CN109887599A - A kind of tcm prescription curative effect deduction method neural network based - Google Patents
A kind of tcm prescription curative effect deduction method neural network based Download PDFInfo
- Publication number
- CN109887599A CN109887599A CN201910137999.7A CN201910137999A CN109887599A CN 109887599 A CN109887599 A CN 109887599A CN 201910137999 A CN201910137999 A CN 201910137999A CN 109887599 A CN109887599 A CN 109887599A
- Authority
- CN
- China
- Prior art keywords
- pulse
- data
- period
- output
- neural network
- 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
Landscapes
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a kind of tcm prescription curative effect deduction methods neural network based, ready data are input in the model of tcm prescription curative effect deduction, by neural network, export the laboratory indexes after patient takes medicine N days (a cycle), by three such periods, come medical drugs in instructing.The neural network is made of, the input of the output of the output layer in previous period and full articulamentum and other data as current period n convolutional layer and a full articulamentum.This method can instruct in medical drugs.
Description
Technical field
It makes up a prescription field more particularly to a kind of tcm prescription curative effect neural network based deduction side the invention belongs to Chinese medicine
Method.
Background technique
Traditional Chinese medicine has thousands of years of developing history, is summary of experience of the numerous working people in treatment disease, Chinese medicine
Diagnoses and treatment process, a substantially diagnosis and treatment based on an overall analysis of the illness and the patient's condition process and it is a deal with to a large body of facts information, extract
The process of rule.Diagnosis and treatment is the essence of traditional Chinese medicine, is theory and means of the Chinese medicine for medical diagnosis on disease and treatment.Dialectical opinion
The process controlled is collection, conversion, processing, processing and the feedback with information to information, adjusts remedy measures, reaches healing
The purpose of disease.Dialectical in traditional Chinese medicine has the characteristics that ambiguity, uncertain, subjective, thus the diagnosis of Chinese medicine and
Treatment and experience, the level of doctor have compared with Important Relations.The research ideas and methods of Chinese medical discrimination be concentrated mainly on experimental study,
Clinical observation, documents management, in summary of experience.The development of current era shows that these methods are inadequate.Neural network have compared with
The ability for obtaining data rule well, being applied to traditional Chinese medicine has feasibility.
The key to solve the above problems is exactly to apply neural network, carries out the deduction of tcm prescription curative effect.
Summary of the invention
The present invention overcomes in place of the deficiencies in the prior art, propose a kind of tcm prescription curative effect neural network based to push away
Drill method, it is intended that medical drugs in guidance.
The present invention is to adopt the following technical scheme that up to foregoing invention purpose
A kind of tcm prescription curative effect deduction method neural network based carries out in accordance with the following steps:
Step (1): prepare the data that tcm prescription curative effect is deduced；
Step (2): ready data being input in the model of tcm prescription curative effect deduction, defeated by neural network
Laboratory indexes after patient's N days (period 1) of medication out；By the inspection of laboratory indexes and prescriptions of traditional Chinese medicine, patient after N days
Data of the index as second round, and the output in a upper periodic network is input to together in the network in next period, it passes through
Cross neural network, the laboratory indexes after output patient's 2N days (second round) of medication；By the laboratory indexes and Chinese medicine after 2N days
Prescription, patient data of the Index for examination as the period 3, and the output of a upper periodic network is input to next week together
In the network of phase, by neural network, exports patient and take medicine the laboratory indexes after 3N days (period 3), Chinese medicine sees a doctor general warp
Go through three periods；
Step (3): training: assigning initialization values to network parameter, the maximum number of iterations Q of network, learning rate λ be arranged,
Ready data set is inputted into network, is trained, if Loss value declines always, continues to train, until iteration Q times
Afterwards, final model is obtained, if Loss value tends towards stability halfway, stops iteration, obtains final model；
Step (4): it deduces: inputting all data of patientJ indicates that jth kind is sick, i the i-th period of expression (i=1,2,
3) it, is input in the model of tcm prescription curative effect deduction, obtains the laboratory indexes in each period.
