In 5 SEEG patients, stimulus-locked and movement-locked templates explained little single-trial activity

We compared stimulus-locked and mouse-movement-locked mean templates as predictors of held-out single-trial 70–150 Hz activity in 50-ms bins from five patients in OpenNeuro ds006253. We used the archived binned epochs and behavior files from contribution 8d5572e9-7dd0-44e5-bb81-6e57581c5c8a, with movement latencies of 0.3–1.0 s, five time-ordered folds, and a planned responsive-channel threshold of R² ≥ 0.05. Only 4 of 396 bipolar channels in 2 patients met the threshold: 2 were stimulus-locked in sub-02 and 2 movement-locked in sub-23. The maximum R² was 0.10. The planned patient-level sign-flip test of median movement-locked minus stimulus-locked R² was not significant (p = 1.0); at threshold 0.02, patient medians ranged from −0.032 to +0.062 (mean 0.0002, p = 1.0). A latency-shuffle control was significant in 47 channels, versus about 20 expected by chance, indicating movement-timing information despite small per-channel template effects. The results do not establish a general stimulus-locking or movement-locking rule. They suggest that movement-locking selection would find few channels in these implants. Limits include selection of 5 of 22 patients by file size, the 0.3–1.0 s latency range, linear mean templates without per-trial amplitude scaling, and 50-ms resolution that can penalize timing jitter.

stereo-EEG · SEEG · brain-computer interface · high-gamma activity · movement onset · stimulus onset · single-trial prediction · mean templates · random-dot motion