A reset reduced mean detection and background commands in eight SEEG tests with rest training under a limit of six commands per eligible calibration minute

We tested a low score reset in eight public SEEG records from OpenNeuro ds004473 version 1.0.2. Both command rules used the same reader trained with rest examples. Each rule had a separate early threshold calibration. The limit was six background commands per eligible calibration minute. In later tests, mean response detection fell from 13.48% to 10.76%. Mean background commands fell from 2.56 to 1.83 per eligible minute. The paired mean detection change was -2.72 percentage points. Its descriptive patient bootstrap 95% interval was -7.07 to -0.27 points. Detection fell in three patients and was equal in five. The exact two-sided sign-flip p value was 0.25. This small reused cohort gives exploratory evidence. Shared calibration limits did not give equal later rates. The reset did not improve mean detection in this test. The package preserves a verified state replay, all inputs, source records, scores, commands, three figures, and repeat instructions. A full public raw data audit checked all 67,309 feature windows. A separate refit and replay checked the effect of float32 rounding differences.

stereoelectroencephalography · brain-computer interface · command detection · threshold calibration · score reset · background commands · patient bootstrap