Optimally-Discriminative Voxel-Based Morphometry Significantly Increases the Ability to Detect Group Differences in Schizophrenia, Mild Cognitive Impairment, and Alzheimers Disease

Tianhao Zhang¹, Christos Davatzikos¹

Voxel-Based Morphometry (VBM) has been widely applied for characterizing brain changes on structural Magnetic Resonance Imaging. However in the conventional VBM methods, Gaussian smoothing, which is always used prior to General Linear Model (GLM) to integrate imaging signals from a region, proves critical due to lack of the spatial adaptivity necessary to optimally match image filtering with an underlying region of interest. In this work, Optimally-Discriminative Voxel-Based Analysis (ODVBA), as a recently-developed method utilizing a new spatially adaptive smoothing scheme to determine group differences, is evaluated in comparison with the conventional VBM method, two other spatially adaptive smoothing methods, and two cluster enhancing methods, in three studies on schizophrenia, mild cognitive impairment, and Alzheimers disease.