CN110403611B - Method, device, computer equipment and storage medium for predicting glycated hemoglobin composition value in blood - Google Patents

Abstract

The present application relates to a method, system, computer equipment and storage medium for predicting the value of glycated hemoglobin in blood. The method includes: acquiring an image of the tongue image to be tested and data of the consultation to be tested of the individual to be tested; inputting the image of the tongue image to be tested and the data of the consultation to be tested of the individual to be tested into a prediction obtained by machine learning a model, and the predicted value of the target blood component of the individual to be tested is obtained through analysis based on the prediction model. The prediction model obtained through the training of the machine learning model makes it possible to obtain the corresponding target blood component prediction value according to the tongue image of the human body and the consultation data, so as to realize the rapid and automatic acquisition of the blood component value, which is beneficial to improve the disease diagnosis efficiency of patients. , has the characteristics of low price, fast and convenient application.

Description

Method, device, computer equipment and storage medium for predicting the value of glycated hemoglobin in blood

technical field

The present application relates to the technical field of blood component value prediction, and in particular, to a method, device, computer equipment and storage medium for predicting the value of glycated hemoglobin in blood.

Background technique

In modern society, people pay more and more attention to health, and ordinary people also hope to know their health status anytime, anywhere, simply, conveniently and cheaply. Various smartphones and health bracelets on the market cater to such needs to some extent. But to gain an in-depth understanding of human health, laboratory data, especially blood component values, are still needed. There are many types of blood components, such as hemoglobin, white blood cells, glycosylated hemoglobin, creatinine, erythrocyte sedimentation rate and so on. The acquisition of blood values, generally by drawing blood for analysis, is a traumatic operation. The analysis also requires related instruments and consumables, which involves a certain cost.

Traditional Chinese medicine has a history of thousands of years in China, and has rich clinical experience in the fields of disease treatment, disease prevention and health preservation. Among them, the diagnosis and treatment methods of traditional Chinese medicine are known for their simplicity, convenience and low cost. "The Yellow Emperor's Classic of Internal Medicine" says: "There are all kinds of shapes in the inside, and they must be in the outside." That is to say, if there is something wrong in the human body, it will definitely appear on the surface of the body. Traditional Chinese medicine uses the four diagnostic methods of looking, smelling, asking, and cutting to collect the manifestations of diseases and understand the health of the human body. Among them, tongue diagnosis and questioning are an important part of the four diagnosis. However, at present, the traditional Chinese medicine four-diagnosis instruments such as those described in CN1561904A or WO2011026305A1 are mainly aimed at collecting the four-diagnosis data. Even if some instruments can output the deduction results of the four-diagnosis data, they are only limited to the diagnosis results of traditional Chinese medicine. , and none of them have the function of predicting blood composition. Furthermore, such instruments are also specially designed, expensive, and only used by professionals, and are difficult to popularize to the general public.

Various studies have revealed that there is a close relationship between tongue image and changes in some components of blood. For example, red blood cell count (RBC) and hemoglobin (HGB) are associated with darker tongue color; decreased hematocrit (HCT) is associated with pale tongue and thin tongue; elevated serum creatinine in renal failure is associated with a variety of Tongue characteristics.

As another example, diabetes is a disease in which the blood sugar concentration is abnormal. Some people have studied the diagnosis of diabetes by tongue images according to the principle of tongue diagnosis in traditional Chinese medicine. But the method used requires special equipment. Such diagnostic methods are not easy to popularize and involve expensive purchase of equipment.

SUMMARY OF THE INVENTION

Based on this, it is necessary to provide a method, device, computer equipment and storage medium for predicting the value of glycated hemoglobin in blood in response to the above technical problems.

A method for predicting a glycated hemoglobin component value in blood, the method comprising:

Obtain the tongue image to be tested and the consultation data to be tested of the individual to be tested;

The image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into the prediction model obtained by training, and the predicted value of the target blood component of the individual to be tested is obtained through analysis based on the prediction model.

In one of the embodiments, before the step of acquiring the tongue image image to be tested and the consultation data to be tested of the individual to be tested, the step further includes:

Collect target blood component data;

collecting consultation data and pre-collected tongue images corresponding to the target blood component data;

Preprocessing the target blood component data, the pre-collected tongue image and the consultation data;

Based on the machine learning model, the pre-collected tongue image and the interview data are parsed, the tongue characteristics and the interview data of the pre-collected tongue image are obtained, and the tongue characteristics of the pre-collected tongue image are obtained , the characteristics of the interview data and the target blood component data are trained to obtain the prediction model.

In one embodiment, the to-be-measured tongue image and the to-be-measured consultation data of the to-be-measured individual are input into a prediction model obtained by training, and the to-be-tested model is parsed and obtained based on the prediction model. The steps of measuring the predicted value of the target blood component of the individual include:

inputting the tongue image image to be tested and the consultation data to be tested of the individual to be tested into the prediction model obtained by training;

Based on the prediction model, analyze the tongue image to be tested and the consultation data to be tested, and obtain the tongue image features of the tongue image to be tested and the features of the consultation data to be tested;

According to the characteristics of the tongue image of the tongue image to be measured and the characteristics of the consultation data to be measured, the predicted value of the target blood component of the individual to be measured is obtained.

In one of the embodiments, the step of obtaining the predicted value of the target blood component of the individual to be measured according to the characteristics of the tongue image of the tongue image to be measured and the characteristics of the consultation data to be measured includes:

Determine the tongue features and the data features of the pre-collected tongue images that are closest to the features of the tongue images to be tested and the features of the data to be tested, and obtain the target of the individual to be tested Predictive value of blood components.

In one embodiment, the step of collecting the consultation data corresponding to the target blood component data and pre-collecting the tongue image includes:

obtaining the consultation data corresponding to the target blood component data;

Get face image;

Parse the face image to obtain the pre-collected tongue image.

A device for predicting the value of glycated hemoglobin in blood, the device comprising:

The data acquisition module to be tested is used to acquire the tongue image to be tested and the medical consultation data to be tested of the individual to be tested;

A predicted value obtaining module, used to input the image of the tongue to be tested and the data of the consultation to be tested of the individual to be tested into the prediction model obtained by training, and to analyze and obtain the individual to be tested based on the prediction model The predicted value of the target blood composition.

In one embodiment, it also includes:

A blood component data acquisition module for collecting target blood component data;

a collection module, used for collecting the consultation data and pre-collected tongue images corresponding to the target blood component data;

a model input module, configured to preprocess the target blood component data, the pre-collected tongue image and the interrogation data;

A prediction model obtaining module is used to analyze the pre-collected tongue image and the interview data based on the machine learning model, and obtain the tongue characteristics and interview data characteristics of the pre-collected tongue image, according to the pre-collection The prediction model is obtained by training the tongue features of the tongue image, the features of the consultation data and the target blood component data.

In one embodiment, the predicted value obtaining module includes:

a data input unit to be tested, configured to input the tongue image image to be tested and the medical consultation data to be tested of the individual to be tested into the prediction model obtained by training;

A feature analysis and obtaining unit, configured to analyze the tongue image to be tested and the consultation data to be tested based on the prediction model, and obtain tongue features of the tongue image to be tested and the consultation data to be tested feature;

A predicted value obtaining unit, configured to obtain the predicted value of the target blood component of the individual to be measured according to the tongue image feature of the tongue image image to be tested and the characteristics of the medical consultation data to be tested.

A computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

Obtain the tongue image to be tested and the consultation data to be tested of the individual to be tested;

The image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into the prediction model obtained by training, and the predicted value of the target blood component of the individual to be tested is obtained through analysis based on the prediction model.

A computer-readable storage medium on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

Obtain the tongue image to be tested and the consultation data to be tested of the individual to be tested;

The image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into the prediction model obtained by training, and the predicted value of the target blood component of the individual to be tested is obtained through analysis based on the prediction model.

The above-mentioned method, device, computer equipment and storage medium for predicting the value of glycated hemoglobin in blood, through the machine learning model, analyzing the tongue image to obtain the characteristics of the tongue image and the characteristics of the consultation data, and establishing the prediction model through the learning of big data, so that it can be based on the human body. The tongue image and consultation data are obtained by analyzing the characteristics of the tongue image and the data of the consultation, and then obtaining the corresponding predicted value of the target blood component, without the need to perform blood test on the individual, so as to realize the rapid and automatic acquisition of the blood component value. It is beneficial to improve the disease diagnosis efficiency of patients, and has the characteristics of low price, quick and convenient application.

Description of drawings

1 is an application scenario diagram of a method for predicting the value of glycated hemoglobin composition in blood in one embodiment;

FIG. 2A is a schematic flowchart of a training acquisition of a prediction model in one embodiment;

2B is a schematic flowchart of a method for predicting the value of glycated hemoglobin in blood in one embodiment;

2C is a schematic flowchart of a training process of a machine learning-based model and a prediction process of the model in one embodiment;

3 is a structural block diagram of a device for predicting the value of glycated hemoglobin in blood in one embodiment;

Fig. 4 is the internal structure diagram of the computer device in one embodiment;

5 is an internal structure diagram of a computer device in another embodiment;

FIG. 6 is a schematic diagram of an inquiry questionnaire in one embodiment;

FIG. 7 is a schematic diagram of the features of consultation data in one embodiment;

8 is a schematic diagram of an original tongue image according to an embodiment;

9 is a schematic diagram of a processed tongue image according to one embodiment;

Figure 10 is the result of testing multiple machine classifiers in one embodiment;

11 is a comparison result diagram of a target blood component predicted value and an actual blood component value obtained by analysis in one embodiment;

Figure 12 is a normalized confusion matrix diagram in one embodiment.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present application more clearly understood, the present application will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application, but not to limit the present application.

The method for predicting the glycated hemoglobin composition value in blood provided by the present application can be applied to the application environment shown in FIG. 1 . The computer 102 communicates with the server 104 through the network through the network. The computer 102 can be, but is not limited to, various personal computers, notebook computers, smart phones, tablet computers and portable wearable devices, and the server 104 can be implemented by an independent server or a server cluster composed of multiple servers. The terminal acquires the target blood component data, pre-collected tongue image and consultation data, and inputs the above-mentioned target blood component data, pre-collected tongue image and consultation data as training data into the machine learning model of the server 104, and the server 104 is based on the machine learning model. Learning the model, parsing the pre-collected tongue image and the interview data, acquiring the tongue characteristics and interview data characteristics of the pre-collected tongue image, and establishing a prediction model.

For example, a method for predicting the value of glycated hemoglobin in blood, the method includes: collecting target blood component data; collecting consultation data and pre-collected tongue images corresponding to the target blood component data; Data, the pre-collected tongue image and the interview data are preprocessed; based on the machine learning model, the pre-collected tongue image and the interview data are parsed to obtain the tongue image of the pre-collected tongue image The characteristics and the characteristics of the consultation data are used to establish a prediction model.

In the above-mentioned embodiment, through the machine learning model, the tongue image is analyzed to obtain the characteristics of the tongue image and the characteristics of the consultation data, and the prediction model is established through the learning of big data, so that the tongue image can be parsed according to the tongue image and the consultation data of the human body. Characteristics and data characteristics of the consultation data, and then obtain the corresponding predicted value of the target blood component, so as to realize the rapid and automatic acquisition of the blood component value, which is beneficial to improve the disease diagnosis efficiency of patients, and has the characteristics of low price and fast and convenient application.

In one embodiment, as shown in FIG. 2A , a method for predicting the value of glycated hemoglobin in blood is provided, and the method includes the following steps:

Step 202, collect target blood component data.

For example, acquire target blood component data. Specifically, the target blood component data is the blood detection data of the human body, for example, the target blood component data is the blood component data of the human body. For example, the target detection data is the glycated hemoglobin data of the human body. The target blood component data is collected and acquired in advance and input into the computer.

Step 204 , collect consultation data and pre-collected tongue images corresponding to the target blood component data.

For example, the consultation data and the pre-collected tongue image corresponding to the target blood component data are acquired. For example, the consultation data corresponding to the target blood component data is acquired, and the pre-collected tongue image corresponding to the target blood component data is acquired. That is, the individual corresponding to the consultation data and the pre-collected tongue image and the individual corresponding to the target blood component data are the same individual.

Specifically, the consultation data is medical record data of the human body, for example, the consultation data is disease data of the human body, for example, the consultation data is disease data of the human body, for example, the consultation data is personal health data. The consultation data records the symptoms of the user's human body, and the consultation data further includes the user's disease name, the user's age, the user's height, the user's weight, and the like. For example, the consultation data is obtained by user input. For example, to obtain the data of the consultation entered into the computer. For example, to obtain the consultation data input on the consultation questionnaire, in this embodiment, by presetting the questions of the consultation questionnaire, the user can directly answer the questions on the consultation questionnaire, and then obtain the user's consultation data , which can effectively improve the efficiency and accuracy of data acquisition.

For example, the consultation data includes the name of the disease suffered by the user, the age of the user, the height of the user, the weight of the user, the subjective symptoms of the user, and the like. In this embodiment, by presetting the questions of the consultation questionnaire based on medical knowledge, the user can directly answer the questions on the consultation questionnaire, and then obtain the consultation data related to the user and the target blood components, which effectively improves the consultation data. acquisition efficiency and accuracy.

The pre-collected tongue image is a pre-collected tongue image, and the tongue image is an image of the human tongue. The representation of the human tongue is closely related to the changes of some components in the human blood, and the components of the human blood are Changes will cause changes in the physical state of the human body, and changes in the physical state of the human body will also lead to changes in the blood composition of the human body. For example, the characterization of the human tongue is related to the changes of glycosylated hemoglobin in human blood. Glycated hemoglobin is the product of the combination of hemoglobin in red blood cells and sugars in serum. It is formed by a slow, continuous and irreversible glycation reaction, and its content depends on the blood sugar concentration and the contact time between blood sugar and hemoglobin. Therefore, the glycated hemoglobin value can effectively reflect the blood sugar control of an individual in the past 1 to 2 months, so it is used as a monitoring index for diabetes control. Because the blood sugar concentration is closely related to the change of the glycated hemoglobin value, and the change of the blood sugar concentration will also cause the change of the human body's physical state. Therefore, different glycated hemoglobin values correspond to different tongue characteristics.

Therefore, if there is a disease or a change in the state of the human body, it will be reflected through the representation of the tongue image. That is to say, changes in the composition of human blood will change the representation of the human tongue, and different components of human blood correspond to different representations of the tongue. For example, pre-collected tongue images are acquired by an image sensor.

In this embodiment, the individual is the detection user, and each individual target blood component data corresponds to a consultation data and a pre-collected tongue image.

