Retinal image-based cardiovascular risk prediction using AI-CRS: a multi-modal deep learning framework.

Area of Science:

• Ophthalmology
• Artificial Intelligence
• Cardiology

Background:

• Traditional cardiovascular risk assessment relies on handcrafted features and single-modality data, limiting accuracy.
• Retinal imaging offers a non-invasive window into systemic vascular health.
• Subtle vascular changes in the retina can indicate cardiovascular disease risk.

Purpose of the Study:

• To introduce AI-CRS, an AI-driven deep learning framework for enhanced cardiovascular risk assessment using retinal images.
• To overcome limitations of traditional methods by integrating multi-modal data and deep learning.
• To improve the accuracy and interpretability of cardiovascular risk stratification.

Main Methods:

• Developed AI-CRS, a deep learning framework combining convolutional neural networks (CNNs), attention mechanisms, and vasculature segmentation.
• Fused vasculature segmentation maps with raw retinal image data for comprehensive analysis.
• Employed attention mechanisms to enhance model interpretability and identify disease-associated vascular patterns.

Main Results:

• AI-CRS demonstrated superior sensitivity, specificity, and accuracy compared to conventional diagnostic techniques.
• The framework successfully detected early-stage cardiovascular disease and subtle vascular anomalies.
• AI-CRS showed strong generalizability across diverse datasets, demographics, and image qualities.

Conclusions:

• AI-CRS provides a non-invasive, scalable solution for cardiovascular risk assessment using retinal imaging.
• The framework aids early diagnosis, personalized care, and population-wide screening, showing potential for hypertension and diabetes assessment.
• AI-CRS enhances clinical workflows and supports precision medicine through automated analysis and improved interpretability.