Site templates increased 20–100 ms shaft log scores over shaft means in five SEEG sessions

We tested electrode shaft readers in five public SEEG sessions from OpenNeuro ds004457 version 1.0.2. A shaft label was the shared letter prefix of an adjacent stimulation pair. The site reader kept one waveform template per site and added site probabilities within each shaft. The comparison reader used one mean waveform per shaft. Both fitted means on the first three pulses per site and a probability scale on the fourth pulse. The final four pulses per site gave 732 test pulses across 183 sites. At 20–100 ms, the site reader increased the shaft log score in all five sessions. The patient mean score increased from 0.785 to 1.204 bits per pulse. The mean increase was 0.419 bits, with a descriptive patient bootstrap interval of 0.305–0.557 bits. The exact two-sided sign-flip reference was 0.0625. Mean accuracy increased from 50.08% to 60.23%. The score benefit remained with a shared covariance and at 30–100 ms. Before-pulse scores were near zero. The result applies to the fixed retrospective site pulse split. Tests of a live brain interface remain necessary. The package includes code, compact voltage data, fitted models, all test probabilities, and four figures. A hash-pinned raw rebuild matched all 30 input arrays and five event tables. A clean replay reproduced the results and figures exactly.

stereoelectroencephalography · SEEG · electrical stimulation · shaft classification · waveform templates · probability calibration