Discharge slowing before seizure offset has a weak, measurement-sensitive patient effect in HUP SEEG

In archived detector outputs for 118 seizures from 32 patients in the HUP SEEG dataset, discharge slowing in the last 20 s before seizure offset showed a moderate within-patient component (ICC(1) = 0.32; bootstrap 95% CI 0.13 to 0.42; permutation p = 0.0002). The corresponding ICC 20–40 s before the end was 0.08 (p = 0.20), and the primary effect remained after regression on seizure duration, baseline noise and initial discharge rate (ICC = 0.38). However, the estimate was sensitive to measurement choices: ICC was 0.13 at threshold K = 3, 0.18 at K = 8, and 0.06 with the second-ranked channel. About 70% or more of seizure-to-seizure variance was within patients. The results support a weak patient component in the archived detector output at the primary setting, but one seizure gives only a weak estimate of a patient's offset slowing. The analysis was limited to one center, an unvalidated detector, one channel per seizure, clinical offset marks, and 2–7 seizures per patient; it does not validate a dynamotype label.

seizure offset · inter-discharge interval · intraclass correlation · stereoelectroencephalography · patient-specificity · epilepsy