In eight patients, stimulation-site offsets improved prediction of intracranial N1 latency more than distance

We analyzed expert-checked intracranial N1 responses to single-pulse electrical stimulation in eight patients from the STReEF dataset (3,302 N1 labels from 335 stimulated pairs). Distance showed a small association with latency (median Theil–Sen slope 0.25 ms/mm; median r² 0.034, or 0.061 after excluding labels ≥90 ms), but a distance model did not improve held-out prediction for a new patient over a constant. Adding a stimulation-site offset reduced median prediction error to 13.3 ms across all labels and 7.4 ms below 90 ms; receiving-contact offsets did not improve on patient offsets. For high-SNR responses (348 N1s), median single-pulse latency jitter was 1.6 ms (IQR 0.6–3.7 ms), compared with 7.7 ms under a noise-only search. Many labels near the 100 ms detector ceiling lacked a local peak in the trial mean. These results suggest that stimulation-site calibration can improve timing estimates within a patient and that early N1s can be precise from pulse to pulse. The study includes one session per patient and no test–retest data; distance is Euclidean rather than fiber length, and jitter estimates are upper bounds under the stated peak-picking method.

SEEG · N1 latency · cortico-cortical evoked potentials · electrical stimulation · brain-computer interface · latency prediction · single-pulse jitter