CN112420153A - Method for improving traditional Chinese medicine prescription based on GAN - Google Patents
Method for improving traditional Chinese medicine prescription based on GAN Download PDFInfo
- Publication number
- CN112420153A CN112420153A CN202011349053.6A CN202011349053A CN112420153A CN 112420153 A CN112420153 A CN 112420153A CN 202011349053 A CN202011349053 A CN 202011349053A CN 112420153 A CN112420153 A CN 112420153A
- Authority
- CN
- China
- Prior art keywords
- chinese medicine
- prescription
- vector sequence
- original
- medicine prescription
- 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.)
- Granted
Links
Images
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
- G16H20/10—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to drugs or medications, e.g. for ensuring correct administration to patients
-
- 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
- G06N3/044—Recurrent networks, e.g. Hopfield 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
- G06N3/04—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
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- 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
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- 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
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/40—ICT specially adapted for the handling or processing of medical references relating to drugs, e.g. their side effects or intended usage
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Physics & Mathematics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Chemical & Material Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medicinal Chemistry (AREA)
- Epidemiology (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Pharmacology & Pharmacy (AREA)
- Toxicology (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Medical Preparation Storing Or Oral Administration Devices (AREA)
Abstract
The application discloses a method for improving a traditional Chinese medicine prescription based on GAN, which is used for solving the problem that a traditional Chinese medicine doctor is assisted to improve the rationality of the traditional Chinese medicine prescription according to the actual needs of a patient in the prior art. The method comprises the following steps: collecting Chinese medicine prescription data, and forming an original Chinese medicine prescription vector sequence according to the Chinese medicine prescription data; inputting the original Chinese medicine prescription vector sequence into a generator of the GAN, training and outputting the generated Chinese medicine prescription vector sequence; inputting the original Chinese medicinal prescription vector sequence and the generated Chinese medicinal prescription vector sequence into a discriminator of the GAN, training and updating parameters in the discriminator, so that the discriminator can distinguish the original Chinese medicinal prescription vector sequence from the generated Chinese medicinal prescription vector sequence; and training the generator and the discriminator alternately to obtain the improved model of the Chinese medicine prescription.
Description
Technical Field
The application relates to the field of deep learning, in particular to a method for improving a traditional Chinese medicine prescription based on GAN.
Background
Generation of a countermeasure network (GAN) is one of the most important methods for unsupervised learning on complex distribution in recent years, and is currently widely applied to the generation field, and GAN can generate impressive results and control smooth and reasonable semantic changes, and becomes the most important generation model framework for learning arbitrary complex data distribution.
The traditional Chinese medicine is accumulated by the ancestors in long-term medical practice and is an important component of ancient excellent cultural heritage in China. Through long-term medical practice, a plurality of medicines are matched and decocted to prepare a decoction, and then a traditional Chinese medicine formula is formed.
Under the traditional situation, the treatment effect of a patient depends on the experience of a traditional Chinese medicine teacher, and more reasonable traditional Chinese medicine formulas can be used by the patient in many cases, but the traditional Chinese medicine teacher often cannot see due to the limitation of the experience and does not have a system for providing various traditional Chinese medicine formulas for the traditional Chinese medicine teacher to select.
Disclosure of Invention
The invention provides a method for improving a traditional Chinese medicine prescription based on GAN, which solves the problem of forming a more reasonable and effective traditional Chinese medicine prescription by effectively utilizing GAN and a deep learning technology and combining the actual condition of a patient to improve the traditional Chinese medicine prescription.
A method for improving a traditional Chinese medicine prescription based on GAN is characterized by comprising the following steps:
collecting Chinese medicine prescription data, and forming an original Chinese medicine prescription vector sequence according to the Chinese medicine prescription data;
inputting the original Chinese medicine prescription vector sequence into a generator of a GAN (genetic adaptive network) network, training and outputting the generated Chinese medicine prescription vector sequence;
inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into a discriminator of the GAN, training and updating parameters in the discriminator, so that the discriminator can distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence;
and training the generator and the discriminator alternately to obtain the improved model of the Chinese medicine prescription.
