Rheumatoid joint traditional Chinese medicine syndrome prediction model construction method and prediction method

Through deep learning and natural language processing technology, a personalized traditional Chinese medicine syndrome prediction model is constructed, which solves the problem of poor training effect of the TCM syndrome auxiliary diagnosis model for rheumatoid arthritis, and improves diagnostic accuracy and system reliability.

CN119008024APending Publication Date: 2024-11-22GUANGANMEN HOSPITAL CHINA ACAD OF CHINESE MEDICAL SCI
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202411086602.3
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-08
Publication Date
2024-11-22

AI Technical Summary

Technical Problem

There is a lack of customized TCM syndrome auxiliary diagnostic models for patients with rheumatoid arthritis in the prior art, resulting in poor training of TCM syndrome prediction models and affecting prediction accuracy.

Method used

Deep learning algorithm combined with natural language processing is used to obtain and screen labeled TCM syndrome data, conduct quantitative evaluation and sample equalization processing, build a personalized TCM syndrome prediction model, and dynamically update training parameters to improve diagnostic accuracy.

Benefits of technology

It realizes an efficient Chinese medicine syndrome prediction method, improves diagnostic accuracy and systematic robustness, ensures that low-quality data does not affect the diagnostic results, and improves the accuracy and reliability of auxiliary diagnosis of Chinese medicine syndromes.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119008024A_ABST
    Figure CN119008024A_ABST
Patent Text Reader

Abstract

The embodiment of the invention discloses a rheumatoid joint traditional Chinese medicine syndrome prediction model construction method and a prediction method. In a specific embodiment, the prediction method comprises the following steps: acquiring a symptom description text of a rheumatoid arthritis patient; extracting features in the symptom description text based on a natural language processing algorithm to obtain feature data; and inputting the feature data into the trained traditional Chinese medicine syndrome prediction model to obtain a traditional Chinese medicine syndrome prediction result. According to the embodiment, the efficient traditional Chinese medicine syndrome prediction method is developed to ensure that the features are accurately extracted from the data source. According to the method, a personalized model training method is adopted, and the training parameters of the traditional Chinese medicine syndrome prediction model are dynamically updated according to the quality score of each medical record, so that the weight of the model for high-quality medical record learning is improved, and the accuracy of diagnosis prediction is improved. Meanwhile, the method achieves the automatic quality evaluation of the sample data through the quantification of each item corresponding to each traditional Chinese medicine syndrome, thereby improving the robustness and reliability of the whole system.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of smart medical technology, and more specifically, to a method for constructing a TCM syndrome prediction model for rheumatoid arthritis and a prediction method. Background Art

[0002] Rheumatoid arthritis (RA) is a chronic inflammatory disease that not only affects joints and causes pain, swelling and stiffness, but may also involve multiple systems throughout the body, leading to a wide range of health problems. In order to improve the quality of TCM treatment of rheumatoid arthritis and standardize the diagnosis and treatment process, in the existing technology, the Chinese Association of Traditional Chinese Medicine, under the leadership of Professor Jiang Quan, has formulated the "Guidelines for the Diagnosis and Treatment of Rheumatoid Arthritis". The guidelines further refine the TCM syndrome classification and diagnostic criteria of RA, and clearly divide the eight main syndrome categories of rheumatism, damp-heat, cold-dampness, phlegm and blood stasis, blood stasis, qi and blood deficiency, liver and kidney deficiency, and qi and yin deficiency, and set clear diagnostic criteria for each syndrome. However, the guidelines currently lack a TCM syndrome auxiliary diagnosis model specifically for RA patients. Existing auxiliary diagnosis systems have problems with accuracy and personalization when dealing with TCM syndromes. These systems are mostly based on electronic medical records and trained using neural network models such as long short-term memory networks (LSTMs) and recurrent neural networks (RNNs). Although these models are capable of processing complex data, they still lack customized training data sets for their specific application in TCM syndromes of RA. In addition, the quality of data in the training data sets is also uneven, which leads to poor training effects of TCM syndrome-assisted diagnosis models (such as TCM syndrome prediction models), which in turn affects the final prediction accuracy of the model. Summary of the invention

[0003] The object of the present invention is to provide a method for constructing a TCM syndrome prediction model for rheumatoid joint and a prediction method having a highly customized data set and measuring the quality of each data in the data set, so as to solve at least one of the problems existing in the prior art.

[0004] In order to achieve the above object, the present invention adopts the following technical solutions:

[0005] The first aspect of the present invention provides a method for constructing a TCM syndrome prediction model for rheumatoid joint, comprising:

[0006] Acquire a plurality of original samples that are marked as corresponding to TCM syndromes and include original entry data corresponding to the TCM syndromes to form an original sample set;

[0007] According to the original sample set, the original items of each TCM syndrome are screened by various screening algorithms to obtain the screening items corresponding to each TCM syndrome;

[0008] Based on the original sample set, multiple item screening samples are further obtained, which are marked as corresponding to TCM syndromes and include screening items corresponding to the TCM syndromes, to form an item screening sample set;

[0009] Each original item corresponding to each TCM syndrome was quantified to evaluate the quality score of each item in screening samples;

[0010] Divide the item screening sample set into a training set and a test set;

[0011] A TCM syndrome prediction model is constructed based on a deep learning algorithm, the TCM syndrome prediction model is trained using the training set, and the training parameters of the TCM syndrome prediction model are updated according to the quality scores of the samples screened for each item to obtain a trained TCM syndrome prediction model.

