CN109087702B - Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis - Google Patents
Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis Download PDFInfo
- Publication number
- CN109087702B CN109087702B CN201810878380.7A CN201810878380A CN109087702B CN 109087702 B CN109087702 B CN 109087702B CN 201810878380 A CN201810878380 A CN 201810878380A CN 109087702 B CN109087702 B CN 109087702B
- Authority
- CN
- China
- Prior art keywords
- information
- tester
- health state
- patient
- representation
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 230000036541 health Effects 0.000 title claims abstract description 39
- 238000004458 analytical method Methods 0.000 title claims abstract description 28
- 238000003745 diagnosis Methods 0.000 title claims abstract description 17
- 239000003814 drug Substances 0.000 title claims abstract description 17
- 238000007500 overflow downdraw method Methods 0.000 title claims abstract description 9
- 238000000034 method Methods 0.000 claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 27
- 208000011580 syndromic disease Diseases 0.000 claims abstract description 19
- 230000003862 health status Effects 0.000 claims abstract description 11
- 238000007689 inspection Methods 0.000 claims abstract description 10
- 230000004927 fusion Effects 0.000 claims abstract description 8
- 238000011156 evaluation Methods 0.000 claims description 21
- 238000004364 calculation method Methods 0.000 claims description 15
- 238000012549 training Methods 0.000 claims description 15
- 238000005457 optimization Methods 0.000 claims description 13
- 238000012360 testing method Methods 0.000 claims description 9
- 238000002372 labelling Methods 0.000 claims description 8
- 238000012512 characterization method Methods 0.000 claims description 7
- 238000012795 verification Methods 0.000 claims description 7
- 239000003550 marker Substances 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 238000011478 gradient descent method Methods 0.000 claims description 3
- 238000012163 sequencing technique Methods 0.000 claims description 2
- 238000002555 auscultation Methods 0.000 claims 1
- 230000036285 pathological change Effects 0.000 abstract description 3
- 231100000915 pathological change Toxicity 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 description 5
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000002405 diagnostic procedure Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 201000010099 disease Diseases 0.000 description 1
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 230000035790 physiological processes and functions Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
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
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
Landscapes
- Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Public Health (AREA)
- Theoretical Computer Science (AREA)
- Medical Informatics (AREA)
- Biomedical Technology (AREA)
- Physics & Mathematics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Evolutionary Biology (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Databases & Information Systems (AREA)
- Evolutionary Computation (AREA)
- Pathology (AREA)
- Life Sciences & Earth Sciences (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Primary Health Care (AREA)
- Medical Treatment And Welfare Office Work (AREA)
- Measuring And Recording Apparatus For Diagnosis (AREA)
Abstract
The four-diagnosis representation information fusion method for the analysis of the health state of the traditional Chinese medicine collects the information of inspection, smell, inquiry, cutting and the like of a patient in clinical treatment, is used for generating multi-source information representation of the patient and marks the syndrome type category to which the patient belongs; respectively analyzing the health state of the tester by using the characteristic representation and the category information of each information source to obtain the auxiliary decision information of the tester by using a plurality of information sources; constructing an information fusion model to maximize decision consistency for returning an optimized health state analysis result; the performance of the proposed algorithm is evaluated by comparing the actual health status of the tester with the corresponding prediction results. The current health state and pathological change essence of the testee can be detected, so that the testee can know the physical condition of the testee and provide reference for making an intervention scheme. The health state analysis result with high precision can be provided, and a basis is provided for health care. The method can fuse the four diagnosis representation information of the clinical patients, and obtain more accurate and reliable state analysis results.
Description
Technical Field
The invention relates to multi-label learning, in particular to a four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis.
Background
The state is a logical starting point of the traditional Chinese medicine health cognition theory, and the health state is a comprehensive state of a morphological structure, a physiological function, a psychological state and the ability of adapting to the external environment in unit time of a human body and reflects a healthy state and situation. The health state analysis is based on the theory of traditional Chinese medicine, and expresses the collected information of inspection, smell, inquiry, cutting and the like in a data form, emphasizes the objective evaluation of the health state and the nature of pathological changes of human bodies, and gives a general judgment to diseases and symptoms (Li Shu Dong. the science of traditional Chinese medicine [ M ]. Beijing: the publishing company of traditional Chinese medicine, 2016).
