In eight SEEG patients, five pulses were typically within 0.02 AUC of all pulses for N1 detection

This study asked how many single electrical pulses per stimulated SEEG pair were needed for an early N1 cortico-cortical evoked potential detector to agree with expert-checked labels, and whether balancing pulse polarity helped. In eight patients from OpenNeuro ds005448 v1.0.0, random subsets of k = 1, 2, 3, 4, 6, 8, and 10 pulses were averaged for 30 draws per pair and scored with a simple N1 z-score. Median AUC was 0.705 at k = 1, 0.743 at k = 2, 0.787 at k = 4, and 0.807 using all pulses. AUC was lower than the all-pulse result in all eight patients for k = 1 to 8 (exact Wilcoxon p = 0.008 to 0.039). The smallest k within 0.02 AUC of all pulses had a median of 5 (range 1 to 8). Polarity-balanced and same-polarity subsets had similar AUC at k = 2 and 4 (p = 0.64 and 0.31); no polarity effect was found. Responses at least 30 mm from the stimulation site had lower sensitivity and needed more pulses. At a fixed z > 3.4 threshold, both sensitivity and false-positive rate increased with pulse count, so this threshold was not calibrated across k. The expert labels came from all-pulse averages, making this an agreement analysis against an optimistic reference rather than a test against ground truth. The simple detector agreed only moderately with experts (all-pulse AUC 0.67 to 0.93), and the study included eight patients at one centre. The archived report and result files disagree on the number of usable stimulated pairs (93, 94, or 96); this count discrepancy is retained and documented in the full text.

SEEG · cortico-cortical evoked potentials · N1 detection · single-pulse stimulation · pulse count · polarity · expert labels · AUC · sensitivity · OpenNeuro