基于iPLS和SiPLS算法的人体血清胆红素含量的可见-近红外光谱建模

发布时间：2018-01-17 21:00

  本文关键词：基于iPLS和SiPLS算法的人体血清胆红素含量的可见-近红外光谱建模　出处：《光电子·激光》2016年10期 　论文类型：期刊论文

  更多相关文章： 可见-近红外(NIR)光谱 间隔偏最小二乘法(iPLS) 协合区间偏最小二乘法(SIPLS) 波段优选 血清胆红素(BR)

【摘要】：为了建立血清胆红素(BR,bilirubin)样品总胆红素(TBIL)、直接胆红素(DBIL)和间接胆红素(IBIL)近红外(NIR)光谱分析最优模型,利用可见-NIR透射光谱技术与间隔偏最小二乘法(iPLS)及协合区间偏最小二乘法(SiPLS)算法相结合对建模区域进行优选,实现血清光谱特征波段选择,建立光谱与血清BR成分之间的定量预测模型,以均方根误差(RMSE)作为模型评价标准。结果表明:SiPLS模型效果更佳,TBIL、DBIL和IBIL的最优建模波长范围分别为400~536nm、1 366~1 502nm和2 324~2 460nm,400~502nm、608~710nm和1 644~1 746nm,400~502nm和1 746~1 848nm;3种BR最优预测模型的RMSE分别为0.598 9、0.207 2和0.386 2μmol/L;波段优选对提高预测结果的准确性有重要的意义;采用SiPLS建立TBIL、DBIL和IBIL定量分析模型,不仅可以提高模型的预测精度,而且克服了iPLS单一区间建模的缺点,优选出的特征谱区还可为设计小型专用光谱分析仪器提供依据。
[Abstract]:In order to establish the serum bilirubin (BR, bilirubin) samples of total bilirubin (TBIL), direct bilirubin (DBIL) and indirect bilirubin (IBIL) near infrared (NIR) spectral analysis of the optimal model, using -NIR visible transmission spectroscopy and interval partial least squares (iPLS) and synergy interval partial least squares (SiPLS) algorithm combining to optimize the modeling area, realize the serum spectral characteristics of band selection, establishment of quantitative prediction model between spectrum and serum BR components, with root mean square error (RMSE) as a model of evaluation criteria. The results show that the better effect of SiPLS model, TBIL modeling, optimal wavelength range of DBIL and IBIL were 400~536nm, 1 366~1 and 2 502nm 324~2 460nm, 400~502nm, 608~710nm and 1 644~1 746nm, 400~502nm 746~1 and 1 848nm; 3 BR optimal prediction model of RMSE were 0.598 9,0.207 2 and 0.3862 mol/L; waveband selection for improving the prediction results. Sex is of great importance. Using SiPLS to establish TBIL, DBIL and IBIL quantitative analysis models can not only improve the prediction accuracy of the model, but also overcome the shortcomings of iPLS single interval modeling, and the optimized characteristic spectral area can also provide a basis for designing small and special spectral analysis instruments.

【作者单位】： 暨南大学光电工程系;暨南大学第一附属医院临床检验中心;
【基金】：国家自然科学基金(31371785) 广东省自然科学基金(S2011040001850) 广东省战略新兴产业核心技术攻关(2012A032300016) 高等学校博士学科点专项科研基金(20124401120005) 广东光阵光电科技有限公司光电信息工程院士工作站(2014B090905001)资助项目
【分类号】：R446.1;O657.33
【正文快照】： tant significance to improve prediction accuracy.The SiPLS-based quantitative analysis model of TBIL,DBIL and IBIL in serum has high prediction accuracy and overcomes shortcomings of the iPLS-basedmodel.Furthermore,the optimized characteristic spectrum r，

本文编号：1437924