Aperiodic-corrected 11–16 Hz power was the most stable NREM SEEG spindle map in 45 subjects

In a prior NREM SEEG study, 11–16 Hz spindle event density was stable between two segments, but a 20–25 Hz control map was equally stable. In this study, we tested whether spindle-band stability remained after controlling for the control-band map and compared event density with aperiodic-corrected power in three OpenNeuro datasets (ds007118, ds007119, ds007120; version 1.0.0). Among 45 subjects with at least six valid bipolar pairs (860 pairs), the median partial Spearman correlation for 11–16 Hz event density between two 120 s segments was 0.44 (95% bootstrap CI 0.28–0.52; 39/44 subjects positive; one-sided Wilcoxon p = 1.9e-6; sign-flip p = 1e-4), controlling for 20–25 Hz event density. Raw spindle-band event density was not more stable than control-band density (median difference -0.04, 15/44 positive, p = 0.34). Aperiodic-corrected 11–16 Hz power was more stable than the corresponding 20–25 Hz control measure (median rho 0.80 vs 0.32; median difference 0.23, 35/43 positive, two-sided p = 1.4e-4). The pair with the highest corrected power in segment A had a median percentile of 0.93 in segment B. Corrected power did not increase at abnormal contacts, while raw event density did (AUC 0.75, p = 0.002). These findings support a spindle-specific stable component and indicate that aperiodic-corrected 11–16 Hz power was the most stable target map in this sample. The two segments came from the same night and run; sleep stage was not checked per epoch, the sample was bandwidth-limited, and FFT filters differed from the planned Butterworth filters.

NREM sleep · SEEG · sleep spindles · signal stability · aperiodic correction · target selection · closed-loop stimulation