基于光声光谱联合主成分回归法的血糖浓度无损检测研究

Non-Invasive Detection of Blood Glucose Concentration Based on Photoacoustic Spectroscopy Combined with Principle Component Regression Method

利用可调谐脉冲激光器激发联合聚焦超声探测器前向探测模式搭建了一套血糖光声无损检测实验装置。为了测试该装置的可靠性,实验中利用532nm泵浦Nd∶YAG调Q脉冲激光器激发不同浓度的葡萄糖水溶液产生实时光声信号;采用脉冲激光在近红外波段1 300~2 300nm内固定间隔波长10nm扫描方式激发不同浓度的葡萄糖水溶液,获取了不同波长下的葡萄糖光声峰峰值,利用差谱方法筛选出了多个葡萄糖的特性波长;然后采用主成分回归算法优选了三个特性波长,并建立了浓度梯度与对应三个优选波长光声峰峰值之间的数学校正模型。实验表明,葡萄糖水溶液的光声信号符合弱吸收介质的柱状光声源模型;利用建立的校正模型对校正集和预测集的葡萄糖浓度预测结果表明,葡萄糖浓度的校正和预测均方根误差均小于10mg·dl^(-1),相似系数为0.993 6。

This paper presents a photoacoustic noninvasive setup of detecting blood glucose based on the tunable pulsed laser cou- pled with the confocal ultrasonic transducer and the forward detection model. To validate the reliability of the setup, in the ex- periments, the different concentrations of glucose aqueous solution are excitated by the Q-switched 532 nm pumped Nd ： YAG pulsed laser to generate the time-resolved photoacoustic signals. And the glucose aqueous solutions are scanned by the tunable pulsed laser in the infrared waveband from 1 300 to 2 300 nm with the interval of 10nm and the photoacoustic peak-to-peak values are gotten. The difference spectral method is used to get the characteristic wavelengths of glucose, and the principle component regression algorithm is used to determine three optimal wavelengths and establish the correction mathematical model between the photoacoustic peak-to-peak values and the concentrations. The experimental results demonstrate that the mechanism of the photoacoustic signal is agreement with the cylindrical model, and the predicted results of the correction and prediction samples based on the established correction model demonstrate that the root-mean-square error of correction and prediction are all less than 10 mg ·dl^-1, the correlation coefficient reaches 0. 993 6.