Four local pulse scores gave a 2.3 percent higher mean upper limit score than donor scores for 20 to 500 ms SEEG responses at an 80 percent target in eight patients

This test compared four local calibration pulse scores with pooled scores from other patients. It used the same first-five-pulse response model, an 80 percent target, and 268 later pulses at 86 selected SEEG readouts in eight patients from OpenNeuro ds004696 version 1.0.1. For the 20 to 500 ms window, the mean upper limit score was 2.262 locally and 2.211 with donors. Lower scores are better. The local score was 2.3 percent higher. The local minus donor difference was 0.051, with a descriptive 95 percent patient bootstrap range of -0.195 to 0.250 after donor limits were refitted. Whole waveform coverage was 80.8 percent locally and 81.9 percent with donors. The range includes gains for either method. This reused cohort does not give clear evidence of a local gain or show equivalence. The rank proof shows that this nonrandomized method needs four scores for a finite 80 percent limit and nine for a finite 90 percent limit, under score exchangeability. The package preserves code, data, provenance, figures, results, checks, and a reusable limit tool. This is an offline research test. It does not establish live BCI performance or clinical safety.

SEEG · electrical stimulation · whole waveform limits · split conformal inference · calibration · OpenNeuro · epilepsy