Traditional Chinese medicine constitution identification method and device, storage medium and server

Abstract

The invention discloses a traditional Chinese medicine constitution identification method and device, a storage medium and a server. The method comprises the following steps: acquiring a human face spectrum sample and blood fat contents of a tested crowd; inputting the human face spectrum sample into a pre-constructed hyperlipidemia risk prediction model to obtain a target human face spectrum sample with hyperlipidemia in the tested crowd and a first target blood fat four-item content corresponding to the target human face spectrum sample; making corresponding labels on the basis of the target face spectrum samples, corresponding first target blood fat four-item contents and traditional Chinese medicine constitution categories, and training the model through training samples obtained by making the corresponding labels to obtain a traditional Chinese medicine constitution identification model; obtaining a to-be-detected human face spectrum sample, and inputting the sample into the traditional Chinese medicine constitution identification model to obtain the content of four second target blood lipids suffering from hyperlipidemia in the detected crowd and a traditional Chinese medicine constitution identification result corresponding to the content of the four second target blood lipids; the technical problem that enough support cannot be accurately provided for treatment and prevention of patients with the lipemia is solved.

Description

Technical Field

[0001] The present application relates to the field of intelligent disease identification, and more particularly, to a traditional Chinese medicine (TCM) constitution identification method, device, storage medium, and server. Background Art

[0002] Research shows that there is a certain correlation between the classification of TCM constitutions and the blood lipid levels of patients with hyperlipidemia. The criteria for determining hyperlipidemia are formulated with reference to the 2007 Chinese Guidelines for the Prevention and Treatment of Dyslipidemia in Adults. Meeting one or more of the following: total cholesterol (TC) ≥ 6.22 mmol / L, triglyceride (TG) ≥ 2.26 mmol / L, low-density lipoprotein cholesterol (LDL-c) ≥ 4.14 mmol / L, high-density lipoprotein cholesterol (HDL-c) ≤ 1.04 mmol / L.

[0003] The nine common TCM constitutions are: balanced constitution, yin-deficient constitution, yang-deficient constitution, phlegm-damp constitution, blood stasis constitution, qi-deficient constitution, qi stagnation constitution, special endowment constitution, and damp-heat constitution. Among them, the balanced constitution belongs to the normal constitution, and the other 8 belong to the biased constitutions.

[0004] In the prior art, the patent with the publication number CN118570539A discloses a TCM constitution identification method based on tongue image features, which relates to the field of image recognition technology. The key points of its technical solution are: the identification method includes: constructing a TCM constitution identification model with expert consensus tongue image features, which includes the following steps: collecting tongue image information, where the tongue image information includes tongue surface image information and sublingual image information; extracting tongue image features with expert consensus through a deep learning model according to the tongue image information in step S1; after fusing the tongue image features and text features, using an artificial neural network model to perform constitution identification on the tongue image features, and constructing a TCM constitution identification model with expert consensus tongue image features.

[0005] This patent can achieve TCM constitution identification by comprehensively using expert consensus tongue image features and text features, and by adopting deep learning and artificial neural network models, and can improve the accuracy and efficiency in TCM constitution identification; however, it cannot achieve TCM constitution identification based on the human face, and does not explore the relevance between hyperlipidemia, the four blood lipid contents, and TCM constitution identification, so it cannot provide sufficient support for the treatment and prevention of patients with hyperlipidemia accurately.

[0006] Regarding the problem that in the related art, TCM constitution identification cannot be achieved based on the human face, and the relevance between hyperlipidemia, the four blood lipid contents, and TCM constitution identification has not been explored, resulting in the inability to provide sufficient support for the treatment and prevention of patients with hyperlipidemia accurately, no effective solution has been proposed yet. Summary of the Invention

[0007] The main objective of this application is to provide a traditional Chinese medicine (TCM) constitution identification method, device, storage medium, and server, aiming to solve the problem that TCM constitution identification cannot be achieved based on human faces, and the lack of exploration of the correlation between hyperlipidemia, the four lipid components, and TCM constitution identification, which results in the inability to provide sufficient support for the treatment and prevention of patients with hyperlipidemia accurately.

[0008] To achieve the above objective, according to one aspect of this application, a TCM constitution identification method is provided.

[0009] The TCM constitution identification method according to this application includes: obtaining the face spectral samples and the four lipid component contents of the tested population; inputting the face spectral samples into a pre-constructed hyperlipidemia risk prediction model to obtain the target face spectral samples of the tested population with hyperlipidemia and their corresponding first target four lipid component contents; making corresponding labels based on the target face spectral samples, their corresponding first target four lipid component contents, and TCM constitution categories, and training the model with the training samples obtained by making corresponding labels to obtain a TCM constitution identification model; obtaining the to-be-tested face spectral samples and inputting them into the TCM constitution identification model to obtain the second target four lipid component contents of the tested population with hyperlipidemia and their corresponding TCM constitution identification results.

