February 17, 2017 (Vancouver, BC) - A team of UBC entrepreneurs have developed a self-learning algorithm for Type I diabetics to see their future glucose values at least one hour in advance.
At any moment in time, a Type I diabetic needs to know what their glucose value is and calculate the required amount of insulin for the next few hours. Without the right amount of insulin a Type I diabetic can go into shock and coma. This is one of the greatest challenges that a Type I diabetic faces in life – a chronic condition in which the pancreas produces little or no insulin. Currently, it is almost impossible to accurately predict future glucose values because of the combination of the internal and external factors involved in glucose metabolism.
Docto, a team working out of the University of British Columbia, uses an approach of studying multiple variables as input for the learning algorithm. The variables are created using data from wearables, and used as proxies for main factors guiding the patient’s metabolism.
“The algorithm integrates participants’ activity, heart rate and blood glucose to create a baseline and monitor for deviations. Users wear two devices, a fitness tracker and a continuous glucose monitor (CGM). The combination of the devices generate more than 3000 data points per day on heart rate, blood glucose, activity and sleep,” says Amir Hayeri Founder, MSc
“The algorithm learns how specific user’s metabolism governs the blood sugar level by creating a custom library of user’s responses and constantly optimizing the use of that library in order to make most accurate predictions into the future. This allows the user to anticipate critical events like hypoglycemia in advance, and thus having enough time to avert the crisis.” says Aleksey Sher, PhD
The team’s work builds upon an emerging field in precision medicine, whose goal is to precisely diagnose and prevent complications for the chronically ill.
The team is currently piloting their app with BC Children’s Hospital.