A TCM disease auxiliary diagnosis system, device and storage medium

Abstract

The present invention proposes a TCM disease auxiliary diagnosis system, equipment and storage medium. The data acquisition module obtains the user's related TCM medical records associated text information; the entity knowledge acquisition module automatically obtains the entity knowledge in the text in the associated text information through the BiLSTM-CRF model; the knowledge graph module constructs a medical diagnosis and treatment knowledge graph database; the TCM diagnosis calculation module calculates the cosine similarity between the entity knowledge and the TCM disease symptom nodes in the knowledge graph to obtain the most relevant TCM disease name; the inspection image recognition module identifies the biochemical examination image documents of the same user and outputs them as standard unified data; the dialectical typing auxiliary module performs dialectical typing prediction on the data output by the inspection image recognition module; the disease analysis report module generates a disease risk report and diagnosis and treatment recommendations based on the disease dialectical typing results. The present invention realizes the effective combination of the objective data of human biological characteristics of Western medicine and the precise diagnosis of TCM diseases, and gives full play to the auxiliary role of the objective indicator data of Western medicine.

Description

Technical Field

[0001] The present invention relates to the technical field of TCM auxiliary diagnosis, and in particular to a TCM disease auxiliary diagnosis system, equipment and storage medium. Background Art

[0002] Traditional Chinese medicine is a brilliant treasure of Chinese civilization. It has thousands of years of history and clinical practice. It has achieved remarkable results in the fields of disease treatment, disease prevention and health preservation, and provides important protection for human health. The diagnostic method of traditional Chinese medicine emphasizes "differentiation and treatment", that is, syndrome analysis is performed based on clinical manifestations, and then prescriptions are prescribed according to different syndromes. The composition of traditional Chinese medicine is complex, and different traditional Chinese medicine agents have different effects on diseases of different syndromes of the same disease. Due to the complexity of the theoretical system of traditional Chinese medicine and the long talent training cycle, the clinical diagnosis and application of traditional Chinese medicine still rely to a large extent on the experience of the ancients, which makes disease diagnosis and even disease classification subject to relatively subjective judgment. Modern medical testing equipment has made great progress, realizing human biological characteristic data, such as blood routine, urine routine, liver and kidney function, electrolytes, immunity, tumor markers, etc., and testing has become very simple and convenient.

[0003] How to effectively combine the objective data of human biological characteristics of Western medicine with the precise diagnosis of diseases in traditional Chinese medicine and give full play to the auxiliary role of objective indicator data of Western medicine has become an important issue that people urgently need to think about and solve. Summary of the invention

[0004] In view of this, the present invention proposes a TCM disease auxiliary diagnosis system to solve the problem of effectively combining objective data of human biological characteristics of Western medicine with accurate diagnosis of TCM diseases.

[0005] The technical solution of the present invention is achieved in this way:

[0006] In a first aspect, the present invention discloses a TCM disease auxiliary diagnosis system, the system comprising the following modules:

[0007] Data acquisition module: used to obtain the user's related TCM medical records associated text information, the associated text information includes the medical records information recorded by the doctor when the user visits the doctor, the medical records information includes the patient's own information, tongue information, complexion information, voice information, pulse information and detailed symptom description;

[0008] Entity knowledge acquisition module: used to automatically acquire entity knowledge in the text through the BiLSTM-CRF model for the associated text information;

[0009] Knowledge graph module: used to construct a TCM diagnosis and treatment knowledge graph database, which includes nodes for TCM disease names, TCM disease symptoms, TCM disease introductions, medicated diets, TCM prescriptions, Western medicine disease names, and TCM disease medical records, and stores the associations between TCM and Western medicine disease names, TCM diseases and TCM symptoms, TCM diseases and basic medical records, TCM diseases and disease classification, classification diseases and medicated diets, and classification diseases and prescriptions;

[0010] TCM diagnosis calculation module: used to represent the entity knowledge in the entity knowledge acquisition module and the knowledge graph in the knowledge graph module using VSM coding, calculate the cosine similarity of the entity knowledge and the TCM disease symptom nodes in the knowledge graph, and then obtain the most relevant TCM disease name for the user;

[0011] Inspection image recognition module: used to identify the user's biochemical inspection image documents and output them as text and numerical data in a standard format and unified latitude;

[0012] Dialectical classification auxiliary module: used to build a multi-classification model based on convolutional neural network to perform dialectical classification prediction on the data output by the inspection image recognition module;

[0013] Disease analysis report module: used to generate disease risk reports and diagnosis and treatment recommendations based on the disease dialectical typing results calculated by the dialectical typing auxiliary module.

