Innovations at the intersection of tech and medicine continue to emerge daily. On Monday, Intel Labs announced an initiative with the University of Pennsylvania (Penn Medicine) that uses artificial intelligence (AI) to identify brain tumors while preserving patient privacy.
The unique approach uses a technology called federated learning. It has funding from the National Institutes of Health (NIH) and the Informatics Technology for Cancer Research (ICTR) program.
Pioneering Advanced Detection
According to data from the American Brain Tumor Association (ABTA), 80,000 individuals will be diagnosed with a brain tumor in 2020. More than 4,600 of those patients will be children. As is the case with all cancers, early detection and treatment is key to ensuring survival.
In recent years, improving artificial intelligence algorithms have grown skilled at identifying tumors and areas with the potential of turning into one. Training these systems requires massive amounts of relevant medical data. Few healthcare organizations have enough data to effectively train an AI cancer recognition system.
As such, Intel Labs is working with Penn Medicine to create a federation of 29 international institutions. Healthcare and research organizations from the United States, Canada, the United Kingdom, Germany, the Netherlands, Switzerland, and India are all represented. Using data from these organizations, the duo will train advanced AI algorithms to recognize brain tumors.
Jason Martin, principal engineer at Intel Labs, says, “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.”
The goal of the partnership is to produce a new, state-of-the-art AI model for detecting brain tumors. By using data from so many organizations, the system will be the best-trained to date.
Of course, any time medical information is a part of research, privacy is a major concern. While Intel Labs and Penn Medicine are taking an innovative approach to cancer detection, they are also doing so with privacy at the forefront.
The duo is using federated learning to accomplish this. In the medical imaging domain, Intel Labs and Penn Medicine were the first to publish a paper about the tech. It demonstrated that federated learning could be used to train an AI model with more than 99 percent accuracy compared to a traditional, non-private training method.
An Intel press release describes federated learning as “a distributed machine learning approach that enables organizations to collaborate on deep learning projects without sharing patient data.”
Essentially, the AI model goes to a facility’s data for training rather than each organization sending their data to a central location. After training on various datasets, the system recombines the data and its learning into one global model.
The project will utilize Intel’s software and hardware to carry out federated learning in a way that ensures privacy protection of the sensitive data being used. For the future of brain tumor screening, the collaboration is noteworthy. It also demonstrates another of the many ways that tech can be used to make healthcare safer, more accurate, and more efficient.