No consistent recovery emerged within 8 s after 43–50 Hz stimulation trains in 30 SEEG recordings

We measured post-train changes in band power and line length in 30 stereo-EEG recordings from 18 patients in OpenNeuro ds006519 v1.0.0. The recordings contained approximately 5 s, 43.2 or 50 Hz stimulation trains that evoked negative motor responses. Train timing came from the stimulation artifact. For non-stimulated bipolar channels, we calculated five log band-power features and line length in 250 ms windows from 0.05 to 8.05 s after the last pulse. We marked windows with |z| > 3 relative to the pre-train baseline as deviant and subtracted a leave-one-out baseline deviant rate. Near channels (<20 mm) had a mean excess deviant fraction of 0.03–0.07 in all epochs, while subject medians were 0–0.02; 4–11 of 18 subjects contributed to the mean excess. Excess did not decrease monotonically with time, and late line-length deviations in far channels (>40 mm) were as large as those near the stimulation site. No feature-by-distance-by-epoch test remained significant after Holm correction over 90 tests (smallest corrected p = 0.34; smallest raw p = 0.0037). Near-channel recovery times varied widely: 19–48% of recordings recovered within 1 s, while 7–33% had not recovered within 8 s. Recovery time did not correlate with stimulation current (|r| ≤ 0.23, p ≥ 0.24; 27 recordings with near channels). These findings do not support one fixed blanking delay for all patients and sites. There was no sham control, and the 8–13 s baseline did not model slow drift.

stereo-EEG · SEEG · brain-computer interface · electrical stimulation · post-stimulation recovery · line length · band power · signal validity