IMG-114. Pre-operative right-hemisphere language network functional connectivity predicts post-operative language impairment

Abstract Post-operative language deficits are common in left-hemisphere glioma patients. While left-hemisphere lateralization of language is well-established, evidence suggests the right hemisphere may play a compensatory role in left-hemisphere glioma patients, though its value in predicting post-operative deficits remains unclear. This study examines whether pre-operative resting-state functional connectivity within the right-hemisphere language network predicts post-operative language impairment in left-hemisphere glioma patients, potentially reflecting resilience to iatrogenic deficits through right-hemisphere functional reorganization. Resting-state fMRI data were acquired in 41 left-hemisphere glioma patients prior to neurosurgery. We calculated the right-hemisphere ROI-to-ROI functional connectivity (RH-RRC) for every unique pair of ROIs in a language atlas. Post-operative language outcomes were obtained from clinical documentation during the in-patient recovery period. A support vector machine (SVM) binary classifier with nested cross-validation predicted whether patients were language-impaired or language-intact based on the pre-operative RH-RRC in the language network. Model performance was evaluated using accuracy and Matthews correlation coefficient (MCC), with significance determined through permutation testing. Pre-operative RH-RRC in the language network predicted post-operative language deficits with 70% accuracy (MCC = 0.46; both p < 0.05). Three control SVMs were constructed using RH-RRC in the default mode, somatomotor, and visual networks; these networks were selected to assess potential confounders (e.g. the control networks differ in anatomical proximity to the language network). We also implemented a control SVM using IDH-mutation, WHO grade, lesion volume, and age as predictors. None of the control SVM models performed better than chance (all accuracy and MCC p > 0.05), emphasizing that the ability to predict post-operative impairments is specific to the pattern of right-hemisphere language network functional connectivity. Pre-operative right-hemisphere language network connectivity predicts post-operative language deficits. The absence of predictive value in control networks and clinical models supports the specificity of this finding, suggesting that pre-operative contralesional functional reorganization uniquely contributes to post-operative language outcomes.