New Paper
13 Apr 2023
Using artificial intelligence (AI) to develop a prediction system for severe courses of covid 19 diseases: Since the beginning of 2021, this task has been the focus of a research network led by the team of biomathematician Prof. Dr. Maik Kschischo from the Remagen RheinAhr Campus of Koblenz University of Applied Sciences. The research team has now successfully completed the project and presented a so-called global trigger tool with great predictive power for the course of disease in hospitalised Covid 19 patients in a scientific publication.
More than 10,000 people in Europe still die every month from or with Covid-19. "Early detection of severe disease progression has the advantage that high-risk patients can be cared for more closely and examined more precisely, and drug intervention can be undertaken at an early stage," the project report states. In addition, scarce resources could be used specifically for patients with high risks.
Against this background, the Koblenz research team developed a pioneering software tool with the help of hospital data together with the Cologne-based start-up company damedic. The first steps involved, among other things, processing anonymised patient data tailored to the project goal. An interdisciplinary team of physicians, IT specialists, coding specialists and the research team of Koblenz University of Applied Sciences - consisting of Prof. Dr. Maik Kschischo, Prof. Dr. Christof Schenkel-Häger, Dr. Jörg Zimmermann, Vanessa Schmitt and Philipp Wendland - worked closely together on the pre-processing of the data.
With the help of the cleaned patient data, the research network was able to create "predictive machine learning models for the prediction of death, transfer to the intensive care unit and the need for mechanical ventilation", the project report continues. The models included laboratory values closely linked to the course of the disease, the age and gender of the patients as predictive variables.
"All in all, the project goal of creating a global trigger tool for severe events in hospitalised Covid 19 patients was achieved," the research consortium concludes. The prediction models for whether a patient will die, be transferred to the intensive care unit or require mechanical ventilation provided reliable results. This also showed a particularly high predictive power of the blood values included in the respective model.
"We will also benefit from our experience with the preparation and analysis of patient data and also the experience with the development of AI methods in several follow-up projects," Kschischo looks ahead. In the future, the research team wants to contribute to using AI to make predictions about acute kidney failure and sepsis.
"This project is an excellent example of the interdisciplinary cooperation between the research groups at our university and also an outstanding example of our research activities that focus on the most pressing challenges. It reflects our commitment to current and socially relevant topics," emphasises Prof. Dr. Antje Liersch, Vice President for Research at Koblenz University of Applied Sciences.
The project at Koblenz University of Applied Sciences, where the "Data-Driven Systems" cluster is a research focus, was funded by the Rhineland-Palatinate Ministry of Science with around 185,000 euros. In total, the Ministry of Science invested 1.3 million euros in research on Corona and AI for nine projects at universities of applied sciences in the state. You can read the scientific publication "Machine learning models for predicting severe COVID-19 outcomes in hospitals" with all the details here: https://doi.org/10.1016/j.imu.2023.101188
Created with Microweber.