Authentication method and authentication device for single traditional Chinese medicine concentrated particles

Abstract

The invention relates to an authentication method and an authentication device for single traditional Chinese medicine concentrated particles. The authentication method comprises the steps that samples recognized as various types of single traditional Chinese medicine concentrated particles in a training sample set are obtained, and the various types of single traditional Chinese medicine concentrated particles are recognized through identity verification of the single traditional Chinese medicine concentrated particles and corresponding types of traditional Chinese medicines; measuring the reference FTIR-ATR spectrum of each sample; and training a first learning model based on a support vector machine by using each reference FTIR-ATR spectrum so as to obtain a mapping relationship between each reference FTIR-ATR spectrum and each type. And during authentication, measuring the FTIR-ATR spectrum of the to-be-authenticated single traditional Chinese medicine concentrated particles, inputting the FTIR-ATR spectrum into the trained first learning model, and judging that the identification type of the to-be-authenticated single traditional Chinese medicine concentrated particles passes authentication when the output prediction type is matched with the identification type of the to-be-authenticated single traditional Chinese medicine concentrated particles. The authentication method provided by the invention can be used for quickly, simply, accurately and reliably inspecting and authenticating the component types of the single traditional Chinese medicine concentrated granules.

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application is based on and claims priority from Hong Kong short-term patent application number 32023081152.5 filed on October 20, 2023, which is incorporated herein by reference in its entirety. Technical Field

[0003] The present application relates to the field of quality management and certification of traditional Chinese medicine, and more specifically, to an authentication method and an authentication device for concentrated granules of a single traditional Chinese medicine. Background Art

[0004] Concentrated Chinese medicine granule (CCMG) is made by pharmaceutical companies from raw Chinese medicines (also known as Chinese herbal medicine slices) through a series of processes such as extraction, concentration, drying and granulation to extract the active ingredients. Patients do not need to boil the medicine, but only need to dissolve the processed granules in hot water to obtain the same efficacy as the corresponding type of Chinese medicine. Therefore, it is easier to take and is also convenient for Chinese medicine doctors and pharmacists to use when prescribing or preparing other medicines.

[0005] Traditionally, Chinese herbal medicines that have not been processed into granules are mainly identified based on their morphological characteristics. However, since the original unique morphology of Chinese herbal medicines will disappear in most cases during the processing of concentrated Chinese herbal medicine granules, the traditional method of identification based on morphology cannot be applied to concentrated Chinese herbal medicine granules. Even professionals such as Chinese medicine doctors or Chinese medicine pharmacists may not be able to distinguish between various types of concentrated Chinese herbal medicine granules. Therefore, when Chinese medicine doctors prescribe concentrated Chinese herbal medicine granules, or when pharmacists prepare concentrated Chinese herbal medicine granules, they can only rely on the ingredients indicated on the drug packaging label. Even if they have doubts about it, there is no means to verify it. In addition, there is a lack of rapid authentication methods in the production process of concentrated granules to control the key links in the manufacturing process of concentrated granules so as to take effective corrective measures. However, any poor management or accidental cross-contamination of concentrated Chinese herbal medicine granules may lead to very serious consequences. In addition, in the existing technology, chemical analysis methods such as liquid chromatography can be used to test the ingredients of traditional Chinese medicine. However, liquid chromatography analysis usually requires expensive machine installation and operation costs, complex sample preparation procedures, demanding elution and washing time, and requires to be performed by well-trained operators. Therefore, it is not convenient enough, which limits its application in rapid authentication in field environments.

[0006] Therefore, there is currently no rapid, simple and reliable testing and authentication method for concentrated Chinese medicine granules that can be provided to Chinese medicine physicians, Chinese medicine pharmacists and pharmaceutical factory technicians. Summary of the invention

[0007] The present application is provided to solve the above problems existing in the prior art. In view of the problem that concentrated Chinese medicine granules cannot be identified by morphological characteristics and the chemical analysis method is too complicated and time-consuming, the present application intends to provide a method and an authentication device that can quickly, simply, accurately and reliably inspect and authenticate the types of ingredients of single concentrated Chinese medicine granules.

[0008] According to the first scheme of the present application, a method for authenticating single-flavor Chinese medicine concentrated granules is provided, wherein the single-flavor Chinese medicine concentrated granules include excipients in addition to the ingredients of the single Chinese medicinal material, and the authentication method comprises: obtaining samples identified as various types of single-flavor Chinese medicine concentrated granules in a training sample set, wherein the various types of single-flavor Chinese medicine concentrated granules are identified by verifying their identity with the corresponding types of Chinese medicinal materials; measuring the baseline FTIR-ATR spectrum of each sample of the various types of single-flavor Chinese medicine concentrated granules; and using the baseline FTIR-ATR spectrum of each sample, training a first learning model constructed based on a support vector machine, wherein the first learning model is configured to learn the mapping relationship between the baseline FTIR-ATR spectrum of each sample and the various types of single-flavor Chinese medicine concentrated granules. The authentication method also includes: when the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated is included in the types in the training sample set, measuring the FTIR-ATR spectrum of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated; and inputting the FTIR-ATR spectrum of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated into a trained first learning model, and when the predicted type output by the trained first learning model matches the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated, determining that the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated has passed the authentication.

[0009] According to the second scheme of the present application, a device for authenticating single-flavor Chinese medicine concentrated granules is provided, wherein the single-flavor Chinese medicine concentrated granules include excipients in addition to the ingredients of the single-flavor Chinese medicine, and the authentication device includes an interface and a processor. The interface is configured to receive the baseline FTIR-ATR spectra of samples identified as various types of single-flavor Chinese medicine concentrated granules in the training sample set, and the FTIR-ATR spectra of samples of the single-flavor Chinese medicine concentrated granules to be authenticated; the processor is configured to execute the steps of the authentication method of the single-flavor Chinese medicine concentrated granules according to the various embodiments of the present application.

[0010] According to the authentication method and authentication device of the single-flavor Chinese medicine concentrated granules of each embodiment of the present application, the first learning model is trained in advance by using the FTIR-ATR spectra of the samples of the single-flavor Chinese medicine concentrated granules of each type in the training sample set, so that the trained first learning model can judge whether the single-flavor Chinese medicine concentrated granules to be authenticated belong to one of the various types in the training sample set according to the FTIR-ATR spectrum of the single-flavor Chinese medicine concentrated granules to be authenticated, thereby realizing the authentication of the identification type of the single-flavor Chinese medicine concentrated granules to be authenticated. According to the authentication method of the present application, without the need for complex and time-consuming chemical analysis and inspection, the type of the single-flavor Chinese medicine concentrated granules can be quickly and accurately authenticated by only using the trained learning model, providing a reliable basis for Chinese medicine doctors to prescribe or pharmacists to dispense medicines, and can avoid adverse effects or even more serious consequences on medication due to type deviation, impure ingredients, poor storage, pollution, etc. of the single-flavor Chinese medicine concentrated granules, thereby improving the safety of the application of single-flavor Chinese medicine concentrated granules.

[0011] The above description is only an overview of the technical solution of the present application. In order to more clearly understand the technical means of the present application, it can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the present application more obvious and easy to understand, the specific implementation methods of the present application are listed below.