The multi-label learning technology is used for processing objects with ambiguity in the real world, and has attracted attention and application in the fields of automatic image labeling, bioinformatics, information retrieval, recommendation systems and the like. Specifically, the syndrome type distribution of the patients in clinical visits is often accompanied by multiple states. Therefore, based on the artificial intelligence technology to solve the problem of analyzing the health state of the traditional Chinese medicine, the multi-mark learning technology is introduced into the analysis of the health state of the traditional Chinese medicine.
According to the principle of 'four diagnostic methods' in traditional Chinese medicine, state analysis is established on the basis of four diagnostic methods information. Considering that the contribution degrees of different information sources to the prediction have differences and the different information sources are mutually associated, the overall information of the clinical patient is collected by a four-diagnosis method, and an information fusion model is further constructed to analyze the health state of the patient.
The big health data of traditional Chinese medicine presents the characteristics of multimode, multi-marking and the like, so that the traditional data analysis theory, method and technology face the serious challenges of effectiveness, accuracy, computability and the like. Therefore, the research of the four-diagnosis representation information fusion method for the health state analysis of the traditional Chinese medicine is beneficial to constructing a more accurate and reliable identification model and exerting the advantages of the artificial intelligence technology to promote the joint development and prosperity of the interdisciplinary.
Disclosure of Invention
The invention aims to provide a four-diagnosis representation information fusion method for health state analysis of traditional Chinese medicine, aiming at the multi-state and diversity of diagnosis information of patients in clinical treatment.
The invention comprises the following steps:
1) collecting information of inspection, smell, inquiry, cutting and the like of a patient in clinical treatment, generating multi-source information representation of the patient, and labeling the affiliated syndrome type category;
2) analyzing the health state of the tester by using the characteristic representation and the category information of each information source to obtain auxiliary decision information of the tester by using a plurality of information sources;
3) constructing an information fusion model to maximize decision consistency for returning an optimized health state analysis result;
4) the performance of the proposed algorithm is evaluated by comparing the actual health status of the tester with the corresponding prediction results.
In step 1), the specific method for collecting information of inspection, smell, inquiry, cutting and the like of a clinical patient for generating multi-source information representation of the patient and labeling the syndrome type category to which the patient belongs may be:
(1) extracting the characteristic information of the four diagnoses of the clinical patients from the electronic medical records to form an information source A; obtaining a tongue picture of a patient by using a inspection instrument, realizing tongue picture segmentation based on a U-Net network model, then obtaining a plurality of characteristic representations of the tongue picture by adopting HSV, LAB and RGB description operators, and respectively forming an information source B, an information source C and an information source D;
(2) the health status of the patient in clinical visit is marked by the doctor as l1,l2,...,lqJ is more than or equal to 1 and less than or equal to q, wherein ljThe jth syndrome type of the patient in clinical visit, and q is the total number of the category markers;
(3) and (3) verifying the algorithm by adopting a ten-fold cross verification method: dividing the processed standardized data into training data and testing data according to a ratio of 9: 1.
In step 2), the specific method for analyzing the health status of the tester by using the characteristic representation and the category information of each information source to obtain the auxiliary decision information of the tester from the plurality of information sources may be:
(1) the health state of the tester is predicted by adopting the SVM, and the calculation formula is as follows:
wherein,
indicating the prediction of the ith tester on the data source a for the jth syndrome,
characteristic characterization information representing the ith tester on the data source A;
(2) the combined characteristic features and corresponding prediction information search Top-k neighbors of the tester in the training set, the neighbor selection is based on the similarity relation between the tester and the training sample, and the calculation formula is as follows:
wherein,
the similarity of the tester and the training sample on the feature space is obtained by calculation by a cosine similarity method;
the similarity of the test person and the training sample in the labeling space is obtained by a Jacard similarity method; beta is a threshold value with the value range of [0, 1%]；
(3) Using similarity relationships simAModeling the correlation between the syndromes to reconstruct the marker space of the tester:
wherein,
indicating the status analysis result of the ith tester on the jth certificate type on the data source A，YzjThe actual value of the ith neighbor of the ith tester on the jth certificate type is shown;
(4) and (4) repeating the steps (1) to (3) to respectively obtain state analysis results based on the information sources B to D.
