Eight channels selected from five early reference peaks matched later voltage landmarks in 21 SEEG records with stable references from all channels

We tested channel selection for a small voltage monitor during brain stimulation. We used public SEEG data from OpenNeuro ds006519 version 1.0.0. Five early peaks from a reference that used all channels supplied channel strengths. We tested later peaks. Selected eight channels matched the final voltage landmark in all 21 primary records from 14 patients at 4096 Hz. The first eight channels in header order matched in 20 records. The equal-patient difference was 2.38 percentage points, with a descriptive 95 percent bootstrap interval of 0 to 7.14 points. The exact sign-flip test gave p = 1.00. The primary comparison does not establish a general advantage for selection. One selected channel also matched all 21 targets in a fixed sensitivity test. The reference came from an offline audit and was stable under its detector checks. The data lack hardware pulse times. These results do not validate train-stop timing or neural recovery. The package preserves exact scores, figures, source hashes, code, and a raw-data retrieval recipe. A fresh extraction reproduced 31 compact files exactly. A separate implementation reproduced 124 peak and run results.

stereoelectroencephalography · SEEG · brain-computer interface · BCI · electrical stimulation · voltage monitoring · channel selection · signal processing