Adding early gamma change increased mean AUC for clinical onset marks in selected bipolar pairs from 16 ds004100 test patients

We tested the added value of early seizure gamma change in public SEEG data from ds004100 version 1.1.3. We fitted fixed models with 16 patients and tested them with 16 other patients. Each patient supplied two seizures. The baseline model used prior power rank and contact number. Adding the gamma change rank raised mean AUC for clinical onset marks from 0.810 to 0.864. The gain was 0.053; its descriptive 95 percent interval was 0.015 to 0.094. The exact patient sign-flip test gave Holm P=0.0429 for two planned tests. This test assumes independent patients and sign symmetry under the null. The gain exceeded a before-onset control by 0.048, but that difference was uncertain: interval -0.007 to 0.105; Holm P=0.128. All six fixed sensitivity checks had positive mean seizure gains. These results apply to the selected bipolar pairs and source clinical marks. Earlier results used all 32 patients. This is an exploratory secondary analysis, with no new external cohort or clinical outcome test. The package preserves all source features, fixed retrieval records, code, fits, predictions, and instructions. A fresh audit checked all 64 source recordings and 757,770 power values.

stereoelectroencephalography · epilepsy · gamma power · seizure onset · ranking · AUC · patient-level validation