Method for identifying production places of traditional Chinese medicinal materials based on Raman spectrum

Abstract

The invention belongs to the technical field of Raman spectrum, and relates to detection of traditional Chinese medicinal materials, in particular to a Raman spectrum-based identification method for producing areas of traditional Chinese medicinal materials. The method comprises the following steps: preparing to-be-detected traditional Chinese medicinal materials into traditional Chinese medicinal material powder, adding the traditional Chinese medicinal material powder into water, performing ultrasonic treatment and standing, and obtaining supernate with a Tyndall effect as a to-be-detected sample; detecting the sample to be detected by adopting a Raman spectrum to obtain original Raman spectrum data; performing residual error processing and residual error smoothing processing on the original Raman spectrum data in sequence, and identifying the producing area of the to-be-detected traditional Chinese medicinal material according to the result of the residual error smoothing processing. According to the identification method provided by the invention, the classification effect can be realized by using a model with a simple basis, the detection operation is simple, convenient and rapid, and the method has the advantages of high accuracy, good reproducibility and the like.

Description

Technical Field

[0001] This invention belongs to the field of Raman spectroscopy technology and relates to the detection of Chinese medicinal materials, specifically to a method for identifying the origin of Chinese medicinal materials based on Raman spectroscopy. Background Technology

[0002] The information disclosed in this background section is intended only to enhance understanding of the overall background of the invention and is not necessarily to be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art.

[0003] Traditional methods for identifying Chinese medicinal herbs mainly rely on macroscopic characteristics such as appearance, color, odor, and texture, as well as microscopic examination of tissue structure and cell morphology, depending on the personal experience and subjective judgment of the identification experts. Other common identification techniques for Chinese medicinal herbs include spectroscopic methods, chromatographic methods, mass spectrometry, and DNA molecular identification. Raman spectroscopy is a scattering spectroscopy technique based on the vibrational-rotational energy level transitions of molecules. Its essence is inelastic scattering of light; when photons collide with molecules, energy loss occurs (i.e., Raman shift). The Raman shift corresponds to the energy level difference between molecular vibrations and rotations, and by analyzing the Raman shift, relevant information about the molecules can be obtained. Furthermore, Raman spectroscopy, with its non-destructive, rapid, and fingerprint-like characteristics, can achieve non-destructive, high-resolution, in-situ analysis of the chemical composition and structure of substances, making its applications extremely wide-ranging. Raman spectroscopy is extremely sensitive to minute changes in molecules. Medicinal herbs from different origins and growing environments will exhibit subtle differences in the types and contents of their secondary metabolites (such as alkaloids, flavonoids, and saponins).

[0004] Currently, there are two main methods for sample pretreatment in Raman spectroscopy for identifying Chinese medicinal materials: 1. directly detecting the medicinal materials by turning them into powder; 2. extracting the medicinal materials and then detecting the extracted samples. However, these two methods of sample pretreatment for Raman spectroscopy identification have the following difficulties: When the medicinal materials are made into powder, the presence of large amounts of impurities such as cellulose in the powder leads to highly unstable signals and poor reproducibility in Raman spectroscopy; while the sample pretreatment method for extracting medicinal materials is complex and time-consuming. Therefore, there is an urgent need for a rapid, simple, accurate, and reproducible method for the identification and classification of Chinese medicinal materials. Summary of the Invention

[0005] To address the shortcomings of existing technologies, the purpose of this invention is to provide a method for identifying the origin of Chinese medicinal materials based on Raman spectroscopy. The identification method provided by this invention can achieve classification results using basic and simple models (PLS-DA, SVM), and the detection operation is simple and fast, with advantages such as high accuracy and good reproducibility.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] A method for identifying the origin of traditional Chinese medicinal materials based on Raman spectroscopy includes the following steps: The Chinese medicinal materials to be tested were made into powder, and the powder was added to water, ultrasonically treated and allowed to stand. The supernatant with Tyndall effect was obtained as the sample to be tested. The sample to be tested was detected using Raman spectroscopy to obtain raw Raman spectral data; The raw Raman spectral data were subjected to residual processing and residual smoothing in sequence. The origin of the Chinese medicinal materials to be tested was identified based on the results of the residual smoothing.

[0008] The Tyndall effect refers to the phenomenon where, when a beam of light passes through a colloid (a dispersion system in which the dispersed particles have diameters between 1 and 100 nanometers), the light forms a bright and clear path within the colloid when viewed from a direction perpendicular to the beam. When a sample solution exhibits the Tyndall effect, it indicates that the molecules in the solution can interact with light, producing significant scattered light. Therefore, this invention uses Raman spectroscopy to detect solutions exhibiting the Tyndall effect, obtaining relevant information about the sample and thus achieving the purpose of identification.

[0009] This invention demonstrates through experiments that when Raman spectroscopy is used to detect supernatants exhibiting the Tyndall effect, the obtained raw Raman spectral data possesses key characteristics, laying the foundation for identifying the origin of Chinese medicinal materials. Experiments show that the raw Raman spectral data not only contains key characteristics but also includes noise, baseline drift residues, or differences between samples. Through residual processing and residual smoothing, random noise in the residuals is effectively suppressed, while more regular fluctuation trends are preserved. This allows for the extraction of characteristic peaks and information from the samples, thereby enabling the identification of the origin of Chinese medicinal materials.

[0010] The beneficial effects of this invention are as follows: 1. The identification method proposed in this invention has a simple pretreatment process, which only requires dispersing the powder in water and ensuring the formation of a colloidal solution. There is no need for extraction, separation and other steps, which greatly saves time and cost. Moreover, the sample adaptability of this method is wide.

[0011] 2. The identification method proposed in this paper selects Raman spectroscopy as the detection means. Raman spectroscopy has strong specificity. Herbs from different places of origin and different growing environments will have slight differences in secondary metabolites, material composition and content. Raman spectroscopy can capture these minute differences.

