Ensuring anatomical specificity in lesion network mapping: a multicentre study of 2,950 stroke patients

As many neurobehavioural functions rely on distributed brain networks, anatomically diverse brain lesions can cause highly similar neurobehavioural deficits. Lesion network mapping (LNM) builds on this principle to identify symptom-specific brain networks by mapping focal lesions onto a normative functional connectome. Recent work, however, questioned the anatomical specificity of LNM, showing that commonly used LNM procedures converge on nonspecific properties of the normative connectome rather than symptom-specific networks. Here, we investigated whether anatomically specific LNM can be achieved through appropriate study design and statistical inference. Using a multicenter dataset of 2,950 stroke patients across 12 cohorts, we compared lesion-derived connectivity maps between patients with and without impairment across six cognitive domains. We evaluated group differences using both permutation of symptom labels and parametric two-sample inference, and additionally examined a one-sample approach similar to what was used in the previous critique. Both two-sample approaches yielded distinct, biologically plausible lesion network maps with evident domain-specific features, and limited correspondence to nonspecific connectome properties. Simulation analyses across 10,000 null studies confirmed valid type I error control of the permutation framework. In contrast, the one-sample analyses, which aggregated lesion connectivity maps across affected patients without a control comparison, reproduced the previously described convergence toward nonspecific connectome properties and showed marked similarity across cognitive domains. These findings demonstrate that connectome-driven convergence is not an inherent limitation of LNM, but depends critically on study design: case-control comparisons with appropriate statistical inference ensure anatomical specificity and provide a robust framework for identifying brain networks associated with distinct neurobehavioural deficits.