Concluded Projects
The following projects are concluded. Unfortunately, you can't join in anymore, but feel free to contact the researchers for references.
EatMaps: Development of Nutritional-Behavior-Maps from Wearables
We use wearable sensors and smartphone data to understand how emotions, stress, and daily contexts (like location and social situations) influence eating behavior in people with diabetes and obesity. By combining physiological signals, activity tracking, and AI analysis of meal photos, the project aims to identify patterns of emotional eating that drive weight gain and provide personalized dietary interventions.
CASSANDRA: Clinical ASSist AND aleRt Algorithms in visceral surgery
Of Germany's 7.1 million annual surgeries, 110,000 are complex abdominal operations with high complication rates affecting one in four patients. The CASSANDRA project develops machine learning algorithms to predict post-surgical complications using real-time patient monitoring data.
We develop methods and technologies for building AI-based IT systems, focusing on healthcare applications like federated learning for privacy-preserving medical research, analyzing health data from everyday wearable devices, combining clinical records with daily-life sensor data, and making AI algorithms more understandable to doctors and patients.
Low-Cost and Unobtrusive Human Motion Analysis in Daily Life
To assess gait, we develop an inexpensive monitoring system that works at home. It uses commercially available sensor devices that do not require markers on the body. This method supports the rehabilitation progress of e.g. stroke or Parkinson's patients.
SensorHub 2.0: An App to collect scientific data from various sensor sources
We simplify health research by collecting data from multiple wearable devices and questionnaires through a single mobile app. It solves the problem of incompatible data formats from different manufacturers. Researchers can design studies through a web interface and participants use the mobile app to collect all data in a standardized format ready for immediate analysis.
WideHealth: Widening Research on Pervasive and eHealth
We contribute to a European research network focused on data-driven healthcare, human factors in digital health, and federated learning. The project trained early-career researchers and administrative staff across North Macedonia, Slovenia, Portugal, Italy, and Germany to develop and adapt new eHealth technologies for their local healthcare systems.
Privacy-Preserving Federated Learning
Federated learning allows machine learning models to be trained across multiple databases without centralizing sensitive data. This makes it ideal for medical applications where patient privacy must be protected. Our method keeps decoder components private and only shares encoder updates between participants.
Fully-Decentralised Machine Learning
Decentralizing machine learning eliminates the need for a central server by having data owners communicate directly through a peer-to-peer network using a directed acyclic graph structure. The system creates specialized model clusters for different data types, and provides resistance against malicious attacks.
DivAirCity: Equation of Social Inequality, Health Conditions and Air Pollution in Cities
We use nature-based solutions and participatory urban development to improve air quality and climate resilience in cities. We include digital health monitoring to track personal exposure to air pollutants, linking environmental data with health outcomes.
Digital Phenotyping for and Beyond Clinical Trials
Modern sensor technologies can continuously monitor vital signs and physiological changes over time. We use this to better understand human health patterns and disease prevention.
INALO: Intelligent Alarm Optimization for the Intensive Care Unit
Up to 99% of ICU vital sign alarms are false positives, leading to staff desensitization and preventable deaths. This project develops AI-powered software that combines patient monitoring data with electronic health records to create patient-specific alarm systems that can filter out false alarms and prioritize truly critical alerts.
AI-Rescue: AI-aided Data Analysis and Simulation of German Emergency Medical Services
We develop artificial intelligence tools to support German emergency medical services in making faster, better decisions during life-threatening situations. We research how mobile sensors can monitor patients' critical health conditions during emergencies and collect data for hospital treatment.
