CN110403611A

CN110403611A - Blood constituent value prediction technique, device, computer equipment and storage medium

Info

Publication number: CN110403611A
Application number: CN201810403962.XA
Authority: CN
Inventors: 张世平
Original assignee: Individual
Current assignee: Individual
Priority date: 2018-04-28
Filing date: 2018-04-28
Publication date: 2019-11-05
Anticipated expiration: 2038-04-28
Also published as: CN110403611B

Abstract

This application involves a kind of blood constituent value prediction technique, system, computer equipment and storage mediums.Method includes: the tongue picture image to be measured for obtaining test individual and interrogation data to be measured；The tongue picture image to be measured of the test individual and the interrogation data to be measured are input to the prediction model obtained through machine learning, the blood constituent predicted value of the test individual is obtained based on prediction model parsing.The prediction model obtained by machine learning model training, make it possible to the tongue picture image and interrogation data according to human body, obtain corresponding blood constituent predicted value, to realize that blood constituent value fast and automatically obtains, be conducive to improve the medical diagnosis on disease efficiency of patient, with cheap, using efficient and convenient feature.

Description

Blood constituent value prediction technique, device, computer equipment and storage medium

Technical field

This application involves blood constituent value electric powder predictions, more particularly to a kind of blood constituent value prediction technique, dress It sets, computer equipment and storage medium.

Background technique

Modern society people increasingly focus on health, and ordinary people also wishes can whenever and wherever possible, simply, conveniently, inexpensively Learn the health condition of oneself in ground.Various smart phones in the market, healthy bracelet have catered to such need to a certain extent It asks.But to the health status in depth understanding human body, then there is still a need for pass through laboratory data, especially blood constituent value.Blood There are many liquid ingredient type, such as hemoglobin, white blood cell, glycosylated hemoglobin, creatinine, erythrocyte sedimentation rate etc..The acquisition of blood values, one As be by extract blood perform an analysis, being one has traumatic operation.Also relevant instrument and consumptive material are needed when analysis, are related to Certain expense.

Traditional Chinese medicine possesses thousands of years of history in China, cure the disease, diseases prevention and health field have clinic very rich Experience.Wherein, the diagnosis and treatment method of Chinese medicine, with it is simple, conveniently, it is inexpensively famous.Huangdi's Internal Classics are said: " have all shapes in interior, must shape in Outside." that is have defect in the body of people, it is bound to display in body surface.Chinese medicine passes through hope, hear, asking, cutting four Method is examined, the performance of acquisition disease is gone, understands the health of human body.Wherein, lingual diagnosis and interrogation, even more the important composition portion of the four methods of diagnosis Part.But the current Chinese medicine four methods of diagnosis quasi-instrument as described in CN1561904A or WO2011026305A1, it is all to acquire the four methods of diagnosis Data as the main purpose, even if certain instruments have output to four methods of diagnosis data deduction as a result, only be confined to Chinese medicine in terms of Diagnostic result, the function that all blood constituent is not predicted.Furthermore this quasi-instrument is also specially to design, and price is high It is expensive, it is used only for professional, it is difficult to spread to general Luo great Zhong.

A variety of researchs are it is disclosed that the variation of tongue picture and the certain ingredients of blood has close relationship.For example, red blood cell meter Numerical value (RBC) and Hemoglobin Value (HGB) are light related with the depth of tongue color；Hematocrite value reduces (HCT) and light tongue and slight of stature Tongue is related；The raising of serum creatinine is related with a variety of tongue picture features when renal failure.

In another example diabetes are a kind of diseases of the blood sugar concentration exception in blood.Someone studies according to traditional Chinese medicine lingual diagnosis Principle, diabetes are diagnosed by tongue picture.But method used needs special instrument.This kind of diagnostic method is neither easy general And further relate to the expensive expense of purchase instrument.

Summary of the invention

Based on this, it is necessary in view of the above technical problems, provide a kind of blood constituent value prediction technique, device, computer Equipment and storage medium.

A kind of blood constituent value prediction technique, which comprises

The tongue picture image to be measured of acquisition test individual and interrogation data to be measured；

The tongue picture image to be measured of the test individual and the interrogation data to be measured are input to and trained acquisition Prediction model obtains the blood constituent predicted value of the test individual based on prediction model parsing.

The tongue picture image to be measured for obtaining test individual and the step of interrogation data to be measured in one of the embodiments, Before further include:

Collect target blood compositional data；

Collect interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture image；

The target blood compositional data, the pre-collecting tongue picture image and the interrogation data are pre-processed；

Based on machine learning model, the pre-collecting tongue picture image and the interrogation data are parsed, the pre-collecting is obtained The tongue picture feature and interrogation data characteristics of tongue picture image, it is special according to the tongue picture feature of the pre-collecting tongue picture image, interrogation data Target blood compositional data of seeking peace training obtains the prediction model.

It is described by the tongue picture image to be measured of the test individual and the interrogation to be measured in one of the embodiments, Data are input to the prediction model trained and obtained, and the blood constituent of the test individual is obtained based on prediction model parsing The step of predicted value includes:

The tongue picture image to be measured of the test individual and the interrogation data to be measured are input to and trained acquisition The prediction model；

Based on the prediction model, the tongue picture image to be measured and the interrogation data to be measured are parsed, are obtained described to be measured The tongue picture feature of tongue picture image and the interrogation data characteristics to be measured；

According to the tongue picture feature of the tongue picture image to be measured and the interrogation data characteristics to be measured, the test individual is obtained Blood constituent predicted value.

The tongue picture feature according to the tongue picture image to be measured and the interrogation number to be measured in one of the embodiments, According to feature, the step of obtaining the blood constituent predicted value of the test individual, includes:

Determination is immediate described pre- with the tongue picture feature of the tongue picture image to be measured and the interrogation data characteristics to be measured The tongue picture feature and interrogation data characteristics for collecting tongue picture image, obtain the blood constituent predicted value of the test individual.

Corresponding with the target blood compositional data interrogation data and advance of collecting in one of the embodiments, Collect tongue picture image the step of include:

Obtain the interrogation data corresponding with the target blood compositional data；

Obtain facial image；

The facial image is parsed, the pre-collecting tongue picture image is obtained.

A kind of blood constituent value prediction meanss, described device include:

Testing data obtain module, for obtain test individual tongue picture image to be measured and interrogation data to be measured；

Predicted value obtains module, for by the tongue picture image to be measured of the test individual and the interrogation data to be measured It is input to the prediction model trained and obtained, the blood constituent prediction of the test individual is obtained based on prediction model parsing Value.

In one of the embodiments, further include:

Blood component data obtains module, for collecting target blood compositional data；

Acquisition module, for collecting interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture figure Picture；

Mode input module, for the target blood compositional data, the pre-collecting tongue picture image and the interrogation Data are pre-processed；

Prediction model obtains module, for being based on machine learning model, parsing the pre-collecting tongue picture image and described asking Data are examined, the tongue picture feature and interrogation data characteristics of the pre-collecting tongue picture image are obtained, according to the pre-collecting tongue picture image Tongue picture feature, interrogation data characteristics and the training of target blood compositional data obtain the prediction model.

