Ten-second block limits from 60-second calibration had 44.0% mean coverage for high-gamma log-power means across interictal SEEG records from 36 patients at a 90% target

We tested limits for mean high-gamma log power across interictal SEEG records. The sample had 36 patients and 2,839 bipolar pairs. Calibration used the first 60 seconds of record A. Targets used 60 seconds later in A or in record B. Ten-second block limits had a 90% coverage target. Their mean separate-record coverage was 44.0%, with a 95% interval of 39.2% to 48.8%. Each patient had equal weight. Coverage was 71.6% later in A. The primary block-minus-independent difference was +0.12 percentage points. Its 95% interval was -1.45 to +1.76 points. This does not show a clear block gain. Stationary simulations showed that finite block resampling can miss the target without a signal change. The package includes all inputs, limits, figures, code, and fixed source checks. These results concern power stability in a clinical cohort. They do not establish intent decoding or stimulation benefit.

intracranial EEG · SEEG · high-gamma power · calibration · block resampling · coverage · brain-computer interface