Four Class Label Prediction and Heart Condition Diagnosis using Multi Structured Pooling Deep Recurring Neural Network

Background: It is important to classify the ECG (Electrocardiogram) signal in order to diagnose heart illness. Simultaneously, the identification of heart illness also depends on the early identification and precise prediction of various cardiac arrhythmias. The goal of the current study is to categorize the four important class labels that define the conditions of the heart, which are Supraventricular Ectopy, Ventricular Ectopy, Normal Sinus Rhythm, and Supraventricular and Ventricular Ectopy. Objectives: The study's objectives are to remove noise from the ECG signal by using an elliptic filter based on empirical mode decomposition, or EMD-EF. Additionally, it makes use of Multi Structured Pooling Deep RNN (MSP-Deep RNN) and Distribution based Whale Optimization Algorithm (D-WOA) to efficiently choose features and anticipate the four crucial class labels. Methodology: First, the suggested EMD-EF processes the input signal and eliminates undesired noise from it. Subsequently, the pre-processed data are extracted for features by taking into account the various signals of the ECG waveform, which include the PR, PT, QT, QRS Complex, and PQRST Amplitude prediction. The D-WOA then chooses just the pertinent elements for additional classification. The next step is to forecast the four class labels using MSP-Deep RNN. Results/Conclusion: To determine the effectiveness of the suggested system, a comparison analysis is conducted using a range of metrics. Positive Predictive Value (PPV), Specificity, Sensitivity, Accuracy, and Error rate are the parameters that are considered. The suggested method outperforms the current methods in terms of efficiency, as demonstrated by its high accuracy rate and little error in identifying the ECG signals and accurately predicting the class labels.