At six CCEP sites, all-contact referencing did not consistently reduce evoked RMS distance dependence

We compared whether an all-contact common average reference (CAR_all) reduced the distance dependence of evoked responses more than a common average reference of the 30% least-responsive contacts (CAR_low). Across six cortico-cortical evoked potential (CCEP) stimulation sites in four SEEG subjects, the mean difference in per-site Spearman rho for distance versus log10 evoked RMS (CAR_all minus CAR_low) was +0.07 (four of six sites positive; range −0.52 to +0.66; exact one-sided sign-flip p = 0.28, Holm-adjusted p = 0.31). The primary hypothesis was not supported. The secondary far/near RMS ratio difference was +0.068 (four of six sites positive; p = 0.16, Holm-adjusted p = 0.31), also not supported. Descriptively, median rho was −0.25 for the original reference, −0.21 for CAR_all, and −0.31 for CAR_low. The mean fraction of contacts more than 40 mm away classified as responsive by a fixed 5 SD rule was 0.002, 0.22, and 0.05, respectively. Thus CAR_all increased far-contact threshold detections but did not change the rank order of response size over distance in a consistent direction at these sites. The analysis used six even test trials per site and Euclidean distance in ACPC coordinates. The sample is small, there was no line-noise filter in the upstream RMS values, the CAR_low subset is a fixed 30% rule rather than the CARLA algorithm, and a moderate real effect is not excluded.

cortico-cortical evoked potentials · stereo-electroencephalography · common average reference · evoked RMS · electrode distance · sign-flip test