近红外无创血糖检测系统信号提取能力的验证

Verification of Signal Extraction Capability of Near-Infrared Non-Invasive Blood Glucose Detection System

近红外无创血糖检测技术的精度目前仍达不到临床应用的需求,主要的原因为:一方面人体血糖信号微弱,并且血液中一些组分的近红外吸收谱带与葡萄糖的吸收谱带存在重叠,为了从光谱中解析葡萄糖浓度信息,通常会采用如偏最小二乘(PLS)等多变量回归的方法。另一方面,在实际测量中,必然会存在光源漂移、测量条件变化等背景干扰,这些背景干扰对测量造成的影响往往大于血糖浓度变化引起的光谱响应,因此在建立血糖预测模型之前必须对这些背景干扰进行有效地控制与消除,否则使用多变量回归方法所建立的血糖预测模型中很可能存在伪相关。因此为了更好地实现无创血糖检测,测量系统本身必须具备很高的血糖检测能力,并且在控制测量条件尽可能稳定的前提下,采用合适的数据处理方法去除绝大部分的背景干扰。为此,对自行研发的无创血糖检测系统的血糖检测能力进行了评估,证明了本系统能够达到较高的测量精度;对三名健康的受试者开展了口服糖耐量试验(OGTT)以及口服水耐量试验(OWTT),对比了本系统在两个不同光源-探测器距离下测得的OGTT与OWTT的光谱数据,发现在两个光源-探测器距离下OGTT吸光度变化量的方差值都大于OWTT,但由于受到背景干扰的影响,三名受试者的吸光度变化量的方差值随波长的分布特点存在较大的差异;对两个光源-探测器距离下的光谱数据进行差分处理,对比分析OGTT与OWTT的差分光谱数据,结果表明OGTT差分吸光度变化量的方差值远大于OWTT,且三名受试者的差分吸光度变化量的方差值随波长分布特性与葡萄糖溶液的吸收特性基本一致,证明使用自行研发的无创血糖检测系统配合差分处理方法,能够有效去除背景干扰,提取血糖浓度信息。

Near-infrared non-invasive blood glucose detection technology still fails to meet the accuracy required for clinical application.The main reason on the one hand is that the human blood glucose signal is weak,and the near-infrared absorption band of some components in blood overlaps with the absorption band of glucose.Multivariate regression methods such as partial least squares(PLS)are usually used to extract the glucose concentration information from spectral data.On the other hand,in the measurement process,background interference such as light source drift and measurement condition changes is inevitable.The impact of background interference on the measurement is often stronger than the spectral response caused by the changes in blood glucose concentration.Therefore,these background disturbances must be effectively controlled and eliminated before establishing the blood glucose prediction model.Otherwise,there will likely be pseudo correlations exiting in the blood glucose prediction model established by a multivariate regression method.Therefore,in order to non-invasively detect the blood glucose signal even better,the measurement system itself should have high blood glucose detection capability,and under the premise of keeping the measurement conditions as stable as possible,appropriate data processing methods should be used to eliminate most of the background interference.To this end,this paper evaluated the blood glucose detection capability of the self-developed non-invasive blood glucose detection system,proved that the system could achieve high detection accuracy.Then oral glucose tolerance test(OGTT)and oral water tolerance test(OWTT)were performed on three healthy subjects,and the spectral data of OGTT and OWTT at two different source-detector distances were compared,and the analysis results showed that the variance of the OGTT absorbance change was much larger than the OWTT under the two source-detector distances,and the wavelength distribution characteristic of the absorbance change’s variance for the three subjects varied greatly.Then the spectra data from the two source-detector distances was differentially processed,and the differential spectral data were compared.The analysis results indicated that the variance of OGTT differential absorbance change was far larger than that of OWTT,and the wavelength distribution characteristics of differential absorbance change’s variance for the three subjects were consistent with the absorption characteristic of glucose solution,which proved that the self-developed non-invasive blood glucose detection system combining the differential processing method could effectively eliminate the background interference and extract the signal of blood glucose.