Schedule of Events | Search Abstracts | Invited Symposia | Symposia | Poster Sessions | Data Blitz

Presurgical right-hemisphere language network connectivity predicts spared language function after left-hemisphere tumor surgery.

Poster Session C - Sunday, March 8, 2026, 5:00 – 7:00 pm PDT, Fairview/Kitsilano Ballroom

Emma Strawderman^1,2 (), Conor O'Hara², Madalina Tivarus², Michelle C. Janelsins², Tyler Schmidt², Kevin A. Walter², Webster H. Pilcher², Bradford Z. Mahon^2,3, Frank E. Garcea^2,4; ¹University of Rochester School of Medicine and Dentistry, ²University of Rochester Medical Center, ³Carnegie Mellon University, ⁴University of Rochester

Although the networks supporting language are typically left-lateralized, the right hemisphere may play a compensatory role in patients with left-hemisphere tumors. We tested whether presurgical right-hemisphere language network functional connectivity predicts post-operative language outcomes in left-hemisphere glioma patients. Resting-state fMRI data were acquired from 55 left-hemisphere glioma patients before neurosurgery. Right-hemisphere ROI-to-ROI connectivity (RH-RRC) was computed for all ROI pairs from the Fedorenko language network atlas. Post-operative language outcomes were extracted from clinical documentation during inpatient recovery. A random forest classifier with nested cross-validation was trained to distinguish language-impaired (n = 27) from language-spared (n = 28) patients using the pattern of RH language RRC. Model performance was evaluated using accuracy, sensitivity for preserved language, and AUC. Permutation testing assessed the statistical significance of model performance and identified features critical to accurate classification. A clinical model using tumor grade, a tumor marker, tumor volume, sex, and age was constructed for comparison. Right-hemisphere language network connectivity accurately predicted post-operative language outcomes (mean accuracy = 73%, sensitivity = 75%, and AUC = 0.75; all p = 0.001), whereas the control model performed at chance levels. Functional connectivity between the right middle frontal gyrus and pars orbitalis emerged as the most influential classifier feature and was significantly elevated in patients with preserved language function after surgery (t = 2.68, p = 0.005). Enhanced coupling between these executive control and semantic language regions may support the reallocation of cognitive resources to compensate for increased language processing demands following surgical disruption of the left-hemisphere.

Topic Area: LANGUAGE: Other