A method to enhance the sensitivity of DTI analyses to group differences: a validation study with comparison to voxelwise analyses

Abstract

Studies of white matter (WM) abnormalities in psychiatric and neurological disorders often use the analysis package Tract-Based Spatial Statistics (TBSS). However, with small samples and/or subtle effects, a study using the standard TBSS approach can be underpowered. For such cases, a new method is presented that summarizes global differences between TBSS-derived fractional anisotropy (FA) images with a single paired t-statistic, estimating the degrees of freedom using spatial autocorrelation. The sensitivity of the method is demonstrated by using well-known aging effects on FA as a proxy for disease effects. Sixty healthy subjects were divided equally into younger- (YA), middle- (MA), and older-aged (OA) groups and significant global differences were demonstrated in the YA versus OA (all N ≥ 4, FA difference≈0.023), MA versus OA (all N≥4, FA difference≈0.017), and YA versus MA (FA difference≈0.005 at N=20) comparisons. In contrast, no significant difference could be detected in the YA versus MA comparison using voxelwise TBSS analysis with the full sample (N=20 per group). This method should facilitate localizing analyses in the direction of a proven group difference while providing clinically relevant information about pathophysiologic processes globally affecting WM.