AI for Patient Diagnosis and Treatment
The overarching theme of our research is to develop causally-informed predictive models for healthcare applications. Analyzing large-scale retrospective Electronic Health Records (EHR) data provides unique opportunities to develop technologies applicable in realistic and challenging clinical settings with a strong impact on individualized care. We develop:
- AI models for prediction of disease development trajectories over time, for
- short- and long-term clinical outcome prediction
- identification of increased-risk patient categories
- Causal models for measuring effectiveness of different treatment options, for
- risk analysis of treatment decisions
- assessing potential improvement in patient recovery for new therapies
In this context, we are interested in quantifying the uncertainty of observations and decisions for both humans and AI, designing robust models in the presence of not-at-random missing data, and designing reliable validation metrics geared towards clinical expectations.
Our team, including colleagues, students, and staff, is now distributed across six sites:
- Malone Center for Engineering in Healthcare at the Johns Hopkins University (Baltimore, USA)
- Computer Science department at the Technical University of Munich
- Institute of Computational Biology at the Helmholtz Center in Munich (Munich, Germany)
- Cardiac Surgery and Emergency Medicine at the Johns Hopkins Hospital (Baltimore, USA)
- Neonate Center at the Ludwig Maximilian University Hospital in Munich
- German Center for Lung Research.
We also collaborate with private healthcare companies across Europe, currently in Norway, Switzerland and Germany.
Apply here:
AI Senior Scientist (CLOSED)
Fully funded PhD positions via Munich School for Data Science (CLOSED)