mm).Whole brain volumes wereacquired for each subject every minute for 180 minutes. Following preprocessing of these volumes, SPM 99 was used to analyse the data andidentify alterations in BOLD contrast. This study was conducted inaccordance with the Animals (Experimental Procedures) Act (1986)and local ethical requirementsResultsPhMRI was sensitive to antidepressant action in the rat brain. Significantalterations in BOLD contrast were observed following both acuteand chronic fluoxetine. Acute treatment produced significant increasesin BOLD contrast in the hypothalamus and significant decreases inbrain regions including the pre-frontal cortex and dorsal and medianraphe nucleus (figure 1 a&b). Chronic treatment produced the oppositeeffect; significant increases in BOLD contrast within the dorsaland raphe nucleus and decreases in the hypothalamic region (figure 2a&b) Therefore a distinct spatio-temporal difference was identifiedwithin brain regions following acute and chronic dosing.abaFigure 2: The effect of chronic fluoxetine dosing. SPM {t}distribution maps of BOLD signal change overlaid onto co-registeredspin echo anatomical templates, a) shows the results for the vehiclegroup, b) shows the results for the drug group. Coloured pixelsrepresent significant correlation (thresholded at p < 0.05 corrected formultiple comparisons, T > 4.28) of signal time course withpharmacokinetic profile of fluoxetine. Red = positive correlation, blue= negative correlation, vehicle group n=10 drug group n=7.ConclusionsPhMRI can successfully identify a distinct change between an acuteand a more clinically relevant chronic dose of the SSRI fluoxetine.This study concurs with findings from laboratories employinginvasive neuroimaging techniques. It further supports the theory thathypothalamic and mesencephalic nuclei regions are integral inserotonergic neurotransmission 6 and the gradual desensitization of the5HT 1A autoreceptor in the dorsal raphe nucleus contributes to the slowonset of therapeutic activity observed in SSRIs. This in vivoneuroimaging tool enables the direct translation from the pre-clinicalto clinical setting and may expedite the development of therapies withincreased clinical efficacy. This study provides the first evidence thatthe differences in acute and chronic effects of an SSRI can beobserved through phMRI.References(1) Elhwuegi, Progress in Neuro-Psychopharmacology &BiologicalPsychiatry 28, 435 – 451 (2004)(2) Czachura and Rasmussen, Naunyn Schmiedeberg’s Arch.Pharmacol. 362, 266 – 275 (2000)(3) Kreiss and Lucki, J. Pharmacol. Exp. Ther. 274, 866 - 76 (1995)(4) Riad et al J. Neuroscience 24, 5420 – 6 (2004)(5) Bandettiini et al Mag. Reson. Med. 25, 390 – 397 (1992)(6) Li et al J Pharmacol. Exp. Ther. 279, 1035 – 1042 (1996)bThis project was funded by a BBSRC Case Studentship in conjunctionwith Eli Lilly & Co. Surrey, U.K.Figure 1: The effects of acute fluoxetine dosing. SPM {t} distributionmaps of BOLD signal change overlaid onto co-registered spin echoanatomical templates, a) shows results for the vehicle group, b) showsthe results for the drug group. Coloured pixels represent significantcorrelation (thresholded at p < 0.05 corrected for multiple comparisons,T > 4.28) of signal timecourse with pharmacokinetic profile offluoxetine. Red = positive correlation, blue = negative correlation,vehicle group n=10 drug group n=8.


Development of a combined microPET®-MR systemP16A.J.Lucas 1,2 , R.C.Hawkes 1 , R.E.Ansorge 2 , G.B.Williams 1 , R.E.Nutt 3 , J.C.Clark 1 , T.D.Fryer 1 , T.A.Carpenter 11 Wolfson Brain Imaging Centre, University of Cambridge, Box 65, Addenbrooke’s Hospital, Cambridge CB2 2QQ; 2 CavendishLaboratory, University of Cambridge, J.J. Thompson Avenue, Cambridge CB3 0HE; 3 Siemens Molecular Imaging Preclinical Solutions,810 Innovation Drive, Knoxville, TN 37932, USAIntroductionA number of different approaches for combining MRI with PET havebeen investigated (1, 2, 3, 4). Our approach (Figure 1) is based on anovel, 1T actively-shielded superconducting magnet with a 80mm gapto accommodate a multi-ring PET detector array based on a microPET®Focus 120 system (Siemens Molecular Imaging PreclinicalSolutions, Knoxville USA). The PET detectors and MR imager viewthe same region in space, which facilitates simultaneous MRI and PETacquisition. We present an overview of the status of the system.MethodsMR Performance A 2D fast spin echo (FSE) sequence (NEX 6TR/TE 2000/50.5ms) with a slice thickness of 0.8mm and an imagematrix of 256 x 192 voxels over a 20.5 x 15.4mm field of view producesin-plane resolution of 80 µm in 9 minutes.Effect of fibre optic bundle length To study the effect of increasingthe optical fibre length on the PET detector sensitivity, position map,and energy resolution the performance of two ‘short’ (10 cm opticalfibre) and two ‘long’ (120 cm optical fibre) detectors were comparedusing a 68 Ge point source imaged in singles mode.Performance of PMT in magnetic field To study the impact of magneticfield on PET sensitivity, position map and energy resolution, asingle short detector was operated in singles mode using a 68 Ge pointsource. The PMT end of the detector was located inside a softiron/permalloy magnetic shield resulting in a nominal 1mT field.ResultsMR performance Examples of mouse brain images acquired usingconventional MRI gradients and RF coils in conjunction with thenovel 1T magnet are shown in Figure 2.Effect of fibre optic bundle length The mean full width half maximumof the photopeak for the individual crystals increases from17.2±0.1% for the short detectors to 27.1±0.5% for the long detectors.The ratio of the energy resolution values indicates that the long detectormeasures 40% of the light measured with the short detector. Positionmaps are shown in Figure 3. The photopeak sensitivity for thelong detectors was 2% higher than for the short detectors.Performance of PMT in magnetic field Direct comparison of thesingles count rate with the PMT positioned in a magnetic field of 1mTwith the singles count rate at ‘low’ magnetic field (~0.05mT) showsthat the photopeak sensitivity (350-650keV) at the 1mT position is98% of that at low field. Position maps acquired with the PMT positionedin low magnetic field and in a nominal field of 1mT are shownin Figure 4. The crystal energy spectra acquired with the PMT at the1mT position shows a mean FWHM of 19.1±0.4% compared with thatof 18.7±0.4% at low field.ConclusionThe light loss from the use of long fibre optic bundles decreases theenergy resolution and hence the scatter fraction will increase. However,the spatial resolution and sensitivity are not degraded. Measurementsindicate that operating the PMT in 1mT will have minimaleffect on image quality.Figure 1: Novel magnet with superimposed schematic of PETsystem.Figure 2: Mouse brain images acquired using conventionalgradient, shim and radiofrequency (RF) hardware on novel 1Tmagnet.Figure 3: Position maps for PET detectors with 10cm (left) and120cm (right) fibre optic bundles.References1. Shao, Y., et al. Phys. Med. Biol. 42, 1965-1970 (1997).2. Grazioso, R., et al. Proc. Intl. Soc. Mag. Reson. Med. 13, 408(2005).3. Handler, W.B., et al. Proc. Intl. Soc. Mag. Reson. Med. 13, 868(2005).4. Pichler, B.J., et al. J. Nucl. Med. 47(4), 639-647 (2006).We acknowledge funding from EPSRC (Grant numberEP/C009096/1). This study was funded in part by the EC-FP6-projectDiMI, LSHB-CT-2005-512146. We would also like to acknowledgethe support of GSK and Bruker BiospinFigure 4: Position maps for a PET detector in low (~0.05mT)magnetic field (left) and a nominal 1mT field (right).


The relationship between brain structure and schizotypal measures in subjects at high risk of developingschizophreniaG .Katherine S. Lymer, Dominic E. Job, T. William J. Moorhead, Andrew M. McIntosh,David G. Owens, Eve C. Johnstone and Stephen M. LawrieDivision of Psychiatry, Royal Edinburgh Hospital, Morningside Park, Edinburgh EH10 5HFP17IntroductionEvaluation of the clinical and neuropsychological measures recordedat baseline in the Edinburgh High-Risk Study (EHRS) showed that theRust Inventory of Schizotypal Cognitions (RISC) (1) and StructuredInventory for Schizotypy (SIS) (2) were able to distinguish betweenthose high-risk (HR) subjects who went on to develop schizophreniafrom those who did not (3). The aim of this study is to identify thebaseline brain structure-function relationships associated with thesebehavioural differences.MethodsSubjects: A total of 132 HR subjects (n=89 without psychotic symptoms(HRwell); n=26 with psychotic or possibly psychotic symptoms(HRsymp); n=17 who were subsequently diagnosed with schizophrenia(HRill)) and 33 controls had complete neuropsychological evaluationand sMRI at baseline. sMRI: Volumetric data required for VBMwas acquired using MPRAGE (TR=10ms, TE=4ms, TI=200ms, relaxationdelay time=500ms, flip angle=12˚, FOV=250mm x 250mm,B 0=1.0T) resulting in 128 contiguous 1.88mm thick ‘slices’. Imageanalysis: VBM was performed using the SPM99 toolbox[http://www.fil.ion.ucl.ac.uk/spm/software/spm99/] and was based onthe grey-matter optimised protocol developed by Good et al (4, 5).Behavioural tests: Battery of clinical and neuropsychological testsincluding the RISC and SIS to obtain measures of schizotypal cognition.Statistical analysis: Multiple regressions were performed inSPM, in each of the HR population sub-groups with the individualscores of RISC and SIS. Where significant correlations were found,the corresponding maximum peak voxels were extracted into SPSS totest for subject group by neuropsychological score interactions.ResultsIn the right pulvinar nucleus, a negative correlation was found withSIS and GMD in the HRwell group (p = 0.023), Figure 1, althoughwhen testing for interactions, no significant difference between thegroups was found.Figure 2 (a): Positive correlation of RISC with GMD in the L STG(BA41) in HRill group. Z-scores are indicated by the colour barFigure 1 Negative correlation of SIS with GMD in the R pulvinar inthe HR well group. Z-scores are indicated by the colour barA significant positive correlation was found in the left superiortemporal gyrus (STG) (p = 0.008) of grey matter density (GMD) inthe HRill group with RISC score, Figure 2a. Testing this relationshipfor interactions between all groups showed the HRill to have thestrongest correlation (p = 0.017), Figure 2b.Figure 2 (b): Strong positive regression at same coordinates as (a)shown using SPSSConclusionChanges in cognition and psychotic symptoms have been previouslyassociated with reductions in the left STG in patients withschizophrenia (6) so it is appropriate that a correlation with RISC inthe same brain region was found in the HRill group. Reductions inpulvinar volume have been reported in a combined group of subjectswith schizophrenia and schizotypal personality disorder compared tohealthy controls (7) suggesting a trait effect. The correlation of SISwith pulvinar nucleus within the HRwell group supports thishypothesis. These results suggest that those HR subjects who becomeill have a different pattern of premorbid structural deficits andassociated behavioural abnormalities than those who do not.References(1) Rust J: Schizophr Bull 14:317-322 (1998)(2) Kendler KS et al, Schizophr Bull 15: 559-571 (1989)(3) Johnstone EC et al, B J Psychiatry 186: 18-25 (2005)(4) Good CD et al, NeuroImage 14:21-36 (2001)(5) Moorhead TWJ el al, NeuroImage 22:188-202 (2004)(6) Shenton ME et al, Schizophr Res 49: 1-52 (2001)(7) Byne W et al, Arch Gen Psychiatry 58: 133-140 (2001)GKSL, DEJ, AMM & SML are supported by the Dr Mortimer andTheresa Sackler Foundation; This work was funded by the UKMedical Research Council


