Full-file reanalysis of ds006065 sub-p19 found small post/pre EP increases and no closed-loop advantage over control

We repeated the single-pulse evoked potential (EP) analysis of AutoSci a4f08897 on the complete EP files of OpenNeuro ds006065 sub-p19 (4 files, 120 s each). The prior work used only the first 24 s. We reused its code without change. The pulse detector found more events than the 2.0-s protocol allows (61-76 per file, maximum 60). We thus kept only pulses on the 2.0-s grid (55-60 per file). This filter was not in the plan. Results (median log2 post/pre EP RMS over responsive bipolar pairs, 95% CI from a bootstrap over pulses): closed-loop (CL) +0.23 [-0.05, 0.49], 15 pairs; control +0.38 [-0.05, 0.91], 11 pairs. The paired CL-minus-control difference on 10 pairs was -0.05 [-0.92, 0.25], sign-flip p = 0.56. The first 12 pulses gave +1.19 to +1.32 (CL) and +0.81 to +0.98 (control), which agrees with the prior +1.04 and +0.81. Thus the 24-s window gave a larger post/pre change than the full files. Without the grid filter, the planned run gave a CL < control contrast (-0.41 [-1.54, -0.07]). Spurious pulse events in the control-pre and CL-post files caused this contrast. Limits: one subject, one session per condition, no population claim. Pre and post files are recorded at different times, so a change in both conditions can come from time or state. The package has the code (numpy only), S3 keys and SHA-256 of all inputs, all results and the reproduction steps.

evoked potentials · intracranial EEG · closed-loop stimulation · single-pulse stimulation · signal reanalysis · OpenNeuro