[0012] 3. The premise for the identification and classification proposed in this paper is that the sample solution has the Tyndall effect. The Tyndall effect proves that the system is a homogeneous colloid, which avoids the shortcomings of Raman peak shift and intensity fluctuation caused by particle agglomeration in "direct powder detection", and further ensures the accuracy and reproducibility of classification. Attached Figure Description

[0013] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0014] Figure 1 Raman spectra of sample solutions of different concentrations in this invention; Figure 2 The following is a correlation diagram of the Raman characteristics of cinnamon in Example 2 of the present invention, wherein a. is the infrared absorption spectrum of cinnamic acid standard (solid powder); b. is the Raman spectrum of cinnamic acid standard (solid powder); c. is the Raman spectrum of aqueous solution of cinnamic acid standard; d. is the residual Raman spectrum of aqueous solution of cinnamic powder. Figure 3 The following are Raman characteristic correlation diagrams of wild chrysanthemum in Example 1 of the present invention; wherein a. is the infrared absorption spectrum of scutellarin standard (solid powder); b. is the Raman spectrum of scutellarin standard (solid powder); c. is the Raman spectrum of scutellarin standard aqueous solution; d. is the residual Raman spectrum of wild chrysanthemum powder aqueous solution; Figure 4 The images show the original Raman spectra (a), residual Raman spectra (b), and residual smoothed Raman spectra (c) of the wild chrysanthemum powder solution in Example 1 of this invention. Figure 5 The images show the Raman spectra of the original data of the cinnamon powder solution (a), the Raman spectra of the residual data of the cinnamon powder solution (b), and the Raman spectra of the smoothed residual data of the cinnamon powder solution (c) in Example 2 of the present invention. Figure 6 Raman spectra of wild chrysanthemum powder from different origins in Example 1 of the present invention, where a represents Sichuan (SC) origin and b represents Hubei (HB) origin; Figure 7 This is a response diagram of the PLS-DA classification results of the original data of wild chrysanthemum samples in Embodiment 1 of the present invention; Figure 8 This is a response graph of the PLS-DA classification results of the residual data of wild chrysanthemum samples in Embodiment 1 of the present invention; Figure 9 The response graph shows the PLS-DA classification results of the residual smoothing data of wild chrysanthemum samples in Embodiment 1 of the present invention. Figure 10This is a response graph of the SVM classification results of the original data of wild chrysanthemum samples in Embodiment 1 of the present invention; Figure 11 This is a response graph of the SVM classification results of the wild chrysanthemum sample residual data in Embodiment 1 of the present invention; Figure 12 This is a response graph of the SVM classification results of the residual smoothing data of wild chrysanthemum samples in Embodiment 1 of the present invention; Figure 13 The above are the confusion matrix and ROC curve of the PLS-DA classification results of the wild chrysanthemum sample raw data in Embodiment 1 of the present invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Sichuan (SC) and Category 2 is Hubei (HB). Figure 14 The above are the confusion matrix and ROC curve of the PLS-DA classification results of the wild chrysanthemum sample residual data in Embodiment 1 of the present invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Sichuan (SC) and Category 2 is Hubei (HB). Figure 15 The above are the confusion matrix and ROC curve of the PLS-DA classification results of wild chrysanthemum sample residual smoothing data in Embodiment 1 of the present invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Sichuan (SC) and Category 2 is Hubei (HB). Figure 16 The above are the confusion matrix and ROC curve of the SVM classification results of the original wild chrysanthemum sample data in Embodiment 1 of the present invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Sichuan (SC) and Category 2 is Hubei (HB). Figure 17 The above are the confusion matrix and ROC curve of the SVM classification results of the wild chrysanthemum sample residual data in Embodiment 1 of the present invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Sichuan (SC) and Category 2 is Hubei (HB). Figure 18 The above are the confusion matrix and ROC curve of the SVM classification results of the wild chrysanthemum sample residual smoothing data in Embodiment 1 of the present invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Sichuan (SC) and Category 2 is Hubei (HB). Figure 19The diagram shows the methodological results in Embodiment 1 of the present invention, where a represents the precision result, b represents the stability, c represents the intra-day precision result, d represents the inter-day precision result, and e represents the reproducibility result. Figure 20 The images show the Raman spectra of cinnamon powders from different origins in Example 2 of this invention, where a represents Guangxi (GX) and b represents Fujian (FJ). Figure 21 This is a response graph of the PLS-DA classification results of the original data of the cinnamon sample in Embodiment 2 of the present invention; Figure 22 This is a response graph of the PLS-DA classification results of the cinnamon sample residual data in Embodiment 2 of the present invention; Figure 23 This is a response graph of the PLS-DA classification results of the residual smoothing data of the cinnamon sample in Embodiment 2 of the present invention; Figure 24 This is a response graph of the SVM classification results of the original cinnamon sample data in Embodiment 2 of the present invention; Figure 25 This is a response graph of the SVM classification results of the cinnamon sample residual data in Embodiment 2 of the present invention; Figure 26 This is a response graph of the SVM classification results of the residual smoothing data of the cinnamon sample in Embodiment 2 of the present invention; Figure 27 The confusion matrix and ROC curve of the PLS-DA classification results of the original number of cinnamon samples in Example 2 of this invention are shown. In the figure, a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Guangxi (GX) and Category 2 is Fujian (FJ). Figure 28 The confusion matrix and ROC curve of the PLS-DA classification results of the residual data of cinnamon samples in Example 2 of this invention are shown. Among them, a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Guangxi (GX) and Category 2 is Fujian (FJ). Figure 29 The confusion matrix and ROC curve of the PLS-DA classification results of the residual smoothing data of cinnamon samples in Example 2 of this invention are shown. Among them, a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Guangxi (GX) and Category 2 is Fujian (FJ). Figure 30The above are the confusion matrix and ROC curve of the SVM classification results of the original data of cinnamon samples in Example 2 of this invention, where a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Guangxi (GX) and Category 2 is Fujian (FJ). Figure 31 The confusion matrix and ROC curve of the SVM classification results of the cinnamon sample residual data in Example 2 of this invention are shown. Here, a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Guangxi (GX) and Category 2 is Fujian (FJ). Figure 32 The confusion matrix and ROC curve of the SVM classification results of the residual smoothing data of cinnamon samples in Example 2 of this invention are shown. Among them, a is the confusion matrix of the calibration set, b is the confusion matrix of the validation set, c is the ROC curve of the calibration set, and d is the ROC curve of the validation set. Category 1 is Guangxi (GX) and Category 2 is Fujian (FJ). Figure 33 The diagram shows the methodological results in Embodiment 2 of the present invention, where a represents the precision result, b represents the stability, c represents the intra-day precision result, d represents the inter-day precision result, and e represents the reproducibility result. Detailed Implementation

