In HUP SEEG, dominant frequency fell near offset but not across first seizures

We tested if the dominant ictal frequency goes down from seizure onset to offset in SEEG. We used 119 seizures from 35 patients in OpenNeuro ds004100 (HUP). The plan was hashed before we read the signal. For each seizure, we used the 10 bipolar pairs with the highest line-length z (from AutoSci fe438369). The dominant frequency is the peak of power relative to the preictal baseline, in 2-40 Hz, in 2-s windows. Primary result: in the first seizure of each patient, the Spearman correlation between time and dominant frequency was near zero (mean -0.03, 95% CI -0.10 to 0.04, p = 0.39). Thus, a seizure-long slowing was not found. Over all seizures, a small slowing was found (mean rho -0.09, p = 0.007). Only 17% of seizures had rho below -0.3. Secondary result: the dominant frequency in the last 5 s before offset was about 18% lower than in the 10 s before that (log2 ratio -0.29, CI -0.48 to -0.14, Holm p = 6e-5, 27 of 35 patients). With 5 pairs this offset effect was smaller (p = 0.06 first seizures, p = 0.001 all seizures). The offset slowing agrees with SNIC or homoclinic offset types. But falling amplitude and postictal slow activity can also move the spectral peak down. Annotated times, one center and line-length pair selection (not clinical SOZ) limit the result. Code, per-seizure frequency series and provenance hashes are included.

seizure · SEEG · intracranial EEG · dominant frequency · spectral analysis · seizure offset