Non-invasive determination of blood glucose level using narrowband microwave sensor

Abstract

The diabetic patients use invasive technique to monitor their Blood Glucose Level (BGL) on a regular basis. This paper mainly presents a system that estimates BGL noninvasively using microwave techniques. A narrowband microstrip antenna resonating at 1.3 GHz is used as a microwave sensor. If human finger containing a specific value of the BGL is placed on a radiating patch of narrowband microstrip patch antenna, microwave sensor structure, then the near field of this radiating patch antenna structure gets interact with human finger and results change in electrical characteristics of the antenna. These electric changes are in relation to blood permittivity changes due to variation in BGL value. The change in electric characteristics of narrow band antenna microwave structure results corresponding frequency shift. The dataset of more than 200 individuals are generated where corresponding frequency shifts for various BGLs are recorded. The prepared dataset provided linear relationship between reference BGL and corresponding frequency shift. Regression analysis for BGL estimation resulted coefficient of determination with value 0.75. The significant improvement in this coefficient of determination is obtained by subband frequency analysis, which resulted in close approximation with expected value. The Surveillance Error Grid (SEG) and Mean Absolute Relative Difference (MARD) analyses are performed on prepared dataset to validate and verify clinical acceptance of microwave based non-invasive BGL estimation system. The developed antenna model has the MARD value of 4.204% and existing split ring resonator non-invasive antenna has 12.5% MARD value for the whole dataset.