[0012] Furthermore, the method of obtaining a plurality of original samples that are marked as corresponding to TCM syndromes and include the original entry data corresponding to the TCM syndromes, and forming an original sample set includes

[0013] Acquire symptom description texts of multiple rheumatoid arthritis patients, wherein the symptom description texts are annotated with corresponding TCM syndromes; subject the symptom description text data of each TCM syndrome to word segmentation, denoising and keyword extraction in turn to extract multiple original entry data corresponding to the TCM syndrome; obtain multiple original samples annotated as corresponding TCM syndromes and including the original entry data corresponding to the TCM syndrome to form an original sample set.

[0014] Furthermore, the screening algorithm is selected from one or more of item distribution examination method, Cronbach's coefficient method, factor analysis method, discrete trend method and expert evaluation method.

[0015] Furthermore, the quantification of each original item corresponding to each TCM syndrome includes

[0016] Calculate the subjective weight and objective weight of each original item in the corresponding TCM syndrome respectively to calculate the comprehensive weight of each original item in the TCM syndrome;

[0017] The quality score of each original sample was calculated based on the comprehensive weight of each original item in each TCM syndrome.

[0018] Furthermore, after quantifying each original item corresponding to each TCM syndrome, the method further includes:

[0019] The importance of each screening item was quantitatively rated based on the Likert rating method.

[0020] Furthermore, before quantitatively rating the importance of each screening item based on the Likert rating method, the method further includes:

[0021] Based on multiple sample balancing processing algorithms, the item screening sample sets are respectively balanced to obtain the balanced sample sets corresponding to each sample balancing processing algorithm;

[0022] Based on the random forest algorithm, the importance of each balanced sample set is sorted respectively, and the feature contribution of each screening item in the balanced sample set corresponding to each sample balancing algorithm is obtained respectively;

[0023] The characteristic contribution of each screening item in the balanced sample set corresponding to each sample equalization processing algorithm is used as the basis of the Likert rating method.

[0024] Furthermore, the sample set equalization processing algorithm includes a random oversampling algorithm, an adaptive synthetic sampling algorithm and a synthetic minority class oversampling algorithm.

[0025] Furthermore, after constructing a TCM syndrome prediction model based on a deep learning algorithm, the method further includes:

[0026] The TCM syndrome prediction model was evaluated through reliability assessment algorithm, validity assessment algorithm and responsiveness assessment algorithm.

[0027] Furthermore, the updating of the training parameters of the TCM syndrome prediction model according to the quality scores of the samples screened by each item includes:

[0028] During the training process of the TCM syndrome prediction model, the loss function or learning rate of the TCM syndrome prediction model is adjusted according to the quality scores of the samples screened for each item.

[0029] The second aspect of the present invention provides a method for predicting TCM syndromes of rheumatoid joints applied to the method for constructing a TCM syndrome prediction model of rheumatoid joints as described in the first aspect, comprising:

[0030] Get the text of symptom descriptions of patients with rheumatoid arthritis;

[0031] Extract features from symptom description text based on natural language processing algorithm to obtain feature data;

[0032] The characteristic data is input into the trained TCM syndrome prediction model to obtain the TCM syndrome prediction results.

[0033] The beneficial effects of the present invention are as follows:

[0034] The present invention develops an efficient TCM syndrome prediction method, which can process symptom description text information uploaded by patients based on natural language processing technology to ensure accurate feature extraction from the data source.

[0035] The present invention proposes a fusion model framework based on deep learning, which integrates the natural language processing model (NLP) for text to achieve comprehensive analysis of symptom descriptions and prediction of TCM syndromes.

[0036] The present invention adopts a personalized model training method to dynamically update the training parameters of the TCM syndrome prediction model according to the quality score of each medical record, so as to increase the model's weight for learning high-quality medical records, thereby improving the accuracy of diagnostic prediction.

[0037] At the same time, the present invention realizes automatic quality assessment of sample data by quantifying each original entry corresponding to each TCM syndrome, ensuring that low-quality data will not affect the accuracy of the diagnosis result, thereby improving the robustness and reliability of the entire system. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The specific implementation modes of the present invention are further described in detail below in conjunction with the accompanying drawings.

[0039] Figure 1 A flow chart showing a method for constructing a TCM syndrome prediction model for rheumatoid joint provided by an embodiment of the present invention.

[0040] Figure 2 A flow chart showing a method for predicting TCM syndromes of rheumatoid joint provided by one embodiment of the present invention.

[0041] Figure 3 A scree plot for factor analysis of rheumatic arthralgia syndrome provided by one embodiment of the present invention is shown.

[0042] Figure 4 The importance ranking results after sample equalization by random oversampling algorithm and random forest algorithm are shown.

[0043] Figure 5 The importance ranking results after sample equalization processing by the synthetic sampling algorithm and random forest algorithm are shown.

