Hybrid Classification Approaches for ECG Signal Detection of Four Cardiac Ailments

To diagnose and classify the different cardiac diseases including arrhythmia, atrial fibrillation, congestive heart failure, and normal sinus rhythm are important. These disorders can be identified using Electrocardiogram (ECG) signals and the recent improvements in deep learning have enabled a higher accuracy in such diagnosis. This paper presents the hybrid classification models based on CNN for the automated identification and categorization of ECG signals for four precise cardiac diseases. The proposed methodology involves noise removal and feature extraction techniques and the discrete wavelet packet transform method, specifically the MODWPT, is used to eliminate noises from the signals and for signal enhancement. CNNs provide an automated feature extraction through its layered architecture and therefore does not require the feature selection process. Thus, the identification of cardiac disorders improves the classification precision, sensitivity, and specificity by using the innovative CNN-AlexNet and CNNVGG16net models. The comparative analysis with more traditional approaches shows that the proposed hybrid deep learning approach is less time-consuming in the detection process and results in higher classification performance. These results support the use of this model in actual clinical situations for real-time analysis of ECGs. The proposed models show the potential for refining ECG signal classification in medical diagnosis in the future.