Regularized least squares regression for calibration of a photoacoustic spectroscopy based non-invasive glucose monitoring system

Near infra red photoacoustic measurements can be used to estimate blood glucose concentration in a non-invasive manner. A multi-wavelength photoacoustic measurement apparatus operating at 905 nm and 1550 nm is used to obtain measurements from glucose solutions with concentration in the physiological blood glucose range (0-500 mg/dl). Amplitude and area based features computed from photoacoustic measurements at both wavelengths are calibrated to obtain the glucose concentration. A Mean Absolute Difference (MAD) of 14.95 mg/dl (0.83 mmol/l), Mean Absolute Relative Difference (MARD) of 8.94 % and a Root Mean Square Error of Prediction (RMSEP) of 18.39 mg/dl (1.02 mmol/l) is obtained using regularised least squares regression. This performance is comparable to other non-invasive measurement methods and can be improved upon to realize a fully non-invasive glucose monitoring system.