Earlier-site variation lowered error divided by first-pulse energy but raised voltage error versus a fixed half-size SEEG forecast from one pulse at 20–500 ms in eight sessions

We tested whether variation from earlier stimulation sites helps forecast SEEG voltage from one pulse at a new site. Public OpenNeuro ds004696 data supplied eight patient sessions, 68 eligible sites, and 688 site-to-readout pairs. Each target had at least two earlier donors. Each donor supplied five pulses before the target first pulse. For centered bipolar voltage at 20–500 ms, the mean patient error reduction from a fixed half-size forecast was 3.48% after division by first-pulse energy. Six patients improved. The descriptive bootstrap interval was 1.00% to 5.91%; the planned exact test gave p = 0.0547. Without an energy divisor, the mean patient error increase was 19.96%. Transfer also had 19.25% higher scaled error than the own five-pulse rule. The gain depends on the error measure. The package preserves a fixed streaming tool, exact products, all scores, figures, source records, and reproduction code.

stereoelectroencephalography · SEEG · neurostimulation · signal processing · brain-computer interfaces · calibration · trial variation · voltage forecasting