Reverse N1 responses showed reciprocity beyond distance and degree in most SEEG patients, but added little predictive AUC

We tested if cortico-cortical evoked potential (CCEP) connections are reciprocal beyond distance, same-shaft position and node degree. We used the expert-checked N1 labels of OpenNeuro ds005448 v1.0.0 (8 SEEG patients, 5 ECoG patients for comparison). For each patient, a logistic model predicts the forward N1 from log distance, same shaft, out-degree, in-degree and the reverse N1. A null shuffles the responses of each source within 5 mm distance bins (100 permutations). Grouped cross-validation gives the AUC gain from the reverse N1. Results for SEEG: the raw forward probability is 0.32-0.65 with a reverse N1 and 0.06-0.28 without it (8/8 patients). Beyond distance and degree, the reverse coefficient is above its null in 7/8 patients and significant in 5/8 (Fisher p = 7.8e-5). The sign-flip test across patients is not significant (p = 0.19). The same 5 patients are significant for pairs at 20 mm or more. The base model already has AUC 0.90-0.97, and the reverse N1 adds a median of 0.0003 AUC (max 0.0065; far pairs max 0.018). N1 latencies in the two directions correlate in 8/8 SEEG patients (median rho 0.35). In ECoG, reciprocity is significant in 5/5 patients and the AUC gain is 0.001-0.029. Thus a reverse CCEP map gives real but small extra information for SEEG stimulation maps. The package contains code, input data with provenance, per-patient results, a figure and reproduction instructions.

cortico-cortical evoked potentials · CCEP · single-pulse electrical stimulation · SPES · SEEG · ECoG · reciprocity · brain connectivity · epilepsy