CN112489769B - Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network - Google Patents
Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network Download PDFInfo
- Publication number
- CN112489769B CN112489769B CN201910776695.5A CN201910776695A CN112489769B CN 112489769 B CN112489769 B CN 112489769B CN 201910776695 A CN201910776695 A CN 201910776695A CN 112489769 B CN112489769 B CN 112489769B
- Authority
- CN
- China
- Prior art keywords
- neural network
- module
- deep neural
- chinese medicine
- diagnosis
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 239000003814 drug Substances 0.000 title claims abstract description 145
- 238000003745 diagnosis Methods 0.000 title claims abstract description 38
- 208000017667 Chronic Disease Diseases 0.000 title claims abstract description 29
- 238000013528 artificial neural network Methods 0.000 title claims abstract description 25
- 208000024891 symptom Diseases 0.000 claims abstract description 38
- 229940079593 drug Drugs 0.000 claims abstract description 34
- 238000000034 method Methods 0.000 claims abstract description 33
- 238000003062 neural network model Methods 0.000 claims abstract description 32
- 239000013610 patient sample Substances 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims description 50
- 230000006870 function Effects 0.000 claims description 39
- 238000007781 pre-processing Methods 0.000 claims description 36
- 230000011218 segmentation Effects 0.000 claims description 35
- 238000011176 pooling Methods 0.000 claims description 28
- 239000000523 sample Substances 0.000 claims description 26
- 238000013527 convolutional neural network Methods 0.000 claims description 18
- 230000008569 process Effects 0.000 claims description 17
- 230000004913 activation Effects 0.000 claims description 15
- 238000013079 data visualisation Methods 0.000 claims description 11
- 239000013598 vector Substances 0.000 claims description 11
- 125000004122 cyclic group Chemical group 0.000 claims description 3
- 238000013500 data storage Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 210000002569 neuron Anatomy 0.000 claims description 3
- 238000012795 verification Methods 0.000 claims description 3
- 238000011156 evaluation Methods 0.000 claims description 2
- 238000011282 treatment Methods 0.000 abstract description 16
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 abstract description 9
- 230000036541 health Effects 0.000 abstract description 9
- 201000010099 disease Diseases 0.000 abstract description 8
- 238000004092 self-diagnosis Methods 0.000 abstract description 7
- 230000001684 chronic effect Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 6
- 238000013461 design Methods 0.000 description 5
- 238000012360 testing method Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 4
- 230000032683 aging Effects 0.000 description 3
- 238000004458 analytical method Methods 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 230000000007 visual effect Effects 0.000 description 3
- 238000010276 construction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000011369 optimal treatment Methods 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000013136 deep learning model Methods 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 230000005802 health problem Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 238000004393 prognosis Methods 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/90—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to alternative medicines, e.g. homeopathy or oriental medicines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/35—Clustering; Classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Public Health (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Databases & Information Systems (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Molecular Biology (AREA)
- Pathology (AREA)
- Alternative & Traditional Medicine (AREA)
- Pharmacology & Pharmacy (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a chronic disease intelligent traditional Chinese medicine diagnosis and medicine recommendation system based on a deep neural network, which comprises a traditional Chinese medicine condition reading module, a diagnosis module and a diagnosis module, wherein the traditional Chinese medicine condition reading module is used for reading symptom information of a patient sample; the invention constructs a database module for storing symptom information of patient samples and obtained drug information; the traditional Chinese medicine condition reading module is used for reading symptom information of a patient sample, predicting input information through the deep neural network model, and obtaining classification results of related medicines to form prediction information; the medicine recommendation module is used for recommending medicines, so that intelligent traditional Chinese medicine diagnosis and medicine recommendation of chronic diseases are realized, doctors are assisted in diagnosis and treatment, and especially, doctors are assisted in providing the best treatment method for patients, so that the cost is saved; the disease diagnosis accuracy rate can be improved, the death rate can be reduced, the hospital pressure can be relieved, and the like; the medical device helps chronic patients to realize self diagnosis and treatment to a certain extent, saves national medical resources and improves the health level of the whole people.
Description
Technical Field
The invention belongs to the technical field of intelligent medical treatment, and particularly relates to a chronic disease intelligent traditional Chinese medicine diagnosis and medicine recommendation system based on a deep neural network.
Background
The aging situation of the population in China is severe, the huge aging population causes serious support burden of the aged, the medical and health expenses are increased, the demands of the aged service in society are expanded, and the social development is affected deeply. The health problem of the old is prominent. On the other hand, the illness state of the old chronic patients also brings a certain burden to carers of the old chronic patients. Studies have shown that "most caregivers have varying degrees of economic, social, physiological and psychological burden". Therefore, the medical problems of the huge and growing senile chronic disease groups are one of the non-negligible problems in the aging problems, and bring great problems to the medical and health service system of China, and the search of the content and the mode of the senile chronic disease service meeting the requirements of Chinese national conditions is urgent.
