Study On Near Infrared Rapid Analysis Method Of Hemoglobin

Human hemoglobin(Hb)is an iron-containing complex allosteric protein,which is composed of four subunits and has the following functions: the carrying oxygen and carbon dioxide,maintain the balance of both blood acid and alkali,and so on.It’s an important clinical biochemical indicator that can be used to judge anemia and the status of iron nutritional and a variety of other diseases.Near infrared(NIR)spectroscopy has the advantages of rapid,non-destructive,reagentless,which has been applied in various fields.In recent years,NIR spectroscopy is used in the hemoglobin analysis in vivo or vitro,and it caused widespread concern due to its noninvasive(or reagentless)green detection methods.However,spectral prediction accuracy has not yet reached the standard of clinical application because of the lack of basic research.This paper studied specially the modeling optimization method of the hemoglobin NIR spectroscopy analysis in the human peripheral blood samples,and which provides valuable reference for further application.A total of 300 human peripheral blood samples were collected and the Hb content was determined by cyanide high-speed iron Hb spectrophotometry method as a reference for modeling and validation.120 samples were randomly selected for validation,and the remaining 180 samples were randomly divided into calibration(100 samples)and predicted(80 samples).In order to avoid saturation absorption of NIR spectra caused by water molecules,the wavebands with absorbance higher than 4 were removed,and the remaining were(400-1880 & 2100-2300nm)were further used for modeling.In order to eliminate the spectral baseline drift and tilt,the Savitzky-Golay(SG)correction was used to establish the partial least squares(PLS)model for the further spectral band.According to the prediction effect,the parameters of optimal SG correction mode like the following: derivatives order d=2,the degree of polynomial p=2,the number of smoothing points m=31,and the spectral prediction effect was improved.In order to further extract the information and improve the model prediction accuracy,the integrated SG-EC-PLS and SG-RRPC-PLS methods are proposed by combining respectively the wavelength selection method of equidistant combination(EC)and the repetition rate priority combination(RRPC).For SG-EC-PLS model,the initial wavelength(I),number of wavelengths(N)and number of wavelength gaps(G)were 1230 nm,71 and 6,respectively,and the RMSEPV,RRMSEPV and RP,V of NIR predicted value and standard method measured value were 3.29gL-1,2.67% and 0.980,respectively.For SG-RRPC-PLS model,the number of wavelengths(N)was 41,the corresponding RMSEPV,RRMSEPV and RP,V were 2.86gL-1,2.31% and 0.983,respectively.The results show that the relationship betiween the predicted value of the two integrated methods and the measured values of the standard method for hemoglobin is high,the deviation is very low and the prediction accuracy is high.In the two methods,the SG-RRPC-PLS model used smaller wavelength and achieved better predictive effect.The established integrated optimization methods can effectively eliminate noise interference and extract information wavelength,and provide valuable reference for further application.The established chemometrics method and research framework are also expected to be applied to other fields.