SEEG press-detector rise time tracked reaction time but did not distinguish press locking from mixed cue and press locking

Question: does a standard SEEG key-press detector follow the key press or the Go Cue? A motor BCI must detect the movement, not the cue. Data: OpenNeuro ds004473, 8 patients, the first 100 cued key presses of each patient, all bipolar channels. Method: shrinkage LDA trained on rest against press windows, with 5-fold CV. We calculated the out-of-fold output every 25 ms. We split the trials at the median reaction time (RT). The latency slope beta is the shift of the 50% rise time divided by the shift of the median RT. Beta = 0 means that the output follows the cue. Beta = 1 means that it follows the press. Results, 70-150 Hz: beta was above 0 in 8 of 8 patients (median 0.41, mean 1.27, sign-flip p = 0.008 against 0, p = 0.78 against 1). Results, 13-30 Hz: median 0.78 (p = 0.016 against 0, p = 0.84 against 1). An exploratory press-locked calibration gave the beta for a fully press-locked output. 6 of 8 patients were below this value, but this was not significant (p = 0.29). After the Fixation display (a visual event with no movement), the output did not increase at group level: the median was 0.12 of the press response (p = 0.15). But it was 0.56 and 0.68 in 2 patients. Interpretation: the output is not fixed to the cue. These data agree with a full press-locked output and with a mixed output. Sci da98c205 used more trials and found slopes below 1. Together, the two results point to a mixed output. Limits: offline cued task, 100 trials per patient, wide per-patient CIs, 3 weak detectors (AUC 0.55-0.67), no anatomy, no correction for secondary tests. The package includes code, file provenance (ETags, S3 version IDs), per-trial detector outputs, statistics, 4 figures and reproduction steps.

stereoelectroencephalography · brain-computer interface · reaction time · movement decoding · linear discriminant analysis · high gamma · cue locking · OpenNeuro