[0015] It should be noted that the following detailed descriptions are exemplary and intended to provide further illustration of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0016] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0017] The main methods for processing the Raman spectral data of the obtained samples are as follows: The principle of residual processing is as follows: 1. Using the Raman spectrum of the blank pure aqueous solution as the independent variable X (dimension m×1) and the Chinese medicinal material sample solution as the dependent variable Y (dimension m×1), linear regression analysis was performed using software to obtain the linear relationship between the independent variable X and the dependent variable Y. X_aug = [ones(length(X), 1),X]; (1) Where X_aug represents adding a column vector with all 1s before the independent variable X; [b, bint, r, rint] = regress(Y, X_aug); (2) b represents the parameter estimate of the least squares method (including the slope and intercept of the regression); bint represents the confidence interval of the regression; r represents the residual between X and Y; rint represents the confidence interval of the residual.

[0018] 2. Obtain the residuals between X and Y based on the linear relationship between X and Y, and obtain the differences between different Chinese medicinal herb powders by comparing the residuals; The principle of Partial Least Squares Discrimination (PLS-DA) is as follows: 1. The Raman spectral data X is divided into samples using the Kennard-Stone method to obtain the calibration set X_train and the validation set X_test; as well as the label data calibration set Y_train and the validation set Y_test; 2. Train the PLS-DA model on the calibration set data to obtain the prediction results of the label data Y calibration set and validation set; (3) (4) In equations (3) and (4), T represents the score matrix, P represents the load matrix of X, Q represents the load matrix of Y, E represents the residual matrix of X, and F represents the residual matrix of Y.

[0019] 3. Draw the confusion matrix and calculate the evaluation metrics of the classification model, including precision, recall, accuracy, and F1 score; (5) (6) (7) (8) TP indicates that the actual instance is positive and the model predicts it to be positive (correct prediction); FP indicates that the actual instance is negative but the model predicts it to be positive (false alarm: a negative instance is treated as a positive instance); TN indicates that the actual instance is negative and the model predicts it to be negative (correct prediction); FN indicates that the actual instance is positive but the model predicts it to be negative (missed detection: a positive instance is treated as a negative instance).

[0020] The principle of Support Vector Machine (SVM) classification is as follows: 1. Data Preparation and Division. The Raman spectroscopy data X and Y are merged into a new matrix W (containing m samples, each sample containing n features and 1 output value); the calibration set and validation set are then obtained by dividing the matrix using the Kennard-Stone method according to a certain ratio. 2. Normalization. The partitioned data is normalized to the [0, 1] interval using mapminmax; the normalization formula is as follows: (9) 3. Establish a support vector machine classification model; 4. Use a linear kernel function to train the support vector machine classification model.

[0021] 5. Draw the confusion matrix and calculate the performance metrics for evaluating the model: precision, recall, accuracy, and F1 score.

[0022] Methodological examination of this invention: Precision test: Select a sample of appropriate concentration (in this invention, this refers to 0.1% wild chrysanthemum or 0.1% cinnamon sample) for testing, measure continuously for 5 times, and record the Raman spectral intensity measurement results. Calculate the precision result RSD < 5%.

[0023] Stability test: To examine the stability of the samples, samples of appropriate concentrations (in this invention, 0.1% wild chrysanthemum or 0.1% cinnamon) were selected for testing. The absorbance intensity was measured at 0, 2, 4, 6, 8, and 12 hours, with each measurement repeated five times. The average value was calculated, and the relative standard deviation (RSD) of the stability was determined to be <5%. Repeatability testing: mainly examines the consistency of measurement results. Repeatability testing includes intra-day precision and progressive precision.

[0024] Intra-day precision: Samples of appropriate concentration (0.1% wild chrysanthemum or 0.1% cinnamon in this invention) were selected for testing. The absorption intensity was measured at 3 time points (2 hours apart) within one day. Each time point was tested 5 times, for a total of 15 times. The change in absorption intensity was recorded. The relative standard deviation (RSD) of intra-day precision was calculated to be <5%.

[0025] Daytime precision: Samples of appropriate concentration (in this invention, this refers to 0.1% wild chrysanthemum or 0.1% cinnamon) were selected for testing, and the tests were conducted continuously for 3 days, 5 times a day, and the changes in absorption intensity were recorded; the relative standard deviation (RSD) of daytime precision was calculated to be <5%.

[0026] Reproducibility test: This test mainly examines the reproducibility of the detection method. Five samples of 0.1% wild chrysanthemum were prepared, and the absorption intensity of each sample was tested. Each sample was tested five times, and the average value was taken. The data were recorded, and the relative standard deviation (RSD) of the reproducibility was calculated to be <5%.

[0027] Given that existing technologies for the identification and classification of Chinese medicinal materials are difficult to simultaneously achieve speed, simplicity, accuracy, and stability, this invention proposes a method for identifying the origin of Chinese medicinal materials based on Raman spectroscopy.

