Design and development of a high-sensitivity microstrip sensor for glucose detection

This study presents a high-sensitivity microstrip sensor designed for non-invasive blood glucose monitoring. Since blood is primarily water, distilled water with dissolved glucose served as a reliable substitute in the experiments. The sensor uses two microstrip antennas operating at three resonant frequencies: 5.8 GHz, 6.9 GHz, and 8.1 GHz. Key parameters, including the transmission coefficient (S21), were recorded to evaluate the sensor’s response to varying glucose concentrations. To ensure experimental stability, a custom housing made of polylactic acid (PLA) was fabricated, providing secure positioning of the sensor components while maintaining environmental consistency. Glucose solutions with values between 5 mg/dL and 500 mg/dL were made. Each sample was tested three times, for a total of 300 frequency response measures. These data served as the foundation for training and validating a multi-layer perceptron (MLP) neural network, which demonstrated remarkable predictive accuracy with a mean relative error (MRE) of 1.14%. The sensor exhibited an average sensitivity of 1.27% per mg/dL across the tested range, underscoring its capability to detect small variations in glucose concentration effectively. These results indicate the sensor’s potential for non-invasive, rapid, and cost-effective glucose monitoring. It also shows versatility for applications in food quality control and industrial processes.