Aperiodic-corrected sigma power did not separate SOZ from normal SEEG pairs in 40 patients

Question: do seizure-onset-zone (SOZ) contacts show a change in sleep-spindle-band (sigma, 11-16 Hz) oscillatory power in NREM sleep? Data: OpenNeuro ds007118, ds007119 and ds007120 (snapshot 1.0.0), one run per patient, seconds 150-270, within-shaft bipolar depth pairs, clinical channel labels. Each EDF byte range has a SHA-256. Method: Welch spectra. An aperiodic line was fitted in log-log space on 5-9 Hz and 18-30 Hz. sigma_excess is the mean residual over 11-16 Hz. Per-patient Mann-Whitney AUC, then group mean AUC with bootstrap CI and sign-flip test. The plan was registered before any sigma value was calculated. Result: 160 SOZ, 160 spiking and 408 normal pairs in 43 patients. Primary test SOZ vs normal: mean AUC 0.517 (95% CI 0.415-0.618, p = 0.746, 40 patients). Spiking vs normal: 0.432 (p = 0.24). SOZ vs spiking: 0.535 (p = 0.58). Spike-epoch removal, slow and fast sub-bands, same-lobe controls and a residual after slope and distance were also null. Raw sigma power was much higher at SOZ pairs (AUC 0.763, p < 1e-5). Flank power at 5-9 and 18-30 Hz gave the same AUC (0.762, exploratory). Thus the raw sigma difference is a broadband offset, not a spindle-peak change. Limits: clinical labels, 120 s per patient, N2 and N3 not separated, no spindle event detector, about half of normal pairs have a clear sigma peak.

sleep · spindles · sigma power · intracranial EEG · SEEG · seizure-onset zone · epilepsy · aperiodic activity