[0028] A typical embodiment of the present invention provides a method for identifying the origin of traditional Chinese medicinal materials based on Raman spectroscopy, comprising the following steps: The Chinese medicinal materials to be tested were made into powder, and the powder was added to water, ultrasonically treated and allowed to stand. The supernatant with Tyndall effect was obtained as the sample to be tested. The sample to be tested was detected using Raman spectroscopy to obtain raw Raman spectral data; The raw Raman spectral data were subjected to residual processing and residual smoothing in sequence. The origin of the Chinese medicinal materials to be tested was identified based on the results of the residual smoothing.

[0029] In some embodiments, the process of establishing a model for identifying the origin of Chinese medicinal materials includes the following steps: At least two kinds of Chinese medicinal materials from different origins were made into Chinese medicinal material powders, and the different Chinese medicinal material powders were added to water, ultrasonically treated and allowed to stand, and the supernatant with Tyndall effect was obtained as the sample solution. The sample solution was analyzed using Raman spectroscopy to obtain raw Raman spectral data; The raw Raman spectral data were subjected to residual processing and residual smoothing in sequence. Based on the results of the residual smoothing, a model for identifying the origin of Chinese medicinal materials was established.

[0030] Specifically, the process of establishing a Chinese medicinal herb origin identification model also includes using Raman spectroscopy to detect the sample solution in a standard solution, where the standard solution serves as a quality marker for the medicinal herb. These quality markers are characteristic compounds representing the quality of the medicinal herb; for example, cinnamic acid is the quality marker for cinnamon, and buddleja glycoside is the quality marker for wild chrysanthemum. During the establishment of the Chinese medicinal herb origin identification model, infrared spectroscopy can also be used to detect the sample solution and the standard solution. Infrared spectroscopy further ensures that characteristic peaks are retained in the aqueous solution of the quality marker, providing a basis for classification.

[0031] In some embodiments, when preparing the sample to be tested, the mass of the medicinal material added per milliliter of water is 0.09~0.11 mg. Studies have shown that the identification effect is better under these conditions.

[0032] In some embodiments, the ultrasonic treatment time is 25-35 minutes. This condition allows for better dispersion of the Chinese herbal medicine powder in water.

[0033] In some embodiments, the settling time after ultrasonic treatment is 25-35 minutes.

[0034] In some embodiments, the model used for residual processing and residual smoothing is partial least squares discriminant analysis or support vector machine. Studies have shown that these two models can identify the origin of Chinese medicinal materials. Research indicates that using support vector machine for residual processing and residual smoothing yields better identification results.

[0035] In some embodiments, residual smoothing is performed using SG smoothing.

[0036] In some embodiments, the medicinal materials are cinnamon and / or wild chrysanthemum. This invention verifies that the method is applicable to the identification of the origin of these two medicinal materials using cinnamon and wild chrysanthemum.

[0037] Since the chemical composition of Chinese medicinal materials from different producing areas may vary, the embodiments of the present invention provide an identification basis for the application of the above identification method in the quality monitoring of Chinese medicinal materials.

[0038] To enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be described in detail below with reference to specific embodiments.

[0039] The identification method for Chinese medicinal materials based on Raman spectroscopy used in the following embodiments includes the following steps: 1. The Chinese medicinal materials are pulverized by breaking down the cell walls, and then large particles are removed by sieving through a 140-mesh pharmacopoeia sieve. Finally, the Chinese medicinal material powder that passes through the 140-mesh pharmacopoeia sieve is obtained. 2. Take an appropriate amount of the above-mentioned medicinal powder, add about 15 mL of purified water, mix well, and sonicate for 30 minutes to allow the medicinal powder to fully dissolve and diffuse in the aqueous solution. 3. After sonicating the sample, let it stand for about 30 minutes, then take an appropriate amount of supernatant into a clean glass test tube. Illuminate the glass test tube with a laser pointer in a dark environment to ensure that the supernatant exhibits the Tyndall effect. 4. Take approximately 400 μL of the supernatant of the sample that produces the Tyndall effect and place it in a 1 mm quartz cuvette. 5. Place the cuvette on the stage of the Raman spectrometer, measure the Raman spectral intensity of the sample, and record the data; 6. Perform appropriate preprocessing on the obtained Raman spectral data, establish a classification model, and realize the classification and identification of Chinese medicinal materials.

[0040] Investigation on the concentration range of Chinese medicinal herb powder aqueous solution Step 1: Take 0.015 g, 0.030 g, 0.045 g, and 0.075 g of Chinese medicinal materials (wild chrysanthemum or cinnamon) powder that have passed through a 140-mesh pharmacopoeia sieve, dissolve them in 15 mL of purified water, mix well, and obtain sample solutions with concentrations of 0.1%, 0.2%, 0.3%, and 0.5%. Sonicate the sample solutions for 30 min. Step Two: After sonicating the sample and allowing it to stand for approximately 30 minutes, take an appropriate amount of the supernatant into a clean glass test tube. Illuminate the glass test tube with a laser pointer in a dark environment and observe whether the supernatant exhibits the Tyndall effect. If the supernatant exhibits the Tyndall effect, continue with Step Three. If the supernatant does not exhibit the Tyndall effect, it indicates that the concentration of this solution is too high and unsuitable for subsequent operations; this concentration should be discarded (see [link to explanation] for the reason). Figure 1 ,Depend on Figure 1 It can be seen that when the concentration is greater than 0.1%, the Raman spectral intensity of the solution is lower than that of the aqueous solution. Moreover, when the concentration is greater than 0.1%, the Raman signal is easily masked by the strong background response of water. After ultrasonic treatment and standing for 30 minutes, the Tyndall effect cannot be generated immediately, making it difficult to achieve the purpose of rapid detection (the Tyndall effect will appear when the standing time is long enough (about 12 hours).

[0041] Step 3: Perform Raman spectroscopy on the supernatant of the sample that produces the Tyndall effect and record the Raman spectral intensity.

[0042] The results are as follows Figure 1 As shown in the figure, when the sample concentration is 0.1%, its Raman spectral intensity differs significantly from that of pure water. When the sample concentration is greater than 0.1%, its Raman spectral intensity is lower than that of water. This may be related to the aggregation effect of solute molecules at high concentrations. As the solute concentration increases, intermolecular interactions strengthen (such as the formation of aggregates), which weakens the Raman scattering efficiency of individual solute molecules. When the sample concentration is less than 0.1%, the Raman scattering signal is too weak due to the low concentration, and the Raman signal is easily masked by the strong background response of water. Therefore, a sample with a concentration of 0.1% was selected for subsequent experiments.

[0043] Relationship between powder solutions and their markers Step 1: Prepare a 0.1% solution of cinnamon powder and wild chrysanthemum powder. Take about 400 μL into a 1 mm quartz cuvette; place the cuvette on the stage of the Raman spectrometer, and measure the Raman spectral intensity of the sample, recording the data.

[0044] Step 2: Take an appropriate amount of quality marker (such as cinnamon: cinnamic acid; wild chrysanthemum: mongholic acid, etc.) and place it on the stage of the Raman spectrometer. Detect the Raman spectral intensity of the samples and record the data.

[0045] Step 3: Take an appropriate amount of quality markers (such as cinnamon: cinnamic acid; wild chrysanthemum: mongholic acid, etc.) and place them on the stage of the infrared spectrometer. Detect the infrared spectral intensity of the samples and record the data.

[0046] Step 4: Take an appropriate amount of quality marker (such as cinnamon: cinnamic acid; wild chrysanthemum: mongholic acid, etc.), prepare an aqueous solution (standard solution) of a certain concentration, and take about 400 μL into a 1 mm quartz cuvette; place the cuvette on the stage of the Raman spectrometer, detect the Raman spectral intensity of the sample, and record the data.

[0047] Step 4: Assign peaks to the Raman and infrared spectra of the quality markers and correlate them with the Raman spectra of their aqueous solutions to ensure that the characteristic peaks of the quality markers are still retained in the aqueous solutions, providing a basis for classification.

[0048] The results of the cinnamon Raman feature correlation diagram are as follows: Figure 2 As shown. By Figure 2 a can be seen to be 3000 cm -1 The nearby peak corresponds to the stretching vibration of the CH group on the benzene ring, 1500-1600 cm⁻¹ -1 The region corresponds to the skeletal vibration of the benzene ring and the stretching vibration of the C=C double bond, 1700 cm⁻¹. -1 The left and right sides correspond to the C=O stretching vibrations of the carboxyl groups, which are typical functional group vibrations of cinnamic acid. Figure 2 b is the Raman spectrum of the cinnamic acid standard. Figure 2 a and Figure 2 b is complementary, and Raman spectroscopy is more sensitive to nonpolar bonds such as C=C double bonds and C=C single bonds, especially in the 1000-1600 cm⁻¹ range. -1 The multiple peaks correspond to benzene ring vibrations and alkene bond vibrations, which are characteristic Raman peaks of cinnamic acid. Figure 2 b and Figure 2 In comparison to c, Figure 2 The c-peak is broader and its intensity distribution is more dispersed. This may be because when cinnamic acid dissolves in water, the carboxyl group ionizes into carboxylate ions (-COOH → -COO). - The formation of hydrogen bonds between cinnamic acid molecules and water molecules alters the vibrational environment of the molecules, leading to changes in the peak position and shape of characteristic peaks. However, in aqueous cinnamic acid solutions, the peak position and shape remain constant within the 1000-1600 cm⁻¹ range. -1 The multiple sets of peaks that appear between them, and the positions of the peaks still correspond to the Raman spectrum peaks of the cinnamic acid standard, therefore Figure 2 c (Raman spectrum of aqueous solution) retains the core Raman characteristic peak of cinnamic acid, indicating that Raman spectroscopy can be used for the qualitative identification of cinnamic acid in aqueous solution. Figure 2 The overall trend of the d-spectrum, the position of the main peak, and the comparison with the cinnamic acid standard ( Figure 2c) The core characteristics overlap, indicating that the cinnamon powder aqueous solution contains cinnamic acid. Figure 2 c and Figure 2 d can verify whether the cinnamon powder contains the target component (cinnamic acid), which is also the basis for Raman spectroscopy to classify.

[0049] The results of the Raman characteristic correlation diagram of wild chrysanthemum are as follows: Figure 3 As shown, Figure 3 a is the infrared spectrum of scutellarin, which shows that at 3500 cm⁻¹... -1 The nearby peaks correspond to the stretching vibration of the hydroxyl group (-OH); 1500-1600 cm⁻¹ -1 The region corresponds to the skeletal vibration of the benzene ring and the stretching vibration of the double bond (C=C); 1500~1000 cm -1 The multiple strong peaks in the region (peaks 4-14) correspond to the skeletal vibrations (C=C, CO, etc.) of the flavonoid nucleus and the characteristic vibrations of the glycosyl group (such as COC, CO stretching vibrations). These peaks are the characteristic infrared absorption of styracin and can be used to identify its structure. Figure 3 b is the Raman characteristic peak chromatogram of the strychnine standard, where 1600~2000 cm⁻¹ -1 The strong peaks in the region (such as peak 8) correspond to the C=C stretching vibration of the flavonoid nucleus, which is a typical Raman signal of flavonoids, 1400~1600 cm⁻¹. -1 The peaks (peaks 4-7) correspond to the coupled vibrations of glycosyl groups and flavonoid rings, and can be used as Raman "fingerprints" of scutellarin for qualitative identification. Figure 3 c is the Raman spectrum of the aqueous solution of styraxanthoxylin and cylindrica, derived from... Figure 3 c and Figure 3 As can be seen from d, the position and intensity of the peaks changed after scutellarin was prepared into an aqueous solution. This change may be due to the formation of hydrogen bonds between the hydroxyl groups of scutellarin and water molecules in the aqueous solution. In addition, some new peaks appeared in the Raman spectrum of the aqueous solution of scutellarin. Figure 3 Peaks 4-7 in c indicate the state of strychnine in aqueous solution. Strychnine aqueous solution has a concentration of 1400-1600 cm⁻¹. -1 The multiple sets of peaks that appear between them, and the positions of the peaks still correspond to the Raman spectrum peaks of the strychnine standard, therefore Figure 3 c retains the core Raman characteristic peak of scutellarin, indicating that Raman spectroscopy can be used for the qualitative identification of scutellarin in aqueous solution. Figure 3 The core peak position of the residual spectrum in d and Figure 3 The c-characteristic peaks clearly overlap, indicating that chrysanthemum powder contains benzoin-like components; however, Figure 3 The spectral peaks of d are more dense and the fluctuations are more complex because the powder is a mixture of multiple components (in addition to styraxin, it also contains polysaccharides, flavonoids, volatile oils and other substances). The superposition of the Raman signals of these components makes the spectrum more complex. Figure 3 c and Figure 3 d can verify whether the wild chrysanthemum powder aqueous solution contains the target component (monotropine), which is also the basis for Raman spectroscopy for classification.

[0050] Steps of Example 1 and Example 2: 1. The wild chrysanthemum and cinnamon were crushed separately by breaking down the cell walls, and then the large particles were removed by sieving through a 140-mesh pharmacopoeia sieve. Finally, the wild chrysanthemum powder and cinnamon powder that passed through the 140-mesh pharmacopoeia sieve were obtained.

[0051] 2. Take 0.015 g of wild chrysanthemum powder and 0.015 g of cinnamon powder respectively, add about 15 mL of purified water, mix well, and sonicate for 30 min to allow the wild chrysanthemum powder and cinnamon powder to fully dissolve and diffuse in the aqueous solution.

[0052] 3. After sonicating the sample, let it stand for about 30 minutes, then take an appropriate amount of supernatant into a clean glass test tube. Illuminate the glass test tube with a laser pointer in a dark environment to ensure that the supernatant exhibits the Tyndall effect.

[0053] 4. Take approximately 400 μL of the supernatant of the sample that produces the Tyndall effect and place it in a 1 mm quartz cuvette.

[0054] 5. Place the cuvette on the stage of the Raman spectrometer and measure the Raman spectral intensity of the sample solution and the standard solution respectively, and record the data.

[0055] 6. Perform residual processing and residual smoothing on the obtained Raman data.

[0056] Depend on Figures 4-5 It can be seen that the original Raman spectrum of the medicinal powder solution is at 1300 nm. -1 There are obvious strong peaks nearby, but the characteristic absorption peaks of the indicative components cannot be obtained; the residuals exhibit significant fluctuations (especially in the low wavenumber region 0-500 cm⁻¹). -1 This indicates that the original spectrum, in addition to the main features, also contains noise, baseline drift residues, or differences between samples. The smoothed residual plot is a noise reduction process for the residuals. By comparing the residual plot and the smoothed residual plot, it can be seen that smoothing effectively suppresses random noise in the residuals and preserves a more regular fluctuation trend. The data processing step from original spectrum → residual → residual smoothing aims to extract characteristic peaks and feature information, thereby achieving the purpose of classification.

[0057] Example 1: Taking wild chrysanthemum powder from Sichuan (SC) and Hebei (HB) as examples, mongholic acid was used as a quality marker to verify the applicability of the method of the present invention.

[0058] 1. The wild chrysanthemums from Sichuan and Hebei were crushed and pulverized, and then the large particles were removed by sieving with a 140-mesh pharmacopoeia sieve. Finally, wild chrysanthemum powder that passed through the 140-mesh pharmacopoeia sieve was obtained.

[0059] 2. Take 0.015g of medicinal powder from different origins, add about 15 mL of purified water, mix well, and sonicate for 30 min to allow the medicinal powder to fully dissolve and diffuse in the aqueous solution.

[0060] 3. After sonicating the sample, let it stand for about 30 minutes, then take an appropriate amount of supernatant into a clean glass test tube. Illuminate the glass test tube with a laser pointer in a dark environment to ensure that the supernatant exhibits the Tyndall effect.

[0061] 4. Take approximately 400 μL of the standard solution (Mongolian glycoside solution, concentration 1%) and the supernatant of the sample that produces the Tyndall effect into a 1 mm quartz cuvette.

[0062] 5. Place the cuvette on the stage of the Raman spectrometer and measure the Raman spectral intensity of the sample solution and the standard solution respectively, and record the data.

[0063] 6. PLS-DA and SVM were used to process the residuals and smooth the residuals of the obtained Raman data, respectively, to establish a classification model and realize the classification and identification of Chinese medicinal materials.

[0064] The medicinal powder prepared in step 1 was directly subjected to Raman spectroscopy for detection, such as... Figure 6 As shown, the signal strength decreases with increasing Raman shift (wavenumber). However, from... Figure 6 No characteristic peaks for mongholic acid were observed in the sample. Therefore, when Raman spectroscopy was performed directly on the raw medicinal powder, there was little spectral characteristic information, which was not conducive to the development of classification methods.

[0065] Table 1 shows the classification results of different classification models in this embodiment.

[0066] Table 1. Classification results of PLS-DA and SVM models for wild chrysanthemum powder solution.

[0067] Note: Residual smoothed data is obtained by performing SG smoothing on residual data, where the SG smoothing parameter is a second-order polynomial and the window is 21. When the response value of a sample is higher than the threshold of a certain category, it is determined to belong to that category; if the response value is higher than the threshold of multiple categories or lower than the threshold of all categories, it is determined to be "undiscriminable".