MRI Based Attenuation Correction for Combined PET/MRP18Ian B Malone 2 , Richard E Ansorge 1,2 , Tim D Fryer 1 , Guy B Williams 1 , T Adrian Carpenter 11Wolfson Brain Imaging Centre, University of Cambridge, Box 65, Addenbrooke’s Hospital, Cambridge CB2 2QQ, UK;2 Cavendish Laboratory, University of Cambridge, J.J.Thomson Avenue, Cambridge CB3 0HE, UKIntroductionAttenuation correction is an essential step in the reconstruction ofquantitative PET images. An important aspect in the development of apractical combined PET and MRI scanner (PET/MR (1,2)) capable ofsimultaneously acquiring PET and MR will be to achieve attenuationcorrection when neither rotating positron-emitter source or CT dataare available. An accurate attenuation map is also required if modelledscatter correction is to be applied. When co-registered MR data isavailable the attenuation map can be estimated by assigningappropriate attenuation coefficients to tissues. Previous work by Zaidiet al. (3) demonstrates that this can be done for the head bysegmenting the images to four brain regions. We report here a methodbased upon a digital brain phantom (BrainWeb (4)) whichimplements a brain model based upon tissue classes, used incombination with known tissue attenuation coefficients.MethodsA synthetic T1 weighted 3D MR scan is simulated for the digital brainphantom and also an attenuation map in the same stereotactic space(Figure 1). The attenuation phantom was generated using:1.05x10 -4 cm -1 background, 9.7x10 -2 cm -1 cereberospinal fluid,9.2x10 -2 cm -1 fat, 1.0x10 -1 cm -1 other soft tissues and 1.72x10 -1 cm -1bone (derived from (5)). To obtain a subject-specific attenuation map,the synthetic T1 weighted image is registered to a 3D T1 weightedMR scan of the subject using non-linear warping using the VTK-CISGtoolkit (6) (Figure 2). The transformation parameters so derived areapplied to the synthetic attenuation map in order to produce thesubject attenuation map. Finally the phantom attenuation map wassmoothed using a 3D isotropic 6mm Gaussian kernel to match theresolution of the measured attenuation.Figure 2: Synthetic T1 weighted volume MR (top left) is nonlinearlyregistered to the subject's T1 weighted 3D MR scan (top right). Theresulting non-linear transform is applied to attenuation map (lowerleft). The measured attenuation map is shown lower right.ResultsWe have tested this approach for MR and PET data acquired on twoseparate scanners, and compared it with attenuation maps acquiredwith 68 Ge rotating rod sources; figure (3) shows that the maps are ingood agreement. There is little spread in phantom attenuation valuesas even after smoothing large regions of the phantom attenuation mapare homogeneous, the noise in the measured attenuation leads to thehorizontal spread of clusters in figure (3).ConclusionThis approach avoids segmenting the MR data directly into tissueclasses where a particular problem is distinguishing air spaces frombone. This method may be used with other common MR sequencessince both proton density and T2 weighted MR volumes can besimulated from the digital brain phantom. The digital brain phantomderived maps have much lower noise than the acquired maps—thiswill lead to less noise propagation into the final images.We intend to perform PET reconstruction using these estimatedattenuation maps to investigate their effect on image quantification,and to extend this technique to other regions of the body (requiring theuse of full body phantoms).Figure 3: 2D histogram of estimated attenuation versus estimatedattenuation for the volume covered by the brainweb phantom. Thescale bar is the number of voxels in each bin. Contrast has been scaledto show the bins containing low counts; bins in the clusters atestimated attenuation around 0cm -1 and 0.10cm -1 have counts from300 to 400.References(1) Grazioso R et al. 13 th Proc. ISMRM (2005)(2) Shaw N et al. 13 th Proc. ISMRM (2005)(3) Zaidi H et al. Med. Phys. 30 937-948 (2003)(4) Collins DL et al. IEEE TMedIm. 17 463-468 (1998)(5) Hubbell JH and Seltzer SM http://physics.nist.gov/xaamdi (2004)(6) Studholme C et al. Pattern Recogn. 32 71-86 (1999)Ian Malone’s student funding by EPSRCFigure 1: Simulated T1 weighted Volume MR (left) andcorresponding attenuation map (right) based on BrainWeb digitalbrain phantom.


T 1 mapping in childhood abdominal tumours – preliminary experienceP19Iosif Mendichovszky 1 , Øystein E. Olsen 2 ,1 RCS Unit of Biophysics, UCL Institute of Child Health, London, UK;2 Radiology Department, Great Ormond Street Hospital for Children NHS Trust, London, UKIntroductionMagnetic resonance (MR) imaging has the potential to provideinformation on changes in tumour size and volume in cancer patientsundergoing anti-tumoural treatment. However, MR signal intensityimages are poor in detecting tumoural response to treatment andtumour volume is an unreliable measure of therapy effectiveness.Furthermore, MR imaging of paediatric abdominal tumours ischallenging due to respiratory movements and the smaller size ofthese patients. A more quantitative approach in the diagnosis andfollow-up of these patients would greatly improve patientmanagement. In this work we describe a quantification method forabdominal tumours in free-breathing paediatric patients based oncomparing T1 maps acquired before and after the injection of aparamagnetic contrast agent.MethodsImaging protocol: Imaging was carried out on a 1.5T Siemens Avantoscanner. To generate the T1 maps we acquired several T1-weightedimages using a respiratory-gated inversion-recovery 2D turbo-FLASHpulse-sequence and six different inversion times (128, 250, 500, 1000,2000, 3000 ms). Respiratory gating was performed using a 2Dnavigator-echo, and 3 slices positioned through the tumour andadjacent tissues were acquired. The imaging protocol was run bothbefore and after the administration of a low dose of a MRI contrastagent (Magnevist) and post-contrast imaging was done after allowingthe agent to fully mix in the body. Pre and post contrast injection T1maps were calculated from the inversion-recovery T1-weightedimages. Several regions of interest were drawn on the T1 maps both innormal and tumoural tissues and mean T1 values were obtained.These were converted into relaxation rate values R = 1/T 1. For lowdoses of gadolinium we can assume that the relationship betweenrelaxation rate (1/T1) and contrast agent concentration is linear andexpressed by the formulaR 1= R 0+ [ Gd]rwhere R 1 is relaxation rate of the tissue after contrast injection, R 0 therelaxation rate of the same tissue before contrast, [Gd] the gadoliniumconcentration in that tissue and r the relaxivity constant. Knowing thepre and post contrast relaxation rates, we were able to generatequantitative maps based on the amount of gadolinium entering thetissue of interest (reflected by the [Gd]r factor).Patient population: Our subjects (6 patients in total) were paediatriconcology patients undergoing routine MRI clinical scans for diagnosisor follow-up purposes.ResultsFigure 1 shows results obtained from an MRI scan of a 3 year-oldpatient. The first two images are IR signal intensity images (TI =250ms) pre (A) and post (B) contrast injection of a slice through theliver. Normal liver tissue as well as multiple nodular tumours can beeasily seen. The second set of images are T1 maps pre (C) and post(D) contrast injection of the same slice calculated from the IR images.As expected, a significant drop in T1 values was observed both innormal and tumoural tissues following contrast injection. For thispatient T1 values for normal hepatic tissue dropped from 450 ms to353 ms after contrast injection (corresponding to an increase inrelaxation rates from 2.2s -1 to 2.83 s -1 ), while the reduction in T1 fortumoural tissue was greater – from 1065 ms to 543 ms (correspondingto an increase in relaxation rates from 0.94s -1 to 1.84s -1 ). Thedifference in the increase of R 1 values between normal tissues andtumour was present in all our patients. The ratio image (E) in figure 1is derived from the pre and post contrast T1 maps and reflects therelative ralaxation rate in a tissue of interest due to the presence ofgadolinium in that tissue ([Gd]r/R 0 ). As R 0 is readily available fromthe pre-injection T1 maps we can determine the [Gd]r product whichgives us direct information on the amount of gadolinium present in theregion of interest.Figure 1: Signal intensity images, T1 maps and ratio image pre andpost contrast injection in a 3 month old child. Important differences innormal and tumoural tissues can be seen pre and post contrastinjection on both the signal intensity images and the T1 maps.ConclusionWe have shown that T1 mapping in childhood abdominal tumours isfeasible in free breathing and can give us quantitative information onthe amount of gadolinium present in a region of interest. An importantstrategy in treating oncological patients is to reduce or block bloodbeing delivered to tumors, thus inducing necrosis and cellular death.As the amount of gadolinium present in a tumour is associated withthe perfusion level, a drop in tumoural perfusion will result in reducedlevels of gadolinium in the tumoural tissue. Thus, this method mightprovide us with a useful clinical tool for oncological treatmentmonitoring and tumoural response to radio/chemotherapy.References1. Higgins DM, Ridgway JP, Radjenovic A, Sivananthan UM,Smith MA. T1 measurement using a short acquisition period forquantitative cardiac applications. Med Phys. 2005 Jun;32(6):1738-46.2. Zhang JL, Koh TS. On the selection of optimal flip angles for T1mapping of breast tumors with dynamic contrast-enhanced magneticresonance imaging. IEEE Trans Biomed Eng. 2006 Jun;53(6):1209-14.3. Cernicanu A, Axel L. Theory-based signal calibration withsingle-point T1 measurements for first-pass quantitative perfusionMRI studies. Acad Radiol. 2006 Jun;13(6):686-93.4. Olsen OE, Sebire NJ. Apparent diffusion coefficient maps ofpediatric mass lesions with free-breathing diffusion-weightedmagnetic resonance: feasibility study. Acta Radiol. 2006Mar;47(2):198-204.5. Messroghli DR, Plein S, Higgins DM, Walters K, Jones TR,Ridgway JP, Sivananthan MU. Human myocardium: single-breathholdMR T1 mapping with high spatial resolution--reproducibilitystudy. Radiology. 2006 Mar;238(3):1004-12.


An increased matrix single echo acquisition MAMBA 2D arrayP20Martyn PALEY 1 , James WILD 1 , Kuan LEE 11 Academic Radiology, University of Sheffield;IntroductionRecently there has been interest in encoding planar images within asingle echo for ultra-rapid imaging. The single echo acquisition (SEA)technique (1) uses a 64 element RF strip array and receiver chainlinked to conventional 1D pulsed gradient frequency encoding togenerate high frame rate images of a thin plane. The MAMBA 2Dtechnique (2) alternatively uses only a single RF channel and volumeRF coil with micro B 0 coils arranged in a 2D array to produce uniquefrequency encoding of a thin plane without any pulsed gradients.Images are formed after a single Fourier Transform and subsequent2D image intensity allocation. The difficulty with the MAMBA 2Dmethod is that the number of turns required for the individual B 0 coilsrises as n 2 with the number of in-plane pixels n. Alternatively, thecoils can be single turns fed with different currents but this becomesvery complex to construct without e.g. a complex silicon chip design.Previously we designed a 5x5 matrix 2D coil using a geneticalgorithm (2) which resulted in a complicated coil layout. A new arrayconcept is introduced which only requires 2n B 0 loops for single echo2D planar encoding of n 2 pixels.MethodsTwo sets of n loops with 1-n turns each are placed in proximity andorthogonal to each other. One set of loops has current I while the otherset of loops has n x I. A MATLAB simulation was developed usingthe Biot-Savart law to model the 2D array. To demonstrate theprinciple, a 10 x 10 matrix MAMBA 2D surface loop array waswound with 15 mm spacing and 150 mm length with turns varyingfrom 1-10 along each axis. The array was located parallel to B 0 in thebody coil adjacent to a uniform cylindrical phantom. Data wasacquired using a spin echo sequence (TR=200 ms, TE=10 ms, matrix=256x256, NEX=1) with phase map reconstruction on a 3T MRsystem (Philips Intera). A current of 120 mA was applied to one axisand 1.2A to the other axis i.e. a factor of 10:1.ResultsFigure 3 plots the field (a.u.) for each of the rows in Fig 2 versusposition along a row in mm.Figure 4 shows the constructed 10x10 MAMBA 2D array.Figure 1 An ideal 10x10 MAMBA 2D field with monotonicallyincreasing B 0 .Figure 2 Biot -Savart simulation of a real 10x10 MAMBA 2D arrayThe surface shows the magnitude of the B 0 fields above each ‘pixel’ ina plane 5 mm above the loops and it can be seen that these vary in therequired way although there is non-linearity near the edges.Figure 5 shows a phase difference map from a uniform circularphantom placed adjacent to the 10x10 B 0 array.The phase shift gradually increases towards the top and left of theimage where the higher number of turns were located as predicted.Discussion and ConclusionThe dual axis stepped loop design generates a two dimensionalvarying B 0 field and provides a method to extend MAMBA 2D tohigher matrix resolution, (dependent on the frequency resolution ofdata acquisition.) while minimising the number of loops. The linearitycould be improved further using additional correction loops at theedges of the array although having unique fields is more importantthan linearity. Further work will acquire instant dynamic data setswith no switched gradients applied. Single echo acquisition in 2Dcould prove useful for imaging thin planar objects when acoustic noiseor eddy current effects cannot be tolerated and where very hightemporal resolution is required.References1. Wright et al., MRM, 2005, 54; 386-392.2. Lee et al., MRI, 2002, 20; 119-125.


