Applications of Machine Learning for Predicting Heart Failure

Abstract

Heart Failure is a major health burden for healthcare systems worldwide. Early diagnosis, prediction and management of patients with these conditions are critical to improve patient health outcome. The availability of large datasets from different sources can be leveraged to build machine learning models that can empower clinicians by providing early warnings and insightful information on the underlying conditions of the patients. In this chapter, we review research work on the application of machine learning methods for the diagnosis and prediction of heart failure, and readmission risk scoring. We present recent work on the use of different clinical modalities such as pathology images, echocardiography reports, electronic health records for building predictive models for heart failure diagnosis and prediction. We will cover the model details from traditional machine learning methods as well as from deep learning. Furthermore, we give a summary of the results and performance of these techniques.