[0010] Further, obtaining the face spectral samples of the tested population includes: scanning the face images of the tested population based on near-infrared imaging technology, using the Dlib model to detect the face target areas in the face images, and generating a face detection box box(x 1 , x 2 , x 3 , x 4 ), where x i is the tuple of pixel point i, representing the pixel values of the row and column where the pixel point is located; cropping the face core area according to the face detection box, averaging and converting at least one scan data of the core area to obtain the face spectral samples.

[0011] Further, obtaining the four lipid component contents of the tested population includes: using the blood detection method to detect the lipid contents in the blood of the tested population to obtain the four lipid component contents; where the four lipid component contents at least include total cholesterol Tc, triglyceride TG, high-density lipoprotein HDL-C, and low-density lipoprotein LDL-C.

[0012] Further, the construction of the hyperlipidemia risk prediction model includes: preprocessing the face video samples; establishing the correspondence between the preprocessed face video samples and the standard four lipid component contents in the hyperlipidemia risk determination criteria; and training the model with reference to the correspondence and the hyperlipidemia risk determination criteria to obtain the hyperlipidemia risk prediction model.

[0013] Further, before inputting the face spectral sample of the population to be tested and the four blood lipid contents into the pre-constructed hyperlipidemia risk prediction model to obtain the target face spectral sample of the population with hyperlipidemia and the corresponding first target four blood lipid contents, it further includes: preprocessing the face spectral sample.

[0014] Further, preprocessing the face spectral sample includes: using a moving average filtering method or a Savitzky-Golay filtering method to eliminate noise in the face spectral sample; using polynomial baseline correction or least squares baseline correction to correct the baseline drift of the face spectral sample; performing normalization processing on the face spectral sample and scaling it to the same range to obtain a preprocessed face spectral sample.

[0015] Further, training the model with the training samples obtained by making corresponding labels to obtain a traditional Chinese medicine constitution identification model includes: adopting a Logistic regression analysis method, inputting the training samples obtained by making corresponding labels into a model composed of multiple binary classification models for learning and training, and constructing a traditional Chinese medicine constitution identification model.

[0016] To achieve the above object, according to another aspect of the present application, a traditional Chinese medicine constitution identification device is provided.

[0017] The traditional Chinese medicine constitution identification device according to the present application includes: an acquisition module for acquiring the face spectral sample of the population to be tested and the four blood lipid contents; a prediction module for inputting the face spectral sample into the pre-constructed hyperlipidemia risk prediction model to obtain the target face spectral sample of the population with hyperlipidemia and the corresponding first target four blood lipid contents; a training module for making corresponding labels based on the target face spectral sample, the corresponding first target four blood lipid contents and the traditional Chinese medicine constitution category, and training the model with the training samples obtained by making corresponding labels to obtain a traditional Chinese medicine constitution identification model; an identification module for acquiring the face spectral sample to be tested and inputting it into the traditional Chinese medicine constitution identification model to obtain the second target four blood lipid contents of the population with hyperlipidemia and the corresponding traditional Chinese medicine constitution identification result.

[0018] To achieve the above object, according to another aspect of the present application, a computer-readable storage medium is provided.

[0019] In the computer-readable storage medium according to the present application, a computer program is stored, wherein the computer program is set to execute the traditional Chinese medicine constitution identification method when running.

[0020] To achieve the above object, according to another aspect of the present application, a server is provided.

[0021] The server according to the present application includes: a memory and a processor, wherein the memory stores a computer program, wherein the processor is configured to run the computer program to execute the TCM constitution identification method.