For example, acquire multiple target blood component data, multiple consultation data, and multiple pre-collected tongue images, and the multiple target blood component data, multiple consultation data, and multiple pre-collected tongue images are in a one-to-one correspondence . Multiple pre-collected images have different representations, and each consultation data is also different, therefore, different representations will correspond to different consultation data, corresponding to the blood component data of different individuals, that is, a target blood component The data corresponds to a consultation data and a pre-collected tongue image.

In this embodiment, obtaining the consultation data is beneficial to further improve the accuracy of parsing the tongue image.

For example, a plurality of consultation data are acquired, for example, each consultation data corresponds to a target blood component data and a pre-collected tongue image. Specifically, the consultation data of the same user corresponds to the target blood component data of the user and a pre-collected tongue image. It is beneficial to improve the accuracy of data analysis.

For example, acquiring the glycated hemoglobin values of several individuals, the interrogation data of several individuals, and the pre-collected tongue images of several individuals, the glycated hemoglobin values of several individuals, the interrogation data of several individuals, and the pre-collected tongue images of several individuals are: One-to-one correspondence. The multiple pre-collected images have different representations and each glycated hemoglobin value is different, so different characterizations will correspond to different glycated hemoglobin values. The corresponding glycated hemoglobin values and pre-collected tongue images correspond to the same individual.

Step 206, preprocessing the target blood component data, the pre-collected tongue image and the consultation data.

For example, after preprocessing, the preprocessed target blood component data, the pre-collected tongue image and the interview data are input into a machine learning model as training data. For example, a machine learning model can be a single model, such as a decision tree model, a linear regression model, a neural network model, etc., or a combination of two or more models.

Specifically, the target blood component data, the pre-collected tongue image and the consultation data are integrated into a data matrix, and the data matrix is input into a machine learning model.

For example, the preprocessing of pre-collected tongue images includes: uniformly organizing the captured tongue images into 512x512 pixel maps, training through machine learning, performing image interception processing with the method of convolutional network, and excluding the images of the outer periphery of the tongue image. to get a new tongue image. In this new tongue image, the pixel values that do not belong to the tongue and its surrounding tissues are set to zero. Then, the new tongue image is converted into a feature vector with a certain length by the following formula. The mathematical model equations used are as follows:

，其中，dst(x,y)为特征向量，src 为转化函数，f_x为图像，f_y是向量的最终维度。

, where dst(x,y) is the feature vector, src is the transformation function, f _x is the image, and f _y is the final dimension of the vector.

For another example, in some embodiments, some tongue image features can be selected as feature vectors instead of tongue image pixel data, for example, the color of the tongue image is used as a feature, or the shape of the tongue is used as a feature, or the The texture of the tongue tissue serves as a feature. The mathematical equations for calculating these features can be referred to using existing published related methods [(Zhang, D., Zhang, H., & Zhang, B.). Tongue Image Analysis (pp. 1-335). Springer. 2017 ].

。使用图像特征代替舌象像素数 据作为特征向量，有助提高预测的准确性。 Depending on the class of tongue image features used, vectors with different dimensions will be obtained. A new feature vector V can be generated by combining various features and indicators.

. Using image features instead of tongue pixel data as feature vectors can help improve the accuracy of prediction.

After converting each of the intercepted tongue images or selected tongue image features into vectors, import the corresponding consultation results, and use binary expansion or one-hot according to different results. code (one hot encoding) to generate different values, creating a final vector containing tongue image feature results and interview results.

The generated data is associated with the corresponding object detection value. For all images and consultation results of the pre-collected data, repeat the above process to establish a training data set. After that, the pre-collected data set is divided into training samples and test samples, the training samples are input into the machine learning model, and the test samples are used to measure the accuracy of the model.

Step 208, based on the machine learning model, parse the pre-collected tongue image and the interview data, obtain the tongue characteristics and interview data characteristics of the pre-collected tongue image, The prediction model is obtained by training on tongue image features, interview data features and target blood component data.

For example, based on the machine learning model, parse the pre-collected tongue image to obtain the tongue characteristics of the pre-collected tongue image, parse the interview data, and obtain the correspondence between the characteristics of the interview data and the target detection value, according to the The prediction model is obtained by learning from the tongue characteristics, interrogation data characteristics and target blood component data of the pre-collected tongue images.

Specifically, the tongue image feature is the representation of the tongue image, or in other words, the tongue image feature is the representation of the tongue image. The feature of the consultation data is the vectorized value of the consultation data.

Specifically, since the characteristics of the tongue image are related to the blood composition of the human body, changes in the blood composition of the human body will cause changes in the characteristics of the tongue image. Different tongue image characteristics correspond to different blood components of the human body. Different target blood component data.

In addition, there is a one-to-one correspondence between the consultation data and the target blood component data. Different blood component data will lead to different individual consultation data. Therefore, the consultation data is associated with the blood component data. Therefore, in this step, the Pre-collected tongue images and interrogation data are analyzed to obtain multiple tongue characteristics and interrogation data features of the pre-collected tongue images. Based on the machine learning method, the tongue characteristics and questions of each pre-collected tongue image are learned and obtained. A prediction model for the correspondence between diagnostic data features and target blood component data. For example, according to the correspondence between the pre-collected tongue images and the characteristics of the interview data and the corresponding target blood component data, learn to obtain the tongue characteristics of each pre-collected tongue image and the corresponding relationship between the interview data and the target blood component data. , so as to build a predictive model.

In this embodiment, after acquiring the target blood component data, the corresponding relationship between the target blood component data and the pre-collected tongue image and the consultation data is established, so that the target blood component data and the pre-collected tongue image and the consultation data can be Then, the changes of the tongue characteristics of the pre-collected tongue images and the relationship between the interrogation data and the target blood component data are obtained, and the tongue characteristics of the pre-collected tongue images and the change rules between the characteristics of the interrogation data and the target blood composition data are obtained. , and establish a prediction model based on the tongue image features of pre-collected tongue images and the relationship between the characteristics of the consultation data and the target blood component data, so that the acquisition of the predicted value of the target blood component is more accurate.

In this embodiment, based on a large number of pre-collected tongue images, interrogation data and target blood component data, big data analysis and learning are performed, and then the tongue characteristics of the pre-collected tongue images, the characteristics of the interrogation data and the target blood are obtained by learning. According to the corresponding rule of the component data, a prediction model based on the tongue characteristics of the pre-collected tongue image and the corresponding relationship between the characteristics of the consultation data and the target blood component data is established. For example, the corresponding relationship can be visualized as a graph, for example, the corresponding relationship between the tongue feature of the pre-collected tongue image and the target blood component data is mapped as a curve. By establishing the correspondence between the tongue image features of the pre-collected tongue image images, in the subsequent detection, the target blood component prediction value can be obtained through the tongue image image and the interrogation data, which is beneficial to assist the diagnosis of diseases.

For example, based on the machine learning model, the pre-collected tongue images are parsed, the tongue features of the pre-collected tongue images are acquired, and the relationship and correspondence between the tongue features of the plurality of pre-collected tongue images are obtained. The change relationship of a plurality of consultation data is obtained, the effective tongue image features are obtained, and the corresponding relationship between the effective tongue image features of the pre-collected tongue image images and the consultation data is established.

The effective tongue image feature is one of multiple features in the tongue image. In this embodiment, the effective tongue image feature is the tongue image feature related to the change of the interrogation data in the pre-collected tongue image. Therefore, through the tongue image The change relationship of the characteristics, combined with multiple consultation data, obtain the effective tongue characteristics, establish the corresponding relationship between the effective tongue characteristics of the pre-collected tongue images and the interview data, so that the tongue characteristics and the question The establishment of the corresponding relationship of the diagnostic data is more accurate, which is beneficial to improve the accuracy of obtaining the predicted value of the target blood component.

