Share Tweet Share Email Print

New diabetes monitor predicts blood sugar spikes 30 minutes faster

Data-driven model could help prevent risks associated with low blood sugar
image of three scientists, standing at table with equipment, looking up at camera and smiling
(Left to right) Clara Mosquera-Lopez, Ph.D., Nichole Tyler, and Peter Jacobs, Ph.D., in their offices at OHSU, May 13, 2019. The three are part of a team that created a new algorithm which allows diabetes monitors to predict blood sugar spikes up to 30 minutes faster. (OHSU/Kristyna Wentz-Graff)

Hypoglycemia, a condition caused by a very low level of blood sugar, is a common occurrence for the approximately 40,000 Americans diagnosed with type 1 diabetes each year. If not immediately treated, symptoms of hypoglycemia can range from fatigue or irritability, to seizure, loss of consciousness, and in some cases, death.

Individuals at risk of hypoglycemic events can manage the impacts through clinician-recommended diabetes management plans, as well as a mechanical device known as a continuous glucose monitor, or CGM. These tools are effect in providing real-time notification of low glucose levels, however, patients must act quickly.

To better facilitate blood sugar level management, researchers at the Artificial Intelligence for Medical Systems Lab at OHSU in Portland, Oregon, have developed an innovative algorithm, called Glucop30, that achieves advanced prediction accuracy of glucose levels 30 minutes sooner.

The details of this advancement are published in the IEEE Journal of Biomedical Health Informatics.

person with diabetes monitor attached, and holding up a phone with the readout of their glucose levels
People with diabetes monitor hypoglycemic events using a device known as a continuous glucose monitor, or CGM. A new algorithm called Glucop30 will allow CGM devices to predict hypoglycemia events more accurately up to 30 minutes in advance, which helps patients better facilitate blood sugar level management. (OHSU/Peter Jacobs)

Based on innovative machine learning methods and a large dataset, acquired from the Tidepool Big Data Donation project, Glucop30 achieves a root-mean-square-error of 7.5 mg/dL, which is approximately 45 percent more accurate than any commercial prediction algorithm currently available. It also forecasts hypoglycemia with a sensitivity of more than 90 percent and only one false alarm per week.

“When integrated into a CGM, Glucop30 will allow individuals the power to anticipate their glucose levels with state-of-the-art precision,” said AIMS principal investigator Peter Jacobs, Ph.D., an assistant professor of biomedical engineering in the OHSU School of Medicine. “The ability to accurately predict a high- or low blood sugar level up to a half hour in advance means that an individual can simply adjust diet or insulin dosage appropriately to eliminate a potential hypoglycemic event before it happens.”

The development of Glucop30 was funded by The Leona M. and Harry B. Helmsley Charitable Trust (Grant 2018PG-T1D001) and the National Institute of Diabetes and Digestive and Kidney Disease of the National Institutes of Health (Grant 1 R01Dk120367-01). 

The technology is currently being considered for licensing by several companies in the field of glucose management.

Previous Story Near-death comeback gives man new lease on life Next Story New eye clinic building meets construction milestone