A 0.2 s movement window was non-inferior to 0.5 s in 15 SEEG patients

In an offline analysis of 15 patients from OpenNeuro ds006253, a 0.2 s high-gamma activity window starting at mouse movement onset had mean test AUC 0.811, compared with 0.833 for the 0.5 s window. The difference was -0.022 (95% bootstrap CI -0.035 to -0.008; 12/15 patients lower). The pre-registered non-inferiority hypothesis was supported at a 0.05 AUC margin (exact one-sided sign-flip p=0.00015; Holm p=0.00031); one patient lost more than 0.05. The difference CI excludes zero, so the shorter window remained slightly lower. A 1.0 s window did not differ from 0.5 s (+0.012, 95% CI -0.013 to +0.035; Holm p=0.375). The descriptive Wolpaw bit-rate upper bound was largest at 0.2 s (1.01 vs 0.49 bits/s at 0.5 s). The pre-registered prediction based on ds004473 failed. Analysis used author 70-150 Hz HGA, shrinkage LDA, and a first-60%/last-40% time split. Results are limited to offline event-locked windows with known movement onset, a fixed 0.5 s rest window, one session per patient, and a bit-rate calculation that counts window time only.

stereo-EEG · movement detection · high-gamma activity · window length · non-inferiority · AUC · shrinkage LDA