The predicted value acquisition module includes: in one of the embodiments,

Testing data input unit, for by the tongue picture image to be measured and the interrogation number to be measured of the test individual According to be input to trained obtain the prediction model；

Feature analysis obtaining unit parses the tongue picture image to be measured and described to be measured for being based on the prediction model Interrogation data obtain the tongue picture feature and the interrogation data characteristics to be measured of the tongue picture image to be measured；

Predicted value obtaining unit, for special according to the tongue picture feature of the tongue picture image to be measured and the interrogation data to be measured Sign, obtains the blood constituent predicted value of the test individual.

A kind of computer equipment, including memory and processor, the memory are stored with computer program, the processing Device performs the steps of when executing the computer program

A kind of computer readable storage medium, is stored thereon with computer program, and the computer program is held by processor It is performed the steps of when row

Above-mentioned blood constituent value prediction technique, device, computer equipment and storage medium pass through machine learning model, solution It analyses tongue picture image and obtains tongue picture feature and interrogation data characteristics, learnt by big data, establish prediction model, make it possible to basis The tongue picture image and interrogation data of human body parse tongue picture feature and interrogation data characteristics, and then obtain corresponding blood constituent Predicted value, to realize that blood constituent value fast and automatically obtains, is conducive to improve disease without carrying out blood test to individual The medical diagnosis on disease efficiency of people, have it is cheap, using efficient and convenient feature.

Detailed description of the invention

Fig. 1 is the application scenario diagram of blood constituent value prediction technique in one embodiment；

Fig. 2A is the flow diagram that the training of the prediction model in one embodiment obtains；

Fig. 2 B is the flow diagram of blood constituent value prediction technique in one embodiment；

Fig. 2 C is the process of the prediction process of the training process and model of the model based on machine learning in one embodiment Schematic diagram；

Fig. 3 is the structural block diagram of blood constituent value prediction meanss in one embodiment；

Fig. 4 is the internal structure chart of computer equipment in one embodiment；

Fig. 5 is the internal structure chart of computer equipment in another embodiment；

Fig. 6 is the schematic diagram of the interrogation questionnaire in one embodiment；

Fig. 7 is the schematic diagram of the interrogation data characteristics in one embodiment；

Fig. 8 is the schematic diagram of the original tongue picture image of one embodiment；

Fig. 9 is the schematic diagram of one embodiment treated tongue picture image；

Figure 10 is the result that multiple machine sort devices are tested in one embodiment；

Figure 11 is the comparison knot of blood constituent predicted value and practical blood constituent value that the parsing in one embodiment obtains Fruit figure；

Figure 12 is the standardization confusion matrix figure in one embodiment.

Specific embodiment

It is with reference to the accompanying drawings and embodiments, right in order to which the objects, technical solutions and advantages of the application are more clearly understood The application is further elaborated.It should be appreciated that specific embodiment described herein is only used to explain the application, not For limiting the application.

Blood constituent value prediction technique provided by the present application, can be applied in application environment as shown in Figure 1.Wherein, Computer 102 is communicated with server 104 by network by network.Wherein, computer 102 can be, but not limited to be various Personal computer, laptop, smart phone, tablet computer and portable wearable device, server 104 can be with independently The server cluster of server either multiple servers composition realize.Terminal obtains target blood compositional data, advance Collect tongue picture image and interrogation data, and by above-mentioned target blood compositional data, pre-collecting tongue picture image and interrogation data As the machine learning model of training data input server 104, server 104 is based on machine learning model, parses described pre- Tongue picture image and the interrogation data are collected, the tongue picture feature and interrogation data characteristics of the pre-collecting tongue picture image is obtained, builds Vertical prediction model.

For example, a kind of blood constituent value prediction technique, which comprises collect target blood compositional data；Collect with The corresponding interrogation data of the target blood compositional data and pre-collecting tongue picture image；To the target blood compositional data, institute It states pre-collecting tongue picture image and the interrogation data is pre-processed；Based on machine learning model, the pre-collecting tongue picture is parsed Image and the interrogation data obtain the tongue picture feature and interrogation data characteristics of the pre-collecting tongue picture image, establish prediction mould Type.

In above-described embodiment, by machine learning model, parses tongue picture image and obtains tongue picture feature and interrogation data characteristics, Learnt by big data, establish prediction model, make it possible to tongue picture image and interrogation data according to human body, parses tongue picture spy It seeks peace interrogation data characteristics, and then obtains corresponding blood constituent predicted value, to realize that blood constituent value fast and automatically obtains , be conducive to improve patient medical diagnosis on disease efficiency, have it is cheap, using efficient and convenient feature.

In one embodiment, as shown in Figure 2 A, provide a kind of blood constituent value prediction technique, include in this way with Lower step:

Step 202, target blood compositional data is collected.

For example, obtaining target blood compositional data.Specifically, which is the blood testing number of human body According to for example, the target blood compositional data is the blood component data of human body.For example, the target detection data are the sugar of human body Change hemoglobin data.The target blood compositional data is to collect to obtain in advance, and be input in computer.

Step 204, interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture image are collected.

For example, obtaining interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture image.For example, obtaining Interrogation data corresponding with the target blood compositional data are taken, and obtain pre-collecting corresponding with the target blood compositional data Tongue picture image.That is, interrogation data and the corresponding individual of pre-collecting tongue picture image are corresponding with target blood compositional data Individual is the same individual.

Specifically, which is the medical record data of human body, for example, the interrogation data are the illness data of human body, example Such as, which is the disease data of human body, for example, the interrogation data are individual health data.The interrogation data recordation The symptom of the human body of user, interrogation data further include the disease name of user, the age of user, the height of user, user body Again etc..It is obtained for example, the interrogation data are inputted by user.For example, obtaining the interrogation data for being input to computer.For example, obtaining The interrogation data being input on interrogation questionnaire in the present embodiment, the problem of by default interrogation questionnaire, allow users to directly Problem on corresponding interrogation questionnaire is answered, and then gets the interrogation data of user, effectively improves the acquisition of interrogation data Efficiency and accuracy.

For example, interrogation data include user suffered from disease name, the age of user, the height of user, the weight of user, The subjective symptoms etc. of user.In the present embodiment, by being based on the problem of medical knowledge presets interrogation questionnaire, allow users to straight The problem connect on corresponding interrogation questionnaire is answered, and then gets user's interrogation data relevant to target blood ingredient, is had Effect improves the acquisition efficiency and its accuracy of interrogation data.

The pre-collecting tongue picture image is the tongue picture image collected in advance, which is the image of the tongue of human body, people The variation close relation of fractions in the characterization and blood of human body of the tongue of body, and the change of the ingredient of the blood of human body Change the variation for the physical condition that will cause human body, the variation of the physical condition of human body will also will lead to the change of human blood components Change.For example, the variation of glycosylated hemoglobin has certain pass in the characterization and blood of human body of the tongue of human body.HbAle egg White is the product that the hemoglobin in red blood cell is combined with the carbohydrate in serum.It is by slow, lasting and irreversible Saccharification reaction is formed, content number depend on blood sugar concentration and blood glucose and hemoglobin time of contact.So saccharification blood Red eggs white value can effectively reflect the case where glycemic control in the past 1~2 month of individual, therefore be used as the prison of blood glucose control Survey index.Since the variation of blood sugar concentration and glycosylated hemoglobin value has close relationship, and the variation of blood sugar concentration also can Cause the variation of the physical condition of human body, therefore, if there are hyperglycemia states for a long time for human body, it will pass through the characterization of tongue picture It embodies.So different glycosylated hemoglobin values corresponds to different tongue characterizations.

Therefore, human body changes if there is disease or state, can all be embodied by the characterization of tongue picture.Also It is to say, the composition transfer of blood of human body will make the characterization of the tongue of human body change, the ingredient of different blood of human body The characterization of corresponding different tongue.For example, obtaining pre-collecting tongue picture image by imaging sensor.

In the present embodiment, individual i.e. detection user, each corresponding interrogation data and one of individual target blood compositional data Pre-collecting tongue picture image.

For example, multiple target blood compositional datas, multiple interrogation data and multiple pre-collecting tongue picture images are obtained, it is multiple Target blood compositional data, multiple interrogation data and multiple pre-collecting tongue picture images are one-to-one relationship.Multiple pre-collecting figures As having different characterizations, and each interrogation data are also different, therefore, the different characterizations interrogation number that correspondence is different According to the blood component data of corresponding Different Individual, that is to say, that the corresponding interrogation data of a target blood compositional data With a pre-collecting tongue picture image.

In the present embodiment, interrogation data are obtained, the accuracy for further increasing the parsing to tongue picture image is conducive to.

For example, multiple interrogation data are obtained, for example, each interrogation data and a target blood compositional data and a pre-collecting Tongue picture image is corresponding.Specifically, the interrogation data of same user and the target blood compositional data of the user and a pre-collecting tongue As image is corresponding.Be conducive to improve the accuracy to data parsing.

For example, obtaining the advance of the glycosylated hemoglobin value of several body, the interrogation data of several body and several body Collect tongue picture image, the pre-collecting tongue of the glycosylated hemoglobin value of the several body, the interrogation data of several body and several body As image is one-to-one relationship.Multiple pre-collecting images have a different characterizations, and each glycosylated hemoglobin value also it is each not It is identical, therefore, the different characterizations glycosylated hemoglobin value that correspondence is different.And corresponding glycosylated hemoglobin value and pre-collecting Tongue picture image then corresponds to same individual.

Step 206, the target blood compositional data, the pre-collecting tongue picture image and the interrogation data are carried out pre- Processing.

For example, after the pre-treatment, by the pretreated target blood compositional data, the pre-collecting tongue picture image and The interrogation data input machine learning model as training data.For example, machine learning model can be single kind of model, example Such as, decision-tree model, linear regression model (LRM), neural network model etc. are also possible to the knot of two or more models It closes.

Specifically, it is by the target blood compositional data, the pre-collecting tongue picture image and the interrogation Data Integration The data matrix is inputted machine learning model by data matrix.

For example, including: that the tongue picture image unification of shooting is arranged as 512x512 to the pretreatment of pre-collecting tongue picture image Pixel map carries out image interception processing by machine learning training in the method for convolutional network, and the image of tongue picture periphery is excluded Fall, obtains a new tongue picture image.In this new tongue picture image, it is not belonging to the pixel value quilt of tongue picture and its perienchyma It sets to zero.Then, by following formula, this new tongue picture image is converted into the feature vector with a certain length. Mathematical model equation used is as follows:

Dst (x, y) × src (f_x(x, y), f_y(x, y)), wherein dst (x, y) is feature vector, and src is conversion function, f_xFor image, f_yIt is the final dimension of vector.

In another example in certain embodiments, can choose certain tongue picture characteristics of image as feature vector, to replace tongue As pixel data, for example, using the color of tongue picture as feature, or using the form of tongue as feature, and or tongue tissue Texture is as feature.The mathematical equation for calculating these features can refer to using the existing correlation technique published [(Zhang,D.,Zhang,H.,&Zhang,B.).Tongue Image Analysis(pp.1-335) .Springer.2017]。

Classification depending on used tongue picture characteristics of image, it will obtain the vector of different dimensions.It can be by more Kind feature and indicator combination generate a new feature vector V., that is, V=<img, feature₁, features₂..., feature_n>.Use characteristics of image to replace tongue picture pixel data as feature vector, helps the accuracy for improving prediction.

Described in each interception tongue picture image or it is chosen after tongue picture characteristics of image be converted into vector after, Corresponding interrogation is imported as a result, according to different as a result, using binary system extension (binary expansion) or one-hot encoding (one hot encoding) generates different numerical value, creation one comprising tongue picture characteristics of image result and interrogation result most Whole vector.

Data generated are associated with corresponding target detection value.For pre-collecting data all images, ask It examines as a result, repeating the above process and establishes training dataset.Later, the data set of pre-collecting is divided into training sample and test specimens This, training sample is input in machine learning model, and test sample is used to the accuracy of measurement model.

Step 208, it is based on machine learning model, parses the pre-collecting tongue picture image and the interrogation data, obtains institute The tongue picture feature and interrogation data characteristics for stating pre-collecting tongue picture image, according to the tongue picture feature of the pre-collecting tongue picture image, are asked It examines data characteristics and the training of target blood compositional data obtains the prediction model.

For example, being based on machine learning model, parses the pre-collecting tongue picture image and obtain the pre-collecting tongue picture image Tongue picture feature parses the interrogation data, the corresponding relationship of interrogation data characteristics and target detection value is obtained, according to described advance Tongue picture feature, interrogation data characteristics and target blood the compositional data study for collecting tongue picture image obtain the prediction model.

Specifically, tongue picture feature is the characterization of tongue picture, and in other words, which is the characterization of tongue picture image.Interrogation Data characteristics is the value after the vectorization of interrogation data.

Specifically, since tongue picture feature is related to the blood constituent of human body, the variation of the blood constituent of human body will cause tongue As the change of feature, different tongue picture features corresponds to the blood constituent of different human bodies, and therefore, different tongue picture features will correspond to Different target blood compositional datas.

In addition, interrogation data and target blood compositional data are one-to-one relationships, different blood component datas will Cause the interrogation data of individual different, therefore, interrogation data are associated with blood component data, therefore, in this step, to advance Collection tongue picture image and interrogation data are parsed, and multiple tongue picture features and the interrogation data for obtaining the pre-collecting tongue picture image are special Sign is based on machine learning method, the tongue picture feature and interrogation data characteristics and target blood of each pre-collecting tongue picture image of study acquisition The prediction model of the corresponding relationship of liquid compositional data.For example, according to pre-collecting tongue picture image and interrogation data characteristics with it is corresponding Target blood compositional data corresponding relationship, study obtain each pre-collecting tongue picture image tongue picture feature and interrogation data with The rule of the corresponding relationship of target blood compositional data, to establish prediction model.

In the present embodiment, after obtaining target blood compositional data, target blood compositional data and pre-collecting tongue picture are established The corresponding relationship of image and the interrogation data enables to target blood compositional data and pre-collecting tongue picture image and asks It examines data correlation, and then obtains variation and interrogation data and the target blood component number of the tongue picture feature of pre-collecting tongue picture image According to relationship, obtain pre-collecting tongue picture image tongue picture feature and interrogation data characteristics and target blood compositional data variation Rule, and tongue picture feature based on pre-collecting tongue picture image and interrogation data characteristics and target blood compositional data are established with this Incidence relation prediction model so that the acquisition of blood constituent predicted value is more accurate.

