Artificial intelligence in carotid ultrasound demonstrated clinically meaningful performance across multiple domains, including IMT classification accuracy exceeding 90% and pooled plaque-classification AUC of 0.95.
Artificial intelligence in carotid ultrasound demonstrates high diagnostic accuracy across multiple domains, advancing from proof-of-concept to clinically relevant decision support.
Carotid artery ultrasound is a widely used, noninvasive modality for assessing intima-media thickness (IMT), plaque burden, and stenosis severity; however, its interpretation remains constrained by operator dependency and inter-observer variability. Artificial intelligence (AI) has emerged as a practical approach to automating image analysis and improving diagnostic reproducibility. This is a focused narrative review of AI applications in carotid ultrasound, organized by clinical domain, with emphasis on representative performance metrics and translational relevance. Representative peer-reviewed studies published between 2019 and 2026 were categorized into six domains: IMT measurement and boundary segmentation; plaque detection, classification, and segmentation; stenosis assessment and cardiovascular/stroke risk prediction; three-dimensional ultrasound and image enhancement; hemodynamic and arterial stiffness analysis; and point-of-care ultrasound (POCUS) integration and procedural support. AI demonstrated clinically meaningful performance across all domains, including IMT classification accuracy exceeding 90%, a plaque-segmentation Dice coefficient of approximately 0.85, pooled plaque-classification AUC of 0.95, stroke-risk prediction accuracy of 93.81%, portable 3D ultrasound accuracy of 80% (sensitivity 71%, specificity 85%), blood-flow velocity error below 3%, and prospective validation of POCUS-based blood pressure and return-of-spontaneous-circulation assessment. AI in carotid ultrasound has advanced from proof-of-concept analysis to clinically relevant decision support. Broader adoption requires multicenter validation, standardized reporting, and workflow-level implementation.
Jin et al. (Sun,) conducted a review in Carotid atherosclerosis. Artificial intelligence was evaluated. Artificial intelligence in carotid ultrasound demonstrated clinically meaningful performance across multiple domains, including IMT classification accuracy exceeding 90% and pooled plaque-classification AUC of 0.95.