In step 3), the information fusion model is constructed to maximize decision consistency, and a specific method for returning an optimized health state analysis result may be:
(1) obtaining a final result of a tester by utilizing a plurality of state results predicted by the four-diagnosis representation information of a clinical patient, and constructing the following optimization objective function to solve:
wherein,
showing the optimization result of the ith tester on the jth syndrome type, wherein the optimization result is obtained by fusing multi-source decision information, and W is { W ═ W1,w2,...,wMIs the weight distribution of M information sources, where M4,
in addition, cmRepresents a set of (i, j) s, and (i, j) satisfies
Alpha is a threshold value and has a value range of [0,1 ]]；
(2) Initializing weights, order
Setting:
(3) fixing W, and solving Y by gradient descent method*The calculation formula is as follows:
(4) fixed Y*And solving W by using a Lagrange multiplier method, wherein the calculation formula is as follows:
(5) repeating the steps (3) and (4) until the optimization target is converged, and returning the optimization result Y of the health state of the tester*。
In step 4), the specific method for evaluating the performance of the proposed algorithm by comparing the actual health status of the tester with the corresponding prediction result may be:
the method is used for predicting the category labels of testers in the test data, and the performance of the algorithm is evaluated by adopting the following five indexes:
(1) hamming loss: the method is used for inspecting the misclassification condition of the sample on a single mark, and the smaller the evaluation index is, the better the evaluation index is;
(2) 1-error Rate: the method is used for inspecting the condition that in the category mark sorting sequence of the sample, the mark at the forefront of the sequence does not belong to a related mark set, and the smaller the evaluation index is, the better the evaluation index is;
(3) coverage rate: the method is used for inspecting the search depth condition required by covering all related marks in the category mark sequencing sequence of the sample, and the smaller the evaluation index is, the better the evaluation index is;
(4) loss of ordering: the method is used for inspecting the situation that the sorting error occurs in the sorting sequence of the class marks of the samples, and the smaller the evaluation index is, the better the evaluation index is;
(5) average precision: the evaluation index is larger and better when the label arranged before the related label in the sorting sequence of the class labels of the samples is still the related label.
Compared with the prior art, the method can detect the current health state and pathological change essence of the tester, so that the tester can know the physical condition of the tester and provide reference for making an intervention scheme.
The invention can provide a high-precision health state analysis result and provide a basis for health care.
The invention can fuse the four-diagnosis characterization information of the clinical patient, thereby obtaining more accurate and reliable state analysis results.
Drawings
FIG. 1 is a schematic view of tongue segmentation.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The embodiment comprises the following steps:
1) collecting information of inspection, smell, inquiry, cutting and the like of 729 clinical patients for generating multi-source information representation of the patients, marking the affiliated syndrome type, and counting 339 syndrome type types;
(1) extracting the characteristic information of the four diagnoses of the clinical patients from the electronic medical records to form an information source A; the tongue picture of the patient is obtained by using a inspection instrument, and the tongue picture segmentation is realized based on a U-Net network model, as shown in figure 1. Then obtaining a plurality of characteristic representations of the tongue picture by adopting HSV, LAB and RGB description operators to respectively form an information source B, an information source C and an information source D;
(2) the health status of the patient in clinical visit is marked by the doctor as l1,l2,...,lqJ is more than or equal to 1 and less than or equal to q. Wherein ljThe jth syndrome type of the patient in clinical visit, and q is the total number of the category markers;
(3) and (3) verifying the algorithm by adopting a ten-fold cross verification method: dividing the processed standardized data into training data and testing data according to a ratio of 9: 1.
