Non-invasive blood glucose measurement technology has gained significant attention. In this project, we developed a non-contact system for measuring blood glucose information using mid-infrared light.  

In this academic year, the following work was contacted.

(1) An infrared light spectrometer was used to measure and collect physical signals linked to glucose. Then the appropriate light frequency is determined by comparing different spectral results. The relationship between mid-infrared transmittance and blood glucose concentration is studied.

(2) The experiment equipment was made, and the data collection was conducted. In the experiment, we used a self-made infrared experimental instrument to obtain the physical signal value and an invasive blood glucose meter available in the market to get the real blood glucose value. 

(3) The machine learning algorithms were used as predictors for non-contact measurement modeling. Furthermore, Clarke Error Grid Analysis (EGA) is adopted to evaluate and present the results. The EGA results show that our predicted values are almost concentrated in regions A and B, which are clinically correct and uncritical decisions. It proves that our non-invasive blood glucose monitoring system, with the convergence of smart devices and AI technology, can provide good technical support to healthcare services.