[0044] Figure 6 The importance ranking results after sample equalization processing by the synthetic minority class oversampling algorithm and random forest algorithm are shown.

[0045] Figure 7 A schematic diagram showing the structure of a computer system for implementing the apparatus provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0046] In order to explain the present invention more clearly, the present invention is further described below in conjunction with embodiments and drawings. Similar components in the drawings are represented by the same reference numerals. It should be understood by those skilled in the art that the content specifically described below is illustrative rather than restrictive, and the protection scope of the present invention should not be limited thereto.

[0047] like Figure 1 As shown, the first embodiment of the present invention provides a method for constructing a TCM syndrome prediction model for rheumatoid joint, comprising:

[0048] Step S11: obtaining a plurality of original samples that are marked as corresponding to TCM syndromes and include original entry data corresponding to the TCM syndromes to form an original sample set;

[0049] Step S12: According to the original sample set, the original items of each TCM syndrome are screened by a plurality of screening algorithms to obtain screening items corresponding to each TCM syndrome;

[0050] Step S13: further obtaining a plurality of item screening samples that are marked as corresponding to TCM syndromes and include screening items corresponding to the TCM syndromes based on the original sample set to form an item screening sample set;

[0051] Step S14: quantify each original item corresponding to each TCM syndrome to evaluate the quality score of each original sample; use the quality score of each original sample as the quality score of the item screening sample corresponding to the original sample;

[0052] Step S15: Divide the item screening sample set into a training set and a test set;

[0053] Step S16: constructing a TCM syndrome prediction model based on a deep learning algorithm, and training the TCM syndrome prediction model using a training set;

[0054] Step S17: updating the training parameters of the TCM syndrome prediction model according to the quality scores of the samples screened for each item, so as to obtain a trained TCM syndrome prediction model.

[0055] The present invention adopts a personalized model training method to dynamically update the training parameters of the TCM syndrome prediction model according to the quality score of each medical record, so as to increase the model's weight for learning high-quality medical records, thereby improving the accuracy of diagnostic prediction.

[0056] At the same time, the present invention realizes automatic quality assessment of sample data by quantifying each original entry corresponding to each TCM syndrome, ensuring that low-quality data will not affect the accuracy of the diagnosis result, thereby improving the robustness and reliability of the entire system.

[0057] In a possible implementation, a plurality of original samples that are marked as corresponding to TCM syndromes and include original entry data corresponding to the TCM syndromes are obtained to form an original sample set including

[0058] Collect electronic medical record data of rheumatoid arthritis patients, which should include information such as symptom descriptions. Obtain symptom description texts of multiple rheumatoid arthritis patients, and the symptom description texts are annotated with corresponding TCM syndromes; perform word segmentation, denoising, and keyword extraction on the symptom description text data of each TCM syndrome in turn to extract multiple original entry data corresponding to the TCM syndrome; obtain multiple original samples annotated as corresponding TCM syndromes and including the original entry data corresponding to the TCM syndrome, and form an original sample set.

[0059] In a possible implementation, the screening algorithm is selected from one or more of item distribution examination method, Cronbach's coefficient method, factor analysis method, discrete trend method and expert evaluation method.

[0060] Item distribution examination method: Analyze the distribution of items in each syndrome, screen out items with high discrimination, and enhance the effect of syndrome classification.

[0061] Discrete trend method: calculate the discrete index of the items (such as variance or standard deviation) and select items with high stability and consistency.

[0062] Expert evaluation method: Combine the experience and opinions of Chinese medicine experts to evaluate the practicality and relevance of the items.

[0063] Cronbach's coefficient method: detect the consistency of items within the same syndrome to ensure measurement reliability.

[0064] Factor analysis: Analyze the correlation between items through statistical methods and optimize the item structure to accurately reflect the multidimensional characteristics of TCM syndromes.

[0065] In one possible implementation, quantifying each original item corresponding to each TCM syndrome includes:

[0066] Step S141, respectively calculating the subjective weight and objective weight of each original item in the corresponding TCM syndrome to calculate the comprehensive weight of each original item in the TCM syndrome;

[0067] Step S142: Calculate the quality score of each original sample according to the comprehensive weight of each original item in each TCM syndrome.

[0068] In a possible implementation, after quantifying each original item corresponding to each TCM syndrome, the method further includes:

[0069] The importance of each screening item was quantitatively rated based on the Likert rating method.

[0070] The present invention uses a subjective and objective comprehensive weighting method: combining quantitative analysis and expert opinions to balance the scientificity and practicality of the items, thereby calculating the quality score of each item screening sample.

[0071] The present invention uses the Likert Scale (Likert) rating method: the Likert scale is used to quantify the importance of the items and simplify the subsequent analysis process such as evaluating the TCM syndrome prediction model.

[0072] In a possible implementation, before quantitatively rating the importance of each screening item based on the Likert rating method, the method further includes:

[0073] Based on multiple sample balancing processing algorithms, the item screening sample sets are respectively balanced to obtain the balanced sample sets corresponding to each sample balancing processing algorithm;

[0074] Based on the random forest algorithm, the importance of each balanced sample set is sorted respectively, and the feature contribution of each screening item in the balanced sample set corresponding to each sample balancing algorithm is obtained respectively;

[0075] The characteristic contribution of each screening item in the balanced sample set corresponding to each sample equalization processing algorithm is used as the basis of the Likert rating method.