In the present embodiment, it is based on a large amount of pre-collecting tongue picture image and interrogation data and target blood compositional data, into The parsing of row big data and study, and then the tongue picture feature of study acquisition pre-collecting tongue picture image and interrogation data characteristics and target The consistent relationship of blood component data, and then establish the tongue picture feature based on the pre-collecting tongue picture image and interrogation data spy The prediction model of sign and the corresponding relationship of the target blood compositional data.For example, the corresponding relationship can turn to curve with image Figure, for example, the tongue picture feature of the pre-collecting tongue picture image and the corresponding relationship of the target blood compositional data are mapped as Curve.By establishing the corresponding relationship of the tongue picture feature of pre-collecting tongue picture image, so that in subsequent detection, Ke Yitong It crosses tongue picture image and interrogation data obtains blood constituent predicted value, be conducive to the diagnosis of aided disease.

For example, being based on machine learning model, the pre-collecting tongue picture image is parsed, the pre-collecting tongue picture image is obtained Tongue picture feature, according between the tongue picture feature of multiple pre-collecting tongue picture images variation relation and corresponding multiple interrogations The variation relation of data obtains effective tongue picture feature, and the effective tongue picture feature for establishing the pre-collecting tongue picture image is asked with described Examine the corresponding relationship of data.

Effective tongue picture feature is one in tongue picture image in multiple features, and in the present embodiment, effective tongue picture feature is pre- The tongue picture feature relevant to the variation of interrogation data in tongue picture image is collected, therefore, passes through the variation relation of tongue picture feature, knot Multiple interrogation data are closed, effective tongue picture feature is obtained, the effective tongue picture feature for establishing the pre-collecting tongue picture image is asked with described The corresponding relationship of data is examined, so that tongue picture feature and establishing for the corresponding relationship of the interrogation data are more accurate, favorably In the accuracy for improving the acquisition of blood constituent predicted value.

For example, being based on machine learning model, multiple pre-collecting tongue picture images and interrogation data are parsed, each institute is obtained The tongue picture feature and each interrogation data characteristics for stating pre-collecting tongue picture image, the tongue picture for establishing each pre-collecting tongue picture image are special The corresponding relationship of sign and each interrogation data characteristics and the target blood compositional data, establishes prediction model.

In the present embodiment, by being parsed respectively to the multiple interrogation data of multiple pre-collecting tongue picture images, obtain advance Collect image tongue picture feature and interrogation data characteristics changing rule so that the changing rule of the tongue picture feature of pre-collecting image with And the changing rule of interrogation data characteristics is associated with target blood compositional data, and then accurately establishes each pre-collecting tongue picture figure The corresponding relationship of the tongue picture feature of picture and each interrogation data characteristics and the target blood compositional data.

For example, being based on machine learning model, the pre-collecting tongue picture image is parsed, the pre-collecting tongue picture image is obtained Tongue picture feature, according between the tongue picture feature of multiple pre-collecting tongue picture images variation relation and corresponding multiple targets The variation relation of blood component data obtains effective tongue picture feature, establishes effective tongue picture feature of the pre-collecting tongue picture image With the corresponding relationship of the target blood compositional data.It is noted that a tongue picture image has multiple and different characterizations, That is, a tongue picture image has multiple tongue picture features, these features can be color, texture, shape etc..These are special Sign can be in the display on the whole of tongue picture image, can also be in local display.And some tongue pictures are characterized in and target blood ingredient Data are relevant, and other tongue picture be characterized in it is incoherent with target blood compositional data.In the present embodiment, effective tongue picture is special Sign is tongue picture feature relevant to the variation of target blood compositional data, therefore, by the variation relation of tongue picture feature, in conjunction with more The variation relation of a target blood compositional data obtains effective tongue picture feature, establishes effective tongue of the pre-collecting tongue picture image As the corresponding relationship of feature and the target blood compositional data, so that tongue picture feature and the target blood compositional data Corresponding relationship establish it is more accurate, be conducive to improve blood constituent predicted value obtain accuracy.

As shown in Figure 2 B, after step 208 further include step 210, obtain test individual tongue picture image to be measured and to Survey interrogation data.

Step 212, the tongue picture image to be measured of the test individual and the interrogation data to be measured are input to and have been instructed Practice the prediction model obtained, the blood constituent predicted value of the test individual is obtained based on prediction model parsing.

For example, being input to the tongue picture image to be measured of the test individual and the interrogation data to be measured through engineering The prediction model obtained is practised, the blood constituent predicted value of the test individual is obtained based on prediction model parsing.This implementation In example, the prediction model that training obtains is the prediction model that machine learning obtains.

For example, the machine sort device optimized by machine learning model.

In the present embodiment, test individual is the user's individual for needing to carry out blood constituent prediction, and tongue picture image to be measured is to need The tongue picture image of the user of illness judgement is carried out, in other words, which is the figure of the tongue of user to be diagnosed Picture, interrogation data to be measured are the interrogation data that test individual is filled in.For example, obtaining tongue picture image to be measured by imaging sensor. For example, the blood constituent predicted value is the blood testing data of the human body of test individual, that is to say, that the blood constituent predicted value It is homogeneous data, the physical examination blood component data of the human body of blood constituent predicted value prediction, example with target blood compositional data Such as, which is the blood testing data of the human body of prediction, for example, the blood constituent predicted value is the people of prediction The blood component data of body.The blood constituent predicted value is predicted based on existing target blood compositional data, should Target blood compositional data is the physical examination data of actually detected acquisition, and blood constituent predicted value is based on existing target blood The data that the corresponding relationship prediction of the tongue picture feature and interrogation data of compositional data and pre-collecting tongue picture image obtains.

In the present embodiment, it is based on existing prediction model, obtains the tongue picture image to be measured and interrogation number to be measured of test individual According to the tongue picture image to be measured of the test individual and the interrogation data to be measured being input in prediction model, and then pre- The blood constituent predicted value for obtaining the test individual is surveyed, without carrying out blood test to individual, so that test individual Blood constituent acquisition is highly efficient, more convenient.

In one embodiment, described by the tongue picture image to be measured of the test individual and the interrogation data to be measured The step of being input to prediction model, obtaining the blood constituent predicted value of the test individual based on prediction model parsing is wrapped It includes: the tongue picture image to be measured of the test individual and the interrogation data to be measured is input to prediction model；Based on described Prediction model parses the tongue picture image to be measured and the interrogation data to be measured, and the tongue picture for obtaining the tongue picture image to be measured is special It seeks peace the interrogation data characteristics to be measured；It is special according to the tongue picture feature of the tongue picture image to be measured and the interrogation data to be measured Sign, obtains the blood constituent predicted value of the test individual.

In the present embodiment, after the tongue picture image to be measured and interrogation data to be measured for obtaining test individual, by test individual Tongue picture image to be measured and interrogation data to be measured are input to prediction model, solve to the tongue picture image to be measured and interrogation data to be measured Analysis, obtains the tongue picture feature and interrogation data characteristics to be measured of the tongue picture image to be measured, is based on the prediction model, obtains blood constituent Predicted value.

In one embodiment, described special according to the tongue picture feature of the tongue picture image to be measured and the interrogation data to be measured The step of levying, obtaining the blood constituent predicted value of the test individual comprises determining that and the tongue picture of the tongue picture image to be measured spy It seeks peace the tongue picture feature and interrogation data characteristics of the immediate pre-collecting tongue picture image of the interrogation data characteristics to be measured, obtains Obtain the blood constituent predicted value of the test individual.

