In one patient, 12 SEEG contacts identified cortical stimulation sites at 25–31% accuracy after 100 ms

In one patient (OpenNeuro ds003708, sub-01), we tested whether responses recorded at 12 unstimulated SEEG contacts identify which of 37 cortical ECoG sites received a single 6 mA pulse. With a corrected per-trial baseline and exact leave-one-trial-out template matching, top-1 accuracy was 0.376 over 10–300 ms (chance 0.027). The 10–30 ms window gave 0.431 accuracy, while windows beginning after 100 ms gave 0.25–0.31, about ten times chance. Without baseline correction, the pre-pulse control classified sites at 0.074 (permutation p=0.01); after correction it was at chance (0.011, p=0.99), identifying a leak caused by a baseline-index error in the earlier analysis. Four contacts on one shaft gave 0.22–0.25 accuracy; single contacts gave 0.06–0.20 without permutation tests. Responses were more similar for nearby stimulation sites (Spearman rho=-0.21; Mantel p=0.001), and wrong choices were nearer than random wrong sites (mean 34 mm versus 41 mm). The analysis used 378 good trials, 37 sites, and 12 contacts. The late response result is less likely to reflect the early stimulation artifact, but slow amplifier recovery is not excluded; the early window does not separate artifact from N1. Pulses at each site were consecutive, so post-pulse state changes may still covary with site. This is one-patient evidence and does not establish generalization.

stereo-EEG · electrical stimulation · cortico-cortical evoked potentials · site identification · template matching · baseline correction · single-patient study