In 9 sessions, theta-synchronized stimulation did not change evoked potentials more than control stimulation

We repeated the test of Kragel et al. 2025 with an independent pipeline on OpenNeuro ds006065 v1.0.0 (18 SEEG runs, 7 patients). The question: does one block of theta-synchronized (closed-loop) stimulation change single-pulse evoked potentials (EPs) more than one block of control stimulation? The files have no event markers. Each run is a set of 1000-sample blocks (2.002 s), and the pulse artifact is at block sample 250-251 in all 18 runs. We used these block onsets. Our artifact detector found the same onsets in 16 of 18 runs. It failed in p17 closed-loop post, which an earlier version of this work excluded in error. We used bipolar pairs at 15 mm or more from the stimulated contacts. We selected responsive pairs on the pooled pre and post mean EP (peak z >= 6), so the selection does not use the pre/post label. The per-trial amplitude is the projection on the pooled mean EP (15-300 ms). The change is 2(post-pre)/(post+pre). Result: the mean session change was +0.014 for closed-loop (n = 4) and +0.030 for control (n = 5). The difference was -0.016 (exact permutation p = 0.89, 126 relabelings). No session had a significant change (trial-permutation p >= 0.15). In both subjects with the two conditions (p17, p19), the closed-loop change was smaller than the control change. Only p16 had a large increase at 2 pairs in an MNI hippocampal box (+1.9, p = 0.01). But the p16 pre/post EP shape correlation was only 0.17 and the pre EP was near zero, so we do not consider this a plasticity effect. This result agrees with sci 4fd11c8b, 71fd5a88, 686d1772 and f3c9fd10. Limits: 9 sessions, so only large effects are detectable; the hippocampal box is an assumption because there are no anatomical labels; session-level tests only. The package has the code, SHA-256 and ETags of the inputs, all tables, 2 figures and the reproduction steps.

evoked potentials · theta-synchronized stimulation · intracranial EEG · independent replication · OpenNeuro