2) Respectively analyzing the health state of the tester by using the characteristic representation and the category information of each information source to obtain the auxiliary decision information of the tester from a plurality of information sources;
(1) the health state of the tester is predicted by adopting the SVM, and the calculation formula is as follows:
wherein,
indicating the prediction of the ith tester on the data source a for the jth syndrome,
characteristic characterization information representing the ith tester on the data source A;
(2) the combined feature characterization and corresponding prediction information are used to search the training set for Top-k neighbors of the tester. The neighbor selection is based on the similarity relation between the tester and the training sample, and the calculation formula is as follows:
wherein,
the similarity of the tester and the training sample on the feature space is obtained by calculation by a cosine similarity method;
the similarity of the test person and the training sample in the labeling space is obtained by a Jacard similarity method; beta is a threshold value with the value range of [0, 1%]；
(3) Using similarity relationships simAModeling the correlation between the syndromes to reconstruct the marker space of the tester:
wherein,
indicating the status analysis result of the ith tester on the jth certificate type, YzjThe actual value of the ith neighbor of the ith tester on the jth certificate type is shown;
(4) the predicted results on Source A were compared with BSVM (M.R. Boutell, J.Luo, X.Shen, C.M.Brown, Learning Multi-label scene classification, Pattern Recognition,2004,37(9): 1757-. The algorithm 1 corresponds to the verification result of the algorithm provided by the invention; the algorithm 2 corresponds to the verification result of LIFT; algorithm 3 corresponds to the verification result of the BSVM. As can be seen from table 1, the present invention is better than other algorithms in most evaluation indexes by considering the marker correlation performance.
TABLE 1
| Algorithm | Loss of hamming | 1-error Rate | Coverage rate | Loss of ordering | Average accuracy |
| 1 | 0.0122±0.0008 | 0.1701±0.0330 | 0.3251±0.0435 | 0.0617±0.0081 | 0.6405±0.0265 |
| 2 | 0.0133±0.0009 | 0.1990±0.0377 | 0.4168±0.0408 | 0.0882±0.0085 | 0.5910±0.0260 |
| 3 | 0.0123±0.0009 | 0.1358±0.0263 | 0.3453±0.0436 | 0.0665±0.0076 | 0.6355±0.0177 |
(5) And (4) repeating the steps (1) to (3) to respectively obtain health state analysis results based on the information sources B to D.
3) Constructing an information fusion model to maximize decision consistency for returning an optimized health state analysis result;
(1) obtaining a final result of a tester by utilizing a plurality of state identification results predicted by the four-diagnosis representation information of a clinical patient, and constructing the following optimization objective function to solve:
wherein,
showing the optimized result of the ith tester on the jth certificate typeThe method is obtained by fusing multi-source decision information. W ═ W1,w2,...,wMIs the weight distribution of M information sources (where M is 4),
in addition, cmRepresents a set of (i, j) s, and (i, j) satisfies
Alpha is a threshold value and has a value range of [0,1 ]]；
(2) The weights are initialized. Order to
Setting:
(3) fixing W, and solving Y by gradient descent method*The calculation formula is as follows:
(4) fixed Y*And solving W by using a Lagrange multiplier method, wherein the calculation formula is as follows:
(5) repeating the steps (3) to (4) until the optimization target is converged, and returning the optimization result Y of the health state of the tester*。
4) Analyzing the health state of the testee in the test data by using the method;
the proposed algorithm is compared to the predicted results for each information source as shown in table 2. As can be seen from Table 2, the algorithm can obtain the optimal result on most evaluation indexes by fusing the information sources A-D.
TABLE 2
| Loss of hamming | 1-error Rate | Coverage rate | Loss of ordering | Average accuracy | |
| Information source A | 0.0122±0.0010 | 0.1701±0.0330 | 0.3251±0.0435 | 0.0617±0.0081 | 0.6405±0.0265 |
| Information source B | 0.0180±0.0014 | 0.5624±0.0657 | 0.4090±0.0443 | 0.0978±0.0071 | 0.3536±0.0203 |
| Information source C | 0.0181±0.0015 | 0.6090±0.0470 | 0.4122±0.0412 | 0.0982±0.0058 | 0.3447±0.0160 |
| Information source D | 0.0181±0.0015 | 0.5968±0.0595 | 0.4075±0.0413 | 0.0968±0.0083 | 0.3516±0.0241 |
| Information fusion | 0.0118±0.0011 | 0.1604±0.0272 | 0.3328±0.0634 | 0.0637±0.0108 | 0.6473±0.0191 |
The proposed algorithm was compared to other fusion algorithms as shown in table 3. The algorithm 1 corresponds to the verification result of the algorithm provided by the invention; the average result of the algorithm 2 is predicted based on all information sources; and (3) correspondingly based on the voting results predicted by all the information sources, an algorithm 4 connects all the information sources in series, and then the SVM is used for classification. As can be seen from table 3, the algorithm proposed by the present invention has the best results.
