LSTM-Based Cardiovascular Disease Detection Using ECG Signal

Out of all diseases, the cardiovascular disease (CVD) kills the largest number of human beings in the world. This has drawn the attention of many researchers to address this vital issue and to suggest appropriate early detection method of heart diseases so that many valuable lives can be saved. The CVDs are mainly detected from various test data as well as signals such as ECGs, heart sounds, and cardiac computerized tomography scans. In this chapter, attempt has been made to propose an efficient diagnose coronary heart disease (CHD) from the ECG recordings of the subjects employing a simple but robust LSTM network method of detection of CHD. The standard PTB diagnostic database version 1.0.0—PhysioNet comprising the ECG signals recordings of 268 subjects are used in the model. Three stages of LSTM network with 256, 128, and 64 modules in each stage and using 20% random dropouts of weights between modules are employed to develop the detection model. Two types of training and validation schemes (80 and 20%) and (70 and 30%) of the datasets have been carried out. The simulation-based experiments of the developed model using the standard ECG signals of healthy and HD patients exhibit the best performance of 86.598%, 93.817%, 88.015%, 71.597%, and 0.915 in terms of accuracy, precision, sensitivity, specificity, and F1-score, respectively. The proposed model can be conveniently used by doctors for early diagnosis of the CVD.