Federated Machine Learning Powers World’s Largest Brain Tumor study without Compromising Patient Confidentiality

The largest brain tumor study ever conducted, using federated machine learning, was conducted without sharing any patient data

Researchers from Penn Medicine and Intel Corporation have led the largest global machine-learning effort to date. They gathered information on brain scans taken at 71 locations around the world of 6,314 patients with glioblastoma. The researchers developed a model which can improve the identification and prediction boundaries of three sub-compartments of tumors without compromising the privacy of the patient. Their findings were published in Nature Communications today.

The Perelman school of medicine at the University of Pennsylvania, said that the senior author, Spyridon B. Bakas, Ph.D. was an assistant professor of Pathology & Laboratory Medicine and Radiology. The more data we feed into machine-learning models, the better they are. This will improve our ability, to treat and remove glioblastoma from patients, with greater precision.

Many researchers studying rare conditions like GBM (an aggressive form of brain tumour) have their patient populations restricted to the institution they work at or a geographical area. Many healthcare providers are unable to collaborate across institutions due to privacy legislation such as Health Insurance Portability and Accountability Act of 1996 in the United States and General Data Protection Regulation in Europe.