[0012] It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed. BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In the drawings, which are not necessarily drawn to scale, the same reference numerals may describe similar components in different views. The drawings generally illustrate various embodiments by way of example and not limitation, and together with the description and claims, serve to illustrate the disclosed embodiments. When appropriate, the same reference numerals are used throughout the drawings to refer to the same or similar parts. Such embodiments are illustrative and are not intended to be exhaustive or exclusive embodiments of the present apparatus or method.

[0014] Figure 1 A schematic flow chart of a method for authenticating concentrated single-flavor Chinese medicine granules according to an embodiment of the present application is shown.

[0015] Figure 2 A schematic diagram of FTIR-ATR spectra of various samples of the same type of single-flavor Chinese medicine concentrated granules after correction according to an embodiment of the present application is shown.

[0016] FIG. 3( a ) is a schematic diagram showing the benchmark FTIR-ATR spectra of various types of single-flavor Chinese medicine concentrated granules before spectrum segment extraction according to an embodiment of the present application.

[0017] FIG3( b ) is a schematic diagram showing PCA scores of the benchmark FTIR-ATR spectra of various types of single-flavor Chinese medicine concentrated granules before and after spectral segment interception according to an embodiment of the present application.

[0018] FIG3( c ) is a schematic diagram showing the benchmark FTIR-ATR spectra of various types of single-flavor Chinese medicine concentrated granules after spectral segment truncation according to an embodiment of the present application.

[0019] Figure 4 A schematic flow chart of a method for authenticating the identified type and identified batch of a single-flavor Chinese medicine concentrated granule to be authenticated according to an embodiment of the present application is shown.

[0020] Figure 5 An example of prediction results of the first learning model according to an embodiment of the present application for single-flavor Chinese medicine concentrated granules under different storage conditions is shown.

[0021] Figure 6 A block diagram showing some components of an authentication device for concentrated single-flavor Chinese medicine granules according to an embodiment of the present application is shown.

[0022] Figure 7 A schematic diagram showing the performance of the first learning model on the training set according to an embodiment of the present application is shown.

[0023] FIG8( a ) is a schematic diagram showing the performance of the first learning model on the training set according to an embodiment of the present application.

[0024] FIG8( b ) shows a schematic diagram of the performance of the first learning model on the test set according to an embodiment of the present application.

[0025] FIG8( c ) shows a schematic diagram of the performance of the first learning model on the prediction set according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the technical solution of the present application, the present application is described in detail below in conjunction with the accompanying drawings and specific implementation methods. The embodiments of the present application are further described in detail below in conjunction with the accompanying drawings and specific implementation methods, but are not intended to limit the present application.

[0027] The words "first", "second" and similar words used in this application do not indicate any order, quantity or importance, but are only used to distinguish. Words such as "include" or "comprise" and similar words mean that the elements before the word include the elements listed after the word, and do not exclude the possibility of also covering other elements. The execution order of the various steps in the method described in this application in conjunction with the accompanying drawings is not intended to be limiting. As long as it does not affect the logical relationship between the various steps, several steps can be integrated into a single step, a single step can be decomposed into multiple steps, and the execution order of the various steps can be changed according to specific needs.

[0028] It should also be understood that the term "and / or" in this application is only a description of the association relationship of associated objects, indicating that three relationships may exist. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this application generally indicates that the associated objects before and after are in an "or" relationship.

[0029] According to an embodiment of the present application, a method for authenticating a single-flavor Chinese medicine concentrated granule is provided. The so-called single-flavor Chinese medicine concentrated granule means that the Chinese medicine concentrated granule is made by extracting a single-flavor Chinese medicine, and only contains the drug components of the corresponding single-flavor Chinese medicine. Taking Codonopsis pilosula Chinese medicine concentrated granules as an example, it can be made from Codonopsis pilosula Chinese medicine (Chinese medicine decoction pieces), and Poria cocos Chinese medicine concentrated granules are made from Poria cocos Chinese medicine. The specific method of extracting the effective ingredients of Chinese medicines may vary depending on the different Chinese medicines, and this application does not limit this.

[0030] In addition to the corresponding Chinese medicinal material ingredients, excipients for particle forming may also be included. For most Chinese medicine concentrated granules, the commonly used excipient is dextrin. In other cases, other excipients may also be included, which are not specifically limited in this application, but their role is only to form the shape, and do not affect the efficacy of the single-flavor Chinese medicine concentrated granules.

[0031] Figure 1 The flowchart of the authentication method of the concentrated single-flavor Chinese medicine granules according to the embodiment of the present application is shown. Figure 1 As shown, in step 101, samples identified as various types of single-flavor Chinese medicine concentrated granules in the training sample set can be obtained, wherein various types of single-flavor Chinese medicine concentrated granules are identified by identity verification with corresponding types of Chinese medicinal materials.

[0032] In some embodiments, the training sample set can be constructed in advance with reference to the types of single-flavor Chinese medicine concentrated granules that are intended to be authenticated in the future. As an example only, commonly used single-flavor Chinese medicine concentrated granules can be selected from prescriptions widely prescribed in public hospitals and private clinics, and different types of Chinese medicine concentrated granules derived from plants and animals can also be included in the samples, as well as samples with different therapeutic function groups such as tonifying deficiency, clearing heat, calming the liver, and removing wind and cold as much as possible, etc. In addition, single-flavor Chinese medicine concentrated granule samples of types that are easily confused in terms of properties or efficacy corresponding to each type in the training sample set can also be included in the training sample set. In other embodiments, Chinese medicine concentrated granule samples with drug toxicity that need to be specially identified and that are easy to cross-contaminate with other Chinese medicine concentrated granules can also be included in the training sample set. As an example only, for example, when Sichuan Fritillaria is included in the training sample set, Zhejiang Fritillaria can be included in the training sample set together, and when Poria is included in the training sample set, Poria and Smilax glabra can be included in the training sample set together, and so on. In this way, the trained first learning model can have the ability to distinguish the types of single-flavor Chinese medicine concentrated granules with similar properties or effects, or with toxicity, so as to better serve Chinese medicine doctors or pharmacists, and further improve the safety of single-flavor Chinese medicine concentrated granules when used. In an embodiment of the present application, the training sample set will include the following 11 types of single-flavor Chinese medicine concentrated granules: Astragali Radix, Codonopsis Radix, FritillariaeCirrhosae Bulbus, Fritillariae Thunbergii Bulbus, Lonicerae Japonicae, Panacis Quinquefolii Radix, Poria, Poria cumRadix Pini, Pseudostellariae Radix, Ginseng Radix et Rhizoma, and Smilacis Glabrae Rhizoma.

[0033] The single-flavor TCM concentrated granules in the training sample set can be purchased from licensed pharmaceutical manufacturers, while the corresponding types of TCM materials can be purchased from qualified TCM pharmacies. TCM experts can further identify the types of TCM materials based on morphological features such as color, texture, pattern and smell, and then verify the identity of each type of single-flavor TCM concentrated granules with the corresponding types of TCM materials, thereby ensuring that the TCM concentrated granules in the training sample set are identified as the correct types.

[0034] As an example only, ultra performance liquid chromatography (UPLC) can be used to verify the identity of each type of single-flavor Chinese medicine concentrated granules with the corresponding type of Chinese medicinal materials under the condition of considering excipients. This is a more advanced chromatographic detection method improved from traditional high performance liquid chromatography (HPLC) and can improve the sensitivity, specificity and accuracy of drug and biomedical analysis.

