ROLE OF CLOUD-DEPLOYED GRAPH NEURAL NETWORKS IN MAPPING CORONARY ARTERY DISEASE PROGRESSION: A SYSTEMATIC REVIEW

Abstract

Coronary artery disease (CAD) is a leading cause of mortality worldwide, demanding more precise diagnostic strategies. Traditional AI algorithms, like CNNs, RNNs, often fail to absorb these complex relational patterns within cardiovascular data. GNNs give an alternative because they can process such dynamic relationships. This work attempts to study GNNs for CAD progression modeling and diagnosis, including the integration of such models within cloud infrastructures for a scalable and real-time deployment. A systematic literature review was performed in accordance with PRISMA guidelines. The databases searched were PubMed, Web of Science, IEEE Xplore, and Google Scholar, yielding 259 articles. After applying inclusion criteria, 32 studies were selected. These were analyzed from the perspective of GNN architecture, CAD application area, strategies for cloud deployment, and diagnostic performance. GNN-based diagnostic models with accuracies of up to 96% and AUCs higher than 0.90 have been reported in the literature. Subsequent cloud deployment of these models allows real-time inference and easy integration into hospital systems, enabling federated learning of new models while preserving patient data privacy. Use cases include coronary imaging, ECG analysis, and behavioral risk profiling. GNNs, combined with cloud technologies, present a transformative approach for precision cardiology, enabling accurate and personalized CAD diagnostics. However, adoption in clinical settings requires further advancements in model explainability, privacy safeguards, and regulatory compliance. Future research should emphasize open CAD graph datasets, improve GNN interpretability, and validate these systems in real-world clinical environments