Towards Automatic Pathology Classification for a 24/7ECG-based Telemonitoring Service

Xenia Klinge, Matthias Nadig, Andrey Girenko, Jan Alexandersson, Boicho Boichev

In: Proceedings of the 6th international Workshop on Sensor-based Activity Recognition and Interaction. International Workshop on Sensor-based Activity Recognition and Interaction (iWOAR-2019) September 16-17 Rostock Germany ISBN 978-1-4503-7714-0/19/09 ACM 2019.


We present work in progress focusing on an extension of a 24/7 monitoring system for persons at cardiac risk, a system that allows patients to move freely in and outside of clinical settings. The system consists of a comprehensive sensor patch, a relay and a monitoring center staffed with clinicians and doctors. Detected anomalies may trigger escalation plans which include activating families, ambulances and clinics. In order for the system to scale up, we demonstrate how anomalies can be automatically detected using machine learning technologies. We describe and evaluate a classifier for detection of Ventricular Extrasystoles. The classifier has been tested on real-world sensor data as well as a standard ECG database and achieves varying recognition rates, depending on many factors. The next steps include an improvement of the detection algorithm, especially its training methods, and introducing the extension into the telemonitoring centre, thus evaluating the user acceptance amongst the cardio experts.


iWOAR2019_paper_18.pdf (pdf, 3 MB)

Deutsches Forschungszentrum für Künstliche Intelligenz
German Research Center for Artificial Intelligence