In eight SEEG patients, selected stimulation sites reduced held-out distant N1 responses versus random sites but matched the nearest responsive sites

We assessed whether clinical single-pulse electrical stimulation (SPES) in SEEG can activate a target contact while limiting N1 responses at distant contacts, and whether low off-target response spread transfers to held-out electrode shafts. The analysis used expert-checked N1 labels for 8 patients (459 targets) from OpenNeuro ds005448, with contact coordinates. For each target, a responsive stimulation site was selected on one random half of the electrode shafts and measured on the other half across 200 splits per patient. At a 15 mm radius, the selected site activated a median 0.113 of distant contacts, versus 0.168 for a random responsive site; the selected fraction was lower in all 8 patients (exact two-sided sign-flip p = 0.0078). Zero off-target N1 occurred for 0.30 of targets with selected sites versus 0.14 with random sites. Results were the same at 10 mm and 25 mm. The selected site was close to the nearest responsive site (0.113 versus 0.114; selected better in 6/8 patients, median gain 0.018, p = 0.039). Even with an in-sample oracle choice, a site with zero distant N1 at 15 mm was available for a patient-median 0.22 of targets. Site footprints were unequal (median Gini 0.66); the larger footprint with white-matter contacts was not significant (5/7 patients, p = 0.094). These results suggest that site-specific spread can help rank sites, but clinical SPES was not selective for most targets. Limits include 8 patients from one center, binary N1 labels, sparse DTI contacts, and held-out splits within a single session; target activation was not held out.

SEEG · single-pulse electrical stimulation · brain-computer interface · N1 response · selectivity · off-target activation · electrophysiology