Meeting Banner
Abstract #1791

Simple, Accurate, Whole-Brain White Matter Segmentation in 3 Seconds

Wenzhe Xue 1,2 , Christine M. Zwart 2 , and Joseph Ross Mitchell 2

1 Biomedical Informatics, Arizona State University, Scottsdale, AZ, United States, 2 Radiology, Mayo Clinic, Scottsdale, AZ, United States

Segmentation of brain white matter could aid assessment of neurological disorders. However, this is not routinely performed in clinical settings, due in part to long computation times and the need for users to carefully tune multiple algorithm parameters to achieve acceptable results. Our goal is to develop a simple, rapid, reliable and accurate technique to segment brain white matter, gray matter and ventricles. We are extending the new parallel level set algorithm proposed by Roberts et. al.. This algorithm efficiently leverages the massive parallelism of commodity graphical processing units (GPUs) to achieve a 14x speed over previous parallel algorithms. Despite this speed advantage, the user is still required to place an initial seed and then tune three parameters to achieve acceptable results. The tuning process requires expertise and increases segmentation time, variability, and inconvenience. Here we report on efforts to automatically select optimal values for the three algorithm parameters when segmenting brain white matter in T1-weighted MR exams.

This abstract and the presentation materials are available to members only; a login is required.

Join Here