A common challenge in the age of ‚big data‘ is the siloed storage of medical data in different centers that hinders the effective training and validation of machine learning models.

Therefore, we strive to connect data sources and research centers to train and cross-validate robust models on heterogenous multi-center data.

This workflow of federated learning in collaboration with multiple data sources preserves the autonomy of individual partners with respect to data management as well as the anonymity of patients.
Further, it eliminates biases in data acquisition at individual centers and increases transferability of models to real-world practice.