The traditional Chinese medicine has unique advantages and rich practical experience for preventing and treating complications of chronic diseases, and the concept of overall concept and diagnosis and treatment is significant in the aspects of treatment and prognosis, preventing and treating complications, improving quality of life and the like. However, the development of Chinese medicine mainly faces three problems: the learning cost of the traditional Chinese medicine is high, and the inheritance difficulty is far higher than that of Western medicine; traditional Chinese medicine lacks a scientific theoretical system like Western medicine; the number of the traditional Chinese medicine medical institutions is small, and the doctors are deficient in resources. Artificial intelligence technology creates social value by studying or manufacturing systems or machines with human intelligence to better assist humans in working. Along with the improvement of the computing capacity of a computer and the perfection of an artificial intelligence science theory, the artificial intelligence technology has made great progress, and the medical diagnosis method based on the artificial intelligence also becomes a research hotspot of students worldwide. The diagnosis of the important clinical indexes can be quantified by intelligent diagnosis, so that doctors can be helped to improve the accuracy of diagnosis in a wide range of medical professions, and the medical service quality can be improved. More and more students in different fields begin to use informationized technical means to research the subjects with extremely rich and complex knowledge content of traditional Chinese medicine, so that the traditional Chinese medicine can be better mutually exchanged and popularized in different subjects, a new solution is provided for the intelligent dialectical problem of the traditional Chinese medicine, and the method has very important significance in dialectical disease analysis of the traditional Chinese medicine, mining of association relations of the traditional Chinese medicine and the western medicine and construction of the intelligent medical system of the traditional Chinese medicine.
Therefore, the inventor builds a deep intelligent machine learning model based on the deep neural network to solve the technical problems of chronic disease diagnosis and medicine recommendation, and creatively provides a chronic disease intelligent traditional Chinese medicine diagnosis and medicine recommendation system based on the deep neural network.
Disclosure of Invention
The invention aims at: in order to solve the technical problems related to the background technology, a chronic disease intelligent traditional Chinese medicine diagnosis and medicine recommendation system based on a deep neural network is provided.
The technical scheme adopted by the invention is as follows:
A chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on a deep neural network comprises a Chinese medicine condition reading module for reading symptom information of a patient sample;
The data preprocessing module is used for preprocessing the data of the read symptom information;
The database module is used for storing symptom information of the patient sample and obtained medicine information;
the deep neural network model is used for predicting input information and obtaining classification results of related medicaments to form prediction information;
the medicine recommending module is used for combining the prediction information of the deep neural network model, the symptom information of the patient sample stored by the database module and the obtained medicine information to analyze and calculate to obtain recommended medicines;
And the data visualization module is used for visually displaying recommended medicines.
As a further technical scheme of the invention, the traditional Chinese medicine condition reading module supports a plurality of data storage forms.
As still further technical scheme of the invention, the data preprocessing module carries out data preprocessing on the read symptom information, the data preprocessing process comprises text word segmentation, stop word removal and text vectorization, wherein the text word segmentation is carried out by using a word segmentation tool, and then the text word segmentation is added into a dictionary library for expansion; the method comprises the steps of generating a term acyclic graph according to the condition of generating words, and finding out the maximum segmentation combination based on word frequency by adopting a dynamic programming algorithm when finding out the maximum probability path;
Removing stop words, and removing the stop words contained in the word segmentation result by using a library to obtain a result of using a stop word list;
Wherein the text vectorization is generated using a trained word2vec trained model.
As still further technical scheme of the invention, the database module selects MySQL database, stores symptom information and classification information data preprocessed by the data preprocessing module, and stores information of different medicines.
As a still further technical scheme of the invention, the deep neural network model is VGGNet convolutional neural network structure model; three Inception modules are adopted in the VGGNet convolutional neural network to replace convolutional layers to extract features, the last pooling layer is changed from original maximum pooling to spatial pyramid pooling, and the total number of models is 23; which contains 10 convolutional layers, 5 pooling layers, 3 inception modules, 3 fully connected layers, 1 input layer, and 1 output layer.
As still further technical scheme of the invention, a batch gradient descent algorithm is adopted in the deep neural network model, and the expression is as follows:
Wherein, batch represents the batch size, the operation is carried out in a matrix mode, and the weight is updated by using the sum of the loss functions of one batch size data sample each time;
The input number of each time, namely the batch size, is set to be 8, the learning rate LEARNING RAGE of the initial model is 0.001, the model is trained once for training all images, then the training is called as one completed epoch training, 30 epochs are trained in total, the learning rate adjustment strategy is multistp, stepvalue is 20 epochs, namely the learning rate is reduced once when the 20 th epochs is completed;
An objective function in the deep neural network model selects an improved cross entropy loss function:
J(θ；Is,ys)＝Lc+ω*LR
Wherein L c is a classification term loss function, which is realized through cross entropy; l R is a region term loss function, which is realized through a Dice loss function; i s,ys is an input variable in the training data respectively; x v denotes the position of the probabilistic feature vector; a representative probabilistic feature vector output; p 1 denotes the probability of the corresponding output; ω is used to adjust the ratio between the two losses, preferably ω is 3.