[0022] In the embodiment of the present application, a method of TCM constitution identification is adopted, by obtaining facial spectral samples and four blood lipid contents of the tested population; the facial spectral samples are input into a pre-constructed hyperlipidemia risk prediction model to obtain target facial spectral samples of hyperlipidemia in the tested population and the first target four blood lipid contents corresponding thereto; corresponding labels are made based on the target facial spectral samples, the first target four blood lipid contents corresponding thereto, and the TCM constitution categories, and the model is trained by the training samples obtained by making the corresponding labels to obtain a TCM constitution identification model; the facial spectral samples of the tested population are obtained, and the TCM constitution is input into the pre-constructed hyperlipidemia risk prediction model to obtain the target facial spectral samples and the first target four blood lipid contents corresponding thereto; corresponding labels are made based on the target facial spectral samples, the first target four blood lipid contents corresponding thereto, and the TCM constitution categories, and the TCM constitution identification model is obtained by training the model with the training samples obtained by making the corresponding labels ... TCM constitution identification model is obtained by The TCM constitution identification model was established to obtain the second target four blood lipid contents of hyperlipidemia in the tested population and the corresponding TCM constitution identification results; the purpose of realizing TCM constitution identification based on human face and determining the correlation between hyperlipidemia, four blood lipid contents and TCM constitution identification was achieved, thereby achieving the technical effect of accurately providing sufficient support for the treatment and prevention of patients with lipidemia, and further solving the technical problem of being unable to accurately provide sufficient support for the treatment and prevention of patients with lipidemia due to the inability to realize TCM constitution identification based on human face and not exploring the correlation between hyperlipidemia, four blood lipid contents and TCM constitution identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The drawings constituting a part of this application are used to provide a further understanding of this application, so that other features, purposes and advantages of this application become more obvious. The schematic embodiment drawings and their descriptions of this application are used to explain this application and do not constitute an improper limitation on this application. In the drawings:

[0024] Figure 1 is a schematic diagram of a process of a method for identifying a constitution of TCM according to an embodiment of the present application;

[0025] Figure 2 It is a schematic diagram of the structure of a TCM constitution identification device according to an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0027] It should be noted that the terms "first", "second", etc. in the description, claims and above-mentioned drawings of this application are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances, so as to implement the embodiments of the present application described herein. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device comprising a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0028] In this application, the orientation or positional relationship indicated by the terms "upper", "lower", "left", "right", "front", "rear", "top", "bottom", "inner", "outer", "middle", "vertical", "horizontal", "lateral", "longitudinal", etc. is based on the orientation or positional relationship shown in the drawings. These terms are mainly used to better describe the present invention and its embodiments, and are not used to limit that the indicated devices, elements or components must have a specific orientation or be constructed and operated in a specific orientation.

[0029] Moreover, in addition to being able to represent the orientation or positional relationship, some of the above terms may also be used to represent other meanings. For example, the term "upper" may also be used to represent a certain attachment relationship or connection relationship in some cases. For those of ordinary skill in the art, the specific meanings of these terms in the present invention can be understood according to specific circumstances.

[0030] In addition, the terms "installed", "set", "provided with", "connected", "connected to", "socketed" should be understood in a broad sense. For example, it can be a fixed connection, a detachable connection, or an integral structure; it can be a mechanical connection or an electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or there can also be internal communication between two devices, elements or components. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0031] It should be noted that, without conflict, the embodiments in this application and the features in the embodiments can be combined with each other. The following will describe this application in detail with reference to the drawings and in combination with the embodiments.

[0032] According to an embodiment of the present invention, a traditional Chinese medicine constitution identification method is provided. As Figure 1 shown, the method includes the following steps S101 to step S104:

[0033] Step S101, obtain the face spectrum sample and the four-item blood lipid content of the tested population;

[0034] Scan the face of the person to be tested based on near-infrared imaging technology and collect the face spectral samples to obtain the contents (concentrations) of the four blood lipids in the blood of the person to be tested.

[0035] Specifically, obtaining the face spectral samples of the population to be tested includes: scanning the face images of the population to be tested based on near-infrared imaging technology, using the Dlib model to detect the face target area in the face images, and generating a face detection box box(x 1 ,x 2 ,x 3 ,x 4 ), where x i is the tuple of pixel point i, representing the pixel values of the row and column where the pixel point is located; cropping the face core area according to the face detection box, and averaging and converting at least one scan data of the core area to obtain the face spectral samples.

[0036] Specifically, obtaining the contents of the four blood lipids of the population to be tested includes: using the blood detection method to detect the blood lipid content in the blood of the population to be tested to obtain the contents of the four blood lipids; where the contents of the four blood lipids at least include total cholesterol Tc, triglyceride TG, high-density lipoprotein HDL-C, and low-density lipoprotein LDL-C.

[0037] By using face detection technology and combined processing and conversion, the face spectral samples are obtained from the face images, providing data guarantee for subsequent traditional Chinese medicine constitution identification based on the face.

[0038] Step S102: Input the face spectral samples into a pre-constructed hyperlipidemia risk prediction model to obtain the target face spectral samples of the population with hyperlipidemia and the corresponding first target contents of the four blood lipids.

[0039] Specifically, after inputting the face spectral samples into the pre-constructed hyperlipidemia risk prediction model, the model will divide the normal and abnormal intervals of the four blood lipids according to the contents (concentrations) of the four blood lipids in the blood of the person to be tested, and construct a data set of the target face spectral samples of the population with hyperlipidemia and the corresponding first target contents of the four blood lipids according to the face spectral samples.

