Learning the Sensing Delay for Personalized Continuous Diabetes Monitoring

Interstitial fluid (ISF) serves as a rich source of biomarkers, enabling minimally invasive continuous health monitoring through ISF sensors. Despite their potential advantages, ISF sensors face a major challenge related to the delay in the transfer of target analytes from blood to ISF, compared to blood-based measurements. Particularly, this delay can vary significantly from subject to subject. While machine learning algorithms have been developed for continuous glucose measurement within ISF, these algorithms have not considered this delay. In this paper, we quantify the delay between ISF and blood-based detection of glucose and ketone bodies in diabetic rats using decision-tree based algorithms. Accounting for this delay can eventually improve the accuracy of ISF sensors for continuous health monitoring of individual patients