Automated segmentation of the canine corpus callosum for the measurement of diffusion tensor imaging

Abstract

The goal of this study was to apply image registration-based automated segmentation methods to measure diffusion tensor imaging (DTI) metrics within the canine brain. Specifically, we hypothesized that this method could measure DTI metrics within the canine brain with greater reproducibility than with hand-drawn region of interest (ROI) methods. We performed high-resolution post-mortem DTI imaging on two canine brains on a 7 T MR scanner. We designated the two brains as brain 1 and brain 2. We measured DTI metrics within the corpus callosum of brain 1 using a hand-drawn ROI method and an automated segmentation method in which ROIs from brain 2 were transformed into the space of brain 1. We repeated both methods in order to measure their reliability. Mean differences between the two sets of hand-drawn ROIs ranged from 4% to 10%. Mean differences between the hand-drawn ROIs and the automated ROIs were less than 3%. The mean differences between the first and second automated ROIs were all less than 0.25%. Our findings indicate that the image registration-based automated segmentation method was clearly the more reproducible method. These results provide the groundwork for using image registration-based automated segmentation methods to measure DTI metrics within the canine brain. Such methods will facilitate the study of white matter pathology in canine models of neurologic disease.