Polarity-label errors produced distance gradients in a 10-session SEEG stimulation analysis

This method demonstration tested whether alternating-polarity stimulation pulses can separate neural responses from stimulation artefact in stereo-electroencephalography (SEEG) recordings. We processed 10 sessions from OpenNeuro ds005169 v1.0.0: one 50 Hz negative control and nine sessions selected from the 43.2 Hz protocol. Only two sessions met both quality criteria (measured rate 43.2 Hz and polarity-label alternation 1.00). In these sessions, 59% and 26% of valid channels had p<0.01, while response z did not correlate with distance (Spearman rho -0.09 in both); artefact RMS decreased with distance (rho -0.54 and -0.85). Across the eight flagged protocol sessions, response z decreased with distance (rho -0.50 to -0.82), with gradients similar to those of artefact, consistent with leakage into the neural estimate when labels were wrong. The 50 Hz control showed z values from about -5 to +7, including many negative values, consistent with mains noise changing the null. Only two clean sessions from two subjects were available, so these results support no population claim. The distance-independent component in clean sessions is unresolved. The findings show that a distance gradient alone is not evidence of neural spread unless pulse rate and polarity labels are verified.

stereo-electroencephalography · intracranial EEG · electrical stimulation · stimulation artefact · polarity alternation · pulse-locked response · neural recording · signal processing