A method for establishing an anti-moisture interference model for hemoglobin based on near-infrared spectroscopy technology

Abstract

The present invention discloses a method for establishing an anti-moisture interference model for hemoglobin based on near-infrared spectroscopy technology, which relates to the fields of near-infrared spectroscopy and medicine. The method includes the following steps: (1) preparing experimental materials; (2) collecting sample spectra; (3) preprocessing the spectra; (4) selecting moisture characteristic bands; (5) removing the moisture characteristic bands; (6) selecting hemoglobin characteristic bands; (7) establishing a hemoglobin content prediction model. Beneficial effects: In view of the characteristics of strong absorption and wide bands of moisture during NIRS analysis, the present invention reasonably applies a variable selection algorithm and uses the moisture content as the measured component in the model to establish the model. After screening out the moisture characteristic bands in the full band and removing them, the moisture interference encountered by Hb during NIRS modeling is eliminated, reducing the interference variables and redundant variables for subsequent Hb model establishment, and improving the operation efficiency and prediction accuracy of the model.

Description

Technical Field

[0001] The present invention relates to the field of near-infrared spectroscopy and medical field, and particularly relates to a method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology. Background Art

[0002] With the rapid development of society and economy, while people's material life is fully satisfied, they begin to pay more and more attention to their own health problems. Therefore, the detection and prevention of diseases are particularly important. Hemoglobin (Hb) is an iron-containing composite allosteric protein in higher organisms, which has functions such as transporting oxygen and carbon dioxide and maintaining the acid-base balance of blood. The hemoglobin content index is the main basis for clinical diagnosis of blood diseases such as anemia. The decrease of hemoglobin content will lead to the reduction of the total number of red blood cells and the specific volume per unit volume, and the oxygen-carrying capacity will be reduced, resulting in hypoxia of various tissues and organs of the body, which greatly endangers physical health.

[0003] At present, there are mainly two methods for measuring hemoglobin: (1) cyanmethemoglobin (HICN) spectrophotometry; (2) sodium dodecyl lauryl sulfate (SLS) hemoglobin determination method. The advantages of these methods are accuracy and stable results, but they have disadvantages such as cumbersome operation, consumption of chemical reagents, and long detection cycle. Therefore, it is of great significance to establish a simple and rapid Hb quantitative detection method without chemical reagents.

[0004] Near-infrared spectroscopy (NIRS) analysis technology can establish a calibration model by comparing the information such as the content and category of the component to be measured carried by the spectrum with the corresponding data measured by the standard method. Subsequently, the model can be used to quickly predict the relevant characteristics of unknown samples. It is a comprehensive analysis technology integrating spectrum measurement technology, chemometrics and computer technology. The main characteristics of NIRS analysis technology are as follows: 1) Non-destructive to samples. The sample to be measured is measured in situ, and the sample is not damaged or consumed during the measurement process; 2) Real-time analysis. Spectrum acquisition and analysis can usually be completed within dozens of seconds, with fast and efficient analysis; 3) No reagents required. Neither the measurement nor the analysis process requires the intervention of chemical reagents and no chemical waste is generated; 4) Convenient operation. The computer completes the data analysis without the need for professional personnel; 5) Simultaneous measurement of multiple components.

[0005] In previous infrared spectroscopy studies, water often acts as a solvent and exhibits strong absorbance. Reducing water interference has been a long-standing challenge in NIRS analysis of protein-containing samples. Due to the association of hydrogen bonds in liquid water, its absorption peaks are all broad spectral bands. The main problem is that during the NIRS analysis process, there are a large number of overlapping absorption regions between water molecules and protein molecules, which will directly affect the quality of the qualitative and quantitative models of proteins. In addition, changes in different water contents will also cause shifts in the absorption spectral lines of the components to be measured, and the differences in sample moisture will affect the accuracy and robustness of the NIRS models for proteins. Therefore, it is crucial to reduce the influence of water on NIRS prediction analysis and establish a robust analysis model. Currently, in the research on methods for establishing a robust NIRS water analysis model, there are still problems such as inaccurate display of water characteristic bands and incomplete deduction.

[0006] The Chinese patent application document with publication number CN114813463A discloses a method for eliminating water interference. This method is mainly applied to the prediction of the basic density of papermaking wood chips by near-infrared spectroscopy. This method can only perform water correction on the spectra of samples to be measured and cannot, in a practical sense, remove the influence of water interference bands on near-infrared spectroscopy modeling. Therefore, further improvement is still needed.

