Ten-pulse 10–90 ms SEEG score thresholds fitted at a 10% source error limit gave 4.3% and 12.6% target false response rates in two recordings

We tested whether a response threshold kept a 10% source false response limit when it moved between two public SEEG recordings. We used OpenNeuro ds004774 version 1.0.0, 63 fixed surface stimulation sites, 630 pulses, six depth readout contacts per recording, and expert DvB labels. The fixed score was the largest absolute ten-pulse mean voltage from 10 to 90 ms, divided by baseline SD. Source negative labels set each threshold. Target labels did not change it. UMCU21 to UMCU25 gave 6 false calls among 138 negative labels (4.35%) and 50.0% sensitivity. The reverse test gave 16 among 127 (12.60%) and 74.6% sensitivity. With source threshold refitting, 95% site bootstrap bounds for false responses were 0.65% to 26.80% and 4.55% to 23.81%. Both include 10%. Starting the response window at 20 ms reversed which direction exceeded 10% in the point estimates. The package includes all eight fixed settings in both directions, three figures, voltage epochs, source pins, labels, predictions, bootstrap draws, code, and repeat instructions. A fresh retrieval checked 630 pinned raw ranges and all 5,034,960 saved voltage values. Separate checks and a clean numerical replay passed. These are expert label errors in two reused clinical recordings. They do not establish a new-patient error rate or live brain-computer interface performance.

SEEG · electrical stimulation · evoked responses · threshold transfer · false response rate · expert labels · site bootstrap