“Prediction of Cardiovascular Disease Using Retinal Images with Deep Learning method”

Roopesh Kumar BN 1, Raghavi CS2,
1Associate Professor, Department of Computer Science and Engineering, KSIT, India

2345 Student, Department of Computer Science and Engineering, KSIT, India

Abstract:
Cardio Sentinel is a multimodal diagnostic framework engineered to quantify cardiovascular risk by integrating non-invasive ocular biomarkers with systemic clinical data. The architecture utilizes a bifurcated neural strategy: a specialized Convolutional Neural Network (CNN) optimized for high-resolution retinal fundus feature extraction, operating in parallel with a Deep Neural Network (DNN) designed to process multidimensional patient vectors, including hemodynamic and metabolic indicators such as blood pressure, lipid profiles, and Body Mass Index (BMI). A critical innovation of this system is its intermediate feature fusion layer, which synthesizes spatial vascular geometry and tabular health records into a unified predictive tensor. To mitigate the common challenge of sparse clinical data, the project incorporates a physiologically constrained generative pipeline to produce synthetic patient profiles based on validated epidemiological distributions. The final system provides a stratified risk assessment—categorized into low, moderate, and high tiers—rendered through an explainable AI (XAI) framework. This decision-support tool identifies subtle microvascular markers often missed during manual inspections, enabling healthcare practitioners to initiate proactive, data-driven interventions before acute clinical events manifest.

Keywords:

Healthcare & System Performance, Bifurcated Neural Topology, Systemic Vascular Integrity.