TABLE 3
| Algorithm | Loss of hamming | 1-error Rate | Coverage rate | Loss of ordering | Average accuracy |
| 1 | 0.0118±0.0011 | 0.1604±0.0272 | 0.3328±0.0634 | 0.0637±0.0108 | 0.6473±0.0191 |
| 2 | 0.0161±0.0019 | 0.2386±0.0350 | 0.3716±0.0578 | 0.0778±0.0102 | 0.5222±0.0190 |
| 3 | 0.0177±0.0020 | 0.4073±0.0621 | 0.3715±0.0578 | 0.0803±0.0101 | 0.4422±0.0242 |
| 4 | 0.0123±0.0007 | 0.1427±0.0409 | 0.3473±0.0427 | 0.0673±0.0083 | 0.6343±0.0158 |
The method comprises the steps of preprocessing information captured by a four-diagnosis acquisition instrument, analyzing the prediction result of each information source to judge the health state of a tester, and fusing the prediction results of a plurality of characteristic characterization information to maximize the consistency of state identification, so that accurate and reliable reference is provided for the tester to formulate an intervention scheme.
Claims (3)
1. The four-diagnosis representation information fusion method for the traditional Chinese medicine health state analysis is characterized by comprising the following steps of:
1) collecting the inspection, auscultation, inquiry and cutting information of a patient in clinical treatment, generating multi-source information representation of the patient, and labeling the affiliated syndrome type;
2) the method comprises the following steps of respectively analyzing the health state of a tester by utilizing the characteristic representation and the category information of each information source to obtain the auxiliary decision information of a plurality of information sources to the tester, and specifically comprises the following steps:
(1) the health state of the tester is predicted by adopting the SVM, and the calculation formula is as follows:
wherein,
indicating the prediction of the ith tester on the information source a for the jth syndrome,
the characteristic representation information of the ith tester on the information source A is represented;
(2) the combined characteristic features and corresponding prediction information search Top-k neighbors of the tester in the training set, the neighbor selection is based on the similarity relation between the tester and the training sample, and the calculation formula is as follows:
wherein,
the similarity of the tester and the training sample on the feature space is obtained by calculation by a cosine similarity method;
the similarity of the test person and the training sample in the labeling space is obtained by a Jacard similarity method; beta is a threshold value with the value range of [0, 1%]；
(3) Using similarity relationships simAModeling the correlation between the syndromes to reconstruct the marker space of the tester:
wherein,
indicating the status analysis result of the ith tester on the information source A with respect to the jth certificate type, YzjThe actual value of the ith neighbor of the ith tester on the jth certificate type is shown;
(4) repeating the steps (1) to (3) to respectively obtain state analysis results based on the information sources B to D;
3) the method comprises the following steps of constructing an information fusion model to maximize decision consistency, and returning an optimized health state analysis result, wherein the specific method comprises the following steps:
(1) obtaining a final result of a tester by utilizing a plurality of state results predicted by the four-diagnosis representation information of a clinical patient, and constructing the following optimization objective function to solve:
wherein,
expressing the optimization result of the ith tester on the jth syndrome type, wherein the optimization result is obtained by fusing multi-source decision information;
representing the state analysis result of the ith tester on the information source m about the jth certificate type; w ═ W1,w2,...,wMIs the weight distribution of M information sources, where M4,
in addition, cmRepresents a set of (i, j) s, and (i, j) satisfies
Alpha is a threshold value and has a value range of [0,1 ]]；
(2) Initializing weights, order
Setting:
(3) fixing W, and solving Y by gradient descent method*The calculation formula is as follows:
(4) fixed Y*And solving W by using a Lagrange multiplier method, wherein the calculation formula is as follows:
(5) repeating the steps (3) and (4) until the optimization target is converged, and returning the optimization result Y of the health state of the tester*；
4) The performance of the proposed algorithm is evaluated by comparing the actual health status of the tester with the corresponding prediction results.
