<p>Cardiovascular diseases remain the foremost cause of global mortality, highlighting the need for accurate, fast, and interpretable diagnostic models. However, Internet of Things (IoT)–enabled wireless sensor networks (WSNs) used in continuous cardiac monitoring often suffer from noisy data, redundant features, and limited generalization, which limit their reliable clinical deployment. This study proposes a Quantum-Adaptive Propagation Graph Network (QAP-GN) to address these challenges and enhance intelligent cardiac disease prediction in WSN-IoT environments. The framework begins with Median and Median Absolute Deviation (MMAD) normalization to suppress outliers and stabilize feature scaling. It then employs a Hybrid Dandelion–Nizar Optimizer (DO-NO) for optimal feature selection, combining global and local search capabilities to minimize redundancy and improve discriminative power. The core QAP-GN architecture integrates adaptive deep graph learning with quantum self-attention, enabling robust spatial–temporal representation of physiological signals. A Newton–Raphson-Based Optimizer (NRBO) further accelerates convergence and enhances learning stability through curvature-based optimization. Experimental validation on the Cleveland and Hungarian heart disease datasets achieved 99.89% accuracy, 98.85% F1-score, and minimal error rates (RMSE = 1.03, MAE = 1.05), with an inference time of 0.02 s. These outcomes surpass those of existing machine and deep learning methods, demonstrating superior diagnostic precision, generalization, and computational efficiency. The proposed QAP-GN framework provides a scalable, real-time, and resource-efficient solution for smart cardiac diagnosis, supporting future integration into IoT-based clinical decision systems and edge-healthcare applications.</p>