A gap gate kept 24 of 160 SEEG response labels correct across a simulated ±9.77 ms grid in eight patients

We tested output rejection in eight public SEEG patients from OpenNeuro ds005448 version 1.0.0. The reader used four fixed bipolar pairs and four stimulation condition labels. Early responses set the model. Early windows before pulses set fixed gate thresholds. Later data supplied 160 responses and 160 paired controls. A gap gate kept 24 response labels correct and accepted at every error in a 41-point native-sample grid from minus 9.77 to plus 9.77 ms. It accepted 23 controls at one or more errors. An absolute distance gate kept no response labels correct and accepted across the full grid, and accepted 31 controls. Both gates met the same early empirical 10 percent control target. The reader without a gate had 74 response labels correct across the grid. Thus the gap gate lost 50 of those labels. The test did not establish a lower false-output rate for the gap gate across patients. The result uses a small, reused offline cohort. It does not establish live brain interface performance. The package preserves three figures, native voltage crops, all models and decisions, source checks, code, and replay instructions.

SEEG · stereoelectroencephalography · brain stimulation · response detection · output rejection · timing error · calibration · robustness