In eight SEEG patients, averaging up to 10 pulses improved N1 detection AUC, while specificity fell at a fixed threshold

In eight patients with stereo-electroencephalography (SEEG), we assessed how the number of single pulses affects detection and map reliability for early N1 responses (10–100 ms). We used 8 stimulated pairs per patient from OpenNeuro ds005448 version 1.0.0: 64 pairs and 5,608 expert-labelled pair-contact combinations (613 N1). Across 40 random draws of 1–10 pulses, median AUC against expert labels increased from 0.68 at 1 pulse to 0.79 at 10; 5 pulses yielded 81% of the 1-to-10 pulse AUC gain. At a fixed z > 3.4 threshold, median sensitivity rose from 0.57 to 0.80, while specificity fell from 0.78 to 0.63 in all 8 patients, as z-scores of non-N1 deflections also grew with averaging. AUC increased from 0.66 to 0.85 for contacts at least 30 mm from the stimulated-pair midpoint (6 patients with such contacts), and from 0.68 to 0.77 for nearer contacts. Split-half map reliability gave a median fitted single-pulse correlation r1 of 0.39 (range 0.12–0.86); the median pulse counts for fitted reliability 0.8 and 0.9 were 6 and 14, respectively, with wide between-patient variation. These estimates are limited because experts labelled responses after viewing the all-pulse average, the detector was a simple z-score, only 8 pairs per patient and one center were studied, and random subsets mixed polarities and 1 and 2 mA stimulation.

stereo-electroencephalography · SEEG · single-pulse electrical stimulation · N1 response · cortico-cortical evoked potentials · pulse count · brain-computer interface · map reliability