In 5 SEEG patients (OpenNeuro ds006253), 0.2 s high-gamma features do not predict mouse speed better than a template of time from movement onset (mean dR2 = +0.0002, p = 0.44)

We decoded mouse speed in 0.2 s bins from -1 s to +1 s around movement onset. The input was 70-150 Hz high-gamma activity (HGA) in 5 SEEG patients of OpenNeuro ds006253. The speed data come from the velocity derivatives of the dataset. Earlier sci packages did not use these files. We verified their layout: about 60 Hz from stimulus onset to the response click. We wrote the plan and recorded its SHA-256 before we computed results. Training used the first 60% of trials in time order. Testing used the last 40%. A template that knows only the time from movement onset gave test R2 = 0.27 to 0.70. Ridge regression on HGA alone gave R2 = -0.03 to 0.10. HGA added to the template did not increase R2: mean dR2 = +0.0002, exact sign-flip p = 0.44. The primary hypothesis is not supported. One patient had a small significant gain (dR2 = +0.003, permutation p = 0.001). The within-trial shape correlation was small (mean r = 0.04, p = 0.06). We excluded two patients before analysis because their velocity rows did not match the HGA trials. Result for BCI: in this task, the apparent success of continuous speed decoding comes from movement timing. Compare speed decoders against a timing template. Limits: 5 patients and 0.2 s mean HGA features only. The package contains code, data provenance, the plan, results and a figure.

SEEG · high-gamma activity · mouse speed · movement onset · neural decoding · ridge regression · brain-computer interface · OpenNeuro