Postprandial peak identification from continuous glucose monitoring time series

Continuous glucose monitors (CGMs) have been mainly used in medical applications to monitor blood glucose and to control insulin doses in Type 1 diabetes (T1D) patients. CGMs are becoming popular in people without diabetes and with Type 2 diabetes (T2D) due to their rising commercial availability and effectiveness. They are a useful tool for understanding individuals’ dynamic blood responses to food. However, before such information can be extracted for further analysis, the peaks must be selected automatically. Published works have focused on detecting the onset of meal intakes and calculating their carbohydrate load to assist glucose control in T1D management. This work aims to develop a threshold-based algorithm for entire postprandial peak identification, including starting and endpoints, from data obtained from people with different glucose tolerance levels. The algorithm achieved promising performance using an individual threshold-based approach, with recall and precision rates of 0.84 and 0.85, respectively