In 58-subject NREM SEEG, 11–16 Hz event ranks were stable across two 120 s segments (median rho 0.65), but 20–25 Hz ranks were equally stable and spiking/SOZ pairs had more events

Question: for a closed-loop spindle stimulator, does the SEEG pair with the most spindle-band events in one NREM segment also have the most events later? Data: OpenNeuro ds007118, ds007119, ds007120 v1.0.0, 58 subjects, 1119 adjacent bipolar pairs, segment A at 20% and segment B at 60% of the run, 120 s each. Method (pre-registered): 11-16 Hz Hilbert envelope, threshold 3 x pair median, 0.5-3 s events. Result: per-subject Spearman rho of density A vs B had median 0.65 (95% CI 0.55-0.71), 51 of 58 positive, Wilcoxon p = 1.3e-8. The A-best pair had median percentile 0.73, 0.79 and 0.83 in B with 30, 60 and 120 s of calibration (random 0.50). It gave 75%, 81% and 90% of the best B density (random 50%). Limits: a 20-25 Hz control band, added after the plan, was equally stable (median rho 0.68, difference p = 0.10). Pairs with spiking or SOZ contacts had more events (median AUC 0.68, p = 1.3e-4). Thus the stable map is partly a channel property and partly epileptiform, not only spindles. There was no sleep staging inside runs and no visual spindle check.

SEEG · sleep · NREM · spindles · event density · bipolar electrodes · target selection · epileptiform activity · closed-loop stimulation