[0007] The Chinese patent application document with publication number CN103134770A discloses a method for correcting the absorbance of soil, designing a normalized moisture absorption coefficient MAI, and using the absorption coefficient MAI to correct the original absorbance of the soil. However, this method can only solve the influence of water on near-infrared spectroscopy detection from the perspective of spectral correction and cannot, like the elimination of water characteristic bands, completely and accurately eliminate the modeling influence brought by water. Therefore, further improvement is still needed. Summary of the Invention

[0008] The technical problem to be solved by the present invention is how to solve the problem that water interference exists in the NIRS analysis of hemoglobin, affecting the accuracy of the model.

[0009] The present invention solves the above technical problems through the following technical means:

[0010] A method for establishing a water-resistant interference model for hemoglobin based on near-infrared spectroscopy technology, comprising the following steps:

[0011] (1) Prepare experimental materials:

[0012] Use the freeze-dried powder of hemoglobin (Hb) to prepare samples, divide the sample set into a calibration set and a prediction set, and calculate the true water content value and the true hemoglobin (Hb) content value in the calibration set; divide the prediction set into a prediction set of hemoglobin (Hb) content and a prediction set of water content;

[0013] (2)Collect the sample spectrum: Use a near-infrared spectrometer to collect the sample spectrum;

[0014] (3)Spectrum preprocessing: Use the Savitzky-Golay convolution smoothing filtering algorithm to process the sample spectrum in step (2) to obtain the original spectrum;

[0015] (4)Select the water characteristic band: Use the variable selection algorithm to select the characteristic band of water with the original spectrum obtained in step (3) as the input;

[0016] (5)Eliminate the water characteristic band: Eliminate the selected water characteristic band from the full wavelength range of the original spectrum;

[0017] (6)Select the hemoglobin characteristic band: Use the variable selection algorithm to select the characteristic band of hemoglobin (Hb) with the spectrum after removing the water characteristic band from the full wavelength range in step (5) as the input;

[0018] (7)Establish a prediction model for hemoglobin content: Establish a NIRS prediction model between the selected characteristic band of hemoglobin (Hb) and the true content of hemoglobin (Hb) in the calibration set. The specific modeling method is partial least squares; then apply the established prediction model to the prediction set of hemoglobin (Hb) content, observe the evaluation index, and evaluate the prediction performance of the model; the evaluation indexes are shown in formulas (a) and (b):

[0019] (a) ;

[0020] (b);

[0021] In the formula, is the root mean square error of cross-validation, is the coefficient of determination of cross-validation, is the root mean square error of the prediction set, is the coefficient of determination of the prediction set, n is the number of samples in the cross-calibration set or prediction set, y is the true value of the sample, is the mean value of the true values of the samples, is the predicted value of the model.

[0022] Beneficial effects: In view of the characteristics of strong absorption and wide wavelength band of water in NIRS analysis, the variable selection algorithm is reasonably used in the present invention, and the water content is used as the measured component of the model to establish the model; then, after screening out the water characteristic band in the full wavelength range and eliminating it, the water interference encountered by Hb in NIRS modeling is eliminated, reducing the interference variables and redundant variables for the subsequent establishment of the Hb model, and improving the operation efficiency and prediction accuracy of the model.

[0023] Preferably, the range of the sample concentration in the step (1) is set to be 1.2 mg / mL - 7 mg / mL.

[0024] Preferably, the concentration interval of the sample in the step (1) is 0.2 mg / mL.

[0025] Preferably, the division method of the sample set in the step (1) is set such that the ratio of the calibration set to the prediction set is 6:4.

[0026] Preferably, the acquisition wavelength range in the step (2) is 7000 - 11000 .

[0027] Preferably, the spectral acquisition method in the step (2) is the transmission mode.

[0028] Preferably, the hemoglobin (Hb) content in the sample spectrum in the step (2) is 0.12% - 0.7%, and the water content is 99.98% - 99.30%.

[0029] Preferably, the variable selection algorithms in the step (4) and the step (6) are both selected from one of the competitive adaptive reweighted sampling method (CARS), the interval random frog leaping algorithm (IRF), the genetic algorithm (GA), and the differential evolution algorithm (DE).

