A paired test with five early pulses did not establish lower SEEG prediction error for voltage selection than shape selection in eight patients

We compared two rules that select one SEEG recording pair at each stimulation site. The voltage rule minimized relative squared voltage error. The shape rule maximized waveform correlation. Both used five early good pulses. We tested the choices on five later good pulses in the same session. OpenNeuro ds004696 version 1.0.1 supplied 84 sites with a readout choice in eight patients. The geometric patient error ratio was 1.101 for voltage selection divided by shape selection. The descriptive 95 percent patient bootstrap interval was 0.992 to 1.314. The paired two-sided sign-flip probability was 0.28125. The test did not establish lower error for voltage selection or equivalence between rules. The rules chose the same pair at 74 of 84 sites. The package preserves exact products for 886 candidate combinations, pulse scores, three figures, code, source hashes, and retrieval instructions. A complete sub-05 source derivative passed direct product and pulse-score checks. The other patients use hash-checked archived products with prior extraction records. The scope is clinical limbic data and same-session prediction, conditional on retrospective source quality labels. Online control and stimulation effects were not tested.

SEEG · brain-computer interface · readout selection · voltage prediction · waveform correlation · pulse response · calibration · intracranial EEG