Ultrasound-Based Radiomics and Artificial Intelligence in Vascular Disease: Current Evidence, Clinical Applications, and Future Directions

Abstract

Ultrasound is central to vascular imaging because of its accessibility, safety, and real-time capabilities, yet conventional duplex interpretation remains limited by operator dependence and qualitative assessment. Ultrasound-based radiomics and artificial intelligence (AI) offer opportunities to extract quantitative imaging biomarkers and improve reproducibility in vascular disease evaluation. A narrative review was conducted using PubMed and Scopus, including studies published between 2018 and 2025 that evaluated ultrasound-based radiomics or AI in vascular disease. Emphasis was placed on duplex ultrasound applications relevant to arterial, venous, and aneurysmal pathology, while computed tomography- or magnetic resonance imaging-only radiomics studies were excluded. Evidence indicates that ultrasound-based radiomics and AI enhance plaque characterization, automate segmentation, and support risk stratification in carotid artery disease, peripheral arterial disease, venous thromboembolism, and abdominal aortic aneurysms. These approaches reduce interobserver variability and enable standardized longitudinal assessment, although most studies remain retrospective. Clinical translation is constrained by acquisition variability, limited external validation, and challenges integrating AI tools into established vascular laboratory workflows. Current evidence supports AI primarily as a decision-support adjunct rather than a replacement for expert interpretation. Ultrasound-based radiomics and AI are evolving adjuncts to vascular ultrasound. With standardized acquisition and validation, these techniques may enhance reproducibility and care.