Adding a dropout layer from the second convolution layer in the deep neural network model, and randomly discarding neurons of different parts with a certain probability in each iteration of the neural network in the training process, wherein the dropout rate p=0.2;
the activation function in the deep neural network model is a sig_ReLU function, and the activation function combines a sigmoid activation function and a ReLU activation function; the expression is as follows:
as a still further technical scheme of the invention, the deep neural network model also comprises a training system, the training system trains through a training set, the recognition precision of the model is tested on a verification set after the model training is finished, and the model is evaluated through the Tanimoto coefficient precision of the model, and the cyclic training is judged through multiple logics, so that the precision requirement is met.
As still further technical scheme of the invention, the medicine recommending module is a multi-interface module, and recommended medicines are sent to the data visualizing module.
As still further technical scheme of the invention, the data visualization module presents the predicted result as the result of the data table.
As still further technical solution of the present invention, the first column of the data table is a sample name, the number 1 is to recommend the drug for the sample, and the number 0 is that the sample does not need the corresponding drug.
The invention discloses an intelligent traditional Chinese medicine recommendation method based on deep learning, which comprises the following steps of:
Step 1: preprocessing patient condition data in the dataset;
Step 2: training a convolutional neural network with a training sample;
step 3: and predicting the test sample by using the trained convolutional neural network model to obtain a final prediction result.
Specifically, in step 1, preprocessing is performed on patient condition data in a data set, the patient condition data in the data set is divided into a training sample and a test sample, and then text word segmentation, stop word removal and text vectorization are performed. When the method aims at unstructured text, a one-step unstructured text parsing process is needed to be added in the preprocessing process. After the text word segmentation is performed by using a word segmentation tool, a plurality of typical dictionary libraries are added for expansion. And constructing a loop-free diagram by utilizing all the conditions of possible word generation, and adopting a dynamic programming algorithm when searching the maximum probability path to find out the maximum segmentation combination based on word frequency. The step of removing stop words uses the library to remove stop words contained in the word segmentation result, namely, the result of using the stop word list. Text vectorization is generated using a trained word2vec trained model.
Specifically, in step 2, the convolutional neural network is trained by using a training sample, the convolutional neural network structure is shown in fig. 4, three Inception modules are used for replacing a convolutional layer on the basis of VGGNet to extract features, the last pooling layer is changed from the original maximum pooling to spatial pyramid pooling, and the total number of models is 23. Which contains 10 convolutional layers, 5 pooling layers, 3 inception modules, 3 fully connected layers, 1 input layer, and 1 output layer. Training the convolutional neural network by using the training set preprocessed in the step 1, carrying out multiple convolutions, maximum pooling and inception on vectors at the bottom layer by using the network, and obtaining a feature map containing disease characteristics through pyramid pooling and 3 full-connection layers, wherein each convolution layer pair corresponds to the convolution layer in the coding network, and outputting by using sig_ReLU as an activation function. Finally, the convolution output characteristics are converted into a prediction result by softmax.
After the model training of the first 2 steps is completed, step 3 can be performed to predict the chronic patients to be tested. The disease data is preprocessed, and then the convolutional neural network is used for prediction, so that the whole process is completely automatic and high in speed.
The invention provides an intelligent Chinese medicine dialectical and medicine recommendation system for chronic diseases, which mainly comprises: the system comprises a traditional Chinese medicine condition reading module, a data preprocessing module, a deep neural network model, a medicine recommending module, a data visualization module and a database module.
The traditional Chinese medicine symptom reading module obtains symptom information of a patient sample; the data preprocessing module performs data preprocessing on the read symptom information, and comprises the following steps: text word segmentation, stop word removal and text vectorization; the deep neural network model module predicts the input information to obtain a classification result of each drug, and further obtains recommended drugs; the three modules are all connected with the database, the database module stores symptom information of patient samples and acquired medicine information, intelligent medicine recommendation can be provided for patients needing diagnosis and treatment through the medicine recommendation module by fusing the symptom information and the medicine information, the patients are helped to realize self diagnosis and treatment to a certain extent, the general health level of the national people is improved, and the national medical resources are saved.
In summary, due to the adoption of the technical scheme, the beneficial effects of the invention are as follows:
1. The invention provides a chronic disease intelligent traditional Chinese medicine diagnosis and medicine recommendation system based on a deep neural network, which is characterized in that a database module for storing symptom information of a patient sample and obtained medicine information is constructed; the traditional Chinese medicine condition reading module is used for reading symptom information of a patient sample, predicting input information through the deep neural network model, and obtaining classification results of related medicines to form prediction information; the recommended medicine is obtained by the medicine recommending module through combination of the prediction information of the deep neural network model, the symptom information of the patient sample stored by the database module and the obtained medicine information, intelligent traditional Chinese medicine diagnosis and medicine recommendation of the chronic diseases are realized, doctors are assisted in diagnosis and treatment, and especially, doctors are assisted in providing an optimal treatment method for the patient, so that the cost is saved; the disease diagnosis accuracy rate can be improved, the death rate can be reduced, the hospital pressure can be relieved, and the like; the medical device helps chronic patients to realize self diagnosis and treatment to a certain extent, saves national medical resources and improves the health level of the whole people.
