SEEG seizure end times were less spread than start times, but sharper end crossings may explain part of the difference

We measured when bipolar stereo-EEG (SEEG) channels started and stopped seizure activity in the HUP iEEG dataset (OpenNeuro ds004100; 132 of 143 ictal runs, 122 seizures, 35 patients). A channel was active when 1–45 Hz line length was more than 5 robust z above a pre-ictal baseline for at least 5 s. The primary outcome was the log2 ratio of the across-channel interquartile range (IQR) of end times to the IQR of start times, averaged by patient and tested with a sign-flip test. The plan was written before any signal was read. End times were less spread than start times: median IQR 5.3 s versus 11.6 s; patient-mean log2 ratio −0.77 (95% CI −1.42 to −0.08, p = 0.035), with a negative ratio in 24 of 35 patients. More channels ended within 2 s of the median end than started within 2 s of the median start (+0.19, Holm-adjusted p = 0.002). Start and end times had a weak negative rank correlation (rho = −0.11, Holm-adjusted p = 0.036), so early channels did not end early. The result was significant at threshold 3 but not threshold 8 (p = 0.17) or in the first-seizure-only analysis (p = 0.073). The raw IQR difference in seconds was not significant. The z-step at the end crossing was about 1.7 times larger than at the start crossing. Thus, sharper end crossings can improve timing precision and explain part of the smaller end spread; this study cannot separate that effect from more synchronous termination. The six runs not read because of a time limit, clinical channel selection, and lack of correction for multiple seizures per patient beyond patient averaging limit interpretation. Code, features, per-run provenance, and results are included.

stereo-EEG · seizure · epilepsy · seizure termination · line length · channel timing · synchrony · OpenNeuro ds004100