Design of non-invasive blood glucose sensor and development of machine learning technique for detection

Abstract

Blood glucose level monitoring advanced over the years due to an increase in health issues related to sugar, blood pressure, and so on. The Basic way is to test by a single drop of blood and other blood tests, but what if you do not need to poke or inject a needle to find an insulin level in your body. In a non-invasive way, we can calculate and estimate the accurate amount of insulin sugar level in the blood glucose by using cutting-edge technology and engineering. This Dissertation aims to develop the mm-wave at 60 GHz as a non-invasive sensor because the easurement method would use a machine learning technique to increase the accuracy of the glucose level estimation and then calculate the insulin level. In this work, we develop the antenna design and the thumb design to simulate the sensor using CST microwave studio. Cole-cole model for the blood and the analysis by varying the blood permittivity related to the glucose level and using machine learning technique to find the glucose level. The results that were obtained were promising and very relatable to the actual readings. The critical factor of this approach that made machine learning necessary to this project is a number of external independent variables that will influence the effect of output to the desired level; that’s where the machine learning helps out with great power, the accuracy of machine learning is very high with respect to your data input and validation set. The results percentage and accuracy are discussed in the conclusion .