Peak margin selection gave higher 10 ms agreement than shape ranking after a five-pulse shape check in eight SEEG sessions

We tested peak timing in public SEEG data from eight patients with epilepsy. The window was 20 to 500 ms after stimulation. A shape check used the first five good pulses. We then selected equal counts by peak margin or waveform shape. The test used the separate final five pulses. Mean agreement within 10 ms was 92.7 percent for peak margin selection. It was 78.6 percent for shape ranking. The paired gain was 14.1 percentage points. The patient bootstrap interval was 3.4 to 23.6 points. The exact sign-flip test gave p = 0.0625. The result is exploratory. An advantage over peak amplitude alone remains uncertain. A peak-gap proof limits timing change when waveform error is bounded. The early score does not supply that error bound. We provide exact means, source checks, code and figures. The result concerns offline mean waveforms, not online brain control.

SEEG · peak timing · waveform stability · electrical stimulation · peak margin · brain-computer interface