[0030] Preferably, the model number of the hemoglobin (Hb) lyophilized powder in the step (1) is H7379, and the purity is ≥98%.

[0031] Preferably, the signal of the near-infrared spectrometer in the step (2) is MB3600.

[0032] The advantages of the present invention are as follows:

[0033] 1. Aiming at the water interference existing in the NIRS analysis of hemoglobin, the present invention proposes a new method to reduce this interference, named the variable selection combined with interference component elimination method (VCIE). In view of the characteristics of strong absorption and wide wavelength band of water in the NIRS analysis, the variable selection algorithm is reasonably used, and the water content is used as the measured component of the model to establish the model. After screening out the water characteristic wavelength bands in the full wavelength range and eliminating them, the water interference encountered in the NIRS modeling of Hb is eliminated, reducing the interference variables and redundant variables for the subsequent Hb model establishment, and improving the operation efficiency and prediction accuracy of the model.

[0034] 2. The present invention converts the water in the actual sample into the water characteristic spectrum generated during the NIRS analysis of the Hb solution, and eliminates the water interference from the spectral analysis level, which can more conveniently, quickly and accurately establish the NIRS anti-water interference model of Hb.

[0035] 3. The present invention ingeniously applies the idea of combining the variable selection algorithm and the water characteristic band, and organizes and applies this idea scheme to the NIRS quantitative analysis of Hb, successfully reducing the influence of water during NIRS modeling.

[0036] 4. The present invention makes full use of the advantages of the variable selection method, accurately selects the water characteristic band, and then eliminates it in the full band. This idea method is also the most core difference point from the existing water elimination methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] Figure 1 It is the original spectrum diagram of Hb in Example 1 of the present invention;

[0038] Figure 2 It is the flow chart of the VCIE method in Example 1 of the present invention;

[0039] Figure 3 It is the water characteristic band selection diagram in Example 1 of the present invention;

[0040] Figure 4 It is the effect diagram of modeling by the VCIE method in Example 1 of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0041] To make the objectives, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work shall fall within the protection scope of the present invention.

[0042] There are a large number of overlapping bands between Hb and water in the NIRS region, and the differences in water in different samples will also affect the spectral position of Hb, all of which have a great interference on the establishment of the Hb quantitative model. To solve the water interference problem encountered by Hb in NIRS analysis, the present invention discloses a method VCIE for establishing an Hb anti-water interference model, which is specifically carried out according to the following steps:

[0043] Definition: The spectral preprocessing methods include baseline correction methods (such as derivative correction methods, including the first derivative (first derivative, Der), etc.); scattering correction methods (mainly including multiplicative scatter correction (MSC) and standard normal variate transform (SNV), etc.); smoothing correction methods (mainly including moving window smoothing (SMOOTH), Savitzky-Golay convolution smoothing filter, etc.).

[0044] Definition: The modeling method is partial least squares (PLS).

[0045] Definition: The evaluation indexes of the model are root mean square error of cross-validation ( ), coefficient of determination of cross-validation ( ), root mean square error of prediction set ( ), and coefficient of determination of prediction set ( ). Generally, / is smaller, / is larger, indicating that the model has high precision and good performance. The specific formulas are as follows:

[0046] (1)

[0047] (2)

[0048] In the formula, n is the number of samples in the cross-validation set or prediction set, y is the true value of the sample, is the mean value of the true values of the samples, is the predicted value of the model. An excellent model usually has a smaller root mean square error and a larger coefficient of determination.

[0049] Definition: The variable selection algorithms based on four different principles are competitive adaptive reweighted sampling method (CARS), interval random frog leaping algorithm (IRF), genetic algorithm (GA), and differential evolution algorithm (DE).

[0050] Example 1:

[0051] A method for establishing an anti-moisture interference model for hemoglobin based on near-infrared spectroscopy technology (i.e., the VCIE method) includes the following steps:

[0052] (1) Prepare experimental materials:

