Statistical analysis based online sensor failure detection for continuous glucose monitoring in type I diabetes

• A statistical analysis based sensor failure detection method is proposed.
• It can work as a high-level monitoring tool to evaluate reliability of CGM.
• Different types of monitoring charts are developed for two important typical disturbances.
• The consistency of the glucose dynamics is checked from different perspectives.
• The feasibility and performance are illustrated for 10 in silico subjects.

Wide use of continuous glucose monitoring (CGM) provides time-series sensor data and knowledge about the underlying correlations of glucose concentrations and their progressing dynamics over time direction. From self-monitoring of blood glucose to continuous glucose monitoring, sensor performance is key for successful clinical use. However, sensor failures have not been analyzed and detected before use of CGM data for glucose control, although they are very common problems in real-world cases and may result in unreliable CGM data and thus bad control performance. In the present work, a statistical analysis based failure detection method is proposed as a high-level glucose monitoring tool to identify sensor problems online only using CGM data. By analyzing CGM data, different types of monitoring charts (i.e., fault detection charts) are developed and sensor abnormality is indicated during online monitoring by comparing monitoring profiles against the predefined confidence limits. Comparison is also made between different monitoring charts for two important typical disturbances. Different from the conventional glucose monitoring which worked as a low-level monitoring tool and focused on direct real-time display of CGM readings, the proposed method focuses on detecting undesirable sensor disturbances by analyzing the underlying time-wise glucose autocorrelations. The feasibility of the proposed method to serve as a completely new glucose monitoring engine is successfully assessed using CGM data collected from the Food and Drug Administration (FDA)-accepted University of Virginia/University of Padova (UVa/Padova) metabolic simulator.