Fluorescent substance characteristic wavelength quantitative method

Abstract

The invention provides a fluorescent substance characteristic wavelength quantitative analysis method based on concentration-dependent red shift induced by an inner filter effect. The method comprises the following steps: acquiring fluorescence spectrums of a series of fluorescent substance solutions with a series of concentrations; preprocessing the spectrum, and extracting characteristic wavelengths (peak wavelength lambdamax and wavelengths lambdal and lambdar corresponding to specific intensity nImax at two sides of a peak value); constructing a linear model of the characteristic wavelength and the concentration, and determining an optimal fitting model by dynamically optimizing a scale factor n and a wavelength combination strategy; and high-precision and wide-range quantitative analysis of the concentration of the solution to be detected is realized by using the optimal model. According to the method, the linear detection range can be expanded to more than 5 times that of a traditional method, the relative average error is smaller than 1%, the method is not influenced by hardware parameter fluctuation, and the range and precision of fluorescence quantitative analysis are remarkably improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of fluorescence quantitative analysis, and particularly relates to a characteristic wavelength quantitative analysis method based on concentration-dependent red shift induced by internal filter effect, which is suitable for wide-range and high-precision fluorescence substance detection in the fields of environmental monitoring, biochemical analysis, food safety and the like. BACKGROUND

[0002] The fluorescence spectrum method has the advantages of high sensitivity, high accuracy, low detection limit, good selectivity, non-destructiveness, simplicity and the like, is one of commonly used and reliable analysis methods, can provide rich information of substance molecular interaction, energy level conversion and concentration change, and is widely applied in the fields of biochemical analysis, material application, clinical medical detection, environmental monitoring, food safety, public security intelligence and the like.

[0003] The traditional fluorescence quantitative analysis is based on the relationship between fluorescence intensity and substance concentration. When the concentration is low, the detected fluorescence intensity F obs is proportional to the concentration c. However, when the concentration is increased, the fluorescence intensity does not linearly increase with the concentration or even decreases, which leads to deviation of linear relationship and distortion of spectrum shape, and the concentration is greatly deviated according to the fluorescence intensity at this time, which is mainly caused by fluorescence internal filter effect (IFE). The IFE is divided into primary internal filter effect (pIFE) and secondary internal filter effect (sIFE). The pIFE is that when the solution concentration is too high, the incident light is strongly absorbed by the front part of the fluorescence substance, so that the light intensity actually used for excitation is weakened, which leads to the decrease of fluorescence intensity. The sIFE usually occurs in the fluorescence solution with large overlap between the absorption spectrum and the fluorescence emission spectrum, and the emitted fluorescence is reabsorbed by the sample.

[0004] Due to the existence of fluorescence internal filter effect (IFE), the conventional fluorescence spectrum analysis method cannot accurately quantify the concentration of the substance. The existing research regards the fluorescence internal filter effect as an error source in fluorescence analysis, which leads to distortion of fluorescence spectrum, distortion of spectrum shape and nonlinearity between fluorescence signal intensity and concentration of the fluorophore. Many researchers try to restore the linear relationship between fluorescence intensity and its concentration by deducing the IFE correction formula. Some researchers correct the formula of fluorescence intensity and concentration of the fluorescence substance solution by using absorbance, and the corrected formula is as follows:

[0005]

[0006] Wherein, F obs is the measured maximum fluorescence intensity, F corr is the maximum fluorescence intensity after correction of the internal filter effect from F obs ; A ex and A emrespectively, where s is the thickness of the excitation beam, g is the distance between the edge of the excitation beam and the edge of the cuvette (in the direction of the receiving end), and d is the width of the cuvette.

[0007] In 1994, Albinsson et al. gave the corrected fluorescence intensity as:

[0008]

[0009] Nettles et al. measured the water Raman signal in a series of reference samples to obtain the change in Raman intensity caused by the inner filter effect in the sample, and determined the effective excitation path length d of the fluorophotometer by a simple linear fitting of the relationship between the water Raman intensity and the absorbance. ex and the emission path length d em Then, the fluorescence and Raman inner filter effects were more reliably corrected using formula (3), where L is the width of the cuvette used to measure the sample absorbance spectrum.

[0010]

[0011] Ma et al. studied the sIFE of the fluorescence excitation spectrum of a rhodamine 6G aqueous solution with varying concentrations and proposed a mathematical model to solve the inflection point of the fluorescence excitation spectrum caused by sIFE.