[0035] In the specific implementation, the corresponding type of Chinese medicinal material solution is used as the standard solution, and the single Chinese medicine concentrated granule solution is used as the test solution. Both will be prepared according to the chromatographic analysis procedures recommended by relevant standards such as the Hong Kong Chinese Medicine Standards or the Pharmacopoeia of the People's Republic of China, among which methanol is used as the solvent to dissolve the two precisely weighed compounds.

[0036] As an example only, for identity verification, ultra-high performance liquid chromatography can be coupled with a triple quadrupole mass spectrometer QQQ-MS / MS (e.g., Agilent's UPLC-QQQ-MS / MS instrument) and run on a Waters ACQUITY UPLC C18 column (2.1x100mm, 1.7μm) at 30°C and a flow rate of 350μl / min. The mobile phase contains a buffer solution (10mM ammonium acetate solution containing 0.1% acetic acid) (A) and acetonitrile (B). The gradient elution rate is as follows: 5-20% B at 0-20 minutes, 20-30% B at 20-30 minutes, 30-50% B at 30-40 minutes, and 50% B at 40-50 minutes, with an injection volume of 1μl. In addition, a guard column with the same stationary phase will be used to improve the durability of the column. In other embodiments, instruments and equipment may also be used to verify the identity of single-flavor Chinese medicine concentrated granules with corresponding types of Chinese medicinal materials under different experimental conditions. This application does not impose any restrictions on this, as long as the accuracy of the types identified by the samples in the training sample set can be ensured.

[0037] In other embodiments, the concentrations of the corresponding types of Chinese medicinal material solutions and the concentrations of the single Chinese medicinal concentrated granule test solution can be adjusted separately, and the identity calibration curves of the two solutions at different concentrations can be used to determine the identity of the single Chinese medicinal concentrated granules and the corresponding Chinese medicinal materials.

[0038] In addition, excipients are often present in the production process of concentrated Chinese medicine granules. Excipients are inactive ingredients in drugs that facilitate drug administration and improve drug stability. In order to verify the impact of the presence of excipients on the sensitivity and specificity of identity certification, this application also conducted experiments on the effectiveness of identity verification between Chinese medicinal materials and single-flavor concentrated Chinese medicine granules in the presence of excipients such as dextrin.

[0039] Taking dextrin as an excipient as an example, the above experiment can be carried out in the following steps: First, the Chinese medicinal material extract is mixed with dextrin of different concentrations, and the Chinese medicinal material solution after mixing with dextrin is used to prepare the Chinese medicinal material reference chromatogram. Specifically, each Chinese medicinal material can be ground into powder, 20g of powder is refluxed with 200ml of distilled water for 1 hour, and after drying the filtrate, the Chinese medicinal material extract is collected and mixed with dextrin of different proportions, such as 0%, 5%, 10%, 20%, 40%, 60%, 80% and 100% (w / w). Then, by drawing a relationship diagram between the FTIR spectrum absorbance and the corresponding concentration of the Chinese medicinal material extract, the calibration curve of the Chinese medicinal material standard is determined, and the chromatographic similarity between the corresponding Chinese medicine concentrated granules and the Chinese medicinal material is compared by calculating the correlation coefficient, thereby completing the identity verification.

[0040] In step 102, the baseline FTIR-ATR spectra of each sample of each type of single-flavor Chinese medicine concentrated granules may be measured.

[0041] FTIR-ATR (Fourier Transform Infrared-Attenuated Total Reflectance Spectroscopy) technology collects the attenuated total reflection signal of infrared light on the sample surface and performs Fourier transform and other processing on the signal, thereby achieving the determination, authentication and analysis of the sample content. FTIR-ATR technology can be used to detect samples in different physical states such as solid, semi-solid and liquid. It has no requirements on thickness, etc., has a wide range of applications, and has a fast detection speed, simple operation and reliable results.

[0042] The inventors of the present application have found through a large number of experiments that the FTIR-ATR spectra of different types of single-herb concentrated granules of traditional Chinese medicine are also different. Therefore, there is a potential to use the FTIR-ATR spectra of single-herb concentrated granules of traditional Chinese medicine to distinguish their types.

[0043] Specifically, the FTIR-ATR spectra of each sample of the type can be tested under the optimized preparation conditions for the single-flavor Chinese medicine concentrated granules of the corresponding type. In addition, any suitable instrument can be used for FTIR-ATR spectrum measurement. For example, the Bruker ALPHA Platinum ATR QuickSnapTM The sampling module is used to generate a baseline FTIR-ATR spectrum of each sample. The ATR sampling module contains a diamond cell with a refractive index of 2.4. Each sample will be scanned at 7500-375 cm -1 The spectral range is 2cm -1 The spectral resolution of the FTIR-ATR spectrum is scanned, etc. The procedure of FTIR-ATR spectrum measurement should be carried out according to the instructions provided by the manufacturer of the instrument used and will not be described in full here.

[0044] Preferably, each sample may be measured multiple times, and then the results of the multiple measurements may be optimized after eliminating outliers, and / or corrected such as by averaging, so as to obtain a reference FTIR-ATR spectrum corresponding to the sample.

[0045] In some embodiments, in order to better train the first learning model in subsequent steps, each type of single-flavor Chinese medicine concentrated granules can be divided into multiple parts in advance, that is, the single-flavor Chinese medicine concentrated granules of the same type in the training sample set will contain multiple samples. According to the working principle of infrared spectroscopy, we know that the absorbance value of a specific peak in the infrared spectrum is proportional to the amount of the contributing chemical component. Therefore, the measurement results of the ATR-FTIR spectrum may vary to a certain extent due to factors such as the size of each sample. Therefore, the FTIR-ATR spectra of each sample of the same type can be calibrated.

[0046] Figure 2 A schematic diagram of FTIR-ATR spectra of various samples of the same type of single-flavor Chinese medicine concentrated granules after correction according to an embodiment of the present application is shown.

[0047] Due to the above reasons, the baseline of each FTIR-ATR spectrum may have a tendency to gradually increase, resulting in differences in the intensity of the FTIR-ATR spectrum of each sample, which all need to be corrected. Just as an example, Figure 2 The standard normal variant (SNV) method was used to normalize the FTIR-ATR spectra of 10 ginseng Chinese medicine concentrated granules samples, that is, the average value of the FTIR-ATR spectra of 10 samples was subtracted from the FTIR-ATR spectrum of each sample and divided by the standard deviation of each spectrum. Figure 2The FTIR-ATR spectra of the 10 ginseng Chinese medicine concentrated granules samples after correction processing shown basically overcome the defect of differences in FTIR-ATR spectrum intensity between samples. In some embodiments, similar deviation processing can be performed on the FTIR-ATR spectra of multiple repeated samples of single-flavor Chinese medicine concentrated granules of various types, so that before entering the subsequent steps, the FTIR-ATR spectra of single-flavor Chinese medicine concentrated granules of various types and samples have a benchmark as close as possible to avoid irrelevant changes caused by an increase in the baseline in the spectrum.

[0048] In step 103, the first learning model constructed based on the support vector machine is trained using the baseline FTIR-ATR spectra of each sample, and the first learning model is configured to learn the mapping relationship between the baseline FTIR-ATR spectra of each sample and each type of single-flavor Chinese medicine concentrated granules.

