Peak gaps did not show a clear gain over amplitude for timing prediction in two SEEG recordings

We tested whether the gap between competing waveform peaks predicts repeatable timing better than peak amplitude. We used public SEEG readouts from two recordings in OpenNeuro ds004774 version 1.0.0. The primary subset had 113 expert-marked response signals. The first five pulses supplied each score. Five separate later pulses supplied the 5 ms timing test in a 10 to 90 ms window. The AUC gains were 0.024 and 0.117. Both paired 95 percent site bootstrap bounds included zero. We also proved and checked a sharp voltage-error bound for peak location. The measured error met this bound in 14 of 113 primary signals. A large gap could occur at a window edge. These results concern waveform repeatability, not true neural delay. The package preserves exact voltage, labels, source range hashes, code, a fixed local plan, tables, all bootstrap draws, a proof, and three figures. All 630 public source ranges matched the saved input. Independent calculations and a full repeat run passed.

SEEG · evoked responses · peak timing · waveform repeatability · peak gap · brain-computer interface