[0012] The correction formulas of the above studies mostly correct the inner filter effect for the geometric structure of the cuvette and the geometric parameters of the light beam, without considering the influence of the solute-solvent system of the fluorescent substance itself on the inner filter effect, or the formation of dimers or polymers of the fluorescent substance with increasing solution concentration. The contribution of dimers or polymers to the fluorescence spectrum and the absorbance spectrum is different, and the influence on the absorption process of the emitted fluorescence is also different. The traditional way of correcting the absorption process cannot meet the correction requirements, and the correction of the fluorescence spectrum intensity decay and the red shift of the fluorescence peak still needs to be improved.

[0013] In addition to optimizing the classic formula, some researchers have constructed a wide range of fluorescence intensity-concentration fitting formula by means of trace element analysis, but this model only considers pIFE, and in systems where sIFE is prevalent, the fitting accuracy is greatly reduced. The fluorescence peak wavelength correction algorithm proposed by Li can improve the relationship between fluorescence intensity and concentration at low concentrations, but it is difficult to deal with high concentration situations. In addition, the strategy of reducing IFE error by optimizing experimental conditions has limitations, such as using microcuvette to shorten the optical path can weaken pIFE, but will cause the attenuation of fluorescence signal intensity and the decline of signal-to-noise ratio; the introduction of high-power laser light source can slow down the fluorescence quenching and widen the linear range, but increases the experimental cost; sample dilution not only reduces the signal-to-noise ratio and introduces artificial errors, but also may change the molecular characteristics of the natural sample, affecting the detection accuracy. The fluorescence intensity ratio method relies on double receiving optical fibers and double spectrometers, which is costly and complex, and the error increases sharply at high concentrations.

[0014] Although IFE is usually considered as an error source of traditional fluorescence spectroscopy, more and more researchers no longer regard it as an interference factor, but as a tool for quantitative analysis of fluorescent substances. Many studies focus on the selective analysis of fluorescent substances, usually by adjusting the concentration of fluorescent substances or absorbing substances to change the degree of IFE, in order to carry out quantitative analysis of certain substances.

[0015] However, there is no technical report in existing research that uses the principle of concentration-dependent red shift induced by inner filter effect to solve the problem of narrow linear range and large error at high concentration of traditional fluorescence intensity method. SUMMARY

[0016] In view of the defects in the prior art, the purpose of the present application is to provide a fluorescence substance characteristic wavelength quantitative analysis method based on the concentration-dependent red shift induced by inner filter effect, to solve the problem of narrow linear range and large error at high concentration of traditional fluorescence intensity method.

[0017] According to one aspect of the present application, a fluorescence substance characteristic wavelength quantitative analysis method is provided, comprising:

[0018] Obtaining the fluorescence spectrum of a series of concentrations of fluorescent substance solution;

[0019] Pretreating the fluorescence spectrum and extracting characteristic wavelengths, the characteristic wavelengths including peak wavelength λ max , peak intensity I max , both sides of specific intensity nI max , corresponding left wavelength λ l , right wavelength λ r , any one or more of them, wherein 0.1<n<1;

[0020] Constructing a linear model of characteristic wavelength and concentration, the linear model being λ lOr λ r Single-wavelength model with λ l as variable r Dual-wavelength model with λ l , λ max and λ r as variable max Triple-wavelength model with λ max , λ max and λ l as variable

[0021] Determine the optimal fitting model by dynamically optimizing the scale factor n and wavelength combination strategy

[0022] Calculate the concentration of the solution to be measured according to the characteristic wavelength of the solution to be measured using the optimal fitting model.

[0023] Optionally, the pre-processing of the fluorescence spectrum includes baseline correction to eliminate dark noise, and digital filtering to smooth the spectral data segment.

[0024] Optionally, the extraction of the characteristic wavelength includes determining nI max and λ max by a peak detection algorithm, and calculating nI max corresponding λ l and λ r using linear interpolation, wherein the step length of n is 0.001.

[0025] Optionally, the linear model of the characteristic wavelength and the concentration is constructed, wherein:

[0026] The single-wavelength model: from the left or right side of the peak wavelength λ max , select the wavelength λ l or λ r corresponding to the current target intensity as a single characteristic variable, which contains the information of the fluorescence peak red shift;

[0027] The dual-wavelength model: extract two wavelengths λ l and λ r corresponding to the target intensity on both sides of the peak wavelength λ max , and combine λ l , λ r as a characteristic variable, which can reflect the change of the fluorescence peak red shift and describe the change of the fluorescence peak shape.

[0028] The triple-wavelength model: extract two wavelengths λ l and λ r corresponding to the target intensity on both sides of the peak wavelength λ max , and combine λ l , λ r , λ maxAs three characteristic variables, a three-wavelength characteristic combination is formed, including the peak wavelength and the specific intensity wavelengths corresponding to both sides of the peak, which further enhances the ability to characterize the fluorescence peak morphology and covers the key position information of the fluorescence peak.

[0029] Optionally, the linear model of characteristic wavelength and concentration is constructed, wherein:

[0030] The dual-wavelength model needs to fit λ l and λ r The relevance of to avoid ambiguous interpretation;

[0031] The three-wavelength model needs to be further fitted with λ max and λ l ,λ r relationships to enhance stability.

[0032] Optionally, in the dual-wavelength model, the fitting peak wavelength λ max The wavelength λ of the target intensity on both sides l and λ r The relationship between them is established, and a one-to-one correspondence is established to ensure the certainty of the wavelength combination;

[0033] In the three-wavelength model, the fitting peak wavelength λ max The target intensity wavelength λ corresponding to both sides max and λ l ,λ r , establish their corresponding relationship, and ensure the accurate correspondence of the three wavelength variables at different concentrations.

[0034] Optionally, determining the optimal fitting model by dynamically optimizing the scaling factor n and the wavelength combination strategy includes:

[0035] For the single wavelength model, the dual wavelength model or the triple wavelength model, the corresponding characteristic variables are calculated based on each n value, where 0.1 <n<1,步长0.001,通过遍历不同n值构建对应的特征矩阵;

[0036] A linear regression model is trained for each n value, the feature matrix is ​​fitted to the sample concentration, the model performance is quantified using an evaluation index, and the best fitting model is determined based on the evaluation index.

[0037] Optionally, the evaluation index includes the determination coefficient R 2 , mean absolute percentage error MAPE and mean square error MSE, the best fitting model is R 2 The model with the highest MAPE and lowest MSE.

[0038] Optionally, obtaining fluorescence spectra of fluorescent substance solutions with a range of concentrations includes:

[0039] The device is composed of an excitation light source, a collimation system, a cuvette, a converging system and a detector, wherein the excitation light emitted by the light source is collimated by the collimation system, then is incident from the center of the incident surface of the cuvette, excites the fluorescent substance solution, and the emitted fluorescence is converged by the converging system and is received by the detector.

[0040] Optionally, the fluorescent substance is any one of Rhodamine B, fluorescein sodium and chlorophyll a, which has a concentration-dependent red shift characteristic.

[0041] Compared with the prior art, the present application has at least one of the following beneficial effects:

[0042] The characteristic wavelength quantitative analysis method of the fluorescent substance based on the concentration-dependent red shift induced by the inner filter effect can expand the linear detection range to more than 5 times of the traditional method by extracting the characteristic wavelength to construct a linear model with the concentration. The relative average error of the method is less than 1%, which significantly improves the accuracy of the fluorescent quantitative analysis. Meanwhile, the characteristic wavelength is hardly affected by the hardware parameters such as light source fluctuation and receiving position offset, and has the stability of not depending on the hardware parameters, which provides a new means for wide-range and high-precision fluorescent substance quantitative analysis, and has important application value in the fields of environmental monitoring and biochemical analysis. BRIEF DESCRIPTION OF DRAWINGS

[0043] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments with reference to the attached drawings:

[0044] Figure 1 The flowchart of a characteristic wavelength quantitative analysis method based on the concentration-dependent red shift induced by the inner filter effect for an embodiment of the present application.

[0045] Figure 2 The fluorescent spectrum diagram of Rhodamine B aqueous solution for an embodiment of the present application.

[0046] Figure 3 The single-wavelength fitting result of Rhodamine B for an embodiment of the present application.

[0047] Figure 4 The double-wavelength fitting result of Rhodamine B for an embodiment of the present application; wherein (a) is the fitting plane of the concentration and λ l , λ r , (b) is the correlation curve of λ l and λ r , and (c) is the spatial curve after constraint.

[0048] Figure 5 The triple-wavelength fitting result of Rhodamine B for an embodiment of the present application. DETAILED DESCRIPTION

[0049] The present application will be described in detail below with specific examples. The following examples will help those skilled in the art to further understand the present application, but do not limit the present application in any form. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These are all within the scope of protection of the present application.

[0050] During the research of the present application, the researchers noticed the concentration-dependent red shift (CDRS) phenomenon in the fluorescence spectrum, which is particularly evident in high-concentration fluorescent solutions, mainly caused by IFE. Based on the CDRS induced by IFE and the full-spectrum analysis, the present application proposes a characteristic wavelength quantitative analysis method (FWQA). With the peak intensity I max as the reference, the peak wavelength λ max and the wavelengths λ max and λ l corresponding to specific intensities nI r (0.1<n<1) on both sides of the peak value are determined, and these wavelengths are determined as characteristic wavelengths. Then a linear fitting model of wavelength and concentration is constructed respectively with single characteristic wavelength, double characteristic wavelengths and three characteristic wavelength combinations. Through the dynamic optimization of the scale factor and the combination strategy, the spectral morphology of different fluorescent substances is adaptively matched to obtain the optimal fitting model.

[0051] Specifically, the embodiment of the present application provides a fluorescent substance characteristic wavelength quantitative analysis method based on the concentration-dependent red shift induced by the inner filter effect, referring to Figure 1 The method comprises:

[0052] S1, obtaining the fluorescence spectrum of a series of concentrations of fluorescent substance solutions;

[0053] S2, preprocessing the fluorescence spectrum and extracting characteristic wavelengths, the characteristic wavelengths including one or more of the peak wavelength λ max , the peak intensity I max , the wavelengths λ l (left side) and λ r (right side) corresponding to specific intensities nI max on both sides of the peak value, wherein 0.1<n<1;

[0054] S3, constructing a linear model of characteristic wavelengths and concentrations, the model being a single-wavelength model (with λ l or λ r as the variable), a double-wavelength model (with λ l and λ r as the variable), or a three-wavelength model (with λ l , λ max and λ rany one of the following: a single wavelength model, a two-wavelength model, and a three-wavelength model.

[0055] S4, determining an optimal fitting model by dynamically optimizing the scale factor n and the wavelength combination strategy (single wavelength, double wavelength, and triple wavelength), the optimal fitting model being one of the single wavelength model, the double wavelength model, and the triple wavelength model.

[0056] S5, calculating the concentration of the to-be-detected solution according to the characteristic wavelength of the to-be-detected solution by using the optimal fitting model.

[0057] The above embodiment of the present application can expand the linear detection range to more than 5 times of that of the traditional method by constructing a linear model of concentration based on the extracted characteristic wavelength, and can significantly improve the accuracy of fluorescence quantitative analysis, and the characteristic wavelength is hardly affected by the fluctuation of the light source, the shift of the receiving position, and other hardware parameters.

[0058] In the above embodiment of the present application, the linear model includes three models, and the complexity of the single model, the double model, and the triple model gradually increases, the single wavelength model has high fitting accuracy for the to-be-tested fluorescent substance, but the double wavelength model and the triple wavelength model can further increase the accuracy. In actual use, any one of the single wavelength model, the double wavelength model, and the triple wavelength model can be selected according to actual needs or conditions. For example, when a fluorescent substance that has not been tested is encountered, the accuracy of the single wavelength model may be insufficient, and at this time, the double wavelength model or the triple wavelength model can be used to increase the accuracy.

[0059] In some embodiments, in step S1, the device for measuring the fluorescence spectrum includes a light source, a collimation system, a cuvette, a converging system, and a detector; wherein the light source can adopt an LED, an LD, a laser, or the like, and the wavelength is selected according to the excitation wavelength of the fluorescent substance; the collimation system can adopt a convex lens, a collimation lens barrel, a fiber-fiber collimation mirror, or the like; the converging system is arranged in the direction perpendicular to the excitation light in the detection plane and located on the center line of the exit surface of the cuvette, and can adopt a convex lens, a fiber-fiber collimation mirror, or the like; and the detector can adopt a spectrometer or the like that can obtain full-spectrum data. The excitation light emitted by the light source is collimated by the collimation system, enters the center of the entrance surface of the cuvette, excites the fluorescent solution, and the emitted fluorescence is converged by the converging system and transmitted into the detector to be received. It should be noted that in actual implementation, the specific devices and structures in the above light source, collimation system, converging system, and detector are not limited to the above several kinds, and can be determined according to application needs.

[0060] In some embodiments, in step S2, all the acquired fluorescence spectra are pre-processed, including: removing the baseline generated by the electrical signal drift and dark current in the spectrometer, and removing the spectra containing gross errors according to the Raman criterion; using a digital filtering method to smooth the spectral data, suppress random noise, improve the quality of the spectral data, and ensure that the peak detection algorithm can accurately lock the peak position. For example, baseline correction to eliminate dark noise, and Savitzky-Golay filtering to smooth the spectral data segment.

[0061] In some embodiments, in step S2, when extracting the characteristic wavelength, first, the peak intensity I max and the corresponding wavelength λ max of the pre-processed spectral data are accurately identified by a peak detection algorithm. Then the target intensity is defined as nI max (0.1 < n < 1, step size 0.001). The range of n is limited to 0.1-1 for two reasons, one is that the lower limit 0.1 ensures that the target intensity is high enough above the background noise (especially for low concentration samples), avoiding wavelength extraction errors caused by random baseline fluctuations; the other is that the upper limit 1 ensures that the target intensity is lower than the peak intensity, covering the entire fluorescence peak profile. The step size of 0.001 is selected to balance the optimization accuracy and computational efficiency, which can capture the subtle changes of the optimal n value of different fluorescent substances, and also avoid excessive computational load affecting practical application. By setting the scaling factor, the uniformity of the target wavelength extraction of the fluorescence peak under different concentrations is ensured, providing a basis for subsequent feature selection and construction.

[0062] In some embodiments, in step S2, the linear interpolation method is used to calculate the corresponding λ max and λ l of nI r . Due to the data acquisition characteristics of the spectrometer, the acquired spectral data is distributed in discrete intervals. If interpolation is not performed, significant errors will be introduced in the target wavelength extraction process due to the limited resolution of the spectrometer, resulting in decreased model accuracy and concentration prediction bias. Linear interpolation is simple and efficient, and is suitable for the characteristics of fluorescence spectral data. In fluorescence spectra, the fluorescence peak only increases or decreases monotonically on both sides of the peak point, and other data points show a monotonous trend, so the error caused by the linear change assumption is small. In actual operation, find the first point (wavelength w1, intensity I1) P1 higher than the target intensity and the first point (wavelength w2, intensity I2) P2 lower than the target intensity on the spectral curve, and then use the linear interpolation formula to calculate the wavelength w corresponding to the target intensity:

[0063]

[0064] Linear interpolation ensures the accuracy of target wavelength extraction, effectively reduces the computational complexity, and provides a reliable data basis for model training and concentration prediction.

[0065] In some embodiments, in step S3, when constructing the linear model:

[0066] Single-wavelength model: the wavelength corresponding to the current target intensity is selected as the single feature variable from the left or right side of the peak wavelength λ max , which contains the information of the red shift of the fluorescence peak.

[0067] Double-wavelength model: two wavelengths λ max and λ l corresponding to the target intensity on both sides of the peak wavelength λ r are extracted, which are combined as the feature variable. This combination not only reflects the change of the red shift of the fluorescence peak, but also more comprehensively describes the change of the shape of the fluorescence peak. In the process of constructing the model, the correlation between λ l and λ r needs to be fitted to establish a one-to-one correspondence, avoiding the model error caused by the uncertainty of the wavelength combination, and improving the stability and prediction accuracy of the model.

[0068] Triple-wavelength model: on the basis of the double-wavelength model, the peak wavelength λ max is introduced as the third feature variable, forming a three-wavelength feature combination containing the peak wavelength and the wavelengths corresponding to the specific intensity on both sides of the peak, further enhancing the characterization ability of the fluorescence peak shape and covering the key position information of the fluorescence peak. When constructing the model, the relationship between the peak wavelength and the wavelengths corresponding to the target intensity on both sides needs to be fitted to ensure the accurate correspondence of the three wavelength variables under different concentrations, and to improve the stability and prediction accuracy of the model. The specific intensity refers to the target intensity, that is, n times the peak value of the current spectrum (n*Imax).

[0069] For the single-wavelength model, the feature wavelength of each concentration sample is λi (where i represents the ith sample), and the feature matrix is a column vector

[0070] [λ1 λ2 … λ m ] T

[0071] For the double-wavelength model, the feature wavelengths of each concentration sample are λ l,i and λ r,i , and the feature matrix is a 2-column matrix

[0072]

[0073] For the triple-wavelength model, the feature matrix is a 3-column matrix

[0074]

[0075] wherein r, l respectively refer to the target wavelengths corresponding to the right and left sides of the peak, and m is the total number of samples. This matrix is used as the input of linear regression to establish the relationship with the concentration. First, the fitting formula in matrix form is established

[0076] C = A - W + b

[0077] wherein C is the concentration vector of the sample ([C1C2… C m ] T ), A is the characteristic matrix, W is the weight matrix of the characteristic wavelengths (for example, [W1W2W3] T in the three-wavelength model), and b is the bias vector ([b1b2… b m ] T ). The fitting formula in matrix form is solved by linear regression to determine W and b.

[0078] For an unknown concentration solution of a fluorescent substance, after obtaining its characteristic wavelengths (λ max , λ l , λ r ), the concentration (c) thereof is calculated using the formula:

[0079] c = [λ l λ max λ r ] - W + b

[0080] For single-wavelength or double-wavelength models, the formula is adjusted by reducing the number of characteristic wavelength terms accordingly.

[0081] In some embodiments, in step S4, for single-, double-, and three-wavelength models, the corresponding characteristic variables are calculated based on each n value (0.1 < n < 1, step size 0.001), and the corresponding characteristic matrix is constructed by traversing different n values. A linear regression model is trained for each n value, and the characteristic matrix is fitted to the sample concentration. The R 2 , mean absolute percentage error (MAPE), and mean squared error (MSE) are used to quantify the performance of the model, and the model with the highest R 2 and the lowest MAPE and MSE among all n values is selected as the optimal model.

[0082] For double-wavelength and three-wavelength models, to further optimize the model performance and ensure the uniqueness of the characteristic variable combination at a single concentration, the relationship between the characteristic variables must be fitted. Specifically, in the double-wavelength model, the relationship between the wavelengths corresponding to the target intensities on both sides of the peak wavelength is fitted to establish a one-to-one correspondence between them, avoiding model errors caused by uncertainty in wavelength combination; in the three-wavelength model, the relationship between the peak wavelength and the wavelengths corresponding to the target intensities on both sides is fitted to ensure the accurate correspondence of the three wavelength variables at different concentrations, improving the stability and prediction accuracy of the model.

[0083] The above embodiment of the present application realizes wide-range and high-precision quantitative analysis of fluorescent substances based on the CDRS phenomenon induced by IFE, by extracting the concentration-dependent characteristic wavelength to construct a linear model. The method overcomes the defects of the traditional fluorescence intensity method, such as large IFE influence and narrow linear range, and the characteristic wavelength is less affected by the fluctuation of hardware parameters, and has high stability.

[0084] To verify the method of the above embodiment of the present application, a series of concentration gradient rhodamine B, fluorescein sodium and chlorophyll a solutions are used for experimental verification and modeling. For the three fluorescent substances, the single, double and triple wavelength models can all obtain excellent prediction models.

[0085] In a specific embodiment, the sample to be measured is a prepared rhodamine B aqueous solution with a concentration range of 40 mg / L-400 mg / L and 10 concentration gradients. In other embodiments, other fluorescent solutions with CDRS can also be selected, including but not limited to rhodamine B ethanol solution, rhodamine B glycerol solution, rhodamine 6G aqueous solution, fluorescein sodium aqueous solution and chlorophyll ethanol solution, etc.

[0086] The device used to measure the fluorescence spectrum of the fluorescent solution includes a light source, a collimating system, a cuvette, a converging system and a detector. Specifically, the light source is selected as a LED with a center wavelength of 460 nm, the collimating system and the converging system are both selected as concave lenses, and the detector is selected as a spectrometer with a model of QE65000 from Ocean Optics. The excitation light emitted by the LED is collimated by the convex lens, is incident from the center of the entrance surface of the four-way light path cuvette with an optical path of 10 mm, excites the fluorescent solution, the convex lens is arranged in the detection plane perpendicular to the direction of the excitation light and on the center line of the exit surface of the cuvette, converges the emitted fluorescence, and the fluorescence is transmitted into the spectrometer through the optical fiber and is received.

[0087] The accurate fluorescent substance characteristic wavelength quantitative analysis method based on the concentration-dependent red shift induced by the inner filter effect includes the following steps:

[0088] Step 1: 200 mg of rhodamine B (AR, Shanghai Maikelin Biochemical Co., Ltd.) powder is weighed using a high-precision electronic balance and is dissolved in a 500 mL volumetric flask to prepare a fluorescein sodium aqueous solution with a concentration of 400 mg / L. The solution is diluted with deionized water to a series of concentration gradients of 40 mg / L-400 mg / L of rhodamine B aqueous solution, and the concentrations are 40 mg / L, 80 mg / L, 120 mg / L, 160 mg / L, 200 mg / L, 240 mg / L, 280 mg / L, 320 mg / L, 360 mg / L and 400 mg / L.

[0089] Step 2: Put the prepared 10 different concentrations of sodium fluorescein aqueous solution into the fluorescence device to measure the fluorescence spectrum, set the integration time to 100 ms, measure the fluorescence spectrum of each concentration sample for 50 times, eliminate the spectrum with large fluctuation, and average the remaining spectrum. The measured fluorescence spectrum is shown in FIG. 1. Figure 2

[0090] Step 3: Pretreat all spectra, remove the baseline generated by the electrical signal drift and dark current in the spectrometer, eliminate the spectrum containing gross error according to the Lillie criterion, and smooth the spectral data by using Savitzky-Golay filter.

[0091] Step 4: Determine the peak wavelength λ max and the peak intensity I max by using the peak detection algorithm, and determine the wavelength λ max (left) and λ l (right) corresponding to the specific intensity nI r on both sides, set the value range of n to be 0.1-1, and the step size to be 0.001.

[0092] Step 5: Construct single-wavelength, double-wavelength and triple-wavelength models respectively, for the double-wavelength model, fit the correlation between λ l and λ r , and for the triple-wavelength model, fit the relationship between λ max and λ l , λ r .

[0093] Step 6: Traverse the value of n, calculate the R 2 , MAPE and MSE of each model, and determine the optimal model. For rhodamine B, in the single-wavelength method, the optimal target intensity is determined to be 0.524I max by optimizing the proportion factor n, the characteristic wavelength corresponding to the concentration fitting model has R 2 = 0.999, MAPE = 0.998%, and MSE = 6.82; for the double-wavelength method, the determination coefficient (R 2 ) of the initial fitting plane is 0.999, the MAPE value is 0.095%, and the MSE value is 0.036, the correlation between λ l and λ r is fitted, the R 2 value is 0.999, the MAPE value is 0.007%, after introducing the constraint, the model has R 2 = 0.999, MAPE = 0.071%, and MSE = 0.021; for the triple-wavelength model, the determination coefficient (R 2 ) is 0.999, the MAPE value is 0.087%, and the MSE value is 0.034.

[0094] ​Step 7: For the unknown concentration of Rhodamine B solution, repeat steps 2-4 to extract the characteristic wavelength, and substitute it into the optimal model to calculate its concentration.

[0095] Result analysis:

[0096] In Figure 2 a series of concentration gradient of Rhodamine B fluorescence spectrum, the observation of a single spectrum found that the fluorescence peak was asymmetric, and the downward trend on the right side of the peak was more gentle than on the left side. By observing all the spectra, it can be seen that as the concentration increases, the shape of the fluorescence peak changes, the fluorescence peak as a whole red shifts, and the fluorescence peak wavelength λ max continuously increases. Considering the asymmetry of the fluorescence peak, we take the wavelengths λ l , λ r corresponding to 1 / 2 peak height on the left and right sides of the peak, and take the wavelength difference between the two as the bandwidth Δλ, which can be found that the three variables also continuously increase with the concentration. At the same time, the fluorescence intensity first increases and then decreases, and the quenching occurs at high concentration, which verifies the rationality of the characteristic wavelength as a quantitative index.

[0097] Figure 3 The single-wavelength model fitting result of Rhodamine B shows that the fitting curve R 2 = 0.999483, and the linear range covers 40-400 mg / L, which is 5 times larger than the traditional intensity method (80 mg / L).

[0098] Figure 4 The double-wavelength model further optimizes the quantitative precision: the λ l , λ r of Rhodamine B shows a strong linear correlation (R 2 = 0.99985), and the spatial curve fitting based on this constraint has a MSE of 0.021, which is 41.7% lower than the unconstrained model; the double-wavelength combination captures the peak red shift and peak shape change, and the characterization ability of the concentration is better than that of a single wavelength, especially in the medium and high concentration segment (200-400 mg / L), the prediction error is less than 0.5%.

[0099] Figure 5 The three-wavelength model of Rhodamine B integrates three-dimensional features, realizes the highest precision, and further improves the indicators of the fitting model.

[0100] In addition, the researchers of the present application also explored the influence of light source intensity fluctuation and receiving position change on the traditional method, and verified the detection stability of the method which does not depend on hardware parameters. In order to solve the problem of insufficient resolution of the spectrometer, the linear interpolation method can be introduced to reduce the error. Compared with the traditional fluorescence spectrum method based on intensity, the method of the present application can greatly expand the range and reduce the prediction error, and is not affected by some interference factors in the traditional method, which provides a new solution for the quantitative detection of high concentration fluorescence.

[0101] The specific embodiments of the present application have been described. It is to be understood that the application is not limited to particular details of the embodiments described, as modifications and variations can be made within the scope of the application by those skilled in the art. Other preferred features and characteristics of the present application will become apparent from the above description, and it is intended to cover any and all adaptations and modifications of variants of the application and with the scope of the claims.

Claims

1. A method for quantitative analysis of characteristic wavelengths of fluorescent substances, characterized in that: Comprising: Obtaining the fluorescence spectra of a series of fluorescent substance solutions with different concentrations; Preprocessing the fluorescence spectra and extracting the characteristic wavelengths; The characteristic wavelength includes a peak wavelength λ max , peak intensity I max Specific intensity on both sides nI max The corresponding left wavelength λ l , right wavelength λ r Any one or more of the following, where n is the scaling factor, 0.1 <n<1; Construct a linear model of characteristic wavelength and concentration, the linear model is based on λ l or λ r The single wavelength model with λ as the variable l and λ r The dual-wavelength model with λ as the variable l ,λ max and λ r Any of the three-wavelength models with ; Determining the optimal fitting model by dynamically optimizing the scaling factor n and the wavelength combination strategy; Using the optimal fitting model to calculate the concentration of the test solution according to its characteristic wavelengths.

2. The method according to claim 1, characterized in that The preprocessing of the fluorescence spectra includes: Baseline correction to eliminate dark noise; Using digital filtering to smooth the spectral data segment.

3. The method according to claim 1, characterized in that The extraction of the characteristic wavelengths includes: Determine nI using a peak detection algorithm max and λ max ; Calculate nI using linear interpolation max The corresponding λ l and λ r , where the step size of n is 0.

001.

4. The method according to claim 1, wherein Constructing a linear model between the characteristic wavelengths and the concentration, where: The single wavelength model: From the peak wavelength λ max Select the wavelength λ corresponding to the current target intensity on the left or right side l or λ r As a single feature variable, this variable contains information about the red shift of the fluorescence peak; The dual-wavelength model: Simultaneously extract the peak wavelength λ max Two wavelengths λ on both sides corresponding to the target intensity l and λ r , change λ l ,λ r The combination is used as a characteristic variable, which can reflect the change of the red shift of the fluorescence peak and describe the change of the fluorescence peak shape. The three-wavelength model: extracting the peak wavelength λ max and the two wavelengths λ on both sides corresponding to the target intensity l and λ r , change λ l ,λ r ,λ max As three characteristic variables, a three-wavelength characteristic combination is formed, including the peak wavelength and the specific intensity wavelengths corresponding to the peak on both sides, which further enhances the ability to characterize the fluorescence peak morphology and covers the key position information of the fluorescence peak.

5. The method according to claim 4, characterized in that Constructing a linear model between the characteristic wavelengths and the concentration, where: The dual-wavelength model needs to fit λ l and λ r The relevance of to avoid ambiguous interpretation; The three-wavelength model needs to be further fitted with λ max and λ l ,λ r relationships to enhance stability.

6. The method according to claim 5, characterized in that In the dual-wavelength model, the fitting peak wavelength λ max The wavelength λ of the target intensity on both sides l and λ r The relationship between them is established, and a one-to-one correspondence is established to ensure the certainty of the wavelength combination; In the three-wavelength model, the fitting peak wavelength λ max The target intensity wavelength λ corresponding to both sides max and λ l ,λ r , establish their corresponding relationship, and ensure the accurate correspondence of the three wavelength variables at different concentrations.

7. The method according to claim 1, characterized in that The determination of the optimal fitting model by dynamically optimizing the scaling factor n and the wavelength combination strategy includes: For the single-wavelength model, the dual-wavelength model or the triple-wavelength model, calculating the corresponding characteristic variables based on each n value, where 0.1 < n < 1 and the step size is 0.001, and constructing the corresponding characteristic matrix by traversing different n values; Training a linear regression model for each n value, fitting the characteristic matrix to the sample concentration, quantifying the model performance using evaluation metrics, and determining the optimal fitting model according to the evaluation metrics.

8. The method according to claim 7, characterized in that The evaluation index includes the determination coefficient R 2 , mean absolute percentage error MAPE and mean square error MSE, the best fitting model is R 2 The model with the highest MAPE and lowest MSE.

9. The method according to any one of claims 1 to 8, characterized in that The obtaining of the fluorescence spectra of a series of fluorescent substance solutions with different concentrations includes: Using a device composed of an excitation light source, a collimation system, a cuvette, a converging system and a detector. The excitation light emitted by the light source is collimated by the collimation system and then enters the center of the incident surface of the cuvette to excite the fluorescent substance solution. The fluorescence emitted and converged by the converging system is transmitted to the detector for reception.

10. The method according to any one of claims 1 to 8, characterized in that The fluorescent substance is a fluorescent substance with a concentration-dependent red-shift characteristic, including any one of rhodamine B, sodium fluorescein, and chlorophyll a.

Images