[0049] The inventors of the present application tried different algorithms during the experiment, such as PCA (Principal Component Analysis), which is a widely used data dimensionality reduction algorithm. Sample labels are not considered during the dimensionality reduction process and it is an unsupervised algorithm. Therefore, although clustering can be achieved when multiple samples are applied with the PCA algorithm, when the data meet certain characteristics, data of different categories may be merged into one category, that is, it cannot be correctly classified. Therefore, it is not suitable for the authentication algorithm in the present application.

[0050] Another type of dimensionality reduction algorithm, such as LDA (Linear Discriminant Analysis) algorithm, can take into account the labels of sample data, and therefore, can be used for the construction of the first learning model in this application, but relative to the machine learning algorithm based on support vector machine (Supported Vector Machine, SVM), through the comparison of the confusion matrix in the experimental results, it is found that its performance in product screening is inferior to the machine learning algorithm based on support vector machine, and therefore, this application will adopt the first learning model constructed based on support vector machine. The specific algorithm of support vector machine can adopt the algorithm and configuration commonly used in the art, which will not be repeated here. In addition, a software package that has been developed to support the algorithm based on support vector machine can also be used, and this application does not limit this.

[0051] When training the first learning model, multiple samples of each type of single-flavor Chinese medicine concentrated granules in the training sample set can be divided into a training set and a test set using a Kennard-Stone (KS) algorithm. The ratio of the number of samples in the training set and the test set can be 1:1, or can be set and adjusted as needed. This application does not limit this, but it is necessary to ensure that the sample data in the training set covers the sources of most variables in the training sample set to ensure that the trained first learning model can represent the entire training sample set to a greater extent.

[0052] Based on the sample set division as described above, the first learning model is trained for multiple rounds, with the goal of learning the mapping relationship between the benchmark FTIR-ATR spectrum of each sample and the various types of single-herb Chinese medicine concentrated granules to construct a corresponding loss function, and the loss function is minimized using methods such as gradient descent, ultimately achieving the goal of correctly outputting the type to which the sample belongs when the benchmark FTIR-ATR spectrum of the single-herb Chinese medicine concentrated granules sample in the test set is input into the first learning model.

[0053] Then, in step 104, when the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated is included in the type in the training sample set, the FTIR-ATR spectrum of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated is measured, wherein the measurement method of the FTIR-ATR spectrum is the same as described above and will not be repeated here.

[0054] Next, in step 105, the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is input into the trained first learning model. When the predicted type output by the trained first learning model matches the identification type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated, it is determined that the identification type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated has passed the authentication.

[0055] It is understandable that when the predicted type output by the trained first learning model does not match the identified type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated, it is determined that the identified type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated does not pass the authentication. Additionally, in order to ensure the reliability of authentication and the safety of the drug, even if the predicted type output by the first learning model is a type other than the identified type, it is not appropriate to determine that the true type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is the predicted type output by the model.

[0056] It is also worth noting that if the identification type of the sample of the single-ingredient Chinese medicine concentrated granules to be authenticated is not included in the category range of the training sample set of the first learning model, then the first learning model will not be able to accurately predict its category, and the output result of the first learning model cannot be used as the basis for authenticating the type of the single-ingredient Chinese medicine concentrated granules.

[0057] Figure 7 A schematic diagram showing the performance of the first learning model on the training set according to an embodiment of the present application is shown. Figure 7 The training sample set of the embodiment in the embodiment includes samples of the 11 types of single-flavor Chinese medicine concentrated granules mentioned above of multiple brands (Brand 1, Brand 2 and Brand 3), wherein the various single-flavor Chinese medicine concentrated granules may include multiple different batches (Batch 1 and Batch 2). In the experiment, each sample of each type, each brand and each batch was replicated 10 times, and the ratio of the number of samples in the training set and the test set was 1:1. Figure 7 The first learning model in the embodiment is implemented using the support vector machine-based algorithm software LIBSVM (V3.25) independently developed by the inventor of the present application, and is run on a chemometric analysis system independently developed by the inventor based on GNU Octave 7.2.0, wherein 1°, 2°, and 3° are the complexities of the first learning model.

[0058] like Figure 7 As shown, the accuracy of the trained first learning models of various complexities on the training set reaches 100%, and the accuracy on the test set is also over 98%. In addition, the training time of the first learning model is very short, usually between 3 seconds and 23 seconds.

[0059] Compared with the prior art which requires complex and time-consuming chemical analysis to identify the ingredients in single-flavor Chinese medicine concentrated granules, the authentication method for single-flavor Chinese medicine concentrated granules according to each embodiment of the present application obtains the FTIR-ATR fingerprint spectra of each type of single-flavor Chinese medicine concentrated granules by using samples of each type of single-flavor Chinese medicine concentrated granules in the training sample set in advance, and enables the first learning model to learn the mapping relationship between the FTIR-ATR fingerprint spectra and the single-flavor Chinese medicine concentrated granules in a very short time through training, thereby, for example, when a doctor or pharmacist is using the single-flavor Chinese medicine concentrated granules to be used, When the source and / or quality of the sample are in doubt, a simple FTIR-ATR spectrum measurement can be performed on the sample to be tested. Only the FTIR-ATR spectrum of the sample to be tested and the trained first learning model can be used to conveniently and accurately authenticate whether the single-flavor Chinese medicine concentrated granules to be tested are consistent with the type identified therein, thereby providing a reliable basis for Chinese medicine physicians to prescribe or pharmacists to dispense medicines, avoiding adverse effects or even more serious consequences on medication due to type deviation, impure ingredients, poor storage, contamination, etc. of the single-flavor Chinese medicine concentrated granules, thereby improving the safety of the application of single-flavor Chinese medicine concentrated granules.

[0060] In some embodiments, the authentication method according to the present application can not only authenticate the type of single-flavor Chinese medicine concentrated granules, but also authenticate the brand of the same type of single-flavor Chinese medicine concentrated granules. In the case where the samples of the first type of single-flavor Chinese medicine concentrated granules in the training sample set include different brands, the first learning model constructed based on the support vector machine is trained using the benchmark FTIR-ATR spectrum of the sample corresponding to the (first type, brand) binary group, and the first learning model is further configured to learn the mapping relationship between the benchmark FTIR-ATR spectrum of each sample and each type of single-flavor Chinese medicine concentrated granules, and the mapping relationship between the benchmark FTIR-ATR spectrum of the first type of sample and the (first type, brand) binary group.

[0061] That is to say, when the first learning model is trained, it carries the brand information associated with the type, so the trained first learning model can identify the type and brand of the single-flavor Chinese medicine concentrated granules at the same time. Therefore, when the (identification type, identification brand) tuple of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is contained in the (first type, brand) tuple of the sample in the training sample set, the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated can be measured first, and then the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated can be input into the trained first learning model. When the (predicted type, predicted brand) tuple output by the trained first learning model matches the (identification type, identification brand) tuple, it is determined that the identification type and identification brand of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated have passed the authentication.