[0076] In a possible implementation, the sample set equalization processing algorithm includes a random oversampling algorithm, an adaptive synthetic sampling algorithm and a synthetic minority class oversampling algorithm.

[0077] In one possible implementation, after building a TCM syndrome prediction model based on a deep learning algorithm, the method further includes:

[0078] The TCM syndrome prediction model was evaluated through reliability assessment algorithm, validity assessment algorithm and responsiveness assessment algorithm.

[0079] The present invention evaluates the effectiveness, reliability and applicability of the TCM syndrome prediction model in scientific research and practical applications through reliability evaluation, validity evaluation and responsiveness evaluation.

[0080] In one possible implementation, updating the training parameters of the TCM syndrome prediction model according to the quality scores of the samples screened for each item includes:

[0081] During the training process of the TCM syndrome prediction model, the loss function or learning rate of the TCM syndrome prediction model is adjusted according to the quality scores of the samples screened for each item.

[0082] The present invention uses the collected and pre-processed training data set to train the model. During the training process, by setting a suitable loss function and optimizer, the model parameters are continuously optimized to improve the prediction accuracy of the model.

[0083] Implement cross-validation and other technical means to evaluate the generalization ability of the model, and adjust the model structure and parameters based on the evaluation results to achieve the best prediction effect.

[0084] like Figure 2 As shown, the second embodiment of the present invention provides a method for predicting TCM syndromes of rheumatoid joints applied to the method for constructing a TCM syndrome prediction model of rheumatoid joints in the first embodiment, comprising:

[0085] Step S21, obtaining a symptom description text of a rheumatoid arthritis patient;

[0086] Step S22: extracting features from the symptom description text based on a natural language processing algorithm to obtain feature data;

[0087] Step S23, inputting the characteristic data into the trained TCM syndrome prediction model to obtain TCM syndrome prediction results.

[0088] In the present invention, a TCM syndrome prediction model is constructed based on deep learning technology, which can predict TCM syndromes based on patient symptom descriptions. The model can be designed by using a natural language processing (NLP) model to process symptom description texts, and finally perform feature vector fusion to output predicted TCM syndromes through a fully connected layer.

[0089] In the present invention, the diagnosis of TCM syndrome is highly subjective, usually depends on the doctor's experience and knowledge, and is susceptible to the doctor's level and experience. Building a model is extremely critical for evaluating the changes in syndromes. It not only helps to monitor the progression of the disease and adjust the treatment plan, but also unifies the diagnostic criteria, reduces human errors, and thus improves the consistency and accuracy of diagnosis. At the same time, different patients can show different syndrome types. Through the model, the syndrome type of the patient can be quickly identified, providing a scientific basis for formulating a personalized treatment plan. And the model can continuously track the changes in the patient's syndrome, helping doctors adjust the treatment method according to the patient's real-time status, better control the condition, and optimize the treatment results.

[0090] In the present invention, the TCM syndrome prediction model can be used as an auxiliary tool for doctors in diagnosing TCM syndromes of rheumatoid arthritis. Through data analysis and machine learning and other technologies, the development trend and influencing factors of RA can be understood and predicted to help doctors make more accurate diagnoses and formulate more effective treatment plans.

[0091] Combining traditional Chinese medicine diagnostic methods with modern technology will not only help inherit and develop Chinese medicine, but also introduce the unique charm of Chinese medicine to the world and promote the international development of Chinese medicine.

[0092] The construction of the model in the present invention loads a large amount of medical data, including the patient's clinical information and imaging data, which requires strict data cleaning and analysis, and can provide rich data resources for traditional Chinese medicine research. These data are helpful for in-depth analysis of the characteristics of traditional Chinese medicine syndromes of rheumatoid arthritis, and provide a valuable empirical basis for the research and development of traditional Chinese medicine.

[0093] The present invention develops an efficient TCM syndrome prediction method, which can process symptom description text information uploaded by patients based on natural language processing technology to ensure accurate feature extraction from the data source.

[0094] The present invention proposes a fusion model framework based on deep learning, which integrates the natural language processing model (NLP) for text to achieve comprehensive analysis of symptom descriptions and prediction of TCM syndromes.

[0095] In one possible implementation, the present invention also introduces a dynamic update and learning mechanism that allows the model to be continuously optimized and adjusted based on newly added case data, ensuring that the diagnostic system can adapt to new data distributions and changes over time and maintain its long-term effectiveness and accuracy.

[0096] Specifically, a plurality of new samples that are marked as corresponding TCM syndromes and include screening item data corresponding to the TCM syndromes are obtained to form a new sample set; the parameters of the TCM syndrome prediction model are optimized and adjusted through the new sample set.

[0097] In one possible implementation, the present invention also develops a set of optimization technologies for mobile application user interface (UI) and user experience (UX) design to ensure that users can easily and quickly upload and input symptom descriptions, and can clearly and easily receive and understand diagnosis results.

