Meeting Banner
Abstract #3893

New Strategy for Registering DW & Non-DW Images Via Tensor Estimation Metric

Cheng Guan Koay1,2, Andrew L. Alexander1, M. Elizabeth Meyerand1

1Department of Medical Physics, University of Wisconsin-Madison, Madison, WI, United States; 2STBB, National Institutes of Health, Bethesda, MD, United States

Registration of DW and non-DW images is a critical data-analytics step. In this study, we proposed a novel DW and non-DW image registration strategy that provides the much needed information on the goodness-of-registration of DWI images to non-DWI (T2-weighted image or template) in terms of tensor estimation metric within the registration process. This strategy requires very minimal modification to the existing acquisition of DTI except two distinct b-values and is built upon weighted linear least squares estimation for computational efficiency consideration. Reduction in registration error as computed from the metric was about 48.6% for the experimental data set we tested.