Prediction of Critical Speech Sites in Glioma-Infiltrated Cortex using Intraoperative, Resting State Electrophysiologic Biomarkers

Poster Session D - Monday, March 31, 2025, 8:00 – 10:00 am EDT, Back Bay Ballroom/Republic Ballroom

Vardhaan Ambati¹, Sanjeev Herr¹, Jasleen Kaur¹, Paul Villalobos¹, Emily Cunningham², Youssef Sibih¹, Sena Oten¹, Alexander Aabedi¹, David Brang², Shawn Hervey-Jumper¹; ¹University of California, San Francisco, ²University of Michigan

Background Diffuse gliomas often invade speech-critical areas, making maximal safe resection difficult. Direct cortical stimulation (DCS) identifies functional (DCS+) vs. nonfunctional (DCS-) cortex but is technically demanding. This study investigates electrophysiologic biomarkers of DCS+ cortex to improve surgical safety and efficiency. Methods Prospectively collected intraoperative local field potentials (Theta [4-8 Hz], Alpha [8-13 Hz], Beta [13-30 Hz], Gamma [30-70 Hz], High Gamma [70-150 Hz]) from subdural arrays in glioma-infiltrated cortex were compared between DCS+ and DCS- sites. Machine learning classifiers (separate for LGG and glioblastoma [GBM]), using stacked logistic regression and XGBoost, were trained (80%) and tested (20%) to predict DCS+ vs. DCS- sites using resting-state power spectral data (frequency x power). Results In total, 1421 cortical sites were studied, with 512 sites aligned to electrodes (49 DCS+) in 91 patients (oligodendroglioma 21, astrocytoma 22, GBM 48). In LGG (oligodendroglioma/astrocytoma), DCS+ cortex exhibited significantly higher Alpha, Beta, Gamma, and High Gamma power (all p<0.05). In contrast, DCS+ sites in GBM (N=20) showed only increased High Gamma power (p<0.05). Classifier accuracy for LGG was 94%, for GBM 92%. Permutation tests confirmed significant classification performance (LGG: p<0.001; GBM: p=0.0001). Conclusion This is the first study to predict DCS status based on resting-state electrophysiologic biomarkers and demonstrates that identification of DCS+ areas based on frequency power may aid in safe glioma resection. Additionally, these results highlight tumor/grade-specific effects on cortical function (LGG vs. GBM DCS+ sites had different spectral power signatures), prompting further investigation into tumor/grade-specific effects on neural circuits.

Topic Area: METHODS: Electrophysiology