Optionally, inputting the original vector sequence of the chinese herbal prescription into a generator of GAN, training, and outputting the generated vector sequence of the chinese herbal prescription, specifically including:
fixing the parameters of the discriminator, disordering the original Chinese medicine prescription vector sequence and inputting the disordering into the generator, updating the parameters of the generator, so that the discriminator can not distinguish whether the Chinese medicine prescription vector sequence is the original Chinese medicine prescription vector sequence or the generated Chinese medicine prescription vector sequence, and finally outputting the generated Chinese medicine prescription vector sequence.
Optionally, the generator specifically includes:
a sequence encoder and a prescription sequence generator;
the sequence encoder is a recurrent neural network and is used for extracting characteristics according to an original Chinese medicine prescription vector sequence to generate an encoding vector, and the prescription sequence generator is a recurrent neural network and is used for inputting the encoding vector, a patient label, a weight condition and a random vector and outputting the generated Chinese medicine prescription vector sequence.
Optionally, the original chinese herbal prescription vector sequence and the generated chinese herbal prescription vector sequence are input into the discriminator of the GAN, and parameters in the discriminator are trained and updated, so that the discriminator can distinguish the original chinese herbal prescription vector sequence from the generated chinese herbal prescription vector sequence, which specifically includes:
fixing the generator network parameters, inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into a discriminator of the GAN, training, obtaining the errors of the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence according to a loss function, reversely spreading the errors, and updating the parameters in the discriminator so that the discriminator can distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence.
Optionally, collecting the data of the Chinese medicinal formulae, and forming an original vector sequence of the Chinese medicinal formulae according to the data of the Chinese medicinal formulae, specifically comprising:
before collecting the data of the traditional Chinese medicine prescription, designing a coding neural network and coding each medicine;
adding labels to the medicinal materials by combining clinical data to generate medicinal material component vectors, and combining the medicinal material component vectors to form an original Chinese medicinal prescription vector sequence;
structuring the Chinese medicinal prescriptions issued by the Chinese physicians to obtain an original Chinese medicinal prescription vector sequence, and inputting the original Chinese medicinal prescription vector sequence into the sequence encoder.
Optionally, the patient label and the weight condition specifically include:
performing structured treatment according to the state of illness of the patient to form a patient label, wherein the label comprises symptoms and severity;
setting weight conditions according to patient specific conditions, wherein the specific conditions comprise: pay attention to price and therapeutic effect.
Optionally, inputting the weight condition, the patient label, the random vector and the original Chinese medicinal prescription vector sequence into the generator to form the generated Chinese medicinal prescription vector sequence;
and executing the steps for multiple times to form multiple groups of the generated vector sequences of the traditional Chinese medicine prescriptions.
Optionally, inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into the discriminator, and outputting a comprehensive score and a total price of each Chinese medicine prescription;
and determining the updating of the Chinese medicinal prescriptions according to the generated Chinese medicinal prescription vector sequences and the discrimination results.
Optionally, according to the recovery condition of the patient after taking the medicine, the result is fed back to form a new original vector sequence of the traditional Chinese medicine prescription, and the improved model of the traditional Chinese medicine prescription is continuously optimized.
Optionally, the vector of the medicinal material components specifically includes:
the code, dosage and price of each medicinal material.
The invention provides a method for improving a traditional Chinese medicine prescription based on GAN, which utilizes GAN and deep learning technology, fully combines clinical data and disease condition data, and constructs an improved network model of the traditional Chinese medicine prescription. Compared with the traditional generation mode technology, the GAN can be used for better generating more reasonable traditional Chinese medicine formulas, mass traditional Chinese medicine formula data are marked with multiple labels, the rationality and the accuracy of the formulas are improved, meanwhile, the price and the component factors of the medicinal materials formed by the traditional Chinese medicine formulas are considered, more serious traditional Chinese medicine formulas can be generated based on a model according to the disease conditions, and the personalized requirements of traditional Chinese medicine doctors and patients can be better met; the model can generate a plurality of groups of Chinese medicinal formulas and provide evaluation to form a plurality of Chinese medicinal formula candidates, and assists a Chinese medical practitioner to improve the rationality of the Chinese medicinal formulas, improve the Chinese medicinal efficacy and curative effect and help a patient to recover as soon as possible. In addition, the data optimization model fed back by the patient is continuously collected, so that the model accuracy is further improved, and the auxiliary effect of the traditional Chinese medicine prescription for the middle-aged doctor is further improved.
