Supplementary data and analysis code for generative AI enhancement of low-field MRI

This record contains the supplementary data and reproducible analysis code for the systematic review and meta-analysis “Generative AI enhancement of low-field MRI”. The archive includes the extracted per-study dataset, supplementary protocol-deviation and robustness documentation, and Python scripts used to reproduce the quantitative analyses reported in the manuscript. The analyses cover image-quality outcomes, agreement with high-field MRI references, diagnostic accuracy, within-study AI-on versus AI-off effects, architecture comparisons, GRADE-style absolute effects, likelihood ratios, Fagan post-test probabilities and permutation meta-regression. The review was registered on PROSPERO under CRD420261373181. Twenty-four studies met inclusion criteria, spanning 6.5 mT to 1.5 T. Meta-analytic pooling was restricted to true low-field MRI studies at ≤0.55 T. Files includedSupplementary_data_and_code.zip The archive contains: README.md extraction_dataset.csv Supplementary_S1.pdf code/meta_compute.py code/meta_corrected.py code/meta_advanced.py code/meta_grade_permutation.py code/meta_architecture.py Versionv1.0.0 Publication date2026-06-05 Keywordslow-field MRI; portable MRI; generative artificial intelligence; deep learning; image enhancement; image reconstruction; diagnostic accuracy; systematic review; meta-analysis; PROSPERO; reproducible research LicenceCreative Commons Attribution 4.0 International (CC BY 4.0) Access rightPublic Related identifiersIs supplement to: Generative AI enhancement of low-field MRI Once the paper receives a DOI, update this Zenodo record and add the article DOI as “Is supplement to”. FundingNo dedicated external funding. NotesAll values were extracted from published studies included in the systematic review. No individual participant-level data are included. The code uses Python 3 with numpy, scipy and matplotlib, and writes outputs to relative results and figures folders.