Motion coherence did not change the SEEG movement detector score in 12 patients within the tested range

We trained a shrinkage linear discriminant analysis (LDA) movement-versus-rest detector on 70–150 Hz high-gamma SEEG window means from 12 patients in OpenNeuro ds006253. In a time-ordered split (first 60% of trials for training, last 40% for testing), motion coherence was not associated with the held-out MOVE-window score after adjustment for a cubic trial-order trend and log stimulus-to-movement time (mean partial Spearman ρ = −0.025, 95% CI [−0.103, 0.054], exact sign-flip p = 0.486). The POST-STIM-window result was also small (ρ = −0.034, 95% CI [−0.115, 0.047], p = 0.367); the Holm-corrected p-value was 0.734. A positive control found an association between MOVE score and peak movement speed (ρ = 0.089, 95% CI [0.020, 0.158], p = 0.022). Within this task and tested range, the confidence interval excludes |ρ| > 0.12, suggesting that this detector does not require difficulty-dependent recalibration and does not provide a usable readout of stimulus difficulty. The analysis is offline, uses one time-ordered split, and removes only a cubic trend from the staircase-linked coherence ranks; a null result is not proof of no effect, and a slower non-cubic drift could still matter.

stereo-EEG · high-gamma activity · movement detection · motion coherence · brain-computer interface · shrinkage LDA · partial Spearman correlation · OpenNeuro ds006253