Intel and UPenn head group using AI to find brain tumors

Intel and Penn Medicine are creating a federation of 29 international research institutions to train artificial intelligence models to identify brain tumors. 

About 80,000 people are diagnosed with a brain tumor each year, and more than 4,600 are children.  Early detection can lead to better outcomes, according to the American Brain Tumor Association.

The approach Intel and Penn Medicine will use is designed to prevent sensitive patient data from the leaving the hospital or research center where patient treatment occurs, according to an Intel blog.  Their approach is called federated learning.

Instead of moving data to a central place, federated workflow machine learning models move to the data at each hospital for training to protect privacy. Later, the models recombine to create a global model

“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 full potential,” said Jason Martin, a principle engineer at Intel Labs.  Intel engineers will work with Intel software and hardware in collaboration with the Perelman School of Medicine at the University of Pennsylvania. Together they will work with the 29 medical centers.

The work will be funded by the National Cancer Institute of the National Institutes of Health through the Informatics Technology for Cancer Research program.  A three-year, $1.2 million grant was awarded to Dr. Spyridon Bakas at the Center for Biomedical Image Computing and Analytics at UPenn.

No single institution can hold all the diverse data needed for such machine learning training, Bakas said in a statement.   The federation will begin developing algorithms to identify brain tumors from an expanded version of the International Brain Tumor Segmentation challenge dataset.

The researchers are based in institutions in the U.S., Canada, UK, Germany, the Netherlands, Switzerland and India.  Penn Medicine and Intel published findings last year that demonstrated the federated learning method could train a model at 99% of the accuracy of a traditional method.

The first phase of the work is being conducted by the Hospital at UPenn, 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.

 

RELATED: Police at Virginia capitol will test threat detection tech