It, will be to for example, the tongue picture feature of tongue picture image to be measured and the tongue picture feature of pre-collecting tongue picture image are compared It surveys interrogation data characteristics to compare with interrogation data characteristics, the tongue picture feature based on pre-collecting tongue picture image and the target blood The corresponding relationship of liquid compositional data, the corresponding relationship based on interrogation data characteristics Yu the target blood compositional data, and then obtain Take blood constituent predicted value.For example, obtaining the tongue picture feature of the tongue picture image to be measured, the tongue of the tongue picture image to be measured is compared As the tongue picture feature of feature and multiple pre-collecting tongue picture images, determination is closest with the tongue picture feature of the tongue picture image to be measured Pre-collecting tongue picture image tongue picture feature, compare interrogation data characteristics to be measured and multiple interrogation data characteristicses, determine with should be to Survey the immediate interrogation data characteristics of interrogation data characteristics, tongue picture feature and interrogation number based on the pre-collecting tongue picture image According to the corresponding relationship of feature and the target blood compositional data, according to the tongue picture feature of immediate pre-collecting tongue picture image with And interrogation data characteristics obtains blood constituent predicted value.

When the sample size of pre-collecting tongue picture image and target blood compositional data is less, for example, passing through tongue picture to be measured The tongue picture feature of image obtains special with the tongue picture of the immediate pre-collecting tongue picture image of tongue picture feature of the tongue picture image to be measured Sign, by interrogation data characteristics to be measured, acquisition and the immediate interrogation data characteristics of interrogation data characteristics to be measured are pre- by this The tongue picture feature and interrogation data characteristics for collecting tongue picture image obtain corresponding target blood compositional data, with the target blood at Divided data is blood constituent predicted value.For example, when pre-collecting tongue picture image and the sample size of interrogation data are less than preset threshold When, by the tongue picture feature and interrogation data characteristics to be measured of tongue picture image to be measured, obtain special with the tongue picture of the tongue picture image to be measured Sign and tongue picture feature and interrogation data characteristics with the immediate pre-collecting tongue picture image of interrogation data characteristics to be measured, it is then logical The tongue picture feature and interrogation data characteristics for crossing the pre-collecting tongue picture image obtain corresponding target blood compositional data, with the target Blood component data is blood constituent predicted value.

In the present embodiment, since the sample size of pre-collecting tongue picture image and target blood compositional data is less, and tongue picture Feature is unable to accurate quantification, therefore, the variation between tongue picture feature be not it is linear, as a result, by the tongue of tongue picture image to be measured As feature can not directly obtain blood constituent predicted value, need to obtain first immediate with the tongue picture feature of tongue picture image to be measured The tongue picture feature of pre-collecting tongue picture image obtains corresponding target blood compositional data as reference with this, and then with the mesh Blood component data is marked as blood constituent predicted value, to improve the accuracy of the acquisition of blood constituent predicted value.

When the sample size of pre-collecting tongue picture image and target blood compositional data is more, for example, passing through tongue picture to be measured The tongue picture feature of image and interrogation data characteristics to be measured obtain tongue picture feature and interrogation data to be measured with the tongue picture image to be measured The immediate target blood compositional data of feature, using the target blood compositional data as blood constituent predicted value.For example, when advance When collecting the sample sizes of tongue picture image and interrogation data and being greater than preset threshold, by the tongue picture feature of tongue picture image to be measured and to be measured Interrogation data characteristics obtains and the tongue picture feature of the tongue picture image to be measured and the immediate target blood of interrogation data characteristics to be measured Compositional data, using the target blood compositional data as blood constituent predicted value.In the present embodiment, due to pre-collecting tongue picture image and The sample size of interrogation data is larger, and therefore, the variation between the tongue picture feature of tongue picture image can be linearly presented, therefore, Then immediate target blood component number can be determined according to the tongue picture feature and interrogation data characteristics to be measured of the tongue picture image to be measured According to, using the target blood compositional data as blood constituent predicted value, so that the acquisition efficiency of blood constituent predicted value is effectively improved, And make the acquisition of blood constituent predicted value more accurate.

In one embodiment, the acquisition target blood compositional data and the step of pre-collecting tongue picture image include: to obtain Take the interrogation data corresponding with the target blood compositional data；Obtain facial image；The facial image is parsed, is obtained The pre-collecting tongue picture image.

For example, obtaining the interrogation data corresponding with the target blood compositional data；It obtains and the target blood The facial image of the corresponding individual of compositional data；The facial image is parsed, the pre-collecting tongue picture image is obtained.

For example, passing through camera detection face using method disclosed in patent publication No. CN106859595A；Work as detection When to face, facial image is obtained；The facial image is parsed, mouth position is obtained；It is obtained according to the mouth position advance Collect tongue picture image.

Specifically, in this step, camera real-time detection face, the i.e. camera are in running order, for example, the camera shooting Head is in shooting state, and for another example, which is in shooting video state, it should be appreciated that, video is multiple consecutive images, i.e., Video is dynamic image.The image taken by camera, whether comprising in face or detection image in detection image Face whether face camera.For example, detecting face using HAAR-Like characteristics algorithm, examined using HAAR-Like characteristics algorithm Face in altimetric image whether face camera.For example, whether detection opens using (App), when detecting using opening, open Camera is opened, camera detection face is passed through.Specifically, using for the application software in terminal.For example, the terminal is mobile whole End, for example, the mobile terminal is mobile phone, for example, the mobile terminal is tablet computer.

When comprising face, or detecting the face in the image of camera shooting in the image for detecting camera shooting It is right against camera, then obtains facial image.For example, shooting face, obtains facial image, for example, captured in real-time face, obtains Multiple facial images obtain multiple facial images for example, captured in real-time face video, in this step, it is determined that the face of user Face camera, for example, shooting face when detecting face, obtains facial image at this point, obtaining the facial image of user. For example, opening flash lamp when detecting face, face is shot, obtains facial image, opens the photograph that flash lamp is conducive to shooting Piece is relatively sharp.

The facial image got is parsed, the position of face upper oral part, mouth on the mouth position, that is, face are obtained Bar position.For example, parsing the facial image using HAAR-Like characteristics algorithm, mouth position is obtained, for example, using HAAR-Like characteristics algorithm carries out the detecting at each position of face to facial image, identifies and confirm the position of oral area, described in acquisition Mouth position.

After getting mouth position, the corresponding region of the mouth position is shot, generates tongue picture image.Specifically Ground, the mouth position are the corresponding position of human body mouth, the tongue that can precisely align user are shot to the position, in turn Tongue picture image can accurately be obtained.

In the present embodiment, facial image is obtained by shooting when camera detection face, and parse facial image and get The mouth position of face, and then the tongue picture image of oral area is obtained automatically, so that the acquisition of tongue picture image is more convenient, used in user When rear camera is shot, alignment oral area is selected without user, and without user according to the Image Adjusting posture of shooting, so that Tongue picture head portrait is more accurate, is easy to use by users.

For example, a kind of blood constituent value prediction technique is also referred to as a kind of blood constituent value prediction side based on machine learning Method is referred to as a kind of method predicted with tongue picture and interrogation data blood constituent value based on machine learning, please tied Close Fig. 2 C comprising following steps:

Step 1 acquires multiple target blood ingredients and corresponding pre-collecting tongue picture image and interrogation data, as instruction Practice data to input and store.

In the present embodiment, target blood ingredient, that is, target blood compositional data.

Corresponding target blood ingredient, pre-collecting tongue picture image and interrogation Data Integration are a data record by step 2.

I.e. each corresponding target blood ingredient, pre-collecting tongue picture image and interrogation Data Integration are a data record, Multiple target blood ingredients, pre-collecting tongue picture image and interrogation Data Integration are multiple data records.

Multiple data records are integrated into data matrix by step 3, and data matrix is imported to the storage of machine learning model In module.

For example, being data matrix by multiple target blood ingredients, pre-collecting tongue picture image and interrogation Data Integration, by data Matrix imports in the memory module of machine learning model.

Step 4 is trained deep neural network based on data matrix.Training method is as follows:

A. machine learning basic framework is set, multiple target blood ingredients, pre-collecting tongue picture image and interrogation data are pressed The data model including input layer, at least one layer of hidden layer and output layer is established according to data characteristics.Input layer includes pre-collecting tongue picture Image and interrogation data；Output layer includes corresponding blood constituent estimated value, and each hidden layer is defeated with upper one layer comprising several Value has the node of mapping corresponding relationship out.

B. each node establishes the data model of the node using math equation, default described using artificial or random device Related parameter values in math equation, the input value of each node is the data characteristics, each hidden layer and output layer in input layer In each node input value be upper layer output valve, the output valve of each node is this node through the math equation operation in every layer Resulting value afterwards.

C. the parameter value A is initialized_i, by the target blood of the output valve of node each in the output layer and corresponding node Ingredient compares, and corrects the parameter value A of each node repeatedly_i, circuit sequentially, final obtain makes each node in the output layer The corresponding each node of output valve of output valve when generating with blood constituent estimated value similarity as local maxima in parameter value A_i。

Step 5, the pre-collecting tongue picture image and interrogation data that will acquire import in the machine learning model, to blood at Estimated value is divided to carry out forecast analysis；

Step 6, by the result of machine learning model output and the specific blood constituent value.

In one embodiment, it is collected in the form of questionnaire without professional knowledge or that can be provided by the object is provided People's health data and subjective symptoms.

In one embodiment, to the parameter value A_iThe method optimized is unsupervised learning method.

In one embodiment, to the parameter value A_iThe method optimized is supervised learning method.

In one embodiment, the math equation is parameter mathematical equation or nonparametric math equation, wherein parameter number Learning equation can be linear model, neuron models or convolution algorithm, and nonparametric math equation can be extreme value operational equation.Convolution mould Type setting means is as follows:

Y=g (X)=f_nO f _n-1O f _n-2O…Of₁(X)

Wherein y is the blood component data in the output layer, dimension M_n, X is trained material data, dimension M₀, f₁ To f_nFor each layer of operational equation of setting, and each layer of Equation f₁Dimension be M_i-1→M_i, such as first layer f₁It is exactly by dimension For M₀X be converted into dimension be M₁Output Z₁, and Z₁Then become second layer Equation f₂Input, and so on, wherein each layer Model f_iThere is matched parameter group A_i。

In one embodiment, the tongue picture image is collected with the intelligent movable device with camera lens, the interrogation number The electric questionnaire of given structure is collected accordingly.

Here is a specific embodiment:

One of implementation case in the present embodiment, specifically by taking glycosylated hemoglobin value as an example, as the acquisition of interrogation data Example:

Variation of the variation of blood sugar concentration in addition to that can cause glycosylated hemoglobin value, can also cause the change of human body state Change, and these variations can be reflected by subjective symptoms.In addition, human body physical condition variation may also with the age, gender, Height, constitution, illnesses etc. are related.Specifically, in the case study on implementation of glycosylated hemoglobin value prediction, which is The personal health data that oneself can generally fill in, for example, the interrogation data be the personal age, gender, height, constitution, The data such as known illnesses；In another example the interrogation data are personal subjective symptoms, dry has been included whether, if having urine Frequently, if having eye-blurred etc..It is obtained for example, the interrogation data are inputted by user.It is input on interrogation questionnaire for example, obtaining Interrogation data.For example, obtaining the interrogation data for being input to computer.In the present embodiment, the problem of by default interrogation questionnaire, make The problem that obtaining user can directly correspond on interrogation questionnaire is answered, and then gets the interrogation data of user, is effectively improved The acquisition efficiency and its accuracy of interrogation data.

The interrogation questionnaire of interrogation data is as shown in fig. 6, obtain interrogation data characteristics, the interrogation data according to the interrogation data Feature is value, that is, interrogation data characteristics by the content vectorization of interrogation, after vectorization as shown in fig. 7, obtaining interrogation data Vectorization after value i.e. obtain interrogation data characteristics, interrogation data characteristics as in Fig. 7 Q1-Q35 arrange shown in.

In the present embodiment, pre-collecting tongue picture image is also obtained.Pre-collecting tongue picture image is obtained, the pre-collecting tongue picture image is such as Shown in Fig. 8, then pre-collecting tongue picture image is handled, the side in step 206 is used to the processing of pre-collecting tongue picture image Method, by the tongue portion intercepts in pre-collecting image, and the other parts in image then use blacking or shade to cover, interception Image is as shown in Figure 9 after processing.Value after the processing after image vector is tongue picture characteristics of image, as the 0-191 of Fig. 7 is arranged It is shown.

The target blood compositional data, the pre-collecting tongue picture image and the interrogation data are pre-processed；Base In machine learning model, the pre-collecting tongue picture image and the interrogation data are parsed, obtains the pre-collecting tongue picture image Tongue picture feature and interrogation data characteristics, establish prediction model.

For example, specifically, collected target blood signal component value, such as glycosylated hemoglobin value and corresponding ask Data and pre-collecting tongue picture image are examined, data set is divided into training dataset and inspection data by composition data collection, such as the stream of Fig. 2 C Shown in journey figure, with the accuracy of testing model.Wherein, by the multiple machine sort devices of test as a result, obtaining the mould of optimization Type, that is, the machine sort device model optimized.As shown in Figure 10, the result to test multiple machine sort devices.Wherein random forest The accuracy rate highest (0.897638) of (Random Forest).

Then, obtain test individual tongue picture image to be measured and interrogation data to be measured；By described in the test individual to It surveys tongue picture image and the interrogation data to be measured is input to prediction model, obtain described to be measured based on prediction model parsing The blood constituent predicted value of body.

By obtain multiple test individuals tongue picture image to be measured and interrogation data to be measured, and by the institute of the test individual It states tongue picture image to be measured and the interrogation data to be measured is input to prediction model, multiple institutes are obtained based on prediction model parsing State the blood constituent predicted value of test individual.

Specifically, for normal person, the normal range (NR) of glycosylated hemoglobin (HbA1c) value 4% and 5.6% it Between.Glycosylated hemoglobin value means that a possibility that subject suffers from diabetes increases between 5.7% and 6.4%, claims glycosuria Sick early period (prediabetes).When glycosylated hemoglobin value reaches 6.5% or higher level, it is meant that subject suffers from There is diabetes (diabetes).Figure 11 is the blood constituent predicted value that the test individual is obtained based on prediction model parsing With the comparing result figure of the practical blood constituent value of test individual.Wherein, number is the number of test individual, glycosylated hemoglobin Value is the actual glycosylated hemoglobin value of test individual, and real label is the practical knot of the judgement of the diabetes of the test individual Fruit, prediction glycosylated hemoglobin value are the blood constituent predicted value of test individual, i.e., the saccharification blood predicted by the above method Red eggs white value, percentage error are the error predicted between glycosylated hemoglobin value and actual glycosylated hemoglobin value, prediction Label is the result of the prediction to the diabetes of test individual.

