CN110223749A - Chinese medical discrimination aid decision-making method based on PCNN network and attention mechanism - Google Patents
Chinese medical discrimination aid decision-making method based on PCNN network and attention mechanism Download PDFInfo
- Publication number
- CN110223749A CN110223749A CN201910517325.XA CN201910517325A CN110223749A CN 110223749 A CN110223749 A CN 110223749A CN 201910517325 A CN201910517325 A CN 201910517325A CN 110223749 A CN110223749 A CN 110223749A
- Authority
- CN
- China
- Prior art keywords
- medical
- attention
- case
- medical record
- pcnn
- 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
- 238000000034 method Methods 0.000 title claims abstract description 31
- 230000007246 mechanism Effects 0.000 title claims abstract description 21
- 208000011580 syndromic disease Diseases 0.000 claims abstract description 72
- 239000013598 vector Substances 0.000 claims abstract description 42
- 208000024891 symptom Diseases 0.000 claims abstract description 33
- 201000010099 disease Diseases 0.000 claims abstract description 21
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 11
- 239000000284 extract Substances 0.000 claims abstract description 7
- 238000005065 mining Methods 0.000 claims abstract description 7
- 238000005516 engineering process Methods 0.000 claims abstract description 6
- 238000012545 processing Methods 0.000 claims abstract description 4
- 230000011218 segmentation Effects 0.000 claims abstract description 4
- 230000004069 differentiation Effects 0.000 claims description 42
- 239000003814 drug Substances 0.000 claims description 19
- 230000008506 pathogenesis Effects 0.000 claims description 14
- 238000011176 pooling Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 5
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000013475 authorization Methods 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 5
- 238000003745 diagnosis Methods 0.000 description 5
- 206010013781 dry mouth Diseases 0.000 description 3
- 210000003734 kidney Anatomy 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 208000000044 Amnesia Diseases 0.000 description 2
- 208000031091 Amnestic disease Diseases 0.000 description 2
- 206010008479 Chest Pain Diseases 0.000 description 2
- 206010033557 Palpitations Diseases 0.000 description 2
- 230000006986 amnesia Effects 0.000 description 2
- 238000007418 data mining Methods 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 206010012601 diabetes mellitus Diseases 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 208000002173 dizziness Diseases 0.000 description 2
- 210000004185 liver Anatomy 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000001356 surgical procedure Methods 0.000 description 2
- 206010063659 Aversion Diseases 0.000 description 1
- 206010010774 Constipation Diseases 0.000 description 1
- 206010011224 Cough Diseases 0.000 description 1
- 206010012735 Diarrhoea Diseases 0.000 description 1
- 208000000059 Dyspnea Diseases 0.000 description 1
- 206010013975 Dyspnoeas Diseases 0.000 description 1
- 208000008454 Hyperhidrosis Diseases 0.000 description 1
- 206010022998 Irritability Diseases 0.000 description 1
- 241000264063 Lethrinus rubrioperculatus Species 0.000 description 1
- 208000019914 Mental Fatigue Diseases 0.000 description 1
- 206010033425 Pain in extremity Diseases 0.000 description 1
- 208000032140 Sleepiness Diseases 0.000 description 1
- 206010041349 Somnolence Diseases 0.000 description 1
- 208000009205 Tinnitus Diseases 0.000 description 1
- 208000031975 Yang Deficiency Diseases 0.000 description 1
- 208000031971 Yin Deficiency Diseases 0.000 description 1
- 230000001684 chronic effect Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 208000013403 hyperactivity Diseases 0.000 description 1
- 230000037315 hyperhidrosis Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 206010029410 night sweats Diseases 0.000 description 1
- 230000036565 night sweats Effects 0.000 description 1
- 206010029446 nocturia Diseases 0.000 description 1
- 230000001717 pathogenic effect Effects 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000002271 resection Methods 0.000 description 1
- 230000001953 sensory effect Effects 0.000 description 1
- 208000013220 shortness of breath Diseases 0.000 description 1
- 230000037321 sleepiness Effects 0.000 description 1
- 231100000886 tinnitus Toxicity 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- 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
- 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
- G16H15/00—ICT specially adapted for medical reports, e.g. generation or transmission thereof
-
- 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)
- Theoretical Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Computing Systems (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Data Mining & Analysis (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Epidemiology (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses a kind of Chinese medical discrimination aid decision-making methods based on point sequence convolutional neural networks (PCNN) and attention (Attention) mechanism, specifically include: carrying out word segmentation processing to every part of case using language technology platform tools；Feature coding is carried out to every part of case using PCNN network, constructing its corresponding feature vector indicates；Syndrome prediction is carried out using Attention mechanism.Point of use sequence CNN network of the present invention automatically extracts the features such as the symptom, the cause of disease, the interpretation of the cause, onset and process of an illness of case, constructs corresponding feature coding, is not required to manually mark feature；It introduces attention mechanism and eliminates noise corpus, distribute different weights to case, improve nicety of grading；The present invention carries out dialectical law mining from many aspects such as the symptoms, the cause of disease, the interpretation of the cause, onset and process of an illness of case data, improves the reliability of dialectical result；Meanwhile the present invention is suitable for the dialectical law discovery of most of diseases, with very high scalability.
