In 18 pediatric SEEG patients, 10-minute STE HFO maps approached full-recording SOZ separation, while MNI maps did not

We asked how much sleep SEEG is needed for an HFO-rate channel map to separate clinical seizure-onset zone (SOZ) contacts and how map stability relates to that separation. We used author-supplied HFO detections for all 18 pediatric SEEG subjects in OpenNeuro ds005398 and counted events in non-overlapping windows from 1 to 45 min. With the full recording, STE-detector maps had median within-patient AUC 0.80 (16/18 patients above 0.5; one-sided sign-flip p = 0.0002); events with spike probability > 0.5 had AUC 0.68 (p = 0.008). All-event and MNI maps did not separate SOZ contacts (AUC 0.57, p = 0.15; and 0.52, p = 0.31). STE window AUC was 0.54 at 1 min, 0.73 at 5 min, 0.77 at 10 min, and 0.81 at 45 min; the 10-min value was 97% of the full-recording value. MNI maps had the highest split-half Spearman stability at 1 min (0.87), although they did not separate the SOZ. STE maps had stability below 0.8 at every tested duration. These findings support about 10 min of sleep calibration with an STE-type detector in this dataset and show that map stability alone is not a sufficient stop rule. Limits include one center, 18 patients, clinical SOZ labels, author-generated detections, no sleep-stage control, and an undocumented spike column. We did not download raw signals or run a new HFO detection.

SEEG · high-frequency oscillations · seizure-onset zone · calibration · responsive stimulation · sleep · OpenNeuro ds005398