Peak confirmation had lower response label AUC point estimates than a peak from ten pulses at 10–90 ms in two SEEG recordings

We tested whether a peak selected from five early stimulation pulses and measured at the same time in five later pulses could rank expert response labels better than a peak from all ten pulses. We used OpenNeuro ds004774 version 1.0.0. The test covered two SEEG recordings, 63 surface stimulation sites, six depth contacts per recording, 630 pulses, and 566 expert judgments on 378 unique cells. In the fixed 10–90 ms window, the response label area under the receiver operating characteristic curve (AUC) fell from 0.905 to 0.870, from 0.767 to 0.755, and from 0.892 to 0.857. Paired 95% site intervals for the differences were -0.063 to -0.010, -0.051 to +0.028, and -0.093 to +0.014. A fixed confirmation gate reduced false positive rates and sensitivity. Many no-response cells kept the early peak sign in later pulses. Thus, repeatability alone did not reproduce expert labels. This result is limited to these recordings, labels, pulse split, source reference, and window. It does not test a live brain-computer interface. The package includes code, raw SEEG epochs, source annotations, version-pinned retrieval records, all scores and draws, four figures, and reproduction instructions. Fresh checks verified all 630 raw byte ranges and reproduced the included arrays exactly.

stereoelectroencephalography · evoked responses · electrical stimulation · peak confirmation · area under the curve · brain-computer interfaces