[0053] Select Hb freeze-dried powder (model H7379, purity ≥ 98%, purchased from Merck KGaA, Germany) and distilled water to prepare Hb aqueous solution, and maintain the temperature at room temperature (25.6 °C). For the preparation of Hb aqueous solution, first add 12 mg of Hb freeze-dried powder weighed by an electronic balance (model PWN124ZH / E, purchased from Ohaus Corporation, USA) to 9.5 mL of distilled water, gently shake until completely dissolved, and then add water to make up the volume to 10 mL to obtain a 1.2 mg / mL Hb aqueous solution. Samples with different moisture contents are prepared according to this method, and the Hb concentration of the samples is 1.2 mg / mL - 7 mg / mL. In order to more clearly see the spectral differences caused by different moisture contents, the same concentration interval is set, and the concentration interval is 0.2 mg / mL. During the preparation of the samples, a magnetic stirrer (model WH260, purchased from VWR International GmbH, Germany) is used to continuously stir to ensure that the samples are fully dissolved; the division method of the sample set is set so that the ratio of the calibration set to the prediction set is 6:4, and the true moisture content value and the true Hb content value in the calibration set are calculated; the prediction set is further divided into a prediction set of Hb content and a prediction set of moisture content;

[0054] (2) Collect sample spectra: Use a near-infrared spectrometer (MB3600, ABB Ltd., Canada) to collect sample spectra. The spectral collection method is transmission, and the resolution is 8 , and the number of scans is 128 times; the collection band is 7000 - 11000 ; during the process of collecting the spectra of the samples, at the same concentration, the samples are placed in a supporting instrument container and oscillated three times, and three spectra are collected. A total of 90 sample spectra are collected, among which the Hb content is 0.12% - 0.7%, the moisture content is 99.98% - 99.30%, and the total number of spectral variables is 1038;

[0055] Note: NIRS reflects the spectral absorption of the components of organic compounds in different band intervals. Therefore, the reasonable selection of the band interval has a direct impact on reflecting the information of the sample components and structure. In the NIRS region, the absorption of water occupies the main part of the substances of interest. Among them, in the region, due to the combination of the asymmetric stretching and bending of water molecules around and the combination of the symmetric and asymmetric stretching modes of water molecules around , very strong absorption peaks will be generated by water molecules. In the case of a single moisture background for the modeling samples, too strong moisture absorption is not conducive to the research. At the same time, Hb has a key absorption band within , that is, the second harmonic absorption of the N-H group. Therefore, the spectral analysis band of this study is set to .

[0056] (3)Spectral preprocessing: Use the Savitzky-Golay convolution smoothing filter algorithm to process the sample spectrum in step (2) to obtain the original spectrum;

[0057] Note: In the present invention, the sample material used is a solution prepared from freeze-dried Hb powder. Therefore, during the sample spectrum acquisition process, various effects such as spectral scattering and drift will exist. Spectral preprocessing of the collected original spectrum can effectively eliminate the influence brought by the sample itself on Hb for NIRS modeling. The preprocessing methods adopted include baseline correction methods (first derivative (DEr)); scattering correction methods (multiplicative scatter correction (MSC) and standard normal variate transform (SNV)); smoothing correction methods (Moving window smoothing (SMOOTH), Savitzky-Golay convolution smoothing filter). After using the above spectral preprocessing methods to process the spectral original data, a PLS regression model for moisture content and Hb content is established. The model prediction results under different preprocessing methods are obtained in the experiment. The experimental results show that the data modeling effect is the best after S-G filter preprocessing.

[0058] (4)Select moisture characteristic bands: Take the original spectrum obtained in step (3) as the input quantity, and use four variable selection algorithms, namely CARS, IRF, GA, and DE, to select the characteristic bands of moisture. The moisture characteristic bands selected by CARS mainly concentrate in , , , ; The moisture characteristic bands selected by IRF mainly concentrate in , , ; The moisture characteristic bands selected by GA mainly concentrate in , ; The moisture characteristic bands selected by DE mainly concentrate in , , , ;

