Poster F101, Tuesday, March 28, 8:00 – 10:00 am, Pacific Concourse
Semi-Automation of a Reliable Method for Measuring Human Insular Cortex
Aliyah Jones1, David Stephenson, M.S.2, Allen L. Reiss, M.D.3, Elliott Beaton, Ph.D.2, Jeremy D. Cohen, Ph.D.1; 1Xavier University of Louisiana, 2University of New Orleans
Insular Cortex, a multimodal region with connectivity throughout the brain, has a role in numerous clinical disorders. Manual morphometry is the ideal means to measure volume of insular cortex, in order to capture subtle inter-subject variability, but is very time consuming. Automated image processing is far more efficient, but is susceptible to losing some of the anatomical variation across subjects. The goal of this study was to combine the accuracy of manual morphometry with the efficiency of an automated algorithm for obtaining measurements of human insular cortex using Advanced Normalization Tools (ANTs). Similar to the previously published protocol for hippocampus, landmarks were placed on insular cortex using MANGO, and ANTs was used to generate automated ROIs. Manual ROIs were used with the automated ROIs to create a correction algorithm that would improve the reliability over the fully automated ROIs. This segmentation adapter overlaps the automated ROI and the manual ROI and corrects the automated tracing (semi-automated). Manual ROIs were used from a previously analyzed sample in Fragile X Syndrome. Intra-class correlation coefficients (ICC) were used to test reliability. Results showed low reliability between manual tracings and ANTs automated tracings indicating that the automated approach alone is not enough. ICC between manual left and semi-automated left measurements was .784, and .864 between manual right and semi-automated right measurements. The semi-automated ROIs were found to be more accurate than just the automated ROIs, indicating this novel ANTs protocol may be a reliable tool for analyzing insular morphometry in larger subject samples.
Topic Area: METHODS: Neuroimaging