Description
Technical Field
The invention relates to a traditional Chinese medicine syndrome differentiation aid decision method, in particular to a traditional Chinese medicine syndrome differentiation aid decision method based on a PCNN (pulse coupled neural network) and an attention mechanism.
Background
The syndrome differentiation and treatment is the main characteristic of the traditional Chinese medicine, namely, the symptoms collected by inspection, smelling, inquiry and resection (four diagnosis) are analyzed and summarized according to the principle of combining four diagnosis and reference, the pathological mechanism, the disease property, the disease position and the relationship between pathogenic factors and vital qi of the disease are determined, so that the disease condition is determined, and corresponding treatment is selected. Syndrome differentiation in traditional Chinese medicine is the theoretical core of disease diagnosis in traditional Chinese medicine and is also a difficult problem in traditional Chinese medicine diagnostics.
In the traditional Chinese medicine dialectical process, a doctor obtains symptoms and physical sign information of a patient through sensory observation and subjective description of the patient, the doctor integrates the patient information according to personal knowledge and experience to obtain a diagnosis result, the accuracy of dialectical differentiation depends on the personal experience, diagnosis skill, cognition level and thinking ability of the doctor to a certain extent, the subjectivity is strong, and the dialectical process is a 'black box theory' and is difficult to interpret. To solve these problems, a large number of dialectical aid decision methods based on data mining technology have been developed, in which the latest research LDA model analyzes clinical cases regarding diabetes to find associated knowledge hidden between symptoms and syndromes, resulting in 7 different syndromes regarding diabetes differentiation, each containing different symptoms.
The existing syndrome differentiation aid decision-making method based on the data mining technology is only suitable for single specific diseases, if the syndrome differentiation result of other diseases is predicted, the syndrome differentiation accuracy is greatly reduced relative to the specific diseases, if the Aprior algorithm is used for mining the syndrome differentiation rule from the clinical medical record data of the Dingshi surgery, the obtained result is only suitable for the disease syndrome differentiation of the Dingshi surgery, and the expansibility is lacked; secondly, when the syndrome differentiation rule is studied according to the medical record data, the study starts from the aspect of symptoms, and the influence of the disease location and pathogenesis on the syndrome differentiation result is ignored, so that the syndrome differentiation result is greatly different from the real result.
Disclosure of Invention
In view of the above, the present invention provides a traditional Chinese medicine syndrome differentiation aid decision method based on a PCNN and an attention mechanism, aiming at the problems of poor expansibility and consideration of only a single influence factor existing in the existing syndrome differentiation decision method, and the method automatically extracts medical plan features based on a point-sorted CNN (PCNN) by combining 3 features of symptoms, causes, pathogenesis and the like influencing syndrome differentiation to predict the syndrome differentiation result; meanwhile, different influence weights of different medical schemes on the syndrome differentiation result are considered, and a Attention (ATT) mechanism is further adopted to carry out weight distribution on all medical schemes, so that a prediction model is optimized.
