Using predictive validity to compare associations between brain damage and behavior

Abstract

Lesion-behavior mapping (LBM) provides a statistical map of the association between voxel-wise brain damage and individual differences in behavior. To understand whether two behaviors are mediated by damage to distinct brain regions, researchers often compare LBM beta weight outputs by either the overlap method (emphasizing differences) or by the correlation method (emphasizing similarity). However, both methods lack a principled way to determine LBM distinctness and are disconnected from a major goal of LBMs, that of predicting behavior from brain damage. Without such criteria, researchers may draw conclusions from numeric differences between LBMs that are irrelevant to predicting behavior from brain damage. Here, we developed and validated a Predictive Validity Comparison method (PVC) that establishes such a criterion by directly comparing two LBMs based on the difference in predictive accuracy: two LBMs are significantly different if they provide unique predictive power for the two behaviors assessed. Applying the PVC to two lesion-behavior stroke data sets (Ding et al., 2020; Pustina et al., 2018) demonstrated the utility of a statistical criterion for determining when individual differences across behaviors result from the same vs. different lesion patterns. Using region-of-interest based simulations derived from proportion damage in Brodmann regions from a large group of participants (n=131), PVC detected when two behaviors were mediated by different regions (high sensitivity) and determined when two behaviors were mediated by the same region (high specificity). By objectively determining whether two behavioral deficits can be explained by a single vs. distinct patterns of brain damage, the PVC provides a critical advance in establishing the brain bases of behavior. We have released a GUI-driven, open-access web app of this method to encourage widespread adoption.