Heiko Krumm’s research while affiliated with TU Dortmund University and other places

In this paper, we present a system that allows patients who require anticoagulation medicine an opportunity to independently manage their dosage concentration with the help of two machine learning algorithms. The basic idea is to predict the next dosage by using a neuronal network and the model predictive control approach, both based on the history of data already available from patients. This machine learning system is expanded by an smartphone application for the patients, and a website for the doctors to support their patients.

Long-term survival after left ventricular assist device (LVAD) implantation in heart failure patients is mainly determined by a sophisticated after-care. Ambulatory visits only take place every 12 weeks. In case of life-threatening complications (pump thrombosis, driveline infection) this might lead to delayed diagnosis and delayed intervention. It is the intention of the international project Medolution (Medical care evolution) to develop new approaches in order to create best structures for telemonitoring of LVAD patients. In the very early period of the project a questionnaire was sent to 180 LVAD patients to evaluate the need and acceptance of telemonitoring. Thereafter, a graphical user interface (GUI) mockup was developed as one of the first steps to improve the continuous contact between the LVAD patient and the physician. As a final goal the Medolution project aims to bundle all relevant informations from different data sources into one platform in order to provide the physician a comprehensive overview of a patient's situation. In the systems background a big data analysis should run permanently and should try to detect abnormalities and correlations as well. At crucial events, a notification system should inform the physician and should provide the causing data via a decision support system. With this new system we are expecting early detection and prevention of common and partially life-threatening complications, less readmissions to the hospital, an increase in quality of life for the patients and less costs for the health care system as well.

Although regular physical activities reduce mortality and increase quality of life many cardiac patients discontinue training due to lack of motivation, lack of time or having health concerns because of a too high training intensity. Therefore, we developed an exergaming based system to enhance long-term motivation in the context of rehabilitation training. We combined different hardware components such as vital sensors, a virtual reality headset, a motion detecting camera, a bicycle ergometer and a motion platform to create an immersive and fun experience for the training user without having to worry about any negative health impact. Our evaluation shows that the system is well accepted by the users and is capable of tackling the aforementioned reasons for an inactive lifestyle. The system is designed to be easily extensible, safe to use and enables professionals to adjust and to telemonitor the training at any time.

Networked customer devices provide a powerful, flexible and cost-efficient implementation environment for cyber-physical system applications but due to their inherent unreliability can directly support limited dependability only. Since in such an environment non-critical functions can collect, process and prepare comprising data which supports the adaptation and optimization so that new and significantly enriched medical functions can be realized, we envisage Big Dependable Systems (BDS), a corresponding upcoming class of systems. Major steps towards their systematic development utilize approaches for communication and dynamic configuration in service-oriented systems, for automated monitoring and repair, and for in-built fault tolerance. We report on respective approaches examined and on a set of pilot applications before the vision of BDS is outlined before that background. The pilot applications employ cyber-physical systems in order to support rehabilitation training sessions which are remotely supervised and take place as well at home as in a virtual reality.

The GlobalSensing system has the aim to supervise (bicycle-) hiking of a group of patients in a cardiologic rehabilitation by taking advantage of user-friendly components. A smartphone application, the patient component, records the vital data and data of the training by using a sensor-broker. These data is transmitted via internet to the group leader and to the cardiologist. This arrangement offers the possibility to monitor the data in real time, to ensure an optimal individual training and to protect the patient against overloads.

The RehaWeb system aims at motivating heart patients in rehab with a combination of social networking features, editorial contents as well as mobile support and monitoring. Hikes on selected routes can be planned with RehaWeb community friends. A smartphone application guides the way and collects vitals. All data is transmitted to the RehaWeb server, where it can be accessed as exercise diary by the patient and medical staff. The personal progress and other topics can be discussed with other user in the forums.