For example, based on the machine learning model, parse a plurality of the pre-collected tongue images and interview data, obtain the tongue characteristics of each of the pre-collected tongue images and the characteristics of each interview data, and establish each of the pre-collected tongue images. A prediction model is established based on the tongue characteristics of the tongue image and the correspondence between the characteristics of each consultation data and the target blood component data.

In this embodiment, by separately analyzing multiple pre-collected tongue images and multiple consultation data, the tongue image features of the pre-collected images and the change rule of the interview data characteristics are obtained, so that the tongue image features of the pre-collected images change. The regularity and the changing regularity of the characteristics of the consultation data are associated with the target blood component data, so as to accurately establish the tongue characteristics of each of the pre-collected tongue images and the corresponding relationship between each of the characteristics of the consultation data and the target blood component data.

For example, based on the machine learning model, the pre-collected tongue images are parsed, the tongue features of the pre-collected tongue images are acquired, and the relationship and correspondence between the tongue features of the plurality of pre-collected tongue images are obtained. The change relationship of the plurality of target blood component data is obtained, and the effective tongue image feature is obtained, and the corresponding relationship between the effective tongue image feature of the pre-collected tongue image image and the target blood component data is established. It is worth mentioning that a tongue image has multiple different representations, that is, a tongue image has multiple tongue features, which can be color, texture, shape, and so on. These features can be displayed in the whole tongue image or in part. While some tongue features are related to the target blood component data, other tongue features are not related to the target blood component data. In this embodiment, the effective tongue image feature is the tongue image feature related to the change of the target blood component data. Therefore, through the change relationship of the tongue image feature and the change relationship of multiple target blood component data, the effective tongue image feature is obtained, Establishing the corresponding relationship between the effective tongue feature of the pre-collected tongue image and the target blood component data, so that the establishment of the corresponding relationship between the tongue feature and the target blood component data is more accurate, which is beneficial to improve the target blood Accuracy of component predictions obtained.

As shown in FIG. 2B , after step 208 , a step 210 is further included, in which an image of the tongue image to be tested and the consultation data to be tested of the individual to be tested are acquired.

Step 212: Input the image of the tongue to be tested and the consultation data to be tested of the individual to be tested into the prediction model obtained by training, and analyze and obtain the target blood composition of the individual to be tested based on the prediction model. Predictive value.

For example, the image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into a prediction model obtained by machine learning, and the target blood components of the individual to be tested are obtained through analysis based on the prediction model Predictive value. In this embodiment, the prediction model obtained by training is the prediction model obtained by machine learning.

For example, an optimized machine classifier is obtained through a machine learning model.

In this embodiment, the individual to be tested is a user individual who needs to perform blood component prediction, and the tongue image image to be tested is the tongue image image of the user who needs to perform disease judgment, or in other words, the tongue image image to be tested is the user to be diagnosed The image of the tongue, and the data to be tested is the data filled in by the individual to be tested. For example, an image of the tongue to be tested is acquired through an image sensor. For example, the predicted value of the target blood component is the blood test data of the human body of the individual to be tested, that is, the predicted value of the target blood component and the data of the target blood component are the same data, and the predicted value of the target blood component predicts the blood of the human body for physical examination. The component data, for example, the predicted value of the target blood component is the blood detection data of the predicted human body, for example, the predicted value of the target blood component is the predicted blood component data of the human body. The target blood component prediction value is predicted based on the existing target blood component data, the target blood component data is the physical examination data obtained by actual detection, and the target blood component prediction value is based on the existing target blood component data and the predicted value. Collect the tongue characteristics of the tongue image and the data obtained by the corresponding relationship prediction of the interview data.

In this embodiment, based on the existing prediction model, the tongue image to be tested and the consultation data to be tested of the individual to be tested are obtained, and the tongue image to be tested of the individual to be tested and the consultation data to be tested are obtained. The data is input into the prediction model, and then the predicted value of the target blood component of the individual to be tested is predicted and obtained, without the need for blood testing of the individual, so that the acquisition of the blood component of the individual to be tested is more efficient and convenient.

In one embodiment, the image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into a prediction model, and the target blood of the individual to be tested is obtained through analysis based on the prediction model The step of component prediction value includes: inputting the tongue image image to be tested and the consultation data to be tested of the individual to be tested into a prediction model; based on the prediction model, parsing the tongue image image to be tested and the data to be tested. The data to be tested is obtained, and the tongue features of the tongue image to be tested and the features of the data to be tested are obtained; according to the features of the tongue image to be tested and the features of the data to be tested , to obtain the predicted value of the target blood component of the individual to be tested.

In this embodiment, after acquiring the tongue image to be tested and the consultation data to be tested of the individual to be tested, the tongue image to be tested and the consultation data to be tested of the individual to be tested are input into the prediction model, and the tongue image to be tested and the consultation data to be tested are input into the prediction model, The image image and the consultation data to be tested are analyzed to obtain the tongue characteristics of the tongue image to be measured and the characteristics of the consultation data to be measured, and based on the prediction model, the predicted value of the target blood component is obtained.

In one embodiment, the step of obtaining the predicted value of the target blood component of the individual to be tested according to the tongue image features of the tongue image image to be tested and the characteristics of the data to be tested includes: The tongue characteristics of the tongue image to be tested are closest to the tongue characteristics of the pre-collected tongue image and the characteristics of the consultation data, and the predicted value of the target blood component of the individual to be tested is obtained. .

For example, the tongue features of the tongue image to be tested are compared with the tongue features of the pre-collected tongue images, and the features of the inquiries data to be tested are compared with the features of the inquiries, based on the tongue features of the pre-collected tongue images The corresponding relationship with the target blood component data is based on the corresponding relationship between the characteristics of the inquiries data and the target blood component data, and then the predicted value of the target blood component is obtained. For example, acquiring the tongue characteristics of the tongue image to be measured, comparing the tongue characteristics of the tongue image to be measured with the tongue characteristics of the plurality of pre-collected tongue images, and determining the tongue characteristics of the tongue image to be measured. The tongue image features of the pre-collected tongue images that are closest to the tongue image features of the The corresponding relationship between the tongue features of the pre-collected tongue images and the features of the interview data and the target blood component data, and the target blood component prediction is obtained according to the tongue features of the closest pre-collected tongue images and the features of the interview data. value.

When the number of samples of the pre-collected tongue image and target blood component data is small, for example, the pre-collected tongue image that is closest to the tongue characteristics of the tongue image to be tested is obtained through the tongue characteristics of the tongue image to be tested. The tongue image feature of the image, through the feature of the consultation data to be tested, obtains the feature of the consultation data that is closest to the feature of the consultation data to be tested, and obtains the corresponding tongue image feature and the feature of the consultation data through the pre-collected tongue image feature. The target blood component data is used as the target blood component prediction value. For example, when the number of samples of pre-collected tongue image and consultation data is less than a preset threshold, a tongue image corresponding to the tongue image to be tested is obtained by using the tongue image features of the tongue image to be tested and the characteristics of the consultation data to be tested. characteristics, and the tongue characteristics and interrogation data characteristics of the pre-collected tongue image that are closest to the characteristics of the interrogation data to be tested, and then obtain the corresponding target blood components through the tongue characteristics and interrogation data characteristics of the pre-collected tongue image The target blood component data is used as the target blood component prediction value.

