Curr Cardiol Rev. 2026 Feb 12. doi: 10.2174/011573403X421729251114113706. Online ahead of print.
ABSTRACT
INTRODUCTION: Cardiovascular diseases (CVDs) remain the leading cause of global mortality, with congenital heart disease (CHD), acquired heart disease (AHD), valvular disorders, and cardiomyopathies contributing significantly to morbidity. Retinal fundus imaging has emerged as a non-invasive modality capable of capturing microvascular alterations that may serve as biomarkers for systemic cardiovascular dysfunction.
METHODS: This review systematically examined literature published between 2015 and 2025 on the use of retinal fundus imaging for predicting structural heart diseases. Databases including PubMed, Scopus, and Web of Science were searched using predefined keywords. Studies were evaluated according to disease focus, imaging modality, analytical methods, and diagnostic performance.
RESULTS: Findings highlight that deep learning and machine learning models applied to retinal fundus images have demonstrated promising accuracy in detecting and classifying CVDs. Convolutional neural networks achieved up to 91% AUC for CHD detection, while hybrid multimodal approaches improved sensitivity in AHD and valvular disease prediction. Cardiomyopathies were associated with vessel tortuosity and microhemorrhages, quantifiable through automated image analysis. Table 1 provides a statistical summary of performance across studies.
DISCUSSION: Emerging approaches, such as transformer-based models and adaptations of the Segment Anything Model (SAM) for medical imaging, offer potential for improving generalizability and interpretability. Challenges remain, including dataset imbalance, limited longitudinal validation, and the black-box nature of AI models.
CONCLUSION: Retinal imaging holds strong potential as a scalable, non-invasive tool for cardiovascular disease prediction. Integrating advanced AI architectures may enhance diagnostic accuracy and accelerate translation into clinical practice.
PMID:41691690 | DOI:10.2174/011573403X421729251114113706