For example, the parsing of the sample by multiple test individuals, obtains multiple glycosylated hemoglobin measured values.By comparison to The glycosylated hemoglobin predicted value of individual and the practical glycosylated hemoglobin value of test individual are surveyed, glycosylated hemoglobin prediction is obtained The accuracy rate of value, as a result as shown in the standard confusion matrix figure of Figure 12.Figure 12 is based on preset to glycosylated hemoglobin range Classification, obtains the generic of the blood constituent predicted value of test individual.Wherein, the standard confusion matrix figure in Figure 12 is shown The machine sort device of optimization is to test sample as a result, blood constituent predicted value (prediction label) of the abscissa for sample, vertical seat It is designated as the actual blood constituent value (real label) of sample, and abscissa and ordinate are all divided into three sections, normal, glycosuria Disease and prediabetes, wherein be normally non-diabetic patients, the glycosylated hemoglobin value of blood constituent predicted value is just Often, it is diabetic that diabetes are corresponding, and the glycosylated hemoglobin value of blood constituent predicted value is higher, prediabetes pair What is answered is the individual for having onset diabetes to be inclined to, and the glycosylated hemoglobin value of blood constituent predicted value has the tendency that higher.

In this way, each coordinate section indicates the accuracy rate of blood constituent predicted value in the section.For example, abscissa zone For diabetes, ordinate is diabetes, and the numerical value in corresponding section is 0.97, then shows the HbAle egg of multiple test individuals White predicted value and actual diabetes glycosylated hemoglobin matching rate, for example, test individual quantity is 200, and HbAle egg The number of the test individual of the deviation of white predicted value and actual blood constituent value within a preset range is 194, then shows blood The accuracy rate of liquid ingredient prediction value is 0.97.The numerical value in other sections and so on, not burdensome description in the present embodiment.

It as shown in Figure 12, can be to predict individual sugar compared with high-accuracy by above-mentioned blood constituent value prediction technique Change Hemoglobin Value, and its clinical classification, and then assist the judgement to diabetic, is conducive to the disease for improving patient Diagnosis efficiency, have it is cheap, using efficient and convenient feature.

In one embodiment, as shown in figure 3, providing a kind of blood constituent value prediction meanss, comprising: testing data obtains Modulus block 302 and predicted value obtain module 304, in which:

Testing data obtain module 302 be used for obtain test individual tongue picture image to be measured and interrogation data to be measured.

Predicted value obtains module 304 and is used for the tongue picture image to be measured and the interrogation number to be measured of the test individual According to the prediction model trained and obtained is input to, the blood constituent for obtaining the test individual based on prediction model parsing is pre- Measured value.

In one embodiment, further includes: blood component data obtains module, acquisition module, mode input module and pre- It surveys model and obtains module, in which:

Blood component data obtains module for collecting target blood compositional data.

Acquisition module is for collecting interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture image.

Mode input module is used for the target blood compositional data, the pre-collecting tongue picture image and the interrogation number According to being pre-processed.

Prediction model obtains module and is used to be based on machine learning model, parses the pre-collecting tongue picture image and the interrogation Data obtain the tongue picture feature and interrogation data characteristics of the pre-collecting tongue picture image, according to the pre-collecting tongue picture image Tongue picture feature, interrogation data characteristics and the training of target blood compositional data obtain the prediction model.

In one embodiment, it includes: testing data input unit, feature analysis acquisition list that the predicted value, which obtains module, Member and predicted value obtaining unit, in which:

Testing data input unit is used for the tongue picture image to be measured and the interrogation number to be measured of the test individual According to be input to trained obtain the prediction model.

Feature analysis obtaining unit is used to be based on the prediction model, parses the tongue picture image to be measured and described to be measured asks Data are examined, the tongue picture feature and the interrogation data characteristics to be measured of the tongue picture image to be measured are obtained.

Predicted value obtaining unit is used for special according to the tongue picture feature of the tongue picture image to be measured and the interrogation data to be measured Sign, obtains the blood constituent predicted value of the test individual.

In one embodiment, the predicted value obtaining unit is also used to determining special with the tongue picture of the tongue picture image to be measured It seeks peace the tongue picture feature and interrogation data characteristics of the immediate pre-collecting tongue picture image of the interrogation data characteristics to be measured, obtains Obtain the blood constituent predicted value of the test individual.

In one embodiment, the acquisition module includes: interrogation data capture unit, facial image acquiring unit and pre- Collect tongue picture image acquisition unit, in which:

Interrogation data capture unit is for obtaining the interrogation data corresponding with the target blood compositional data.

Facial image acquiring unit is for obtaining facial image.

Pre-collecting tongue picture image acquisition unit parses the facial image, obtains the pre-collecting tongue picture image.

In one embodiment, the machine learning module is also used to parse multiple described pre- based on machine learning model Tongue picture image is collected, the tongue picture feature of each pre-collecting tongue picture image is obtained, establishes each pre-collecting tongue picture image The corresponding relationship of tongue picture feature and the target blood compositional data.

Specific restriction about blood constituent value prediction meanss may refer to above for blood constituent value prediction technique Restriction, details are not described herein.Modules in above-mentioned blood constituent value prediction meanss can be fully or partially through software, hard Part and combinations thereof is realized.Above-mentioned each module can be embedded in the form of hardware or independently of in the processor in computer equipment, It can also be stored in a software form in the memory in computer equipment, execute the above modules in order to which processor calls Corresponding operation.

For example, a kind of blood constituent value prediction meanss based on machine learning, including can be by tongue picture image and interrogation data It imports the input module for calculating equipment and respectively or concentrates tongue picture image and interrogation data be stored in local device or cloud The memory module of storage.Blood constituent value prediction meanss further include store in advance machine learning training module, will with it is described The blood constituent value of tongue picture image and interrogation data matches, and the blood constituent value is exported to object generated.It is described Input module includes camera and the software for driving the camera.The output device includes intelligent movable device.

For example, a kind of blood constituent value prediction meanss based on machine learning, including blood constituent value prediction module and Multiple information exchange modules.One of information exchange module is connect with the tongue picture image collection module, for obtaining tongue picture Feature, and tongue picture feature is exported, another information exchange module and tongue picture feature output module and interrogation data export mould Block connection, is input to blood constituent value model for data.

For example, the blood constituent value prediction meanss based on machine learning include blood constituent value prediction result output module, The blood constituent value prediction result output module predicts operation mould according to tongue picture image and interrogation data, by blood constituent value Type generates blood constituent value results of measuring.

In one embodiment, a kind of computer equipment is provided, which can be server, internal junction Composition can be as shown in Figure 4.The computer equipment include by system bus connect processor, memory, network interface and Database.Wherein, the processor of the computer equipment is for providing calculating and control ability.The memory packet of the computer equipment Include non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system, computer program and data Library.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating The database of machine equipment is used for storage of blood signal component value prediction data.The network interface of the computer equipment is used for logical with server Cross network connection communication.To realize a kind of blood constituent value prediction technique when the computer program is executed by processor.

