Personalized & Smartphone-based Solution for Arrhythmia Detection

According to the World Health Organization (WHO)'s report, cardiovascular diseases (CVD) are the leading cause of death in the World [1]. Cardiac arrhythmias are one of the prevalent groups of CVDs referred to as abnormalities and irregularities in heartbeats. These cardiac arrhythmias are associated with an increased risk of serious problems and death [2]. Atrial fibrillation (AF) is one of the prevalent forms of atrial arrhythmia. According to the United States Centers for Disease Control and Prevention (CDC), it is estimated that by 2030 around 12.1 million will suffer from AF in the United States [3]. The estimated average health expenditures of CVDs in the U.S. is 363.4 billion dollars and as the prevalence of CVDs grow the costs are also predicted to double by 2035 [4]. In order to prevent the further advancement of heart disease and stroke, early detection of arrhythmias is crucial. The prevalence of these personal mobile devices has led to rapid growth in the development of medical software applications that provide a conduit to many growing issues in healthcare. Arrhythmia detection using smartphone applications is one of the foremost interests in medical research today. Photoplethysmography (PPG) sensors detect the rate of blood flow by using a light-based technology to determine the electrical signals of the heart. The importance of personalized healthcare technologies specifically for AF management has been emphasized in recent studies [5-7]. The personalized AF detection provides an opportunity to identify each individual's status, which would result in personalized treatment and medication at the right time and the correct dose. This research aims to propose a novel system that can be used for personalized arrhythmia detection using smartphones. The smartphone photoplethysmogram (PPG) signals were gathered from a sample of patients attending a cardiology clinic at Texas Tech and used to detect atrial fibrillation (AF) which is the most common cardiac arrhythmia affecting millions of people worldwide. The AF and normal heart rhythm signals were used to extract the personalized features for each patient. These features were used as the input of the proposed machine learning algorithm to detect the AF in a personalized way. The preliminary clinical results indicate that our proposed system can be used for personalized AF detection and management.