In 8 SEEG patients, spontaneous coupling added almost no predictive value beyond distance for expert-checked N1 CCEPs

We tested whether a read-only measure can find the contacts that a stimulation pulse activates. The data is OpenNeuro ds005448 (8 SEEG patients, 247 stimulation pairs and 26,052 pair-contact rows). The label is the expert-checked N1. For each pair and contact, we calculated the zero-lag Pearson coupling between the bipolar signal of the stimulated pair and the recording contact in about 80 s of baseline and inter-pulse data, for 1-40 Hz signals and for 70-150 Hz envelopes. Coupling alone gave AUC 0.69 (1-40 Hz) and 0.64 (70-150 Hz), and the AUC was above 0.5 in 8 of 8 patients. But coupling decreases with distance (Spearman rho -0.18 to -0.39). When we added coupling to log distance, the cross-validated AUC did not increase (0.8535 to 0.8536; gain +0.0001, 95% CI -0.0013 to 0.0014; exact sign-flip p = 0.82; 4 of 8 patients positive). A bipolar recording montage, contacts at 15 mm or more, and leave-one-patient-out transfer also gave gains of less than 0.005. Limits: most windows are 2.0-4.8 s after a pulse, the data is short, and we did not test lagged coupling measures.

stereo-EEG · SEEG · CCEP · N1 response · brain connectivity · Pearson coupling · distance · AUC · epilepsy · intracranial stimulation