[0098] The present invention develops a user-friendly mobile application interface, where patients can input symptom descriptions through their mobile phones. The trained auxiliary diagnosis model is integrated into the mobile application to ensure that the data uploaded by the patient can be correctly processed by the model and the prediction results are presented to the user in an easy-to-understand manner.

[0099] After completing application development and model integration, conduct comprehensive testing, including unit testing, integration testing, and user acceptance testing, to ensure application stability and user experience.

[0100] Applications that have passed the test will be deployed to the application market for patients to download and use, and a feedback mechanism will be established to collect user feedback to guide subsequent application optimization and upgrades.

[0101] In one possible implementation, the present invention also implements advanced privacy protection and data security technologies to ensure that personal health information uploaded by users is strictly protected, prevents data leakage, and meets the requirements of relevant laws and regulations.

[0102] In one embodiment of the present invention, the invention uses HTTPS or TLS / SSL protocol to encrypt all network transmitted data to ensure that the data is not stolen or tampered with during transmission. The personal health information stored on the server is encrypted, and even if the data is illegally obtained, its content cannot be directly read. A strong encryption algorithm such as AES can be used.

[0103] In a specific embodiment, step S11: the data obtained are as follows: the number of samples of this experimental data is 11039, of which 1852 are rheumatism arthralgia syndrome, 1950 are liver and kidney deficiency syndrome, 2519 are cold and dampness arthralgia syndrome, 507 are qi and blood deficiency syndrome, 256 are qi and yin deficiency syndrome, 2940 are damp-heat arthralgia syndrome, 731 are phlegm and blood stasis arthralgia syndrome, and 283 are blood stasis and collateral obstruction syndrome.

[0104] The original data set contains many samples with missing values. To ensure the accuracy of the experiment, these samples were uniformly eliminated, and outliers and atypical data were removed. After processing, 1,895 samples remained, which were used as the final data of this experiment. The features in the data are complex and there are multiple overlaps. They are processed with unique hot encoding, and the absence of the symptom is marked as 2, and the presence of the symptom is marked as 1. Some items contained in the features are extracted to meet the subsequent statistical requirements.

[0105] For all the experimental data, there were 16 main rheumatoid symptom items in all samples, such as "joint pain, swelling, wandering", "pain worse at night", etc., and 30 secondary rheumatoid symptom items, such as "aversion to wind, or sweating", "headache", etc. For the experiment, 3 items were retained for tongue color, 3 items were retained for tongue shape, 7 items were retained for tongue coating, 4 items were retained for tongue coating color, and 12 items were retained for pulse condition.

[0106] The contents of the rheumatic arthralgia syndrome after screening in step S12 are as follows:

[0107] This experiment mainly uses Cronbach's coefficient method, factor analysis method, and discrete trend method to verify items for rheumatic arthralgia syndrome. The item distribution observation method verifies items by examining the item frequency. However, the data in this experiment is unbalanced and one-hot encoding is used. For a few syndromes and items, the requirements cannot be met, so this method is not used to examine the items.

[0108] Table 1 is a table of Cronbach coefficients for rheumatic arthralgia syndrome. The Cronbach coefficient method is to observe whether the Cronbach Alpha coefficient is greater than 0.7. When the coefficient is greater than 0.7, the system has a high reliability. At the same time, observe the impact of deleting each item on the coefficient. According to the table, deleting the slippery pulse item will have a small increase in the Cronbach Alpha index.

[0109] Table 1 Cronbach coefficient table of rheumatic arthralgia syndrome

[0110]

[0111] The discrete trend method calculates the coefficient of variation by comparing the standard deviation and mean of each item. The coefficient of variation is the standard deviation divided by the mean. When the standard deviation is less than 0.75 or the coefficient of variation is less than 0.15, the item passes the test.

[0112] Table 2 is the data table of rheumatic arthralgia syndrome obtained by discrete trend method. Each item in it meets the requirements, so no item removal is considered.

[0113] Table 2 Data table of rheumatic arthralgia syndrome obtained by discrete trend method

[0114]

[0115] The items were screened by factor analysis. Table 3 is the rotated component matrix of the factor analysis of rheumatic arthralgia syndrome. Figure 1 This is the scree plot of factor analysis of rheumatic arthralgia syndrome. The factor analysis method uses a small number of factors to explain the original data. When the KMO measure in the KMO test during the factor analysis is greater than 0.6 and the significance of the Bartlett sphericity test is less than 0.05, it means that factor analysis can be performed. The larger the KMO measure, the more suitable it is for factor analysis. The factor analysis screening items require that each item has an association with any factor, that is, when the load of the item on different factors is less than 0.4, consider deleting this item. The rotated component matrix is ​​obtained by the maximum variance method. The scree plot is used to select the number of factors, which indicates the size of the eigenvalues ​​corresponding to each factor. When the eigenvalue does not change significantly with the factor, that is, the scree plot tends to be flat, the number of factors is selected.

