A shared early response update gave no clear 100–500 ms SEEG forecast gain after five calibration pulses in eight sessions

We tested whether an early SEEG response can improve a later voltage forecast after electrical stimulation. Public OpenNeuro ds004696 v1.0.1 supplied eight patients, 86 sites, and 886 site/readout instances. Five distinct pulses per site fit templates. Five later pulses tested them. At 100 ms, one shared coefficient used only completed calibration responses in the same patient. It updated the centered 100–500 ms mean waveform from the centered 20–100 ms input. Mean patient error reduction from a scaled template baseline was 0.046%. The descriptive 95% patient bootstrap interval was -0.026% to 0.116%. Separate readout updates increased mean error by 0.848% and increased error in all eight sessions. Pre-pulse and shuffled-input controls did not show a clear benefit specific to the early trace. This exploratory test reused a cohort and did not test new patients or a live interface. The package preserves exact native voltage products, three figures, all controls, result tables, source hashes, and method descriptions. GPL analysis and extraction code is excluded, so the exact calculation implementation is not publicly replayable from this package. Two full public source files were rebuilt exactly. Native voltage checks and full numeric replay passed.

stimulation-evoked potentials · SEEG · brain-computer interfaces · voltage forecasting · calibration · time-ordered evaluation