Early calibration from other people selected strength 1 for SEEG forecasts from five pulses at 20 to 500 ms in eight people

We tested whether early SEEG pulses from other people could tune a fixed noise correction rule. The source was OpenNeuro ds004696 1.0.1. Each forecast used the first five pulses of a stimulation condition. Every later pulse was a test target. The main test used 317 conditions, eight people, and 1,933 later pulses at 20 to 500 ms. Each target person used only early pulses from the seven other people to select strength. All eight donor fits selected strength 1 from ten fixed candidates. This was the fixed comparator, so mean error improvement was exactly 0.000%. The central 95% range from 20,000 whole-person refits was -0.332% to 0.037%. It is a descriptive range. Other fixed windows and scoring checks showed no clear gain. The result does not establish the best continuous strength or a BCI control benefit. The package contains code, compact pulse products, all candidate scores, source pins, retrieval instructions, and four figures. A new check of all 64 selected source channels matched all 4,732 pulse matrices exactly. A separate run reproduced all numeric results.

SEEG · brain-computer interface · stimulation response · calibration · noise correction · voltage forecast