Main Article Content
A Recurrent Neural Network (RNN) models are widely used for automation among the latest emerging technologies of Artificial Intelligence. Because of their ability to process temporal data they are applied for Cardiac ailments detection based on the hand-crafted feature set obtained by processing Electrocardiogram (ECG) of the ailments. In this work ECG data from MIT-BIH database of three cardiac ailments such as Arrhythmia, Atrial Fibrillation and Congestive heart failure were considered to detect and classify these ailments from normal signals using algorithms of RNN. Long short-term memory (LSTM), Gated recurrent unit (GRU) and Bi-directional LSTM (Bi-LSTM) are the algorithms of RNN considered in this work to classify the cardiac ailments. Maximal overlap discrete wavelet packet transform (MODWPT) is used to transform and process the ECG signal of ailments and detect the characteristic waves (P, QRS, T) of the ECG signal. Various hand-crafted features are determined from characteristic waves and given as input to train Recurrent Neural Network algorithms such as LSTM, GRU and Bi-LSTM. These algorithms classify the considered cardiac ailments with Accuracy and Mean of MCC (Mathew’s Correlation coefficient) of around 96.11% and 94.83% for LSTM model and 97.50% and 96.68% for GRU and 98.33% and 97.80% for Bi-LSTM algorithms respectively. Of these Bidirectional LSTM (Bi-LSTM) performance is better than other algorithms of RNN. The proposed technique performs better in terms of accuracy with current methods that are discussed in this work.