2. The four-diagnosis characterization information fusion method for health status analysis of traditional Chinese medicine according to claim 1, wherein in step 1), the specific method for collecting the inspection, smell, inquiry and cutting information of the clinical patients for generating the multi-source information representation of the patients and labeling the syndrome type categories to which the patients belong is as follows:
(1) extracting the characteristic information of the four diagnoses of the clinical patients from the electronic medical records to form an information source A; obtaining a tongue picture of a patient by using a inspection instrument, realizing tongue picture segmentation based on a U-Net network model, then obtaining a plurality of characteristic representations of the tongue picture by adopting HSV, LAB and RGB description operators, and respectively forming an information source B, an information source C and an information source D;
(2) the health status of the patient in clinical visit is marked by the doctor as l1,l2,...,lqJ is more than or equal to 1 and less than or equal to q, wherein ljThe jth syndrome type of the patient in clinical visit, and q is the total number of the category markers;
(3) and (3) verifying the algorithm by adopting a ten-fold cross verification method: dividing the processed standardized data into training data and testing data according to a ratio of 9: 1.
3. The four-diagnostic-diagnosis characterization information fusion method for health status analysis of traditional Chinese medicine according to claim 1, wherein in step 4), the specific method for evaluating the performance of the proposed algorithm by comparing the actual health status of the testee with the corresponding prediction result is:
the category labels of testers in the test data are predicted by using the algorithm, and the performance of the algorithm is evaluated by adopting the following five indexes:
(1) hamming loss: the method is used for inspecting the misclassification condition of the sample on a single mark, and the smaller the evaluation index is, the better the evaluation index is;
(2) 1-error Rate: the method is used for inspecting the condition that in the category mark sorting sequence of the sample, the mark at the forefront of the sequence does not belong to a related mark set, and the smaller the evaluation index is, the better the evaluation index is;
(3) coverage rate: the method is used for inspecting the search depth condition required by covering all related marks in the category mark sequencing sequence of the sample, and the smaller the evaluation index is, the better the evaluation index is;
(4) loss of ordering: the method is used for inspecting the situation that the sorting error occurs in the sorting sequence of the class marks of the samples, and the smaller the evaluation index is, the better the evaluation index is;
(5) average precision: the evaluation index is larger and better when the label arranged before the related label in the sorting sequence of the class labels of the samples is still the related label.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810878380.7A CN109087702B (en) | 2018-08-03 | 2018-08-03 | Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810878380.7A CN109087702B (en) | 2018-08-03 | 2018-08-03 | Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109087702A CN109087702A (en) | 2018-12-25 |
| CN109087702B true CN109087702B (en) | 2021-07-16 |
Family
ID=64833581
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810878380.7A Active CN109087702B (en) | 2018-08-03 | 2018-08-03 | Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109087702B (en) |
Families Citing this family (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111260619A (en) * | 2020-01-14 | 2020-06-09 | 浙江中医药大学 | Tongue body automatic segmentation method based on U-net model |
| CN112200091A (en) * | 2020-10-13 | 2021-01-08 | 深圳市悦动天下科技有限公司 | Tongue region identification method and device and computer storage medium |
| CN112530584A (en) * | 2020-12-15 | 2021-03-19 | 贵州小宝健康科技有限公司 | Medical diagnosis assisting method and system |
| CN113409938A (en) * | 2021-06-30 | 2021-09-17 | 海南医学院 | Modeling method and system of traditional Chinese medicine syndrome type prediction model of systemic lupus erythematosus |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101129261A (en) * | 2007-02-09 | 2008-02-27 | 北京中医药大学 | Device and method for acquiring and recognizing pulsation information and tongue inspection information |
| CN101647696A (en) * | 2008-08-13 | 2010-02-17 | 上海经路通中医药科技发展有限公司 | Intelligent system for health diagnosis and treatment |
| CN104766068A (en) * | 2015-04-20 | 2015-07-08 | 江西中医药大学 | Random walk tongue image extraction method based on multi-rule fusion |
| CN105528529A (en) * | 2016-02-20 | 2016-04-27 | 成都中医药大学 | Data processing method of traditional Chinese medicine clinical skill evaluation system based on big data analysis |
