Generating Data-Driven ROIs for Quantifying Structural Brain Changes in Frontotemporal Dementia

• Activity Number: 360
• Type: Contributed
• Date/Time: Tuesday, August 11, 2015 : 10:30 AM to 12:20 PM
• Sponsor: Section on Statistics in Imaging
• Author(s): Aleksandr Pankov* and Richard Binney and Suneth Attygalle and John Kornak and Howard Rosen
• Companies: and Temple University and UC San Francisco and UC San Francisco and UC San Francisco
• Keywords: Cross-validation ; Voxel Based Morphometry ; Frontotemporal Dementia ; Image Analysis

Abstract

Frontotemporal lobular dementia (FTLD) is a neurodegenerative disorder that is associated with an earlier age of onset and a more rapid rate of decline than Alzheimer's disease. Due to the rush to develop treatments for FTLD, it is of critical importance to have strong biomarkers of FTLD progression. Accordingly, the goal of this project is to find a data-driven region-of-interest (ROI) that leads to the largest effect size when looking at longitudinal structural brain changes with MRI in FTLD patients. To achieve this goal, a cross-validation approach is taken, where maps of estimated change across a population are thresholded at various levels in the training-set. The corresponding effect sizes for the supra-threshold regions (the ROI) applied in the test-set are then calculated. The optimal ROI corresponds to largest cross-validated effect size. The approach will be demonstrated on MRI datasets from a cohort of low (1.5T) and high (3T) magnetic field strength datasets. The cross-validated effect sizes from the data-driven ROIs at each field strength will then be compared with those of conventional anatomically defined ROIs.