[0116] like Figure 3 As shown in the figure, the horizontal axis is the number of corresponding factors, and the vertical axis is the size of the eigenvalue corresponding to the corresponding factor. The eigenvalue of the scree plot tends to be flat after selecting 3 factors, so the first three factors are considered as the component factors of rheumatic arthralgia syndrome. Table 3 is the component matrix after rotation of the factor analysis of rheumatic arthralgia syndrome, in which each item has a related factor, so the items are tested by factor analysis and are not considered for deletion.

[0117] Table 3 Rotated component matrix of factor analysis of rheumatic arthralgia syndrome

[0118]

[0119]

[0120] The following formula is the factor expression extracted from the factor analysis component matrix, where the coefficients are the loadings of each item on different factors, Y n Indicates the final score of each factor, X n Indicates the numerical value of the item. The relationship expression between the total score and each factor is calculated through the variance explanation table of the rheumatic arthralgia syndrome components in Table 4. The final score result and the corresponding weight proportion of each item are obtained by simplifying the expression of each factor, as shown in the score table of each item of rheumatic arthralgia syndrome in Table 5.

[0121] Y1=0.825X1+0.819X2+0.853X3+0.464X4+0.782X5+0.492X6+0.472X7+0.382X8+0.264X9+0.586X 10

[0122] Y2=-0.127X1-0.105X2-0.137X3-0.225X4-0.085X4+0.271X6+0.390X7+0.589X8-0.786X9+0.244X 10

[0123] Y3=-0.105X1+0.001X2-0.027X3+0.516X4-0.192X4+0.137X6+0.230X7+0.405X8+0.170X9-0.610X 10

[0124] Table 4 Variance explanation table of components of rheumatic arthralgia syndrome

[0125]

[0126]

[0127] Y=0.476X1+0.493X2+0.503X3+0.322X4+0.445X5+0.390X6+0.417X7+0.430X8+0.022X9+0.329X 10

[0128] Y is the final score of Fengshi Bidu syndrome in factor analysis, X n These are the data for each item of rheumatic arthralgia syndrome.

[0129] Table 5 Scoring table for each item of rheumatic arthralgia syndrome

[0130]

[0131] On the basis of the above, the process of screening original items in step S12 of other seven main syndrome categories such as damp-heat arthralgia, cold-damp arthralgia, phlegm-stasis arthralgia, blood stasis, deficiency of both qi and blood, insufficiency of liver and kidney, deficiency of both qi and yin is similar to the above, and will not be repeated here. After carrying out the above screening, comprehensive analysis of each syndrome result is performed. Screen out multiple syndrome overlapping items, for such overlapping items in multiple syndromes as fur color and fur quality, it will have an impact on syndrome discrimination and analysis, such as light red tongue color, red and thin fur quality, greasy, etc., and subsequent refinement adjustment is needed to make differences between different syndromes. Simultaneously, other processing such as merging for some special items is carried out to reduce data dimensions, such as less fur without fur, and different results in pulse are merged and refined.

[0132] In a specific implementation, step S141, respectively calculating the subjective weight and objective weight of each original item in the corresponding TCM syndrome to calculate the comprehensive weight of each original item in the TCM syndrome; step S142, calculating the quality score of each original sample according to the comprehensive weight of each original item in each TCM syndrome, includes:

[0133] (1) Rheumatic arthralgia syndrome:

[0134] The objective weight is normalized by the expression calculated for the syndrome above. The following is the score expression calculated above:

[0135] Y=0.476X1+0.493X2+0.503X3+0.322X4+0.445X5+0.390X6+0.417X7

[0136] +0.430X8+0.022X9+0.329X 10

[0137] Y is the final score of Fengshi Bidu syndrome in factor analysis, X n These are the data of each original item of rheumatic arthralgia syndrome.

[0138] The coefficients are normalized to obtain the new score expression as follows:

[0139] Y a =0.944X1+0.979X2+X3+0.624X4+0.879X5+0.765X6+0.821X7

[0140] +0.848X8+0.638X 10

[0141] Y aIt is the objective weight score of rheumatic arthralgia syndrome.

[0142] The subjective weight was calculated using the expert scoring table. The weight was the ratio of the mean expert score of each item to the total mean score of each syndrome item as the weight coefficient. Normalization was performed at the same time. The calculated score expression was as follows:

[0143] Y b =X1+0.850X2+0.618X3+0.638X5+0.713X6+0.713X7+0.850X8

[0144] +0.488X9+0.382X 10

[0145] Y b It is the objective weight score of rheumatic arthralgia syndrome.

[0146] The subjective and objective weights are integrated, and the weighted average fusion strategy with equal subjective and objective weights is initially adopted for fusion. The formula is as follows:

[0147]

[0148] The final score is calculated as follows:

[0149] Y z =0.972X1+0.915X2+0.809X3+0.312X4+0.759X5+0.739X6+0.767X7

[0150] +0.849X8+0.244X9+0.510X 10

[0151] Y z It is the comprehensive weighted score of rheumatic arthralgia syndrome.