[0062] Figure 8 (a), Figure 8 (b) and Figure 8 (c) respectively show the performance diagrams of the first learning model according to the embodiment of the present application on the training set, the test set and the prediction set. In Figure 8 (a), Figure 8 (b) and Figure 8 (c), the training sample set contains samples of the 11 types of single-flavor Chinese medicine concentrated granules mentioned above, and 2-3 different brands are selected for each type of single-flavor Chinese medicine concentrated granules. In the experiment, each sample of each type and each brand was replicated 10 times, and the ratio of the number of samples in the training set and the test set was 1:1. The brands of the 11 types of single-flavor Chinese medicine concentrated granules samples in the prediction set are randomly selected from the trained brands, wherein each sample in the prediction set is accompanied by (identification type, identification brand), and each indicator in Figure 8 (a), Figure 8 (b) and Figure 8 (c) is calculated based on the comparison results of the (prediction type, prediction brand) binary output by the trained first learning model and the (identification type, identification brand) binary.

[0063] The meaning and calculation method of each performance indicator in the figure are as follows:

[0064] Sensitivity = TP / (TP+FN)

[0065] Specificity = TN / (TN+FP)

[0066] Accuracy = TP / (TP+FP)

[0067] Among them, TP represents the number of true positive samples, FN represents the number of false negative samples, TN represents the number of true negative samples, and FP represents the number of false positive samples. The above "true" represents a correct prediction, and "false" represents an incorrect prediction; "positive" means that the (predicted category, predicted brand) tuple is completely consistent with (identified category, identified brand), and "negative" means that at least one item in the (predicted category, predicted brand) tuple is inconsistent with (identified category, identified brand).

[0068] As can be seen from Figure 8 (a), the sensitivity, specificity and accuracy of the trained first learning model for each type and brand of single-flavor Chinese medicine concentrated granule samples in the training set all reach 1.00; as can be seen from Figure 8 (b), the sensitivity, specificity and accuracy of the trained first learning model for most of the single-flavor Chinese medicine concentrated granule samples in the test set can also reach 1.00, except for the sensitivity of Poria cocos of brand 3, which is 0.80, and the specificity and accuracy of Poria cocos of brand 1 are 0.99 and 0.91 respectively; as can be seen from Figure 8 (c), the prediction set and the test set maintain a basically consistent trend, except for a few types of Poria cocos, Poria cocos and Codonopsis pilosula, there are certain deviations, and the remaining types can reach 1.00 in terms of sensitivity, specificity and accuracy. It can be seen that the first learning model trained according to the authentication method provided in the embodiment of the present application has good performance in terms of sensitivity, specificity and accuracy for single-flavor Chinese medicine concentrated granules of different types and brands participating in the training.

[0069] The inventors of the present application also discovered during the experiment that the moisture in the single-herb concentrated granules may affect the FTIR-ATR spectrum of the single-herb concentrated granules. Therefore, in some embodiments, in order to avoid the influence of the moisture contained in each sample in the training sample set on the FTIR-ATR spectrum of the single-herb concentrated granules and ensure the consistency of the spectrum measurement standards to obtain reliable and high-quality spectra, before measuring the baseline FTIR-ATR spectrum of each sample of each type of single-herb concentrated granules, a first pretreatment including a drying treatment can be performed on each sample, and then the baseline FTIR-ATR spectrum of each sample of each type of single-herb concentrated granules after the first pretreatment can be measured.

[0070] In some embodiments, after measuring the baseline FTIR-ATR spectrum of each sample in the training sample set, the spectrum diagram of the full spectrum segment is not necessarily used directly for the training of the first learning model, but the spectrum segment used for the training of the first learning model can be selected according to certain criteria, so that its training efficiency is higher and the performance is better. In other words, the baseline FTIR-ATR spectrum of each sample can be subjected to a second preprocessing including spectrum segment interception, and the baseline FTIR-ATR spectrum of each sample after the second preprocessing can be used to train the first learning model constructed based on the support vector machine, and the first learning model can learn the mapping relationship between the baseline FTIR-ATR spectrum of each sample after the spectrum segment interception and each type of single-flavor Chinese medicine concentrated granules. On this basis, when the identification type of the sample of the single-herb concentrated granules to be authenticated is included in the types in the training sample set, after measuring the FTIR-ATR spectrum of the sample of the single-herb concentrated granules to be authenticated, the corresponding spectral segment of the FTIR-ATR spectrum of the sample of the single-herb concentrated granules to be authenticated is similarly intercepted, and the FTIR-ATR spectrum of the authenticated sample of the single-herb concentrated granules after the spectral segment is intercepted is input into the trained first learning model. When the predicted type output by the trained first learning model matches the identification type of the sample of the single-herb concentrated granules to be authenticated, it is determined that the identification type of the sample of the single-herb concentrated granules to be authenticated has passed the authentication.

[0071] In other embodiments, when the types of samples of single-flavor Chinese medicine concentrated granules included in the training sample set are different, the spectral segments of the intercepted benchmark FTIR-ATR spectra can also be different, as long as the benchmark FTIR-ATR spectra of each type of single-flavor Chinese medicine concentrated granules in the intercepted spectral segment have greater distinguishability relative to the benchmark FTIR-ATR spectra of each type of single-flavor Chinese medicine concentrated granules in other spectral segments. In this case, it is shown that these types in the sample set have greater differences in the spectral segment, and the clustering effect is better. Therefore, when the benchmark FTIR-ATR spectra of each sample after the spectral segment is intercepted for training, different types of single-flavor Chinese medicine concentrated granules can be better distinguished. The method for specifically judging the distinguishability between the benchmark FTIR-ATR spectra of each type and each sample can, for example, use the aforementioned PCA algorithm to judge the clustering degree of the benchmark FTIR-ATR spectra of each type of sample, so as to select the spectral segment with the best distinguishability. In other embodiments, any other applicable algorithm may be used to determine the discrimination of the benchmark FTIR-ATR spectra of various types of single-flavor Chinese medicine concentrated granules in different spectral bands, and the present application does not limit this.

[0072] Taking the PCA algorithm and the embodiment in which the training sample set contains 11 kinds of single-flavor Chinese medicine concentrated granules as an example, FIG3(a) shows a schematic diagram of the benchmark FTIR-ATR spectra of each kind of single-flavor Chinese medicine concentrated granules before the spectral segment is cut according to the embodiment of the present application. FIG3(b) shows a schematic diagram of the PCA scores of the benchmark FTIR-ATR spectra of each kind of single-flavor Chinese medicine concentrated granules before the spectral segment is cut and after the different spectral segments are cut according to the embodiment of the present application. FIG3(c) shows a schematic diagram of the benchmark FTIR-ATR spectra of each kind of single-flavor Chinese medicine concentrated granules after the spectral segment is cut according to the embodiment of the present application.

