NIR光谱结合稳定等效MW-PLS方法用于高脂血症指标分析

NIR Spectroscopy Combined with Stability and Equivalence MW-PLS Method Applied to Analysis of Hyperlipidemia Indexes

基于稳定性、等效性对移动窗口偏最小二乘(MW-PLS)方法进行改进,应用于高脂血症指标总胆固醇(TC)、甘油三酯(TG)无试剂近红外光谱分析的波长优化。提出兼顾随机性、稳定性的定标、预测、检验框架。从全体人血清样品(阴性145、阳性158,共303)中,随机选取103个为检验集(阴性44、阳性59),余下为建模集(阴性101、阳性99,共200),再将建模集随机划分为定标集(阴性51、阳性49,共100)、预测集(阴性50、阳性50,共100)共50次;基于所有划分的平均预测效果优选模型参数,使得模型具有稳定性;采用样品优选的模型进行检验。TC,TG最优MW-PLS波段分别为1 556~1 852,1 542~1 866nm;为了解决由于材料性能、成本的因素对仪器分光系统设计的制约,提出等效模型集,并得到TC和TG等效模型集的唯一公共波段1 542~1 852nm。检验结果表明:采用最优MW-PLS波段,检验样品的TC和TG预测均方根误差(V_SEP)分别为0.177和0.100mmol L-1、相关系数(V_RP)分别为0.988和0.996,高脂血症的灵敏度、特异性分别为95.0%和90.5%;基于公共等效波段,TC和TG的V_SEP值分别为0.177和0.101mmol·L-1、V_RP值分别为0.988和0.996,灵敏度、特异性分别为92.7%和90.3%。结论:近红外光谱结合稳定等效MW-PLS方法提供了一种有潜力的大人群血脂检测工具。

Moving window partial least square(MW-PLS)method was improved by considering the stability and equivalence,and was used for the wavelength optimization of reagent-free near-infrared(NIR)spectroscopic analysis of total cholesterol(TC)and triglycerides(TG)for hyperlipidemia.A random and stability-dependent framework of calibration,prediction,and validation was proposed.From all human serum samples(negative 145 and positive 158,a total of 303sample),103samples(negative44and positive 59)were randomly selected for the validation set,the remaining samples(negative 101 and positive 99,a total of200sample)were used as modeling set;then the modeling set was randomly divided into calibration set(negative 51 and positive49,a total of 100sample)and prediction set(negative 50 and positive 50,a total of 100sample)by 50 times.To produce modeling stability,the model parameters were optimized based on the average prediction effect for all divisions;the optimized models were validated by using the validation samples.The obtained optimal MW-PLS wavebands were 1 556~1 852 nm for TC and1 542~1 866 nm for TG.In order to solve the problem that instrument design typically involves some limitations of position and number of wavelengths because of cost and material properties,the equivalent model sets were proposed,and a unique public waveband 1 542~1 852 nm of the equivalent model sets for TC,TG was found.The validation results show that:using the optimal MW-PLS wavebands,validation samples' root mean square error of prediction(V_SEP)for TC,TG were 0.177,0.100mmol·L-1,the correlation coefficient of prediction(V_RP)for TC,TG were 0.988,0.996,and the sensitivity and specificity for hyperlipidemia achieved 95.0%,90.5%,respectively;using the public equivalent wavebands,the V_SEP for TC,TG were0.177,0.101mmol·L-1),the V_RPfor TC,TG were 0.988,0.996,and the sensitivity and specificity achieved 92.7%,90.3%,respectively.Conclusion:NIR spectroscopy combined with the stability and equivalence-improvement MW-PLS method can provide a potential tool for detecting hyperlipidemia for large population.