Mean time pattern agreement stayed positive after a voltage mask with 0.5-second margins in eligible 1–25 Hz SEEG from two epilepsy cohorts

We tested whether large voltage events account for mean time pattern agreement in public SEEG. The study used 36 HUP patients and 19 NIH or JHU patients with epilepsy. Signals were bipolar, passed 1–25 Hz, and had a sample rate of 64 Hz. A fixed mask removed samples beyond six times a MAD-based scale and a 0.5-second margin on each side. Pairs had to retain at least half their valid triples in both records or intervals. All 55 patients and 677 of 680 pairs qualified. Quiet mean agreement was 0.00022135 in HUP and 0.00036414 in NIH or JHU. Both 95 percent patient bootstrap intervals were above zero. The mean fell by 4.7 percent and 8.6 percent against the same pairs without removal. Independent uniform shifts of a mask preserve expected agreement when all triples have strict ranks. We proved this reference and checked every primary signal. Large voltages and nearby samples contributed to the score, but the retained samples still had positive mean agreement. The mask does not identify clinical spikes. The result is limited to these processed signals and this selection rule. The package contains data, masks, counts, code, source hashes, raw retrieval recipes, and all fixed checks.

SEEG · ordinal patterns · voltage mask · time pattern agreement · intracranial EEG · epilepsy · signal processing