13.07.2015 Views

Official Proceedings - AIUM

Official Proceedings - AIUM

Official Proceedings - AIUM

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

American Institute of Ultrasound in Medicine <strong>Proceedings</strong> J Ultrasound Med 32(suppl):S1–S134, 20131543325 Novel Quantitative Conventional-Frequency Detection ofCell Death In Vivo With Neoadjuvant Chemotherapy forLocally Advanced Breast CancerGregory Czarnota, 1 * Ali Sadeghi-Naini, 1 Omar Falou, 1 SaraIradji, 1 William Tran, 1 Michael Kolios 2 1 Radiation Oncology/Physical Sciences, Sunnybrook Health Sciences Center,Toronto, Ontario, Canada; 2 Physics, Ryerson University,Toronto, Ontario, CanadaObjectives—We have previously demonstrated that highfrequencyultrasound and spectral analysis can detect cell death. Here weinvestigated whether quantitative conventional-frequency (7-MHz) ultrasoundincorporating spectral analysis and textural parameters may be usedfor the same purpose in vivo in human patients receiving chemotherapy.Methods—A clinical study was undertaken investigating theefficacy of ultrasound to quantify cell death in tumor responses with cancertreatment. Patients (n = 60) with locally advanced breast cancer receivedanthracyline- and taxane-based chemotherapy treatments. Datacollection consisted of acquiring tumor images and radiofrequency dataprior to treatment onset and at 4 times during treatment (weeks 0, 1, 4, 8,and preoperatively). Digital low-frequency ultrasound data were collectedand sampled with a 15-bit dynamic range using an Ultrasonix-RP devicewith a 7-MHz central frequency (3–10 MHz, –6-dB range). Whole-mounthistology was obtained for all samples.Results—Data indicated that spectral ultrasound changes weresignificant at 4 weeks after the start of treatment. Increases of approximately9 dBr (±1.67) in ultrasound backscatter were observed in patientswho responded to treatment. Patients assessed as responding poorlydemonstrated significantly lower increases (2.3 ± 1.7 dBr). Increases in 0-MHz intercept followed similar trends, while increases in spectral slopewere observed from tumor regions demonstrating increases in tissueechogenicity. Textural analysis of parametric images indicated that featuressuch as homogeneity and contrast could detect responses as early as1 week after the start of treatment.Conclusions—This study demonstrates the potential of ultrasoundto quantify changes in tumors in response to cancer treatment administrationin a clinical setting. This approach may assist in thecustomization of cancer treatments facilitating switching from ineffectivetreatments to efficacious therapies.1540914 Nonlinear Modeling of the Canine Liver With IncreasingHepatic PressureVeronica Rotemberg, 1 * Brett Byram, 1 Mark Palmeri, 1,2Michael Wang, 1 Kathryn Nightingale 1 1 Biomedical Engineering,Duke University, Durham, North Carolina USA;2Anesthesiology, Duke Medical Center, Durham, North CarolinaUSAObjectives—Elevated hepatic venous pressure is associatedwith leading causes of death from advancing liver disease and is currentlymonitored using an invasive and expensive method. The mechanical behaviorof the liver during pressurization is not well understood. In thiswork, liver strain during hepatic pressurization was characterized usingsuccessively acquired 3D B-mode volumes and was compared with concurrentshear wave speed (SWS) estimates. An experiment was designedto elucidate liver nonlinear material properties by using volumetric imagingin unconstrained ex vivo canine livers to estimate the change in liverstrain with pressurization and compare this change in strain with simultaneousSWS estimates. The concurrent strain and SWS estimate informationis actively being used to develop a nonlinear material model forhepatic behavior with increasing portal venous pressure.Methods—Hepatic pressure was increased stepwise from 0 to20 mm Hg with 3D B-mode acquisition during each step at 3.2 volumes/susing a Siemens Acuson SC-2000 scanner and 4z1c matrix array transducer(Mountain View, CA). Displacements were calculated using 3Dcross-correlation and a 2.88 × 0.60 × 0.68-mm tracking kernel. 