Abstract/Summary

Introduction. Continuous glucose monitoring (CGM) for diabetes combines non-invasive glucose biosensors, continuous monitoring, cloud and analytics to connect and simulate a hospital setting in a person’s home. CGM systems inspired analytics methods to measure glycemic variability yet existing glycemic variability analytics methods disregard glucose trends and patterns, hence they fail to capture entire temporal patterns nor provide granular insights of glucose fluctuations. Aim. To propose a machine learning-based framework for blood glucose fluctuation pattern recognition which enables a more comprehensive representation of glycemic variability profiles that could present detailed fluctuation information; be easily understood by clinicians; and to provide insights on patient groups based on time in blood fluctuation patterns. Methods. 1.5 million measurements from 126 patients in the UK with type 1 diabetes were collected, and prevalent blood fluctuation patterns were extracted using dynamic time warping. The patterns were further validated in 225 patients in US with type 1 diabetes. Hierarchical clustering was then applied on time in patterns to form four clusters of patients. Patient groups were compared through statistical analysis. Results. Six patterns depicting distinctive glucose levels and trends were found and validated. Based on which, four glycemic variability profiles of type 1 diabetes patients were found. They were significantly different in terms of glycemic statuses such as diabetes duration, HbA1c and time in range, and thus had different management needs. Conclusion. The proposed method can analytically extract existing blood fluctuation patterns in CGM data. Thus, time in patterns can capture a richer view of patients’ glycemic variability profile. Its conceptual resemblance with time in range along with rich blood fluctuation details makes it more scalable, accessible, and informative to clinicians.