Drawings
The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:
FIG. 1 is a flow chart of a method for improving a GAN-based Chinese medicinal formulation provided in an embodiment of the present application;
fig. 2 is a structural diagram of an improved model of a traditional Chinese medicine prescription provided in the embodiment of the present application.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be described in detail and completely with reference to the following specific embodiments. It should be apparent that the described embodiments are only some of the embodiments of the present application, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present application.
In the prior art, the artificial intelligence technology is developed rapidly, the commercialization speed of the artificial intelligence technology is beyond expectation, the artificial intelligence brings subversive changes to the whole society, and the artificial intelligence technology becomes an important development strategy for countries in the future. Particularly, the algorithm evolution taking deep learning as a core has the super-strong evolutionary capability, and under the support of big data, a large-scale neural network similar to a human brain structure is obtained through training and construction, so that various problems can be solved.
Generation of a countermeasure network (GAN) is one of the most important methods for unsupervised learning on complex distribution in recent years, and is currently widely applied to the generation field, and GAN can generate impressive results and control smooth and reasonable semantic changes, and becomes the most important generation model framework for learning arbitrary complex data distribution.
In the world cultural science and technology history, the traditional Chinese medicine is the only cultural and technological wonderful trace which can be vigorous and vigorous after more than 2000 years. The traditional Chinese medicine is accumulated by the ancestors in long-term medical practice and is an important component of ancient elegant cultural heritage in China. Because the traditional Chinese medicines are mainly prepared from plant medicinal materials and are most commonly used, the medicine is called 'herbal' in ancient times, and the herbal has achieved unprecedented achievements with the development of modern natural science and technology and national economy. Through long-term medical practice, a plurality of medicines are combined and decocted to prepare a decoction, so that a traditional Chinese medicine formula is formed.
The traditional Chinese medicine prescription is an intelligent crystal and component of traditional Chinese medicine culture, a traditional Chinese medicine doctor prescribes the prescription according to the actual condition of a patient, and the treatment effect of the patient depends on the experience of the traditional Chinese medicine doctor to a certain extent. Under the circumstances, how to effectively utilize the GAN and the deep learning technology and improve the traditional Chinese medicine prescription by combining the actual condition of the patient becomes a problem to be solved urgently.
The solution of the present application can solve the above problems, which will be described in detail below.
Fig. 1 is a flow chart of an improved GAN-based herbal formulation provided in an embodiment of the present application, which may include the following steps:
a method for improving a traditional Chinese medicine prescription based on GAN is characterized by comprising the following steps:
collecting Chinese medicine prescription data, and forming an original Chinese medicine prescription vector sequence according to the Chinese medicine prescription data;
inputting the original vector sequence of the Chinese medicinal prescription into a generator of the GAN, training and outputting the generated vector sequence of the Chinese medicinal prescription;
inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into a recognizer of the GAN, training and updating parameters in the recognizer, and enabling the recognizer to distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence;
alternately training the generator and the discriminator to obtain the improved model of the Chinese medicine prescription.
In an embodiment of the present application, a large amount of data of a chinese medicine prescription, for example, data of a chinese medicine prescription prescribed by a chinese medicine doctor or data of a chinese medicine prescription imported from a database, is collected and structured, a neural network is designed according to the composition and amount of medicinal materials in the chinese medicine prescription, and the neural network is encoded to form a vector set, and simultaneously, a corresponding label is marked for a patient's condition of actual medication in combination with clinical data to form a data set of a chinese medicine prescription to be trained, that is, an original chinese medicine prescription vector sequence.
As shown in fig. 2, GAN is used as a core structure of an improved model of a traditional Chinese medicine prescription, an original traditional Chinese medicine prescription vector sequence is input into a generator G of GAN for training, parameters of a discriminator D are fixed during training of the generator G, the original traditional Chinese medicine prescription vector sequence is disturbed, the traditional Chinese medicine prescription sequence vector is composed of medicinal material component vectors, namely, the same medicinal material component vector is input into the generator G for multiple times according to different sequences, so that the number of training samples is increased, the parameters of the generator G are updated, the discriminator D cannot distinguish whether the traditional Chinese medicine prescription vector sequence is the original traditional Chinese medicine prescription vector sequence or the generated traditional Chinese medicine prescription vector sequence, and the generated traditional Chinese medicine prescription vector sequence is finally output.
