After line-noise removal, 15 pulses predicted a reliable SEEG CCEP map in 14 patients

We estimated how many single-pulse stimulations are needed for a reliable cortico-cortical evoked potential (CCEP) response map on SEEG depth contacts in 14 patients from UMC Utrecht. For each patient, we analyzed one run and 643 stimulation sites, using contacts at least 10 mm from both stimulated contacts. The median one-pulse split-half map correlation was 0.22 (95% bootstrap CI 0.18–0.37) before line-noise removal and 0.38 after removing 50, 100, and 150 Hz components. Spearman–Brown extrapolation predicted 33 pulses (CI 16–40) for a split-half correlation of 0.9 before removal and 15 after removal. The raw-data estimate was close to an independent-center result of 17 pulses. Removing line noise increased reliability in all 14 patients, so phase-locked mains noise did not inflate it. Reliability varied across patients (raw one-pulse r 0.12–0.87). N1 detection from 1–5 pulses agreed poorly with held-out pulses. Results are limited by one run per patient, few SEEG contacts, no expert N1 labels, pooled polarity, possible stimulus-artifact tails after 10 ms, and the assumption that pulses are exchangeable.

cortico-cortical evoked potentials · SEEG · single-pulse stimulation · split-half reliability · Spearman–Brown prediction · line noise · N1 detection · brain-computer interface