The patient bootstrap interval included zero added SOZ rank gain from post-seizure gamma in 29 ds004100 SEEG patients under independent, exchangeable patient sampling

We tested whether gamma decrease 10 to 30 seconds after seizure end adds seizure onset zone (SOZ) rank information beyond onset gamma and contact number. We used public OpenNeuro ds004100, version 1.1.3, with two records per patient. The main test used 29 patients and 472 adjacent bipolar pairs on 62 shafts with both labels. Each test patient was absent from fitting and training scaling. Training used run 01 from the other patients. Testing used run 02 from the excluded patient. A fixed pairwise model gave mean patient AUC 0.790 for onset gamma and contact number, and 0.856 after early gamma decrease was added. The mean gain was 0.066. Its descriptive 95 percent patient bootstrap interval was -0.006 to +0.178 after model refitting in 2,000 draws. The interval assumes independent, exchangeable patient sampling and a stable fixed learner. A second seed also gave an interval that included zero. Nine patients improved, 16 had no change, and four declined. One patient had one scored comparison and a gain of one. The evidence does not establish added benefit. This is a secondary analysis of a known cohort from one center. Pair selection uses clinical labels, so the result is retrospective. We rebuilt all 64 source records, checked the main fits with a separate solver, and obtained identical numerical files in a clean repeat. The package preserves code, source versions, power data, predictions, fitted models, bootstrap draws, all planned checks, and reproduction instructions.

stereoelectroencephalography · seizure onset zone · gamma power · postictal · patient bootstrap · feature addition