The Effects of Amygdalar Size Normalization on Group Analysis in Late-Life Depression

Structural MRI has been utilized in numerous ways to measure morphologic characteristics of subcortical brain regions. Volumetric analysis is frequently used to quantify the size of brain structures to ultimately compare size differences between individuals. In order to make such comparisons, inter-subject variability in brain and/or head size must be taken into consideration. A heterogeneous set of methods are commonly used to normalize regional volume by brain and/or head size yielding inconsistent findings making it diffcult to interpret and compare results from published volumetric studies. This study investigated the effect that various volume normalization methodologies might have on group analysis. Specifically, the amygdalae were the regions of interest in elderly, healthy and depressed individuals. Normalization methods investigated included spatial transformations, brain and head volume, and tissue volume techniques. Group analyses were conducted with independent t-tests by dividing amygdalar volumes by various volume measures, as well as with univariate analysis of covariance (ANCOVA) analyses by using amygdalar volumes as dependent variables and various volume measures as covariates. Repeated measures ANOVA was performed to assess the effect of each normalization procedure. Results indicate that volumetric differences between groups varied based on the normalization method utilized, which may explain, in part, the discrepancy found in amygdalar volumetric studies. We believe the findings of this study are extensible to other brain regions and demographics, and thus, investigators should carefully consider the normalization methods utilized in volumetric studies to properly interpret the results and conclusions.