Training the discriminator D after the generator G finishes training, wherein the core of the discriminator D is a binary classifier, and an original Chinese medicine prescription vector sequence and a patient label are input, so as to distinguish whether the Chinese medicine prescription vector sequence is from the original or is generated by the generator and meet the requirement of the patient disease condition label, so that the discriminator D can distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence, thereby improving, fixing the network parameters of the generator G, inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into the discriminator D of the GAN, training the discriminator D, reversely propagating errors, updating the parameters in the discriminator D, and enabling the discriminator D to distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence.
The training of the Chinese medicinal prescription improvement model is completed through the alternate training generator G and the discriminator D, and the model is finally formed for improving the Chinese medicinal prescription prescribed by the Chinese medical practitioner, so that the Chinese medical practitioner is assisted to improve the rationality of the Chinese medicinal prescription, improve the Chinese medicinal efficacy and curative effect, and help the patient to recover as soon as possible.
Optionally, inputting the original vector sequence of the chinese herbal prescription into a generator of GAN, training, and outputting the generated vector sequence of the chinese herbal prescription, specifically comprising:
fixing the parameters of the discriminator, disordering the original Chinese medicine prescription vector sequence and inputting the disordering into the generator, updating the parameters of the generator, so that the discriminator can not distinguish whether the Chinese medicine prescription vector sequence is the original Chinese medicine prescription vector sequence or the generated Chinese medicine prescription vector sequence, and finally outputting the generated Chinese medicine prescription vector sequence.
In an embodiment of the application, GAN is used as a core structure of a traditional Chinese medicine prescription improvement model, an original traditional Chinese medicine prescription vector sequence is input into a generator of GAN for training, parameters of a discriminator are fixed during training of the generator, the original traditional Chinese medicine prescription vector sequence is disordered, the traditional Chinese medicine prescription vector is composed of medicinal material component vectors, namely, the same medicinal material component vector is input into the generator for multiple times according to different sequences, so that the number of training samples is increased, the generator parameters are continuously trained, the discriminator cannot distinguish whether the traditional Chinese medicine prescription vector sequence is the original traditional Chinese medicine prescription vector sequence or the generated traditional Chinese medicine prescription vector sequence, and finally the generated traditional Chinese medicine prescription vector sequence is output.
Optionally, the generator specifically includes:
a sequence encoder and a prescription sequence generator;
the sequence encoder is a recurrent neural network and is used for extracting characteristics according to an original Chinese medicine prescription vector sequence to generate an encoding vector, and the prescription sequence generator is a recurrent neural network and is used for inputting the encoding vector, a patient label, a weight condition and a random vector and outputting the generated Chinese medicine prescription vector sequence.
As shown in fig. 2, in an embodiment of the present application, the generator G includes a sequence encoder SeqEnc and a prescription sequence generator SeqGen, and is responsible for outputting a generated vector sequence of a chinese medicine prescription through a generator model according to an input vector sequence of a chinese medicine prescription, a patient disease label and a weight condition, where the sequence encoder SeqEnc is a recurrent neural network, and the recurrent neural network has memory, parameter sharing and Turing completion (training completion), so that it has certain advantages in learning a non-linear characteristic of a sequence. Extracting the characteristics of the vector sequence of the original traditional Chinese medicine prescription to generate a coding vector; the prescription sequence generator SeqGen is a circulating neural network, and a coding vector generated by an input sequence encoder SeqEnc is input into the prescription sequence generator SeqGen for training by combining a patient disease label, a weight condition and a random vector, and the generated Chinese medicine prescription vector sequence is output by the SeqGen.
Optionally, inputting the original vector sequence of the chinese herbal medicine prescription and the generated vector sequence of the chinese herbal medicine prescription into a discriminator of the GAN, training and updating parameters in the discriminator, so that the discriminator can distinguish the original vector sequence of the chinese herbal medicine prescription from the generated vector sequence of the chinese herbal medicine prescription, specifically including:
fixing generator network parameters, inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into a GAN discriminator, training, obtaining the errors of the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence according to a loss function, reversely spreading the errors, and updating the parameters in the discriminator so that the discriminator can distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence.
