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

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.