113 Novel Resting-State Electrophysiologic Detection of Speech Networks in Glioma-Infiltrated Cortex
Summary
No summary yet. VIRAT can read this paper and write a plain-language summary of its aim, methods, key findings and limitations.
Summarise with VIRAT Free to try with your Science Catalogs account.Abstract
INTRODUCTION: Awake language mapping is essential for safe glioma resection, but the current gold standard, direct cortical stimulation (DCS), requires patient cooperation and significant resources. Resting-state electrophysiologic tools could offer rapid, scalable alternatives, especially for patients unable to tolerate awake mapping or in resource-limited settings. METHODS: Resting-state ECoG recordings were analyzed from 81 prospective glioma patients (36 IDH-mutant; 45 IDH–wildtype) undergoing awake craniotomy with DCS mapping. Electrodes were aligned to DCS-mapped sites. Analyses were stratified by IDH status. Power spectral density was computed to extract electrophysiological features: broadband (2-150 Hz) deviation and narrowband oscillations (theta to high gamma). These features were used to train multivariate classifiers (LASSO, Ridge, Elastic Net, logistic regression) to predict DCS status. Model generalization was evaluated in an external patient (IDH-mutant; 3 DCS+; 5 DCS– sites). RESULTS: In the IDH-mutant cohort, DCS+ sites showed significantly elevated broadband deviation and stronger oscillatory activity across theta to high gamma bands, revealing key electrophysiologic biomarkers of IDH-mutant DCS+ sites (all p0.05). Given IDH-specific differences, we trained multivariate classifiers to predict DCS+ sites in IDH-mutant patients. Classifiers achieved high accuracy (85.5%–90.3%) and AUROC (0.746–0.865) in the internal dataset. Further, models were highly generalizable in the external dataset (accuracy: 87.5%, AUROC: 0.933). CONCLUSIONS: This is the largest, prospective electrophysiologic analysis of glioma-infiltrated cortex and of DCS-mapped language sites. We show for the first time that resting-state biomarkers from subdural ECoG reliably differentiate language-critical cortex identified by stimulation, and are generalizable in an external dataset. These findings offer an immediate, actionable tool to augment/substitute DCS, particularly in low-resource settings or when awake mapping is not clinically feasible. Further, IDH status-specific differences suggest that language network reorganization may vary by molecular subtype.
Fields of Study & Subjects
Glioma Diagnosis and Treatment
Neural dynamics and brain function
EEG and Brain-Computer Interfaces
Publication details
- Journal / source
- Neurosurgery
- Volume
- 72
- Issue
- Supplement_1
- Pages
- 14–15
- Publication date
- 26 March 2026
- Language
- EN