Network Based Quantum Inspired Deep Learning Framework for Cardiovascular Disease Prediction Using Interpretable Feature Interaction Networks

Abstract Cardiovascular disease (CVD) remains the leading cause of global mortality, accounting for approximately 17.9 million deaths annually. It is required that the development of sophisticated network-based predictive algorithms that can preserve the multifaceted feature interaction in the early detection of the disease and clinical decision support. This research introduces a new hybrid scheme of quantum-inspired feature interaction networks and deep neural network structures to predict CVD. We use our methodology to build a full graph of feature interactions based on ZZ-feature maps and the weight of edges is a pairwise quantum-inspired correlation between clinical features. This network-based representation is further provided to a multi-layered neural network with ensemble decision-making. Experimental validation on the UCI Cleveland Heart Disease dataset demonstrates (303 patients, 13 clinical attributes), our hybrid architecture has an accuracy of 90.5% (95% CI: 89.491.6), precision of 89.1, recall of 91.24%, F1-score of 89.9%, and ROC-AUC of 0.936, and a 5-fold stratified cross-validation. The statistical significance analysis confirms that the quantum-inspired network transformation yields statistically significant performance improvements over the baseline approaches (p < 0.01, Cohen’s d > 0.8). However, we note that control experiments isolating the ZZ-feature map contribution from the dimensionality expansion (20→40 dimensions) and ensemble effects have not been performed in the current study and are reserved for future work. The results of SHAP-based network interpretability analysis indicate that thalassemia, the major vessel count, and chest pain type are the most powerful nodes in the prediction network, which aligns with traditional cardiovascular diagnostic guidelines. The suggested framework provides a reproducible methodology, with transparent limitation reporting, that serves as a basis for network-driven clinical decision support systems and preserves paths for translating quantum hardware.