In order to solve the technical problems, the invention discloses a traditional Chinese medicine syndrome differentiation auxiliary decision method based on a point sorting CNN network and an attention mechanism, which specifically comprises the following steps:
step 1, pretreatment: performing word segmentation processing on each medical case by using a Language Technology Platform (LTP) tool;
step 2, medical record coding: performing feature coding on each medical case by using a point sorting CNN (PCNN) network to construct a corresponding feature vector representation;
and 3, predicting the syndromes by using an Attention authorization mechanism.
Further, step 2 specifically comprises: firstly, each medical case is converted into a feature vector to be represented, and then the corresponding medical case feature vector is constructed by utilizing convolution, sectional maximum pooling and nonlinear conversion.
Further, each medical record is converted into a feature vector for representation, and the specific steps are as follows:
medical case subset M1 extracts: searching the original medical record set for the medical record containing the symptom set S according to the complete word matching method to form a medical record subset M1 ═ M1，…，ml，…，md}；
For medical case M in medical case subset M1l＝{w1，w2，…，wmThe vectorization expression is carried out on (1 ≦ l ≦ d), wherein wiThe symptoms, the causes and the pathogenesis of the medical record contained in the medical record are represented after the medical record is pretreated;
word wi(1. ltoreq. i. ltoreq.m) vector representationThe word code + the position code,
wherein the word codes are as follows: word w with word2veciConverting into a vector form; position coding: according to the word wiWith the first entity w1And tail entity wmConstructing a two-dimensional vector;
medical record mlThe vectorization of (c) is represented as follows:
wherein m is medical table mlThe number of unrepeated features in the above list, i.e., the symptoms, causes, and pathogenesis.
Further, convolution: performing local feature extraction on the medical plan by using convolution operation, specifically:
selecting a convolution kernel: select 3 convolution kernels Wi(i is more than or equal to 1 and less than or equal to 3), randomly initializing convolution kernel WiDynamically adjusting the value of the convolution kernel according to the classification result, and selecting the optimal symptom characteristic, disease location characteristic and disease characteristic;
sliding length: l — 3, step size: 1, stride is equal to 1;
the sliding window is set to q, as follows:
q＝{wi-l+1，wi-l+2，…，wi} (2)
and (3) convolution operation:
wherein p isiTo use a convolution kernel WiThe extracted local features are compared with the local features,is a bias vector.
Further, segmented maximum pooling: combining the local features extracted by the 3 convolution kernels to obtain a vector of fixed length consisting of q +1 parts, performing maximal pooling on each part as follows:
[x]ij＝max(pij) (4)
wherein, [ x ]]ijRepresenting the jth key feature extracted by the ith convolution kernel.
Further, nonlinear transformation constructs the corresponding medical case feature vector:
wherein,is represented by 3 convolution kernels Wi(i is more than or equal to 1 and less than or equal to 3) extracted medical schemeIs represented by the key features of (1).
Further, step 3 uses an Attention mechanism to predict syndromes, specifically:
and (3) weight distribution: all of the cases in the case subset M1 are assigned different weights,
all cases containing this symptom set are assigned a weight:
wherein, αlShow medical caseWeight assignment of elShow medical caseFor the syndromeThe degree of matching is as follows:
in order to show the effect of each medical record during the mining of syndrome differentiation rules, the medical record subset M1 is re-expressed by using weights as follows:
in order to make the output number of the network the same as the number of the syndrome tags for classification of the softmax layer, it is dimension-processed using equation (9):
wherein A is a matrix composed of all syndrome vectors,is a bias vector;
predicting medical case using softmax functionThe probability over syndrome R is as follows:
wherein K is the number of all syndromes in the medical record set, and P represents the medical recordVector representation over K syndromes.
