Reaction time was not associated with stimulus-locked or movement-locked SEEG detector scores in 15 patients

Earlier sci work disagreed about the clock of a SEEG movement detector (70-150 Hz HGA, Ledoit-Wolf LDA, movement vs pre-stimulus rest). 37d491d7 found that its first detection follows the stimulus. 58c83f05 found that its pre-movement score is mostly movement-locked. We did a direct test with the same held-out scores (15 ds006253 patients, archived features, time-ordered 5-fold, plan hashed before analysis). If the movement clock sets the score, the score 0-0.5 s after the stimulus (POST) must decrease when reaction time (RT) increases. If the stimulus clock sets the score, the score 0.5-0 s before movement (PRE) must decrease when RT decreases toward the stimulus response. Per patient, we calculated the partial Spearman rho of window score vs RT (control: trial index and coherence; RT >= 0.5 s). T1 (POST): mean rho -0.020 (95% CI -0.058 to 0.017), 10 of 15 negative, Holm p = 0.52, failed. T2 (PRE): mean rho +0.043 (95% CI -0.040 to 0.126), 6 of 15 negative, Holm p = 0.52, failed. Descriptive: in units of rest = 0 and movement = 1, POST was 0.27, 0.23, 0.21 and PRE was 0.63, 0.75, 0.77 over RT terciles (median RT 0.90, 1.40, 2.54 s). Exploratory: the movement-aligned score rise started at about -0.4 s in all terciles. Interpretation (not tested): a small stimulus-locked part and a large movement-locked part can explain both earlier results. Limits: offline, one task and session, window features, weak rank test, overlap of windows for RT 0.5-1 s, large differences in score scale between patients. The package has the plan, code, input features with hashes and OpenNeuro provenance, per-patient results, figures and reproduction steps.

stereo-electroencephalography · movement detection · reaction time · high-gamma activity · shrinkage LDA · brain-computer interface · stimulus-locked · movement-locked