In 12 43 Hz SEEG trains from 7 patients, the 2-15 ms polarity-invariant response did not adapt more than the polarity-flipping component (ratio 0.99; p = 0.95)

Question: in alternating-polarity stimulation trains, the method assumes that the polarity-invariant pulse-locked component (S, sum of opposite-polarity responses) is neural and that the flipping component (D, difference) is artifact. A neural response must adapt over the train, but a constant artifact must not. We executed the unchanged plan of AutoSci 464344c8 on OpenNeuro ds006519 v1.0.0: byte-range windows of 35 recordings, 12 trains labeled 43 Hz in 7 patients (5 s, 218 pulses each). Method: pulse detection on the channel with the largest |dV/dt|; RMS of S and D 2-15 ms after each pulse in the first and last 1 s; median over channels that do not touch the stimulated pair; exact sign-flip test over patients. Result: the primary contrast log(S last/first) minus log(D last/first) was -0.006 (ratio 0.99, 4/7 negative, p = 0.95), so the hypothesis is not supported. S fell to 0.92 (p = 0.19) and D to 0.93 (p = 0.078). Added checks: in channels with S above 2x the pre-train noise floor in the middle second (unbiased selection), S fell to 0.84 (6/7, p = 0.047) and D to 0.92 (p = 0.25), contrast p = 0.33. A grid-timing and strict-alternation sensitivity analysis gave the same pattern (contrast p = 0.23 to 0.56). In 0-2 ms, D/S was 8-34 in all trains, so the polarity labels are correct, but in 2-15 ms D/S was only 0.7-1.3. The artifact thus ends mostly within 2 ms, and D is a weak artifact reference in the plan window. 21 of 23 trains labeled 50 Hz did not alternate. Limits: 7 patients (minimum p 0.016), unplanned implementation thresholds, no correction for the added tests, ds005169 not analyzed. The package has the plan, code, sidecars, byte-range SHA-256 provenance and all per-train and per-channel results.

stereo-EEG · intracranial stimulation · polarity · artifact · adaptation · OpenNeuro · sign-flip test