Broadband Dielectric Spectroscopy for Quantitative Analysis of Glucose and Albumin in Multicomponent Aqueous Solution

We show the measurement of the physiological range of glucose in a multicomponent aqueous solution through multivariate analysis of broadband dielectric spectra from 500 MHz to 50 GHz. To obtain higher glucose selectivity, we apply spectral preprocessing to the dielectric loss spectra to extract feature values of glucose. Using regression models derived from different concentrations of glucose in a solution with bovine serum albumin (BSA), we analyze solutions considering the physiological range of both components. Prediction errors for the glucose and BSA concentrations are estimated to be 54 and 83 mg/dL, respectively, even for varied concentrations of each component. We also investigate the dependence of the glucose prediction error on the solution temperature. The prediction error for the glucose concentrations is estimated to be 99 mg/dL at a difference of +/- 1 K. This technique will be ease to implement with a broadband microwave sensor or with biomedical sensors that require the capability to analyze multiple components in solutions.