In one embodiment of the present application, the identifier D is trained after the generator G is trained, the core of the identifier D is a binary classifier, which inputs an original vector sequence of the chinese herbal medicine prescription and a patient label, and aims to distinguish whether the vector sequence of the chinese herbal medicine prescription is from the original or from the generator, and simultaneously satisfies the requirement of the patient's disease label, so that the identifier D can distinguish the original vector sequence of the chinese herbal medicine prescription from the generated vector sequence of the chinese herbal medicine prescription, thereby improving, fixing the network parameters of the generator G, inputting the original vector sequence of the chinese herbal medicine prescription and the generated vector sequence of the chinese herbal medicine prescription into the identifier D of GAN, training the identifier D, obtaining the errors of the original vector sequence of the chinese herbal medicine and the generated vector sequence of the chinese herbal medicine according to a loss function, propagating the error backwards, updating the parameters in the identifier D, so that the discriminator D can discriminate the original vector sequence of the chinese medicinal prescription from the generated vector sequence of the chinese medicinal prescription.
Optionally, collecting the data of the chinese medicinal formulae, and forming an original vector sequence of the chinese medicinal formulae according to the data of the chinese medicinal formulae, specifically including:
before collecting the data of the traditional Chinese medicine prescription, designing a coding neural network and coding each medicine;
combining clinical data, adding labels to medicinal materials to generate medicinal material component vectors, and combining the medicinal material component vectors to form an original Chinese medicinal prescription vector sequence.
Structuring the Chinese medicinal prescriptions issued by the Chinese physicians to obtain an original Chinese medicinal prescription vector sequence, and inputting the original Chinese medicinal prescription vector sequence into a sequence encoder.
In an embodiment of the application, medicinal materials are used as data and are trained through a neural network, the medicinal materials are coded firstly, the more similar the medicinal materials are, the closer the distance obtained by coding is, the coding neural network is designed, each medicinal material is coded, medicinal material labels are added, such as medicinal material name labels, effect labels and the like, each medicinal material generates a medicinal material component vector, the medicinal material component vector is formed by adopting a v (c, w, p) triple sequence, c represents medicinal material coding, w represents dosage, and p represents price; taking Xiaoqinglong decoction as an example, the medicinal materials comprise ephedra (c1), peony (c2), asarum (c3), dried ginger (c4), liquorice (c5), cassia twig (c6), schisandra (c7) and pinellia ternate (c8), one dose comprises ephedra (9g), peony (9g), asarum (3g), dried ginger (5g), liquorice (6g), cassia twig (6g), schisandra (6g) and pinellia ternate (9g), and a vector sequence of medicinal material components is formed, wherein v1 is a vector sequence
(c1,9, p1), v2(c2,9, p2), v3(c3,3, p3), v4(c4,5, p4), v5(c5,6, p5), v6(c6,6, p6), v7(c7,6, p7), v8(c8,9, p8), the sequences are combined to form the original Chinese medicine prescription vector sequence, and the medicine component vectors can be input into a sequence encoder in a disorganized sequence.
Optionally, the patient label and the weight condition specifically include:
carrying out structuring treatment according to the state of illness of a patient to form a patient label, wherein the label comprises symptoms and severity;
setting a weight condition according to the specific condition of the patient, wherein the specific condition comprises the following steps: pay attention to price and therapeutic effect.
In one embodiment of the present application, the patient label may be labeled with a label that reflects the patient's condition, including symptoms, severity of the condition, category of the condition, patient's physical condition, etc.; the patient label is exemplified by fever, including the code and extent of its fever, e.g., greater than 40 high fever (f1, 9), low fever (f1, 1). The weight condition indicates the weight of the prescription, such as the preference of the patient to pay attention to price, treatment effect, children medication, low toxicity of the medicinal materials and the like.
Optionally, inputting the weight condition, the patient label, the random vector and the original Chinese medicine prescription vector sequence into a generator to form a generated Chinese medicine prescription vector sequence;
and (4) executing the steps for multiple times to form multiple groups of generated vector sequences of the Chinese medicinal prescriptions.
In an embodiment of the present application, the random vector is a vector formed by a random number, so as to ensure diversity of output results, and by adding the random number, a plurality of groups of vector sequences of the traditional Chinese medicine formula generated by the formula sequence generator SeqGen can be formed, and the outputted vector sequences of the plurality of groups of traditional Chinese medicine formula are used for a traditional Chinese medicine practitioner to select, so that the prescribed formula is more reasonable, for example, the patient has a serious illness, but only paying attention to the price due to the money problem, and then the traditional Chinese medicine practitioner selects a more appropriate and reasonable formula according to the generated vector sequences of the plurality of groups of traditional Chinese medicine formula.
