Cardiovascular Care Using A Hybrid Ml/Dl Framework for Early Detection and Real-Time Monitoring

Abstract

Cardiovascular diseases (CVDs) are a leading cause of death globally, and it is important to ensure that arrhythmias are detected in good time to intervene successfully. This paper proposes a hybrid CNN-BLSTM stacked ensemble model with a squeeze-and-excitation residual network (SE-ResNet) to detect and monitor cardiovascular disease from ECG signals. The network uses CNN layers for spatial features, BLSTM networks for temporal dependencies, and ensemble learning for robust classification. Several benchmark datasets were statistically evaluated using multiple INCART and MIT-BIH arrhythmia recordings, comprising more than 200,000 heartbeats from normal and abnormal classes. The proposed model performed better on all major metrics: accuracy of 96.8, precision of 96.5, recall of 97.2, F1-score of 96.8, and a Matthews Correlation Coefficient (MCC) of 0.95 which is significantly better as compared to standalone CNN (89.3), LSTM (90.5) and CNN-LSTM (92.8) models. The statistical analysis shows that the hybrid framework reduces false positives for rare arrhythmias, including premature ventricular contractions (PVCs), by over 30% compared to baseline models, and improves sensitivity and robustness in skewed populations. Comparative performance also reveals an absolute improvement of 37% across all performance measures over traditional methods, demonstrating the effectiveness of combining spatial-temporal feature extraction with attention-based residual learning and an ensemble classifier. The results indicate the clinical potential of hybrid ML/DL frameworks for real-time cardiovascular monitoring and early arrhythmia detection, enabling proactive intervention measures. The further work will engage larger, multi-institutional datasets to improve statistical generalization, streamline the framework to accommodate both the IoMT and edge devices, and incorporate explainable AI methods to present healthcare practitioners with interpretable predictions.