Calibration range limits reduced 20 to 500 ms donor SEEG error at penalty one after five fit pulses in eight people, but a fixed template had lower error

We tested recovery of a hidden SEEG shaft after stimulation in eight people from public OpenNeuro ds004696 version 1.0.1. Five early pulses fit each model. All later pulses tested it. A new rule kept each donor score within its five-pulse fit range. At a fixed relative slope penalty of one, equal-person mean normalized error fell from 1.5230 to 1.2326 in the 20 to 500 ms window. The reduction was 19.1 percent, with a descriptive person bootstrap interval of 12.1 to 25.9 percent. Error fell in all eight people. The fixed template had error 1.1731. Clipped error remained 5.1 percent greater than template error. Correct pulse matching gave little benefit over shuffled donor input. The test had 317 conditions and 1,183 condition/readout instances. All 4,732 reused signal matrices matched a fresh raw reconstruction exactly. The package preserves code, matrices, source event maps, model coefficients, pulse errors, three figures, raw retrieval receipts, and checks. This was an offline test of simulated known shaft loss on reused people. It did not test future prediction, actual shaft failure, or a live BCI.

SEEG · intracranial EEG · stimulation response · calibration range · signal reconstruction · brain-computer interface