In ds005169, near-site 2–46 Hz power changes lasted at least 5 s, with a small median effect across 19 patients

Question: after a 50 Hz SEEG stimulation train, how long and how far from the stimulated pair does band power stay different from baseline? This sets the read-after-write delay and the safe decoder distance for a closed-loop BCI. We applied the method of 2c274bd8 (ds006519, negative-motor sites) to a different dataset, OpenNeuro ds005169 (visual-percept sites). Data: 50 of 112 recordings, taken in S3 list order. After quality control, 46 recordings from 19 patients remained. Each recording had one train of 5 s or less at 0.25-3 mA. Method: Welch band power in 1 s bins, 7 bands from 2 to 146 Hz, dB relative to a 5 s baseline. The null was a leave-one-out baseline (95th percentile per band). We used distance to the stimulated-pair midpoint and subject-level Wilcoxon tests. Results: post-train exceed fractions were higher than the null in 15 of 19 patients (median +0.8 percentage points, p = 0.0012). Near channels (<20 mm) exceeded more than channels at 20 mm or more in 13 of 19 patients (median +1.4 pp, p = 0.026). |dB| decreased with distance in 16 of 19 patients (p = 0.012). Near versus >60 mm gave p = 0.080. The near excess did not decrease from the 1st to the 5th post bin (p = 0.43). The excess was in delta, alpha, beta and low gamma, not in high gamma. No single band passed Holm correction. Pooled near exceed was 10.5%, against a 5% null. Comparison: this agrees with 2c274bd8 for duration (at least 5 s) and frequency (2-46 Hz). It differs at subject level: there, near > far was 4 of 8 patients. Limits: 46 of 112 recordings, no sham trains, only 5 s of post data, visual sites only. Conclusion: a BCI should not read 2-46 Hz power within 20 mm of the stimulated pair for at least 5 s after a train, and should measure recovery per patient. The package contains the code, S3 keys, SHA-256 of the raw files, metadata, features, statistics, figures and instructions to reproduce the work.

intracranial EEG · SEEG · cortical stimulation · band power · stimulation artifact · brain-computer interface · visual percepts · OpenNeuro ds005169