Further, the data preparation in step (1) is as follows:
Collect the prescriptions of traditional Chinese medicine of a kind disease, the inspection of patient and analysis data, wherein prescriptions of traditional Chinese medicine includes the weight of n kind drug
(in grams) and medicining times (m times/day) are measured, check that data include the height, weight, body temperature, blood pressure, tongue of patient
Tongue fur, pulse condition, wherein tongue fur be divided into thin and whitish fur, whiten tongue fur, white and greasy tongue coating, powder-like fur, white dry tongue fur, thin and yellowish fur, yellowish greasy tongue, yellow rough tongue fur,
Totally 14 kinds of basic tongue furs are divided into surface pulse which can be felt when touched only lightly, full pulse, soft pulse, scattered as, pulse condition for the sliding tongue fur of Huang, sallow tongue fur, deep yellow tongue fur, grey and black coat, grayish fur, black tongue fur
Arteries and veins, hollow pulse, Ge Mai, eddp pulse, deep-sited pulse, firm pulse, weak pulse, retarded pulse, slow arteries and veins, weak thready, nodus, rapid pulse, abrupt pulse, swift pulse, artery, void
Arteries and veins, scarcely perceptible pulse, thready pulse, slow, short arteries and veins, forceful pulse, smooth pulse, tense pulse, long pulse, taut pulse totally 28 kinds of basic pulse conditions, by tongue fur situation and arteries and veins
As using one-hot coding as input data respectively；Analysis data includes physiochemical indice, urine index, liver function index, kidney function
Energy index, blood lipid, electrolyte, amylase, analysis data are that the output of current period is also the input in next period.
The composition of neural network is as follows in step (2):
Neural network is made of multiple convolutional layers, by dataBe input in convolutional neural networks, by n convolutional layer and
One full articulamentum obtains output o(1)；Using the index of period 1 output, the output of full articulamentum and other data as the
The input data of two cyclesOutput o is obtained by n convolutional layer and a full articulamentum(2)；The finger that second round is exported
Input data of the output and other data of mark, full articulamentum as the period 3Connect entirely by n convolutional layer and one
It connects layer and obtains output o(3), wherein j indicates the type of disease.
The loss function of training is as follows in step (3):
Wherein y(t)It is the practical survey of each period
The laboratory indexes (t=1,2,3) measured, o(t)It is the laboratory indexes (t=1,2,3) that each periodic network training obtains.
Detailed description of the invention
Fig. 1 is the flow chart of tcm prescription curative effect deduction method neural network based.
Fig. 2 is the network model of tcm prescription curative effect deduction method neural network based.
Specific embodiment
The present invention will be further explained below with reference to the attached drawings.
In the present embodiment, as shown in Figure 1, tcm prescription curative effect deduction method neural network based of the invention, specifically
Realization mainly includes the following steps:
Step (1): prepare the data that tcm prescription curative effect is deduced；
Step (2): ready data being input in the model of tcm prescription curative effect deduction, defeated by neural network
Laboratory indexes after patient's 10 days (period 1) of medication out；By the inspection of laboratory indexes and prescriptions of traditional Chinese medicine, patient after N days
Data of the index as second round, and the output in a upper periodic network is input to together in the network in next period, it passes through
Cross neural network, the laboratory indexes after output patient's 20 days (second round) of medication；By the laboratory indexes and Chinese medicine after 20 days
Prescription, patient data of the Index for examination as the period 3, and the output of a upper periodic network is input to next week together
In the network of phase, by neural network, exports patient and take medicine the laboratory indexes after 30 days (period 3), Chinese medicine sees a doctor general warp
Go through three periods；
Step (3): training: initialization values are assigned to network parameter, the maximum number of iterations Q=50000 of network is set, are learned
Ready data set is inputted network, is trained, if Loss value declines always, continues to train by habit rate λ=0.01,
After iteration 50000 times, final model is obtained, if Loss value tends towards stability halfway, stops iteration, is obtained final
Model；
Step (4): it deduces: inputting all data of patientJ indicates that jth kind is sick, i the i-th period of expression (i=1,2,
3) it, is input in the model of tcm prescription curative effect deduction, obtains the laboratory indexes in each period.
Further, the data preparation in step (1) is as follows:
Collect the prescriptions of traditional Chinese medicine of a kind disease, the inspection of patient and analysis data, wherein prescriptions of traditional Chinese medicine includes the weight of n kind drug
(in grams) and medicining times (m times/day) are measured, check that data include the height, weight, body temperature, blood pressure, tongue of patient
Tongue fur, pulse condition, wherein tongue fur be divided into thin and whitish fur, whiten tongue fur, white and greasy tongue coating, powder-like fur, white dry tongue fur, thin and yellowish fur, yellowish greasy tongue, yellow rough tongue fur,
Totally 14 kinds of basic tongue furs are divided into surface pulse which can be felt when touched only lightly, full pulse, soft pulse, scattered as, pulse condition for the sliding tongue fur of Huang, sallow tongue fur, deep yellow tongue fur, grey and black coat, grayish fur, black tongue fur
Arteries and veins, hollow pulse, Ge Mai, eddp pulse, deep-sited pulse, firm pulse, weak pulse, retarded pulse, slow arteries and veins, weak thready, nodus, rapid pulse, abrupt pulse, swift pulse, artery, void
Arteries and veins, scarcely perceptible pulse, thready pulse, slow, short arteries and veins, forceful pulse, smooth pulse, tense pulse, long pulse, taut pulse totally 28 kinds of basic pulse conditions, by tongue fur situation and arteries and veins
As using one-hot coding as input data respectively；Analysis data includes physiochemical indice, urine index, liver function index, kidney function
Energy index, blood lipid, electrolyte, amylase, analysis data are that the output of current period is also the input in next period.
The composition of neural network is as follows in step (2):
Neural network is made of multiple convolutional layers, by dataIt is input in convolutional neural networks, by 10 convolutional layers
Output o is obtained with a full articulamentum(1)；Using the period 1 output index, full articulamentum output and other data as
The input data of second roundOutput o is obtained by 10 convolutional layers and a full articulamentum(2)；By second round output
Input data of the output and other data of index, full articulamentum as the period 3By 10 convolutional layers and one
Full articulamentum obtains output o(3), wherein j indicates the type of disease.
The loss function of training is as follows in step (3):
Wherein y(t)It is the practical survey of each period
The laboratory indexes (t=1,2,3) measured, o(t)It is the laboratory indexes (t=1,2,3) that each periodic network training obtains.
Claims (4)
1. a kind of tcm prescription curative effect deduction method neural network based, it is characterized in that carry out in accordance with the following steps:
Step (1): prepare the data that tcm prescription curative effect is deduced；
Step (2): ready data being input in the model of tcm prescription curative effect deduction, and by neural network, output is suffered from
Laboratory indexes after person's N days (period 1) of medication；By the Index for examination of laboratory indexes and prescriptions of traditional Chinese medicine, patient after N days
It is input in the network in next period together as the data of second round, and the output in a upper periodic network, by mind
Laboratory indexes through network, after exporting patient's 2N days (second round) of medication；By after 2N days laboratory indexes and prescriptions of traditional Chinese medicine,
Data of the Index for examination of patient as the period 3, and the output of a upper periodic network is input to together the net in next period
In network, by neural network, exports patient and take medicine the laboratory indexes after 3N days (period 3), Chinese medicine, which is seen a doctor, generally undergoes three
Period；
Step (3): training: assigning initialization values to network parameter, the maximum number of iterations Q of network, learning rate λ be arranged, will be quasi-
The data set input network got ready, is trained, if Loss value declines always, continues to train, after iteration Q times, obtain
Stop iteration if Loss value tends towards stability halfway to final model, obtains final model；
Step (4): it deduces: inputting all data of patientJ indicates jth kind disease, and i indicates the i-th period (i=1,2,3), defeated
Enter in the model deduced to tcm prescription curative effect, obtains the laboratory indexes in each period.