Impact of inconsistent resolution on VBM studies: an example using Semantic DementiaP21João M. S. Pereira 1 , Peter J. Nestor 2 , Guy B. Williams 1 ,1 Wolfson Brain Imaging Centre, Dept. of Clinical Neurosciences, University of Cambridge; 2 Neurology Unit, Dept. of ClinicalNeurosciences, University of CambridgeIntroductionFor longitudinal studies, historical data is often available which doesnot match current acquisition protocols. This may be due to scannerupgrade or replacement, or changes to the scan parameters; and it maynot be possible to control for this over the project timescale. In thisstudy we consider the limited case of a change in scan resolution. Weshow that it may be possible to use a mixture of voxel sizes in ananalysis without significant loss of statistical power provided that theeffect is balanced across groups. As an examplar, we consider a studyof semantic dementia (SD), with voxel depth of 1.5 mm and repeat thestudy after degradation of the data with a 1.8 mm reslicing.MethodsFourteen SD subjects (10 male and 4 female, average age 61.9) wereincluded in this study along with fourteen age-matched subjects (7male and 7 female, average age 66.6). All subjects were scanned usinga 1.5T GE MRI scanner and the images were acquired using a T1-weighted 3D spoiled gradient echo sequence (echo time of 4.2 ms,inversion time of 650 ms, flip angle of 20º). All scans had voxel dimensionsof 0.84x0.84x1.5 mm 3 . Corresponding degraded scans with1.8 mm voxel depth were obtained from the two original samplegroups through the interpolation of the original scans, using a normalisedsinc interpolation (1). Subsequent modulated Voxel-Based Morphometry(VBM) studies were performed using SPM5 on the segmentedgrey matter images smoothed with a 8 mm gaussian kernel.Heterogeneous samples were studied that combined both 1.5 mm and1.8mm voxel depth scans. The proportion of degraded scans in eachgroup was varied systematically and the output contrasted in eachanalysis to the benchmark in which all scans had 1.5 mm depth, bothvisually (Figure 1) and through the computation of the root meansquare error difference between the t-maps (Figure 2). The statisticalthreshold was set at a Family Wise Error (FWE) correction ofP(corrected) = 0.05 and the extent threshold was set at k = 200.ResultsIn spite of some variability, all VBM studies showed the pattern ofcortical degeneration expected from SD subjects,Figure 1: (a) Glass brain image of the benchmark VBM study, (b) fulldegradation study with all subjects with 1.8 mm voxel depth, (c) unbalancedstudy with all SD subjects with 1.8 mm voxel depth andcontrols with 1.5 mm and vice-versa (d)Figure 2: T-maps root mean square error difference, computed for thewhole brain, when compared to benchmarkwith greatest atrophy on the anterior left temporal lobe (2). The resultswere dependent on the balance of the number of scans with the samevoxel depth in each group, as seen in Figure 2. In balanced cases, i.e.with the same number of scans with a given voxel depth in both thecontrol group and in the SD group, few changes were observed in thestatistical map compared to benchmark, which is apparent when contrastingFigure 1(a) with Figure 1(b). Despite less sensitivity, thelocation of the greater significance peaks was otherwise identical tothe benchmark. The same was true where all SD subjects had 1.8 mmdepth while maintaining the controls at 1.5 mm (Figure 1(c)) and viceversa(Figure 1(d)). Nonetheless, these latter analyses produced outputsthat were visually more distant from the reference study, with adistinct loss of sensitivity (Figure 1(d)) and an increase in false positives(Figure 1(c)).ConclusionThe results show that, provided the proportion of subjects with differentvoxel depth in each group is maintained, there are no great differencesin the VBM conclusions. However, in unbalanced analysis theremay be a loss of sensitivity. Nevertheless, the significance peaksremain very similar throughout these changes, suggesting that criticalgroup differences remain largely unaffected by such alterations. It istherefore possible to conduct viable VBM studies on mixed samplesby accepting the trade-off between sample size and minor accuracyissues. Naturally, it still is more sensible to only use 1.5 mm (or less)scans when available, but this study shows that concerns about thevalidity of VBM analysis where mixed samples are used can be minimisedif balanced groups are used. Furthermore, degrading all scansfrom 1.5 mm to 1.8 mm voxel depth did not have a significant impacton VBM results.References(1) N. A. Thacker, A. Jackson, D. Moriarty and B. Vokurka, RenormalisedSinc Interpolation, Tina Memo No. 1999-005 (2000)(2) G. B. Williams, P. J. Nestor and J. R. Hodges, Neural correlates ofsemantic and behavioural deficits in frontotemporal dementia, Neuro-Image 24, 1042-1051 (2005)


Ultra-Efficient Shielded Dome Gradient CoilsP22Michael POOLE, Richard BOWTELLSir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, University Park,Nottingham, England, NG7 2RD (correspondance to Michael Poole: ppxmp@nottingham.ac.uk)IntroductionInsert gradient coils built specifically for head imaging are useful formany purposes such as fMRI, diffusion imaging, and high-resolutionanatomical imaging. Such coils inherently produce strong gradientfields because they are smaller and have wires closer to the region ofuniformity (ROU). The geometry of the coils described here aim toexploit this property to the full by placing the conducting surface asclose to the head as possible. A boundary element method (BEM)after Lemdiasov and Ludwig [1] was used to design shielded X, Y,and Z gradient coils on this surface that produce magnetic fieldgradients that are as strong as possible whilst minimising theirinductance, resistance, and imposing torque balancing.MethodsThe boundary element method [1] allows gradient coils to be designedon conducting surfaces with arbitrary geometry. It works bydiscretising the stream-function of the current density on the surfaceinto a weighted set of divergence-free basis functions. The inductance,resistance, and torque of the coil as well as the magnetic flux densityat any point can be derived in terms of these basis functions, allowinga functional reflecting the coil characteristics to be optimised. Surfacemeshes were created in 3D Studio MAX® (Autodesk) and importedinto Matlab® (Mathworks) for calculation of the basis-functionweights. A 3D-contouring algorithm was written to generate the wirepathsfrom these weights. A Biot-Savart calculation was performed onthe wire-paths to obtain the magnetic flux density distributions, andmultipole expansion analysis was used to model the inductances andresistances using FastHenry© [2]. Fig. 1 shows the geometry of thecoils, the inner surface is designed to be separated by roughly 50 mmfrom the head to allow space for the RF coil. The spheroid in thecentre is the region of uniformity (ROU), and the triangles show halfof the discretised surface of the primary (red), secondary (blue)surfaces. The ROU contains 587 evenly distributed points, and thesurface is discretised into 2162 basis-functions. The region ofshielding (ROS) (not shown in Fig. 1), at which the magnetic fluxleakage was minimised, are a set of points distributed on the surfaceof a 550 mm long cylindrical surface with a diameter of 640 mm. Aprototype X-gradient coil is being constructed on a rapid-prototypeformer, and will be tested.Figure 2 a) primary and b) secondary surfaces wire-paths for the Xgradient.Figure 3 The field produced by the X gradient in the y=0 plane.Table 1. The properties of the X, Y, and Z gradient coils.Figure 1. The geometry of the coils.ResultsAll coils were designed to have a field homogeneity of less than 5% ofthe maximum field value in the ROU. Fig. 2 shows the wire-paths ofthe X gradient, where red wires indicate reversed current flow withrespect to blue. Fig. 3 is a contour map (green) of the field that itproduces with the 5% field homogeneity contour (black), the coilsurface (blue dashed), and ROU and ROS (orange). Table 1 shows theproperties of the X, Y, and Z gradient coils, resistances andinductances in brackets were modelled using 3mm diameter wire withFastHenry©.ConclusionA shielded gradient coil set has been designed to have very highefficiency using a boundary element method [1]. The gradient fieldsproduced by these coils are 2.0, 1.8, and 2.4 times stronger thanprevious X, Y, and Z gradients of equal inductance, have superiorhomogeneity and are also shielded [3]. The results have beencorroborated using Biot-Savart calculation and multipole expansionimpedance extraction calculation with FastHenry© [2].Refernces[1] R. A. Lemdiasov, R. Ludwig. Concepts Magn. Reson. B. 26B, 67-80 (2005). [2] M. Kamon, M. J. Tsuk, J. K. White. IEEE Trans. onMTT. 42, 1750-1758 (1994). [3] D. C. Alsop, T. J. Connick. Magn.Reson. Med. 35 875-886 (1996).AcknoledgementsWe wish to thank Magnex Scientific Ltd. (partof the Varian, Inc. group) for the support of a studentship


Application of Higher-Order Boundary Element Method to Gradient Coil DesignP23C. Cobos Sanchez 1 ,L. Marin 2, R.W. Bowtell 1 , H. Power 2 ,P.M. Glover 1 , A.A. Becker 2 & I.A. Jones 21 Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, The University of Nottingham, Nottingham Park,Nottingham NG7 2RD, UK2 School of Mechanical, Materials and Manufacturing Engineering, The University of Nottingham, Nottingham Park,Nottingham NG7 2RD, UKIntroductionThe boundary element method (BEM) is a numerical computationalmethod of solving linear partial differential equations which havebeen formulated as integral equations. BEM provides the basis of apowerful approach for the design of gradient coils of arbitrary geometry.Here, BEM has been used to design cylindrical and hemisphericalhead gradient coils, incorporating a rectangular window (from whichwires are excluded) so as to allow visual interaction with the subject.The efficacy of employing curved rather than flat element geometryhas also been explored.MethodsIn the quasistatic regime, the magnetic vector potential A satisfiesPoisson’s equation. Using the free-space Green’s function, integralrepresentations over the coil surface of A, and hence of the magneticinduction, B, can be formed [1]. When a desired variation of B over aset of points in a region of interest (ROI) is imposed, the integralequation for B becomes an inverse problem, whose solution yields adivergence-free current density that generates the desired fieldvariation.To solve this integral equation, a discretisation process is carried outby meshing the coil surface into triangular elements. Shape functionsare used to define the geometry of these elements, which may be flat(linear approximation) [2] or curved (quadratic or cubic approximation).,where the vertices of the triangles are the nodes.A divergence-free current distribution is defined in terms of a linearcombination of tangential vectors to every point in an element withunknown weights related to every node [3]. The spatial variation ofthe current density over an element depends on the geometry of theelement and the nodal values. The inverse problem is ill-posed andcannot be solved by direct methods [4]. Additional information aboutthe desired solution is therefore imposed via the requirements that theinductance is minimised and the torque on the coil balanced. Usingminimization or regularization techniques, the required continuouscurrent density solution can then be obtained. The coil wire arrangementis produced by contouring the stream function of the currentdensity (Figs.2b,2a)ResultsCylindrical, transverse and longitudinal head gradient coils of radius0.175m and height 0.350 m incorporating a window of height 0.1mand length 0.2m, with a spherical ROI of radius 0.065 cm (Fig.1) weredesigned using both linear and quadratic elements, where the regularizationparameter is adjusted to produce a coil of minimum inductancewith less than 5% relative field error in the ROI (Figs 2.a and 2.b)Hemispherical coils of radius 0.175m with and without cylindricalextension of height 0.175 m were designed as well in the same way(Figs.3 and 4).With a window of height 0.10 m and length 0.20 m forthe both cases.Biot-Savart calculations of the field homogeneity, torque and inductanceare in good agreement with the theoretical results (Table.1).Quadratic and linear solutions converge to the same value for a givenproblem (Fig.6).Computational times for quadratic approximation are bigger thanlinear, being in any case a few hours up to couple of thousands ofelementsConclusionBEM provides a solid mathematical framework for gradient coildesign, allowing the production of torque-balanced coils of minimumindictance with a wide variety of shapes and geometries. Quadraticand constant BEMs both produce effective solutions. The best choiceof element type will depend on the particular geometry of the coildesign.References(1) M.P. Morse and Feshbach Methods of Theoretical Physics (1953)(2) R. Lemdiasov and R. Ludwig, Concepts Magn. Reson. B, 26B,67-80 (2005).(3) L. Marin, H. Power, R.W. Bowtell, C. Cobos Sanchez, A.A.Becker, P. Glover and I.A. Jones, Divergent-free current densityhigher-order boundary element method for an inverse MRI problem,submitted to Proceedings of the Royal Society. Part A: Mathematical,Physical & Engineering Sciences(4) A.N. Tiknonov, and V.Y. Yersin Method for Solving Il-PosedProblems. Nauka (1986)


