A 90% nested rule for top 20% STE HFO lists gave 69% mean agreement with later 30-minute samples in 18 SEEG sleep records

We tested a fixed stopping rule on public SEEG HFO marks from OpenNeuro ds005398, version 1.1.3. The rule used top 20% channel lists from the released STE detector. It required 90% agreement for three recent cumulative updates. All 18 sleep records met the rule. Mean internal agreement was 95.8%. Mean agreement with a separate later 30-minute sample was 69.3%. The mean gap was 26.5 percentage points, with a 95% patient bootstrap interval of 20.6 to 33.2 points. Whole-minute order permutations gave a mean gap of 22.3 points. The observed excess was 4.2 points, with an interval of -0.1 to 9.0 points. The primary conditional test did not reject uniform minute order (p = 0.082). This result applies to the tested rule and records. It does not set a clinical recording time. The package contains source bytes, pinned retrieval code, new analysis code, all fixed checks, results, and repeat instructions. Independent checks and full repeats from archived and fresh public inputs passed. The cohort was reused from cited prior work.

high-frequency oscillations · stereoelectroencephalography · intracranial EEG · stopping rule · sample agreement · channel ranking