[0059] Description: A near-infrared spectroscopy quantitative analysis model is established between the selected moisture characteristic bands and the actual moisture content in the calibration set, and the established prediction model is applied to the prediction set of moisture content, and the evaluation indicators of the model are observed. The use of variable selection methods in the process of Hb quantitative analysis modeling can simplify the model, improve the operating efficiency of the model, and enhance the interpretability of the model. More importantly, when redundant or nonlinear variables are eliminated, a calibration model with strong predictive ability and good robustness can be obtained, specifically referring to a quantitative model established between the content of the measured component and the band selected by the variable selection algorithm. In order to prevent the randomness of the final result caused by a single variable selection algorithm, the present invention uses four different principles of variable selection algorithms, namely competitive adaptive reweighted sampling (CARS), interval random frog leaping algorithm (IRF), genetic algorithm (GA) and differential evolution algorithm (DE) for horizontal comparison: Among them, the variable selection algorithm with the best moisture content prediction model is GA, and the RMSEP of the prediction model is , is 0.9950; followed by DE, the RMSEP of the prediction model is , is 0.9941; then IRF, the RMSEP of the prediction model is , is 0.9884; finally, CARS, the RMSEP of the prediction model is , The main reason for the different prediction effects of the four variable selection algorithms is that due to the association of hydrogen bonds, the absorption peaks of liquid water are all broad bands, among which the first and second harmonic absorptions of OH stretching vibration appear at 6944 and 760 respectively. and 10420 Nearby, its combined frequency absorption band appears at 8197 nearby. Figure 3 Compared with GA and DE algorithms, CARS and IRF algorithms did not select the liquid water in the above 8197 The nearby combined frequency absorption bands have a weaker ability to screen out characteristic bands of moisture, and the final model effect is poor.

[0060] (5) Eliminating moisture characteristic bands: After step (4) is completed, the selected moisture characteristic bands are eliminated from the original spectrum.

[0061] Note: The significance of this step is that the water bands selected in step (3) are actually interference bands in the process of Hb quantitative analysis modeling. These characteristic bands will seriously affect the NIRS modeling analysis of Hb, so they need to be eliminated;

[0062] (6) Select the characteristic bands of hemoglobin: Use the spectrum processed in (5) as the input, and select the Hb characteristic interval bands by using four variable selection algorithms of CARS, IRF, GA, and DE respectively. Among them, CARS selects 30 characteristic bands; IRF selects 239 characteristic bands; GA selects 260 characteristic bands; DE selects 243 characteristic bands.

[0063] (7) Establish a prediction model for hemoglobin content: Establish a NIRS prediction model between the Hb characteristic bands selected in step (6) and the true Hb content in the calibration set. The specific modeling method is partial least squares; then apply the established prediction model to the prediction set of Hb content, observe the evaluation index, and evaluate the prediction performance of the model. The results show that in the CARS algorithm, RMSEP is , is 0.9931; in the IRF algorithm, RMSEP is , is 0.9903; in the GA algorithm, RMSEP is , is 0.9979. In the DE algorithm, RMSEP is , is 0.9983.

[0064] Figure 1 is the original Hb spectrogram. It can be seen from the figure that A is the spectrum of the collected sample, and B is the spectrum after the best preprocessing method, that is, S-G filtering.

[0065] Figure 2 is the flow chart of the VCIE method;

[0066] Figure 3 is the moisture characteristic band selection diagram. It can be seen from the figure that the red bands are the moisture characteristic bands selected by using the variable selection algorithm. A is the moisture characteristic band selected by using the CARS algorithm; B is the moisture characteristic band selected by using the IRF algorithm; C is the moisture characteristic band selected by using the GA algorithm; D is the moisture characteristic band selected by using the DE algorithm.

[0067] Figure 4 is the effect diagram of the VCIE method modeling. It can be seen from the figure that among the four different variable selection algorithms, the best effect is DE, followed by GA, and the effects of IRF and CARS are about the same. The difference is actually related to the moisture removal effect.

[0068] Comparative Example 1:

[0069] A method for establishing a prediction model for Hb content, including the following steps:

[0070] (1) Prepare experimental materials: same as step (1) of Example 1;

[0071] (2) Sample the sample spectrum: same as step (2) of Example 1;

[0072] (3) Preprocess the spectrum: same as step (3) of Example 1;

[0073] (4) Establish an Hb content prediction model: Use the same variable selection algorithm as in Example 1 for the sample spectrum after the spectral preprocessing method. Take the Hb content as the prediction component of the model, select the characteristic bands of Hb, and then establish the NIRS quantitative model of Hb. The results show that among the four variable selection algorithms, the model using DE has the best prediction effect, and the RMSEP is , being 0.9951; the model using GA has the second-best prediction effect, and the RMSEP is , being 0.9932; then comes CARS, with RMSEP being , being 0.9901; the worst is IRF, with RMSEP being , being 0.9870.