Detection of the Stria of Gennari using Turbo Spin Echo Imaging at 3TP24Rosa Sanchez, Susan Francis, Richard Bowtell,Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and AstronomyUniversity of Nottingham, University Park, Nottingham, NG7 2RDIntroductionThe striate cortex in the calcarine sulcus can be distinguished fromneighbouring regions by the presence of the stria of Gennari, a denseband of myelination within the cortical grey matter. MRI sequencesproviding contrast between grey and white matter potentially allowdetection of the stria, but this is only possible with very highresolution imaging, since the thickness of the myelin band is onlyabout 0.3 mm (1). In previous work detection of the stria hasgenerally been accomplished by using fast, three dimensional spoiledgradient echo sequences applied with a partial inversion prior toacquisition of each plane of k-space, so as to yield optimal contrastbetween grey and white matter (1-4). Here, we show that multi-slice,turbo spin echo (TSE) imaging (5), used in conjunction with inversionrecovery at 3 T provides an alternative route to the identification ofthe stria of Gennari in a reasonable measurement time.MethodsImaging was performed on a 3T Philips Achieva System using an 8-channel SENSE head coil. Multi-slice TSE images were acquired inthe occipital lobe with the slices aligned perpendicular to the plane ofthe cortex, using a SENSE-factor of 2 (RL). For these experiments,the echo train length (ETL) was 17, the refocusing angle was 140 o andlinear phase encoding order was used. It was found that the bestcontrast between grey and white matter was obtained by preceding theTSE acquisition with an inversion recovery, so as to produce mixed T 1and T 2 contrast. Optimal contrast was found for an inversion recoverytime of 220 ms and an echo time of 70 ms, with a repetition time of3500 ms and a TFE shot duration of 132 ms. Image data setscomprising 5 slices, each with a slice thickness of 1 mm and an inplaneresolution of 0.35 x 0.35 mm 2 were acquired on three subjects(FOV=180mm x 180mm; matrix size = 512 x 512). A total number of20 averages were performed, grouped into 5 individual data sets of 4averages. Each of these data sets was acquired in 161 s, leading to atotal scan time of 13.5 minutes. Motion correction was applied beforeaveraging together each of the individual data sets. 3D fast spoiledgradient echo image data with two different contrasts were alsoacquired so as to allow segmentation of brain tissue. One subject wasalso scanned using a refocusing angle of 120 o to obtain 10 slices of 1.5mm thickness. On this subject, 4 scans of 4 averages (5 minutesduration) each were acquired, giving a total acquisition time of 20minutes for the 16 averages.Results and DiscussionFigure 1 shows IR-TSE data obtained from the one subject scannedwith 1.5 mm slice thickness. The highest signal intensity occurs inCSF, with grey matter appearing with significantly higher intensitythan white matter. The image on the left is a magnification of theregion where the stria was visible; a dark band can clearly be seenwithin parts of the grey matter along the calcarine sulcus.AWMCSFBWMFigure 1: High resolution images (0.35x 0.35 x 1.5 mm 3 ) showing ahypointense band (stria of Gennari) within the grey matter andprofiles through sulci along the lines marked on the images.AWMCWMCBCSFFigure 2: High resolution images (0.35 x 0.35 x 1 mm 3 ) of the threesubjects scanned, showing the hypointense band within the greymatter (indicated by the arrows).Intensity profiles were measured across the sulci at different locationswhere the stria was most evident. Profiles A and C were measuredfrom the edge of white matter (WM) to the edge of the CSF (from leftto right) and B from the edge of the white matter across the CSF to theedge of the white matter. A dip in signal intensity, shown by thearrow, is present in all examples, indicating a band of myelination.Figure 2 shows results obtained from the three different subjectsscanned using a 1 mm slice thickness. Magnified views of the markedregions are shown on the right. A hypointense band, as indicated bythe arrows, can clearly be seen within the cortical grey matter in thefirst and second subject. In the third subject a short band ofmyelination along the flank of the arrowed sulcus can bedistinguished.The mean ratio of the contrast between grey and white matter to noisewas measured as 4.0 in these images.ConclusionWe have shown that IR-TSE can be used at 3T to identify themyelination patterns of the striate cortex in reasonable measurementtimes. A hypointense band of signal intensity in grey matter of theoccipital lobe was seen all three subjects scanned. In future work, weexpect to produce images at higher resolution by an adequateoptimization of the parameters of the IR-TSE sequence at 7T.References(1) Clark C et al Cereb. Cortex 2 417-424, 1992.(2) Walters N et al PNAS 100, 2981-2986, 2003(3) Bridge H et al. Journal of Vision. 5 , 93-102 (2005)(4) Clare S, Bridge H, Human Brain Mapping. 26 , 240-250 (2005)(5) Thomas DL et al. Mag.Res.Med. 51, 1254-1264 (2004)


P25Automated Techniques for Murine Phenotyping in Huntington’s DiseaseS. J. Sawiak 1 , G. B. Williams 1 , N. I. Wood 2,3 , A. J. Morton 3 , T. A. Carpenter 11 Wolfson Brain Imaging Centre, Box 65, Addenbrooke’s Hospital, Hills Road, Cambridge CB2 2QQ2 Cambridge Centre for Brain Repair, E. D. Adrian Building, Forvie Site, Robinson Way, Cambridge CB2 2PY3 Department of Pharmacology, Tennis Court Road, Cambridge CB2 1PDIntroductionVoxel-based morphometry (VBM) techniques are well-established inhuman subjects to find both structural and functional differencesbetween subject groups under various conditions [1]. With the mousegenome now complete and the ease of preparing transgenic models forhuman disease, the use of mice as models of human disease is on theincrease. Currently most of the resulting morphometric data ismanually processed (see, for example, [2]): a time-consuming anderror-prone process liable to the bias of particular reviewers. We hopeto develop techniques such as voxel-based morphometry to work withmurine studies with the eventual goal of providing protocols forlongitudinal in-vivo studies tracking the development of disease overtime.In collaboration with the Cambridge MRC Centre for Brain Repair 2and Department of Pharmacology 3 , we are focusing initial efforts onthe R6/2 mouse model of Huntington’s disease.MethodsSpecimen PreparationA total of 93 brains (51/42 male/female, 47/46 transgenic/wildtype)were placed into 13mm-diameter glass tubes and bathed in FluorInertFC-77 (3M chemicals), as a proton-free susceptibility matchingfluid[3]. Gauze was placed on top of the brain to reduce motion andkeep the brain below the FluorInert level.Image acquisitionImages have been acquired using a 1T Bruker spectrometer andreconstructed by ParaVision 3.0.2. A custom built solenoid coil wasused as the probe. An FSE sequence with ETL 4, NEX 4 was usedwith TR/TE 2000/50.5ms, a field of view 1.79×1.34×0.90cm 3 over amatrix size 256×192×128 to give a final isotropic resolution of 70µm.Final imaging time was a little under 14 hours.Data analysisImage registration techniques based on mutual information have beenused for both linear and non-linear registration. SPM5[4] (StatisticalParametric Mapping, v5) was used to prepare preliminary maps ofstatistical deviations between groups. Analyze 7 (Mayo Clinic) wasused for manual morphometric analysis for region of interest (ROI)studies.ResultsImages show good contrast between grey and white matter with mostmajor brain structures clearly distinguished (Figure 1). Signal to noise(measured in the hippocampus area) is approximately 30 in each of 93images.Clear changes are seen in the transgenic mice compared to thewildtype mice, most significantly in the left/right ventricles (figure 2).Details of the SPM process such as segmentation and registrationparameters remain to be resolved.Figure 1 Sagittal mouse brain slice. Contrast is largely proton-density weightedRegions of interest likely to be affected by the disease (based onprevious histological studies [5]) have been selected for manualanalysis. Significant changes have been found so far in the globuspallidum (p < 0.0006, Kolmogorov-Smirnov test), left/right ventricles(p < 0.0009) and caudate putamen (p < 0.03). No significant effectswere found in the hippocampus (p < 0.5) or internal capsule(p < 0.51).Figure 3 Relative ventricular size for each group in the manualregion of interest studyConclusionThe MR images shown show excellent detail and show cleardifferences between the brain groups of the study. When these havebeen completed to include the amygdala, a reasonable set of data withwhich to test and develop automated methods, such as VBM will beavailable. These early results clearly show the viability of MRtechniques in this important area.References[1] Ashburner et al. NeuroImage (2000) 11:805-821[2] Koshibu et al. NeuroImage (2004) 22:1636-1645[3] Ma et al. Neuroscience (2006) 4:135[4] Ashburner et al. NeuroImage (2005) 26:839-851[5] Stack et al. J. Comp Neurology (2005) 490:354-370S. J. Sawiak would like to thank the MRC for funding.Figure 2 Wildtype brain (left) transgenic brain (right),enlargement of the ventricles is clearly seen


The influence of task demand on the BOLD response in a complex power-grip taskP27Annette Sterr 1 , Shan Shen 1 , Gang Gao 1 , Wengshen Hou 2 , Andre Szameitat 11 Department of Psychology, University of Surrey, Guildford, UK; 2 Department of Biomedical Sciences, University of Chongquing,Chongquing, ChinaIntroductionThe present experiment used a visually guided force-feedback task 1 tostudy the relationship of task demand and brain activation measuredwith fMRI. Participants were asked to keep the strength of the powergrip within a constantly moving target range for a period of 30seconds (figure 1). Thereby the level of task difficulty wasmanipulated in two ways, the number and speed of force level changes(trail of target bar; fast vs slow trail (T F , T S )), and the level of accuracyprovided by the feedback (sensitivity of target range; high vs low (A H ,A L )). Behavioural pilot data suggested an interaction of theseexperimental factors with highest error levels in fast/high trails, andlowest error levels in slow/low trails. With the present experiment weaimed to specify the neural activations involved in this task, and towhat extend the increase in task demand is reflected in enhancedresponses.Methods11 right-handed volunteers (aged 20-35) were scanned in a 3TSiemens Trio scanner. The visual information was projected through astandard rear-projection system. Power grip strength was recorded inreal time and visually displayed as feedback (in-house designedsystem, Figure 1). The experiment consisted of two sections, eachcomprising five randomised conditions (T F A H ; T S A H ; T F A L ; T S A L ; andbaseline (fixation cross), 12 minutes recording time). For thefunctional scans, imaging parameters were TE = 30ms, TR=2000ms,FOV=192×192mm 2 , matrix size = 64×64, voxel size = 3×3×4mm.Structural images were obtained from T1-weighted 3D MPrage scans(TE=5.57ms, TR=1900ms, FA=11º, FOV=256×256mm 2 , matrixsize=256×256, voxel size = 1×1×1mm). Brainvoyager QX was usedfor data preprocessing and statistical image analysis. Raw imageswere motion corrected, spatially (FWHM = 6mm) and temporally (3seconds high-pass) filtered, and subsequently transformed to Talairachspace. Statistical maps obtained from random-effects GLM analysiswere superimposed on a 3D inflated cortex model.ResultsFour condition-to-baseline contrasts were calculated to assess theactivation pattern evoked by the feedback task. This analysis revealedconsistent activations in contralateral motor regions, somatosensoryassociation areas and secondary visual areas. The condition contrastsfurther revealed greatest activation differences in for the contrastbetween the conditions with poorest performance. Theactivation/behaviour relationship was further reflected inhighlysignificant correlations between the signal strength and the errorindexFigure 2: Functional activations during for different task demandscompare to resting baselineConclusionFour condition-to-baseline contrasts were calculated to assess theactivation pattern evoked by the feedback task. This analysis revealedconsistent activations in contralateral motor regions, somatosensoryassociation areas and secondary visual areas. The condition con-trastsfurther revealed greatest activation differences in for the contrastbetween the conditions with poorest performance. Theactivation/behaviour relationship was further reflected in highlysignificant correlations be-tween the signal strength and the errorindex.1000Figure 1: Illustration of task set-up


