At Euclidean distances ≥10 mm in 8 SEEG patients, exponential and power-law models fit CCEP N1 response probability equally well

We tested if the probability of a CCEP N1 response in human SEEG follows the exponential distance rule that tract tracing shows in macaque and mouse. We used OpenNeuro ds005448 v1.0.0 (STReEF) with expert-checked N1 detections. The input files have verified SHA-256 values. For each patient, we fitted p(d) = c*exp(-lambda*d) and p(d) = c*d^(-alpha) with binomial maximum likelihood, for recording contacts 10 mm or more from the midpoint of the stimulated pair. We wrote the analysis plan before the statistics. In the primary test (8 SEEG patients, exact sign-flip test), the mean log-likelihood difference was +0.0005 for each observation (4/8 positive, p = 0.80). Thus the two models fit equally well, and the exponential rule is not supported over the power law. The results are the same without contacts on the stimulated lead (p = 0.80) and for d of 20 mm or more (p = 0.72). The mean lambda in SEEG was 0.074 per mm (95% CI 0.057 to 0.096), which is a decay length of approximately 13.5 mm. This is 2.5 to 3 times slower per mm than the macaque value, but Euclidean distance is not fiber distance. In the top distance quartile, there were more responses than the exponential fit predicted (all 13 patients, 11/13, p = 0.012; SEEG, 7/8, p = 0.0625), but fewer than the power-law fit predicted in 6/8 SEEG patients (post-hoc). For N1 amplitude, log amplitude was fitted better by ln(d) than by d in 8/8 SEEG patients (p = 0.0078). The limits are one centre, n = 8 SEEG patients, Euclidean distance, an N1 detection threshold, and no correction for the secondary tests. The package contains the code (numpy only), the input data, the provenance table, the results and the reproduction instructions.

cortico-cortical evoked potentials · CCEP · stereo-EEG · SEEG · exponential distance rule · power law · N1 response · effective connectivity