Time clustering of the six largest 80–150 Hz power windows gave a mean local seizure onset difference of −0.48 percentage points. Tests used two 60-second clips per patient in 30 SEEG patients.

We tested whether large power windows occur together more often at seizure onset pairs. The source was public SEEG data from OpenNeuro ds004100, version 1.1.3. Tests used two 60-second clips per patient in 30 patients. The comparison used 196 seizure onset pairs and 341 other pairs in 69 contact name groups. For each pair and clip, we selected the six largest one-second 80–150 Hz power windows. The band excluded 118–122 Hz. The cluster score was the number of selected neighbors divided by five. Groups, clips, and patients had equal weight at each level. The mean seizure onset minus other-pair score was −0.48 percentage points. The 95 percent patient bootstrap interval was −3.34 to +2.46 points. The planned two-sided sign-flip probability was 0.754. The test assumes independent patients and symmetric differences under the zero null model. The mean local area under the curve was 0.494. Uniform time order gives an exact mean cluster score of 0.1. Both recorded class means exceeded this benchmark. The time-order control was descriptive. We retrieved all 72 pinned source prefixes again. All hashes, power arrays, and contact selection records matched. Independent checks and a full offline repeat reproduced the calculations. The package preserves code, inputs, selected windows, random assignments, exact counts, results, and reproduction instructions. The result does not prove equal class means. One-second power windows do not identify individual neural oscillations. The study used short clips and clinical labels from one dataset.

stereoelectroencephalography · interictal SEEG · high-frequency power · time clustering · seizure onset labels · window selection