A comparison of modified k-t BLAST and a k-t variable-density under-sampling technique forretrospectively gated MR flow measurementYuehui Tao 1 , Ian Marshall 11 Medical Physics and SHEFC Brian Imaging Research Centre for Scotland, University of Edinburgh, Edinburgh, UKP28IntroductionMR flow measurement has many important applications such asmeasuring blood flow velocity to estimate wall shear stress in arteries.However 3D flow encoding takes a long time. Many techniques suchas k-t BLAST (1) have been developed to accelerate such scan byacquiring only part of the data that should normally be collected.Retrospective gating, which is often applied with MR flowmeasurement, makes it difficult to implement those reducedacquisition techniques because the sampling pattern does not conformto a Cartesian grid. Normally the solution is to use gridding tointerpolate acquired data into an equivalent Cartesian data set (2). Asthe acquired data set is under-sampled, errors will be introduced bythe gridding operation. Here a modification is made to k-t BLAST soit can directly accept data from retrospectively gated under-samplingwithout the need for gridding. The modification will also greatlyincrease reconstruction speed of k-t BLAST for both retrospective andprospective gating. As an alternative to k-t BLAST, a k-t variabledensityunder-sampling technique (ktVDUST) for retrospective gatingis proposed. Data acquisitions using sampling strategies of bothmethods have been simulated and reconstruction results from the twomethods have been compared for same acceleration factors.MethodsModifications to k-t BLAST: Basically gridding is an inverse problemand reconstruction from under-sampled data is another inverseproblem. So it is possible to combine these two into one inverseproblem. In the original k-t BLAST a linear system equation set is setup in x-f space (the inverse Fourier transform of k-t space) to describethe relationships between full data and sampled data. It requires aFourier transform of the under-sampled data into x-f space. Thismakes gridding necessary. However an equivalent system equation setcan be set up between full Cartesian data in x-f space and undersamplednon-Cartesian data in k-t space, thus eliminates the need forgridding. The parameters of the new equation set will only takeaccounts of two steps. The first is Fourier transform of full data in x-fspace into k-t space. The second is an interpolation according to thesampling pattern in k-t space. The interpolation step will not introduceerrors because it is form full Cartesian data. In this way k-t BLASTcan directly process non-Cartesian data sampled in k-t space. The newequation set has another advantage. It has full rank, that is, the numberof equations in this set equals the number of samples in k-t space. Theoriginal k-t BLAST system equation set is always ill-conditioned andits size equals the full resolution of k-t space, in case of 4-foldacceleration that is 4 times the size of the new equation set. So thenew equation set can be solved more efficiently. In ourimplementation an 8-fold acceleration of reconstruction has been seenfor a 4-fold under-sampling. This good property will also work forCartesian sampling pattern in k-t space.A k-t variable-density under-sampling technique for retrospectivelygated under-sampling: In retrospective gating the number of samplesof each phase-encoding position can be customized although thecontrol of timing of each sampling is not available. So k-t space canbe sampled with variable sampling density by changing the number ofsamples for each k-space position. The sampling pattern can bedesigned according to the desired acceleration factor. Central regionsof k-t space is sampled more frequently than the outer regions, asshown in Figure 1(d). Data can be recovered for each k-space positionseparately. For each k, full data is assumed to have limited temporalfrequencybandwidth which equals the number of samples in this k-space position. Most real MRI data conforms to this assumption. Asdata is recovered for each k separately, the computation time is veryshort and can be ignored. This technique is quite simple because allneed to do is to decide how many samples should be acquired for eachphase encoding position. It collects data efficiently. All sampled datais directly used to provide information and there is no need for anextra training scan.ResultsAs for all the techniques each k-t slice can be treated separately, hereonly a fully sampled k-t slice from real data is used as a scan subjectto be sampled with different sampling strategies mentioned above.Figure 1(a) shows how this slice is extracted from real data as an y-tslice. The magnitudes of this y-t slice is also shown in Figure 1(b).Simulations are carried out for both methods with the sameacceleration factor of 8/3 and 4 (in k-t BLAST central 1/8 k-t space isfully sampled as training data). Reconstruction errors are measured bythe temporal average of the magnitudes of the differences betweenrecovered image and true image.Figure 1: Simulations of retrospectively gated under-sampling andreconstruction using modified k-t BLAST and ktVDUST(b) shows the image is quite sharp. In (c) and (d) only half of allthe sampling patters are displayed. Black spots show the locations ofsamples in k-t space. (e) shows for an acceleration factor of 4 resultsfrom the two methods are close. Errors from ktVDUST is slightlysmaller than k-t BLAST. In the high intensity region, errors aresmaller than 3% for both methods. (f) shows for an acceleration factorof 8/3 ktVDUST gives better results than k-t BLAST. Both methodsshow better results for a smaller acceleration factor.Conclusionk-t BLAST is modified to directly accept non-Cartesian data from aretrospectively gated under-sampling and need for gridding iseliminated. k-t BLAST reconstruction is also greatly accelerated. A k-t variable-density under-sampling technique for retrospective gatinghas been proposed and according to simulation results, ktVDUST cangive better performance than k-t BLAST. Both techniques giveacceptable results for an acceleration factor of 4 in the simulations.References(1) Tsao J et al. Magn Reson Med. 50, 1031-1042 (2003)(2) Hansen MS et al. Magn Reson Med. 52, 1175-1183 (2004)


Slice profile effects in variable flip angle HP 3 He MRIK. Teh, K.J. Lee, J.M. WildAcademic Radiology, University Of Sheffield, Sheffield, UKP29Introduction Unlike thermally polarised NMR,hyperpolarised (HP) M o is non-recoverable and is progressivelydiminished after application of each RF pulse. Hence HP gasMRI is very sensitive to flip angle deviations and differentmethods have been implemented (1, 2) to maximise theutilisation of the non-constant M 0 . To minimise blurring throughk-space filtering (2) a constant transverse magnetisation can becreated using the variable flip angle scheme (1): θ(n)=tan -1 (1/√(N-n)) Eq.[1] where N is the total number of excitations, nis the nth excitation pulse and θ is the flip angle. In this work theperformance of the variable flip angle scheme was investigated insingle slice and multi-slice 2-D spoiled gradient echo imagingexperiments with HP 3 He gas. The slice profile in HP gas MRIhas previously been shown to be variable from RF view – RFview as a result of the non-ideal flip angle distribution causingdifferential rates of magnetisation depletion across the slice. Thiscan result in a “rabbit ear” profile shape as described previously(2). A strategy to compensate for non-uniformity inmagnetisation response caused by this non-ideal view dependentslice profile by using a variable slice select gradient betweenviews, is presented here.Methods Measurements were conducted on a 1.5Twhole body MRI system (Eclipse-Philips Medical System).Studies were performed using a mixture of (100ml 3 He and900ml N 2 ) contained in a 1 litre Tedlar plastic bag. The 3 He gas(Spectra Gases) was polarised on site to around 20% with opticalpumping rubidium spin exchange apparatus (GE Health). Aquadrature high quality factor (Q) T/R birdcage was used for allexperiments with homogeneous flip angle distribution, importantin experiments sensitive to flip angle changes. An interleavedsequence was used with N=4 pulses. Eq.[1] was used todetermine the theoretical amplitude of RF pulses needed per viewto maintain constant transverse magnetisation without any sliceselection scaling. Scaling for non-uniform slice profile effects,introduced by the 2D slice selection process, was thenimplemented by determining the scale factor for RF pulseamplitude. To do this the transverse magnetisation, M xy (z) wassummed over a) the whole FOV, this represents the single 2-Dexperiment and b) by summing M xy (z) over the FWHM limitswhich represents the multi slice experiment - see Fig.[1]. Thelimits of ± FWHM/2 represent the scenario in multi-slicing dueto magnetisation burnout on either sides of the excited slicethrough multi slicing. By taking limits of FWHM as shown inFig.[1] it was assumed that above and below the limits ofFWHM the sum of the magnetisation is zero. Diffusion effectsbetween slices are negligible in these low N experiments due tothe short TR. The sequence was run with 256 samples, 13 mmslice thickness and a TR of 80ms.-FOV/2-FWHM/2FWHM/2FWHMFig.[1]- limits of integration of M xy (z) for scenarios a) and b).Results Fig.[2a] shows the plot of normalised magnetisationversus pulse number, n, as calculated by summing M xy (z) overFOV limits for the theoretical flip angles (30°,35°,45°,90°) asgiven by Eq.[1]. It is obvious that a non-steady magnetisationresponse is found despite using the variable flip angle scheme –ZFOV/2we attribute this to view dependent slice profile effects (2).Fig.[2b] has the same magnetisation response plotted over theFWHM limits of integration. The plot enables determination ofthe slice selection scaling to achieve constant M o . From this datathe slice thickness was scaled per view as illustrated in Fig.[3a]where the slice profile changes with n. Notice changes in themeasured slice profile width due to slice selection scaling.Fig.[3b] is the corresponding curve depicting magnetisationchanges as a function of n with an integration limit of the wholeFOV. Despite the variable flip angle and slice select scaling anon-steady magnetisation response is still seen. Fig.[4a] showsplot of slice profile, M xy (z), over the limits of the FWHM andFig.[4b] shows corresponding normalised magnetisationresponse from integrating between these limits –note the almostflat response in this case.Normalised MoMoMo1.41.351.31.251.21.151.11.05FOV / 2∑−FOV/ 2M xy( z)11 2 3 4n pulsesFig.[2a]2.5 x 105 21.5-6.5mm10.50-FOV/2ZFig.[3a]14 x 104121086420-∆Z/2ZFig[4a]6.5mmFOV/2∆Z/2Normalised MoNormalised MoNormalised Mo1.21.181.161.141.121.11.081.061.041.020.980.960.940.920.90.8811 2 3 4n pulses1Fig.[2b]0.861 2 3 4n pulses1.041.0210.980.960.940.920.90.88FWHM / 2FWHM / 2Fig.[3b]0.861 2 3 4n pulsesFig[4b]Conclusion In this work it was demonstrated that variableflip angle techniques previously used in 2D HP gas MRI areprone to error and can provide a non-ideal non-constant magnetisationresponse caused by the non-uniform distribution of flipangles across the slice (2). By implementing a variable slice selectgradient in conjunction with variable flip angle techniquesenables us to maintain constant M o throughout the multi-pulseexperiment. The results presented here are directly applicable tothe high acceleration factor parallel imaging techniques that canbe achieved with HP gas using low numbers of RF excitations(3). Future experiments will include implementation in 2D imagingsequences with more pulses (higher N), and also the possibilityof tailoring the slice profile of each pulse in the sequence.References[1] Zhao L et al ,J Magn Reson Ser B 1996;113:179-183[2] Wild JM et al, Magn Reson Med 2002;47:687-695[3] Lee R et al Magn Reson Med 2006;55:1132-1141.Acknowledgement EPSRC #GR/S81834/01(P), GE Health,Spectra Gases, Aerosol Society UK∑−FWHM/ 2FOV / 2∑−FOV/ 2∑−FWHM/ 2M xy( z)M xy( z)M xy( z)


Quantitative Magnetisation Transfer of Low Grade GliomasP30Daniel J. Tozer, Christopher E. Benton, Paul S. Tofts and Jeremy R. Rees,Institute of Neurology, University College London, Queen Square, London, WC1N 3BGIntroductionGliomas are primary brain tumours that can exist for many years in alow grade state (WHO grade II), before they transform into themalignant grade III and IV tumours which cause rapid deteriorationand death. There are many types of glioma which respond differentlyto treatment (1) and so non-invasive determination of histological typecould provide important information for treatment planning withoutthe morbidity and mortality of biopsy.Quantitative analysis of magnetisation transfer imaging (qMT) is arecent development of traditional MT techniques, which has allowed afuller analysis of the MT phenomenon, including the estimation ofmore fundamental parameters such as the fraction of protons existingin macromolecules such as myelin (2, 3). This parameter has beenshown to be altered in the multiple sclerosis lesions and white matter(3) and the white matter of Alzheimer’s sufferers (4). A single gliomahas also been imaged (5) using a variation of the above techniques andit was shown that glioma has a lower bound proton fraction than whitematter and a reduced cross relaxation rate between the free and boundpools of protons.The purpose of this work is to assess the efficacy of qMT in low gradegliomas and ultimately to assess any differences between the tumourtypes.MethodsFour patients (3 female) were scanned using a 1.5 T clinical machine(GE, Milwaukee, WI, USA). The patients all gave informed writtenconsent and the study was passed by the local ethics committee. Atthis stage the numbers are too small to attempt a histologicalcomparison of the data.qMT images were acquired using the 3D sequence described in (6),briefly this is a fast SPGR acquisition with TR/TE = 30.7 ms/5.3 ms.Excitation flip angle = 5°, in order to minimise T1-weighting. MTsaturation was achieved using Gaussian pulses (standard deviation =2.98 ms, duration = 14.6 ms), applied once every TR prior to RF excitation.Ten uniquely MT weighted points were collected using twoMT pulse powers (the powers can be described by their equivalent onresonanceflip angles: 212° and 843°). For each power, five differentfrequency offsets, ranging from 400 Hz to 20000 Hz, were used. Thefrequencies were separated by a constant logarithmic step of ~ 0.4. Amatrix of 256x96x32 over a field of view of 240x180x160 mm 3 wasused, for a total scan time of approximately 15 min. Data were thenreconstructed to an in-plane matrix of 256x256 over a 24x24 cm 2 fieldof view, while the 32 through plane phase encode steps were used toreconstruct 28 5 mm slices (with 4 slices being discarded to minimisewrap around effects). In addition 3D SPGR volumes (TR = 13.1 ms,TE = 4.2 ms) were acquired with three different excitation flip angles(α = 25°, 15°, 5°) to estimate the T 1 of the various tissues. Lastly FastSpin Echo FLAIR images were acquired to define the tumour boundaries.Imaging parameters were: TR/TI/TE = 8774/2192/161 ms; pixel0.94x0.94 mm 2 ; slice thickness 5 mm; gap 1.5 mm. These imagesallow the following parameters to be obtained from the data: gM 0 A -thearbitrary scanner gain (g) multiplied by the magnetisation of the freepool (M 0 A ), RM 0 A -the cross relaxation rate between the pools (R)multiplied by M 0 A , 1/R A T 2A , the ratio of the relaxation times of thefree pool, T 2B -the transverse relaxation time of the bound pool, via theassumption of a super-Lorentzian lineshape for this pool, f/R A (1-f)where f is the bound proton fraction defined as M 0 B /(M 0 A +M 0 B ). Fromthese parameters and the estimate of T 1 , f, R A and T 2A can be derived.The fitting was performed using the Levenberg–Marquardt fittingalgorithm and a look up table for the super-Lorentzian lineshape.Images were then transferred to a workstation (Sun Microsystems,Mountain View, California, USA) for further processing. Each MTweighted dataset was then registered to the first as were the three T 1weighted datasets. The FLAIR dataset was interpolated and also registeredto the first qMT dataset.Regions were then drawn on the interpolated, registered FLAIR imagesas these provide the best delineation of the tumour. Regions weredrawn around the tumour as well as contra-lateral normal appearingwhite matter and in the thalamus. In addition pixel-by-pixel maps ofthe whole brain were obtained.ResultsFigure 1a shows a FLAIR image with the tumour clearly visible as thearea of high signal intensity, 1b show the corresponding f map,indicating the low bound proton fraction seen in the tumour.Figure 1a1bTable 1 shows the fitted parameter values (±standard deviation) forthe following regions, normal appearing white matter (NAWM),thalamus and whole tumour. Table 2 shows the three derivedparameters from the same regions.TissuegM 0ARM 0AT 2B1/R A T 2A f/R A (1-f) (s)(µs)NAWM 606±75 81±19 8.6±0.3 9.8±1.2 0.053±0.003Thalamus 671±95 60±12 8.6±0.4 11.6±0.7 0.055±0.003Tumour 710±96 90±46 7.0±0.4 7.7±1.7 0.035±0.006Table 1Tissue R A (s -1 ) T 2A (ms) f (p.u.)NAWM 1.2±0.1 87±10 5.9±0.6Thalamus 0.88±0.13 99±12 4.6±0.5Tumour 0.46±0.12 306±106 1.6±0.6Table 2As there are only 4 patients at this stage no statistical analysis wasattempted, however it is clear that there is a reduction in f in low gradegliomas and that the T 2 of the free pool of protons is significantlyincreased in the tumour. There is also a possible reduction in the T 2 ofthe bound proton pool; however this will require more subjects toconfirm.ConclusionAlthough this wok is in the early stages it is apparent that qMT mayhave some utility in imaging low grade gliomas. The reduction in thesize of the bound proton pool indicates that macromolecules are beingdestroyed or displaced by the tumour, or that there is an increase infree water. The possible reduction in T 2B would indicate a change inenvironment for the bound protons or preferential destruction of somespecies. Changes in inflammation or a progressive destruction oftissue could mean that qMT parameters may be a marker of malignanttransformation.References(1) Kitange GJ et al. Expert Rev. Anticancer. Ther. 1, 595–605 (2001)(2) Sled JG, Pike GB, Magn. Reson. Med. 46, 923-931 (2001)(3) Davies GR et al. Mult. Scler. 10 607-613 (2004)(4) Ridha BH et al. 9 th International Conference on Alzheimer'sDisease and Related Disorders. Number IC-P-107 (2004)(5) Yanakh VL. Magn. Reson. Med. 47, 929-939 (2002)(6) Cercignani M et al. Neuroimage. 27, 436-441 (2005)This work was funded by the Samantha Dickson Research Trust(Astro fund). DJT is supported by the MS society of GB and NI