[0040] In this embodiment, preferably, the construction of the hyperlipidemia risk prediction model includes: preprocessing the face video samples; establishing the corresponding relationship between the preprocessed face video samples and the standard contents of the four blood lipids in the hyperlipidemia risk determination standard; and training the model with reference to the corresponding relationship and the hyperlipidemia risk determination standard to obtain the hyperlipidemia risk prediction model.

[0041] It should be understood that if one or more of the four standard blood lipid contents in the risk determination criteria for hyperlipidemia are met, there is a risk of developing hyperlipidemia (abnormal blood lipids): total cholesterol (TC) ≥ 6.22 mmol / L, triglyceride (TG) ≥ 2.26 mmol / L, low-density lipoprotein cholesterol (LDL-c) ≥ 4.14 mmol / L, and high-density lipoprotein cholesterol (HDL-c) ≤ 1.04 mmol / L. If not, it is determined that the blood lipids are normal.

[0042] In this embodiment, in order to improve the quality of the facial spectral samples, after obtaining the facial spectral samples of the population to be tested and the four blood lipid contents, before inputting the facial spectral samples into a pre-constructed hyperlipidemia risk prediction model to obtain the target facial spectral samples of the population with hyperlipidemia and the corresponding first target four blood lipid contents, it further includes:

[0043] Preprocessing the facial spectral samples.

[0044] Specifically, preprocessing the facial spectral samples includes: using a moving average filtering method or a Savitzky-Golay filtering method to eliminate the noise in the facial spectral samples; using polynomial baseline correction or least squares baseline correction to correct the baseline drift of the facial spectral samples; and performing normalization processing on the facial spectral samples to scale them to the same range to obtain the preprocessed facial spectral samples.

[0045] Step S103: Make corresponding labels based on the target facial spectral samples, the corresponding first target four blood lipid contents, and the traditional Chinese medicine constitution categories, and train the model with the training samples obtained by making the corresponding labels to obtain a traditional Chinese medicine constitution identification model;

[0046] Training the model with the training samples obtained by making the corresponding labels to obtain a traditional Chinese medicine constitution identification model includes:

[0047] Adopting the Logistic regression analysis method, inputting the training samples obtained by making the corresponding labels into a model composed of multiple binary classification models for learning and training to construct a traditional Chinese medicine constitution identification model.

[0048] Specifically, the model construction steps are as follows:

[0049] First step: Use a non-linear sigmoid function for classification prediction, and its functional formula is:

[0050]

[0051] Let the feature vector affecting the prediction result be X(x 1 ,x 2 ...,x n), and the regression coefficients are θ (θ 0 , θ 1 , …, θ n ), then

[0052]

[0053] Construct the prediction function as:

[0054]

[0055] Step 2: Use the gradient descent algorithm to solve for θ, and the steps are as follows:

[0056] Let the number of training samples in the training dataset be m, then:

[0057] P(y|x; θ) = (h θ (x)) y (1 - h θ (x)) 1-y

[0058] Take the maximum likelihood function as:

[0059]

[0060] Take the logarithm of the likelihood function as:

[0061]

[0062] The maximum likelihood estimation finds the value of θ that maximizes the value of l(θ). Let the loss function J(θ) be:

[0063]

[0064] Because of multiplying by the coefficient -1 / m, so the θ when J(θ) takes the minimum value is the required optimal coefficient. The update process of θ for batch gradient descent to find the minimum value is:

[0065]

[0066] Among them: α is the learning rate. The result of taking the partial derivative of J(θ) is:

[0067]

[0068] The update process of batch gradient descent θ can be written as:

[0069]

[0070] Among them: y i -h θ (x i ) represents the deviation between the result value and the predicted value of each sample.

[0071] Step 3: Convert binary logistic regression to multi-class logistic regression, and set the result value of class i of the logistic regression coefficient θ i to 1, and set the result values of other classes to 0. Use the sample data to calculate θ i , so as to obtain the regression coefficient θ i for each class. If there are k classes (in this cardiovascular and cerebrovascular disease classification model, there are a total of 5 classes, that is, k = 5), there will be n groups of regression coefficients (θ 1 , θ 2 , …, θ k ). Substitute θ i into h θ (x), calculate each predicted value of x one by one, and take the class with the largest predicted value as the predicted result.

[0072] Step S104: Obtain the spectral sample of the face to be measured and input it into the traditional Chinese medicine constitution identification model to obtain the second target four blood lipid contents of hyperlipidemia in the tested population and the corresponding traditional Chinese medicine constitution identification results.