[0014] The present invention constructs a TCM disease auxiliary diagnosis system through the above method, effectively combines the objective data of human biological characteristics of Western medicine with the precise diagnosis of TCM diseases, and plays the auxiliary role of the objective indicator data of Western medicine to help achieve accurate TCM disease identification and diagnosis.

[0015] On the basis of the above technical solution, preferably, the entity knowledge specifically includes:

[0016] Patient identity information: including gender and age;

[0017] Diagnosis and treatment information: including disease site, complexion, tongue, pulse, voice, work and rest, diet, TCM disease symptoms, and disease duration.

[0018] The present invention adopts the above method,

[0019] On the basis of the above technical solution, preferably, when the entity knowledge acquisition module acquires the TCM disease symptoms in the associated text information, the BiLSTM layer in the BiLSTM-CRF model outputs the score of each label, i.e., the TCM disease symptom, and sets the score as a weight to divide the main symptom and the auxiliary symptom;

[0020] The CRF layer defines a score as the output tag sequence y corresponding to the input sequence X, where each score corresponds to a complete path; the Viterbi algorithm is used to predict and solve the optimal path. The optimal path is the final prediction result of the output sequence, and the final prediction result is the probability value of each word mapped to the label.

[0021] The present invention uses the BiLSTM-CRF model through the above method to complete automatic real-time extraction of entity knowledge. It not only takes into account the correlation before and after sequence annotation through the CRF layer, but also has the powerful feature extraction and fitting capabilities of the deep network.

[0022] Based on the above technical solution, preferably, the nodes and association relationships in the knowledge graph are stored in a Neo4j graph database.

[0023] The present invention constructs a knowledge graph based on TCM disease information through the above method. The knowledge graph covers the relevant knowledge of authoritative professionals in the field of TCM and is stored in the Neo4j graph database.

[0024] On the basis of the above technical solution, preferably,

[0025] The TCM diagnosis calculation module converts the entity knowledge terms obtained by the entity knowledge acquisition module into a space vector V d =D(d 1 , d 2 , …d n ), where n represents the vector dimension, d n Indicates the n-th dimension feature words;

[0026] The TCM diagnosis calculation module converts the association between TCM diseases and TCM symptoms, and TCM diseases and basic medical records in the knowledge graph module into a spatial vector set K={k 1 , k 2 , …k m}, k m Represents the space vector corresponding to the mth TCM disease, namely T(t 1 , t 2 ,…t n ), calculate V d The cosine similarity with each vector in the vector set K, where the TCM disease corresponding to the maximum similarity is regarded as the most relevant TCM disease name inferred based on the entity knowledge acquired by the entity knowledge acquisition module.

[0027] The present invention uses the above method to perform reasoning diagnosis on the entity knowledge through a TCM diagnosis calculation module, and obtains the name of the TCM disease that the patient is most likely to suffer from based on the patient's TCM disease symptom information.

[0028] On the basis of the above technical solution, preferably,

[0029] The inspection image recognition module converts the biochemical inspection document image into a grayscale image, determines whether the character arrangement direction in the text area in the grayscale image is horizontal, and if it is horizontal, automatically determines the foreground color of each single-line text area image to obtain a binary image; segments the binary image to obtain a character fragment sequence; performs text line recognition on the segmented single-line binary image; and groups the inspection values ​​in a two-tuple format according to the inspection item name, and returns the data in a json format;

[0030] Before image recognition, abnormal situations in which the detection values ​​are not numerical need to be processed intelligently. For example, the detection values ​​represented by symbolic arrows such as increase and decrease should be occluded and transformed and switched to "0" or "1".

[0031] The present invention extracts data from biochemical examination documents through the above method, thereby obtaining objective data of human biological characteristics obtained by western medical examination.