[0073] Figure 3(a) shows the -1 The full-spectrum FTIR-ATR spectra of 11 types of single-flavor Chinese medicine concentrated granules were processed using the PCA algorithm. The PCA scores of each type of sample are shown in the upper left figure in Figure 3(b). In Figure 3(b), PC1, PC2, and PC3 are the first principal component axis, the second principal component axis, and the third principal component axis, respectively. The percentages indicated in brackets are the proportions of the contribution of each principal component. As can be seen from the upper left figure in Figure 3(b), after applying the PCA algorithm to the full spectrum, the clustering of the FTIR-ATR spectra of each type is not good, and samples of different categories cannot be completely distinguished from each other. Similarly, the upper right and lower right figures in Figure 3(b) show the interception of 3600-2200cm -1 and 1150-750cm -1 When the spectrum is cut, the PCA scores of samples of different categories are obtained. It can be seen that the above two spectrum cut methods cannot distinguish samples of different categories well. In some embodiments, for example, the spectrum range can be optimized by combining the observation method and the traversal algorithm. As an example, the lower left figure in Figure 3 (b) shows that when the 1700-1150cm -1 In the case of the spectral segment, the PCA scores of each type of samples after applying the PCA algorithm show that under this spectral segment interception method, the clustering effect of each type of samples is very good, that is, it has a good degree of discrimination. It can be seen that in the case of the training sample set containing 11 types of single-flavor Chinese medicine concentrated granules, including Astragalus, Codonopsis, Fritillaria, Fritillaria, Honeysuckle, American Ginseng, Poria, Poria, Pseudostellaria, Ginseng and Smilax glabra, it is preferred to intercept the 1700-1150cm -1 The spectrum in the spectral range is used as the training data for the first learning model. Figure 3(c) shows the 1700-1150cm range shown at 301 in Figure 3(a) cut out from the entire FTIR-ATR spectrum range. -1 The spectrum of the range is used as the training data for the first learning model.

[0074] It is worth noting that even single-flavor Chinese medicine concentrated granules produced by the same pharmaceutical manufacturer may have differences in different batches. The main reasons for the differences include, but are not limited to: differences in the Chinese medicinal materials used as raw materials for producing single-flavor Chinese medicine concentrated granules, including differences in origin and harvest seasons, and different manufacturing processes of single-flavor Chinese medicine concentrated granules, etc. Therefore, in other embodiments, it may also be necessary to authenticate batches of single-flavor Chinese medicine concentrated granules. According to the authentication method of the present application, not only can the type of single-flavor Chinese medicine concentrated granules be authenticated, but also batches of single-flavor Chinese medicine concentrated granules of the same type can be authenticated at the same time. Since batches are usually relative to the same brand, in order to better distinguish batches, single-flavor Chinese medicine concentrated granules of the same type of different brands can be regarded as different types of single-flavor Chinese medicine concentrated granules, and each sample in the training sample set is marked with a (type, batch) tuple label, and the (type, batch) tuple is authenticated as a whole.

[0075] Figure 4 A schematic flow chart of a method for authenticating the identified type and identified batch of a single-flavor Chinese medicine concentrated granule to be authenticated according to an embodiment of the present application is shown.

[0076] like Figure 4 As shown, in step 401, samples of single-flavor Chinese medicine concentrated granules of various types are obtained from the training sample set, wherein each type of single-flavor Chinese medicine concentrated granules is identified by identity verification with the corresponding type of Chinese medicinal materials; in step 102, the baseline FTIR-ATR spectrum of each sample of each type of single-flavor Chinese medicine concentrated granules is measured. Steps 401 and 402 are respectively related to Figure 1 The steps 101 and 102 in the above description correspond to each other, so the details are not repeated here.

[0077] In step 403, when the samples of the second type of single-herb concentrated granules of the second type in the training sample set include different batches, the correlation coefficients between the batches of the second type are identified using the baseline FTIR-ATR spectra of the samples of the batches of the second type. When the correlation coefficients between the batches are all lower than the first threshold, the first learning model constructed based on the support vector machine is trained using the baseline FTIR-ATR spectra of the samples corresponding to the (second type, batch) binary group. The first learning model is further configured to learn the mapping relationship between the baseline FTIR-ATR spectra of the samples and the types of the single-herb concentrated granules, and the mapping relationship between the baseline FTIR-ATR spectra of the samples of the second type and the (second type, batch) binary group.

[0078] It can be seen that when it is desired to identify and authenticate different batches of the same type of single-flavor Chinese medicine concentrated granules, it is first necessary to determine whether the batches included in the category (i.e., the first category) have sufficiently large differences from each other, so that they can be distinguished as a subcategory. Therefore, in step 403, the reference FTIR-ATR spectra of the samples of the second category batches can be used, for example, through various applicable correlation algorithms, to identify the correlation coefficients between the batches of the second category. Then, when the correlation coefficients between the batches are all lower than the first threshold, it can be considered that the batches in the second category have sufficient discrimination, and (second category, first batch), (second category, second batch), ..., can be regarded as subcategories of the second category, respectively. Therefore, the annotation of the samples of the second category can be converted into the form of (second category, batch), and then the reference FTIR-ATR spectra of the corresponding samples with the (second category, batch) binary identification are used to train the first learning model, so as to further learn the mapping relationship between the reference FTIR-ATR spectra of the samples of the second category and the (second category, batch) binary through training. The method for calculating the correlation coefficient can be any method, and this application does not limit this. Regarding the first threshold, it can be set based on comprehensive considerations of performance indicators such as sensitivity, specificity, and accuracy to be achieved in implementing the authentication, and this application does not limit this.

[0079] In step 404, when the (identification type, identification batch) tuple of the sample of the single Chinese medicine concentrated granules to be authenticated is included in the (second type, batch) tuple of the sample in the training sample set, the FTIR-ATR spectrum of the sample of the single Chinese medicine concentrated granules to be authenticated is measured. That is, only the (second type, batch) tuple that passes the correlation coefficient determination in step 403 can be used to authenticate the batch of the single Chinese medicine concentrated granules to be authenticated.

[0080] Finally, in step 405, the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is input into the trained first learning model, and when the (predicted type, predicted batch) tuple output by the trained first learning model matches the (identified type, identified batch) tuple, it is determined that the identified type and identified batch of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated have passed the authentication. It is understandable that, in order to ensure the reliability of the authentication, except for the above situation, it cannot be determined that the identified type and identified batch of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated have passed the authentication at the same time.

[0081] Through the above-mentioned embodiments of the present application, not only the type of single-ingredient Chinese medicine concentrated granules can be authenticated, but also the batches can be authenticated under certain circumstances. It can be further understood that, combined with the method of authenticating the (type, brand) binary group and (type, batch) mentioned above, in fact, when the batch conditions can be met (that is, the correlation coefficients between each batch are lower than the first threshold), the authentication method according to the embodiment of the present application can support the simultaneous authentication of the (type, brand, batch) triplet of single-ingredient Chinese medicine concentrated granules, thereby providing more convenient services for Chinese medicine doctors, pharmacists, etc.

[0082] In other embodiments, the authentication method according to the embodiments of the present application can also authenticate whether the single-flavor Chinese medicine concentrated granules meet the required storage conditions.

[0083] The inventors of the present application have found through a series of experiments that inappropriate storage conditions may cause certain active ingredients in single-flavor Chinese medicine concentrated granules to degrade, mold, or produce other undesirable changes in properties or efficacy under favorable temperature and moisture conditions, and the first learning model obtained by training the FTIR-ATR spectrum of single-flavor Chinese medicine concentrated granules under good storage conditions can usually only accurately identify the type of single-flavor Chinese medicine concentrated granules or (second type, batch) binary groups with storage conditions within a certain range. Therefore, the authentication method according to the embodiment of the present application also includes, when the identification type of the single-flavor Chinese medicine concentrated granule sample to be authenticated is included in the type in the training sample set, the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granule to be authenticated is input into the trained first learning model, and when the predicted type output by the trained first learning model matches the identification type of the sample of the single-flavor Chinese medicine concentrated granule to be authenticated, it is determined that the identification type of the sample of the single-flavor Chinese medicine concentrated granule to be authenticated has passed the authentication, and the storage conditions of the sample are within a predetermined range.

