Volume 50, Issue 3 Fall 2026 JVU 503-cc | Ultrasound-Based Radiomics and Artificial Intelligence in Vascular Disease: Current Evidence, Clinical Applications, and Future Directions-Ultrasound-Based Radiomics and Artificial Intelligence in Vascular Disease: Current Evidence, Clinical Applications, and Future Directions

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

Back to course

Pdf Summary

This narrative review examines ultrasound-based radiomics and artificial intelligence (AI) for vascular disease, synthesizing 43 studies published between 2018 and 2025. Radiomics converts ultrasound images into quantitative measures of intensity, texture, and shape, while machine-learning and deep-learning models can automate image segmentation, classification, and risk prediction.<br /><br />Reported applications span carotid disease, peripheral artery disease (PAD), venous thromboembolism, abdominal aortic aneurysms (AAAs), and emerging areas such as dialysis access, transplant assessment, and microvascular disease. In carotid disease, quantitative plaque analysis may identify vulnerability and improve stroke-risk assessment beyond stenosis measurements alone. In PAD, AI can classify Doppler waveforms and quantify plaque or arterial stiffness. In DVT, AI and elastography may improve detection and distinguish acute from chronic thrombus. For AAAs, models combining imaging, biomechanical, and clinical features may predict growth or rupture more effectively than diameter alone. Across these applications, AI may improve consistency, reduce observer variability, and support individualized monitoring.<br /><br />Clinical adoption remains limited by ultrasound’s sensitivity to operator, equipment, and acquisition settings; inconsistent preprocessing and segmentation; small, retrospective, often single-center datasets; and limited external validation. Deep-learning models may also be difficult to interpret, while biased datasets and privacy concerns raise ethical and regulatory issues.<br /><br />The review recommends standardized image acquisition and feature extraction, transparent reporting, and prospective multicenter validation in diverse populations. Tools should integrate smoothly into vascular laboratory workflows and be evaluated for their effect on patient outcomes. At present, radiomics and AI are best used as decision-support aids—not replacements for sonographers’ image-quality control or clinicians’ expert judgment. Their eventual value will depend on reliable validation, interpretability, and demonstrated benefit to patients.

Keywords

ultrasound radiomics

artificial intelligence

vascular disease

machine learning

deep learning

carotid plaque

peripheral artery disease

deep vein thrombosis

abdominal aortic aneurysm

image segmentation