Meeting Banner
Abstract #3552

Effects of Diffusion Weighted Image Interpolation for Motion and Distortion Correction on Tensor Statistics

Mustafa Okan Irfanoglu1, 2, Lindsay Walker1, 3, Raghu Machiraju4, Carlo Pierpaoli1

1NIH, NICHD, Bethesda, MD, United States; 2Center for Neuroscience and Regenerative Medicine, Uniformed Services University of the Health Sciences, Bethesda, MD, United States; 3Center for Neuroscience and Regenerative Medicine, Uniformed Services University of the Health Sciences, Bethesda, MD, United States; 4Computer Sciences & Engineering, The Ohio State University, Columbus, OH, United States

Diffusion imaging data generally requires additional postprocessing in order to correct for artifacts and distortions. Typical correction methodologies involve the use of image registration techniques, which employ some form of interpolation to generate the final corrected images. The effects of these interpolation steps have generally been disregarded in the community. In this work, we show that even with the same data and fixed registration correction parameters, the choice of interpolation technique can have a profound effect on the distribution of tensor derived scalar quantities, hence can affect the outcomes of histogram based analysis.

Abstract PDF E-Poster Video

Keywords

addition additionally affected amount analytical analytically anisotropy applications applied approaches artifacts behavior belonging body brain channel clinical coil combined commonly community computed computer consecutive consisted contaminated continuous contrary corpus corrected correction corrections created cubic dataset datasets decrease decreased deformation derived design determined differ diffusion discontinuities discrete displays disregarded distortion distortions distributions draw eddy elastically employ employed emulate engineering errors even excite experimental features field final finally fractional framework frequencies functions generally generate generated grid ground health healthy hence highest histogram histograms induced interfaces interpolated interpolation involved kernels linear lowest maps matrix mostly motion nearest needed neighbor observe outcomes partially performance phantom plotted predicted probabilities processing produces properties quantities real regenerative registering registration registrations requires scanner scheme sciences several significantly situations slice smoothing space spatially splines statistics step still strategies structural subject subsequently suffers suitable susceptibility synthetic tend tensor tensors tissue trace transformation transformed truth uniformed versus walker white whole yield yields