In 30 SEEG patients, unmatched SOZ–SOZ connectivity replicated, but shaft-pair matching was inconclusive

Question: AutoSci 57a74009 found in 34 HUP SEEG patients that inter-shaft edges between two seizure-onset-zone (SOZ) bipolar pairs are stronger than length-matched non-SOZ edges (1-40 Hz |r| +0.031, 65-115 Hz envelope r +0.028). Is this also true in an independent cohort? Data: OpenNeuro ds003876 v1.0.2, 30 SEEG patients (NIH 11, Miami 10, JHU 8, UMF 1), one 60-s interictal segment each, read with byte ranges and SHA-256 hashes. The dataset has no coordinates. Thus the pre-registered distance control compared SOZ-SOZ (ss) and non-SOZ (nn) edges only inside the same pair of shafts. Sign-flip tests over patients, Holm over 4 tests. Results: without matching, ss edges are stronger than nn edges (1-40 Hz +0.072, 15/17 patients, p = 0.012; envelope +0.046, 15/17, p = 0.002). With shaft-pair matching, the excess is smaller and not significant (H1 1-40 Hz +0.023, 95% CI -0.035 to +0.097, 8/15, Holm p = 1; H2 envelope +0.037, CI -0.003 to +0.104, 9/15, Holm p = 1). In the same 15 patients, the unmatched 1-40 Hz excess is +0.085 (14/15). SOZ-to-non-SOZ edges are equal to nn edges, as in HUP (1-40 Hz -0.0005, CI -0.009 to +0.009, n = 28). Interpretation: the raw SOZ-SOZ excess replicates, and approximately 70% of the 1-40 Hz excess is explained by shaft geometry. The matched test is inconclusive, not a refutation: the CIs include 0 and the HUP effect sizes, and the power for the HUP effect sizes is only 0.14-0.15 (approximately 150 patients are necessary). Limits: no coordinates (contact index is only a proxy for distance; ss edges stay slightly shorter by index), only 15 patients have SOZ on 2 or more shafts, one short segment without artifact rejection, clinical SOZ labels. Deviations from the plan (segment length, task name fallback, EDF label clean-up) are listed in the README.

stereo-EEG · epilepsy · seizure onset zone · interictal connectivity · intracranial EEG · replication · shaft-pair matching