In one embodiment, a kind of computer equipment is provided, which can be terminal, internal structure Figure can be as shown in Figure 5.The computer equipment includes processor, the memory, network interface, display connected by system bus Screen and input unit.Wherein, the processor of the computer equipment is for providing calculating and control ability.The computer equipment is deposited Reservoir includes non-volatile memory medium, built-in storage.The non-volatile memory medium is stored with operating system and computer journey Sequence.The built-in storage provides environment for the operation of operating system and computer program in non-volatile memory medium.The calculating The network interface of machine equipment is used to communicate with external terminal by network connection.When the computer program is executed by processor with Realize a kind of blood constituent value prediction technique.The display screen of the computer equipment can be liquid crystal display or electric ink is aobvious Display screen, the input unit of the computer equipment can be the touch layer covered on display screen, be also possible to computer equipment shell Key, trace ball or the Trackpad of upper setting can also be external keyboard, Trackpad or mouse etc., can also be camera, The camera is for acquiring tongue picture image.

It will be understood by those skilled in the art that structure shown in Fig. 5, only part relevant to application scheme is tied The block diagram of structure does not constitute the restriction for the computer equipment being applied thereon to application scheme, specific computer equipment It may include perhaps combining certain components or with different component layouts than more or fewer components as shown in the figure.

In one embodiment, a kind of computer equipment, including memory and processor are provided, which is stored with Computer program, the processor perform the steps of when executing computer program

The tongue picture image to be measured of acquisition test individual and interrogation data to be measured.

In one embodiment, it is also performed the steps of when processor executes computer program

Collect target blood compositional data.

Collect interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture image.

The target blood compositional data, the pre-collecting tongue picture image and the interrogation data are pre-processed.

The tongue picture image to be measured of the test individual and the interrogation data to be measured are input to and trained acquisition The prediction model.

Based on the prediction model, the tongue picture image to be measured and the interrogation data to be measured are parsed, are obtained described to be measured The tongue picture feature of tongue picture image and the interrogation data characteristics to be measured.

Obtain the interrogation data corresponding with the target blood compositional data.

Obtain facial image.

In one embodiment, a kind of computer readable storage medium is provided, computer program is stored thereon with, is calculated Machine program performs the steps of when being executed by processor

In one embodiment, it is also performed the steps of when computer program is executed by processor

Collect target blood compositional data.

Obtain facial image.

Those of ordinary skill in the art will appreciate that realizing all or part of the process in above-described embodiment method, being can be with Relevant hardware is instructed to complete by computer program, the computer program can be stored in a non-volatile computer In read/write memory medium, the computer program is when being executed, it may include such as the process of the embodiment of above-mentioned each method.Wherein, To any reference of memory, storage, database or other media used in each embodiment provided herein, Including non-volatile and/or volatile memory.Nonvolatile memory may include read-only memory (ROM), programming ROM (PROM), electrically programmable ROM (EPROM), electrically erasable ROM (EEPROM) or flash memory.Volatile memory may include Random access memory (RAM) or external cache.By way of illustration and not limitation, RAM is available in many forms, Such as static state RAM (SRAM), dynamic ram (DRAM), synchronous dram (SDRAM), double data rate sdram (DDRSDRAM), enhancing Type SDRAM (ESDRAM), synchronization link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic ram (DRDRAM) and memory bus dynamic ram (RDRAM) etc..

Each technical characteristic of above embodiments can be combined arbitrarily, for simplicity of description, not to above-described embodiment In each technical characteristic it is all possible combination be all described, as long as however, the combination of these technical characteristics be not present lance Shield all should be considered as described in this specification.

The several embodiments of the application above described embodiment only expresses, the description thereof is more specific and detailed, but simultaneously It cannot therefore be construed as limiting the scope of the patent.It should be pointed out that coming for those of ordinary skill in the art It says, without departing from the concept of this application, various modifications and improvements can be made, these belong to the protection of the application Range.Therefore, the scope of protection shall be subject to the appended claims for the application patent.

Claims

1. a kind of blood constituent value prediction technique, which is characterized in that the described method includes:

The tongue picture image to be measured of the test individual and the interrogation data to be measured are input to the prediction trained and obtained Model obtains the blood constituent predicted value of the test individual based on prediction model parsing.

2. the method according to claim 1, wherein the tongue picture image to be measured of test individual and to be measured of obtaining Before the step of interrogation data further include:

Collect target blood compositional data；

Based on machine learning model, the pre-collecting tongue picture image and the interrogation data are parsed, obtains the pre-collecting tongue picture The tongue picture feature and interrogation data characteristics of image, according to the tongue picture feature of the pre-collecting tongue picture image, interrogation data characteristics and The training of target blood compositional data obtains the prediction model.

3. the method according to claim 1, wherein the tongue picture image to be measured by the test individual It is input to the prediction model trained and obtained with the interrogation data to be measured, is obtained based on prediction model parsing described to be measured Individual blood constituent predicted value the step of include:

The tongue picture image to be measured of the test individual and the interrogation data to be measured are input to and trained described in acquisition Prediction model；

Based on the prediction model, the tongue picture image to be measured and the interrogation data to be measured are parsed, obtain the tongue picture to be measured The tongue picture feature of image and the interrogation data characteristics to be measured；

According to the tongue picture feature of the tongue picture image to be measured and the interrogation data characteristics to be measured, the blood of the test individual is obtained Liquid ingredient prediction value.

4. according to the method described in claim 3, it is characterized in that, the tongue picture feature according to the tongue picture image to be measured and The interrogation data characteristics to be measured, the step of obtaining the blood constituent predicted value of the test individual include:

Determining tongue picture feature and the immediate pre-collecting of interrogation data characteristics to be measured with the tongue picture image to be measured The tongue picture feature and interrogation data characteristics of tongue picture image, obtain the blood constituent predicted value of the test individual.

5. the method according to claim 1, wherein the collection is corresponding with the target blood compositional data The step of interrogation data and pre-collecting tongue picture image includes:

Obtain facial image；

6. a kind of blood constituent value prediction meanss, which is characterized in that described device includes:

Predicted value obtains module, for inputting the tongue picture image to be measured of the test individual and the interrogation data to be measured To the prediction model obtained has been trained, the blood constituent predicted value of the test individual is obtained based on prediction model parsing.

7. the apparatus according to claim 1, which is characterized in that further include:

Acquisition module, for collecting interrogation data corresponding with the target blood compositional data and pre-collecting tongue picture image；

Mode input module, for the target blood compositional data, the pre-collecting tongue picture image and the interrogation data It is pre-processed；

Prediction model obtains module, for being based on machine learning model, parses the pre-collecting tongue picture image and the interrogation number According to the tongue picture feature and interrogation data characteristics of the pre-collecting tongue picture image being obtained, according to the tongue of the pre-collecting tongue picture image As the training of feature, interrogation data characteristics and target blood compositional data obtains the prediction model.

8. the apparatus according to claim 1, which is characterized in that the predicted value obtains module and includes:

Testing data input unit, for the tongue picture image to be measured of the test individual and the interrogation data to be measured is defeated Enter to the prediction model for having trained acquisition；

Feature analysis obtaining unit parses the tongue picture image to be measured and the interrogation to be measured for being based on the prediction model Data obtain the tongue picture feature and the interrogation data characteristics to be measured of the tongue picture image to be measured；

Predicted value obtaining unit, for the tongue picture feature and the interrogation data characteristics to be measured according to the tongue picture image to be measured, Obtain the blood constituent predicted value of the test individual.

9. a kind of computer equipment, including memory and processor, the memory are stored with computer program, feature exists In the step of processor realizes any one of claims 1 to 5 the method when executing the computer program.

10. a kind of computer readable storage medium, is stored thereon with computer program, which is characterized in that the computer program The step of method described in any one of claims 1 to 5 is realized when being executed by processor.