A Personal Device for Automatic Evaluation of Health Status during Physical Training
With the advent of smart phones, continuous health care delivery has become easier. Today, there are several vital parameter-sensing platforms that use the mobile phone as a gateway (and display), and have the capability to directly alert the doctor or hospital in case of a problem with the health condition. There is a recent explosion of wearable devices, such as the Apple Watch and Jawbone UP3, which are collecting a plethora of continuous vital signs, providing an opportunity to identify and prevent non-communicable diseases effectively. While continuous mobile health monitoring opens up new healthcare delivery mechanisms, it also creates new challenges. One of the most important challenges is to obtain relevant and meaningful health information from the continuous stream of vital signs. Automatic correlation of different vital signs is still a widely unexplored territory in which new opportunities for evaluating health of human beings can be found. In this project, we aim at developing a set of algorithms aimed at evaluating the health conditions of human beings. We will focus on developing methods to derive the health score, as well as detect any abnormalities in the vital signs of people from a wearable device that obtains key physiological signals: ElectroCardioGram and heart rate, pulse, blood pressure and the level of physical activity. The methods can be used to track and improve the health of normal (healthy) people, as well as provide early warnings for people under high risk of cardiac and syncope events. We will develop and test the methods on healthy, as well as people undergoing cardiac rehabilitation after a cardiac event, performing sports at the EPFL Sports Center. This evaluation will also be enriched by the detection and short-time prediction of certain critical conditions that may require urgent medical assistance (e.g., a Paroxysmal Atrial Fibrillation), developed using anonymyzed patient data. The goal is to implement these techniques in the wearable device.
- Bianchi F. M., Livi L., Ferrante A., Milosevic J., Malek M. (2018) Time series kernel similarities for predicting Paroxysmal Atrial Fibrillation from ECGs. IJCNN 2018 : International Joint Conference on Neural Networks. IEEE. Rio, Brazil