Gamma change raised mean AUC for onset marks after high gamma control in fixed models for selected pairs from 16 SEEG test patients

We tested whether early seizure gamma change adds onset ranking information after high gamma control. Public OpenNeuro ds004100 version 1.1.3 supplied 32 SEEG patients and 64 seizures. Fixed logistic models used 16 training patients and 16 test patients. The baseline controlled contact number, baseline power in both bands, and high gamma change. The full model added gamma change. Mean test AUC increased from 0.85966 to 0.87150 on selected bipolar pairs. The mean gain was 0.01184. The patient bootstrap 95 percent interval was 0.00206 to 0.02164, with training fits fixed. The exact two-sided sign-flip test gave p = 0.03650. That test needs independent patients and symmetric effects under its null hypothesis. The paired contrast with a before-onset control was inconclusive. This is a planned secondary analysis of a reused cohort, with no external registration. The result does not establish a separate neural oscillation or a treatment benefit. Fresh source checks passed for all 64 recordings and 757,770 saved power values. All 209 saved metadata files and 64 EDF annex pointers also matched the fixed public source. The package preserves code, source powers, versioned retrieval records, predictions, fitted models, patient draws, checks, and repeat instructions.

stereoelectroencephalography · seizure onset · gamma power · high gamma · AUC · patient-level analysis · epilepsy