2. The deep neural network model is comprehensive in construction, high in intelligent degree, high in model processing efficiency and high in prediction result accuracy.
3. The data visualization module further presents the prediction result as the result of the data table; the first column of the data table is a sample name, the number 1 is the recommended medicine for the sample, the number 0 is the medicine which is not needed by the sample, and the prediction result is intuitively and clearly displayed.
4. The invention further provides a data preprocessing module for preprocessing the data of the read symptom information, wherein the data preprocessing process comprises text word segmentation, stop word removal and text vectorization, and the text word segmentation is added into a dictionary library for expansion after word segmentation by a word segmentation tool; the method comprises the steps of generating a term acyclic graph according to the condition of generating words, and finding out the maximum segmentation combination based on word frequency by adopting a dynamic programming algorithm when finding out the maximum probability path; removing stop words, and removing the stop words contained in the word segmentation result by using a library to obtain a result of using a stop word list; text vectorization is generated by using a trained word2vec trained model; the pretreatment effect is good, and the subsequent prediction analysis is convenient.
Drawings
FIG. 1 is a block diagram of a system architecture of the present invention;
FIG. 2 is a model design of the training system of the present invention;
FIG. 3 is a diagram of a sig-ReLU activation function in accordance with the present invention;
FIG. 4 is a network architecture diagram of a deep neural network model of the present invention;
FIG. 5 is a graph of visual predictive outcome data in accordance with the invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Referring to fig. 1, the intelligent chronic disease Chinese medicine dialectical and drug recommendation system provided by the invention mainly comprises: the system comprises a traditional Chinese medicine condition reading module, a data preprocessing module, a deep neural network model, a medicine recommending module, a data visualization module and a database module.
The data traditional Chinese medical condition reading module, the data preprocessing module, the deep neural network model, the drug recommending module and the data visualizing module are sequentially connected, and the data preprocessing module, the deep neural network module and the drug recommending module are respectively connected with the database module.
The traditional Chinese medicine symptom reading module obtains symptom information of a patient sample; the data preprocessing module performs data preprocessing on the read symptom information, and comprises the following steps: text word segmentation, stop word removal and text vectorization; the deep neural network model module predicts the input information to obtain a classification result of each drug, and further obtains recommended drugs; the three modules are all connected with the database, the database module stores symptom information of patient samples and acquired medicine information, intelligent medicine recommendation can be provided for patients needing diagnosis and treatment through the medicine recommendation module by fusing the symptom information and the medicine information, the patients are helped to realize self diagnosis and treatment to a certain extent, the general health level of the national people is improved, and the national medical resources are saved.
Further described, the deep neural network is a neural network model under TensorFlow framework, the framework has flexible portability, extremely fast compiling speed, powerful visual components, a visual network structure and a training process, and the full flow of an application machine can be realized;
Further described, the database module is a MySQL database.
The system can assist doctors to provide the optimal treatment method for patients, so that the cost is saved; the disease diagnosis accuracy rate can be improved, the death rate can be reduced, the hospital pressure can be relieved, and the like; helps patients to realize self diagnosis and treatment to a certain extent, and improves national health level.
Aiming at the existing problems, the invention provides an intelligent traditional Chinese medicine diagnosis and medicine recommendation platform combining a deep neural network model, which is used for making effective auxiliary diagnosis and treatment decisions on the basis of playing the advantages of traditional Chinese medicine.
In the system provided by the invention, the traditional Chinese medicine condition reading module supports various data storage modes, and is convenient for a user to operate.
The data preprocessing module needs to preprocess the symptom information of the read patient symptom information. The preprocessing process mainly comprises three steps of text word segmentation, word stopping and text vectorization. When aiming at unstructured text, a one-step unstructured text analysis process is needed to be added in the preprocessing process. After the text word segmentation is performed by using a word segmentation tool, a plurality of typical dictionary libraries are added for expansion. And constructing a loop-free diagram by utilizing all the conditions of possible word generation, and adopting a dynamic programming algorithm when searching the maximum probability path to find out the maximum segmentation combination based on word frequency. The step of removing stop words uses the library to remove stop words contained in the word segmentation result, namely, the result of using the stop word list. Text vectorization is generated using a trained word2vec trained model.
The deep neural network module mainly comprises a convolution layer, a pooling layer, inception, a full-connection layer and an activation function. The use of convolutional neural network models in computer languages has great advantages over conventional algorithms. Compared with the shallow artificial neural network, the feature data obtained by the deep learning model learning can represent the original data more essentially, the classification and the visualization can be realized conveniently, and the problem that the deep neural network is difficult to train to be optimal can be solved by adopting a layer-by-layer training method, so that the deep neural network model is used for processing. The design of the model is shown in fig. 2, and mainly comprises two parts of network layer structure design and training parameter design, wherein after the design is completed, firstly the divided training set is trained, after the model training is completed, the recognition precision of the model is tested on the verification set, the evaluation means mainly comprises the Tanimoto coefficient precision of the model, if the model does not meet the precision requirement, the training parameters are further adjusted to retrain, if the model meets the precision requirement, the model is detected by the test set, if the model does not meet the precision requirement, the network structure is adjusted to carry out cyclic training again, and if the precision requirement is met, the training is completed.
