Across 14 alternating-polarity SEEG trains from 9 patients, the 2–15 ms polarity-invariant response did not decrease reliably more than the polarity-flipping component

During bipolar SEEG stimulation with alternating pulse polarity, we separated pulse-locked epochs into a polarity-invariant component S (assumed mostly neural) and a polarity-flipping component D (assumed mostly artifact). We analyzed all 35 recordings in OpenNeuro ds006519 v1.0.0, finding 14 trains (12 at 43.2 Hz and 2 at 50 Hz) that alternated and lasted at least 2.5 s, across 9 patients. The primary measure was A = log(RMS S in the last 1 s / RMS S in the first 1 s) − log(RMS D in the last 1 s / RMS D in the first 1 s), for epochs 2–15 ms after each pulse. Its patient mean was −0.268 (exact sign-flip test, one-sided p = 0.096; two-sided p = 0.19), so the decision rule in the plan classified the result as undetermined. With symmetric channel inclusion to avoid regression-to-the-mean bias, A was −0.102 (p = 0.31). D was stable. Across trains, the median fraction of channels with S above three times the pre-train noise floor was 4.9%; log S had a weak relation to log D (r = 0.22). The direction was negative in most analyses, but the effect was small and not reliable. The only p < 0.05 result was in the 43.2 Hz subgroup with planned inclusion; the symmetric rule removed it (p = 0.16). Limits include 9 patients, 1–8 included channels per train, use of a pre-train noise floor, and possible polarity-dependent neural response in D or nonlinear artifact in S. OpenNeuro ds005169 was prepared but not analyzed because the download rate was too slow. The package includes source code, per-channel and per-train results, file hashes, deviations from the plan, and reproduction steps.

stereo-EEG · intracranial EEG · electrical stimulation · alternating polarity · pulse-locked response · neural adaptation · stimulation artifact · signal decomposition