A meta-heuristic aided arrhythmia classification model using advanced deep learning technique with multiple feature extraction mechanisms

Cardiac arrhythmia poses an important threat to human life; hence it is an urge to diagnose properly. There are numerous mechanisms deployed for the identification of arrhythmias; yet, most of the techniques have been utilized sources such as Electrocardiogram (ECG). The ECG-based manual evaluation by the medical analysts is inaccurate. Some experiments have been concentrated on the accuracy and the speed of the learning method by utilizing Artificial Intelligence (AI), and pattern detection in the classification model. However, there are two primary limitations in the conventional mechanisms; the models demand large training time and demand feature selection on a manual basis. Hence, an intellectual arrhythmia classification model using deep learning is introduced to identify the irregular heartbeat. In the beginning, the required signals are accumulated from standard sources. Further, three different kinds of features are extracted for an efficient automatic classification process of arrhythmia. At first, the deep features are extracted by applying the Conditional Autoencoder, and these features are considered as feature set 1. Further, wave features and spectral features are retrieved from the input signal and these features are considered as feature set 2. Subsequently, the signals are converted into spectrogram images and the Graph Convolutional Neural Network (GCNN) technique is employed to retrieve the feature set 3 from those images. Further, the ensemble feature fusion process takes place to combine all three sets of features. Ensemble features are provided as input for the Optimal Dense Recurrent neural network with Attention Mechanism (ODR-AM) for classifying the arrhythmia. The classifier's performance is boosted by optimizing the parameters using the Augmented Random value of Giant Armadillo Optimization (ARGAO). This model is useful to know about the specific type of arrhythmia. Finally, the simulation findings of the presented model are analyzed with other conventional models.