Study On NIR/MIR Spectroscopic Analysis Model Of Human Serum TC

Total Cholesterol?TC?is an important clinical indicator of various cardiovascular and cerebrovascular diseases.Traditional measurement methods require professional operation of reagents and·Laboratories,and are not suitable for rapid detection in daily health screening and treatment of·Large populations.Near-infrared?NIR?spectroscopic analysis has the advantage of no reagents,fast and simple.Fourier transform infrared/attenuated total reflection?FTIR/ATR?spectroscopy technology has achieved spectral measurement of ultra-short optical path·Liquid samples.They have all been applied to the rapid quantitative analysis of blood clinical indicators such as serum TC,and the related optimization methods have been the hotspot in the field.In this paper,serum total cholesterol analysis was used as the research object.The same batch of serum samples were used to carry out quantitative analysis modeling and comparative study of NIR and FTIR/ATR spectrocopy,and related chemometric methods and experimental studies were carried out.The main results are as follows:1.Methodological research1)Calibration-prediction-test design for multi-division modeling:Modeling samples were randomly divided into calibration and prediction sample sets 20 times.Modeling based on all partitioning,and optimizing parameters based on average prediction effects.2)Norris derivative filter?NDF?:Based on partial·Least squares?PLS?regression,the NDF parameter optimization platform of parameter full cycle is constructed to determine the optimal NDF parameter of serum total cholesterol PLS model.3)Equal-wavelength combination and Wavelength Step-by-step Phasing:Construct Equal-wavelength Combination PLS?EC-PLS?modeling platform to achieve a wide range of wavelength models.Further.The Wavelength Step-by-step Phasing?WSP?method is used to improve and select the first 100 EC-PLS models,eliminating interference wavelengths,and simplifying the wavelength model.2.NIR and FTIR/ATR spectroscopic analysis of serum TC1)NIR spectral model:The Norris-PLS model is established after saturated absorption spectrum region is eliminated.The deviation between the model prediction and the clinical measured value?SEP⁺?is 0.342 mmol·L^-1;EC-PLS is further established.SEP⁺of the optimal model was improved to 0.167 mmol·L^-1,and the deviation between the predicted and clinical measured values?SEP?,the correlation coefficient?R_P?,and the sample deviation and predicted deviation ratio?RPD?were 0.169 mmol·L^-1,0.990,6.9;The first 100 EC-PLS models were improved and optimized by WSP method.The optimal EC-WSP-PLS model contained 18wavelengths,and SEP⁺was 0.151 mmol·L^-1.SEP,R_P,and RPD were 0.151 mmol·L^-1,0.993,and7.7.The wavelength model is simplified and the prediction accuracy is improved.2)FTIR/ATR spectral model:SEP⁺of the optimal Norris-PLS model is 0.963 mmol·L^-1.EC-PLS models was further established,SEP⁺of the optimal EC-PLS model was improved to0.678 mmol·L^-1.SEP,R_P,and RPD were 0.846 mmol·L^-1,0.712,and 1.4.The first 100 EC-PLS models were improved and selected by WSP method.The optimal EC-WSP-PLS model contained 20 wavelengths and SEP⁺was 0.578 mmol·L^-1.SEP,R_P,and RPD were 0.808mmol·L^-1,0.734,and 1.4.The overall spectral prediction effect is still far from the NIR spectral model.This paper systematically studies spectral preprocessing and wavelength model optimization methods for near/mid-infrared analysis of serum total cholesterol.NIR spectroscopy combined with EC-WSP-PLS method to obtain high-precision spectral prediction,and its serum total cholesterol quantitative analysis ability is significantly better than FTIR/ATR spectrum.The proposed WSP-PLS is a quadratic optimization method for wavelength models.It can further optimize any wavelength model obtained through other strategies?continuous,discrete?for simpler and better predictions.The various integrated optimization methods proposed in this paper can provide valuable reference for the improvement and clinical application of blood spectroscopy.