Confirmed seizure activity reached 10% of sampled SEEG pairs within a median 2.5–5 s

We measured the time from seizure activity on bipolar SEEG pairs to recruitment of 10% of sampled pairs in 143 focal seizures from 37 patients at one center. The analysis used 20–100 Hz power in 1-s windows, with activity defined as a z-score above 3 for three consecutive windows relative to a -40 to -25 s baseline. With the first active pair as the start, median D10 was 7.5 s (patient median 7.0 s; 95% CI 5.0–10.0); 67% of seizures had D10 of at least 5 s. D10 showed within-patient consistency (F = 3.0, permutation p < 1e-4; ICC approximately 0.34), and the first seizure's D10 predicted the median of later seizures (Spearman 0.44, p = 0.009). First-active pairs were enriched in the seizure-onset zone by +0.15 above the base rate (p = 0.0004). However, sham pre-onset windows had at least one active pair in 50% of seizures. In post-hoc anchor sensitivity analyses, median D10 was 5.0 s from the second active pair and 2.5 s from the third; from clinical onset it was 3.0 s. The findings suggest a short detection window for responsive stimulation, but are limited by one-center sampling, 1-s resolution, spatial sampling of SEEG, a conventional 10% threshold, and post-hoc multi-pair anchors. The recordings were read from the live OpenNeuro S3 mirror; ETags and source keys are preserved, but the exact immutable dataset release is uncertain.

SEEG · focal seizures · seizure propagation · responsive stimulation · high-frequency power · ictal onset · seizure-onset zone · brain-computer interfaces