Identifying Cardiovascular Disorders Through ECG Image Analysis

An electrocardiogram (ECG) is a crucial tool for assessing the electrical and muscular functions of the heart. It plays a key role in detecting cardiovascular diseases, which can be challenging to diagnose. Traditionally, ECG charts were maintained in paper form, making manual examination time-consuming and requiring specialized expertise. This research proposes a methodology to digitize ECG records, leveraging them to train machine learning models for cardiovascular disease detection. The primary focus is on identifying four major cardiac abnormalities: abnormal heartbeat, myocardial infarction, history of myocardial infarction, and a normal person class. Due to limited datasets, other categories cannot be classified with high accuracy. Various techniques, including Gaussian filtering, thresholding, and grayscale, are employed for data cleaning and feature formation. Preprocessed images are resized into 13 leads, facilitating the examination of ECG reports converted into 1D signals. Multiple algorithms such as SVM (Support Vector Machine), KNN (K-nearest neighbors), and Logistic Regression are implemented, and then ensembled using a classifier i.e., Voting-Based Classifier. The final model selection is based on criteria such as accuracy, precision, recall, and F1-score. The proposed Ensembled Technique using a Voting-based Classifier outperforms the existing works; it achieves 92.5% accuracy, 91% recall, 92% precision, and 92% f1-score overall performance in identifying four major cardiac abnormalities. In essence, this study demonstrates how machine learning, when applied to a substantial dataset, can efficiently identify and predict specific categories of diseases within the realm of cardiovascular health.