##plugins.themes.bootstrap3.article.main##

Francis Densil Raj V

Aravind Babu L R

Sivakannan S

Abstract

Cardiovascular disease (CVD) remains a leading cause of mortality and morbidity worldwide. Furthermore, it is projected that worldwide, cardiovascular disease is the primary cause of death and loss of disability-adjusted life years. Over time, wealthy countries have seen a decrease in the rates of cardiovascular death, while the burden of cardiovascular disease has significantly increased in low-income and middle-income countries. The main clinical method for identifying irregularities in cardiac function is through the utilization of a standard 12-lead electrocardiogram (ECG) apparatus. The general public can be screened and physicians can receive additional evaluations through automated 12-lead ECG machines. Nonetheless, a manual ECG interpretation necessitates both expertise and time. In today's rapidly changing world, accurate diagnoses of cardiac abnormalities are crucial for patients' well-being. This paper mainly focuses on the classification of Normal ECG, Atrial flutter, Anteroseptal infarct, and Aortic Valve cardiovascular diseases using Artificial Intelligence techniques. For this study, a dataset of 19140 ECG signals was extracted from the MIMIC-IV-ECG dataset. The original dataset contains more than 15 classes whereas for this study and the authors considered the following four classes namely Normal ECG, Atrial flutter, Anteroseptal infarct, and Aortic Valve. Each class identified represents the diagnosis of a particular set of ECG signals. Pre-processing was done on each ECG signal to prepare it for feature extraction using wavelet transform. A correlation matrix was used to select features after the feature extraction. Four algorithms namely “Random Forest (RF), K-Nearest Neighbour (KNN), Gradient Boosting (GB), and Artificial Neural Network (ANN)” have been compared on the dataset to analyze the performance. ANN scored better than the other models. The performance accuracy is RF - 93.15%, KNN = 91.90%, GB – 92.42%, and ANN – 94.50% respectively.

##plugins.themes.bootstrap3.article.details##