Deep Neural Network Framework for Predicting Cardiovascular Diseases from ECG Signals

Introduction: Cardio Vascular Disease (CVD), a primary cause of death worldwide, includes a variety of heart-related disorders like heart failure, arrhythmias, and coronary artery disease (CAD), where plaque buildup narrows the heart muscle's blood vessels and causes angina or heart attacks. Genetics, congenital anomalies, bad diet, lack of exercise, smoking, and chronic diseases including hypertension and diabetes can cause cardiac disease. Method: The symptoms can range from chest pain and shortness of breath to exhaustion and palpitations and diagnosis usually involves a medical history, physical examination, and electrocardiograms (ECGs), and stress testing. Lifestyle adjustments, medicines, angioplasty, and bypass grafts or heart transplants are possible treatments. Preventive measures include healthy living, risk factor management, and frequent checkups, which are few measures, whereas advanced algorithms can analyze massive volumes of ECG and MRI data to find patterns and anomalies that humans may overlook. Results: The deep learning models increase arrhythmia, coronary artery disease, and heart failure diagnosis accuracy and speed. They enable predictive analytics, early intervention, and personalized treatment programs, increasing cardiac care results. The proposed DNN model consists of a 3-layer architecture having input, hidden, and output layers. In the hidden layer, 2 layers, namely, the dense layer and batch normalization layer are added to enhance its accuracy. Conclusion: Three different optimizers namely Adam, AdaGrad, and AdaDelta are tested on 50 epochs and 32 batch sizes for predicting cardiovascular disease. Adam optimizer has the highest accuracy of 85% using the proposed deep neural network.