[0152] Based on the above content, the process of executing step S141 and step S142 for the other seven major syndromes, such as damp-heat obstruction, cold-dampness obstruction, phlegm-stasis obstruction, blood stasis obstruction, qi and blood deficiency, liver and kidney deficiency, and qi and yin deficiency, is similar to the above, and will not be repeated here. The coefficients of the total score formula of each syndrome are normalized and calculated into the comprehensive weight score formula of the RA scale. Table 6 is a table of the comprehensive weight score formulas of each syndrome, as shown in Table 6:

[0153] Table 6 Comprehensive weight score formula for each syndrome

[0154]

[0155]

[0156] In a specific implementation, step S1431, based on multiple sample equalization processing algorithms, respectively, equalization processing is performed on the item screening sample sets to obtain the equalization sample sets corresponding to each sample equalization processing algorithm; step S1432, based on the random forest algorithm, respectively, the importance of each equalization sample set is sorted to obtain the feature contribution of each screening item in the equalization sample set corresponding to each sample equalization processing algorithm; the steps include:

[0157] Machine learning is performed based on the samples to rank the feature importance of the eight syndromes. Random forest is used for feature importance training. Due to the uneven distribution of multiple syndromes in the samples, three equalization processing methods are used for sample equalization, namely random oversampling algorithm, adaptive synthetic sampling algorithm and synthetic minority class oversampling algorithm.

[0158] The sample is balanced by the random oversampling algorithm. The importance ranking results after the random forest algorithm are as follows Figure 4 As shown in Table 7, the Chinese and English comparison table of the entries is as follows:

[0159] Table 7 Chinese-English comparison table of items

[0160]

[0161]

[0162] It can be seen that the random over-sampling algorithm is used to balance the samples. Among them, "purple darkness" has the highest contribution, reaching 25.6%, followed by "the heat alleviated", "joint pain and swelling" and "shortness of breath", accounting for 19.2%, 13.2% and 10.8% respectively. It can be seen that the characteristics of tongue color: purple darkness and reduced joint heat pain can be used as important indicators to judge the type of rheumatoid arthritis.

[0163] After the samples are balanced by the adaptive synthetic sampling algorithm, the importance ranking results after the random forest algorithm are as follows: Figure 5 As shown in the figure, the synthetic sampling (ADASYN) algorithm is used to balance the samples. Among them, "purpledarkness" has the highest contribution, reaching 27.3%, followed by "white", "the heat is alleviated" and "lightred", accounting for 17.3%, 13.2% and 10.8% respectively. It can be seen that tongue color: purple dark and tongue color: white can be used as important indicators to judge the type of rheumatoid arthritis.

[0164] After the sample is balanced by the synthetic minority class oversampling algorithm, the importance ranking results after random forest algorithm processing are as follows Figure 6 As shown in the figure, the synthetic minority oversampling algorithm (SMOTE) is used to equalize the samples. Among them, "purpledarkness" has the highest contribution, reaching 26.1%, followed by "aggravated by cold", "white" and "joint is tingling", accounting for 19%, 12.2% and 9.1% respectively. It can be seen that tongue color: purple dark and other features can be used as important indicators for judging rheumatoid category.

[0165] like Figure 7 As shown, a computer system suitable for implementing a method for constructing a prediction model of rheumatoid joint TCM syndrome and a prediction method provided by the above-mentioned embodiment includes a central processing module (CPU), which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) or a program loaded from a storage portion into a random access memory (RAM). In the RAM, various programs and data required for the operation of the computer system are also stored. The CPU, ROM and RAM are connected to each other via a bus. An input / output (I / O) interface is also connected to the bus.

[0166] The following components are connected to the I / O interface: an input part including a keyboard, a mouse, etc.; an output part including a liquid crystal display (LCD), etc. and a speaker, etc.; a storage part including a hard disk, etc.; and a communication part including a network interface card such as a LAN card, a modem, etc. The communication part performs communication processing via a network such as the Internet. A drive is also connected to the I / O interface as needed. Removable media, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., are installed on the drive as needed so that the computer program read therefrom is installed into the storage part as needed.

[0167] In particular, according to this embodiment, the process described in the flowchart above can be implemented as a computer software program. For example, this embodiment includes a computer program product, which includes a computer program tangibly contained on a computer-readable medium, and the computer program includes program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through a communication part, and / or installed from a removable medium.

[0168] The flow chart and schematic diagram in the accompanying drawings illustrate the possible architecture, function and operation of the system, method and computer program product of the present embodiment. In this regard, each square box in the flow chart or schematic diagram can represent a part of a module, program segment or code, and a part of the above-mentioned module, program segment or code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two square boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the schematic diagram and / or flow chart, and the combination of the square boxes in the schematic diagram and / or flow chart can be implemented with a dedicated hardware-based system that performs a specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0169] In the description of the present invention, it should be noted that the terms "upper", "lower", etc. indicate positions or positional relationships based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific position, be constructed and operated in a specific position, and therefore cannot be understood as a limitation on the present invention. Unless otherwise clearly specified and limited, the terms "installed", "connected", and "connected" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be an indirect connection through an intermediate medium, or it can be a connection between the two elements. For those of ordinary skill in the art, the specific meanings of the above terms in the present invention can be understood according to the specific circumstances.

[0170] It should also be noted that, in the description of the present invention, relational terms such as first and second, etc., are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "comprise", "include" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device. In the absence of further restrictions, an element defined by the statement "comprises a ..." does not exclude the presence of other identical elements in the process, method, article or device including the element.

