SEEING THE HEART THROUGH THE EYE: CARDIOVASCULAR RISK ASSESSMENT USING DEEP LEARNING TECHNIQUES

Abstract

It is known that cardiovascular diseases (CVDs) are the leading cause of mortality worldwide, and millions of deaths per year are attributable to these diseases..Early and non-invasive detection of such conditions is essential to improving patient outcomes and reducing the burden on global healthcare systems. Traditional diagnostic procedures are often costly, invasive, and inaccessible to resource-limited regions. Recent advancements in artificial intelligence and medical imaging have opened new avenues for non-invasive screening using retinal fundus images, which reflect vascular health and offer a window into systemic diseases.
In this paper, we present a hybrid deep learning model that is composed of four popular convolutional neural networks, MobileNet and DenseNet, predicting the risk of cardiovascular diseases using retinal fundus images. This combination takes advantage of MobileNet's network efficiency and features that are re-used in DenseNet. The model is trained on known datasets like Kaggle. Our hybrid approach achieves superior accuracy, precision, recall, and F1-score compared to single-model baselines. This research highlights the potential of hybrid CNN models in democratizing heart health diagnostics through a non-invasive, efficient, and scalable method.

Keywords: ,