EPISTAR Pulsed Arterial Spin Labelling Perfusion Imaging at 7 TeslaP31Roman WESOŁOWSKI 1 , Alex Gardener 1 , Penny Gowland 1 , and Susan Francis 1 .1 Sir Peter Mansfield Magnetic Resonance Centre, University of Nottingham, Notthingham, NG7 2RD.IntroductionAt ultra-high field arterial spin labelling (ASL) benefits greatly fromthe overall increased signal to noise ratio and extended T 1 relaxationtime of blood. At lower field strengths ASL is often hampered by lowSNR and by the fact that the T 1 of blood is not much longer than thebiological transit time. At high field the long T 1 of blood makes longerdelay times (TI’s) accessible for imaging. Here the EPISTAR (SignalTargeting with Alternating Radiofrequency) pulsed ASL sequence isimplemented at 7 T.MethodsThe EPISTAR method of PASL uses adiabatic 180 0 hyperbolic secantpulses followed by an EPI readout. The inversion pulses alternatebetween being applied below the image slices to label in-flowingblood (tag/label) and being applied above the image slices to controlfor magnetisation transfer effects (control/non-label). This is shownby Figure 1. In this study a 100 mm inversion slab, which was centred75 mm from the central image slice of the multi-slice set, was used.The inversion pulse was preceded by two saturation pulses applied tothe image slices in order to suppress stationary tissue and overcomeerrors due to the imperfections in the inversion slab profile. The profilesof the pulses used were assessed on phantoms prior to performinghuman studies.Data was acquired on a 7 Tesla Philips Achieva scanner. Threehealthy human adult volunteers were studied, who gave their informedconsent.Five GE-EPI images with an in-plane resolution of 2.5 x 2.5 mm 2 and3 mm slice thickness, with a 1 mm slice gap, were acquired using amatrix size of 64 x 64. An echo time of 35 ms was used. For eachsubject five delay times (TI’s) were used, of 800, 1200, 1400, 1700and 2000 ms. Vascular crushing using a 5 ms bipolar gradient pulsebetween slice-selection and the EPI read-out with a b-factor of 1.5smm -2 and critical velocity of 27 mms -1 was used to attenuate thesignal from blood flowing in the larger vessels.Sixty EPISTAR pairs (60 tag and 60 control dynamics) were acquiredat each TI with a repetition time (TR) of 2.5 s. The first five dynamicswere discarded during processing to allow the magnetisation toachieve a steady state.In addition, for each scan session a M 0 resting multi-slice data set andan inversion recovery T 1 -weighted data set comprising eleven TI’swere also acquired. Image registration of the data sets was then performedusing the AIR algorithm in MEDx. Perfusion weighted imageswere then generated by subtracting the tag image from the controlimage, and subsequent perfusion analysis was performed using inhouseprograms written in Matlab and C.ResultsA simulation of the change in grey matter signal for increasing TI dueto perfusion was performed, using values of tissue T 1 = 1.5 s, bloodT 1 = 1.95 s and a perfusion rate, f = 90ml/100g/min. An ROI was thenpositioned over grey matter for the average perfusion weighted images(figure 2). Signal changes in this ROI are shown in red, withsimulated data shown in blue.ConclusionsPerfusion weighted imaging using EPISTAR PASL has beensuccessfully implemented at 7 T. Future work will investigateperfusion quantification and functional perfusion experiments.References(1) Khoo VS et al. Radiother. Oncology. 42, 1-5 (1997)(2) Lemieux L, Barker G Med. Phys. 25 (6), 1049-1054 (1998)(3) Tanner SF et al. Phys. Med. Biol. 45(8), 2117-2132 (2000)(4) Chang H, Fitzpatrick JM, IEEE TMedIm. 11, 319-329 (1992)This project funded by the Cancer Research Campaign(SP1780/0131).ControlMulti-slice imagesTagFLOWFigure 1: STAR label scheme. Alternate tag and control pulses areapplied.TI = 800 msTI = 1200 msTI = 1400 msTI = 1700 msTI = 2000 msFigure 2: Presents five average perfusion weighted images froma single subject.Percentage Change (%)1.81.61.41.210.80.60.40.20Sim GMROI GM0 500 1000 1500 2000 2500TI (ms)Figure 3: Simulated and measured grey matter % signal changevs. delay time (TI).


Preliminary Data for Quantitative Two-Dimensional HRMAS Spectroscopy of Human Brain TumoursAlan WRIGHT 1 , Kirstie Opstad 1 , John Griffiths 1 , B. Anthony Bell 2 and Franklyn Howe 11 CRUK Biomedical Magnetic Resonance Research Group , Division of Basic Medical Sciences and 2 Department of Neurosurgery, StGeorge’s (University of London), SW17 0RE.P32IntroductionHigh resolution magic angle spinning (HRMAS) NMR spectroscopyis being used for metabolomic analysis of tumour biopsies and hasadvantages of minimal sample preparation and subsequenthistopathological analysis. Despite the improvement in 1D spectralresolution achieved by performing MAS on tissue samples (1), 2DTOCSY has even greater peak separation compared to 1D spectra(Figure 1) thus improving peak assignment and measurement.However, quantitation using a 2D cross-peak volume must accountfor: J coupling; relaxation during the mixing time; T1 saturation dueto the short TR used in 2D acquisitions. A method of quantitative 2DTOCSY HRMAS is proposed and demonstrated for phosphocholine(PCh) and glycero-phosphocholine (GPC), two metabolites of interestin phospholipid metabolism.levels, including creatine of metastases and meningiomas, over a 3hour period. Changes in choline compounds are summarised in Table2. These show that PCh and GPC concentrations decrease during the2D experiment time while free Cho levels increase.Tumour Quantification PCh GPC RatioGPC/PChMeningioma 1D LCmodel 0.43±0.4 0.59±0.5 1.382D TOCSY 0.56±0.3 0.29±0.3 0.52Metastases 1D LCmodel 1.76±1.3 0.37±0.1 0.212D TOCSY 1.59±1.2 0.53±0.3 0.34Table 1: Comparison of estimated concentrations of PCh and GPC,and the GPC/PCh ratio, as quantified from 1D and 2D spectra4.0 3.5 3.0 2.5 2.0 1.5 1.0 0.5 0.0Ω−ppmFigure 2: An overlay of two 1D presaturation spectra from differentmetastases. Peak heights are normalised to the highest peak in eachspectrum. The differences in relative peak heights between the twospectra is typical of the variation seen between differrent metastasisbiopsies.Figure 1: Regions of 1D presaturation and 2D TOCSY spectrashowing the improved separation of PCh and GPC peaks with theextra dimensionMethodsTumour samples were taken from histopathologically-verified humantumour biopsies (4 meningiomas and 3 metastases) that had beenremoved and snap-frozen in liquid N2 during routine surgicalexcision. HRMAS spectra were acquired at 4°C using a Bruker600MHz system. For each sample a 1D presaturation spectra wasacquired at the start and end of the NMR protocol (3hrs total time)which consisted of: a 2D TOCSY experiment, CPMG spectra at shortand long TR to estimate T1 saturation; a pulse-acquire experiment tomeasure tissue water signal for a concentration reference. 2D TOCSYof standard metabolites solutions were acquired and used for crosspeakcalibration referenced to Cr. Metabolite-concentration estimateswere made from the 1D data using LCmodel. The change inmetabolite concentration over time was quantified from the 1Dpresaturation spectra acquired at the beginning and end of the NMRexperiment set.ResultsEstimated metabolite concentrations, for meningiomas and metastases,as measured from the 1D presaturation and 2D TOCSY spectrum, aregiven in Table 1. These data show reasonable agreement between thetwo methods though these average results show high standarddeviations about the mean. These large standard deviations betweentissue samples reflect the ‘heterogenous’ nature of the NMR spectrawithin each tumour classification (see Figure 2). The 1D presaturationspectra indicated changes of


P33T 1 Measurements for Cortical Grey Matter, White Matter and Sub-Cortical Grey Matter at 7TP. J. Wright 1 , A. Peters 1 , M. Brookes 1 , R. Coxon 1 , P. Morris 1 , S. Francis 1 , R. Bowtell 1 , P. Gowland 11 Sir Peter Mansfield Magnetic Resonance Centre, School of Physics and Astronomy, University of Nottingham, NottinghamIntroductionKnowledge of the relaxation times in tissue is important foroptimization of image contrast and for choosing optimal sequenceparameters for quantitative measurements, such as those used inmeasuring perfusion. The aim of this study was to measure thelongitudinal relaxation time (T 1 ) at 7T, at high spatial resolution, inthe cortical grey matter, white matter and sub-cortical grey matternuclei, specifically the putamen and caudate nucleus.MethodImage acquisition: 12 subjects age 26 ± 8.1 (mean ± s.d.) years werescanned on a Philips Achieva 7T and 3T MR scanner using aninversion recovery, turbo spin echo sequence with TSE factor = 10,TR = 5000ms and TE = 10 ms. Images were acquired for inversiontimes of 120, 200, 400, 600, 800, 1000, 1500, 1700 and 2000ms innon-sequential order. Acquisition matrix 256 x 256 and voxel size0.78x0.78x3mm 3 was used. 6 contiguous transaxial slices with 1mmslice separation were acquired. Finally a high resolution 0.3x0.3x1.5mm 3 anatomical image was acquired from 6 slices with this sequenceusing a TI of 120 ms, a TR of 3.5 s in 5 minutes 22 seconds (14averages). Pre-processing: Masks were produced in Analyze usingthe volume acquired at TI = 2000ms and motion correction wasperformed using medx®. Fitting: It was expected that a full inversionwould not be achieved in all regions of the head, requiring that thedegree of inversion was fitted for. T 1 was fitted to an expressiontaking account of all RF pulses in steady state [1,2] Measurement ofT 1 : For each subject, T 1 was measured from the T 1 maps in 3 regionsof cortical grey and 4 regions in cortical white matter. T 1 was alsomeasured from ROI’s that were drawn around the left and rightputamen, caudate nucleus and red nucleus (N=6) of each subject.ResultsFigure 3. Figure 4.Figure 1 shows the T 1 measurements for cortical grey matter andwhite matter for each subject. The mean ± standard deviation oversubjects for white matter was 1005 ± 67ms and for grey matter was1640 ± 53ms. Figure 2 shows T 1 measurements for the sub-corticalgrey matter regions investigated, averaged over all subjects. Figure 3shows the high resolution anatomical data set and Figure 4 shows acorresponding T 1 map, apparently showing some partial volume errorsat the interface between grey and white matter. It should be noted thatthe images displayed a drop-off in signal intensity in the temporallobes, apparently due to RF inhomogeneities.Discussion and ConclusionThe cortical grey and white matter T 1 is longer than at 3T as expected,and shows considerable dispersion between brain tissues. Theseresults are consistent with those previously reported on a cadaver at8T [2]. Sub-cortical regions show a lower T 1 value compared to thecortical grey matter, probably related to the higher iron content ofthese areas, which is also known to reduce their T 2 relaxation times[4]. This preliminary data will be used to optimise sequence timings toallow measurements to be made with shorter repetition times, andhence at higher spatial resolution with less sensitivity to partialvolume effects.References[1] Bakker et al, Phys. Med. Biol. 29, 12, 1511 – 1525, 1984.[2] Press et al, Numerical Recipes in C, Cambridge University Press.[3] Mitchell et al, Proc. Intl. Soc. Mag. Reson. Med. 11, 1089, 2003.[4] Haacke et al, Magn. Recon. Imag. 23, 1, 2005