[0171] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not limitations on the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is impossible to list all the embodiments here. All obvious changes or modifications derived from the technical solution of the present invention are still within the protection scope of the present invention.

Claims

1. A method for constructing a TCM syndrome prediction model for rheumatoid joint, characterized in that: include: Acquire a plurality of original samples that are marked as corresponding to TCM syndromes and include original entry data corresponding to the TCM syndromes to form an original sample set; According to the original sample set, the original items of each TCM syndrome are screened by various screening algorithms to obtain the screening items corresponding to each TCM syndrome; Based on the original sample set, multiple item screening samples are further obtained, which are marked as corresponding to TCM syndromes and include screening items corresponding to the TCM syndromes, to form an item screening sample set; Each original item corresponding to each TCM syndrome is quantified to evaluate the quality score of each original sample; the quality score of each original sample is used as the quality score of the item screening sample corresponding to the original sample; Divide the item screening sample set into a training set and a test set; A TCM syndrome prediction model is constructed based on a deep learning algorithm, the TCM syndrome prediction model is trained using the training set, and the training parameters of the TCM syndrome prediction model are updated according to the quality scores of the samples screened for each item to obtain a trained TCM syndrome prediction model.

2. The method for constructing a TCM syndrome prediction model for rheumatoid arthritis according to claim 1, characterized in that: The method of obtaining a plurality of original samples that are marked as corresponding to TCM syndromes and include original entry data corresponding to the TCM syndromes, forming an original sample set including Acquire symptom description texts of multiple rheumatoid arthritis patients, wherein the symptom description texts are annotated with corresponding TCM syndromes; subject the symptom description text data of each TCM syndrome to word segmentation, denoising and keyword extraction in turn to extract multiple original entry data corresponding to the TCM syndrome; obtain multiple original samples annotated as corresponding TCM syndromes and including the original entry data corresponding to the TCM syndrome to form an original sample set.

3. The method for constructing a TCM syndrome prediction model for rheumatoid arthritis according to claim 1, characterized in that: The screening algorithm is selected from one or more of item distribution inspection method, Cronbach's coefficient method, factor analysis method, discrete trend method and expert evaluation method.

4. The method for constructing a TCM syndrome prediction model for rheumatoid arthritis according to claim 1, characterized in that: The quantification of each original item corresponding to each TCM syndrome includes Calculate the subjective weight and objective weight of each original item in the corresponding TCM syndrome respectively to calculate the comprehensive weight of each original item in the TCM syndrome; The quality score of each original sample was calculated based on the comprehensive weight of each original item in each TCM syndrome.

5. The method for constructing a TCM syndrome prediction model for rheumatoid joint according to claim 4, characterized in that: After quantifying each original item corresponding to each TCM syndrome, the method further includes The importance of each screening item was quantitatively rated based on the Likert rating method.

6. The method for constructing a TCM syndrome prediction model for rheumatoid joint according to claim 5, characterized in that: Before quantitatively rating the importance of each screening item based on the Likert rating method, the method also includes Based on multiple sample balancing processing algorithms, the item screening sample sets are respectively balanced to obtain the balanced sample sets corresponding to each sample balancing processing algorithm; Based on the random forest algorithm, the importance of each balanced sample set is sorted respectively, and the feature contribution of each screening item in the balanced sample set corresponding to each sample balancing algorithm is obtained respectively; The characteristic contribution of each screening item in the balanced sample set corresponding to each sample equalization processing algorithm is used as the basis of the Likert rating method.

7. The method for constructing a TCM syndrome prediction model for rheumatoid joint according to claim 6, characterized in that: The sample set equalization processing algorithm includes a random oversampling algorithm, an adaptive synthetic sampling algorithm and a synthetic minority class oversampling algorithm.

8. The method for constructing a TCM syndrome prediction model for rheumatoid joint according to claim 1, characterized in that: After constructing a TCM syndrome prediction model based on a deep learning algorithm, the method further includes evaluating the TCM syndrome prediction model through a reliability assessment algorithm, a validity assessment algorithm, and a responsiveness assessment algorithm.

9. The method for constructing a TCM syndrome prediction model for rheumatoid joint according to claim 1, characterized in that: The training parameters of the TCM syndrome prediction model are updated according to the quality scores of the samples screened for each item, including During the training process of the TCM syndrome prediction model, the loss function or learning rate of the TCM syndrome prediction model is adjusted according to the quality scores of the samples screened for each item.

10. A method for predicting TCM syndromes of rheumatoid joints applied to the method for constructing a TCM syndrome prediction model of rheumatoid joints as claimed in any one of claims 1 to 9, characterized in that: include: Get the text of symptom descriptions of patients with rheumatoid arthritis; Extract features from symptom description text based on natural language processing algorithm to obtain feature data; The characteristic data is input into the trained TCM syndrome prediction model to obtain the TCM syndrome prediction results.

Citation Information

Cited By

  • Traditional Chinese medicine syndrome prediction method fusing large language model and knowledge graph

    CN119811689A

  • Pre-examination triage method and device, equipment and storage medium

    CN119851905A