[0084] Figure 5 The following is an example of the prediction results of the first learning model according to the embodiment of the present application for single-flavor Chinese medicine concentrated granules under different storage conditions, wherein the storage conditions are usually a combination of temperature and humidity. Figure 5As shown in the figure, under some storage conditions, the first learning model will not be able to correctly predict the types of some types of single-flavor Chinese medicine concentrated granules. For example, under the conditions of (20°C, 40%RH), (20°C, 60%RH), (20°C, 80%RH), (30°C, 40%RH), (30°C, 60%RH), and (30°C, 80%RH), the first learning model cannot correctly predict the types of honeysuckle concentrated granules. However, for Sichuan Fritillaria Chinese medicine concentrated granules and American ginseng Chinese medicine concentrated granules, the first learning model can correctly predict their types under the storage conditions given by the test. It can be seen that different types of single-flavor Chinese medicine concentrated granules can correspond to the same or different storage conditions. In other words, the range of the combination of temperature and humidity of the "appropriate" storage conditions that are not necessarily the same for each different type of single-flavor Chinese medicine concentrated granules can be pre-set. The range may not be continuous, but a number of segmented intervals, and then provide authentication of whether the above-mentioned pre-set "appropriate" storage conditions are met. In addition, the authentication of the above-mentioned storage conditions can be combined with batch authentication, which will not be repeated here.

[0085] According to the authentication method of the embodiment of the present application, by using the trained first learning model, the type, or type and batch of single-flavor Chinese medicine concentrated granules can be quickly and accurately authenticated without the need for complex and time-consuming chemical analysis and testing. At the same time, it can also determine whether the storage conditions are appropriate, providing a reliable basis for Chinese medicine doctors to prescribe or pharmacists to dispense medicines. When the authentication fails, it can be effectively prompted that the single-flavor Chinese medicine concentrated granules do not belong to the type it identifies, or do not match the identified batch, and the sample to be authenticated has not been stored under appropriate conditions. As a result, Chinese medicine doctors or pharmacists can avoid adverse effects or even more serious consequences in medication due to type deviation, impure ingredients, poor storage, contamination, and the presence of toxic ingredients in the single-flavor Chinese medicine concentrated granules, thereby improving the safety of the application of single-flavor Chinese medicine concentrated granules.

[0086] According to an embodiment of the present application, an authentication device for a single-flavor Chinese medicine concentrated granule is also provided, wherein the single-flavor Chinese medicine concentrated granule includes an excipient in addition to the ingredients of the single Chinese medicinal material. Figure 6 A block diagram showing some components of an authentication device for concentrated single-flavor Chinese medicine granules according to an embodiment of the present application is shown.

[0087] like Figure 6As shown, the authentication device 600 may include at least an interface 601 and a processor 602. The interface 601 may be configured to receive, for example, a reference FTIR-ATR spectrum of a sample identified as each type of single-flavor Chinese medicine concentrated granules in a training sample set, wherein each type of single-flavor Chinese medicine concentrated granules is identified by verifying the identity with the corresponding type of Chinese medicinal materials, and a FTIR-ATR spectrum of a sample of the single-flavor Chinese medicine concentrated granules to be authenticated.

[0088] The processor 602 may be configured to execute the authentication method of the single-flavor Chinese medicine concentrated granules as described in various embodiments of the present application, and the specific examples are as follows:

[0089] The processor 602 uses the measured benchmark FTIR-ATR spectra of each sample of each type of single-flavor Chinese medicine concentrated granules to train a first learning model constructed based on a support vector machine, wherein the first learning model is configured to learn a mapping relationship between the benchmark FTIR-ATR spectra of each sample and each type of single-flavor Chinese medicine concentrated granules;

[0090] The processor 602 determines whether the identification type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is included in the type range of the training sample set. If so, the measured FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is input into the trained first learning model, and the trained first learning model outputs the authentication result, wherein when the predicted type output by the trained first learning model matches the identification type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated, it is determined that the identification type of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated has passed the authentication. In some embodiments, the interface 601 can receive the benchmark FTIR-ATR spectra of the samples of the single-flavor Chinese medicine concentrated granules identified as various types in the training sample set and the FTIR-ATR spectrum data of the samples of the single-flavor Chinese medicine concentrated granules to be authenticated manually imported by the user through the user interface, or directly receive from other test systems, and the present application does not limit this. Interface 601 may include a network adapter, a cable connector, a serial connector, a USB connector, a parallel connector, a high-speed data transmission adapter (such as optical fiber, USB 3.0, Thunderbolt interface, etc.), a wireless network adapter (such as a WiFi adapter), a telecommunications (3G, 4G / LTE, etc.) adapter, etc., but this application does not impose any restrictions on this.

[0091] In some embodiments, the processor 602 may be a processing unit including one or more general-purpose processing devices, such as a microprocessor, a central processing unit (CPU), a graphics processing unit (GPU), etc. More specifically, the processor 602 may be a complex instruction set computing (CISC) microprocessor, a reduced instruction set computing (RISC) microprocessor, a very long instruction word (VLIW) microprocessor, a processor running other instruction sets, or a processor running a combination of instruction sets. In some embodiments, the processor 602 may also be one or more special-purpose processing devices, such as an application-specific integrated circuit (ASIC), a field programmable gate array (FPGA), a digital signal processor (DSP), a system on a chip (SoC), etc.

[0092] In other embodiments, the authentication device 600 may also include a memory (not shown) for storing the first learning model and data such as the FTIR-ATR spectrum associated with the training sample set. The memory may be, for example, a read-only memory (ROM), a random access memory (RAM), a phase change random access memory (PRAM), a static random access memory (SRAM), a dynamic random access memory (DRAM), an electrically erasable programmable read-only memory (EEPROM), other types of random access memory (RAM), a flash disk or other forms of flash memory, a cache, a register, a static memory, a compact disk read-only memory (CD-ROM), a digital versatile disk (DVD) or other optical memory, a cassette or other magnetic storage device, or any other possible non-temporary medium used to store information or instructions that can be accessed by a computer device.

[0093] The above description is intended to be illustrative rather than restrictive. For example, the above examples (or one or more of their solutions) may be used in combination with each other. Other embodiments may be used by those of ordinary skill in the art when viewing the above description. Moreover, in the above detailed description, various features may be combined together to simplify the application. This should not be interpreted as intending to make the disclosed features that are not required to be protected indispensable for any claim. Instead, the subject matter of the invention may be a combination of features less than all the features of a disclosed embodiment. Therefore, the claims are thus incorporated into the specific implementation as examples or embodiments, wherein each claim is independently a separate embodiment, and it is conceivable that these embodiments may be combined with each other in various combinations or permutations. The scope of the present invention should be determined with reference to the claims and the full scope of equivalents to these claims.

