In 52 subjects, transient rejection removes the apparent excess of sigma events at abnormal SEEG contacts

We measured 11–16 Hz sigma-band events in 120-second NREM sleep SEEG clips from 67 subjects and 1,352 contacts in OpenNeuro ds007118–ds007120 (v1.0.0). Within each subject with both normal and clinically abnormal (spiking or seizure-onset-zone, SOZ) contacts (n=52), event rate without transient rejection was higher at abnormal contacts (median AUC 0.58, 95% CI 0.50–0.65; Wilcoxon p=0.00025). We rejected events coinciding with a 25–45 Hz envelope above three times its channel median. This retained a median 5% of events at abnormal contacts and 29% at normal contacts (n=28; p=0.0018); after rejection, the rate comparison showed no detected difference (median AUC 0.50; p=0.20). Relative sigma power was higher at abnormal contacts (AUC 0.67; p=0.014). These data do not show fewer sigma events at epileptic contacts. They indicate that a simple amplitude-threshold detector mixes spindle-like activity with pathological or artifactual transients, especially at abnormal contacts. For spindle-based sleep decoding or closed-loop stimulation on SEEG, transient rejection or excluding these contacts may be needed. The inference is limited by 120-second clips that were not sleep-staged, low event counts after rejection, an unvalidated heuristic, monopolar reference, and the possibility that rejection also removes genuine spindles. The comparison therefore does not establish equal spindle rates.

SEEG · sleep spindles · sigma band · epilepsy · transient rejection · intracranial EEG