In this embodiment, since the number of samples of pre-collected tongue image and target blood component data is small, and the tongue features cannot be accurately quantified, the changes between the tongue features are not linear. The tongue characteristics of the tongue image cannot directly obtain the predicted value of the target blood components. It is necessary to first obtain the tongue characteristics of the pre-collected tongue images that are closest to the tongue characteristics of the tongue image to be tested as a reference, and then obtain the corresponding tongue characteristics. The target blood component data is obtained, and the target blood component data is further used as the target blood component prediction value, so as to improve the accuracy of acquiring the target blood component prediction value.

When the number of samples of pre-collected tongue image and target blood component data is large, for example, through the tongue image features of the tongue image to be tested and the characteristics of the consultation data to be tested, the tongue image features and the characteristics of the tongue image to be tested are obtained. For the target blood component data with the closest characteristics of the data to be tested, the target blood component data is used as the predicted value of the target blood component. For example, when the number of samples of pre-collected tongue images and consultation data is greater than a preset threshold, a tongue image corresponding to the tongue image to be tested is obtained through the tongue characteristics of the tongue image to be measured and the characteristics of the consultation data to be measured. The target blood component data whose characteristics are the closest to the characteristics of the consultation data to be tested are used as the target blood component prediction value. In this embodiment, since the number of samples of pre-collected tongue image and consultation data is relatively large, the changes between the tongue characteristics of the tongue image can be presented linearly. Therefore, according to the tongue image to be tested, the The tongue image features and the characteristics of the data to be tested are used to determine the closest target blood component data, and the target blood component data is used as the target blood component prediction value, so as to effectively improve the acquisition efficiency of the target blood component prediction value, and make the target blood component Obtaining predicted values is more accurate.

In one embodiment, the step of acquiring target blood component data and pre-collecting the tongue image includes: acquiring the consultation data corresponding to the target blood component data; acquiring a face image; parsing the face image , to obtain the pre-collected tongue image.

For example, the consultation data corresponding to the target blood component data is obtained; the face image of the individual corresponding to the target blood component data is obtained; the face image is analyzed to obtain the pre-collected tongue image.

For example, using the method disclosed in Patent Publication No. CN106859595A, a camera is used to detect a face; when a face is detected, a face image is obtained; the face image is parsed to obtain the position of the mouth; Collect tongue images.

Specifically, in this step, the camera detects the face in real time, that is, the camera is in a working state, for example, the camera is in a shooting state, and for example, the camera is in a video shooting state, it should be understood that the video is a plurality of continuous images, namely Videos are dynamic images. Through the image captured by the camera, it is detected whether the image contains a face, or whether the face in the image is facing the camera. For example, the HAAR-Like feature algorithm is used to detect faces, and the HAAR-Like feature algorithm is used to detect whether the face in the image is facing the camera. For example, to detect whether the application (App) is open, when it is detected that the application is open, the camera is turned on, and the face is detected through the camera. Specifically, the application is application software on the terminal. For example, the terminal is a mobile terminal, for example, the mobile terminal is a mobile phone, for example, the mobile terminal is a tablet computer.

When it is detected that the image captured by the camera contains a human face, or it is detected that the human face in the image captured by the camera is facing the camera, a face image is acquired. For example, photographing a face, obtaining a face image, for example, photographing a face in real time, obtaining multiple face images, for example, photographing a face video in real time, obtaining multiple face images, in this step, it is determined that the user's face is facing The camera, at this time, obtains a face image of the user, for example, when a face is detected, it shoots the face and obtains the face image. For example, when a human face is detected, turn on the flash, photograph the face, and obtain an image of the face, and turning on the flash is beneficial for clearer photos.

The obtained face image is parsed to obtain the position of the mouth on the face, where the mouth position is the position of the mouth on the face. For example, the HAAR-Like feature algorithm is used to analyze the face image, and the position of the mouth is obtained. the mouth position.

After the mouth position is acquired, the area corresponding to the mouth position is photographed to generate a tongue image. Specifically, the position of the mouth is the position corresponding to the human mouth, and shooting the position can accurately align the user's tongue, and then can accurately acquire the tongue image image.

In this embodiment, a face image is captured and obtained by a camera when a face is detected, and the position of the mouth of the face is obtained by analyzing the face image, and then the tongue image of the mouth is automatically obtained, so that the acquisition of the tongue image is more convenient. When the user uses the rear camera to shoot, there is no need for the user to choose to align the mouth, and there is no need for the user to adjust the posture according to the captured image, so that the tongue portrait is more accurate and convenient for the user to use.

For example, a method for predicting the value of glycated hemoglobin in blood is also called a method for predicting the value of glycated hemoglobin in blood based on machine learning. The method for prediction, please refer to Figure 2C, which includes the following steps:

Step 1: Collect multiple target blood components and corresponding pre-collected tongue images and consultation data, input and store them as training data.

In this embodiment, the target blood component is the target blood component data.

Step 2: Integrate the corresponding target blood components, pre-collected tongue images and consultation data into one data record.

That is, each corresponding target blood component, pre-collected tongue image and consultation data are integrated into one data record, and multiple target blood components, pre-collected tongue image and consultation data are integrated into multiple data records.

Step 3: Integrate the multiple data records into a data matrix, and import the data matrix into the storage module of the machine learning model.

For example, multiple target blood components, pre-collected tongue images and consultation data are integrated into a data matrix, and the data matrix is imported into the storage module of the machine learning model.

The fourth step is to train the deep neural network based on the data matrix. The training method is as follows:

a. Set the basic framework of machine learning, and build a data model including an input layer, at least one hidden layer and an output layer according to the data characteristics of multiple target blood components, pre-collected tongue images and interrogation data. The input layer contains pre-collected tongue images and interrogation data; the output layer contains the corresponding estimated blood components, and each hidden layer contains several nodes that have a mapping relationship with the output values of the previous layer.

b. Each node uses a mathematical equation to establish the data model of the node, and uses manual or random methods to preset the relevant parameter values in the mathematical equation. The input value of each node in the input layer is the data feature, and each hidden layer The input value of each node in the output layer is the output value of the upper layer, and the output value of each node in each layer is the value obtained by this node after the mathematical equation is operated.

c. Initialize the parameter value A _i , compare the output value of each node in the output layer with the target blood component of the corresponding node, repeatedly correct the parameter value A _i of each node, and cycle in turn, and finally obtain the The output value of each node in the output layer generates the parameter value A _i in each node corresponding to the output value when the blood component estimated value similarity is the local maximum.

Step 5, import the acquired pre-collected tongue image and consultation data into the machine learning model, and perform predictive analysis on the estimated value of blood components;

Step 6: The machine learning model outputs the result of the specific blood component value.

In one embodiment, personal health data and subjective symptoms that can be provided by the subject without expertise or examination are collected in the form of a questionnaire.

In one embodiment, the method for optimizing the parameter value A _i is an unsupervised learning method.

In one embodiment, the method for optimizing the parameter value A _i is a supervised learning method.

In one embodiment, the mathematical equation is a parametric mathematical equation or a nonparametric mathematical equation, wherein the parametric mathematical equation can be a linear model, a neuron model or a convolution operation, and the nonparametric mathematical equation can be an extreme value operation equation. The convolution model is set as follows:

where y is the blood component data in the output layer, the dimension is M _n , X is the training material data, the dimension is M ₀ , f ₁ to f _n are the set operation equations of each layer, and each layer equation f The dimension of ₁ is M _i-1 → M _i . For example, the first layer f ₁ converts X with dimension M ₀ into output Z ₁ with dimension M ₁ , and Z ₁ becomes the input of the second layer equation f ₂ , and so on, where each layer of model f _i has a matching parameter set A _i .