[0032] On the basis of the above technical solution, preferably,

[0033] The syndrome differentiation auxiliary module collects biological sign data of each syndrome type under the TCM disease in a targeted manner, performs typing prediction, inputs biological sign index data corresponding to each syndrome type under the TCM disease during the network model training, and performs linear function normalization processing on the index data;

[0034] A convolutional neural network model is constructed, including 1 input layer, 5 convolutional layers, 5 pooling layers, 1 fully connected layer and 1 output classification layer; the convolution kernel size of the first convolutional layer in the convolutional neural network model is 5*5, the convolution kernel size of other convolutional layers is 3*3, and the activation function adopts an unsaturated nonlinear function f(x)=max(0,x); the pooling layer size is 1*2, and the maximum downsampling method is adopted; the number of neurons in the fully connected layer is 512; the output layer uses a Softmax classifier, and the output obtains all probabilities of each syndrome type under the TCM disease;

[0035] The convolutional neural network model parameters are randomly initialized, and the constructed multi-classification model is used to learn and train the training sample set data. The sample labels are the TCM disease dialectical classification.

[0036] Using cross entropy cost function The Adam method is used to update the parameters and optimize the convolutional neural network model, where j is the number of output vectors; y is the target output vector; and x is the actual output vector;

[0037] Use the optimized convolutional neural network model to perform dialectical typing prediction on the JSON format data of the same user returned by the inspection image recognition module.

[0038] Through the above method, the present invention effectively combines the objective data of human biological characteristics of Western medicine with the precise diagnosis of diseases in Traditional Chinese Medicine, and plays the role of Western medicine diagnostic examination related data in assisting traditional Chinese medicine diagnosis.

[0039] On the basis of the above technical solution, preferably,

[0040] The disease analysis report module searches the atlas database for related content under the disease classification according to the dialectical classification results of TCM diseases, obtains the related medicinal diet and TCM prescriptions corresponding to the results, generates TCM disease risk reports and diagnosis and treatment recommendations, and pushes them.

[0041] The present invention obtains the TCM diseases that the user may suffer from through the above method, and selects appropriate medicated food, TCM prescriptions, etc. accordingly to achieve symptomatic treatment.

[0042] In a second aspect of the present invention, an electronic device is disclosed, comprising: at least one processor, at least one memory, a communication interface and a bus; wherein the processor, memory and communication interface communicate with each other via the bus; the memory stores a working program of a TCM disease auxiliary diagnosis system that can be executed by the processor, and the working program of a TCM disease auxiliary diagnosis system is configured to implement a working program of a TCM disease auxiliary diagnosis system as described in the first aspect of the present invention.

[0043] In a third aspect of the present invention, a computer-readable storage medium is disclosed, on which a working program of a TCM disease auxiliary diagnosis system is stored. When the working program of a TCM disease auxiliary diagnosis system is executed, the working program of a TCM disease auxiliary diagnosis system as described in the first aspect of the present invention is implemented.

[0044] The TCM disease auxiliary diagnosis system method of the present invention has the following beneficial effects compared with the prior art:

[0045] (1) The system analyzes and processes the user's medical record information and combines it with relevant Western medicine and chemical examination data to automatically complete the dialectical classification of the disease in the medical record, thereby ensuring the selection of appropriate medicinal diets and Chinese medicine prescriptions, etc., overcoming the defects of low credibility and poor repeatability of TCM diagnosis results caused by personal subjectivity in traditional TCM manual diagnosis;

[0046] (2) By effectively combining the objective data of human biological characteristics in Western medicine with the precise diagnosis of diseases in Traditional Chinese Medicine, the correlation between Western medicine chemical examination data and the dialectical classification of diseases in Traditional Chinese Medicine is fully utilized, and the role of Western medicine diagnostic examination data in assisting traditional Chinese medicine diagnosis is brought into play, which helps to improve the accuracy of TCM disease diagnosis. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0048] Figure 1 This is a module diagram of a TCM disease auxiliary diagnosis system of the present invention. DETAILED DESCRIPTION

[0049] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0050] Example

[0051] The workflow of the TCM disease auxiliary diagnosis system of the present invention is shown in Figure 1 , the processing steps are as follows:

[0052] The first step is to obtain the user's related TCM medical records through the data acquisition module. The related text information includes the medical records recorded by the doctor when the user visits the doctor. The medical records include the patient's own information, tongue information, complexion information, voice information, pulse information and detailed symptom description. Go to the second step.

[0053] It should be understood that TCM medical records are usually basic information, tongue information, complexion information, voice information, pulse information, and detailed symptom descriptions written down by professional TCM doctors when users go to professional TCM hospitals for treatment. For clinical diagnosis of TCM diseases, doctors only need to pay attention to the entity knowledge in the medical record text, such as age, gender, diet, sleep, complexion, tongue, pulse, symptoms, etc.

[0054] In the second step, the BiLSTM-CRF model in the entity knowledge acquisition module is used to automatically acquire entity knowledge in the text for the associated text information. Then, the process proceeds to the third step.

[0055] It should be understood that when the entity knowledge acquisition module acquires the TCM disease symptoms in the associated text information, the BiLSTM layer in the BiLSTM-CRF model outputs the score of each label, i.e., the TCM disease symptom, and sets the score as a weight to divide the main symptom and the auxiliary symptom;

[0056] The CRF layer defines a score as the output tag sequence y corresponding to the input sequence X, where each score corresponds to a complete path; the Viterbi algorithm is used to predict and solve the optimal path. The optimal path is the final prediction result of the output sequence, and the final prediction result is the probability value of each word mapped to the label.

[0057] The present invention uses the above scheme and utilizes the sequence labeling algorithm CRF in the BiLSTM-CRF model, which can be used for tasks such as word segmentation, part-of-speech tagging, and named entity recognition, to output the respective scores of all labels of each word, which is equivalent to the probability value of each word being mapped to the label; the BiLSTM-CRF model automatically extracts entity knowledge in real time, and its accuracy rate can reach 96.3% based on test data. It not only takes into account the correlation before and after sequence labeling through the CRF layer, but also has the powerful feature extraction and fitting capabilities of the deep network.

[0058] The third step is to construct a TCM diagnosis and treatment knowledge graph database through the knowledge graph module. The TCM diagnosis and treatment knowledge graph includes nodes for TCM disease names, TCM disease symptoms, TCM disease introductions, medicated diets, TCM prescriptions, Western medicine disease names, and TCM disease medical records. It stores the associations between TCM and Western medicine disease names, TCM diseases and TCM symptoms, TCM diseases and basic medical records, TCM diseases and disease classification, classification diseases and medicated diets, and classification diseases and prescriptions. The nodes and associations in the knowledge graph are stored in the Neo4j graph database. Go to the fourth step.

[0059] It should be understood that the knowledge graph is a networked knowledge system built with a "semantic network" as the framework, which aims to describe the concepts, entities, events and their relationships in the objective world. The present invention collects and processes the above-mentioned graph nodes, graph relationships and other diagnosis and treatment knowledge covered by the "Guidelines for Clinical Research of New Chinese Medicines", "Guidelines for Research of New Chinese Medicines", "Part on Syndrome Types of Clinical Diagnosis and Treatment Terminology of Traditional Chinese Medicine", "Standards for Diagnosis and Efficacy of Traditional Chinese Medicine Diseases", and "Guidelines for Clinical Diagnosis and Treatment of Traditional Chinese Medicine", and stores them in the Neo4j graph database.

[0060] In the fourth step, the TCM diagnosis calculation module uses VSM coding to represent the entity knowledge in the entity knowledge acquisition module and the knowledge graph in the knowledge graph module, and calculates the cosine similarity of the entity knowledge and the TCM disease symptom nodes in the knowledge graph, thereby obtaining the most relevant TCM disease name for the user. Go to the fifth step.

[0061] It should be understood that the primary task of Chinese computing in the field of natural language is to digitize text. It is very important to uniformly convert the computational text from a global perspective and maintain the consistency of Chinese semantics. The vector space model (VSM) has become a commonly used text representation model in natural language processing, and was first used in the SMART information retrieval system. Since the entity knowledge acquired by the entity knowledge acquisition module and the graph nodes in the knowledge graph module are words rather than text sentences, and the content of each part is neither repeated nor considers the correlation between words, it meets the assumptions of the vector space model.

