基于谱分解的无创血糖检测新方法研究

Determination of New Non-Invasive Blood Glucose Detection Method Based on Spectral Decomposition

糖尿病是一种表现为高血糖的糖代谢异常疾病,如果血液中的葡萄糖水平长时间保持非常低或非常高,则可能导致包括组织损伤、中风、心脏病、失明和肾衰竭等严重疾病。根据世界卫生组织(WHO)的数据,目前全球约有4.5亿糖尿病患者。随着糖尿病患者的数量增加,对于血糖检测仪器的需求也日益迫切,由于目前普及的有创侵入式血糖检测仪器会给患者带来不便和疼痛,甚至会引发感染,长此以往会给病人带来不可避免的生理及心理压力,因此实现无创血糖检测具有重要的临床应用价值。光电容积脉搏波(PPG)包含丰富的人体心血管生理病理信息,针对于时域难以观察的PPG信号中与血糖浓度变化相关的频谱信息,提出基于谱分解的无创血糖检测新方法,利用连续小波变换(CWT)对脉搏波信号进行分解,从对应尺度和细节分量的频谱中获取与血糖浓度值变化相关的频谱分量幅值信息,研究发现PPG信号频谱分量幅值的变化与血糖浓度变化值之间存在较高的相关性,通过口服葡萄糖耐糖实验(OGTT)对检测的血糖浓度值与获取的相关PPG信号频谱分量幅值进行偏最小二乘回归建模,并对建立的模型进行评估,校正集的最大RMSEC为12.47 mg·dL^(-1)即0.69 mmol·L^(-1),预测集的最大RMSEP为6.21 mg·dL^(-1)即0.35 mmol·L^(-1),模型的血糖浓度预测值与参考值之间的一致率为96.00%。OGTT实验结果表明,采用谱分解法可以有效分离出血糖分子基团振动特征吸收光谱,血糖频谱分量建模可以最大程度地减小生理变异性和各种环境条件的影响。模型的预测结果符合国家检测标准(>95%)。克拉克网格误差分析结果表明,该方法检测的结果可以用于患者日常血糖监测。

Diabetes is a disease of abnormal glucose metabolism that manifests as hyperglycemia.If the glucose level in the blood remains very low or very high for long periods,it could cause serious diseases including tissue damage,stroke,heart disease,blindness and kidney failure.According to the World Health Organization(WHO),currently there are around 450 million cases of diabetes in the world.With the increase in the number of diabetic patients,the demand for glucose measuring equipment has become increasingly urgent.As the currently popular invasive blood glucose measuring equipment will cause inconvenience and pain to patients and may even cause infections,it will inevitably bring psychological and physiological pressure to patients in the long term.Therefore,the realization of non-invasive blood glucose measurement has important clinical application value.Photoplethysmography(PPG)pulse wave contains abundant information about human cardiovascular physiology and pathology.This paper proposes a new method for non-invasive blood glucose detection based on spectral decomposition,aiming at the spectral information related to blood glucose concentration changes in the PPG signal that is difficult to observe in the time domain Continuous Wavelet Transform(CWT)is used to decompose the PPG signal from the corresponding scale and details in order to obtain the spectral component amplitude information related to the change of blood glucose concentration.Studies have found that there is a higher correlation between the change inthe amplitude of the PPG signal spectral components and the changes in the blood glucose concentration.Through the Oral Glucose Tolerance Test(OGTT),the detected blood glucose concentration and the obtained relevant PPG signal spectral components amplitude is modeled by partial least square regression,and the established model is evaluated.The Root Mean Square of Calibration(RMSEC)of the calibration set is 12.47 mg·dL^(-1),which is 0.69 mmol·L^(-1),and the Root Mean Square Error of Prediction(RMSEP)of the prediction set is 6.21 mg·dL^(-1),which is 0.35 mmol·L^(-1).The agreement between the predicted value of model’s blood glucose concentration and the reference value is 96.00%.The results of the OGTT experiments show that the spectral decomposition method can effectively separate the vibration characteristic absorption spectra of blood glucose molecular groups,and the blood glucose spectral component modeling can minimize the impact of physiological variability and various environmental conditions.The prediction results of the model meet the national testing standards(>95%).Clark grid error analysis show that the results predicted by this method can be used for daily blood glucose monitoring of patients.