According to the American Diabetes Association, Type 1 diabetes affects approximately 3 million Americans. This disease is characterized by an inability to regulate blood glucose, caused by diminished function of the pancreas to produce insulin.

One current method of monitoring blood glucose is through continuous glucose monitoring, which measures patients’ blood glucose levels at regular five-minute intervals via a skin patch sensor. Patients have an increased risk of mortality caused by dipping into the hypoglycemic (< 70 mg/dL) range, so careful accurate monitoring is critical to the well-being of Type 1 diabetes patients. Analysis of the continuous glucose monitoring signal can be challenging, as it is difficult to separate behavioral changes from purely physiological changes in the signal.

In this study, conducted by MSDS students Lev Zadvinskiy, Christian Wheeler and Gregory Gardner, individual patients’ glucose variability was analyzed based on features created from the signal, including glycemic state, risk index calculations, and blood glucose rate of change.

Results show that a known change-point caused by an inpatient study could be identified in 7 out of 15 patients. This is important as glucose variability is reflective of a patient’s physiological state, and an understanding of changes in glucose variability can lead to increased ability to maintain optimal euglycemic levels.

Default placeholder headshot.
Gregory Gardner
Default placeholder headshot.
Christian Wheeler
Default placeholder headshot.
Lev Zadvinskiy
Default placeholder headshot.
Abigail Flower
School of Data Science
Completed in: