A MIPAV plugin for the visual and quantitative comparison of FreeSurfer subcortical segmentation results

Wednesday, October 26, 2011 — Poster Session III
10:00 a.m. – Noon	Natcher Conference Center	NIMH	IMAG-18

Authors

• F Lalonde
• R Cheng
• J Senseney
• M McAuliffe

Abstract

FreeSurfer is an automated structural brain imaging pipeline that includes the segmentation of subcortical structures. Factors such as scanner manufacturer, magnetic field strength, head placement relative to isocenter, and pulse sequence have been shown to affect the reliability of the subcortical segmentation. Reliability of segmentation results is often measured using intra-class correlations. These correlations, however, do not provide any information as to where the subcortical structures differ. A new MIPAV plugin was created to enable visual comparison of segmentations of a specified subcortical structure across multiple MRI sessions. The plugin provides an automatic segmentation, registration and overlap comparison pipeline for specified pairings of a selected subcortical structure across sessions. Parallel processing on a multi-core CPU node is implemented in Java. The local based registration algorithm is multi-threading enabled and allows a user specified number of parameters (6, 9, 12). In addition, MIPAV’s command line interface enables the plugin to run simultaneously over hundreds of nodes on a cluster computer (Biowulf). The clustering, local based multithreading, and efficient memory management form a powerful computing infrastructure to run the pipeline on large MRI datasets. Once scanner test-retest reliability has been established, the technique can be used to measure longitudinal brain changes.

back to top