Research Brief: Evaluating use of new AI technology in diagnosing COVID-19
'Federated learning' could provide important solution for the future of AI healthcare
MINNEAPOLIS/ST. PAUL (11/17/2022)—A team of University of Minnesota Twin Cities researchers led a study evaluating the use of federated learning, an artificial intelligence (AI) technique, to diagnose COVID-19 in chest x-rays. Their findings could lead to more accurate, unbiased machine learning models for diagnosing diseases and improving patient care in real-world healthcare settings.
The study is published in the Journal of the American Medical Informatics Association, a peer-reviewed scientific journal covering research in the field of medical informatics.
The project is co-led by Ju Sun, an assistant professor in the University of Minnesota Department of Computer Science and Engineering, and Christopher Tignanelli, an associate professor in the University of Minnesota Medical School. Both are leaders of the Medical School's Program for Clinical AI in the Center for Learning Health System Sciences.
“Federated learning is an important future solution for AI in healthcare,” Tignanelli said. “As all machine learning methods benefit greatly from the ability to access data that provides closer to a true global distribution, federated learning is a promising approach to obtain powerful, accurate, safe, robust and unbiased models.”
Federated learning enables multiple parties to develop and train AI models collaboratively without the need to exchange or centralize data sets, which helps protect sensitive medical data and may open new research and business avenues to improve patient care. Using the technique in diagnostic x-rays could allow for improved medical image and text analysis, collaborative and accelerated drug discovery, decreased cost and time-to-market for pharmaceutical companies, among other benefits.
“We’re proud to be among the first teams implementing and further refining federated learning in real-world healthcare settings, with the strong support of industrial partners including Nvidia and Cisco,” said Sun. “Data is the oil for modern AI, and federated learning makes the perfect oil refinery to advance AI for healthcare.”
State-of-the-art algorithms are usually evaluated on carefully curated data sets originating from only a few sources, rather than truly representative data. This can introduce biases where demographics or technical imbalances skew predictions and adversely affect the accuracy for certain groups or sites. Researchers say to capture subtle relationships between disease patterns, socio-economic and genetic factors, and complex and rare cases, it is crucial to expose a model to diverse cases.
“We truly believe the potential impact on precision medicine and ultimately improving medical care is very promising,” Tignanelli said.
This research is a collaboration between the U of M, M Health Fairview, Emory University, Indiana University School of Medicine and University of Florida.
Part of the funding for this research was provided by Cisco.
Read the full news release on the U of M Medical School website.