P6: Probabilistic Learning Approaches for Complex Disease Progression Based on High-Dimensional MRI Data

The Project 6 focuses on modeling disease progression in longitudinal biomedical imaging data. We combine deep learning with Bayesian statistical learning to probabilistically predict the evolution of disease over time and improve the quantification of uncertainty in patient outcomes. Our aim is to dynamically model complex structural dependencies in biomedical patient data, while also considering other variables such as gender, age, and ethnicity that may impact disease progression and bias analyses. Our models will be developed and tested for two neurological diseases, Alzheimer’s disease, and multiple sclerosis.

We will collaborate with P5 and P7 on semi-structured deep regression models and with P3 on Bayesian DL models and uncertainty quantification. In addition, we will investigate the Bayesian variable selection procedures for disease progression developed in P5 based on neuroimaging data.

Project Researchers

Clara Hoffmann (KIT)

Moritz Seiler (Charité)