In one embodiment, the tongue image is collected by a mobile smart device with a lens, and the consultation data is collected by an electronic questionnaire with a predetermined structure.

The following is a specific example:

In this embodiment, the glycated hemoglobin value is taken as an example, as one of the implementation cases of consultation data collection:

Changes in blood glucose concentration will not only cause changes in the glycated hemoglobin value, but also cause changes in the physical state of the human body, and these changes will be reflected through subjective symptoms. In addition, changes in the physical state of the human body may also be related to age, gender, height, physique, and diseases. Specifically, in the implementation case of glycated hemoglobin value prediction, the consultation data is the health data of the individual that can generally be filled in by oneself, for example, the consultation data is the individual's age, gender, height, physique, and known diseases. Other data; another example, the consultation data is the individual's subjective symptoms, including whether there is dry mouth, whether there is frequent urination, whether there is blurred vision, etc. For example, the consultation data is obtained by user input. For example, the consultation data entered on the consultation questionnaire is acquired. For example, to obtain the data of the consultation entered into the computer. In this embodiment, by presetting the questions of the consultation questionnaire, the user can directly answer the questions on the consultation questionnaire, thereby obtaining the user's consultation data, which effectively improves the efficiency and accuracy of the obtaining of the consultation data.

The consultation questionnaire of the consultation data is shown in Figure 6. The consultation data feature is obtained according to the consultation data. The consultation data feature is the vectorization of the content of the consultation, and the vectorized value is the consultation data feature. As shown in FIG. 7 , by obtaining the vectorized value of the consultation data, the features of the consultation data are obtained, and the features of the consultation data are shown in the columns Q1-Q35 in FIG. 7 .

In this embodiment, pre-collected tongue images are also acquired. Acquire a pre-collected tongue image, the pre-collected tongue image is shown in Figure 8, then process the pre-collected tongue image, adopt the method in step 206 to process the pre-collected tongue image, The tongue part is intercepted, while other parts in the image are blacked out or shaded. The image after interception processing is shown in Figure 9. The quantized value of the processed image vector is the tongue image feature, as shown in columns 0-191 of FIG. 7 .

Preprocessing the target blood component data, the pre-collected tongue image and the consultation data; based on a machine learning model, parse the pre-collected tongue image and the interview data to obtain the pre-collected image Tongue characteristics and interrogation data characteristics of tongue images are used to establish prediction models.

For example, specifically, the collected target blood component values, such as the glycated hemoglobin value, and the corresponding consultation data and pre-collected tongue images, form a data set, and the data set is divided into training data set and test data, such as The flowchart in Figure 2C is shown to verify the accuracy of the model. Among them, by testing the results of multiple machine classifiers, an optimized model is obtained, that is, an optimized machine classifier model. Figure 10 shows the results of testing multiple machine classifiers. Among them, Random Forest has the highest accuracy (0.897638).

Then, the tongue image to be tested and the consultation data to be tested of the individual to be tested are acquired; the tongue image to be tested and the consultation data to be tested of the individual to be tested are input into a prediction model, based on the prediction The model analysis obtains the predicted value of the target blood component of the individual to be tested.

By acquiring the tongue images to be tested and the consultation data to be tested of a plurality of individuals to be tested, and inputting the tongue images to be tested and the consultation data to be tested of the individuals to be tested into the prediction model, based on the The prediction model is analyzed to obtain a plurality of predicted values of target blood components of the individual to be tested.

Specifically, for normal people, the normal range of glycated hemoglobin (HbA1c) values is between 4% and 5.6%. A glycated hemoglobin value between 5.7% and 6.4% means a subject has an increased likelihood of developing diabetes, known as prediabetes. When the glycated hemoglobin value reaches a level of 6.5% or higher, it means that the subject has diabetes. FIG. 11 is a diagram showing a comparison result between the predicted value of the target blood component of the individual to be tested and the actual blood component value of the individual to be tested obtained by analyzing the prediction model based on the prediction model. The number is the number of the individual to be tested, the glycated hemoglobin value is the actual glycated hemoglobin value of the individual to be tested, the real label is the actual result of the determination of diabetes of the individual to be tested, and the predicted glycated hemoglobin value is the target of the individual to be tested. The predicted value of blood components, that is, the glycated hemoglobin value predicted by the above method, the error percentage is the error between the predicted glycated hemoglobin value and the actual glycated hemoglobin value, and the prediction label is the predicted result of the diabetes of the individual to be tested.

For example, a plurality of glycated hemoglobin measurements are obtained by analyzing samples of a plurality of individuals to be tested. By comparing the predicted glycated hemoglobin value of the individual to be tested with the actual glycated hemoglobin value of the tested individual, the accuracy of the predicted value of glycated hemoglobin is obtained, and the result is shown in the standard confusion matrix diagram in FIG. 12 . FIG. 12 shows the category to which the predicted value of the target blood component of the individual to be tested belongs based on the preset classification of the glycated hemoglobin range. Among them, the standard confusion matrix diagram in Figure 12 shows the results of the optimized machine classifier on the test sample, the abscissa is the predicted value of the target blood component (predicted label) of the sample, and the ordinate is the actual blood component value of the sample (true Label), and the abscissa and ordinate are divided into three intervals, normal, diabetic and pre-diabetic. Among them, normal is a non-diabetic patient, the glycated hemoglobin value of the predicted value of the target blood component is normal, and diabetes corresponds to a diabetic patient , the glycated hemoglobin value of the predicted value of the target blood component is high, and the prediabetes corresponds to an individual with a tendency to develop diabetes, and the glycated hemoglobin value of the predicted value of the target blood component has a tendency to be high.

In this way, each coordinate interval represents the accuracy of the predicted value of the target blood component in the interval. For example, if the horizontal axis is diabetes, the vertical axis is diabetes, and the value of the corresponding interval is 0.97, it means that the predicted value of glycated hemoglobin of multiple individuals to be tested matches the actual diabetic glycated hemoglobin. For example, the number of individuals to be tested is 200 , and the number of individuals to be tested whose deviation between the predicted value of glycated hemoglobin and the actual blood component value is within the preset range is 194, indicating that the accuracy of the predicted value of the target blood component is 0.97. Numerical values in other intervals are deduced by analogy, which is not described redundantly in this embodiment.

It can be seen from Fig. 12 that the above-mentioned method for predicting the glycated hemoglobin composition value in blood can predict the glycated hemoglobin value of an individual and its clinical classification with a high accuracy rate, thereby assisting the judgment of diabetic patients, which is conducive to improving the patient's glycated hemoglobin value. The disease diagnosis efficiency has the characteristics of low price and quick and convenient application.

In one embodiment, as shown in FIG. 3 , a device for predicting the value of glycated hemoglobin in blood is provided, including: a data acquisition module 302 to be measured and a predicted value acquisition module 304, wherein:

The data acquisition module 302 to be tested is configured to acquire the tongue image to be tested and the medical consultation data to be tested of the individual to be tested.

The predicted value obtaining module 304 is configured to input the image of the tongue to be tested and the data of the consultation to be tested of the individual to be tested into the prediction model obtained by training, and analyze and obtain the individual to be tested based on the prediction model The predicted value of the target blood composition.

In one embodiment, it further includes: a blood component data acquisition module, a collection module, a model input module and a prediction model acquisition module, wherein:

The blood component data acquisition module is used to collect target blood component data.

