Pulse pooling gave lower observed mean normalized SEEG error at 20–500 ms after contact order reversal in eight recordings

We tested pulse pooling for response calibration after the stimulation contact order changed. We used public STReEF SEEG data from eight people, with 96 conditions and 6,803 readout instances. Three new pulses formed a template. Two later pulses were fixed test targets. Adding five earlier pulses reduced mean normalized error from 1.232 to 1.133 at 20 to 500 ms. The paired mean change was -0.100. The descriptive 95% person bootstrap interval was [-0.176, -0.015]. The exact two-sided person sign-flip test gave p = 0.0703. This result does not establish a population benefit. The mean absolute voltage error increased. In 668 readouts selected by earlier repeatability, pooling increased mean normalized error from 0.346 to 0.457. That secondary change was also uncertain. Before-pulse and fixed half-gain controls limit an interpretation based on stable response transfer. We verified all 960 public raw pulse epochs and 272,120 direct waveform forecast errors. The package contains three figures, source code, numerical inputs, results, a fixed plan, source versions and hashes, and reproduction instructions. Contact order and time changed together. The test does not establish a polarity cause or live BCI benefit.

stereoelectroencephalography · brain-computer interface · response calibration · pulse pooling · forecast error · intracranial stimulation