New effort to map brain tumors puts an AI approach to the test

Intel, Penn Medicine's project is based on a technique known as federated learning.

Researchers from Intel Labs and the Perelman School of Medicine at the University of Pennsylvania (Penn Medicine) are cooperating to create a new Artificial Intelligence or AI model that can spot brain tumors after the success of AIs in spotting skin cancer, breast cancer, and lung cancer.

According to a report by Engadget, Penn Medicine will be leading 29 different international research and healthcare institutions in creating the AI model with the help of Intel's hardware and software. The AI model will be trained on the most massive brain tumor dataset, but patients don't have to worry as they will do it without sharing any sensitive patient data, including the patient's name and address, to guarantee patient confidentiality.

Furthermore, the project is based on a technique known as federated learning. How this works is that an AI model is trained across decentralised servers that allow hospitals and medical experts to work together without sharing patient data. This will then create a more massive data set than anything an institute can hold on its own, which would then be accessible in different places across the world from the United States to Canada to Switzerland, the Netherlands, Germany, and India.

Institutions expected to join in the first phase of the project include the Hospitals of the University of Pennsylvania, King's College London, the University Pittsburgh Medical Center, Washington University in St. Louis, and 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," said Intel's principal engineer.Jason Martin in a press release, 

The National Institutes of Health (NIH) supports the project and is offering the group with a three-year $1.2m grant. The grant is awarded to the principal researcher, Dr. Spyridon Bakas, from the Center for Biomedical Image Computing and Analytics (CBICA) of the University of Pennsylvania.

Nevertheless, it is unclear when the project will be completed and released, but it will build on the initial research that was first presented in 2018 during the International Conference on Medical Image Computing and Computer-Assisted Intervention.

It made sense to see Intel and Penn Medicine working together for this AI model as they were among the first to lead research on the benefit of using federated learning in healthcare. In their study, they demonstrated how federated learning could be used to acquire 99% accuracy.

Apple and Google have been using federated learning before to help them improve the quality of words, phrases, emoji, and music predictions, but the technique is now being used in the medical domain to help experts in various fields.

This partnership is announced today to coincide with Brain Tumor Awareness Month, according to VentureBeat. According to the American Brain Tumor Association (ABTA), over 700,000 Americans are living with brain tumors today, and there will be another 80,000 people who will be diagnosed this year. They have also predicted that 16,000 people with brain tumors will die this year due to their condition.

Similar articles