The acquisition module is used to collect the consultation data corresponding to the target blood component data and the pre-collected tongue image.

The model input module is used for preprocessing the target blood component data, the pre-collected tongue image and the interrogation data.

The prediction model obtaining module is used to analyze the pre-collected tongue image and the interrogation data based on the machine learning model, obtain the tongue characteristics and interrogation data characteristics of the pre-collected tongue image, and according to the pre-collected tongue image The prediction model is obtained by training on the tongue image features of the image, the features of the consultation data and the target blood component data.

In one embodiment, the predicted value obtaining module includes: a data input unit to be measured, a feature analysis obtaining unit and a predicted value obtaining unit, wherein:

The data input unit to be tested is configured to input the image of the tongue image to be tested and the data of the medical consultation to be tested of the individual to be tested into the prediction model obtained by training.

The feature analysis and obtaining unit is configured to analyze the tongue image to be tested and the consultation data to be tested based on the prediction model, and obtain the tongue features of the tongue image to be tested and the features of the consultation data to be tested .

The predicted value obtaining unit is configured to obtain the predicted value of the target blood component of the individual to be measured according to the characteristics of the tongue image of the image to be measured and the characteristics of the data to be measured.

In one embodiment, the predicted value obtaining unit is further configured to determine a tongue image of the pre-collected tongue image that is closest to the tongue image feature of the tongue image image to be tested and the characteristics of the interview data to be tested. The characteristics and the characteristics of the interview data are used to obtain the predicted value of the target blood component of the individual to be tested.

In one embodiment, the acquisition module includes: an interrogation data acquisition unit, a face image acquisition unit, and a pre-collected tongue image acquisition unit, wherein:

The interrogation data acquisition unit is configured to acquire the interrogation data corresponding to the target blood component data.

The face image acquisition unit is used to acquire the face image.

The pre-collected tongue image acquisition unit analyzes the face image to acquire the pre-collected tongue image.

In one embodiment, the machine learning module is further configured to parse a plurality of the pre-collected tongue images based on the machine learning model, obtain tongue characteristics of each of the pre-collected tongue images, and establish each of the pre-collected tongue images. The correspondence between the tongue features of the tongue image and the target blood component data is collected.

For the specific limitation of the device for predicting the glycated hemoglobin composition value in blood, reference may be made to the above limitation on the method for predicting the glycated hemoglobin composition value in blood, which will not be repeated here. Each module in the above-mentioned apparatus for predicting the value of glycated hemoglobin composition in blood can be implemented in whole or in part by software, hardware and combinations thereof. The above modules can be embedded in or independent of the processor in the computer device in the form of hardware, or stored in the memory in the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.

For example, a device for predicting the value of glycated hemoglobin in blood based on machine learning includes an input module that can import tongue image and consultation data into a computing device, and separately or centrally save the tongue image and consultation data in a local device or Storage module for cloud storage. The device for predicting the glycated hemoglobin component value in blood also includes a pre-stored module for machine learning training, which matches the blood component value of the tongue image and the consultation data, and outputs the blood component value to the generated object. The input module includes a camera and software for driving the camera. The output device includes a mobile smart device.

For example, a device for predicting the value of glycated hemoglobin in blood based on machine learning includes a blood component value prediction module and a plurality of information interaction modules. One of the information interaction modules is connected to the tongue image acquisition module for acquiring tongue characteristics and outputting the tongue characteristics. Input to the blood component value model.

For example, the device for predicting the value of glycated hemoglobin in blood based on machine learning includes a blood component value prediction result output module, the blood component value prediction result output module generates a blood Component value calculation results.

In one embodiment, a computer device is provided, and the computer device may be a server, and its internal structure diagram may be as shown in FIG. 4 . The computer device includes a processor, memory, a network interface, and a database connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The nonvolatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium. The computer device's database is used to store blood component value prediction data. The network interface of the computer device is used to communicate with the server through the network connection. The computer program, when executed by the processor, implements a method for predicting the value of glycated hemoglobin in blood.

In one embodiment, a computer device is provided, and the computer device may be a terminal, and its internal structure diagram may be as shown in FIG. 5 . The computer equipment includes a processor, memory, a network interface, a display screen, and an input device connected by a system bus. Among them, the processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium, an internal memory. The nonvolatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the execution of the operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal through a network connection. The computer program, when executed by the processor, implements a method for predicting the value of glycated hemoglobin in blood. The display screen of the computer equipment may be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment may be a touch layer covered on the display screen, or a button, a trackball or a touchpad set on the shell of the computer equipment , it can also be an external keyboard, touchpad or mouse, etc., and it can also be a camera, which is used to collect tongue images.

Those skilled in the art can understand that the structure shown in FIG. 5 is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer equipment to which the solution of the present application is applied. Include more or fewer components than shown in the figures, or combine certain components, or have a different arrangement of components.

In one embodiment, a computer device is provided, including a memory and a processor, the memory stores a computer program, and the processor implements the following steps when executing the computer program:

Obtain the tongue image to be tested and the medical consultation data of the individual to be tested.

The image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into the prediction model obtained by training, and the predicted value of the target blood component of the individual to be tested is obtained through analysis based on the prediction model.

In one embodiment, the processor further implements the following steps when executing the computer program:

Collect target blood component data.

Inquiry data and pre-collected tongue images corresponding to the target blood component data are collected.

The target blood component data, the pre-collected tongue image and the interview data are preprocessed.

Based on the machine learning model, the pre-collected tongue image and the interview data are parsed, the tongue characteristics and the interview data of the pre-collected tongue image are obtained, and the tongue characteristics of the pre-collected tongue image are obtained , the characteristics of the interview data and the target blood component data are trained to obtain the prediction model.

In one embodiment, the processor further implements the following steps when executing the computer program:

The to-be-measured tongue image and the to-be-measured consultation data of the to-be-measured individual are input into the trained prediction model.

Based on the prediction model, the tongue image to be tested and the consultation data to be tested are parsed, and the tongue image features of the tongue image to be tested and the features of the consultation data to be tested are acquired.

According to the characteristics of the tongue image of the tongue image to be measured and the characteristics of the consultation data to be measured, the predicted value of the target blood component of the individual to be measured is obtained.

In one embodiment, the processor further implements the following steps when executing the computer program:

Determine the tongue features and the data features of the pre-collected tongue images that are closest to the features of the tongue images to be tested and the features of the data to be tested, and obtain the target of the individual to be tested Predictive value of blood components.

In one embodiment, the processor further implements the following steps when executing the computer program:

The consultation data corresponding to the target blood component data is acquired.

Get face images.

Parse the face image to obtain the pre-collected tongue image.

In one embodiment, a computer-readable storage medium is provided on which a computer program is stored, and when the computer program is executed by a processor, the following steps are implemented:

Obtain the tongue image to be tested and the medical consultation data of the individual to be tested.

The image of the tongue to be tested and the consultation data to be tested of the individual to be tested are input into the prediction model obtained by training, and the predicted value of the target blood component of the individual to be tested is obtained through analysis based on the prediction model.

In one embodiment, the computer program further implements the following steps when executed by the processor:

Collect target blood component data.

Inquiry data and pre-collected tongue images corresponding to the target blood component data are collected.

The target blood component data, the pre-collected tongue image and the interview data are preprocessed.