[0073] Specifically, that is, using the above regression classification model, quickly and accurately identify the traditional Chinese medicine constitution category associated with the face spectral sample data to be tested, and intelligently determine the correlation between hyperlipidemia, four blood lipid contents and traditional Chinese medicine constitution identification, which can provide sufficient support for the treatment and prevention of hyperlipidemia patients. After collecting and processing the data, standardized sample data are obtained to meet the input requirements of the model. These standard sample data can quickly obtain classification and identification results through the previously trained regression classification model.

[0074] For example, the TC mean of phlegm-dampness and damp-heat constitution was significantly higher than that of other constitutions (P<0.01), while the TC mean of yang deficiency constitution was significantly lower than that of other constitutions (P<0.01); the TG and LDL-c mean of blood stasis constitution and phlegm-dampness and damp-heat constitution was increased, significantly higher than that of other constitutions (P<0.01); the HDL-c mean of yang deficiency constitution was decreased, significantly lower than that of other constitutions (P<0.01). In addition, in the special group of primary dyslipidemia, when LDL-C was increased, the possibility of qi deficiency constitution and yang deficiency constitution was high. In addition, compared with the peaceful constitution, the biased constitution is correlated with the blood lipid level, among which TC, HDL-C and LDL-C are negatively correlated with the yang deficiency constitution; the phlegm-damp constitution, damp-heat constitution and blood stasis constitution have a greater impact on dyslipidemia, and are dangerous constitutions for dyslipidemia (P<0.05 or P<0.01), and the qi stagnation constitution is a protective constitution for dyslipidemia; the phlegm-damp constitution is positively correlated with serum triglycerides, and is a dangerous constitution for increased TG (P<0.05); the peaceful constitution, yang deficiency constitution and yin deficiency constitution are positively correlated with increased total cholesterol, and are dangerous constitutions for increased TC (P<0.05); the yin deficiency constitution is negatively correlated with serum HDL, and is a protective factor for reduced high-density lipoprotein (P<0.05).

[0075] From the above description, it can be seen that the present invention achieves the following technical effects:

[0076] In the embodiment of the present application, a method of TCM constitution identification is adopted, by obtaining facial spectral samples and four blood lipid contents of the tested population; the facial spectral samples are input into a pre-constructed hyperlipidemia risk prediction model to obtain target facial spectral samples of hyperlipidemia in the tested population and the first target four blood lipid contents corresponding thereto; corresponding labels are made based on the target facial spectral samples, the first target four blood lipid contents corresponding thereto, and the TCM constitution categories, and the model is trained by the training samples obtained by making the corresponding labels to obtain a TCM constitution identification model; the facial spectral samples of the tested population are obtained, and the TCM constitution is input into the pre-constructed hyperlipidemia risk prediction model to obtain the target facial spectral samples and the first target four blood lipid contents corresponding thereto; corresponding labels are made based on the target facial spectral samples, the first target four blood lipid contents corresponding thereto, and the TCM constitution categories, and the TCM constitution identification model is obtained by training the model with the training samples obtained by making the corresponding labels ... TCM constitution identification model is obtained by The TCM constitution identification model was established to obtain the second target four blood lipid contents of hyperlipidemia in the tested population and the corresponding TCM constitution identification results; the purpose of realizing TCM constitution identification based on human face and determining the correlation between hyperlipidemia, four blood lipid contents and TCM constitution identification was achieved, thereby achieving the technical effect of accurately providing sufficient support for the treatment and prevention of patients with lipidemia, and further solving the technical problem of being unable to accurately provide sufficient support for the treatment and prevention of patients with lipidemia due to the inability to realize TCM constitution identification based on human face and not exploring the correlation between hyperlipidemia, four blood lipid contents and TCM constitution identification.

[0077] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0078] According to an embodiment of the present invention, there is also provided a device for implementing the above-mentioned traditional Chinese medicine constitution identification method, as Figure 2 shown, the device includes:

[0079] An acquisition module 10, configured to acquire a face spectral sample and the content of four blood lipids of the tested population;

[0080] Based on near-infrared imaging technology, scan the face of the tested person and collect the face spectral sample, and obtain the content (concentration) of four blood lipids in the blood of the tested person.

[0081] Specifically, acquiring the face spectral sample of the tested population includes: scanning the face image of the tested population based on near-infrared imaging technology, using the Dlib model to detect the face target area in the face image, and generating a face detection box box(x 1 , x 2 , x 3 , x 4 ), where x i is the tuple of pixel point i, representing the pixel values of the row and column where the pixel point is located; crop the face core area according to the face detection box, and perform average processing and conversion on at least one scan data of the core area to obtain a face spectral sample.

[0082] Specifically, acquiring the content of four blood lipids of the tested population includes: using a blood detection method to detect the blood lipid content in the blood of the tested population to obtain the content of four blood lipids; wherein, the content of four blood lipids at least includes total cholesterol Tc, triglyceride TG, high-density lipoprotein HDL-C, and low-density lipoprotein LDL-C.

[0083] By using face detection technology combined with processing and conversion, a face spectral sample is obtained from the face image, providing data guarantee for subsequent traditional Chinese medicine constitution identification based on the face.

[0084] A prediction module 20, configured to input the face spectral sample into a pre-constructed hyperlipidemia risk prediction model to obtain a target face spectral sample of the tested population suffering from hyperlipidemia and the corresponding first target content of four blood lipids;

[0085] Specifically, after inputting the face spectral sample into the pre-constructed hyperlipidemia risk prediction model, the model will divide the normal and abnormal intervals of the four blood lipids according to the content (concentration) of the four blood lipids in the blood of the tested person, and construct a dataset of the target face spectral sample of the tested population suffering from hyperlipidemia and the corresponding first target content of four blood lipids according to the face spectral sample.

[0086] In this embodiment, preferably, the construction of the hyperlipidemia risk prediction model includes: preprocessing the face video samples; establishing the corresponding relationship between the preprocessed face video samples and the standard four blood lipid contents in the hyperlipidemia risk determination criteria; and training the model with reference to the corresponding relationship and the hyperlipidemia risk determination criteria to obtain the hyperlipidemia risk prediction model.

[0087] It should be noted that if one or more of the standard four blood lipid contents in the hyperlipidemia risk determination criteria are met, there is a risk of suffering from hyperlipidemia (abnormal blood lipids): total cholesterol (TC) ≥ 6.22 mmol / L, triglyceride (TG) ≥ 2.26 mmol / L, low-density lipoprotein cholesterol (LDL-c) ≥ 4.14 mmol / L, high-density lipoprotein cholesterol (HDL-c) ≤ 1.04 mmol / L. If not, it is determined that the blood lipids are normal.

[0088] In this embodiment, in order to improve the quality of the face spectral samples, after obtaining the face spectral samples and the four blood lipid contents of the tested population, before inputting the face spectral samples into the pre-constructed hyperlipidemia risk prediction model to obtain the target face spectral samples with hyperlipidemia in the tested population and the corresponding first target four blood lipid contents, it further includes:

[0089] Preprocessing the face spectral samples.

[0090] Specifically, preprocessing the face spectral samples includes: using the moving average filtering method or the Savitzky-Golay filtering method to eliminate the noise in the face spectral samples; using polynomial baseline correction or least squares baseline correction to correct the baseline drift of the face spectral samples; and normalizing the face spectral samples and scaling them to the same range to obtain the preprocessed face spectral samples.

[0091] The training module 30 is used to make corresponding labels based on the target face spectral samples, the corresponding first target four blood lipid contents and the traditional Chinese medicine constitution categories, and train the model with the training samples obtained by making the corresponding labels to obtain the traditional Chinese medicine constitution identification model;

[0092] Training the model with the training samples obtained by making the corresponding labels to obtain the traditional Chinese medicine constitution identification model includes:

[0093] Adopting the Logistic regression analysis method, inputting the training samples obtained by making the corresponding labels into a model composed of multiple binary classification models for learning and training to construct the traditional Chinese medicine constitution identification model.

[0094] Specifically, the model construction steps are as follows:

[0095] Step 1: Use the non-linear sigmoid function for classification prediction. Its functional form is:

[0096]

[0097] Let the feature vector affecting the prediction result be X(x 1 , x 2 ..., x n ), and the regression coefficients be θ(θ 0 , θ 1 , …, θ n ). Then

[0098]

[0099] Construct the prediction function as:

[0100]

[0101] Step 2: Use the gradient descent algorithm to solve for θ. The steps are as follows:

[0102] Let the number of training samples in the training dataset be m. Then:

[0103] P(y|x; θ) = (h θ (x)) y (1 - h θ (x)) 1-y

[0104] Take the maximum likelihood function as:

[0105]

[0106] Take the logarithm of the likelihood function as:

[0107]

[0108] The maximum likelihood estimation finds the value of θ that maximizes the value of l(θ). Let the loss function J(θ) be:

[0109]

[0110] Because of the multiplication by the coefficient -1 / m, the θ when J(θ) takes the minimum value is the required optimal coefficient. The update process of θ for finding the minimum value by the batch gradient descent method is:

[0111]

[0112] Where: α is the learning rate. The result of taking the partial derivative of J(θ) is:

[0113]

[0114] The update process of the batch gradient descent θ can be written as:

[0115]

[0116] where: y i -h θ (x i ) represents the deviation between the result value and the predicted value of each sample.

