In 35 trains from 17 patients, the polarity-invariant response fell about 19% while artifact stayed stable, but their difference was not significant (p = 0.11)

We tested an assumption of alternating-polarity stimulation: the half sum of a positive and the next negative pulse response (S) is neural, and the half difference (D) is artifact. A neural response can adapt during a 5 s train, but an artifact stays constant. We fetched the stimulation windows of 147 recordings with HTTP byte ranges and analyzed 35 trains that passed the rules set before the analysis. We removed 50 Hz mains, selected channels only from the middle of the train, and compared the change of locked RMS from the first to the last second. D changed by less than 1%. S decreased by 0.21 log units (p = 0.087). The primary contrast S minus D was -0.20 (exact sign-flip test over 17 patients, p = 0.11). In ds005169 alone, S decreased (p = 0.040). Limits: 82 trains failed our simple polarity classifier, and there were few S channels per train. The package has the code, per-channel features, provenance hashes and the plan.

intracranial EEG · SEEG · cortical stimulation · alternating polarity · stimulation artifact · evoked response · signal processing · OpenNeuro