Neighbourhood tractography: a new approach to seed point placement for group fibre trackingJonathan D. CLAYDEN 1 , Mark E. Bastin 2 , Amos J. Storkey 31 Neuroinformatics Doctoral Training Centre, School of Informatics;2 Medical and Radiological Sciences (Medical Physics);3 Institute for Adaptive and Neural Computation, School of Informatics; University of Edinburgh, Edinburgh, UKP34IntroductionOne area in which diffusion MRI (dMRI) based fibre trackingtechniques have strong potential is in the segmentation of individualwhite matter structures (tracts) from dMRI images. The segmentedareas can be used as regions of interest (ROIs) for studying tractspecific effects of pathology [1]. This kind of tractography basedsegmentation is advantageous over more established ROI methodsbecause the regions are calculated algorithmically, removing observersubjectivity; and because the regions can be arbitrarily shaped in threedimensions, matching the anatomy of the underlying fasciculus.However, fibre tracking algorithms typically require as input aseed point, a location in dMRI space from which the algorithm beginsto reconstruct a tract. The resultant segmentation can be very stronglydependent on the exact location of this point. This sensitivity can beproblematic when trying to consistently segment a specific tract fromseveral brain volumes. In this work we demonstrate the inconsistencyof segmentations derived from registration based seed pointplacement, and describe an alternative approach that is based onmaximising output similarity to a reference tract within a seedingneighbourhood in native image space.MethodsData acquisition: Six normal volunteers (2 male, 4 female; mean age27 ± 3.4 years) were recruited for this study. Each subject underwent adMRI protocol on a GE Signa LX 1.5 T clinical scanner (GE MedicalSystems, Milwaukee, WI, USA). The protocol was a single-shot spinechoecho-planar imaging sequence with 51 noncollinear diffusionweighting gradient directions, at a b-value of 1000 s mm -2 , and 3 T2weighted scans. TR was 17 s per volume and TE was 94.3 ms.Registration based seed placement: We placed a seed point in thesplenium of the corpus callosum in a standard (MNI) brain volume,and transferred it to each subject’s native dMRI space using theFLIRT registration tool (FMRIB, Oxford, UK), with the MNI whitematter map as a weighting mask [2]. The BEDPOST/ProbTrack fibretracking algorithm [3] was then seeded at these locations in each brainvolume. This algorithm models one fibre direction per voxel.Neighbourhood tractography: One of the tracts produced by theregistration based seeding method described above was chosen as thereference tract for the purposes of this study. For every other brainvolume, the BEDPOST/ProbTrack algorithm was then seeded at everyvoxel location within a 7x7x7 voxel neighbourhood centred at theseed point chosen by registration. The similarity of each of these“candidate” tracts to the reference tract was then calculated using atract similarity measure that we have developed and previouslydescribed [4], and only the tract with the greatest similarity wasretained. The associated seed point is thus the best match to thereference tract seed point, based on the evidence of output similarity.ResultsFig. 1 shows axial projections of the tracts produced by applyingregistration based seed placement to one brain volume from eachsubject. The tract output is probabilistic, with red indicating areas oflow likelihood of connection to the seed point, and yellow indicatinghigh likelihood. Fig. 2 shows the equivalent projections for the tractschosen as most similar to the reference tract (a), within the seedingneighbourhood. In both figures, all tract data have been thresholded atthe 1% level to ignore areas of very low connection likelihood; andthe underlying greyscale images show the slice of the subject’sanisotropy map in plane with the seed point.DiscussionThere is a general improvement in consistency, in terms of tract shape,in the neighbourhood tractography results shown here, relative to theresults from a purely registration based seeding process. Fig. 2(b) and2(c) are clearly better matches to 2(a) than their equivalents in Fig. 1,while the remaining tracts at least do not represent worse matches. Itshould be noted that in any given seeding neighbourhood, there is noguarantee that a good match to the reference tract will be available. Inaddition, the efficacy of neighbourhood tractography will depend onthe capabilities of both the fibre tracking algorithm and the similaritymeasure being used.Figure 1: Tract output produced by using registration to place seeds.Figure 2: Tract output chosen by the neighbourhood tractographymethod, using (a) as the reference tract.ConclusionNeighbourhood tractography is a novel approach to an importantproblem in group fibre tracking—that of consistent tract segmentation.It neither manipulates the dMRI data nor constrains the fibre trackingalgorithm, instead simply providing a preference for seed points thatproduce tracts with high shape and length similarity [4] to a referencetract. Whilst for the purposes of this study the reference tract wastaken from our dataset, in general it would be preferable to work froma standard set of reference tracts, and we would advocate the creationof such a set. We believe that this technique could have considerablepromise, particularly as similarity measures are refined.References[1] Kanaan et al. (2006). Psych Res: Neuroimaging 146(1):73–82.[2] Clayden et al. (2005). Proc ESMRMB 22, number 508.[3] Behrens et al. (2003). Magn Reson Med 50(5):1077–1088.[4] Clayden et al. (2006). Proc ISMRM 14, number 2742.


P35VALIDATION OFDNP-MR (DYNAMIC NUCLEAR POLARISATION-MAGNETIC RESONANCE)MEASUREMENT OF SUBSTRATE METABOLISM AND UPTAKE IN HEARTLowri Cochlin, Marie Schroeder, Damian Tyler, George Radda, Kieran ClarkeCardiac Metabolism Research Group, Department of Physiology, Anatomy and Genetics, Sherrington Building, University of Oxford, UK.Introduction“Molecular Imaging”, the coupling of imaging technologies withspecific molecular probes may revolutionize detection, diagnosisand treatment of disease.Whilst MRI is widely used in healthcare for measuring structureand function, the application of MR for metabolic imaging is limitedby intrinsically low sensitivity originating from the low magneticenergy of nuclear spins compared with their thermal energyat room temperature. In most MRI, the weak nuclear polarizationis compensated by the high concentration of hydrogen nuclei inbiological samples. Unfortunately, magnetic nuclei such as 13 Chave low natural abundance yet, such nuclei can provide considerablemetabolic information. To this end, a method for obtaininghighly polarized nuclear spins in solution has been designed byArdenkjær-Larsen et al 1 , opening up the possibility of imaging 13 Clabelled metabolites in-vivo and more importantly their conversionto other species. Ardenkjær-Larsen et al 1 built upon the theory ofDynamic Nuclear Polarisation (DNP) 2 ; a method of enhancingnuclear polarization in the solid state, to perform DNP-MRS andDNP-MRI.Here we have used DNP coupled with Ardenkjær-Larsen’s dissolutionmethod 1 to produce hyperpolarized for study in-vitro and invivo.Surface coil localised spectroscopy has been used monitorthe uptake and conversion of 13 C labelled pyruvate in the heart andchest of rats.MethodsIn this study 13 C 1 -pyruvic acid (isotope enriched to 99%) waspolarized at 1.2 K using microwaves at 93.88 GHz and 150 mW.Polarization build-up was monitored over 60 minutes until steadystatepolarization was reached.Hyperpolarized 13 C 1 -pyruvic acid was dissolved with sufficientaqueous solution of NaOH and TRIS buffer to yield a 79 mMsolution of 13 C 1 -pyruvate with pH 7.6 at approximately 38°C. Thedissolved solution was injected into an empty glass vessel insidethe magnet for in-vitro experiments, or via a rat tail vein cannula tostudy uptake in-vivo. Spectroscopy was performed in a 7T horizontalbore magnet interfaced to a Varian Inova console (VarianInc., Palo Alto, CA) and a butterfly RF coil.For in-vitro experiments, spectra were acquired with one 12° pulseevery 2 seconds for 2 minutes from the beginning of injection. Forin-vivo experiments, spectra were acquired with 5° pulses every 3seconds for 1 minute from the heart and chest region of four femaleWistar rats. All investigations conformed to Home OfficeGuidance on the Operation of the Animals (Scientific Procedures)Act, 1986 (HMSO) and to institutional guidelines.ResultsHyperpolarized 13 C 1 -pyruvate was injected into a glass vial toobserve the hyperpolarized signal, followed by a multi-averagedacquisition to determine the signal available when the sample hasreturned to thermal equilibrium. Figure 1 shows the 22100 timessignal enhancement achieved after dissolution compared to thermalequilibrium polarization. This corresponds to 0.0006% thermalpolarization, and 13.3% hyperpolarized polarization.Dissolution, transport and injection were achieved within 25 seconds,including 5-10 second injection duration. Accumulation ofhyperpolarized tracer and subsequent decay of signal was observedby real-time acquisition of spectra.Hyperpolarized 13 C 1-pyruvateThermal equilibrium 13 C 1-pyruvateFigure 1: Top: thermal equilibrium spectrum of 13 C 1 -pyruvateSNR 1.1 in 3 hours (SNR 0.025 corrected for 2048 averages)Bottom: hyperpolarized SNR 480 in 0.5 seconds (no averages).Figure 2 shows the arrival (during injection) and decay of thepyruvate signal injected into a glass vial. Figure3 shows the bolusof polarized pyruvate arriving at the heart and its subsequent metabolisminto lactate, alanine and bicarbonate.Peak heightexperimentalcalculated0 20 40 60 80 100 120 Time / sFigure 2: 12° pulse, 0.5 second acquisition (2048 complex points)repeated every 2 seconds for 2 minutes. Injection duration was 10seconds.Pyruvate-hydrateLactateTime (every 3s)AlaninePyruvateBicarbonate10 5 0 -5 -10Frequency (ppm)Figure 3: Heart and chest spectra, 5° pulse, 0.5second acquisition(2048 complex points) repeated every 3 seconds for 1 minute.Injection of 1 cm 3 over 8 seconds into the tail vein of a femaleWistar rat.ConclusionWe have demonstrated high levels of polarisation in-vitro and invivowhich has enabled us to visualise the uptake of polarizedpyruvate in-vivo and its metabolism to lactate, alanine and bicarbonatein real-time. Future work will focus on spectroscopic imaging,kinetic modelling of the metabolic processes demonstrated inFigure 3, and the eventual study of real-time metabolism in normaland diseased heart.References1. Ardenkjaer-Larsen, J.H., et al. Proc Natl Acad Sci U S A,2003. 100(18): p. 10158-63.2. Jeffries. Phys Rev. 106, 164-165 (1957).Acknowledgements:This study was supported by the British Heart Foundation andGeneral Electric.