Claims

1. A method for authenticating a single-flavor Chinese medicine concentrated granule, wherein the single-flavor Chinese medicine concentrated granule comprises an excipient in addition to the ingredients of the single-flavor Chinese medicine, characterized in that: The authentication method comprises: Obtain samples identified as various types of single-flavor Chinese medicine concentrated granules in the training sample set, wherein the various types of single-flavor Chinese medicine concentrated granules are identified by identity verification with corresponding types of Chinese medicinal materials; Measure the baseline FTIR-ATR spectra of each sample of each type of single herbal concentrated granules; Using the benchmark FTIR-ATR spectra of each sample, a first learning model constructed based on a support vector machine is trained, wherein the first learning model is configured to learn a mapping relationship between the benchmark FTIR-ATR spectra of each sample and each type of single-flavor Chinese medicine concentrated granules; In the case where the identification type of the sample of the single-flavor Chinese medicine concentrated granule to be authenticated is included in the types in the training sample set, Measuring the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated; and The FTIR-ATR spectrum of the sample of the single-herb concentrated granules to be authenticated is input into the trained first learning model. When the predicted type output by the trained first learning model matches the identified type of the sample of the single-herb concentrated granules to be authenticated, it is determined that the identified type of the sample of the single-herb concentrated granules to be authenticated has passed the authentication.

2. The authentication method according to claim 1, characterized in that: The authentication method further comprises: In the case where the samples of the first category of the single-flavor Chinese medicine concentrated granules in the training sample set include different brands, the first learning model constructed based on the support vector machine is trained using the benchmark FTIR-ATR spectra of the samples corresponding to the (first category, brand) binary group, wherein the first learning model is further configured to learn the mapping relationship between the benchmark FTIR-ATR spectra of each sample and each category of the single-flavor Chinese medicine concentrated granules, and the mapping relationship between the benchmark FTIR-ATR spectra of the samples of the first category and the (first category, brand) binary group; When the (identification type, identification brand) tuple of the single-flavor Chinese medicine concentrated granule sample to be authenticated is included in the (first type, brand) tuple of the sample in the training sample set: Measuring the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated; and The FTIR-ATR spectrum of the sample of the single-ingredient Chinese medicine concentrated granules to be authenticated is input into the trained first learning model. When the (predicted type, predicted brand) bigram output by the trained first learning model matches the (identification type, identification brand) bigram, it is determined that the identification type and identification brand of the sample of the single-ingredient Chinese medicine concentrated granules to be authenticated have passed the authentication.

3. The authentication method according to claim 1 or 2, characterized in that: The authentication method further includes: before measuring the baseline FTIR-ATR spectrum of each sample of each type of single-flavor Chinese medicine concentrated granules, performing a first pretreatment including a drying treatment on each sample; Performing a second preprocessing including spectral segment interception on the benchmark FTIR-ATR spectrum of each sample, and using the benchmark FTIR-ATR spectrum of each sample after the second preprocessing to train a first learning model constructed based on a support vector machine, wherein the first learning model is configured to learn a mapping relationship between the benchmark FTIR-ATR spectrum of each sample after spectral segment interception and each type of single-flavor Chinese medicine concentrated granules; When the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated is included in the types in the training sample set, the FTIR-ATR spectrum of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated is measured, and the FTIR-ATR spectrum of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated is intercepted with the same spectral segment as the benchmark FTIR-ATR spectrum of each sample in the training sample set, and then input into the trained first learning model. When the predicted type output by the trained first learning model matches the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated, it is determined that the identification type of the sample of the single-herb concentrated granules of Chinese medicine to be authenticated has passed the authentication.

4. The authentication method according to claim 1 or 2, characterized in that: The authentication method further comprises: In the case where the samples of the second category of single-flavor concentrated Chinese herbal medicine granules in the training sample set include different batches, using the benchmark FTIR-ATR spectra of the samples of each batch of the second category to identify the correlation coefficients between the batches of the second category, and in the case where the correlation coefficients between the batches are all lower than the first threshold, using the benchmark FTIR-ATR spectra of the samples corresponding to the (second category, batch) binary group to train the first learning model constructed based on the support vector machine, the first learning model is further configured to learn the mapping relationship between the benchmark FTIR-ATR spectra of each sample and the categories of the single-flavor concentrated Chinese herbal medicine granules, and the mapping relationship between the benchmark FTIR-ATR spectra of the samples of the second category and the (second category, batch) binary group; When the (identification type, identification batch) tuple of the single-flavor Chinese medicine concentrated granule sample to be authenticated is included in the (second type, batch) tuple of the sample in the training sample set: Measuring the FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated; and The FTIR-ATR spectrum of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated is input into the trained first learning model. When the (predicted type, predicted batch) tuple output by the trained first learning model matches the (identification type, identification batch) tuple, it is determined that the identification type and identification batch of the sample of the single-flavor Chinese medicine concentrated granules to be authenticated have passed the authentication.

5. The authentication method according to claim 1, characterized in that: The authentication method further comprises: In the case where the identification type of the single-flavor Chinese medicine concentrated granule sample to be authenticated is included in the types in the training sample set, The FTIR-ATR spectrum of the sample of the single-flavor concentrated Chinese herbal granules to be authenticated is input into the trained first learning model, and when the predicted type output by the trained first learning model matches the identification type of the sample of the single-flavor concentrated Chinese herbal granules to be authenticated, it is determined that the identification type of the sample of the single-flavor concentrated Chinese herbal granules to be authenticated has passed the authentication, and the storage condition of the sample is within a predetermined range; wherein, The storage conditions include a combination of temperature and humidity, and different types of single-flavor Chinese medicine concentrated granules correspond to different storage conditions.

6. The authentication method according to claim 1 or 2, characterized in that: The authentication method further includes: including in the training sample set samples of single-flavor Chinese medicine concentrated granules of types that are easily confused in terms of properties or efficacy corresponding to each type.

7. The authentication method according to claim 3, characterized in that: The authentication method further comprises: When the types of samples of single-herb concentrated granules of Chinese medicine included in the training sample set are different, the spectral segments of the intercepted benchmark FTIR-ATR spectra are different, so that the benchmark FTIR-ATR spectra of each type of single-herb concentrated granules of Chinese medicine in the intercepted spectral segments have greater distinguishability than the benchmark FTIR-ATR spectra of each type of single-herb concentrated granules of Chinese medicine in other spectral segments.

8. The authentication method according to claim 3, characterized in that: The authentication method further comprises: When the types of single-herb concentrated granules of Chinese medicine included in the training sample set include 11 types, namely, Astragalus, Codonopsis, Fritillaria, Fritillaria, Honeysuckle, American ginseng, Poria, Poria, Pseudostellaria, Ginseng and Smilax glabra, the spectra of each sample in the spectral range of 1700-1150cm-1 are intercepted to train the first learning model.

9. The authentication method according to claim 1 or 2, characterized in that: The authentication method further comprises: Ultra-high performance liquid chromatography was used to verify the identity of each type of single-herb concentrated granules with the corresponding type of Chinese medicinal materials, taking into account the excipients.

10. An authentication device for a single-flavor Chinese medicine concentrated granule, wherein the single-flavor Chinese medicine concentrated granule includes an excipient in addition to the ingredients of the single-flavor Chinese medicine, characterized in that: The authentication device comprises: An interface configured to: receive a baseline FTIR-ATR spectrum of a sample identified as each type of single-flavor Chinese medicine concentrated granules in a training sample set, wherein each type of single-flavor Chinese medicine concentrated granules is identified by identity verification with a corresponding type of Chinese medicinal material; and an FTIR-ATR spectrum of a sample of a single-flavor Chinese medicine concentrated granules to be authenticated; A processor configured to execute the authentication method for concentrated single-flavor Chinese medicine granules as described in any one of claims 1-9.

Images