[0068] Table 1 shows that SVM outperforms PLS-DA overall. For all data types (original data, residual data, and residual-smoothed data), SVM achieves a 1.00 accuracy, specificity, and sensitivity on the calibration set, while its error rate and undiscriminable rate are all 0.00. This indicates that SVM performs better on the training data, with no prediction errors or undiscriminable cases. In contrast, PLS-DA's calibration set exhibits slight errors (e.g., an accuracy of 0.99 on the original data) and a small number of undiscriminable samples (e.g., an undiscriminable rate of 0.03 on the original data). Furthermore, SVM performs better on the validation set, especially with residual-smoothed data. SVM achieves a 0.98 accuracy and a mere 0.02 error rate on the validation set, while PLS-DA's validation set accuracy is 0.98, but its calibration set accuracy is lower. Regardless of whether it's PLS-DA or SVM, the validation set performance is improved after residual smoothing, indicating that smoothing residual data reduces data noise and enhances the model's ability to identify samples. This is the reason for choosing SG smoothing.

[0069] To more clearly demonstrate the predictive performance of the classification model, a graph of the model's computer response was plotted as follows: Figures 7-12 As shown. From Figures 7-12 The SVM model can be found to outperform PLS-DA in classification. Figure 7 , 8 Although the different categories of samples represented by circles and triangles in 9 showed some clustering, the response values ​​of some samples were close to the classification threshold, indicating that the category boundaries were not clear enough and there was a risk of misclassification. Figure 10 , 11 The samples in categories 12 (circular and triangular) clustered more closely, with almost no overlap between the two categories. All sample response values ​​were far from the classification threshold, indicating clearer category discrimination and more reliable classification results. Therefore, the SVM model outperformed PLS-DA in terms of accuracy, generalization ability, and stability in the classification of wild chrysanthemums, and can more reliably distinguish between different categories of wild chrysanthemum samples.

[0070] To further evaluate the performance of the classification model, receptor operating curves (ROC) and confusion matrices were used to comprehensively and accurately assess its performance and determine whether the model could reliably distinguish samples. Figures 13-18 As shown. The confusion matrix and ROC curve of the PLS-DA model are shown below. Figure 13 , 14As shown in Figure 15, the confusion matrix reveals some misclassification in the PLS-DA model on the calibration set. For example, the calibration set accuracy for the original wild chrysanthemum sample data is 93.12%, while the accuracy for the residual and smoothed residual data is only 90%. The validation set performs better, with a 100% accuracy for the residual and smoothed residual data, indicating good model adaptability to new samples. However, the numerous misclassifications on the calibration set suggest underfitting during training. The ROC curves show that the AUC values ​​for both class 1 and class 2 are close to 1, indicating that PLS-DA has a strong ability to distinguish between positive and negative classes. The confusion matrix and ROC curve of the SVM model are shown below. Figure 16 , 17 As shown in Figure 18, the accuracy on the calibration set is 100%, indicating a perfect fit to the training data; the minimum accuracy on the validation set is 95%, demonstrating stable generalization ability. The ROC curves show that the AUC values ​​for both class 1 and class 2 are close to 1, indicating that SVM has extremely strong discriminative power against different classes. Therefore, it can be concluded that SVM is more stable and reliable, and more suitable for this wild chrysanthemum classification task.

[0071] The methodological results of this embodiment are as follows: The occurrence of RSD > 5% in the low wavenumber range is mainly attributed to noise interference from the instrument itself. In the low wavenumber region, the measured signal is inherently weak, amplifying the impact of noise on the measurement results. This leads to significant fluctuations in repeated measurements, ultimately manifesting as a higher RSD. Methodological results are as follows: Figure 19 As shown.

[0072] Example 2: Taking cinnamon powder from Guangxi (GX) and Fujian (FJ) as examples, cinnamic acid was used as an indicator component to verify the applicability of this method.

[0073] 1. Cinnamon from Guangxi and Fujian provinces was crushed and pulverized, and then large particles were removed by sieving through a 140-mesh pharmacopoeia sieve. Finally, cinnamon powder that passed through the 140-mesh pharmacopoeia sieve was obtained.

[0074] 2. Take 0.015g of medicinal powder from different origins, add about 15 mL of purified water, mix well, and sonicate for 30 min to allow the medicinal powder to fully dissolve and diffuse in the aqueous solution.

[0075] 3. After sonicating the sample, let it stand for about 30 minutes, then take an appropriate amount of supernatant into a clean glass test tube. Illuminate the glass test tube with a laser pointer in a dark environment to ensure that the supernatant exhibits the Tyndall effect.

[0076] 4. Take approximately 400 μL of the standard solution (cinnamic acid solution, concentration 1%) and the supernatant of the sample that produces the Tyndall effect above into 1 mm quartz cuvettes.

[0077] 5. Place the cuvette on the stage of the Raman spectrometer and measure the Raman spectral intensity of the sample solution and the standard solution respectively, and record the data.

[0078] 6. PLS-DA and SVM were used to process the residuals and smooth the residuals of the obtained Raman data, respectively, to establish a classification model and realize the classification and identification of Chinese medicinal materials.

[0079] The medicinal powder prepared in step 1 was directly subjected to Raman spectroscopy for detection, such as... Figure 20 As shown, the signal strength decreases with increasing Raman shift (wavenumber). However, from... Figure 20 No characteristic peaks for cinnamic acid were observed in a and b in the spectrum, therefore, directly using the Raman spectra of the original medicinal powder is not conducive to the study of classification methods.

[0080] Table 2 shows the classification results of different classification models in this embodiment.

[0081] Table 2. Classification results of cinnamon powder solution by PLS-DA and SVM classification models.

[0082] Note: Residual smoothing data is obtained by performing SG smoothing on the residual data, where the second-order polynomial has a window size of 21. When the response value of a sample is higher than the threshold of a certain category, it is determined to belong to that category; if the response value is higher than the thresholds of multiple categories or lower than the thresholds of all categories, it is determined to be "undiscriminable".

