Niklas Probul1*, Mohammad Bakhtiari1, Mohammad Kazemi Majdabadi1, Balázs Orbán3, Sándor Fejér3, Supratim Das1, Julian Klemm1, Christina C Saak1, Nina K Wenke1, and Jan Baumbach1,2
1 Institute for Computational Systems Biology, University of Hamburg, Hamburg, Germany
2 Institute of Mathematics and Computer Science, University of Southern Denmark, Odense, Denmark
3 Gnome Design SRL, Sfântu Gheorghe, Romania
niklas.probul [at] uni-hamburg.de
Abstract
AI in biomedicine has been a central research topic in recent years. Although there are many different techniques and strategies, the majority rely on data that is of both high quality and quantity. Despite the steady growth in the amount of data generated for patients, it is frequently difficult to make that data useful for research because of strong restrictions through privacy regulations such as the GDPR. Through federated learning (FL), we are able to use distributed data for machine learning while keeping patient data inside the respective hospital. Instead of sharing the patient data, like in traditional machine learning, each participant trains an individual machine learning model and shares the model parameters and weights. Existing FL frameworks, however, frequently have restrictions on certain algorithms or application domains, and they frequently call for programming knowledge.
With FeatureCloud, we addressed these limitations and provided a user-friendly solution for both developers and end-users. FeatureCloud greatly simplifies the complexity of developing federated applications and executing FL analyses in multi-institutional settings. Additionally, it provides an app store that makes it easy for the community to publish and reuse federated algorithms. Apps can be chained together to form pipelines and executed without programming knowledge, making them ideal for flexible clinical applications. Apps on FeatureCloud can receive certification from both internal and external reviewers to guarantee safety. FeatureCloud effectively separates local components from sensitive data systems by utilizing containerization technology, making it robust to execute in any system environment and guaranteeing data security. To further ensure the privacy of data, FeatureCloud incorporates privacy-enhancing technologies and complies with strict data privacy regulations, such as GDPR.
Keywords: federated learning, biomedicine, privacy-preserving machine learning, patient privacy
Acknowledgement: The FeatureCloud project has received funding from the European Union’s Horizon 2020 research and innovation program under grant agreement No. 826078. This publication reflects only the authors’ view and the European Commission is not responsible for any use that may be made of the information it contains. This work was developed as part of the FeMAI project funded by the German Federal Ministry of Education and Research (BMBF) under grant number 01IS21079. This work was further funded by the German Federal Ministry of Education and Research (BMBF) under grant number 16DTM100A.
