Using swarm learning to diagnose cancer
Wednesday April 27th 2022
British researchers have reported a new method of diagnosing cancer using artificial intelligence in the laboratory.
The approach was created by a team at the University of Leeds, UK, and was reported in *Nature Medicine* on Monday.
"Artificial intelligence (AI) can predict the presence of molecular alterations directly from routine histopathology slides," the authors explain.
In order to train reliable AI systems, large datasets are needed, which raises "practical, ethical and legal obstacles", they write, such as the risk of accidental release of confidential patient data.
These obstacles could be overcome with swarm learning, they believe, in which AI models are trained in a decentralised way without combining information in a central location. A resulting algorithm without patient information is then sent to a central computer.
They found that swarm learning could be successfully used for large datasets of histopathology images from over 5,000 patients, highlighting signs of colorectal cancer.
"We trained AI models on three patient cohorts from Northern Ireland, Germany and the United States, and validated the prediction performance in two independent datasets from the UK," they write.
"Our data show that swarm learning-trained AI models outperform most locally trained models. In the future, swarm learning can be used to train distributed AI models for any histopathology image analysis task, eliminating the need for data transfer."
Research lead Dr Jakob Nikolas Kather said: “Based on data from over 5,000 patients, we were able to show that AI models trained with swarm learning can predict clinically relevant genetic changes directly from images of tissue from colon tumours.”
Saldanha, O. L. et al. Swarm learning for decentralized artificial intelligence in cancer histopathology. *Nature Medicine* 25 April 2022; doi: 10.1038/s41591-022-01768-5
Tags: Cancer | Pharmaceuticals | UK News
Printer
friendly page