Alpha-power maps were stable across six 2-s windows and matched full-night STE HFO maps in 18 pediatric SEEG recordings

We read six 2-s windows (at 10-90% of the recording) from each of 18 SEEG patients in OpenNeuro ds005398 with HTTP range reads (65.5 MB). For each window, we calculated per-channel log band power and the within-patient AUC for SOZ vs non-SOZ channels (1285 channels, 181 SOZ). Alpha power (8-13 Hz, the pre-specified band) gave a median AUC of 0.73-0.78 in each window. 104 of 108 patient-windows had AUC > 0.5, and 16 of 18 patients had AUC > 0.5 in all windows. The median Spearman rho between window maps was 0.86 (gamma and ripple: 0.93). The 6-window mean (12 s) gave a median AUC of 0.79 (17/18, sign-flip p = 1e-4). This is equal to the full-night STE HFO-rate map (0.80, paired p = 0.71) and better than the MNI HFO-rate map (0.52, p = 0.001). Power plus STE rank gave 0.82, better than STE alone (14/18, p = 0.006) but not significantly better than power alone (p = 0.13). After we removed the STE rate by regression, power still gave AUC 0.69 (p = 1.5e-4). A gray-matter-only control gave the same result. Limits: no sleep staging, clinical SOZ labels, one pediatric center, bipolar montage, region and contact properties not controlled. The package contains the code, per-window features, provenance (ETags, byte hashes), results and figure.

seizure-onset zone · SEEG · alpha power · high-frequency oscillations · sleep · intracranial EEG · map stability