TelemonKI: AI-Based Telemonitoring and Therapy Evaluation for Chronic Diseases

AI-Based Telemonitoring and Therapy Evaluation for Chronic Diseases

Projectpartner: Qurasoft GmbH Koblenz

The aim of this project is to develop a software system based on causal neural ordinary differential equations (NODEs) that can predict critical health events in patients with chronic diseases at an earlier stage. This system integrates real-time telemonitoring data with clinical information, enabling physicians to intervene sooner and improve patient care processes.

A key feature of the system is its ability to perform counterfactual comparisons between actual and alternative therapies. This enables a data-driven evaluation of treatment strategies by answering questions such as: “How would the disease have progressed under a different therapy?” Achieving this requires advanced methods from causal machine learning as well as the integration of medical domain knowledge using pretrained models—an approach known as Informed Machine Learning.

The project builds on the SaniQ OS telemonitoring platform developed by Qurasoft, which allows physicians to collect vital parameters (e.g., ECG, blood pressure, weight) from chronic patients in real time and link them with other clinical data. This provides a robust data foundation for the AI models.

Methodologically, the project leverages research from the Biomathematics Group at the University of Koblenz on combining NODEs with causal inference methods to analyze irregular medical time series. This allows for the prediction of disease trajectories and critical events, as well as individualized therapy simulations.

The system will be validated through retrospective secondary data analysis and, subsequently, in a prospective pilot study in real-world settings. This validation process will assess the reliability, trustworthiness, and practical utility of the system—paving the way for potential medical device certification and evaluation under the EU Artificial Intelligence Act