After five convolutional neural networks AlexNet, googLeNet, VGGNet, resNet and DenseNet are compared, VGGNet is selected as a basic model of the invention. To accelerate the training process, the present invention uses a batch gradient descent algorithm, which appears to be expressed as follows:
in the formula, the batch represents the batch size, the calculation is performed in a matrix mode, the weight is updated by using the sum of the loss functions of one batch size data sample each time, the iteration times required by the convergence of the network model can be reduced, the network identification effect is ensured, and meanwhile, the training efficiency is improved.
According to the batch gradient descent algorithm, the number of inputs each time, namely batch size, is set to be 8, the learning rate LEARNING RAGE of an initial model is set to be 0.001, after the model finishes training all images in a training set, the model is called as one completed epoch training, 30 epochs are trained in total, the learning rate adjustment strategy is multistp, stepvalue is 20 epochs, namely, the learning rate is reduced once when the 20 th epochs is completed.
The objective function of the present invention selects an improved cross entropy loss function:
J(θ；Is,ys)＝Lc+ω*LR
Wherein L c is a classification term loss function, which is realized through cross entropy; l R is a region term loss function, which is realized through a Dice loss function; i s,ys is an input variable in the training data respectively; x v denotes the position of the probabilistic feature vector; A representative probabilistic feature vector output; p 1 denotes the probability of the corresponding output; ω is used to adjust the ratio between the two losses, ω=3 being chosen in the present invention.
In order to further avoid overfitting, a dropout layer is added in each convolution module (except the first one), neurons of different parts are randomly discarded with a certain probability in each iteration of the neural network in the training process, so that the diversity of a model sample data set can be enhanced, the sparsity among layers is also enhanced, the disturbance resistance of the model to unknown data is stronger, the robustness is higher, and the dropout rate p=0.2 is adopted.
The activation function employed by the present invention is the sig_relu function as shown in fig. 3, which combines the sigmoid activation function and the ReLU activation function. The expression is as follows:
The improved function is a Sigmoid function in the range of x < 0, so that the condition that the gradient of the ReLU explodes or disappears can be avoided, and the characteristics of rapid convergence and sparsity of the ReLU function are maintained.
The pooling method adopted by the invention is space pyramid pooling, and the method utilizes multi-scale information of a pooling area, can map feature vectors with different sizes to the same dimension, avoids scaling operation on original vectors, reduces loss of feature vector information to a certain extent, and improves feature expression capability of network pairs.
Referring to fig. 4, the present invention replaces the convolutional layer with three Inception modules based on VGGNet to extract the features, the last pooling layer is changed from the original maximum pooling to spatial pyramid pooling, and the total number of models is 23. Which contains 10 convolutional layers, 5 pooling layers, 3 inception modules, 3 fully connected layers, 1 input layer, and 1 output layer.
The database module uses MySQL database, which is a widely used relational database system at present, and in the system, the database module stores the preprocessed symptom information, classification information and other data, and simultaneously stores information of different medicines, thereby providing information for intelligent Chinese medicine dialectics and medicine recommendation of chronic diseases.
The medicine recommending module is a multi-interface module, predicts the input symptom information, and then sends the recommended medicine to the data visualizing module.
The data visualization module presents the prediction result as the result of the data table, as shown in fig. 5, wherein the first row in the figure is the recommended drug name, the first row is the sample name, the number 1 is the recommended drug for the sample, and the number 0 is that the sample does not need the corresponding drug.
Through traditional chinese medical science disorder reading module, data preprocessing module, degree of depth neural network model module, medicine recommendation module and data visualization module, realized a complete chronic disease intelligent traditional chinese medical science dialectical and medicine recommendation system, provide intelligent traditional chinese medical science dialectical and medicine recommendation for chronic disease patient, help doctor to diagnose, help chronic disease patient to realize self-diagnosis and treat to a certain extent, save national medical resources, promote the health level of the whole people.
The invention discloses an intelligent traditional Chinese medicine recommendation method based on deep learning, which comprises the following steps of:
Step 1: preprocessing patient condition data in the dataset;
Step 2: training a convolutional neural network with a training sample;
step 3: and predicting the test sample by using the trained convolutional neural network model to obtain a final prediction result.
Specifically, in step 1, preprocessing is performed on patient condition data in a data set, the patient condition data in the data set is divided into a training sample and a test sample, and then text word segmentation, stop word removal and text vectorization are performed. When the method aims at unstructured text, a one-step unstructured text parsing process is needed to be added in the preprocessing process. After the text word segmentation is performed by using a word segmentation tool, a plurality of typical dictionary libraries are added for expansion. And constructing a loop-free diagram by utilizing all the conditions of possible word generation, and adopting a dynamic programming algorithm when searching the maximum probability path to find out the maximum segmentation combination based on word frequency. The step of removing stop words uses the library to remove stop words contained in the word segmentation result, namely, the result of using the stop word list. Text vectorization is generated using a trained word2vec trained model.