Optionally, inputting the original Chinese medicinal prescription vector sequence and the generated Chinese medicinal prescription vector sequence into a discriminator, and outputting the comprehensive score and the total price of each Chinese medicinal prescription;
and determining the updating of the Chinese medicinal prescriptions according to the vector sequences of the Chinese medicinal prescriptions generated by the groups and the discrimination result.
In one embodiment of the present application, the chinese herbal prescriptions are continuously updated according to the patient's condition, and the identifier D compares the generated chinese herbal prescription vector sequence with the original chinese herbal prescription vector sequence, scores the similarity, and marks the price of each outputted chinese herbal prescription vector sequence, and selects the appropriate one according to the replaceability of the herbs. And determining the updating of the Chinese medicinal prescriptions according to the vector sequences of the plurality of groups of generated Chinese medicinal prescriptions and the discrimination result of the similarity of the vector sequences of the two Chinese medicinal prescriptions.
Optionally, according to the recovery condition of the patient after taking the medicine, the result is fed back to form a new original vector sequence of the traditional Chinese medicine prescription, and the improved model of the traditional Chinese medicine prescription is continuously optimized.
In one embodiment of the present application, the recovery status of the patient after taking the medicine is known, and the results are fed back to form a new data set to be trained, i.e. a new original vector sequence of the prescription of Chinese medicine, to continuously optimize the prescription improvement model.
Optionally, the vector of the medicinal material components specifically includes:
the code, dosage and price of each medicinal material.
In one embodiment of the application, the vector of the medicinal material components is formed by a v (c, w, p) triple sequence, wherein c represents the medicinal material code, w represents the dosage, and p represents the price; other attribute elements such as drug properties, drug effects and the like can be added into the vector to form a multi-element vector, and the formation of a specific vector is determined according to specific conditions.
The above are merely examples of the present application and are not intended to limit the present application. Various modifications and changes may occur to those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.
Claims (10)
1. A method for improving a traditional Chinese medicine prescription based on GAN is characterized by comprising the following steps:
collecting Chinese medicine prescription data, and forming an original Chinese medicine prescription vector sequence according to the Chinese medicine prescription data;
inputting the original Chinese medicine prescription vector sequence into a generator for generating an antagonistic network GAN, training and outputting the generated Chinese medicine prescription vector sequence;
inputting the original Chinese medicinal prescription vector sequence and the generated Chinese medicinal prescription vector sequence into a discriminator of the GAN, training and updating parameters in the discriminator, so that the discriminator can distinguish the original Chinese medicinal prescription vector sequence from the generated Chinese medicinal prescription vector sequence;
and training the generator and the discriminator alternately to obtain the improved model of the Chinese medicine prescription.
2. The method of claim 1, wherein inputting the original vector sequence of chinese herbal prescriptions into a GAN generator, training, and outputting the generated vector sequence of chinese herbal prescriptions comprises:
fixing the parameters of the discriminator, disordering the original Chinese medicine prescription vector sequence and inputting the disordering into the generator, updating the parameters of the generator, so that the discriminator can not distinguish whether the Chinese medicine prescription vector sequence is the original Chinese medicine prescription vector sequence or the generated Chinese medicine prescription vector sequence, and finally outputting the generated Chinese medicine prescription vector sequence.
3. The method according to claim 1, wherein the generator specifically comprises:
a sequence encoder and a prescription sequence generator;
the sequence encoder is a recurrent neural network and is used for generating an encoding vector according to the extracted features of the original Chinese medicine prescription vector sequence, and the prescription sequence generator is a recurrent neural network and is used for inputting the encoding vector, the patient label, the weight condition and the random vector and outputting the generated Chinese medicine prescription vector sequence.
4. The method of claim 1, wherein the original sequence of chinese herbal prescription vectors and the generated sequence of chinese herbal prescription vectors are input into a discriminator of the GAN, and parameters in the discriminator are trained to update, so that the discriminator can distinguish between the original sequence of chinese herbal prescription vectors and the generated sequence of chinese herbal prescription vectors, specifically comprising:
fixing the generator network parameters, inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into the identifier of the GAN, training, obtaining the errors of the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence according to a loss function, reversely spreading the errors, and updating the parameters in the identifier so that the identifier can distinguish the original Chinese medicine prescription vector sequence from the generated Chinese medicine prescription vector sequence.