Based on the machine learning model, the pre-collected tongue image and the interview data are parsed, the tongue characteristics and the interview data of the pre-collected tongue image are obtained, and the tongue characteristics of the pre-collected tongue image are obtained , the characteristics of the interview data and the target blood component data are trained to obtain the prediction model.

In one embodiment, the computer program further implements the following steps when executed by the processor:

The to-be-measured tongue image and the to-be-measured consultation data of the to-be-measured individual are input into the trained prediction model.

Based on the prediction model, the tongue image to be tested and the consultation data to be tested are parsed, and the tongue image features of the tongue image to be tested and the features of the consultation data to be tested are acquired.

According to the characteristics of the tongue image of the tongue image to be measured and the characteristics of the consultation data to be measured, the predicted value of the target blood component of the individual to be measured is obtained.

In one embodiment, the computer program further implements the following steps when executed by the processor:

Determine the tongue features and the data features of the pre-collected tongue images that are closest to the features of the tongue images to be tested and the features of the data to be tested, and obtain the target of the individual to be tested Predictive value of blood components.

In one embodiment, the computer program further implements the following steps when executed by the processor:

The consultation data corresponding to the target blood component data is acquired.

Get face images.

Parse the face image to obtain the pre-collected tongue image.

Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be implemented by instructing relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage In the medium, when the computer program is executed, it may include the processes of the above-mentioned method embodiments. Wherein, any reference to memory, storage, database or other medium used in the various embodiments provided in this application may include non-volatile and/or volatile memory. Nonvolatile memory may include read only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM), or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous chain Road (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM) and so on.

The technical features of the above embodiments can be combined arbitrarily. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described. However, as long as there is no contradiction in the combination of these technical features It is considered to be the range described in this specification.

The above-mentioned embodiments only represent several embodiments of the present application, and the descriptions thereof are specific and detailed, but should not be construed as a limitation on the scope of the invention patent. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the scope of protection of the patent of the present application shall be subject to the appended claims.

Claims (10)

1. A method for predicting a glycated hemoglobin component value in blood, the method comprising:

collecting target blood component data; wherein the target blood component data is glycated hemoglobin data of a human body;

collecting inquiry data corresponding to the target blood component data and pre-collecting tongue picture images;

preprocessing the target blood component data, the pre-collected tongue image and the interrogation data; wherein the inquiry data comprises the age, sex, height, constitution, known diseases and subjective symptoms of the individual, and the subjective symptoms of the individual comprise whether the mouth is dry or not, whether the frequency of urination exists or not and whether the vision is blurred or not;

analyzing the pre-collected tongue picture image and the inquiry data based on a machine learning model, obtaining tongue picture characteristics and inquiry data characteristics of the pre-collected tongue picture image, and training according to the tongue picture characteristics, the inquiry data characteristics and target blood component data of the pre-collected tongue picture image to obtain a prediction model;

analyzing the pre-collected tongue picture image based on a machine learning model to obtain tongue picture characteristics of the pre-collected tongue picture image, and analyzing the characteristics of the inquiry data based on the machine learning model, wherein the characteristics of the inquiry data are vectorized values of the inquiry data, and the tongue picture characteristics are vectorized values of the processed image; obtaining effective tongue picture characteristics according to the variation relations among the tongue picture characteristics of the pre-collected tongue picture images and the variation relations of the characteristics of the corresponding inquiry data, and establishing the corresponding relations between the effective tongue picture characteristics of the pre-collected tongue picture images and the inquiry data characteristics and the target blood component data; on the basis of a machine learning method, learning to obtain a prediction model of the corresponding relation between the combination of the tongue picture characteristics and the inquiry data characteristics of each pre-collected tongue picture image and the target blood component data;

acquiring a tongue picture image to be detected and inquiry data to be detected of an individual to be detected;

inputting the tongue picture image to be tested of the individual to be tested and the inquiry data to be tested into a trained prediction model, and analyzing and obtaining a target blood component prediction value of the individual to be tested based on the prediction model.

2. The method of claim 1, wherein the effective tongue image characteristic is one of a plurality of characteristics in the tongue image, or wherein the effective tongue image characteristic is a tongue image characteristic associated with changes in the interrogation data in the pre-collected tongue image.

3. The method according to claim 1, wherein the step of inputting the tongue image to be tested and the inquiry data to be tested of the individual to be tested into a trained prediction model, and the step of obtaining the predicted value of the target blood component of the individual to be tested based on the analysis of the prediction model comprises the following steps:

inputting the tongue picture image to be detected and the inquiry data to be detected of the individual to be detected into the trained prediction model;

analyzing the tongue picture image to be detected and the inquiry data to be detected based on the prediction model, and acquiring the tongue picture characteristics of the tongue picture image to be detected and the inquiry data characteristics to be detected;

and obtaining a target blood component prediction value of the individual to be tested according to the tongue picture characteristics of the tongue picture image to be tested and the inquiry data characteristics to be tested.

4. The method according to claim 3, wherein the step of obtaining the predicted value of the target blood component of the individual to be tested according to the tongue image characteristics of the tongue image to be tested and the inquiry data characteristics to be tested comprises:

and determining the tongue picture characteristics and the inquiry data characteristics of the pre-collected tongue picture image which are closest to the tongue picture characteristics and the inquiry data characteristics of the tongue picture image to be detected, and obtaining the target blood component predicted value of the individual to be detected.

5. The method of claim 1, wherein the step of collecting the interrogation data corresponding to the target blood component data and pre-collecting tongue image data comprises:

obtaining the interrogation data corresponding to the target blood component data;

acquiring a face image;

and analyzing the face image to obtain the pre-collected tongue picture image.

6. An apparatus for predicting glycated hemoglobin component value in blood, which is realized by the method according to any one of claims 1 to 5, comprising:

the system comprises a to-be-detected data acquisition module, a to-be-detected data acquisition module and a to-be-detected data acquisition module, wherein the to-be-detected data acquisition module is used for acquiring a to-be-detected tongue picture image and to-be-detected inquiry data of an individual to be detected;

and the predicted value obtaining module is used for inputting the tongue picture image to be tested of the individual to be tested and the inquiry data to be tested into a trained predicted model, and analyzing and obtaining a target blood component predicted value of the individual to be tested based on the predicted model.

7. The apparatus of claim 6, further comprising:

a blood component data acquisition module for collecting target blood component data;

the acquisition module is used for collecting inquiry data corresponding to the target blood component data and pre-collected tongue picture images;

the model input module is used for preprocessing the target blood component data, the pre-collected tongue picture images and the inquiry data;

and the prediction model obtaining module is used for analyzing the pre-collected tongue picture image and the inquiry data based on a machine learning model, obtaining tongue picture characteristics and inquiry data characteristics of the pre-collected tongue picture image, and training according to the tongue picture characteristics, the inquiry data characteristics and target blood component data of the pre-collected tongue picture image to obtain the prediction model.

8. The apparatus of claim 6, wherein the predictor obtaining module comprises:

the to-be-tested data input unit is used for inputting the to-be-tested tongue picture image and the to-be-tested inquiry data of the to-be-tested individual into the trained prediction model;

a feature analysis obtaining unit, configured to analyze the tongue image to be detected and the inquiry data to be detected based on the prediction model, and obtain tongue features of the tongue image to be detected and characteristics of the inquiry data to be detected;

and the predicted value obtaining unit is used for obtaining the predicted value of the target blood component of the individual to be tested according to the tongue picture characteristics of the tongue picture image to be tested and the inquiry data characteristics to be tested.

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor implements the steps of the method of any one of claims 1 to 5 when executing the computer program.

10. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 5.

Images