Improved study of myocardial infarction using contrast enhanced MRIGang Gao 1, 2 , Paul Cockshott 1 , Bjoern Groenning 1 , Annette Sterr 21. University of Glasgow, UK 2. University of surrey, UKP36IntroductionAt present myocardial viability is assessed by visualevaluation of wall motion abnormalities fromcinematographic (CINE) magnetic resonance (MR)images in combination with presence or absence of latehyper enhancement (LE) on contrast enhanced MR(ceMR) images. Currently, ceMRI and cine MRI dataare collected separately, and subsequently comparedeach other to derive a viability score 1 . This approachmay face co-registration problem due to image misalignbetween the two sets of breath-hold images. We set outto develop and validate image processing techniquescapable of reducing the observer dependence andimproving accuracy in the diagnosis of viablemyocardium.MethodsData acquisition20 patients with chest pain who were both troponin Iand LE positive were recruited. MR was performed at amedian (range) of 69 (16-120) hrs on a Siemens Sonata1.5T system using a phased array chest coil. LVdimensions were evaluated by CINE trueFISP breathholdsequence. ceMR was performed 10 minutes afterinjection of 0.2 mmol/kg gadolinium-DTPA using abreath hold segmented turboFLASH inversion-recoverysequence. For both cine MRI and ceMRI, images wereacquired from 9 cross-sectional slices (slice thickness =8mm) to cover the entire heart.CINE-LE MR imagesReference images were selected manually at a timepoint t during a cardiac cycle from all 9 slices. Firstly, amutual information based 3D affine image registrationalgorithm 2 was used to correct the motion between thereference images and the ceMR images. Secondly, thecorrected reference images were registered with theother CINE images by a 3D non-rigid image registrationalgorithm 2,3,4 , producing a set of 3D array, whichspecify for each pixel in the reference images, thelocation of its corresponding point in the other CINEimages. Using the arrays, it is possible to animate theLE images in a process known as "warping" creatingwarped CINE-LE images.ResultsFrom the 20 patients, data were manually extractedfrom both original and warped images using in-houseoff line software. Mean (SD) LV end diastolic volume(EDV) was measured by planimetry (original/ warped)= (original / warped) = 206 / 215 ml. The t-test showsthere is no significant difference between the LV EDVof the original image and the warped image (p = 0.35 >0.05). Left ventricle end systolic volume (ESV) = 129 /123 ml, p = 0.33. Examples are presented in figure 1and 2.Figure 1. The end diastolic and the end systolic frames of a cinecontrast enhanced sequence. The arrows suggested the potentialhibernating myocardium.Figure 2. An other example of CINE-LE MR images.ConclusionA novel method for cardiac viability assessment usingcombined CINE and ceMR images has been developed.We have shown that there are no significant differencesin left ventricle end diastole volume and end systolevolume between the original CINE and ceMR images.The method therefore appears to be promising as animproved viability assessment tool.References1. Kim RJ, et al, “Relationship of MRI delayed contrastenhancement to irreversible injury, infarct age, andcontractile function”, Circulation, 1999;100:1992-20022. LG Brow, “A Survey of Image RegistrationTechniques”, ACM Computing Surveys, pp.325-376,January, 19923. Thevenaz, U.E., et al., “A pyramid approach tosubpixel registration based on intensity,” IEEE Trans.Image Processing, vol.7, pp. 27-41, 1998.4. H. Mahrholdt, et al., “Relationship of contractilefunction to transmural extent of infarction in patientswith chronic coronary artery disease”, J Am CollCardiol. 2003 Aug 6;42(3):505-12.


_êáíáëÜ=`Ü~éíÉê=çÑ=íÜÉ=fpjojNOíÜ=^ååì~ä=jÉÉíáåÖråáîÉêëáíó=çÑ=pìêêÉóI=dìáäÇÑçêÇ==OPÓOR=^ìÖìëíI=OMMSOR=^ìÖìëíI=OMMSpmlkplo=kbtpibqqboïïïKéÜKëìêêÉóK~ÅKìâLÄÅ|áëãêãMS


dçäÇ=péçåëçê


Spectroscopy BasicsWhat is MRSpectroscopy ?A Method fordetecting andquantifying biologicalcompounds otherthan water.How does it differfrom normal MRI ?AmplitudeLess electrondensity aroundnucleusFrequencyMore electrondensity aroundnucleusSpectroscopy uses a standard MRI scanner but instead of frequencyencoding a spatial image it encodes chemical information.How does the scanner frequency encode chemical information ?Hydrogen (or other) Nuclei in different molecules resonate at differentfrequenciesWhy ?The resonant frequency is proportional to the magnetic fieldDoes the magnetic vary for different nuclei in different molecules?Yes it is dependent on the local climate.What is an MR Spectrum?It’s a graph of signal amplitude versus frequency.Frequency is expressed Hz or “Parts per Million” ppm .The Frequencyshift due to the differences in the chemical environment is called“Chemical Shift”.Generally the spectrum consists of water, fat and metabolites. Theconcentration ratio of water: Metabolites is greater than 10,000:1.Therefore water suppression is applied globally.1) Frequencies arespecific to thelocal chemicalenvironment2) Amplitudes areproportional tothe concentrationConcentrationdeterminesheightCompositiondeterminesposition (ppm)Which application areas are important ?Head, Body: Prostate, muscle, bone marrow, breast & liver, lungs & many otherdeveloping areas of interest. The real strength is that it allows clinicians to performdiagnostic biochemistry in vivo and define (new) features in the pathophysiology ofdisease. The key to achieving this is that we acquire accurate, reproducible spectra inshort acquisition times, compatible across all field strengthsand platforms and integrated into the routine clinicalarchitecture of the systems.


Clinical Neurospectroscopy:In this field which is one of the most well known applications of MR Spectroscopywe tend to use Single or Multiple Voxel Studies of the brain using Probe ( ProtonBrain Examination ) and ProbeQ (which adds semiquantative data).Single Voxel Technique:• STEAM ( Stimulated echo acquisition mode)o A localization method which uses 3 water suppression pulses ( 90 – TE/2-90 –TM-90-TE/2-ECHO) – 3 RF pulses select box.o Results in Probe Spectrum of box.o Choice of traditional spectroscopists - using STEAM libraries.• PRESS ( Point Resolved Spectroscopy)o A double spin echo technique (90-TE1/2-180- TE1/2-TE2/2- 180 – TE2/2 –ECHO). 3 RF pulses select a voxelo Results in Probe Spectrum of voxelo PRESS has 2x SNR of STEAM!o Metabolite peaks look differentVariation with Age Down’s & Dementia Metabolic Disease


Prostate Spectrum1H Prostate MRS - PROstate imaging & SpectroscopyExamCommonality with PROBE-P CSI•PRESS localization•3D CSI Acquisition•FuncTool VisualizationUnique Features of PROSE•Water and Lipid Suppression (Dual bandspatial refocusing pulses)•Over-Prescription (reduction of chemical shifterror)•Enhanced Out-of-Volume Suppression•Endo-rectal coilPolyaminesCreatineCholineCitratespectralPPM 3.02.52.0Prostate Image with Ratio Map andSpectral Grid OverlayOverlay of MRS data on MR ImageCholine / Citrate RatioMetabolite ConcentrationCancer(high choline)• choline, creatine, and citrate resonancesat ~3.2, ~3.0, and ~2.6 ppm, respectively.• lipid resonance around ~2.0 ppm.•red signifies high choline tocitrate ratio, which indicatescancerNormal(high citrate)


Breast SpectrumBREASE - Breast Spectroscopy•Recent literature appears to show that spectroscopy has a useful role in breast imaging with thepresence of choline (3.25ppm) being a strong indicator of malignancy. A potential role has also beenindicated in the assessment of early chemotherapy response.BREASE was developed as a problem solving tool to be used in cases of equivocal contrast uptake orindeterminate morphology. A typical voxel size of 2cm x 2cm x 2cm at 1.5T takes approx. 5 minutes.BREASE is a TE-averaged PRESS sequence designed to cancel/reduce the artifacts associated with lipidsidebands potentially overlyingthe region of cholineAdipose signal is quite large withTE-averagingBaseline is ‘flat’ withno TE-averagingAt varying TE’s, the lipid sidebands are modulated differently Averagingdifferent TE’s results in a cancellation/reduction of sidebands.1.5T Axial VIBRANT on HD Breast arrayBREASE single voxel examCholine peak visibleAxial MIPFor information contact:GE HealthcareCoolidge House,352 Buckingham AvenueSlough, Berks, SL1 4ER, UKT +44 1753 874000www.gehealthcare.comVoxel size: 12mm x 13mm x 17mmScantime: approx. 5 minutes


páäîÉê=péçåëçê


VARIAN, INC.Varian MRI SystemsSCALABLE PARALLEL ARCHITECTURE SENSE Reconstructed ImageParallel imagingUsing SENSE reconstruction, four images (top row) arecombined to yield an image of similar quality as an imageobtained with conventional imaging techniques but witha two-fold reduction in scan time. 4T human system.Data courtesy of Hoby Hetherington, Albert Einstein College of Medicine.Transmit/Receive ImageTwo-fold scan time reductionDistortion reduction in EPI using SENSEEPI images obtained using SENSEreconstruction (third row, 2 shots, R=2) areless distorted than conventional EPI images(second row, 2 shots). Consequently, thefMRI activation (auditory stimulation) isshifted out toward the edges of the brain asseen when overlaid on anatomical images(top row, conventional, and bottom row,SENSE), resulting in a spatially more accurateactivation map. 4T human system.Data courtesy of R. Allen Waggoner, RIKEN – BrainScience Institute, Saitama, Japan.Simultaneous detection of multiple samplesOne slice each from four 3D mouse imagessimultaneously acquired. 7T/400 horizontalanimal system.Data courtesy of Mark Henkelman Mouse Imaging Centre(MICe), Hospital for Sick Children, Toronto, Canada


VARIAN, INC.Magnet Technology Center• Manufacturing of MRI magnets from 4.7T to 16.4T• Premier manufacturer of 7T clinical magnets• New European imaging applications laboratoryThe Varian Inc. Magnet Technology Centre is based 5 milesfrom the centre of the city of Oxford in England


páäîÉê=péçåëçê


IDEAIntegrated Development Environment for ApplicationsEasy pulse programming and image calculation development on Siemens systems.- can be implemented on the main console or on a standalone PC- developing, coding, debugging and testing of sequences offline- Maximising the efficient use of time on the system itself.Siemens run two courses:- one-week sequence programming course (Monday – Friday)- 3.5-day image calculation course (Monday – Thursday)Scheduling allows you to attend both courses back-to-back if required.The dates for the future courses for 2006 are:Sequence Programming CoursesImage Calculation CoursesJuly 31 – August 4, 2006 August 7– 10, 2006September 11 – 15, 2006 September 18 – 21, 2006October 30 – November 3, 2006 November 6 – 9, 2006A thorough knowledge of MR physics and sequence design is assumed and knowledge of C++programming language is beneficial. A nondisclosure form regarding proprietary materials will be signed byall participants on the first day of the course.The IDEA environment enables- programming of arbitrary loop structures- arbitrary k space trajectories- new protocol parameters- configuration of the user interfaceFor further details regarding the IDEA environment and courses,please contact jo.reid@siemens.com.


páäîÉê=péçåëçê


PulseTeqSolutions for MR Clinical ResearchPulseTeq LimitedNew MillsWotton-under-EdgeGloucestershire GL12 8JREnglandPhone:+44 (0)1453 855 381Fax:+44 (0)1453 855 901Silver Sponsor of the 12th Annual Meeting of theBritish Chapter of the ISMRMFounded in 2002, in Guildford in the UK, and selling directly to end-users or toOEMs, PulseTeq Limited develops hardware and accessories for magnetic resonanceimaging and spectroscopy, with a particular focus on RF coils, creatingnew technology and new solutions for researchers and clinicians, often workingclosely with academic groups in collaborative projects. It also offers custom RFcoil design and contract development of systems, system electronics, electromechanicalsubsystems and RF coils.In October 2005 PulseTeq entered into a partnership with Renishaw plc, a UKbased,global leader in industrial metrology, and in February 2006 established asecond UK operating site within Renishaw’s facilities in Gloucestershire.New Mills, PulseTeq’s new registeredoffice in Gloucestershire following itspartnership with Renishaw plc.Contact UsPulseTeq Office LocationsRegistered Office:New MillsWotton-under-EdgeGloucestershire GL12 8JREnglandGloucestershire Office:Bath RoadWoodchesterGloucestershire GL5 5EYEnglandPhone: +44(0)1453 855 381Fax: +44(0)1453 855 901Email: sales@pulseteq.comSurrey Office:Studio 6Kiln HouseFarm LaneBadshot LeaFarnhamSurrey GU9 9HYEnglandPhone: +44(0)1252 338 368


PulseTeqProduct ListSolutions for MR Clinical ResearchPulseTeq LimitedNew MillsWotton-under-EdgeGloucestershire GL12 8JREnglandPhone:+44 (0)1453 855 381Fax:+44 (0)1453 855 901To contact us call:+44(0)1453 855 381Multi-Element Neuro Coil for Imaging and Spectroscopy• Designed for maximum signal-to-noise in the brain• Close fitting with an open structure• Excellent access for fMRI accessoriesMR Simulator• Highly cost-effective solution• Familiarizes patients or fMRI subjects with the MR environment without using valuable scanner time• Reduces the amount of scanner time needed for the development of new fMRI paradigmsCoils for Hyperpolarized Studies• Transmit/receive coil for lung imaging• High B1 homogeneity over a large field of view: 30cm axial: 30cm radial; 25cm vertical• Split design for easy patient positioning1H Decoupled Surface Coils• Compact package designed for multiple applications• 7cm 13C coil and quadrature 1H decoupling coilRigid and Flexible Surface Coils• Rigid 12.5cm diameter coil for various nuclei• Flexible 20cm diameter coil for various nuclei• Field strengths of 1.5T and 3TMultichannel PIN Diode Driver• Capable of driving PIN diode switched transmit/receive coils or T/R switches• Four receiver outputs with independent active decoupling during receive• Output voltages of 15 Volts or 28 VoltsQuadrature Diplexer• Narrowband T/R switch with quadrature combiner and low-noise preamplifier• Operating frequency 128MHz with a bandwidth of 1MHz• Peak power up to 16kW(*) For Investigational Use Only


_êçåòÉ=péçåëçê

More magazines by this user
Similar magazines