Our project aims to develop an advanced artificial intelligence (AI) system that combines medical images and clinical text to support cardiovascular disease diagnosis and prevention. We will create a universal multimodal foundation model that mimics how clinicians analyze medical images, focusing on carotid ultrasound imaging to predict vascular biological age and cardiovascular risk. Cardiovascular disease remains a leading cause of death globally, with early detection crucial for effective intervention. Our research addresses four key questions: (1) How can AI models integrate medical images and clinical text to mimic clinicians’ diagnostic processes? (2) How can carotid ultrasound imaging combined with clinical records predict vascular biological age? (3) What is the relationship between carotid ultrasound features and vascular aging? (4) To what extent can these measurements predict clinical outcomes? The system will analyze carotid ultrasound images to assess plaque characteristics and arterial wall thickness (Intima-Media Thickness), which are established indicators of vascular health. By integrating this imaging data with clinical information from structured records and medical notes, our model will generate comprehensive vascular aging assessments that can identify individuals at elevated risk before symptoms appear. Our innovative approach involves developing a “medical assistant” AI that focuses on specific anatomical regions, generating detailed diagnostic reports while supporting natural language interaction with healthcare providers. This will enable more precise identification of concerning areas in images while providing contextually rich, expert-level interpretations.
The health-related impact of this research is substantial. Early identification of accelerated vascular aging can trigger preventive interventions before cardiovascular events occur. This could significantly reduce the burden of stroke and heart disease on both patients and healthcare systems. Additionally, automated analysis would increase access to specialized cardiovascular assessments in underserved areas facing specialist shortages. In the public interest, our system will enhance the accessibility and efficiency of cardiovascular care. By automating routine aspects of image analysis, clinicians can focus on complex cases and patient interaction. The technology could democratize access to high-quality cardiovascular risk assessment, particularly benefiting communities with limited healthcare resources. This three-year project will develop clinical decision support tools that provide personalized cardiovascular risk assessments, improving early detection and intervention strategies while making specialized cardiovascular diagnostics more widely available to the public.