[0117] Step 3: Convert the binary logistic regression to a multi-class logistic regression. Set the result value of class i of the logistic regression coefficient θ i to 1, and set the result values of other classes to 0. Use the sample data to calculate θ i , so as to obtain the regression coefficient θ i for each class. If there are k classes (in this cardio-cerebrovascular disease classification model, there are 5 classes in total, that is, k = 5), there will be n groups of regression coefficients (θ 1 , θ 2 , …, θ k ). Substitute θ i into h θ (x), calculate each predicted value of x one by one, and select the class with the largest predicted value as the predicted result.

[0118] The identification module 40 is used to obtain the spectral samples of the face to be tested and input them into the traditional Chinese medicine constitution identification model to obtain the second target lipid four-item content of hyperlipidemia in the tested population and the corresponding traditional Chinese medicine constitution identification result.

[0119] Specifically, that is, using the above regression classification model, quickly and accurately identify the traditional Chinese medicine constitution category associated with the face spectral sample data to be tested, intelligently determine the correlation between hyperlipidemia, lipid four-item content and traditional Chinese medicine constitution identification, and can provide sufficient support for the treatment and prevention of hyperlipidemia patients. After collecting and processing the data, standardized sample data are obtained to meet the input requirements of the model. These standard sample data can quickly obtain the classification and identification results through the previously trained regression classification model.

[0120] For example, the TC mean of phlegm-dampness and damp-heat constitution was significantly higher than that of other constitutions (P<0.01), while the TC mean of yang deficiency constitution was significantly lower than that of other constitutions (P<0.01); the TG and LDL-c mean of blood stasis constitution and phlegm-dampness and damp-heat constitution was increased, significantly higher than that of other constitutions (P<0.01); the HDL-c mean of yang deficiency constitution was decreased, significantly lower than that of other constitutions (P<0.01). In addition, in the special group of primary dyslipidemia, when LDL-C was increased, the possibility of qi deficiency constitution and yang deficiency constitution was high. In addition, compared with the peaceful constitution, the biased constitution is correlated with the blood lipid level, among which TC, HDL-C and LDL-C are negatively correlated with the yang deficiency constitution; the phlegm-damp constitution, damp-heat constitution and blood stasis constitution have a greater impact on dyslipidemia, and are dangerous constitutions for dyslipidemia (P<0.05 or P<0.01), and the qi stagnation constitution is a protective constitution for dyslipidemia; the phlegm-damp constitution is positively correlated with serum triglycerides, and is a dangerous constitution for increased TG (P<0.05); the peaceful constitution, yang deficiency constitution and yin deficiency constitution are positively correlated with increased total cholesterol, and are dangerous constitutions for increased TC (P<0.05); the yin deficiency constitution is negatively correlated with serum HDL, and is a protective factor for reduced high-density lipoprotein (P<0.05).

[0121] From the above description, it can be seen that the present invention achieves the following technical effects:

[0122] In the embodiment of the present application, a method of TCM constitution identification is adopted, by obtaining facial spectral samples and four blood lipid contents of the tested population; the facial spectral samples are input into a pre-constructed hyperlipidemia risk prediction model to obtain target facial spectral samples of hyperlipidemia in the tested population and the first target four blood lipid contents corresponding thereto; corresponding labels are made based on the target facial spectral samples, the first target four blood lipid contents corresponding thereto, and the TCM constitution categories, and the model is trained by the training samples obtained by making the corresponding labels to obtain a TCM constitution identification model; the facial spectral samples of the tested population are obtained, and the TCM constitution is input into the pre-constructed hyperlipidemia risk prediction model to obtain the target facial spectral samples and the first target four blood lipid contents corresponding thereto; corresponding labels are made based on the target facial spectral samples, the first target four blood lipid contents corresponding thereto, and the TCM constitution categories, and the TCM constitution identification model is obtained by training the model with the training samples obtained by making the corresponding labels ... TCM constitution identification model is obtained by The TCM constitution identification model was established to obtain the second target four blood lipid contents of hyperlipidemia in the tested population and the corresponding TCM constitution identification results; the purpose of realizing TCM constitution identification based on human face and determining the correlation between hyperlipidemia, four blood lipid contents and TCM constitution identification was achieved, thereby achieving the technical effect of accurately providing sufficient support for the treatment and prevention of patients with lipidemia, and further solving the technical problem of being unable to accurately provide sufficient support for the treatment and prevention of patients with lipidemia due to the inability to realize TCM constitution identification based on human face and not exploring the correlation between hyperlipidemia, four blood lipid contents and TCM constitution identification.

[0123] Obviously, those skilled in the art should understand that the above-mentioned modules or steps of the present invention can be implemented by a general-purpose computing device. They can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Optionally, they can be implemented by program codes executable by the computing device. Thus, they can be stored in a storage device and executed by the computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module to implement. In this way, the present invention is not limited to any specific combination of hardware and software.

[0124] The above are only the preferred embodiments of the present application and are not intended to limit the present application. For those skilled in the art, various changes and modifications can be made to the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included within the protection scope of the present application.

Claims

1. A method for identifying constitution in traditional Chinese medicine, characterized in that: include: Obtain facial spectrum samples and four blood lipid levels of the tested population; Inputting the face spectrum sample into a pre-built hyperlipidemia risk prediction model to obtain a target face spectrum sample of a person suffering from hyperlipidemia in the tested population and the first target four blood lipid contents corresponding thereto; Based on the target face spectrum sample, the first target four blood lipid contents and the TCM constitution category corresponding thereto, corresponding labels are made, and the model is trained by the training samples obtained by making corresponding labels to obtain a TCM constitution recognition model; The spectrum sample of the face to be tested is obtained and input into the TCM constitution identification model to obtain the second target blood lipid four items content of the tested population suffering from hyperlipidemia and the corresponding TCM constitution identification result.

2. The method for identifying constitution in traditional Chinese medicine according to claim 1, characterized in that: Obtain facial spectrum samples of the tested population, including: Based on near-infrared camera technology, the face images of the tested people are scanned, and the Dlib model is used to detect the face target area in the face image to generate a face detection box (x1, x2, x3, x4), where x i is a tuple of pixel i, indicating the row and column pixel value of the pixel; The core area of ​​the face is cropped according to the face detection frame, and at least one scanning data of the core area is averaged and converted to obtain a face spectrum sample.

3. The method for identifying constitution in traditional Chinese medicine according to claim 1, characterized in that: Obtain four blood lipid levels of the tested population, including: The blood lipid content in the blood of the tested population is tested using a blood test method to obtain four blood lipid contents; wherein the four blood lipid contents at least include total cholesterol Tc, triglycerides TG, high-density lipoprotein HDL-C, and low-density lipoprotein LDL-C.

4. The method for identifying constitution in traditional Chinese medicine according to claim 1, characterized in that: The construction of the hyperlipidemia risk prediction model includes: Preprocessing the face video sample; Establish the correspondence between the pre-processed face video samples and the four standard blood lipid levels in the hyperlipidemia risk assessment criteria; Model training is performed with reference to the corresponding relationship and the hyperlipidemia risk determination standard to obtain a hyperlipidemia risk prediction model.

5. The method for identifying constitution in traditional Chinese medicine according to claim 1, characterized in that: After obtaining the facial spectrum samples and the four blood lipid contents of the tested population, the facial spectrum samples are input into a pre-built hyperlipidemia risk prediction model to obtain the target facial spectrum samples of hyperlipidemia in the tested population and the first target four blood lipid contents corresponding thereto, and the method further includes: The face spectrum sample is preprocessed.

6. The method for identifying TCM constitution according to claim 5, characterized in that: Preprocessing the face spectrum sample includes: Using a sliding average filter method or a Savitzky-Golay filter method to eliminate noise in the face spectrum sample; Correcting the baseline drift of the face spectrum sample by using polynomial baseline correction or least squares baseline correction; The face spectrum samples are normalized and scaled to the same range to obtain preprocessed face spectrum samples.

7. The method for identifying constitution in traditional Chinese medicine according to claim 1, characterized in that: The model is trained by training samples obtained by corresponding labels to obtain a TCM constitution identification model, including: The Logistic regression analysis method is used to input the training samples obtained by corresponding labels into a model composed of multiple binary classification models for learning and training, and a TCM constitution identification model is constructed.

8. A TCM constitution identification device, characterized in that: include: An acquisition module is used to obtain facial spectrum samples and four blood lipid contents of the tested population; A prediction module, used for inputting the face spectrum sample into a pre-built hyperlipidemia risk prediction model to obtain a target face spectrum sample of a person suffering from hyperlipidemia in the tested population and the first target four blood lipid contents corresponding thereto; A training module is used to make corresponding labels based on the target face spectrum sample, the first target four blood lipid contents and the TCM constitution category corresponding thereto, and train the model through the training samples obtained by making corresponding labels to obtain a TCM constitution recognition model; The recognition module is used to obtain the face spectrum sample to be tested and input it into the TCM constitution recognition model to obtain the second target blood lipid four items content of the tested population suffering from hyperlipidemia and the corresponding TCM constitution recognition results.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program is configured to execute the TCM constitution identification method according to any one of claims 1 to 7 when running.

10. A server, comprising: A memory and a processor, characterized in that a computer program is stored in the memory, wherein the processor is configured to run the computer program to execute the TCM constitution identification method according to any one of claims 1 to 7.

Images