An early SEEG gate had a 2.44-point mean utility gain with a 95% interval across zero at a 0.05 delay cost in five sessions

We tested whether an early score can decide when to wait for a later SEEG response. We used public OpenNeuro ds004457 version 1.0.2, with 138 remote stimulation sites and 552 test pulses in five patients. Four early pulses per site fitted the reader, the late score weight, and the gate. Four later pulses per site tested them. At a delay cost of 0.05 for 400 ms, mean accuracy was 30.35% with the gate and 28.84% with a calibration-selected fixed rule. Mean nominal sample cutoff was 185.42 ms and 260.00 ms. The mean utility gain was 2.44 points, with a descriptive 95% patient interval from -0.65 to 7.50 points. The exact patient sign-flip p value was 0.75. The gain over always reading early was only 0.30 points. One patient drove most of the gain. The package supplies a two-stage reader, source voltage bins, fitted models, predictions, and three figures. Fresh source extraction reproduced all 2,453,000 voltage bins exactly. This is an offline test within each site sequence. It does not show a reliable population gain or measured system latency.

stereoelectroencephalography · SEEG · electrical stimulation · brain-computer interface · adaptive readout · decision gate · covariance cosine · OpenNeuro