Altered hierarchical brain structure–function coupling in episodic and chronic migraine

Migraine is increasingly recognized as a network disorder involving distributed structural and functional brain alterations. However, regional changes in the relationship between resting-state activity and white-matter organization remain unclear. We used the structural decoupling index (SDI), a graph signal processing measure of structure–function alignment, to characterize regional coupling alterations in migraine. After quality control, 41 participants with episodic migraine (EM), 31 with chronic migraine (CM), and 36 healthy controls (HCs) underwent multimodal magnetic resonance imaging. Regional SDI was calculated across 246 Brainnetome regions using tractography-defined graphs weighted by mean radial diffusivity (RD) along streamlines; sensitivity analyses used mean diffusivity or fiber-number weighting. Covariate-adjusted analyses with false discovery rate correction assessed group differences, and partial correlations tested clinical associations. Exploratory linear support vector machine classification used fully nested cross-validation, with all analytical steps confined to training folds. Exploratory partial least squares (PLS) regression tested the spatial correspondence between the migraine–HC RD-SDI difference map and normative regional gene expression using spatially autocorrelation-preserving surrogate maps. Compared with HCs, patients with migraine had lower SDI in the left superior frontal gyrus and higher SDI in the right inferior temporal gyrus. Three-group analyses additionally identified differences in the left lateral occipital cortex, where CM showed lower SDI than EM; lower values were associated with more monthly migraine days. The principal regional findings were reproduced in sensitivity analyses using both MD-weighted and FN-weighted structural graphs. Regional RD-SDI features showed preliminary internally cross-validated discrimination, strongest for CM versus HCs (AUC = 0.869, 95% CI: 0.784–0.944). PLS1 explained 8.76% of the variance in the migraine–HC RD-SDI difference map but was not significant under the spatially constrained null model ( P spatial = 0.725); no PLS component survived FDR correction. This first application of SDI to migraine revealed bidirectional hierarchical structure–function disruptions across prefrontal, temporal, and occipital regions. Lower lateral occipital SDI was associated with more monthly migraine days, while regional RD-SDI features showed preliminary internally cross-validated discrimination, particularly for CM versus HCs. SDI may provide a useful framework for characterizing hierarchical brain network dysfunction in migraine.