[0074] Comparing the results of Example 1 and Comparative Example 1:

[0075] In the CARS algorithm, it is increased from 0.9901 to 0.9931, and the RMSEP is decreased from to , and the modeling accuracy is improved by 16.41%;

[0076] In the IRF algorithm, it is increased from 0.9870 to 0.9903, and the RMSEP is decreased from to , and the modeling accuracy is improved by 13.64%;

[0077] In the GA algorithm, it is increased from 0.9932 to 0.9979, and the RMSEP is decreased from to , and the modeling accuracy is improved by 44.84%;

[0078] In the DE algorithm, it is increased from 0.9951 to 0.9983, and the RMSEP is decreased from to , and the modeling accuracy is improved by 40.37%.

[0079] The results show that the VCIE method of the present invention can effectively reduce the influence of water interference encountered by Hb in NIRS detection, and at the same time improve the prediction accuracy of the Hb content prediction model.

[0080] The above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present invention.

Claims

1. A method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology, characterized in that, it includes the following steps: (1) Prepare experimental materials: Use freeze-dried powder of hemoglobin (Hb) to prepare samples, divide the sample set into a calibration set and a prediction set, and calculate the true water content value and the true hemoglobin (Hb) content value in the calibration set; divide the prediction set into a prediction set of hemoglobin (Hb) content and a prediction set of water content; (2) Collect sample spectra: Collect sample spectra with a near-infrared spectrometer; (3) Spectral preprocessing: Use the Savitzky-Golay convolution smoothing filter algorithm to process the sample spectra in step (2) to obtain the original spectra; (4) Select water characteristic bands: Use the original spectra obtained in step (3) as the input, and use a variable selection algorithm to select the characteristic bands of water; (5) Eliminate water characteristic bands: Eliminate the selected water characteristic bands from the full wavelength range of the original spectra; (6) Select hemoglobin characteristic bands: Use the spectra after removing the water characteristic bands from the full wavelength range in step (5) as the input, and use a variable selection algorithm to select the characteristic bands of hemoglobin (Hb); (7) Establish a hemoglobin content prediction model, establish a NIRS prediction model between the selected hemoglobin (Hb) characteristic bands and the true hemoglobin (Hb) content in the calibration set, and the specific modeling method is partial least squares; then apply the established prediction model to the prediction set of hemoglobin (Hb) content, observe the evaluation indicators, and evaluate the prediction performance of the model; the evaluation indicators are shown in formulas (a) and (b): （a） ； （b）； Wherein, is the root mean square error of cross-validation, is the coefficient of determination of cross-validation, is the root mean square error of the prediction set, is the coefficient of determination of the prediction set, n is the number of samples in the cross-validation set or the prediction set, y is the true value of the sample, is the mean value of the true values of the samples, is the predicted value of the model; In steps (4) and (6), the variable selection algorithm is selected from one of the competitive adaptive reweighted sampling method, the interval random frog leaping algorithm, the genetic algorithm, and the differential evolution algorithm.

2. The method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology according to claim 1, characterized in that, the range of the sample concentration prepared in step (1) is set to 1.2 mg / mL - 7 mg / mL.

3. The method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology according to claim 2, characterized in that, the concentration interval of the samples in step (1) is 0.2 mg / mL.

4. The method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology according to claim 1, characterized in that, the division method of the sample set in step (1) is set so that the ratio of the calibration set to the prediction set is 6:

4.

5. The method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology according to claim 1, characterized in that, The acquisition band in step (2) is 7000 - 11000 .

6. The method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology according to claim 1, characterized in that, the spectral collection method in step (2) is the transmission mode.

7. The method for establishing a water interference-resistant model for hemoglobin based on near-infrared spectroscopy technology according to claim 1, characterized in that, In step (2), the hemoglobin content of the sample spectrum is 0.12% - 0.7%, and the moisture content is 99.98% - 99.30%.

8. The method for establishing an anti-moisture interference model for hemoglobin based on near-infrared spectroscopy according to claim 1, characterized in that, in steps (4) and (6), the variable selection algorithm is selected from the competitive adaptive reweighted sampling method.

9. The method for establishing an anti-moisture interference model for hemoglobin based on near-infrared spectroscopy according to claim 1, characterized in that, in step (1), the model number of the hemoglobin lyophilized powder is H7379, and the purity is ≥98%.

10. The method for establishing an anti-moisture interference model for hemoglobin based on near-infrared spectroscopy according to claim 1, characterized in that, in step (2), the signal of the near-infrared spectrometer is MB3600.

Images