Dynamic functional connectivity in left frontal and temporal tumors.

Abstract

This study examined dynamic resting-state functional connectivity and graph-theoretic measures in 71 patients with left intra-axial tumors (40 left frontal lobe tumors and 31 left temporal lobe tumors) and 53 age- and sex-matched healthy controls. Dynamic functional connectivity was estimated across 8 80-s sliding windows, region of interest-to-region of interest Fisher-z connectivity values from 132 atlas regions were harmonized using ComBat, and graph metrics were calculated at a 20% proportional density. The primary within-study comparison was restricted to left frontal lobe tumors and left temporal lobe tumors to avoid the confounding between diagnosis and the public healthy control datasets. In restricted maximum likelihood mixed-effects models with AR(1) within-subject covariance, adjusted for z-scored age, categorical TR, and window, none of the 11 graph metrics showed an overall tumor-location effect (P = 0.219 to 0.928; partial eta-squared < 0.001 to 0.019). However, tumor location-by-window interactions survived correction across 11 metrics for clustering coefficient, transitivity, modularity, and assortativity, at T8, adjusted values were higher in left temporal lobe tumors than left frontal lobe tumors for all 4 metrics. Grade-stratified analyses provided no corrected evidence that grade moderated these temporal location effects, while adjustment for peritumoral edema or standardized log lesion volume did not materially alter the primary tumor-location findings. Secondary comparisons indicated lower integration and altered path-based and centrality measures in tumor groups relative to healthy controls, but diagnosis was partially confounded with acquisition protocol; healthy control-only models identified residual acquisition-batch effects for betweenness centrality and eccentricity after correction. Global connectivity laterality index differences did not survive multiple-comparison correction (minimum q = 0.334) and are interpreted descriptively. These findings distinguish temporally specific topology differences from generalized tumor-location effects and require confirmation in uniformly acquired, clinically characterized cohorts.