Abstract
In this article we address analytic challenges inherent in brain volumetrics (i.e., the study of volumes of brains and brain regions). It has sometimes been assumed in the literature that deviations in regional brain size in clinical samples are directly related to maldevelopment or pathogenesis. However, this assumption may be incorrect; such volume differences may, instead, be wholly or partly attributable to individual differences in overall dimension (e.g., for head, brain, or body size). What quantitative approaches can be used to take these factors into account? Here, we provide a review of volumetric and nonvolumetric adjustment factors. We consider three examples of common statistical methods by which one can adjust for the effects of body, head, or brain size on regional volumetric measures: the analysis of covariance, the proportion, and the residual approaches. While the nature of the adjustment will help dictate which method is most appropriate, the choice is context sensitive, guided by numerous considerations—chiefly the experimental hypotheses, but other factors as well (including characteristic features of the disorder and sample size). These issues come into play in logically framing the assessment of putative abnormalities in regional brain volumes.
Notes
*Brain volumes are commonly derived by separating (“segmenting”) gray from white matter, and then identifying the boundaries of specific brain regions (e.g., thalamus, hippocampus); volumes are then typically derived by a count of voxels. The discussion that follows thus does not directly bear on brain structure measurement techniques (e.g., voxel-based morphometry) that do not specifically yield measurements of regional volumes.
*The Passingham study used an age range of 18 to 45, whereas the Pakkenberg and Voigt study used an age range of 28 to 41.