Baseline division reduced mean early gamma AUC for clinical seizure onset labels in 31 SEEG patients with a common average reference

We tested baseline division of early gamma power in 31 SEEG patients from OpenNeuro ds004100 version 1.1.3. The sample had 62 seizures and 3,674 matched contacts. The primary analysis used a common average reference. It used 30–80 Hz power with 55–65 Hz removed. The seizure window was the first ten seconds after the source onset. The baseline window was 20 to 10 seconds before onset. Mean patient AUC for clinical seizure onset labels fell from 0.807 to 0.678 after division. The mean change was −0.129. The 95% patient bootstrap interval was −0.181 to −0.080. AUC fell in 27 patients. The planned sign-flip test gave p = 0.0000070. This test assumes independent patients and symmetric contrast signs under the null. The decrease also appeared within electrode name groups and after a stricter signal quality check. Fresh pinned signal downloads reproduced all 62 feature sets. The package preserves code, powers, source metadata, results, figures, and retrieval instructions. This is a new analysis of a prior cohort. The result measures existing clinical label agreement; it does not establish diagnostic accuracy or clinical benefit.

stereoelectroencephalography · epilepsy · gamma power · baseline normalization · seizure onset zone · area under the curve · signal processing