Accuracy Enhancement for Noninvasive Glucose Estimation Using Dual-Wavelength Photoacoustic Measurements and Kernel-Based Calibration

Frequent monitoring of blood glucose levels is an essential part of diabetes care, but the invasiveness of current devices deters regular measurement. Noninvasive measurement techniques are painless to implement and rely on changes in sample properties to estimate glucose concentration. However, such methods are affected by the presence of different biomolecules, resulting in an increased estimation error and necessitating calibration to obtain accurate glucose concentration estimates. The use of photoacoustic spectroscopy for continuous noninvasive glucose monitoring is studied through measurements on different sample media. In vitro photoacoustic measurements taken from aqueous glucose solutions, solutions of glucose and hemoglobin, and whole blood samples at multiple excitation wavelengths show amplitude and area-based signal features to rise with the increase in sample glucose concentration. The calibration of photoacoustic measurements from glucose solutions using Gaussian kernel-based regression results in a root mean square error (RMSE), mean absolute difference (MAD), and mean absolute relative difference (MARD) of 7.64 mg/dl, 5.23 mg/dl, and 2.07%, respectively. Kernel-based calibration also performs well on solutions of glucose and hemoglobin, and whole blood samples, resulting in lower estimation errors than that of previous efforts and with glucose estimates being in the acceptable zones of a Clarke error grid (CEG). It allows for individual calibration of photoacoustic measurements in vivo, resulting in an RMSE, MAD, and MARD of 19.46 mg/dl, 10.79 mg/dl, and 7.01%, respectively, with 89.80% of the estimates being within Zone A of the CEG. The improvement in estimation accuracy with dual-wavelength photoacoustic measurements and kernel-based calibration would enable continuous noninvasive glucose monitoring, facilitating improved diabetic care.