Study of Noninvasive Glucose Monitoring Based on FBG for Ballistocardiogram

In this article, we propose and validate a noninvasive glucose monitoring method based on fiber optic sensing for heart rate variability ballistocardiography (HRVBCG). The aim was to monitor changes in blood glucose levels by analyzing the HRV parameters derived from the BCG signal. To avoid noise interference with BCG signal extraction, we introduced a variational mode decomposition parameter selection algorithm (S-VMD), which incorporates improvements in center frequency and peak index. Compared with electrocardiogram (ECG) signals, the maximum relative error in BCG signal feature point localization was 2.63%, and the maximum relative error between HRVECG and HRVBCG parameters was 7.5659%. These results indicate a high degree of consistency in the localization of characteristic points between BCG and ECG signals. This demonstrates that the proposed VMD parameter selection algorithm can effectively extract the BCG signal and retain the signal characteristic point data. To carry out the next step of correlation analysis of blood glucose value and HRV change amount. Experimental results revealed that when volunteers consumed food, their blood glucose levels increased, accompanied by an acceleration in heart rate and a decrease in HRV parameters such as standard deviation of NN (SDNN) intervals and root mean square of successive differences (RMSSDs) began to decrease. The overall correlations between the fit regression models for the HRV parameters and the amount of change in blood glucose values were all greater than 0.85, and the mean square errors were all less than 0.12. These findings suggest that the inversion of blood glucose parameters through BCG signal monitoring is feasible in the field of noninvasive blood glucose monitoring.