Intel, UPenn partner with 29 health organizations to train AI to spot brain tumors

Health and research organizations from around the globe will collaborate on building a robust AI model to identify brain tumors using the privacy-preserving method of federated learning.

Intel and the University of Pennsylvania on Monday announced they're launching a federation with 29 other healthcare and research institutions dedicated to training artificial intelligence models that can identify brain tumors. 

The group plans on training robust models using the largest brain tumor dataset to date. By employing the privacy-preserving technique of federated learning, the organizations will be able to contribute to that dataset without actually sharing their patient data. 

executive guide

What is AI? Everything you need to know about Artificial Intelligence

A guide to artificial intelligence, from machine learning and general AI to neural networks.

Read More

"AI shows great promise for the early detection of brain tumors, but it will require more data than any single medical center holds to reach its full potential," Jason Martin, a principal engineer for Intel Labs, said in a statement.  

Training robust neural networks for healthcare applications is easier said than done. Health researchers can't simply contribute data to collaborative efforts, since patient information must remain secure and private. That's largely limited the size of datasets that researchers have had to work with.  

Federated learning has offered a way around this problem. First introduced by Google in 2017, it's a learning paradigm based on decentralized data. Rather than relying on data pooled together in a single location, an algorithmic model is trained in multiple iterations at different sites. In the healthcare sector, this offers a degree of privacy for hospitals and other organizations that want to pool their resources to train a deep learning model without actually sharing their data or letting it out of their possession. 

In 2018, Intel began collaborating with the Center for Biomedical Image Computing and Analytics at the University of Pennsylvania to produce the first proof-of-concept application of federated learning to real-world medical imaging. The team demonstrated they could use federated learning to train a deep learning model to 99 percent of the accuracy of the same model trained with a traditional data-sharing method. 

The new initiative builds on this research, using Intel software and hardware to implement federated learning in a manner that Intel says provides additional privacy protections to both the model and the data.The participating organizations will build an expanded version of the  International Brain Tumor Segmentation (BraTS) challenge dataset.

The participating institutions are based in the US, Canada, the UK, Germany, Switzerland and India. Some of the institutions expected to participate in the first phase of the federation include the Hospital of the University of Pennsylvania, Washington University in St. Louis, the University of Pittsburgh Medical Center, Vanderbilt University, Queen's University, Technical University of Munich, University of Bern, King's College London and Tata Memorial Hospital.

The initiative will be funded by a three-year $1.2 million grant awarded to UPenn's Dr. Spyridon Bakas by the Informatics Technology for Cancer Research (ITCR) program within the National Institutes of Health (NIH).