An Automated System Pre-Trained to Identify Cardiovascular Disorders from ECG Pictures

Globally, heart disorders, often recognized as cardiovascular diseases, are among the major causes of death. Additional lives might be protected the earlier they are identified and anticipated. Cardiovascular disease can be identified with an electrocardiogram (ECG), a simple, low-cost, non-invasive method of detecting the electrical functioning of the heart. Employing a public dataset of ECG images from heart patients, deep learning algorithms were utilized in this study to calculate the four main cardiac abnormalities: abnormal heartbeat, myocardial infarction history, coronary artery disease, and typical individual categories. Initially, the transfer method for learning was studied using the previously learned deep neural networks, VGG16, RESNET50, and InceptionV3. Classical methods of machine learning such as Support Vector Machine, K-nearest relatives, Random Forest, and Naïve Bayes were able to identify features employing the models that were already trained previously stated. For the study, ECG images from people who had or did not have a diagnosis of CVD were collected. Using these preprocessed pictures, relevant features including time-domain, frequency-domain, including nonlinear features were extracted. The collected features were then used to train various machine learning computations, which helped identify the prevalence of CVD. The efficiency of each algorithm was evaluated using metrics such as precision, specificity, and consciousness. In addition, the research employed Flask technology in Python to create an online user experience for the model, facilitating patients' and healthcare practitioners' easy access. ECG pictures can be uploaded through the interface, and the model analyzes them to determine whether CVD is present. These algorithms can also offer a reliable diagnosis, which lowers the possibility of human error.