Hybrid CNN-LSTM deep learning model and ensemble technique for automatic detection of myocardial infarction using big ECG data

Automatic and accurate prognosis of myocardial infarction (MI) using electrocardiogram (ECG) signals is a challenging task for the diagnosis and treatment of heart diseases. MI is also referred as a Heart Attack, which is the most fatal cardiovascular disease. In many cases, MI does not show any symptoms, hence it is also called a silent heart attack. In such cases, patients do not get time to prepare themselves. Hence this disease is more dangerous and fatal with a high mortality rate. Hence, we have proposed an automated detection system of MI using electrocardiogram (ECG) signals by a convolutional neural network (CNN), hybrid CNN- long short-term memory network (LSTM), and ensemble technique to choose the optimum performing model. In this work, we have used 123,998 ECG beats obtained from the PTB diagnostic database (PTBDB) and MIT-BIH arrhythmia database (MITDB) to develop the model. The experiment is performed in two stages: (i) using original and unbalanced datasets and (ii) using a balanced dataset, obtained from synthetic minority oversampling technique (SMOTE) data sampling technique. We have obtained the highest classification accuracy of 99.82 %, 99.88 %, and 99.89 % using CNN, hybrid CNN-LSTM, and ensemble techniques, respectively. Hence the proposed novel data balancing technique (SMOTE-Tomek Link) not only solves the imbalanced data problem but also increases the minority class accuracy significantly. Now our developed model is ready for the clinical application that can be installed in hospitals for the detection of MI.

Automated summary

An automated detection system for myocardial infarction using electrocardiogram signals achieved a classification accuracy of 99.89% with the use of convolutional neural network (CNN), hybrid CNN-LSTM, and ensemble techniques. A novel data balancing technique, SMOTE-Tomek Link, was introduced to solve the problem of imbalanced data and significantly increase accuracy in the minority class.