[0062] Therefore, the TCM diagnosis calculation module converts the entity knowledge terms acquired by the entity knowledge acquisition module into a space vector V d =D(d 1 , d 2 , …d n ), where n represents the vector dimension, d n Indicates the n-th dimension feature words;

[0063] The TCM diagnosis calculation module converts the association between TCM diseases and TCM symptoms, and TCM diseases and basic medical records in the knowledge graph module into a spatial vector set K={k 1 , k 2 , …k m}, k m Represents the space vector corresponding to the mth TCM disease, namely T(t 1 , t 2 ,…t n ), calculate V d The cosine similarity with each vector in the vector set K, where the TCM disease corresponding to the maximum similarity is regarded as the most relevant TCM disease obtained by reasoning based on the entity knowledge obtained by the entity knowledge acquisition module, and the corresponding TCM disease name is obtained through the association relationship in the knowledge graph.

[0064] Step 6: Use the inspection image recognition module to identify the user's biochemical inspection image receipt and output it as text and numerical data in a standard format and unified latitude. Go to step 7.

[0065] It should be understood that the biochemical examination image documents include blood routine, urine routine, liver and kidney function examination sheets; quickly and accurately identifying the uploaded examination images as text data is a prerequisite for the subsequent dialectical typing module. Therefore, in order to achieve efficient recognition, the examination image recognition module converts the biochemical examination document image into a grayscale image, determines whether the character arrangement direction in the text area in the grayscale image is horizontal, and if it is horizontal, automatically determines the foreground color for each single-line text area image to obtain a binary image; segment the binary image to obtain a character fragment sequence; perform text line recognition on the segmented single-line binary image; group the test values ​​in a binary format according to the test item name, and return the data in json format;

[0066] Before image recognition, abnormal situations in which the detection values ​​are not numerical need to be processed intelligently. For example, the detection values ​​represented by symbolic arrows such as increase and decrease should be occluded and transformed and switched to "0" or "1".

[0067] Step 7: The dialectical classification auxiliary module constructs a multi-classification model based on a convolutional neural network to perform dialectical classification prediction on the data output by the inspection image recognition module. Go to step 8.

[0068] It should be understood that the syndrome differentiation auxiliary module collects the biological sign data of each syndrome under the TCM disease in a targeted manner, performs typing prediction, inputs the biological sign index data corresponding to each syndrome under the TCM disease during the network model training, and performs linear function normalization processing on the index data;

[0069] The syndromes include: hypertension of liver fire hyperactivity type, phlegm dampness stagnation type, yin deficiency yang hyperactivity type, yin and yang deficiency type; hyperlipidemia of phlegm turbidity obstruction type, spleen and kidney yang deficiency type, liver and kidney yin deficiency type, yin deficiency yang hyperactivity type, qi stagnation and blood stasis type; gout of phlegm stasis type, damp heat stasis type, spleen deficiency dampness type, qi and blood deficiency type, etc.

[0070] The index data include: blood pressure (SBP, DBP), blood sugar (Glu, HbA1c), blood lipids (TC, TG, LDL-C, HDL-C, ApoA1, ApoB, apoA1 / apoB, ApoE, LP-a, VLDL-C), liver and kidney function (AST, ALT, GGT, ALP, CHE, TBilL, DB, TBA, TP, Cr, Alb, GLO, A / G, pa, BUN, GFR, MALB), blood routine (RBC, RDW, WBC, GR, LY, MO, PLT, Hb, NLR, MPV, EOS, Baso, neut, NEUT, LY, CRP), urine routine (RBC_U, U-WBC, 24h urine protein quantification, β2-MG), myocardial enzymes (HBDH, LDH), electrolytes (na, k, cl, Ca), etc.;

[0071] A convolutional neural network model is constructed, including 1 input layer, 5 convolutional layers, 5 pooling layers, 1 fully connected layer and 1 output classification layer; the convolution kernel size of the first convolutional layer in the convolutional neural network model is 5*5, the convolution kernel size of other convolutional layers is 3*3, and the activation function adopts an unsaturated nonlinear function f(x)=max(0,x); the pooling layer size is 1*2, and the maximum downsampling method is adopted; the number of neurons in the fully connected layer is 512; the output layer uses a Softmax classifier, and the output obtains all probabilities of each syndrome type under the TCM disease;

[0072] The convolutional neural network model parameters are randomly initialized, and the constructed multi-classification model is used to learn and train the training sample set data. The sample labels are the TCM disease dialectical classification.

