Electrocardiographic Machine Learning to Predict Mitral Valve Prolapse in Young Adults

Abstract

Mitral valve prolapse (MVP), known as balloon mitral valve, accounts for 2-4% of cases in the general population and is associated with several cardiac sequelae. A few studies have shown suboptimal results using electrocardiographic (ECG) machine learning to identify MVP in middle-or old-aged individuals; however, no studies have focused on young adults. The aim of this study is to develop an ECG-based system through machine learning to predict MVP in young adults. In a large military population of 2,206 males, aged 17-43 years, support vector machine (SVM), logistic regression (LR) and multilayer perceptron (MLP) classifiers are used as machine learning techniques for 26 ECG features and additional 6 simple biological parameters to link the output of MVP compared with a traditional ECG criterion of a negative T-axis in inferior limb leads. In the parasternal long-axis view of echocardiography, MVP is defined as a displacement of the anterior or posterior leaflet of the mitral valve to the mid portions of the annular hinge point >2 mm. The values of the area under the receiver operating characteristic curve are 74.59%, 74.16% and 73.02% in the proposed SVM, LR and MLP classifiers, respectively, which are better than 38.13% in the traditional ECG criterion for MVP. Our machine learning system provides a novel tool for screening MVP among young male adults. The proposed method can be an adjuvant to the physical findings for early detection of MVP prior to a confirmation by echocardiography for young male adults.