In 15 SEEG patients, pre-stimulus high-gamma activity did not reliably predict movement latency in the planned time-ordered test

We tested if 70-150 Hz SEEG activity in the 0.5 s before the stimulus predicts the movement latency of the same trial. We used the author high-gamma features of 15 patients from OpenNeuro ds006253 (archived in sci ceba42cd). We wrote and hashed the plan before the analysis. Per patient, we removed the effects of trial number (linear and quadratic), motion coherence and stimulus side with a model fit on the first 60 % of trials. A ridge decoder on log channel activity then predicted the residual log latency in the last 40 % of trials. The primary group test (one-sided exact sign-flip on Fisher-z r) gave mean r = 0.069, 10 of 15 patients above 0, p = 0.088. Thus, the planned test did not show a reliable effect. Post-stimulus activity (trials with latency >= 0.6 s) gave mean r = 0.089, p = 0.023 raw and 0.069 after Holm correction. The mean of all channels did not predict latency. The ridge predictions did not reduce test squared error. The planned quadratic drift term extrapolated badly into the late trials. A post hoc model with a linear drift term gave mean r = 0.098, 11 of 15 above 0, p = 0.005. This result is exploratory and needs a new planned test. The package has the code (numpy only), input copies with SHA-256, per-patient results and a figure.

stereo-EEG · high-gamma activity · movement latency · ridge regression · time-ordered validation · session drift · OpenNeuro ds006253