Program Information
Evaluating Symmetric Diffeomorphic Image Registration with Mutual Information On Voxelwise Kinetic Analysis of Free Breathing Liver Dynamic Contrast Enhanced Magnetic Resonance Imaging
M Liu*, R Al-Senan , B Peng , T Lin , S Zhang , S Jambawalikar , Columbia University New York Presbyterian Hospital, New York, NY
Presentations
MO-L-GePD-I-4 (Monday, July 31, 2017) 1:15 PM - 1:45 PM Room: Imaging ePoster Lounge
Purpose: Quantitative kinetic parameter maps derived from un-motion-compensated images in free breathing abdominal imaging suffer from poor fitting. The purpose is to quantify the effects of diffeomorphic algorithms used to correct for nonlinear abdominal motion on pharmacokinetic parameterization of dynamic contrast enhanced magnetic resonance imaging (DCE-MRI)
Methods: DCE-MRI was performed on patients with hepatocellular carcinoma in a Philips Achieva 3T scanner. Oblique Coronal T1 weighted 2D FSPGR sequence was acquired over a single 7mm thick slice with TE/TR 1.5/4.96 ms, 288x180 matrix, and 1s temporal resolution over 280 free-breathing dynamics. Dynamics were registered to the first time point using symmetric diffeomorphic registration with mutual information and b-spline interpolation made available in the ANTs toolbox. Voxelwise pharmacokinetic maps for extended toft model were generated using OLEA software on both the registered and unregistered datasets. An ROI was drawn on enhancing tumor to extract pharmacokinetic parameters and chi squared from both unregistered and registered sets of parameter maps.
Results: Visual comparisons of the two image sets showed reduced liver motion in the registered images. Tumor margins were sharpened and heterogeneity increased in parameter maps derived from the motion corrected datasets. Between unregistered and registered data sets, percent change in (Ktrans, Kep , Vp, Ve) were (-44%,-12%,-50%,+18%) for first and (+52%,+44%,+50%,-36%) for second patient. Chi squared decreased by 87% in the first patient and decreased by 53% in the second patient.
Conclusion: Kinetic parameters fluctuated wildly between unregistered and registered data sets. An 87% and 53% decrease in chi squared in the first and second patient respectively indicates that the application of a symmetric diffeomorphic image registration with mutual information to each time point greatly increases goodness of fit when fitting kinetic parameters voxelwise to 4D data.
Contact Email: