Intel Labs is partnering with University of Pennsylvania’s Perelman School of Medicine (Penn Medicine) to co-develop artificial intelligence (AI) models that can identify brain tumors and protect patients’ privacy at the same time. The collaboration was announced on 11 May and a federation of 29 other healthcare and research institutions across the US, Canada, Europe, and India are also be involved. The study is funded by the Informatics Technology for Cancer Research (ITCR) program under the National Institutes of Health (NIH) through a three-year, $1.2 million grant.

Federated Learning to ascertain data access and privacy

Based on the figures provided by the American Brain Tumor Association (ABTA), there are close to 80,000 newly diagnosed brain tumor cases every year and more than 4600 of them are children. Researchers turned to AI for assistance in early detection and better patient outcomes. However, according to the project’s Principal Investigator Dr. Spyridon Bakas, generally, the scientific community acknowledges that machine learning needs to be fed on “ample and diverse data” which “no single institution can hold”. As such, the involvement of 29 institutions and getting hold onto a large dataset are vital.

Besides, medical and other sensitive data need to remain private and protected. So, the research alliance will be leveraging on Intel’s hardware and software to create brand new AI models using the largest brain tumor dataset to date – the International Brain Tumor Segmentation (BraTS) challenge dataset. Federated learning will be implemented in way which harnesses additional privacy protection to both the models and data.

Federated Learning is a machine learning technique whereby an algorithm is being trained across various decentralized entities holding onto their respective sets of data without the need to share any information or in this case, patients’ medical records. It will “allow medical researchers access to vastly greater amounts of healthcare data while protecting the security of that data”.

AI’s assistance for early detection and better patient outcomes

Jason Martin, Principal Engineers at Intel Labs added in the press release, “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”.

Back in 2018, at the International Conference on Medical Image Computing and Computer Assisted Intervention (MICCAI), Intel Labs and Penn Medicine presented on the use of federated Learning in medical imaging. They claimed to be the first in the field to do so and model trained using federated learning is able to achieve more than 99% accuracy as compared to those trained via typical machine learning approach. Related findings were also detailed in a paper last January.

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 will be the first institutions to participate in initiating the first phase of the study.


Author Bio

Hazel Tang A science writer with data background and an interest in the current affair, culture, and arts; a no-med from an (almost) all-med family. Follow on Twitter.