[0073] Using cross entropy cost function The Adam method is used to update the parameters and optimize the convolutional neural network model, where j is the number of output vectors; y is the target output vector; and x is the actual output vector;

[0074] The optimized convolutional neural network model is used to perform dialectical typing prediction on the json format data of the same user returned by the examination image recognition module, so as to give full play to the role of western medicine diagnosis and examination related data in assisting traditional Chinese medicine diagnosis.

[0075] Through the above steps, the present invention effectively combines the objective data of human biological characteristics of Western medicine with the precise diagnosis of diseases in Traditional Chinese Medicine, fully utilizes the correlation between Western medicine chemical examination data and TCM disease dialectical classification, and plays the role of Western medicine diagnostic examination related data in assisting traditional TCM diagnosis.

[0076] In the eighth step, the disease analysis report module generates a disease risk report and diagnosis and treatment recommendations based on the disease dialectical typing results calculated by the dialectical typing auxiliary module.

[0077] It should be understood that the disease analysis report module searches the atlas database for related content under the disease classification based on the results of the dialectical classification of TCM diseases, obtains the related medicinal foods and Chinese medicine prescriptions corresponding to the results, generates a TCM disease risk report and treatment recommendations, and pushes them.

[0078] The first seven steps in the embodiment of the present invention realize the combination of traditional Chinese medicine knowledge and Western medicine data to automatically complete the disease dialectical classification based on the user's medical records. The eighth step selects appropriate medicated food, Chinese medicine prescriptions, etc. through the knowledge graph, generates a disease risk report and diagnosis and treatment recommendations, and achieves symptomatic treatment.

[0079] The present invention also discloses an electronic device, comprising: at least one processor, at least one memory, a communication interface and a bus; wherein the processor, memory and communication interface communicate with each other via the bus; the memory stores a working program of a TCM disease auxiliary diagnosis system that can be executed by the processor, and the working program of a TCM disease auxiliary diagnosis system is configured to implement a working program of a TCM disease auxiliary diagnosis system as described in an embodiment of the present invention.

[0080] The present invention also discloses a computer-readable storage medium, on which is stored a working program of a TCM disease auxiliary diagnosis system. When the working program of a TCM disease auxiliary diagnosis system is executed, a working program of a TCM disease auxiliary diagnosis system as described in an embodiment of the present invention is implemented.

[0081] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A TCM disease auxiliary diagnosis system, characterized in that: The TCM disease auxiliary diagnosis system comprises the following modules: Data acquisition module: used to obtain the user's related TCM medical records associated text information, the associated text information includes the medical records information recorded by the doctor when the user visits the doctor, the medical records information includes the patient's own information, tongue information, complexion information, voice information, pulse information and detailed symptom description; Entity knowledge acquisition module: used to automatically acquire entity knowledge in the text through the BiLSTM-CRF model for the associated text information; Knowledge graph module: used to construct a TCM diagnosis and treatment knowledge graph database, which includes nodes for TCM disease names, TCM disease symptoms, TCM disease introductions, medicated diets, TCM prescriptions, Western medicine disease names, and TCM disease medical records, and stores the associations between TCM and Western medicine disease names, TCM diseases and TCM symptoms, TCM diseases and basic medical records, TCM diseases and disease classification, classification diseases and medicated diets, and classification diseases and prescriptions; TCM diagnosis calculation module: used to represent the entity knowledge in the entity knowledge acquisition module and the knowledge graph in the knowledge graph module using VSM coding, calculate the cosine similarity of the entity knowledge and the TCM disease symptom nodes in the knowledge graph, and then obtain the most relevant TCM disease name for the user; Inspection image recognition module: used to identify the user's biochemical inspection image documents and output them as text and numerical data in a standard format and unified latitude; Dialectical classification auxiliary module: used to build a multi-classification model based on convolutional neural network to perform dialectical classification prediction on the data output by the inspection image recognition module; Disease analysis report module: used to generate disease risk reports and diagnosis and treatment recommendations based on the disease dialectical typing results calculated by the dialectical typing auxiliary module; The inspection image recognition module converts the biochemical inspection document image into a grayscale image, determines whether the character arrangement direction in the text area in the grayscale image is horizontal, and if it is horizontal, automatically determines the foreground color of each single-line text area image to obtain a binary image; segments the binary image to obtain a character fragment sequence; performs text line recognition on the segmented single-line binary image; and groups the inspection values ​​in a two-tuple format according to the inspection item name, and returns the data in a json format; Before image recognition, the abnormal situation of non-numerical detection values ​​needs to be processed intelligently. The detection values ​​represented by the symbolic arrows should be blocked and transformed to "0" or "1" to indicate the increase or decrease. The syndrome differentiation auxiliary module collects biological sign data of each syndrome type under the TCM disease in a targeted manner, performs typing prediction, inputs biological sign index data corresponding to each syndrome type under the TCM disease during the network model training, and performs linear function normalization processing on the index data; A convolutional neural network model is constructed, including 1 input layer, 5 convolutional layers, 5 pooling layers, 1 fully connected layer and 1 output classification layer; the convolution kernel size of the first convolutional layer in the convolutional neural network model is 5×5, the convolution kernel size of other convolutional layers is 3×3, and the activation function adopts an unsaturated nonlinear function f(x)=max(0,x); the pooling layer size is 1×2, and the maximum downsampling method is adopted; the number of neurons in the fully connected layer is 512; the output layer uses a Softmax classifier, and the output obtains all probabilities of each syndrome type under the TCM disease; The convolutional neural network model parameters are randomly initialized, and the constructed multi-classification model is used to learn and train the training sample set data. The sample labels are the TCM disease dialectical classification. Using cross entropy cost function , the Adam method is used to update the parameters and optimize the convolutional neural network model, where j is the number of output vectors; y is the target output vector; x is the actual output vector; Use the optimized convolutional neural network model to perform dialectical typing prediction on the JSON format data of the same user returned by the inspection image recognition module.

