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

A multi-stage real-time pipeline for intraoperative iEEG processing and functional mapping

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

Teruaki Kimishima¹ (), Emily Cunningham¹, David Brang¹, Shawn Hervey-Jumper²; ¹University of Michigan, ²University of California, San Francisco

Background: Intraoperative mapping during brain tumor resection is essential for preserving critical cognitive and functional regions. In addition to direct electrical stimulation (DES), electrocorticography (ECoG) or intracranial EEG (iEEG) has been increasingly adopted for mapping due to its millisecond-level temporal resolution. However, current real-time systems are often limited to visualizing raw signals without automated quality assessment or functional analysis, making it difficult for neurosurgeons to make rapid, informed decisions during surgery. Methods: We present a two-stage MATLAB/Simulink pipeline for intraoperative iEEG processing. The first stage automatically evaluates signal quality by identifying extreme variability and detecting common artifacts—line, flatline, drift, baseline-shift, and electrode-popping noise—every 500 ms using robust statistical metrics. The second stage computes accumulating, time-locked high-gamma (70–150 Hz) activity in real time from preprocessed signals, generating an evolving evoked average aligned to stimulation events to continuously track cognitive and motor responses. Results: Using simulated and patient-derived datasets, the artifact detection stage achieved high accuracy in detecting each noise type (>90%). The high gamma response estimation stage showed strong correlation with offline gold-standard analysis (r = 0.94, p < 0.01). System latency remained below 0.01 s, confirming suitability for real-time monitoring. Conclusion: This integrated real-time iEEG pipeline demonstrates robust noise rejection and reliable time-locked cognitive-signal detection, supporting its potential for intraoperative application. By unifying these approaches, the pipeline strengthens the precision and confidence of functional mapping during surgery. Its functionality can be extended to generate DES priority maps derived from evoked responses and predictive machine learning models of accumulating resting-state data.

Topic Area: METHODS: Electrophysiology