Time since the previous key press did not change the rest-window score of an SEEG key-press detector (ds004473, 8 subjects, gaps 3-7.4 s)

Question: do rest windows that start soon after a key press get more move-like scores from an SEEG movement detector? If so, a BCI needs a hold-off time after each detection. Method: we used archived HG and beta features of ds004473 (8 subjects). We trained a shrinkage LDA on trials 1-100 and tested on trials 101+. The pre-registered test was the partial Spearman rho between the gap and the rest-window score, with control for trial index, and an exact sign-flip test over subjects. Result: there was no effect (mean rho -0.013, 3 of 8 positive, p = 0.68, post-hoc 95% interval -0.08 to 0.06). HG-only and beta-only decoders also showed no effect. The rest-vs-move AUC was the same for short and long gaps. The move-window score increased with the press-to-press interval (8 of 8, Holm p = 0.023). A post-hoc test showed that this effect followed the reaction time of the present trial (p = 0.016) and not the gap (p = 0.23). Limits: the gaps were 3.0-7.4 s and the range in each subject was narrow. The data are offline and cue-locked. Thus the result does not apply to gaps shorter than 3 s.

SEEG · brain-computer interface · key-press detection · shrinkage LDA · high-gamma · beta · rest-window score · reaction time