Compared with the prior art, the invention can obtain the following technical effects:
1) the method uses a point sorting CNN (PCNN) to automatically extract the characteristics of symptoms, causes, pathogenesis and the like of the medical record, constructs corresponding characteristic codes and does not need to manually label the characteristics;
2) the invention introduces Attention (ATT) mechanism to eliminate noise corpus, distributes different weights to medical cases and improves classification precision;
3) the invention carries out the mining of syndrome differentiation rules from the aspects of symptoms, causes, pathogenesis and the like of medical record data, thereby improving the reliability of the syndrome differentiation result; meanwhile, the invention is suitable for the syndrome differentiation rule discovery of most diseases and has high expandability.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention and not to limit the invention. In the drawings:
FIG. 1 is a flowchart illustrating the overall syndrome differentiation process according to an embodiment of the present invention;
fig. 2 is a schematic diagram illustrating the principle of case encoding using the PCNN network in the embodiment of the present invention.
Detailed Description
The following embodiments are described in detail with reference to the accompanying drawings, so that how to implement the technical features of the present invention to solve the technical problems and achieve the technical effects can be fully understood and implemented.
The invention carries out syndrome differentiation rule research on the basis of medical record data, starts with 3 characteristics of symptoms, causes and pathogenesis, and converts an original medical record into a vector form formed by the characteristics by using a point-sorting CNN (PCNN); then, using an attention mechanism (ATT) to perform weight distribution on all cases; finally, the syndrome differentiation is performed on the medical scheme by using the softmax function, and the whole process of syndrome differentiation is shown in fig. 1.
Data set: the medical record set M takes all syndromes appearing in the medical record set as labels, and the number of syndromes is assumed to be K.
M＝{m1，m2，m3，…，mn}
Inputting: chronic nephrotic symptom set S ═ S1，s2，…，sv}；
And (3) outputting: a K-dimensional vector, each dimension representing the probability of a respective syndrome label.
The invention discloses a traditional Chinese medicine syndrome differentiation auxiliary decision method based on a point sorting CNN network and an attention mechanism, which specifically comprises the following steps:
step 1, pretreatment: performing word segmentation processing on each medical case by using a Language Technology Platform (LTP) tool; the results are shown below:
leg pain, dizziness, dry mouth, soreness of waist, loose stool, nocturia, deficiency of liver, kidney, liver and kidney, etc.
Step 2, medical record coding: and (3) performing feature coding on each medical case by using a point sorting CNN (PCNN) network to construct a corresponding feature vector representation.
Firstly, each medical case is converted into a feature vector for representation, and then the corresponding medical case feature vector is constructed by utilizing convolution, sectional maximum pooling and nonlinear conversion, and a schematic diagram of the principle is shown in fig. 2.
Converting each medical case into a feature vector for representation, and specifically comprising the following steps:
1) medical case subset M1 extracts: searching the original medical record set for the medical record containing the symptom set S according to the complete word matching method to form a medical record subset M1 ═ M1，…，ml，…，md}；
For medical case M in medical case subset M1l＝{w1，w2，…，wmThe vectorization expression is carried out on (1 ≦ l ≦ d), wherein wiThe symptoms, the causes and the pathogenesis of the medical record contained in the medical record are represented after the medical record is pretreated;
word wi(1. ltoreq. i. ltoreq.m) vector representationThe word code + the position code,
wherein the word codes are as follows: word w with word2veciConverting into a vector form; position coding: according to the word wiWith the first entity w1And tail entity wmConstructing a two-dimensional vector;
medical record mlThe vectorization of (c) is represented as follows:
wherein m is medical table mlThe number of unrepeated features in the above list, i.e., the symptoms, causes, and pathogenesis.
2) Convolution: performing local feature extraction on the medical plan by using convolution operation, specifically:
selecting a convolution kernel: select 3 convolution kernels Wi(i is more than or equal to 1 and less than or equal to 3), randomly initializing convolution kernel WiDynamically adjusting the value of the convolution kernel according to the classification result, and selecting the optimal symptom characteristic, disease location characteristic and disease characteristic;
sliding length: l — 3, step size: 1, stride is equal to 1;
the sliding window is set to q, as follows:
q＝{wi-l+1，wi-l+2，…，wi} (2)
and (3) convolution operation:
wherein p isiTo use a convolution kernel WiThe extracted local features are compared with the local features,is a bias vector.
