Preliminary Experiment Results of Left Ventricular Remodelling Prediction Using Machine Learning Algorithms

Abstract

Left ventricular remodelling involves changes in the ventricular size, shape and function where abnormalities eventually lead to heart failure. Early prediction of left ventricular remodelling can help in enhancing clinical decision making in cardiac health management and reducing cardiovascular mortality. Although cardiac magnetic resonance imaging is increasingly being used in clinical assessment of cardiovascular diseases, there is scarce study on predicting the presence of left ventricular remodelling given the derived data from cardiac magnetic resonance images. Four parameters namely left ventricular end diastolic volume, left ventricular end systolic volume, ejection fraction and occurrence/absence of oedema are used for prediction. A preliminary experiment is conducted where multi-layer perceptron and support vector machine are trained with the parameters obtained from cardiac magnetic resonance images in predicting between patients with left ventricular remodelling or normal. The preliminary experimental results indicated that support vector machine model performed better than multi-layer perception.