2. a kind of tcm prescription curative effect deduction method neural network based as described in claim 1, which is characterized in that described
Data preparation is as follows in step (1):
Collect the prescriptions of traditional Chinese medicine of a kind disease, the inspection of patient and analysis data, wherein prescriptions of traditional Chinese medicine includes the weight of n kind drug
(in grams) and medicining times (m times/day), check data include the height of patient, weight, body temperature, blood pressure, tongue fur,
Pulse condition, wherein tongue fur is divided into thin and whitish fur, whitens tongue fur, white and greasy tongue coating, powder-like fur, white dry tongue fur, thin and yellowish fur, yellowish greasy tongue, yellow rough tongue fur, Huang
Totally 14 kinds of basic tongue furs are divided into surface pulse which can be felt when touched only lightly, full pulse, soft pulse, scattered as, pulse condition for sliding tongue fur, sallow tongue fur, deep yellow tongue fur, grey and black coat, grayish fur, black tongue fur
Arteries and veins, hollow pulse, Ge Mai, eddp pulse, deep-sited pulse, firm pulse, weak pulse, retarded pulse, slow arteries and veins, weak thready, nodus, rapid pulse, abrupt pulse, swift pulse, artery, void
Arteries and veins, scarcely perceptible pulse, thready pulse, slow, short arteries and veins, forceful pulse, smooth pulse, tense pulse, long pulse, taut pulse totally 28 kinds of basic pulse conditions, by tongue fur situation and arteries and veins
As using one-hot coding as input data respectively；Analysis data includes physiochemical indice, urine index, liver function index, kidney function
Energy index, blood lipid, electrolyte, amylase, analysis data are that the output of current period is also the input in next period.
3. a kind of tcm prescription curative effect deduction method neural network based as described in claim 1, which is characterized in that described
The composition of neural network is as follows in step (2):
Neural network is made of multiple convolutional layers, by dataIt is input in convolutional neural networks, by n convolutional layer and one
Full articulamentum obtains output o(1)；Using the index of period 1 output, the output of full articulamentum and other data as second week
The input data of phaseOutput o is obtained by n convolutional layer and a full articulamentum(2)；By the index of second round output, entirely
The input data of the output of articulamentum and other data as the period 3By n convolutional layer and a full articulamentum
Obtain output o(3), wherein j indicates the type of disease.
4. a kind of tcm prescription curative effect deduction method neural network based as described in claim 1, which is characterized in that described
The loss function of training is as follows in step (3):
Wherein y(t)It is that each period actual measurement obtains
The laboratory indexes (t=1,2,3) arrived, o(t)It is the laboratory indexes (t=1,2,3) that each periodic network training obtains.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910137999.7A CN109887599B (en) | 2019-02-25 | 2019-02-25 | Neural network-based traditional Chinese medicine prescription curative effect deduction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910137999.7A CN109887599B (en) | 2019-02-25 | 2019-02-25 | Neural network-based traditional Chinese medicine prescription curative effect deduction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109887599A true CN109887599A (en) | 2019-06-14 |
| CN109887599B CN109887599B (en) | 2021-07-06 |
Family
ID=66929318
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910137999.7A Active CN109887599B (en) | 2019-02-25 | 2019-02-25 | Neural network-based traditional Chinese medicine prescription curative effect deduction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109887599B (en) |
Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2004152029A (en) * | 2002-10-30 | 2004-05-27 | Fujitsu Ltd | Function prediction support method, function prediction support program and function prediction support device |
| CN1580282A (en) * | 2003-08-13 | 2005-02-16 | 上海第二医科大学附属瑞金医院 | Method for predicting chemical therapy effect for acute medullary cell leucosis patient |
| US7058616B1 (en) * | 2000-06-08 | 2006-06-06 | Virco Bvba | Method and system for predicting resistance of a disease to a therapeutic agent using a neural network |
| CN1977270A (en) * | 2004-06-30 | 2007-06-06 | 布雷克成像有限公司 | Clinical trial phase simulation method and clinical trial phase simulator for drug trials |
| CN102778548A (en) * | 2012-06-21 | 2012-11-14 | 北京工业大学 | Method for forecasting sludge volume index in sewage treatment process |
| CN104346521A (en) * | 2013-08-07 | 2015-02-11 | B·布莱恩·阿维图姆股份公司 | Device and method for predicting intradialytic parameters |
| CN104915560A (en) * | 2015-06-11 | 2015-09-16 | 万达信息股份有限公司 | Method for disease diagnosis and treatment scheme based on generalized neural network clustering |
| CN105335612A (en) * | 2015-10-28 | 2016-02-17 | 南昌大学 | Prediction method for Sorafenib concentration in kidney area on basis of artificial neural network |
| CN107491638A (en) * | 2017-07-28 | 2017-12-19 | 深圳和而泰智能控制股份有限公司 | A kind of ICU user's prognosis method and terminal device based on deep learning model |
| WO2018048575A1 (en) * | 2016-09-07 | 2018-03-15 | Elekta, Inc. | System and method for learning models of radiotherapy treatment plans to predict radiotherapy dose distributions |
| CN108325020A (en) * | 2018-03-09 | 2018-07-27 | 燕山大学 | A kind of Intravenous Anesthesia multi-parameter index supervisory system of closed |
| US20180240549A1 (en) * | 2015-08-19 | 2018-08-23 | University Of Exeter | Computer-Implemented Apparatus And Method For Predicting Performance Of Surgical Strategies |
| CN108956876A (en) * | 2018-07-12 | 2018-12-07 | 浙江大学 | A kind of measurement time delay correcting method of flue gas on-line continuous monitoring system |
| CN108984811A (en) * | 2017-06-05 | 2018-12-11 | 欧阳德方 | A kind of pharmaceutical preparation prescription virtual design and the method and system of assessment |
| CN109171756A (en) * | 2018-06-18 | 2019-01-11 | 广州普麦健康咨询有限公司 | Diabetes index prediction technique and its system based on depth confidence network model |
-
2019
- 2019-02-25 CN CN201910137999.7A patent/CN109887599B/en active Active
Patent Citations (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7058616B1 (en) * | 2000-06-08 | 2006-06-06 | Virco Bvba | Method and system for predicting resistance of a disease to a therapeutic agent using a neural network |
| JP2004152029A (en) * | 2002-10-30 | 2004-05-27 | Fujitsu Ltd | Function prediction support method, function prediction support program and function prediction support device |
| CN1580282A (en) * | 2003-08-13 | 2005-02-16 | 上海第二医科大学附属瑞金医院 | Method for predicting chemical therapy effect for acute medullary cell leucosis patient |
| CN1977270A (en) * | 2004-06-30 | 2007-06-06 | 布雷克成像有限公司 | Clinical trial phase simulation method and clinical trial phase simulator for drug trials |
| CN102778548A (en) * | 2012-06-21 | 2012-11-14 | 北京工业大学 | Method for forecasting sludge volume index in sewage treatment process |
| CN104346521A (en) * | 2013-08-07 | 2015-02-11 | B·布莱恩·阿维图姆股份公司 | Device and method for predicting intradialytic parameters |
| CN104915560A (en) * | 2015-06-11 | 2015-09-16 | 万达信息股份有限公司 | Method for disease diagnosis and treatment scheme based on generalized neural network clustering |
| US20180240549A1 (en) * | 2015-08-19 | 2018-08-23 | University Of Exeter | Computer-Implemented Apparatus And Method For Predicting Performance Of Surgical Strategies |
| CN105335612A (en) * | 2015-10-28 | 2016-02-17 | 南昌大学 | Prediction method for Sorafenib concentration in kidney area on basis of artificial neural network |