3) Sectional type maximum pooling: combining the local features extracted by the 3 convolution kernels to obtain a vector of fixed length consisting of q +1 parts, performing maximal pooling on each part as follows:
[x]ij＝max(pij) (4)
wherein, [ x ]]ijRepresents the jth key feature extracted by the ith convolution kernel, as represented by the principal symptoms in the medical record.
4) Constructing corresponding medical case characteristic vectors through nonlinear transformation:
wherein,is represented by 3 convolution kernels Wi(i is more than or equal to 1 and less than or equal to 3) extracted medical schemeThe key features of (1) indicate the main symptoms, main causes, main pathogenesis and the like.
And 3, predicting the syndromes by using an Attention authorization mechanism.
The method specifically comprises the following steps:
1) and (3) weight distribution: all of the cases in the case subset M1 are assigned different weights,
all cases containing this symptom set are assigned a weight:
wherein, αlShow medical caseWeight assignment of elShow medical caseFor the syndromeThe degree of matching is as follows:
2) in order to show the effect of each medical record during the mining of syndrome differentiation rules, the medical record subset M1 is re-expressed by using weights as follows:
3) in order to make the output number of the network the same as the number of the syndrome tags for classification of the softmax layer, it is dimension-processed using equation (9):
wherein A is a matrix composed of all syndrome vectors,is a bias vector;
4) predicting medical case using softmax functionThe probability over syndrome R is as follows:
wherein K is the number of all syndromes in the medical record set, and P represents the medical recordVector representation over K syndromes.
The invention uses the point sorting CNN network (PCNN) to automatically extract the characteristics of symptoms, causes, pathogenesis and the like of the medical record, constructs the corresponding characteristic code and does not need to manually label the characteristics.
The invention introduces Attention (ATT) mechanism to eliminate noise corpus, distributes different weights to medical cases and improves classification precision.
Compared with the latest research method, the invention carries out the dialectical rule mining on the aspects of symptoms, causes, pathogenesis and the like of medical record data, thereby improving the reliability of the dialectical result; meanwhile, the invention is suitable for the syndrome differentiation rule discovery of most diseases and has high expandability.
TABLE 1 results of syndrome differentiation
| Symptom group | The syndrome/condition |
| Shortness of breath, chest distress, palpitation, cough, hyperhidrosis | Lung-kidney qi deficiency pattern |
| Chest pain, palpitation, night sweat, dizziness, tinnitus, dry mouth, and constipation | Syndrome of yin deficiency of heart and kidney |
| Mental fatigue, sleepiness, amnesia, aversion to cold and cold limbs | Spleen-kidney yang deficiency syndrome |
| Irritability, amnesia, reddish complexion and red ears, dry mouth and dry tongue | Syndrome of hyperactivity of heart-liver fire |
| … | … |
Table 1 shows the syndromes predicted by the method of the present invention, and the TCM physicians in school hospitals of electronic science and technology university have been invited to analyze the syndrome differentiation results in Table 1, and the results show that 92.17% of symptoms under each syndrome can be verified by the TCM internal medicine.
While the foregoing description shows and describes several preferred embodiments of the invention, it is to be understood, as noted above, that the invention is not limited to the forms disclosed herein, but is not to be construed as excluding other embodiments and is capable of use in various other combinations, modifications, and environments and is capable of changes within the scope of the inventive concept as expressed herein, commensurate with the above teachings, or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the invention as defined by the appended claims.
Claims (7)
1. A traditional Chinese medicine syndrome differentiation auxiliary decision method based on a PCNN (pulse coupled neural network) and an attention mechanism is characterized by specifically comprising the following steps:
step 1, pretreatment: performing word segmentation processing on each medical case by using a language technology platform tool;
step 2, medical record coding: performing feature coding on each medical case by using a point-ordered convolutional neural network (PCNN) to construct a corresponding feature vector representation;
and 3, performing syndrome prediction by using an Attention (Attention) mechanism.
