Improved Classification of Coronary Artery Disease through a Machine Learning Algorithm
Coronary artery disease is the leading cause of heart attacks in the United States, attributable to over 50% of heart attacks every year. However, the main method of diagnosing this disease relies on the electrocardiogram (ECG), which has a very low true positive (specificity) and true negative (sensitivity) rate. As a result, patients most often undergo more intrusive tests like an angiogram or an x-ray to accurately diagnose coronary artery disease. This project proposes the use of a support vector machine algorithm to model a database of patients' vitals to classify coronary artery disease more accurately.
Overall, the support vector machine algorithm improved upon the low accuracy and high intrusiveness of current coronary artery disease detection methods. The algorithm performed with a specificity of 91.8% compared to the current specificity of 37.5% through an ECG test and trained the dataset with an accuracy of 92.3%. Additionally, the algorithm can be implemented in the future in conjunction with an ECG for real-time monitoring by patients.
This algorithm won Honorable Mention (3rd Place) Category Award in Physical Sciences and Engineering.