Specifically, in step 2, the convolutional neural network is trained by using a training sample, the convolutional neural network structure is shown in fig. 4, three Inception modules are used for replacing a convolutional layer on the basis of VGGNet to extract features, the last pooling layer is changed from the original maximum pooling to spatial pyramid pooling, and the total number of models is 23. Which contains 10 convolutional layers, 5 pooling layers, 3 inception modules, 3 fully connected layers, 1 input layer, and 1 output layer. Training the convolutional neural network by using the training set preprocessed in the step 1, carrying out multiple convolutions, maximum pooling and inception on vectors at the bottom layer by using the network, and obtaining a feature map containing disease characteristics through pyramid pooling and 3 full-connection layers, wherein each convolution layer pair corresponds to the convolution layer in the coding network, and outputting by using sig_ReLU as an activation function. Finally, the convolution output characteristics are converted into a prediction result by softmax.
After the model training of the first 2 steps is completed, step 3 can be performed to predict the chronic patients to be tested. The disease data is preprocessed, and then the convolutional neural network is used for prediction, so that the whole process is completely automatic and high in speed.
The invention provides an intelligent Chinese medicine dialectical and medicine recommendation system for chronic diseases, which mainly comprises: the system comprises a traditional Chinese medicine condition reading module, a data preprocessing module, a deep neural network model, a medicine recommending module, a data visualization module and a database module.
The traditional Chinese medicine symptom reading module obtains symptom information of a patient sample; the data preprocessing module performs data preprocessing on the read symptom information, and comprises the following steps: text word segmentation, stop word removal and text vectorization; the deep neural network model module predicts the input information to obtain a classification result of each drug, and further obtains recommended drugs; the three modules are all connected with the database, the database module stores symptom information of patient samples and acquired medicine information, intelligent medicine recommendation can be provided for patients needing diagnosis and treatment through the medicine recommendation module by fusing the symptom information and the medicine information, the patients are helped to realize self diagnosis and treatment to a certain extent, the general health level of the national people is improved, and the national medical resources are saved.
Finally, it is noted that the above embodiments are only for illustrating the technical solution of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that modifications and equivalents may be made thereto without departing from the spirit and scope of the technical solution of the present invention, which is intended to be covered by the scope of the claims of the present invention.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (8)
1. A chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on a deep neural network is characterized in that:
the system comprises a traditional Chinese medicine condition reading module, a diagnosis module and a diagnosis module, wherein the traditional Chinese medicine condition reading module is used for reading symptom information of a patient sample;
The data preprocessing module is used for preprocessing the data of the read symptom information;
The database module is used for storing symptom information of the patient sample and obtained medicine information;
the deep neural network model is used for predicting input information and obtaining classification results of related medicaments to form prediction information;
The deep neural network model is VGGNet convolutional neural network structure model; three Inception modules are adopted in the VGGNet convolutional neural network to replace convolutional layers to extract features, the last pooling layer is changed from original maximum pooling to spatial pyramid pooling, and the total number of models is 23; the method comprises 10 convolution layers, 5 pooling layers, 3 inception modules, 3 full-connection layers, 1 input layer and 1 output layer, wherein a batch gradient descent algorithm is adopted in a deep neural network model, and the expression is as follows:
Wherein, batch represents the batch size, the operation is carried out in a matrix mode, and the weight is updated by using the sum of the loss functions of one batch size data sample each time;
The input number of each time, namely the batch size, is set to be 8, the learning rate LEARNING RAGE of the initial model is 0.001, the model is trained once for training all images, then the training is called as one completed epoch training, 30 epochs are trained in total, the learning rate adjustment strategy is multistp, stepvalue is 20 epochs, namely the learning rate is reduced once when the 20 th epochs is completed;
An objective function in the deep neural network model selects an improved cross entropy loss function:
J(θ；Is,ys)＝Lc+ω*LR
Wherein L c is a classification term loss function, which is realized through cross entropy; l R is a region term loss function, which is realized through a Dice loss function; i s,ys is an input variable in the training data respectively; x v denotes the position of the probabilistic feature vector; a representative probabilistic feature vector output; p 1 denotes the probability of the corresponding output; omega is used to adjust the ratio between the two losses, omega being 3;
Adding a dropout layer from the second convolution layer in the deep neural network model, and randomly discarding neurons of different parts with a certain probability in each iteration of the neural network in the training process, wherein the dropout rate p=0.2;
the activation function in the deep neural network model is a sig_ReLU function, and the activation function combines a sigmoid activation function and a ReLU activation function; the expression is as follows:
the medicine recommending module is used for combining the prediction information of the deep neural network model, the symptom information of the patient sample stored by the database module and the obtained medicine information to analyze and calculate to obtain recommended medicines;
And the data visualization module is used for visually displaying recommended medicines.
2. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases, as claimed in claim 1, is characterized in that: the traditional Chinese medicine condition reading module supports various data storage forms.
3. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases, as claimed in claim 1, is characterized in that: the data preprocessing module performs data preprocessing on the read symptom information, wherein the data preprocessing process comprises text word segmentation, stop word removal and text vectorization, and the text word segmentation is performed by using a word segmentation tool and then added into a dictionary library for expansion; the method comprises the steps of generating a term acyclic graph according to the condition of generating words, and finding out the maximum segmentation combination based on word frequency by adopting a dynamic programming algorithm when finding out the maximum probability path;
Removing stop words, and removing the stop words contained in the word segmentation result by using a library to obtain a result of using a stop word list;
Wherein the text vectorization is generated using a trained word2vec trained model.
4. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases, as claimed in claim 1, is characterized in that: the database module selects MySQL database, stores symptom information and classified information data preprocessed by the data preprocessing module, and stores information of different medicines.
5. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases, as claimed in claim 1, is characterized in that: the deep neural network model also comprises a training system, the training system trains through a training set, the recognition precision of the model is tested on a verification set after model training is completed, evaluation is carried out through the Tanimoto coefficient precision of the model, and the cyclic training is judged through multiple logics, so that the precision requirement is met.
6. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases according to any one of claims 1 to 5, wherein the system is characterized in that: the medicine recommending module is a multi-interface module, and recommended medicines are sent to the data visualization module.
7. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases, as claimed in claim 6, is characterized in that: the data visualization module presents the predicted results as the results of a data table.
8. The deep neural network-based intelligent traditional Chinese medicine diagnosis and drug recommendation system for chronic diseases, as claimed in claim 7, is characterized in that: the first column of the data table is the sample name, the number 1 is the recommended medicine for the sample, and the number 0 is the sample does not need the corresponding medicine.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910776695.5A CN112489769B (en) | 2019-08-22 | 2019-08-22 | Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910776695.5A CN112489769B (en) | 2019-08-22 | 2019-08-22 | Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112489769A CN112489769A (en) | 2021-03-12 |
| CN112489769B true CN112489769B (en) | 2024-05-17 |
Family
ID=74920206
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910776695.5A Active CN112489769B (en) | 2019-08-22 | 2019-08-22 | Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112489769B (en) |
Families Citing this family (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112988953B (en) * | 2021-04-26 | 2021-09-03 | 成都索贝数码科技股份有限公司 | Adaptive broadcast television news keyword standardization method |
| CN113284577B (en) * | 2021-05-24 | 2023-08-11 | 康键信息技术（深圳）有限公司 | Medicine prediction method, device, equipment and storage medium |
| CN113628716A (en) * | 2021-08-05 | 2021-11-09 | 翼健(上海)信息科技有限公司 | Prescription recommendation system |
| CN113782142B (en) * | 2021-08-20 | 2024-04-16 | 中国中医科学院中医药信息研究所 | Traditional Chinese medicine decoction piece formula recommendation system based on integrated neural network |
| CN114969557B (en) * | 2022-07-29 | 2022-11-08 | 之江实验室 | Propaganda and education pushing method and system based on multi-source information fusion |
| CN116092627B (en) * | 2023-04-04 | 2023-06-27 | 南京大经中医药信息技术有限公司 | Intelligent prescription system based on syndrome differentiation of traditional Chinese medicine pathogenesis |
| CN117275660B (en) * | 2023-10-19 | 2024-05-07 | 广州市莲田医药科技有限公司 | Full-link AI auxiliary method for inquiry to prescription |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853325A (en) * | 2009-03-31 | 2010-10-06 | 江启煜 | Computer-aided analysis method for traditional Chinese medicinal clinical information |
| WO2015158017A1 (en) * | 2014-04-16 | 2015-10-22 | 深圳市易特科信息技术有限公司 | Intelligent interaction and psychological comfort robot service system |
| CN106203432A (en) * | 2016-07-14 | 2016-12-07 | 杭州健培科技有限公司 | A kind of localization method of area-of-interest based on convolutional Neural net significance collection of illustrative plates |
| WO2017113232A1 (en) * | 2015-12-30 | 2017-07-06 | 中国科学院深圳先进技术研究院 | Product classification method and apparatus based on deep learning |
| CN106933994A (en) * | 2017-02-27 | 2017-07-07 | 广东省中医院 | A kind of core disease card relation construction method based on knowledge of TCM collection of illustrative plates |
| CN107330876A (en) * | 2017-06-12 | 2017-11-07 | 济南浪潮高新科技投资发展有限公司 | A kind of image automatic diagnosis method based on convolutional neural networks |