2. The PCNN-network-and-attention-mechanism-based traditional Chinese medicine syndrome differentiation aid decision-making method according to claim 1, wherein the step 2 specifically comprises: firstly, each medical case is converted into a feature vector to be represented, and then the corresponding medical case feature vector is constructed by utilizing convolution, sectional maximum pooling and nonlinear conversion.
3. The PCNN-network-and-attention-mechanism-based traditional Chinese medicine syndrome differentiation aid decision-making method of claim 2, wherein each medical scheme is converted into a feature vector for representation, and the specific steps are as follows:
medical case subset M1 extracts: searching the original medical record set for the medical record containing the symptom set S according to the complete word matching method to form a medical record subset M1 ═ M1，…，ml，…，md}；
For medical case M in medical case subset M1l＝{w1，w2，…，wmThe vectorization expression is carried out on (1 ≦ l ≦ d), wherein wiThe symptoms, the causes and the pathogenesis of the medical record contained in the medical record are represented after the medical record is pretreated;
word wi(1. ltoreq. i. ltoreq.m) vector representationThe word code + the position code,
wherein, the word coding: word w with word2veciConverting into a vector form; position coding: according to the word wiWith the first entity w1And tail entity wmConstructing a two-dimensional vector;
medical record mlThe vectorization of (c) is represented as follows:
wherein m is medical table mlThe number of unrepeated features in (A), i.e., symptoms,Etiology and pathogenesis.
4. The PCNN-network-and-attention-mechanism-based traditional Chinese medicine syndrome differentiation aid decision-making method of claim 3, wherein said convolution: performing local feature extraction on the medical plan by using convolution operation, specifically:
selecting a convolution kernel: select 3 convolution kernels Wi(i is more than or equal to 1 and less than or equal to 3), randomly initializing convolution kernel WiDynamically adjusting the value of the convolution kernel according to the classification result, and selecting the optimal symptom characteristic, disease location characteristic and disease characteristic;
sliding length: l — 3, step size: 1, stride is equal to 1;
the sliding window is set to q, as follows:
q＝{wi-l+1，wi-l+2，…，wi} (2)
and (3) convolution operation:
wherein p isiTo use a convolution kernel WiThe extracted local features are compared with the local features,is a bias vector.
5. The PCNN-network-and-attention-mechanism-based traditional Chinese medicine syndrome differentiation aid decision-making method of claim 4, wherein the segmented maximum pooling comprises: combining the local features extracted by the 3 convolution kernels to obtain a vector of fixed length consisting of q +1 parts, performing maximal pooling on each part as follows:
[x]ij＝max(pij) (4)
wherein, [ x ]]ijRepresenting the jth key feature extracted by the ith convolution kernel.
6. The PCNN-network-and-attention-mechanism-based traditional Chinese medicine syndrome differentiation aid decision-making method of claim 5, wherein the non-linear transformation constructs corresponding medical case feature vectors:
wherein,is represented by 3 convolution kernels Wi(i is more than or equal to 1 and less than or equal to 3) extracted medical schemeIs represented by the key features of (1).
