Fixed pre-pulse power weights did not show lower normalized SEEG mean forecast error at 20 to 500 ms with five calibration pulses in eight sessions

We tested whether pre-pulse voltage power can improve a five-pulse SEEG response estimate. The public source was OpenNeuro ds004696 version 1.0.1. We used eight sessions, 86 stimulation sites, and 886 readout pairs from an existing compact dataset. Five early good pulses set fixed weights. The final five good pulses supplied the mean waveform target. The primary window was 20 to 500 ms. Each error was divided by early pulse energy. The mean patient error change was +1.49 percent relative to the equal mean. The descriptive 95 percent patient bootstrap interval was -0.62 to +3.63 percent. Five of eight sessions had higher error. The rule did not establish a gain. An exact identity separated assignment cost from matching gain. The cost was 3.85 percent, and the gain was 2.36 percent. A gain over arbitrary weight assignments was therefore insufficient to beat the equal mean. Independent checks covered 850,560 assignments. Two complete source files were retrieved and checked anew. The package preserves exact inputs, source hashes, all results, four figures, and methods. GPL analysis and extraction scripts are excluded, so the archived results cannot be rerun directly from the package. This was an exploratory test in a reused clinical cohort. It did not test a live brain interface or clinical benefit.

SEEG · electrical stimulation · pulse response · pre-pulse power · waveform forecast · weighting · calibration · bootstrap · sign-flip test · intracranial EEG