Protecting medical data and enabling collaborative big data analysis are two cornerstones of making the use of AI a success story in healthcare. To do that, researchers led by the lab of Helmholtz AI associate Joachim Schultze have now developed Swarm Learning (SL) - a novel AI approach that combines blockchain technology with decentralized machine learning (ML).
Identifying patients with severe diseases fast and accurately is a major opportunity offered by artificial intelligence-driven data analysis. To use machine learning (ML) for this task, large data sets are required to train the models; unfortunately, the clinical data sets at any one institution alone are often not sufficiently large, so collaborations with other institutions become necessary. However, this approach poses a challenge, because the strict privacy and protection standards for medical data need to be adhered to.
Therefore, the lab led by Helmholtz AI associate Joachim Schultze (German Center for Neurodegenerative Diseases (DZNE)), together with researchers that include Fabian Theis (scientific director of Helmholtz AI and director of the Institute of Computational Biology at Helmholtz Zentrum München (HMGU)) and members of the Deutsche COVID-19 Omics Initiative (DeCOI) have developed a new AI approach called Swarm Learning (SL). SL combines blockchain technology with ML in a swarm-like network architecture that protects the privacy of the data, but also enables decentralized, collaborative big data analyses. In their publication, the researchers were able to show that SL successfully identified disease classifiers for several medical conditions (COVID-19, tuberculosis, leukaemia and lung pathologies) using blood transcriptomes and chest X-ray images as use cases.