1 Strainswere estimated in a 20 × 12 × 12-mm region of interest (ROI). Six acousticradiation force impulse–based SWS estimates in the same ROI were generatedat the end of each step.Results—Increases in SWS and axial strain as a function of hepaticpressure as well as the relationship between axial strain and SWSestimates were developed. During portal venous pressurization, 10% increasesin axial strain corresponded to a 1.25-m/s increase in hepatic SWSestimates above baseline estimates at 0 to 5 mm Hg.Conclusions—Increases in axial strain and SWS estimates observedwith increasing hepatic pressure support the development of a nonlinearmechanical material model for the pressurized liver. This materialmodel may lead to noninvasive hepatic pressure characterization usingstiffness metrics.References1. Byram BC, et al. 3-D phantom and in vivo cardiac speckle trackingusing a matrix array and raw echo data. IEEE Trans Ultrason FerroelectrFreq Control 2010; 57:839–854.1514461 UroImage: A Prostate Tissue Characterization/ClassificationSystem Using Grayscale FeaturesGyan Pareek, 1 Rajendra Acharya, 2,3 Swapna Goutham, 4Vinitha Sree, 5 Ratna Yantri, 2 Roshan Martis, 2 Luca Saba, 6Ganapathy Krishnamurthi, 7 Giorgio Mallarini, 6 AymanEl-Baz, 8 Shadi Al Ekish, 1 Michael Beland, 9 Jasjit Suri 5,10 *1Section of Minimally Invasive Urologic Surgery, Warren AlpertMedical School, Brown University, Providence, Rhode IslandUSA; 2 Electronics and Computer Engineering, Ngee Ann Polytechnic,Singapore; 3 Biomedical Engineering, University ofMalaya, Kuala Lumpur, Malaysia; 4 Applied Electronics and Instrumentation,Government Engineering College, Kozhikode,Kerala, India; 5 Global Biomedical Technologies, Roseville,California USA; 6 Radiology, Azienda Ospedaliero Universitariadi Cagliari, Cagliari, Italy; 7 Mayo Clinic, Rochester, MinnesotaUSA; 8 Bioengineering, Speed School of Engineering,University of Louisville, Louisville, Kentucky USA; 9 Ultrasound,Rhode Island Hospital, Providence, Rhode Island USA;10Biomedical Engineering, Idaho State University, Pocatello,Idaho USAObjectives—Prostate transrectal ultrasound (TRUS) images canbe easily acquired in real time at lower cost and hence are widely used forprostate cancer (CaP) diagnosis. However, the prostate regions in TRUSimages are characterized by a weak texture, speckle, short grayscaleranges, and shadow regions. There is a need for image analysis frameworksthat effectively quantify the subtle textural changes in cancerousand noncancerous TRUS prostate images to accurately detect CaP. In thiswork, we have proposed an online computer-aided diagnostic systemcalled “UroImage” that classifies a TRUS image into cancerous or noncancerouswith the help of nonlinear higher-order spectra (HOS) featuresand discrete wavelet transform (DWT) coefficients.Methods—The UroImage system consists of an online systemwhere 5 significant features (1 DWT-based feature and 4 HOS-based features)are extracted from the test image. These online features are transformedby the classifier parameters obtained using the training data set todetermine the class label of the test image. We trained and tested 6 classifiers.The data set used for evaluation had 144 TRUS images, which weresplit into training and testing sets, and cross-validation was adapted fortraining and estimating the accuracy of the classifiers. The ground truthused for training was obtained using the biopsy results.Results—Among the 6 classifiers, using 3- and 10-fold crossvalidationtechniques, support vector machine and fuzzy Sugeno classifierspresented the best classification accuracy of 97.95% with equally highvalues for sensitivity, specificity, and positive predictive value.S61

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!