Distance predicts expert-checked SEEG responses, while binary DTI connectivity adds no held-out AUC in 8 patients

An independent reproduction used new code and new downloads from OpenNeuro ds005448 v1.0.0 to assess whether binary diffusion tensor imaging (DTI) structural connectivity (SC) adds held-out predictive value beyond distance for expert-checked N1 responses to single-pulse stimulation. In 8 SEEG patients and 22,328 observations, median distance-only AUC was 0.894 (range 0.761–0.936), and median SC-only AUC was 0.661. Adding SC to distance and same-shaft features changed leave-one-patient-out AUC by a mean of +0.0007 (p = 0.17); the SC coefficient was positive in 6/8 patients (p = 0.045), and 4/8 patient-level SC permutation tests had p < 0.05. In a new subset of 17,776 other-shaft observations more than 20 mm from the stimulated pair, SC-only AUC was 0.546 (median), and adding SC changed leave-one-patient-out AUC by a mean of −0.0001 (p = 0.95). SC = 1 contacts had higher response rates within the 20–30 mm bin (0.246 vs 0.162), but only 6.5% of subset observations had SC = 1. The median amplitude distance exponent was −0.81 and median latency slope was 0.45 ms/mm. Events files indicated 1 or 2 mA, 1 ms pulses at 0.2 Hz, which differs from the nominal 8 mA reported in the reproduced publication; this discrepancy did not change these analyses. Limits include 8 patients from one center, binary SC for the DTI contact subset, use of the dataset authors’ N1 checks rather than redetection from raw signals, and AUC as a ranking measure.

SEEG · CCEP · brain stimulation · DTI · structural connectivity · distance · N1 response · independent reproduction