7. The PCNN network and Attention mechanism-based traditional Chinese medicine syndrome differentiation aid decision-making method according to any one of claims 1-6, wherein step 3 uses an Attention authorization mechanism for syndrome prediction, specifically:
and (3) weight distribution: all of the cases in the case subset M1 are assigned different weights,
all cases containing this symptom set are assigned a weight:
wherein, αlShow medical caseWeight assignment of elShow medical caseFor the syndromeThe degree of matching is as follows:
in order to show the effect of each medical record during the mining of syndrome differentiation rules, the medical record subset M1 is re-expressed by using weights as follows:
in order to make the output number of the network the same as the number of the syndrome tags for classification of the softmax layer, it is dimension-processed using equation (9):
wherein A is a matrix composed of all syndrome vectors,is a bias vector;
predicting medical case using softmax functionThe probability over syndrome R is as follows:
wherein K isThe number of all syndromes appearing in the medical record set, P represents the medical recordVector representation over K syndromes.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910517325.XA CN110223749B (en) | 2019-06-14 | 2019-06-14 | Traditional Chinese medicine syndrome differentiation auxiliary decision-making method based on PCNN (pulse coupled neural network) and attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910517325.XA CN110223749B (en) | 2019-06-14 | 2019-06-14 | Traditional Chinese medicine syndrome differentiation auxiliary decision-making method based on PCNN (pulse coupled neural network) and attention mechanism |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110223749A true CN110223749A (en) | 2019-09-10 |
| CN110223749B CN110223749B (en) | 2022-12-20 |
Family
ID=67817189
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910517325.XA Active CN110223749B (en) | 2019-06-14 | 2019-06-14 | Traditional Chinese medicine syndrome differentiation auxiliary decision-making method based on PCNN (pulse coupled neural network) and attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110223749B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111477302A (en) * | 2020-03-07 | 2020-07-31 | 深圳问止中医健康科技有限公司 | Traditional Chinese medicine dialectical algorithm of data |
| CN111477295A (en) * | 2020-04-10 | 2020-07-31 | 电子科技大学 | Traditional Chinese medicine formula recommendation method and system based on latent semantic model |
| CN112002408A (en) * | 2020-07-24 | 2020-11-27 | 北京科技大学 | Clinical auxiliary diagnosis device for special diseases of traditional Chinese medicine |
| CN113593698A (en) * | 2021-08-03 | 2021-11-02 | 电子科技大学 | Traditional Chinese medicine syndrome type identification method based on graph attention network |
| CN115631852A (en) * | 2022-11-02 | 2023-01-20 | 北京大学重庆大数据研究院 | Certificate type recommendation method and device, electronic equipment and nonvolatile storage medium |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105912846A (en) * | 2016-04-07 | 2016-08-31 | 南京小网科技有限责任公司 | Intelligent medical aid decision making system on basis of cloud computing technique and medical knowledge base technique |
| CN106295139A (en) * | 2016-07-29 | 2017-01-04 | 姹ゅ钩 | A kind of tongue body autodiagnosis health cloud service system based on degree of depth convolutional neural networks |
| CN108536754A (en) * | 2018-03-14 | 2018-09-14 | 四川大学 | Electronic health record entity relation extraction method based on BLSTM and attention mechanism |
| WO2018184204A1 (en) * | 2017-04-07 | 2018-10-11 | Intel Corporation | Methods and systems for budgeted and simplified training of deep neural networks |
| CN109700433A (en) * | 2018-12-28 | 2019-05-03 | 深圳铁盒子文化科技发展有限公司 | A kind of tongue picture diagnostic system and lingual diagnosis mobile terminal |
| CN109817333A (en) * | 2019-03-21 | 2019-05-28 | 赵宗耀 | A kind of syndrome Differentiation System of Traditional Chinese Medicine and its method of four methods of diagnosis information weight analysis |
-
2019
- 2019-06-14 CN CN201910517325.XA patent/CN110223749B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105912846A (en) * | 2016-04-07 | 2016-08-31 | 南京小网科技有限责任公司 | Intelligent medical aid decision making system on basis of cloud computing technique and medical knowledge base technique |
| CN106295139A (en) * | 2016-07-29 | 2017-01-04 | 姹ゅ钩 | A kind of tongue body autodiagnosis health cloud service system based on degree of depth convolutional neural networks |
| WO2018184204A1 (en) * | 2017-04-07 | 2018-10-11 | Intel Corporation | Methods and systems for budgeted and simplified training of deep neural networks |