| CN108182967A (en) * | 2017-12-14 | 2018-06-19 | 华南理工大学 | A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method |
| CN109344921A (en) * | 2019-01-03 | 2019-02-15 | 湖南极点智能科技有限公司 | A kind of image-recognizing method based on deep neural network model, device and equipment |
| CN110070119A (en) * | 2019-04-11 | 2019-07-30 | 北京工业大学 | A kind of handwritten numeral image recognition classification method based on binaryzation deep neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP3306617A1 (en) * | 2016-10-06 | 2018-04-11 | Fujitsu Limited | Method and apparatus of context-based patient similarity |
-
2019
- 2019-08-22 CN CN201910776695.5A patent/CN112489769B/en active Active
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101853325A (en) * | 2009-03-31 | 2010-10-06 | 江启煜 | Computer-aided analysis method for traditional Chinese medicinal clinical information |
| WO2015158017A1 (en) * | 2014-04-16 | 2015-10-22 | 深圳市易特科信息技术有限公司 | Intelligent interaction and psychological comfort robot service system |
| WO2017113232A1 (en) * | 2015-12-30 | 2017-07-06 | 中国科学院深圳先进技术研究院 | Product classification method and apparatus based on deep learning |
| CN106203432A (en) * | 2016-07-14 | 2016-12-07 | 杭州健培科技有限公司 | A kind of localization method of area-of-interest based on convolutional Neural net significance collection of illustrative plates |
| CN106933994A (en) * | 2017-02-27 | 2017-07-07 | 广东省中医院 | A kind of core disease card relation construction method based on knowledge of TCM collection of illustrative plates |
| CN107330876A (en) * | 2017-06-12 | 2017-11-07 | 济南浪潮高新科技投资发展有限公司 | A kind of image automatic diagnosis method based on convolutional neural networks |
| CN108182967A (en) * | 2017-12-14 | 2018-06-19 | 华南理工大学 | A kind of traditional Chinese medical science medicinal material based on deep neural network recommends method |
| CN109344921A (en) * | 2019-01-03 | 2019-02-15 | 湖南极点智能科技有限公司 | A kind of image-recognizing method based on deep neural network model, device and equipment |
| CN110070119A (en) * | 2019-04-11 | 2019-07-30 | 北京工业大学 | A kind of handwritten numeral image recognition classification method based on binaryzation deep neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112489769A (en) | 2021-03-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112489769B (en) | Chronic disease intelligent Chinese medicine diagnosis and medicine recommendation system based on deep neural network | |
| RU2703679C2 (en) | Method and system for supporting medical decision making using mathematical models of presenting patients | |
| JP6522161B2 (en) | Medical data analysis method based on deep learning and intelligent analyzer thereof | |
| CN106778014A (en) | A kind of risk Forecasting Methodology based on Recognition with Recurrent Neural Network | |
| CN111180068A (en) | Chronic disease prediction system based on multi-task learning model | |
| Dabowsa et al. | A hybrid intelligent system for skin disease diagnosis | |
| Mall et al. | Heart diagnosis using deep neural network | |
| El Hamdaoui et al. | Improving heart disease prediction using random forest and adaboost algorithms | |
| Bishnoi et al. | Artificial intelligence techniques used in medical sciences: a review | |
| Kamada et al. | Knowledge extraction of adaptive structural learning of deep belief network for medical examination data | |
| CN113673434A (en) | Electroencephalogram emotion recognition method based on efficient convolutional neural network and contrast learning | |
| Ma et al. | Construction and evaluation of intelligent medical diagnosis model based on integrated deep neural network | |
| CN110299194B (en) | Similar case recommendation method based on comprehensive feature representation and improved wide-depth model | |
| Thilakarathne et al. | Artificial Intelligence-Enabled IoT for Health and Wellbeing Monitoring | |
| Deb et al. | An outcome based analysis on heart disease prediction using machine learning algorithms and data mining approaches | |
| Yousif | Classification of mental disorders figures based on soft computing methods | |
| Mehrankia et al. | Prediction of heart attacks using biological signals based on recurrent GMDH neural network | |
| Ambrews et al. | Ensemble based machine learning model for heart disease prediction | |
| CN114496231A (en) | Constitution identification method, apparatus, equipment and storage medium based on knowledge graph | |
| Abbas et al. | An Efficient Stacked Ensemble Model for Heart Disease Detection and Classification. | |
| Nagarajan et al. | An optimized sub group partition based healthcare data mining in big data | |
| Komalavalli et al. | An Effective Heart Disease Prediction Using Machine Learning | |
| Dai et al. | Exploring body constitution in traditional Chinese medicine with k-means clustering | |
| Verma et al. | Ontology based personalized modeling for chronic disease risk analysis: An integrated approach | |
| Saranya et al. | Bd-Mdl: bipolar disorder detection using machine leanring and deep learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information |
Country or region after: China Address after: Building 5, 1st Floor, No. 2 Xiyuanba Road, Xihu District, Hangzhou City, Zhejiang Province, 310030 Applicant after: Zhejiang Yuantu Technology Co.,Ltd. Address before: 310030 west side, 4th floor, building 1, No. 10, xiyuansan Road, Sandun Town, Hangzhou City, Zhejiang Province Applicant before: ZHEJIANG YUANTU INTERCONNECTION TECHNOLOGY CO.,LTD. Country or region before: China |
|
| CB02 | Change of applicant information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |