SmartCare: Detecting Heart Failure and Diabetes Using Smartwatch

Abstract

Busy lifestyles of people which resulted in an increase in non-communicable diseases have demanded a revolution in the healthcare system. This has prompted active research in developing smart sensing devices to automatically monitor the health status of a user with less human intervention. This could be more challenging when the disease is asymptomatic, hence smart solutions for early detection of such diseases are vital to help people to maintain a healthy and long life. In this study, we focus on the most common non-communicable diseases, Heart Failure, and Diabetes which are asymptomatic in their early stages. We propose a SmartCare solution for the real-time detection of heart failure and diabetes disease using a smartwatch. Data collected through a smartwatch along with health data provided by the user are used to detect heart failure, severity levels of the heart failure, diabetes disease, and types of diabetes. Random Forest and Logistic Regression algorithms are used to develop the four prediction models. Extensive evaluations performed on patients' data collected from local hospitals show our SmartCare system can detect the heart failure, severity levels of the heart failure, diabetes disease, and types of diabetes with an F1 score of 0.72, 0.7, 0.72, and 0.86 respectively.