Absolute Hurst deviations separated treatment labels within regions in 20 iEEG patients; fMRI evidence from 10 patients was uncertain

We tested whether absolute Hurst deviations mark treated contacts within brain regions in two separate human patient groups. Public electrical and fMRI estimates came from one epilepsy release. The fixed electrical test used patients without fMRI. The BOLD test used other patients. Each score compared treated and spared contacts within one patient, hemisphere, and region. Regions and patients had equal weights. In 20 iEEG patients, mean AUC was 0.6963, with a 95% patient interval of 0.5866 to 0.7896. In 10 other fMRI patients, mean AUC was 0.6032, with interval 0.4572 to 0.7545. Exact patient sign flips gave Holm probabilities of 0.00240 and 0.11914. The fixed shared prediction failed. An alternative electrical estimator gave uncertain evidence. Six fMRI processing variants gave means from 0.416 to 0.736 in the same people. The variant with the largest estimate did not pass correction across those six tests. An added reference-person bootstrap did not resolve fMRI uncertainty. We preserve every negative score and unavailable poor-outcome check. This is an anatomy-conditional benchmark and source audit. It is not independent-study replication, a clinical targeting system, or causal evidence. Small samples, selected contacts, different cohorts, finite references, and shared source harmonization limit inference. The package includes source tables, terms, hashes, plans, code, all results, figures, and repeat instructions.

Hurst exponent · intracranial EEG · fMRI · epilepsy · normative mapping · treatment localization · regional AUC · negative result