5. The method of claim 1, wherein collecting chinese prescription data and forming an original chinese prescription vector sequence based on the chinese prescription data comprises:
before collecting the data of the traditional Chinese medicine prescription, designing a coding neural network and coding each medicine;
adding labels to the medicinal materials by combining clinical data to generate medicinal material component vectors, and combining the medicinal material component vectors to form an original Chinese medicinal prescription vector sequence;
structuring the Chinese medicinal prescriptions issued by the Chinese physicians to obtain an original Chinese medicinal prescription vector sequence, and inputting the original Chinese medicinal prescription vector sequence into the sequence encoder.
6. The method according to claim 3, wherein the patient label, the weighting condition, specifically comprises:
performing structured treatment according to the state of illness of the patient to form a patient label, wherein the label comprises symptoms and severity;
setting weight conditions according to patient specific conditions, wherein the specific conditions comprise: pay attention to price and therapeutic effect.
7. The method of claim 3, further comprising:
inputting the weight condition, the patient label, the random vector and the original Chinese medicinal prescription vector sequence into the generator to form the generated Chinese medicinal prescription vector sequence;
and executing the steps for multiple times to form multiple groups of the generated vector sequences of the traditional Chinese medicine prescriptions.
8. The method of claim 7, further comprising:
inputting the original Chinese medicine prescription vector sequence and the generated Chinese medicine prescription vector sequence into the discriminator, and outputting a comprehensive score and the total price of each Chinese medicine prescription;
and determining the updating of the Chinese medicinal prescriptions according to the generated Chinese medicinal prescription vector sequences and the discrimination results.
9. The method of claim 1, further comprising:
and feeding back the result according to the recovery condition of the patient after the medicine is taken to form a new original Chinese medicine prescription vector sequence and continuously optimize the Chinese medicine prescription improvement model.
10. The method of claim 5, wherein the vector of medicinal material components specifically comprises:
the code, dosage and price of each medicinal material.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011349053.6A CN112420153B (en) | 2020-11-26 | 2020-11-26 | Method for improving traditional Chinese medicine prescription based on GAN |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011349053.6A CN112420153B (en) | 2020-11-26 | 2020-11-26 | Method for improving traditional Chinese medicine prescription based on GAN |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112420153A true CN112420153A (en) | 2021-02-26 |
| CN112420153B CN112420153B (en) | 2022-11-15 |
Family
ID=74842178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011349053.6A Active CN112420153B (en) | 2020-11-26 | 2020-11-26 | Method for improving traditional Chinese medicine prescription based on GAN |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112420153B (en) |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109243567A (en) * | 2018-08-14 | 2019-01-18 | 山东科技大学 | A kind of drug recommended method excavated based on prescription data |
| CN109509056A (en) * | 2018-10-16 | 2019-03-22 | 平安科技（深圳）有限公司 | Method of Commodity Recommendation, electronic device and storage medium based on confrontation network |
| CN110689939A (en) * | 2019-08-27 | 2020-01-14 | 天津开心生活科技有限公司 | Recommendation method and device for medication sequence, readable medium and electronic equipment |
| CN111312356A (en) * | 2020-01-17 | 2020-06-19 | 四川大学 | Traditional Chinese medicine prescription generation method based on BERT and integration efficacy information |
| CN111883263A (en) * | 2020-07-29 | 2020-11-03 | 济南浪潮高新科技投资发展有限公司 | Method, device, equipment and storage medium for assisting in judging drug effect of traditional Chinese medicine prescription |
| CN111968715A (en) * | 2020-06-30 | 2020-11-20 | 厦门大学 | Drug recommendation modeling method based on medical record data and drug interaction risk |
| CN113378991A (en) * | 2021-07-07 | 2021-09-10 | 上海联影医疗科技股份有限公司 | Medical data generation method and device, electronic equipment and storage medium |
-
2020
- 2020-11-26 CN CN202011349053.6A patent/CN112420153B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109243567A (en) * | 2018-08-14 | 2019-01-18 | 山东科技大学 | A kind of drug recommended method excavated based on prescription data |
