A scoping review identified 38 open-access cardiovascular imaging datasets, predominantly CT (68%), highlighting notable gaps in peripheral vasculature, normal anatomy, and cross-modality coverage.
This scoping review provides a comprehensive catalog of 38 open-access cardiovascular imaging datasets, highlighting significant gaps in peripheral vasculature and normal anatomy data to guide future AI research.
AbstractObjectives Open-access datasets are a cornerstone of artificial intelligence research in medical imaging, yet structured catalogs remain scarce and cardiovascular imaging lacks any comprehensive map of its open-access landscape. This scoping review aimed to identify and characterise publicly available cardiovascular imaging datasets, map their properties across anatomical domains and modalities, and highlight gaps to guide future dataset development. Materials & Methods A structured search of the scientific literature and major public repositories (Zenodo, Figshare, Kaggle, and others) was conducted up to March 2026. Datasets were included if they contained cardiovascular imaging data acquired by CT, MRI, or ultrasound and were freely accessible. Two reviewers extracted data including anatomical domain, imaging modality, dataset size, annotation type, licensing, and hosting platform. Results Thirty-eight open-access datasets were identified spanning six anatomical domains: heart (n=16), coronary arteries (n=5), aorta (n=4), pulmonary arteries (n=6), peripheral arteries (n=3), and multi-organ structures (n=4). CT was the dominant modality (68%), followed by ultrasound (24%) and MRI (18%). Dataset sizes ranged from 27 to over 23,000 cases. Licensing varied from fully permissive (Apache 2.0, CC BY 4.0) to restrictive data use agreements. Over half of datasets (53%) were pathology-focused, with normal imaging data largely absent. Conclusion The open cardiovascular imaging dataset landscape remains heterogeneous and incomplete, with notable gaps in peripheral vasculature, normal anatomy, and cross-modality coverage. A freely accessible catalog is provided to support benchmarking and future dataset development.
Touray et al. (Fri,) conducted a review in Cardiovascular imaging (n=38). Open-access datasets was evaluated on Dataset characteristics across anatomical domains and modalities. A scoping review identified 38 open-access cardiovascular imaging datasets, predominantly CT (68%), highlighting notable gaps in peripheral vasculature, normal anatomy, and cross-modality coverage.