Noninvasive blood glucose monitoring using a dual band microwave sensor with machine learning.
Journal:
Scientific reports
PMID:
40346256
Abstract
The potential for continuous non-invasive blood glucose monitoring has attracted a lot of interest in the field of medical diagnostics. This paper provides a new shape of a dual-band bandpass filter (DBBPF) acting as a microwave transmission line sensor for continuous non-invasive blood glucose monitoring operating at 2.45 and 5.2 GHz. The proposed system uses the interaction between biological tissues and microwave signals to correctly assess blood glucose levels. The proposed dual-band bandpass filter (DBBPF), comprises three split ring resonator (SRR) cells with different dimensions. It is designed to operate as a sensor with improved sensitivity, compact dimensions, and a high-quality factor. It also ensures a reasonable bandwidth for lower and higher bands of 8.6 and 2%, respectively in the industrial, scientific, medical band, and the wireless local area network (ISM and WLAN) Bands. A dual-band filter enhances measurement sensitivity and specificity by targeting specific frequency ranges where glucose exhibits distinctive dielectric responses, thereby providing redundant data points for accurate glucose level determination. Glucose concentrations can be evaluated by measuring the changes in the dielectric properties of blood by sending microwave waves through the body and assessing the collected S-parameter signals. The measurement parameters encompass the reflection, phase, magnitude, as well as transmission parameters. This yields multiple evaluations of the glucose-induced alterations. Simulations are validated through laboratory measurements incorporating a phantom finger model for capturing realistic outcomes. Machine learning models are employed to analyze the sensor data, improving the accuracy of diabetes detection. Simulations are validated through laboratory measurements incorporating a phantom finger model for capturing realistic outcomes. A Cole-Cole model, implemented using MATLAB, is utilized for the phantom finger model. The main results reveal the success of the proposed transmission-based microwave glucose sensing, with a remarkable sensitivity of 1 ~ 1.5 dB for glucose level change up to 200 mg/dL.