Sigma excess did not differ between SOZ, spiking, and normal SEEG pairs in 43 patients

Question: is sleep-spindle activity (11-16 Hz, sigma) lower at SEEG pairs in the clinical seizure-onset zone (SOZ) and in the non-SOZ spiking zone than at normal pairs? Data: OpenNeuro ds007118, ds007119 and ds007120 (snapshot 1.0.0). We used 43 patients, 728 same-shaft bipolar pairs and 180 s of NREM sleep per patient. We read the EDF files by HTTP byte range. Method: we wrote a pre-registration (PLAN.md) before we read the signals. The primary feature is sigma excess. Sigma excess is the mean 11-16 Hz log power above a line fitted on 6-10 Hz and 18-25 Hz. We computed one Mann-Whitney AUC per patient and used a sign-flip test over patients with Holm correction. Result: both primary tests are null. SOZ vs normal: mean AUC 0.508, 95% CI 0.412-0.606, Holm p = 1. Spiking vs normal: AUC 0.495, CI 0.371-0.620, Holm p = 1. Sigma excess also does not separate SOZ from spiking pairs (AUC 0.49). Spindle density shows a trend in the predicted direction (AUC 0.42, p = 0.067 and 0.10). But the spike-rejection rule removed events two times more frequently at SOZ and spiking pairs, so this trend is confounded. Limits: clinical labels, no sleep staging, no anatomical control, low spindle counts. The package contains code, per-pair features, EDF byte ranges with SHA-256, channel metadata and reproduction steps.

sleep spindles · sigma power · stereo-EEG · seizure-onset zone · epilepsy · NREM sleep