[0083] Table 2 shows that for both PLS-DA and SVM, residual smoothing significantly improves model performance. In the PLS-DA model, with both raw and residual data, the accuracy on the calibration and validation sets is only 0.83–0.88. After residual smoothing, the accuracy, specificity, and sensitivity on both sets reach 1.00, while the error rate and undiscriminable rate are both 0.00. In the SVM model, with both raw and residual data, the validation set accuracy is only 0.78–0.85. After residual smoothing, all metrics on both sets reach 1.00, with no errors and no undiscriminable rates. Furthermore, SVM exhibits more stable training and fitting capabilities. SVM achieves a calibration set metric of 1.00 across all data types, indicating a stronger learning ability for training samples. In contrast, PLS-DA only achieves perfect calibration set performance with residual smoothing; its performance with raw and residual data is generally average, suggesting a higher dependence on data quality. Therefore, smoothing the processed residual data combined with the SVM classification model yields the best classification results. It can also be seen that noise fluctuations in Raman spectra have a significant impact. Smoothing preprocessing can suppress or reduce the influence of noise and improve the modeling performance of Raman spectra.

[0084] To more clearly demonstrate the predictive performance of the classification model, a graph of the model's computer response was plotted as follows: Figures 21-26 As shown. From Figures 21-26 The SVM model can be found to outperform PLS-DA in classification. Figure 21 , 22 Figure 23 shows the response diagram corresponding to PLS-DA. The circles and triangles represent different categories of cinnamon samples with poor clustering and obvious category overlap. The response values ​​of some samples are close to the classification threshold, or even cross the threshold distribution, indicating that the distinction boundary between different categories of cinnamon by PLS-DA is not clear enough and there is a risk of misjudgment. Figure 24 , 25 Figure 26 shows the response map corresponding to SVM. The circular and triangular samples exhibit strong clustering with no overlap between the two classes. All sample response values ​​are far from the classification threshold, indicating that SVM has a very clear distinction between different categories of cinnamon, resulting in more stable and reliable classification results. Therefore, considering the distribution characteristics of the response map, SVM's performance is significantly better than PLS-DA.

[0085] To further evaluate the performance of the classification model, receptor operating curves (ROC) and confusion matrices were used to comprehensively and accurately assess the model's performance and determine whether the model could reliably distinguish samples. Figures 27-32 The confusion matrix and ROC curve of the PLS-DA model, as shown below. Figure 27 , 28 As shown in Figure 29, the confusion matrix reveals that the PLS-DA model exhibits some misclassification on the calibration set. The calibration / validation set accuracy of PLS-DA is only 82.50%–87.50%, indicating significant misclassification. Furthermore, the AUC of the ROC curve is below 1, suggesting weak classification ability. After smoothing the residual data, the accuracy of both the calibration and validation sets of PLS-DA reaches 100%, and the AUC of the ROC curve is 1.000, demonstrating better classification ability. The confusion matrix and ROC curve of the SVM model are shown below. Figure 30 , 31 As shown in Figure 32, the SVM achieves a consistent 100% accuracy on the calibration set, demonstrating strong fitting ability. However, its validation set accuracy is only 77.50%–85.00%, indicating moderate generalization ability. After residual smoothing, the SVM achieves 100% accuracy on both the calibration and validation sets, with an AUC of 1.00 on the ROC curve. This demonstrates that residual smoothing reduces data noise, significantly improving the generalization ability of the SVM. Therefore, residual smoothing combined with SVM is a better choice for cinnamon classification.

[0086] The methodological results of this embodiment are as follows: Figure 33 As shown.

[0087] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for identifying the origin of traditional Chinese medicinal materials based on Raman spectroscopy, characterized in that, Includes the following steps: The Chinese medicinal materials to be tested were made into powder, and the powder was added to water, ultrasonically treated and allowed to stand. The supernatant with Tyndall effect was obtained as the sample to be tested. The sample to be tested was detected using Raman spectroscopy to obtain raw Raman spectral data; The raw Raman spectral data were subjected to residual processing and residual smoothing in sequence. The origin of the Chinese medicinal materials to be tested was identified based on the results of the residual smoothing.

2. The identification method as described in claim 1, characterized in that, The process of establishing a model for identifying the origin of Chinese medicinal materials includes the following steps: At least two kinds of Chinese medicinal materials from different origins were made into Chinese medicinal material powders, and the different Chinese medicinal material powders were added to water, ultrasonically treated and allowed to stand, and the supernatant with Tyndall effect was obtained as the sample solution. The sample solution was analyzed using Raman spectroscopy to obtain raw Raman spectral data; The raw Raman spectral data were subjected to residual processing and residual smoothing in sequence. Based on the results of the residual smoothing, a model for identifying the origin of Chinese medicinal materials was established.

3. The identification method as described in claim 2, characterized in that, The process of establishing the Chinese medicinal material origin identification model also includes using Raman spectroscopy to detect the sample solution with a standard solution, which serves as a quality marker for the Chinese medicinal material.

4. The identification method as described in claim 1, characterized in that, When preparing the test sample, the mass of Chinese medicinal materials added per milliliter of water is 0.09~0.11 mg.

5. The identification method as described in claim 1, characterized in that, The ultrasonic treatment time is 25~35 min.

6. The identification method as described in claim 1, characterized in that, The time for standing after ultrasonic treatment is 25-35 minutes.

7. The identification method as described in claim 1, characterized in that, The model used for residual processing and residual smoothing is a partial least squares discriminant or support vector machine.

8. The identification method as described in claim 1, characterized in that, SG smoothing for residual smoothing.

9. The identification method as described in claim 1, characterized in that, The medicinal materials mentioned are cinnamon and / or wild chrysanthemum.

10. The application of the identification method according to any one of claims 1 to 9 in the quality control of Chinese medicinal materials.

Images