Photo by John Sattler

Efforts to bring artificial intelligence to insulin pump therapy could make life easier for patients with type 1 diabetes—and their doctors.
Frank Schwartz, M.D., associate professor of endocrinology and director of the Appalachian Rural Health Institute (ARHI) Diabetes/Endocrine Center, is working with diabetologist Jay Shubrook, D.O. (’96), assistant professor of family medicine, and Cindy Marling, Ph.D., a computer scientist at the Russ College of Engineering and Technology, to design the smarter pump therapy.

Today’s insulin pumps measure glucose levels, which project a dizzying array of data difficult even for an endocrinologist to interpret. They do not adjust insulin levels automatically or make treatment suggestions, but the software Schwartz, Shubrook and Marling are developing will do just that. “The technology will remember patterns and interpret daily lifestyle information to make suggestions about how patients can better manage their glucose,” Schwartz says.

In last year’s initial study, 20 patients documented their treatments and lifestyles—diet, stress, sleep and exercise habits—in an online database for six weeks. Schwartz and Shubrook reviewed the data to propose action plans for each patient. Based on the outcomes of these plans, the researchers began teaching the insulin pump therapy program to recognize problematic patterns and possible solutions.
“If someone’s glucose levels are high at night, Dr. Schwartz may tell them to increase their pump basal rates of insulin infusion,” Marling says. If the recommendation works, they plug it into the computer program as a solution to this problem.

The team has identified 50 problems and solutions so far. This year, they will follow up with two studies. The first, which involves 60 patients, will test and validate the software. The second will expand the software with the help of 28 new patients.

The project receives support from OU-COM, the Russ College and the ARHI Diabetes Research Initiative, and equipment from Medtronic MiniMed, an insulin pump manufacturer.
Initial findings will appear in the Journal of Diabetes Science and Technology, Computational Intelligence, and the proceedings of the European Conference on Case-based Reasoning.

“We’ve registered a patent application, and we’re meeting with insulin pump companies in hopes of forming a collaborative partnership,” Schwartz says. “Putting the technology in the hands of a national manufacturer would help it reach more patients.”

A version of this story originally appeared in the Autumn/Winter 2007 issue of Perspectives magazine.