In eight SEEG patients, N1 latency rose with distance but geometry poorly predicted held-out patients

We examined whether Euclidean distance from a stimulated electrode pair predicts the latency of the evoked N1 response at a recording contact, and whether a short pulse train measures latency repeatably. In eight patients from OpenNeuro ds005448, we analyzed 2,990 expert-checked N1s from the first SPESclin run. Latency was positively associated with distance in seven of eight patients (median Spearman rho 0.25; Wilcoxon p = 0.023); median Theil-Sen slope was 0.26 ms/mm. Distance explained little variance, and a model trained on seven patients did not improve held-out-patient prediction over a constant (median gain 0.58 ms in five of eight patients; p = 0.20). In a raw-data subset of 613 N1s from eight random stimulated pairs per patient, odd- and even-pulse latency estimates differed by a median 0.98 ms. However, 55% of subset and 62% of all checked expert N1 peaks were positive in the stored re-referenced data: a negative-only search matched expert latency within 1 ms for 31%, while a sign-matched search did so for 71%. The findings suggest that distance has a detectable but weak association with N1 latency, and does not provide useful held-out-patient prediction in this sample. Limits include eight patients, low power for group tests, uncorrected P values, Euclidean distance rather than fibre path length, and post hoc polarity and edge analyses.

SEEG · N1 latency · CCEP · SPES · electrical stimulation · intracranial EEG · distance · repeatability · polarity · held-out prediction