| CN108536754A (en) * | 2018-03-14 | 2018-09-14 | 四川大学 | Electronic health record entity relation extraction method based on BLSTM and attention mechanism |
| CN109700433A (en) * | 2018-12-28 | 2019-05-03 | 深圳铁盒子文化科技发展有限公司 | A kind of tongue picture diagnostic system and lingual diagnosis mobile terminal |
| CN109817333A (en) * | 2019-03-21 | 2019-05-28 | 赵宗耀 | A kind of syndrome Differentiation System of Traditional Chinese Medicine and its method of four methods of diagnosis information weight analysis |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111477302A (en) * | 2020-03-07 | 2020-07-31 | 深圳问止中医健康科技有限公司 | Traditional Chinese medicine dialectical algorithm of data |
| CN111477295A (en) * | 2020-04-10 | 2020-07-31 | 电子科技大学 | Traditional Chinese medicine formula recommendation method and system based on latent semantic model |
| CN111477295B (en) * | 2020-04-10 | 2022-06-03 | 电子科技大学 | Traditional Chinese medicine formula recommendation method and system based on latent semantic model |
| CN112002408A (en) * | 2020-07-24 | 2020-11-27 | 北京科技大学 | Clinical auxiliary diagnosis device for special diseases of traditional Chinese medicine |
| CN113593698A (en) * | 2021-08-03 | 2021-11-02 | 电子科技大学 | Traditional Chinese medicine syndrome type identification method based on graph attention network |
| CN113593698B (en) * | 2021-08-03 | 2023-04-18 | 电子科技大学 | Traditional Chinese medicine syndrome type identification method based on graph attention network |
| CN115631852A (en) * | 2022-11-02 | 2023-01-20 | 北京大学重庆大数据研究院 | Certificate type recommendation method and device, electronic equipment and nonvolatile storage medium |
| CN115631852B (en) * | 2022-11-02 | 2024-04-09 | 北京大学重庆大数据研究院 | Certificate type recommendation method and device, electronic equipment and nonvolatile storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110223749B (en) | 2022-12-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110223749A (en) | Chinese medical discrimination aid decision-making method based on PCNN network and attention mechanism | |
| Cai et al. | Hierarchical domain adaptation projective dictionary pair learning model for EEG classification in IoMT systems | |
| CN109086805B (en) | Clustering method based on deep neural network and pairwise constraints | |
| Nanehkaran et al. | Analysis and comparison of machine learning classifiers and deep neural networks techniques for recognition of Farsi handwritten digits | |
| Kumar et al. | Covid-19 detection from chest x-rays using trained output based transfer learning approach | |
| CN114724623A (en) | Method for predicting drug-target affinity of protein multi-source feature fusion | |
| Li et al. | Dating ancient paintings of Mogao Grottoes using deeply learnt visual codes | |
| Zhou et al. | Shoulder implant x-ray manufacturer classification: Exploring with vision transformer | |
| Yu et al. | Multilevel dual-direction modifying variational autoencoders for hyperspectral feature extraction | |
| Arora et al. | FKPIndexNet: An efficient learning framework for finger-knuckle-print database indexing to boost identification | |
| Afrianty et al. | Back-Propagation neural network for gender determination in forensic anthropology | |
| Wang et al. | COVID-19 detection by wavelet entropy and self-adaptive pso | |
| Hao et al. | Radiographs and texts fusion learning based deep networks for skeletal bone age assessment | |
| Li et al. | MTS-LOF: medical time-series representation learning via occlusion-invariant features | |
| CN113537412A (en) | Lung disease auscultation system based on convolutional neural network, signal processing method and equipment | |
| Lan et al. | A Novel Skin Cancer Assisted Diagnosis Method based on Capsule Networks with CBAM | |
| Madhurshalini et al. | Automatic identification of skin lesions using deep learning techniques | |
| Huang et al. | Tongue size and shape classification fusing segmentation features for traditional Chinese medicine diagnosis | |
| Li et al. | A perception-driven framework for predicting missing odor perceptual ratings and an exploration of odor perceptual space | |
| Aktas et al. | Graph classification via heat diffusion on simplicial complexes | |
| Nayak et al. | Learning a sparse dictionary of video structure for activity modeling | |
| Debnath et al. | A step towards automated functional assessment of activities of daily living | |
| Alsadoon et al. | Deep learning models for human age prediction to prevent, treat and extend life expectancy: DCPV taxonomy | |
| Xu et al. | Multi-spectral palmprint recognition with deep multi-view representation learning | |
| CN117219285B (en) | Application of data distribution intelligent analysis conversion in medical data mining |
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 |