In 28 SEEG patients, adaptive mapping did not improve 3-site coverage; at 5 sites, known coverage drove the gain

We tested whether a stimulation-site order that uses measured N1 maps reaches more contacts than a static distance-only order in archived N1 tables from 28 SEEG patients in three OpenNeuro CCEP datasets. The analysis was offline and included clinically tested sites, recording contacts outside white matter (SEEG contacts only for ds004080), and distances of 10–80 mm. The prespecified primary endpoint, coverage at three sites for ADAPT minus MODEL, was +0.019 (6 better, 4 worse; two-sided patient sign-flip p = 0.21). At five sites the difference was +0.068 (14/2; Holm p = 0.043), while FEEDBACK alone, which excludes already-covered contacts from the score, contributed +0.052; the additional model update contributed +0.015 (p = 0.33). The adaptive order did not reduce the number of sites needed for 90% coverage (mean ADAPT−MODEL difference −0.006, p = 0.76). Near-site map weights rose with more measured sites, but this did not improve the order. N1 rules, stimulation currents, and electrode types differed among datasets; the analysis used only tested sites, and model weights and kernel width were not tuned. The five-site result was secondary; the primary endpoint was not significant.

CCEP · SEEG · electrical stimulation · N1 response · brain-computer interface · coverage · adaptive mapping