Prediction of Arrhythmia using Hybrid CNN

Cardiac arrhythmia, marked by irregular heart rhythms, poses significant health risks, necessitating accurate and timely detection to prevent complications like stroke. Conventional electrocardiogram (ECG) analysis, reliant on manual interpretation, is laborintensive and error-prone, while traditional machine learning models, such as support vector machines, and standard convolutional neural networks (CNNs) often fail to capture complex temporal patterns or handle imbalanced datasets effectively. This study introduces a hybrid CNN architecture for arrhythmia prediction, incorporating feature fusion and attention mechanisms to enhance spatial and temporal feature extraction from ECG signals. The model is developed and tested using the MIT-BIH Arrhythmia Database, comprising 48 ECG recordings sampled at 360 Hz. Preprocessing involves noise filtering, normalization, and segmentation into 2-second windows around R-peaks. The proposed model achieves an accuracy of 98.2%, sensitivity of 98.0%, and specificity of 98.5%, surpassing baseline machine learning and deep learning methods by 2–5% in key metrics. The attention mechanism improves classification of minority classes, mitigating data imbalance. This hybrid CNN offers a robust solution for automated arrhythmia detection, with promising potential for real-time monitoring and integration into wearable ECG devices for continuous cardiac assessment.