A semiology-driven framework for explorations of temporal lobe epilepsy

Area of Science:

• Neurosurgery
• Epileptology
• Computational Neuroscience

Background:

• Drug-resistant temporal lobe epilepsy (TLE) necessitates surgical intervention, but broad resections risk neuropsychological deficits.
• Stereo-electroencephalography (SEEG) aids in mapping seizure networks for hypothesis-driven exploration.
• Seizure semiology, imaging, scalp EEG, and neuropsychology guide SEEG implantation but semiology alone is not a sole classifier.

Purpose of the Study:

• To define common SEEG implantation patterns in TLE from a stereotactic functional-anatomic viewpoint.
• To test the reproducibility of these SEEG implantation patterns.
• To correlate semiology and electrode coverage with surgical outcomes.

Main Methods:

• Analysis of 72 patients with drug-resistant TLE undergoing SEEG.
• Identification of four reproducible SEEG implantation patterns: mesial-lateral temporal, temporal-basal-occipital, anterior perisylvian, and perisylvian.
• Machine learning classifiers trained on semiological features and electrode coverage to predict implantation patterns and outcomes.

Main Results:

• Semiology alone had moderate predictive power for outcomes (precision 66%, recall 65%).
• Electrode coverage significantly improved prediction (precision 87%, recall 85%).
• Semiology strongly correlated with implantation patterns (precision 84%, recall 82%), confirming its role in guiding strategy. 73% seizure freedom at 1 year.

Conclusions:

• Reproducible SEEG implantation patterns in TLE are characterized and informed by semiology.
• Computational analysis validates these patterns, linking implantation strategies to network localization and surgical approaches.
• SEEG-guided therapy led to high seizure freedom rates, with verbal memory decline observed in 26% of resection patients.