Determination of photoacoustic glucose characteristic wavelengths based on synergy interval partial least square algorithm

Abstract

Determination of blood glucose characteristic absorption wavelengths is one of most important steps in the photoacoustic measurement of diabetes mellitus. In this paper, a kind of custom-built photoacoustic measurement system of glucose solutions based on OPO pulsed laser and ultrasonic detector with central frequency of 10MHz was established. The photoacoustic peak-to-peak values of glucose solutions with six different concentrations were obtained under the scanning wavelengths from 1300 nm to 2300nm. To obtain the characteristic wavelengths of glucose, the differential spectrals between the photoacoustic peak-to-peak spectrals with different concentrations and that of the pure water were firstly gotten. Then, the synergy interval partial least square (siPLS) algorithm was used to optimally choose the characteristic wavelengths of glucose. The mean method was used to pre-process the photoacoustic peak-to-peak data. The photoacoustic peak-to-peak spectrals of glucose were divided into different sub-intervals, and different components and different segments were chosen. In addition, different synergies of sub-intervals were test to get the maximum correction coefficients for the calibration samples. After computing and compare with each other, 30 sub-intervals, 7 components, 4 segments, and 3 synergies sub-intervals were chosen because the root mean square error of cross validation (RMSECV) and the prediction bias are least. Finally, combined with the differential spectral, the optimal characteristic absorption wavelengths of glucose were determined.