Waveform means gave higher mean RMD AUROC than score means with four pulses for withheld shafts in five offline SEEG sessions at 20–100 ms

We tested two ways to combine repeated SEEG responses to electrical stimulation. We used OpenNeuro ds004457 version 1.0.2. The test had five patients, 183 sites, and 38 withheld-shaft folds. Each fold excluded a complete stimulation shaft from model fitting. Both rules used the same four test pulses and the same fixed relative Mahalanobis distance model. We compared the score of the mean waveform with the mean of the four pulse scores. At 20 to 100 ms, mean AUROC was 0.756 and 0.669, respectively. The paired patient difference was 0.087. All five patient means improved. The descriptive 95% patient bootstrap interval was 0.055 to 0.116. The exact two-sided sign-flip p value was 0.0625. Four pulses spanned a median of 14.2 seconds. Three groups spanned more than 30 seconds. An unplanned check without these groups gave a difference of 0.086. We also proved an exact identity for the two score rules. The package has three figures, source code, derived data, models, group scores, and checked source hashes. This exploratory test reused a prior cohort. It did not test a live decision threshold or a future session.

stereoelectroencephalography · electrical stimulation · waveform averaging · relative Mahalanobis distance · AUROC · brain-computer interface · withheld-shaft validation