2. A TCM disease auxiliary diagnosis system as claimed in claim 1, characterized in that: The entity knowledge specifically includes: Patient identity information: including gender and age; Diagnosis and treatment information: including disease site, complexion, tongue, pulse, voice, work and rest, diet, TCM disease symptoms, and disease duration.

3. A TCM disease auxiliary diagnosis system as claimed in claim 2, characterized in that: When the entity knowledge acquisition module acquires the TCM disease symptoms in the associated text information, the BiLSTM layer in the BiLSTM-CRF model outputs the score of each label, i.e., the TCM disease symptom, and sets the score as a weight to divide the main symptom and the auxiliary symptom; The CRF layer defines a score as the output tag sequence y corresponding to the input sequence X, where each score corresponds to a complete path; the Viterbi algorithm is used to predict and solve the optimal path. The optimal path is the final prediction result of the output sequence, and the final prediction result is the probability value of each word mapped to the label.

4. A TCM disease auxiliary diagnosis system as claimed in claim 1, characterized in that: The nodes and relationships in the knowledge graph are stored in the Neo4j graph database.

5. A TCM disease auxiliary diagnosis system as claimed in claim 1, characterized in that: The TCM diagnosis calculation module converts the entity knowledge terms obtained by the entity knowledge acquisition module into space vectors , where n represents the vector dimension, Indicates the n-th dimension feature words; The TCM diagnosis calculation module converts the association between TCM diseases and TCM symptoms, and TCM diseases and basic medical records in the knowledge graph module into a spatial vector set , Represents the space vector corresponding to the mth TCM disease, namely , respectively calculated Vector collection with The cosine similarity of each vector in the entity knowledge acquisition module is calculated, and the TCM disease corresponding to the maximum similarity is regarded as the most relevant TCM disease name obtained by reasoning based on the entity knowledge obtained by the entity knowledge acquisition module.

6. A TCM disease auxiliary diagnosis system as claimed in claim 1, characterized in that: The disease analysis report module retrieves the related content under the disease classification in the atlas database based on the dialectical classification results of the TCM disease, obtains the related medicinal diet and Chinese medicine prescriptions corresponding to the results, generates a TCM disease risk report and diagnosis and treatment suggestions, and pushes them.

7. An electronic device, characterized in that: It includes at least one processor, at least one memory, a communication interface and a bus; wherein the processor, memory and communication interface communicate with each other through the bus; the memory stores a working program of a TCM disease auxiliary diagnosis system that can be executed by the processor, and a method program of a TCM disease auxiliary diagnosis system is configured to implement a working program of a TCM disease auxiliary diagnosis system as claimed in any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The storage medium stores a working program of a TCM disease-aided diagnosis system, which, when executed, implements a working program of a TCM disease-aided diagnosis system as claimed in any one of claims 1 to 6.

Images