| WO2018048575A1 (en) * | 2016-09-07 | 2018-03-15 | Elekta, Inc. | System and method for learning models of radiotherapy treatment plans to predict radiotherapy dose distributions |
| CN108984811A (en) * | 2017-06-05 | 2018-12-11 | 欧阳德方 | A kind of pharmaceutical preparation prescription virtual design and the method and system of assessment |
| CN107491638A (en) * | 2017-07-28 | 2017-12-19 | 深圳和而泰智能控制股份有限公司 | A kind of ICU user's prognosis method and terminal device based on deep learning model |
| CN108325020A (en) * | 2018-03-09 | 2018-07-27 | 燕山大学 | A kind of Intravenous Anesthesia multi-parameter index supervisory system of closed |
| CN109171756A (en) * | 2018-06-18 | 2019-01-11 | 广州普麦健康咨询有限公司 | Diabetes index prediction technique and its system based on depth confidence network model |
| CN108956876A (en) * | 2018-07-12 | 2018-12-07 | 浙江大学 | A kind of measurement time delay correcting method of flue gas on-line continuous monitoring system |
Non-Patent Citations (1)
| Title |
|---|
| 杜文斌: "基于神经网络的冠心病证候诊断标准与药效评价模型研究", 《中国优秀博硕士学位论文全文数据库（博士）医药卫生科技辑》 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109887599B (en) | 2021-07-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Boukhennoufa et al. | Wearable sensors and machine learning in post-stroke rehabilitation assessment: A systematic review | |
| CN104207931B (en) | Face facial point precise positioning and acupuncture and moxibustion prescription, acumoxibustion prescription learning method | |
| CN106228023B (en) | A kind of clinical path method for digging based on ontology and topic model | |
| CN108596069A (en) | Neonatal pain expression recognition method and system based on depth 3D residual error networks | |
| CN103876708B (en) | Wireless mobile information exchange system for pain stress and subjective feeling and pain treatment | |
| Jain et al. | Improving the prediction rate of diabetes using fuzzy expert system | |
| CN109448855A (en) | A kind of diabetes glucose prediction technique based on CNN and Model Fusion | |
| CN104545912A (en) | Cardiac and pulmonary impedance measuring method and device | |
| CN105512489B (en) | A kind of modeling method based on multiple dimensioned heart Thimthy syndrome pathogenesis | |
| CN101647696A (en) | Intelligent system for health diagnosis and treatment | |
| Tan et al. | Short paper: Using BSN for tele-health application in upper limb rehabilitation | |
| CN109326350A (en) | Cerebral infarction clinic diagnosis system | |
| CN107742533A (en) | A kind of method based on dietary nutrient prediction gout | |
| Visco et al. | Artificial intelligence as a business partner in cardiovascular precision medicine: an emerging approach for disease detection and treatment optimization | |
| CN103700052A (en) | Medical information system on basis of government regulation | |
| CN109545374A (en) | A kind of disease based on body gait feature big data automatically tracks, treats and supplying system | |
| CN116312947A (en) | Immersive ankle and foot rehabilitation training method based on upper limb movement signals and electronic equipment | |
| Putri et al. | Human walking gait classification utilizing an artificial neural network for the ergonomics study of lower limb prosthetics | |
| CN105096227A (en) | Visualization processing method for health data | |
| CN110575178A (en) | Diagnosis and monitoring integrated medical system for judging motion state and judging method thereof | |
| CN109887599A (en) | A kind of tcm prescription curative effect deduction method neural network based | |
| CN103761436A (en) | Scientific research data extraction system based on electronic medical records | |
| CN107066816A (en) | Medical treatment guidance method, device and server based on clinical data | |
| Jin et al. | “Medical-and-care wisdom linkage” pension model research and exploration | |
| Suantai et al. | Numerical computation of SEIR model for the Zika virus spreading |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |