Machine Learning Approach for Non-Invasive Detection of Blood Glucose Concentration using Microwave

Abstract

A non-invasive blood glucose measurement method based on microwave transmission and applying machine learning technique to the data obtained, is proposed for monitoring the patients' blood glucose level. Using this technique, non-invasive measurement of the blood glucose concentration at the earlobe portion can be realized and put into use by analysing the reflected microwave signals. A third order Cole-Cole equation is derived to model the dielectric properties of human tissues. Particle swarm optimisation technique is used to determine the coefficients for the glucose concentration dependent equations. With these estimated dielectric values of human tissues, human earlobe portion is modelled and tested at a wide range of frequencies to analyse for the region of linearity. It was observed that frequency range from 6–8 GHz shows the linearity and calculating the complex permittivity values with the help of transmission parameters. Applying Machine-learning technique to the above process can facilitate a real-time processing which in turn is able to alert patients during hyperglycemia conditions, and can suggest a precise dose of insulin to intake.