Estimation of blood glucose levels by sweat sensing based on biophysical modeling of glucose transport

Monitoring glucose concentration in sweat might represent a non-invasive alternative to traditional invasive blood sampling for diabetic patients. The relationship between glucose concentration in blood and in sweat is largely unknown, and methods that can estimate blood glucose levels from measured sweat glucose levels are crucial. In this paper, we present a novel method that was developed by first estimating sweat glucose concentration from blood inputs. Such a method builds on a sweat gland model proposed by La Count et al., additionally considering the dilution effect of different sweat rates between the interstitial space and sweat glands on glucose concentration. The sweat glucose concentration estimated by our model shows an average root mean square percentage error (RMSPE = 11%± 6%), smaller than the original model (RMSPE=21%± 9%). This enables a more accurate estimation of the relationship between glucose levels in sweat and blood. Secondly, solving the inverse problem by an iterative optimization method, we obtained the average RMSPE of blood glucose concentration estimated from the sweat glucose concentration equal to 16.7%± 9.2%. These results show satisfactory prediction accuracy. Our study is the first to realize the estimation of blood glucose changes with high precision based on known sweat glucose concentrations. Furthermore, this research could be significant for the implementation of semi-continuous and prolonged diabetes monitoring by sweat sensing technology.