In 19 qualifying SEEG trains, the neural-assumed component did not decrease significantly more than the artifact component

The alternating-polarity method separates pulse-locked SEEG into a polarity-invariant component (S), assumed to be neural, and a polarity-flipping component (D), assumed to be stimulation artifact. If S is neural, it should adapt during a train while D remains stable. We tested this prediction in public OpenNeuro ds006519 and ds005169 data. From 12 ds006519 recordings and 36 of 112 ds005169 recordings processed, 19 trains from 8 dataset-by-subject patients met the prespecified alternation and duration rules. The mean patient-level S ratio (last/first) was 0.925 and the D ratio was 0.973. Their primary contrast was not significant (log difference -0.051, 95% CI -0.122 to 0.023, exact sign-flip p = 0.27). S amplitude correlated with D amplitude across channels (Spearman rho +0.42, 95% CI 0.23 to 0.58, p = 0.016), consistent with part of S scaling with artifact. Secondary subsets selected on initial S showed opposite changes, but regression to the mean explains this pattern; they do not demonstrate adaptation. The result is limited by 8 patients, a simple polarity classifier that excluded most trains, incomplete processing of ds005169, possible patient overlap between datasets, no amplifier saturation check, pulse timing uncertainty of +/- 1 sample (0.24 ms), and no physiological reference. A decrease in S would be necessary but not sufficient evidence of neural origin.

stereo-EEG · intracranial stimulation · alternating polarity · stimulation artifact · neural signal · adaptation · signal separation