A normal prediction cutoff from five early controls gave 4.7 percent baseline calls in eight offline SEEG sessions

We tested a fixed cutoff for reading voltage after electrical brain stimulation. Public OpenNeuro ds004696 version 1.0.1 supplied eight SEEG sessions and 86 sites. Five early pulses fit each detector. Five later pulses tested it. A normal prediction cutoff reduced the equal-patient mean baseline call rate from 14.00 to 4.72 percent. The paired reduction was 9.28 percentage points. Its descriptive patient bootstrap interval was 7.31 to 11.13 points. Mean post-pulse calls fell from 88.45 to 78.19 percent. The new baseline rate interval was 3.06 to 6.62 percent. The data did not establish a 5 percent upper limit. A proof showed that this cutoff must exceed the maximum of five controls. Normal simulations checked the prediction rule and a failure after control reuse for selection. All repeated results matched. Two complete public source files passed fresh hash checks. Their selected voltage arrays and all 1680 checked decisions matched. The package preserves exact inputs, code, three figures, a proof, source records, and repeat instructions. This reused cohort supports an offline operating tradeoff. It does not establish neural origin or a live BCI benefit.

SEEG · electrical brain stimulation · normal prediction cutoff · baseline calls · response monitoring · calibration · patient bootstrap · OpenNeuro ds004696