High-Sensitivity Non-Invasive Microwave Glucose Sensor with ZnO/CNT Composite Optimized by Deep Learning for Wearable Medical Devices.

Against the backdrop of the continuously rising global prevalence of diabetes, early detection and dynamic monitoring of the condition have become increasingly critical. This study presents a microwave sensor for detecting standard human blood glucose levels (3.9-6.1 mmol/L). The sensor integrates high-frequency electromagnetic coupling with a zinc oxide/carbon nanotube composite sensitive material, which significantly enhances detection sensitivity and response characteristics. Highly significant differences (p ≤ 0.01) were observed across the measured intervals. Furthermore, addressing the limitations of conventional methods, such as complex procedures and invasiveness, this work focuses on exploring the potential of microwave sensing for noninvasive glucose monitoring. We innovatively employed the electromagnetic simulator Sim4Life to construct human tissue models for theoretical analysis and simulation validation, assessing the sensor's detection capability by analyzing electric and magnetic fields at the resonant frequency (3 GHz). Furthermore, practical tests involving multiple volunteers were conducted to evaluate the sensor's performance under real physiological glucose variations. To overcome the challenge of low accuracy in noninvasive detection, a convolutional neural network (CNN) was introduced for data training and prediction, achieving high-precision glucose level output (R 2 = 0.98) under noninvasive conditions. This research provides a new and efficient solution for early screening and continuous monitoring of diabetes, showing significant potential for application in the field of wearable medical devices.