A novel AI-enhanced microwave sensor employing defected ground structure for non-invasive glucose monitoring

Abstract

Diabetes is a chronic disease that affects millions of people worldwide and significantly reduces quality of life. One of the most critical aspects of managing this condition is the accurate, continuous, and reliable monitoring of blood glucose levels. Fluctuations in glucose concentration can lead to both short-term complications and long-term irreversible organ damage. Currently, traditional glucose monitoring methods rely mainly on blood samples obtained through finger-pricking. While these methods are accurate, their invasive nature reduces user comfort, causes pain, poses a risk of infection, and negatively affects patient adherence in the long run. In response to these limitations, non-invasive glucose monitoring technologies, particularly those based on microwave and radio frequency (RF) sensor systems, have gained increasing attention. However, most reported systems still face challenges in achieving high sensitivity and stability under realistic physiological conditions. In this study, we introduce a novel hexagonal microstrip patch antenna with a chaotic Defected Ground Structure (DGS) based on a Duffing chaotic attractor, specifically designed for non-invasive blood glucose sensing. Unlike conventional DGS-based sensors, our chaotic DGS approach enhances tissue penetration and significantly improves sensitivity to subtle dielectric variations caused by glucose concentration changes. The sensor, optimized for finger placement, operates in the 4–5 GHz range to ensure effective tissue coupling. Experimental validation using multi-layer tissue-mimicking phantoms demonstrated the sensor’s ability to differentiate clinically relevant glucose levels (50–200 mg/dL), achieving a high sensitivity of 0.950 MHz/(mg/dL).