| CN109509056A (en) * | 2018-10-16 | 2019-03-22 | 平安科技（深圳）有限公司 | Method of Commodity Recommendation, electronic device and storage medium based on confrontation network |
| CN110689939A (en) * | 2019-08-27 | 2020-01-14 | 天津开心生活科技有限公司 | Recommendation method and device for medication sequence, readable medium and electronic equipment |
| CN111312356A (en) * | 2020-01-17 | 2020-06-19 | 四川大学 | Traditional Chinese medicine prescription generation method based on BERT and integration efficacy information |
| CN111968715A (en) * | 2020-06-30 | 2020-11-20 | 厦门大学 | Drug recommendation modeling method based on medical record data and drug interaction risk |
| CN111883263A (en) * | 2020-07-29 | 2020-11-03 | 济南浪潮高新科技投资发展有限公司 | Method, device, equipment and storage medium for assisting in judging drug effect of traditional Chinese medicine prescription |
| CN113378991A (en) * | 2021-07-07 | 2021-09-10 | 上海联影医疗科技股份有限公司 | Medical data generation method and device, electronic equipment and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112420153B (en) | 2022-11-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN104834735B (en) | A kind of documentation summary extraction method based on term vector | |
| CN106228023B (en) | A kind of clinical path method for digging based on ontology and topic model | |
| CN108630322A (en) | Drug interaction modeling and methods of risk assessment, terminal device and storage medium | |
| CN109920540A (en) | Construction method, device and the computer equipment of assisting in diagnosis and treatment decision system | |
| CN107016438A (en) | A kind of system based on Chinese medical discrimination artificial neural network algorithm model | |
| Pradeep et al. | Predictive analysis of diabetes using J48 algorithm of classification techniques | |
| Tang et al. | A comparative study of medical data classification methods based on decision tree and system reconstruction analysis | |
| CN111785366B (en) | Patient treatment scheme determination method and device and computer equipment | |
| CN115050481B (en) | Traditional Chinese medicine prescription efficacy prediction method based on graph convolution neural network | |
| CN108509419A (en) | Ancient TCM books document participle and part of speech indexing method and system | |
| CN109273098A (en) | A kind of pharmaceutical effectiveness prediction technique and device based on intelligent decision | |
| CN106845118A (en) | A kind of auxiliary based on electronic health record shared document examines medication system and method | |
| CN111883263A (en) | Method, device, equipment and storage medium for assisting in judging drug effect of traditional Chinese medicine prescription | |
| CN112420153B (en) | Method for improving traditional Chinese medicine prescription based on GAN | |
| CN113539412A (en) | Chinese herbal medicine recommendation system based on deep learning | |
| Reitz et al. | Using the unified medical language system to expand the operative stress score–first use case | |
| Zhang et al. | Epileptic EEG signals recognition using a deep view-reduction tsk fuzzy system with high interpretability | |
| Wang et al. | Application of physical examination data on health analysis and intelligent diagnosis | |
| CN111180045B (en) | Method for mining relation between drug pairs and efficacy from prescription information | |
| CN115240873A (en) | Medicine recommendation method based on machine learning, electronic equipment and computer-readable storage medium | |
| Tanuku et al. | Liver Disease Prediction Using Ensemble Technique | |
| Park et al. | Interdisciplinary papers supported by disciplinary grants garner deep and broad scientific impact | |
| Wang et al. | Identifying Chinese Herbal Medicine by Image with Three Deep CNNs | |
| Thai et al. | UIT-ViCoV19QA: A Dataset for COVID-19 Community-based Question Answering on Vietnamese Language | |
| Junejo et al. | RETRACTED ARTICLE: Molecular communication networks: drug target scalability based on artificial intelligence prediction techniques |
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 | ||
| TA01 | Transfer of patent application right | ||
| TA01 | Transfer of patent application right |
Effective date of registration: 20220916 Address after: 250101 building S02, 1036 Chaochao Road, high tech Zone, Jinan City, Shandong Province Applicant after: Shandong Inspur Scientific Research Institute Co.,Ltd. Address before: Floor 6, Chaochao Road, Shandong Province Applicant before: JINAN INSPUR HIGH-TECH TECHNOLOGY DEVELOPMENT Co.,Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |