Improving Accuracy and Precision of Glucose Sensor Profiles: Retrospective Fitting by Constrained Deconvolution

Frequent and accurate reference measurements of blood-glucose (BG) concentration are key for modeling and for computing outcome metrics in clinical trials but difficult, invasive, and costly to collect. Continuous glucose monitoring (CGM) is a minimally-invasive technology that has the requested temporal resolution to substitute BG references for such a scope, but still lacks of precision and accuracy. In this paper, we propose an algorithm that retrospectively reconstructs a reliable continuous-time BG profile for the aforementioned purposes, by simultaneously exploiting the high accuracy of (possibly sparse) BG references and the high temporal resolution of CGM data. The algorithm performs a constrained semiblind deconvolution in two steps: first, it estimates the unknown parameters of a model accounting for plasma–interstitum diffusion and sensor inaccurate calibration; then, it estimates BG performing a regularized deconvolution of CGM data, subject to the additional constraint that the reconstructed BG profile has to lay within the confidence interval of the available BG references. The algorithm was tested on 24 datasets collected in a 20 h clinical trial where CGM records and a median of 13 BG samples per day were available. Mean absolute relative deviation was reduced (from 15.71% to 8.84%) with respect to unprocessed CGM and so did the error in the evaluation of the outcomes metrics (e.g., halved the error in the time-in-hypo assessment). The reconstructed BG profile, in view of its improved accuracy and precision, is suitable for clinical trial assessment, modeling and other offline applications.