Conus radiomics does not detect pediatric tethered cord syndrome — de-identified radiomic feature table and analysis code

Reproducibility archive for the study “Conus radiomics does not detect pediatric tethered cord syndrome: a reproducible, leakage-controlled, equivalence-tested deep-learning pipeline.” It contains a de-identified, study-level IBSI-compliant radiomic feature table (1,456 examinations; no patient identifiers), the list of 68 reproducibility-filtered features, and Python code that reproduces the main result — the equivalence-tested discrimination AUC with bootstrap 95%/90% CI, the positive control, and the learning curve — directly from the feature table. Raw MRI images and patient identifiers are not included and cannot be shared publicly because of ethics and patient-privacy restrictions. See the README for the data dictionary and run instructions.