With uniform tie resolution, a fixed HFO rule gave less than 90 percent later agreement in 17 of 18 SEEG sleep records

This study tested whether early cumulative HFO list stability could predict later list agreement. It used released STE detector marks from 18 SEEG sleep records in OpenNeuro ds005398 version 1.1.3. The top 20 percent list used uniform, independent cutoff tie resolution. The fixed rule required three adjacent cumulative agreements of at least 90 percent. All records passed early, at a mean of 13.1 minutes. However, 17 had less than 90 percent agreement with a separate later 30-minute sample. Mean local agreement was 94.4 percent. Mean later agreement was 70.0 percent. The mean gap was 24.4 percentage points; its patient bootstrap interval was 18.91 to 30.30 points. Six fixed sensitivity settings also had positive mean gaps. A proof gives expected chance overlap after any early stopping rule when later channel identities are independent and exchangeable. A diagnostic that removes persistent channel identity still passed early in 85.9 percent of draws. This reference is not a biological null. The result concerns one fixed rule and one reused cohort. It does not separate count noise from neural change or set a clinical recording duration. The package preserves source tables, hashes, code, proof, all draws, figures, and independent verification.

high-frequency oscillations · stereo-EEG · intracranial EEG · epilepsy · stopping rule · channel-map agreement · tie resolution · temporal sampling