A Comparison of algebraic and dynamic models to estimate blood glucose and A1C from CGM time series

Continuous glucose monitoring (CGM) traces provide a discrete representation of an individual’s interstitial fluid glucose concentration (ISF). The continuous blood glucose concentration (BG) however, is a more useful and desired physical quantity to be measured. Traditionally, BG was the primary metric for diagnosing Diabetes Mellitus (DM). There are however, no clear and reliable method for constructing the continuous BG from the measured discrete ISF. It is still an active area of research. We developed a method based on trajectory optimization where we assume a simple differential equation describing the BG-ISF dynamics and estimate the continuous BG that would generate discrete ISF measurements as close as possible to the CGM trace.