| CN106874655A (en) * | 2017-01-16 | 2017-06-20 | 西北工业大学 | Traditional Chinese medical science disease type classification Forecasting Methodology based on Multi-label learning and Bayesian network |
| CN108198621A (en) * | 2018-01-18 | 2018-06-22 | 中山大学 | A kind of database data synthesis dicision of diagnosis and treatment method based on neural network |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11235062B2 (en) * | 2009-03-06 | 2022-02-01 | Metaqor Llc | Dynamic bio-nanoparticle elements |
-
2018
- 2018-08-03 CN CN201810878380.7A patent/CN109087702B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101129261A (en) * | 2007-02-09 | 2008-02-27 | 北京中医药大学 | Device and method for acquiring and recognizing pulsation information and tongue inspection information |
| CN101647696A (en) * | 2008-08-13 | 2010-02-17 | 上海经路通中医药科技发展有限公司 | Intelligent system for health diagnosis and treatment |
| CN104766068A (en) * | 2015-04-20 | 2015-07-08 | 江西中医药大学 | Random walk tongue image extraction method based on multi-rule fusion |
| CN105528529A (en) * | 2016-02-20 | 2016-04-27 | 成都中医药大学 | Data processing method of traditional Chinese medicine clinical skill evaluation system based on big data analysis |
| CN106874655A (en) * | 2017-01-16 | 2017-06-20 | 西北工业大学 | Traditional Chinese medical science disease type classification Forecasting Methodology based on Multi-label learning and Bayesian network |
| CN108198621A (en) * | 2018-01-18 | 2018-06-22 | 中山大学 | A kind of database data synthesis dicision of diagnosis and treatment method based on neural network |
Non-Patent Citations (2)
| Title |
|---|
| Computational drug repositioning using collaborative filtering via multi-source fusion;Jia Zhang.etc;《Expert Systems With Applications》;20171030;第84卷;全文 * |
| 面向认知的多源数据学习理论和算法研究进展;杨柳，等;《软件学报》;20171130;第28卷(第11期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109087702A (en) | 2018-12-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109087702B (en) | Four-diagnosis representation information fusion method for traditional Chinese medicine health state analysis | |
| CN110956626B (en) | Image-based prognosis evaluation method and device | |
| CN110352389A (en) | Information processing unit and information processing method | |
| KR102639558B1 (en) | Growth analysis prediction apparatus using bone maturity distribution by interest area and method thereof | |
| Dong et al. | Cervical cell classification based on the CART feature selection algorithm | |
| CN116705300A (en) | Medical decision assistance method, system and storage medium based on sign data analysis | |
| Weng et al. | Multi-label symptom analysis and modeling of TCM diagnosis of hypertension | |
| Chiwariro et al. | Comparative analysis of deep learning convolutional neural networks based on transfer learning for pneumonia detection | |
| Sujatha et al. | Densenet201: a customized DNN model for multi-class classification and detection of tumors based on brain MRI Images | |
| Saputra et al. | Implementation of machine learning and deep learning models based on structural MRI for identification autism spectrum disorder | |
| Ganeshkumar et al. | Two-stage deep learning model for automate detection and classification of lung diseases | |
| CN113706518A (en) | Femoral head necrosis stage analysis method and system based on artificial intelligence | |
| James et al. | Classification of x-ray covid-19 image using convolutional neural network | |
| Sun et al. | Liver tumor segmentation and subsequent risk prediction based on Deeplabv3+ | |
| CN117195027A (en) | Cluster weighted clustering integration method based on member selection | |
| Mohapatra et al. | Automated invasive cervical cancer disease detection at early stage through deep learning | |
| Kong et al. | Tooth-marked tongue recognition based on mask scoring R-CNN | |
| Iannello et al. | A slightly supervised approach for positive/negative classification of fluorescence intensity in hep-2 images | |
| CN114004821A (en) | Intestinal ganglion cell auxiliary identification method based on cascade rcnn | |
| CN112842273A (en) | Method for distinguishing first schizophrenia from depression | |
| Was et al. | Verification of results in the acquiring knowledge process based on IBL methodology | |
| Liu | Convolutional Siamese network-based few-shot learning for monkeypox detection under data scarcity | |
| CN111920430A (en) | Automatic bone age assessment method for weak supervised deep learning | |
| Sulaiman et al. | Classification of healthy and white root disease infected rubber trees based on relative permittivity and capacitance input properties using LM and SCG artificial neural network | |
| Delgarmi et al. | Automatic Landmark Detection of Human Back Surface from Depth Images via Deep Learning |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |