Construction method and application of rheumatoid arthritis condition prediction model
By constructing a disease prediction model for rheumatoid arthritis and utilizing cross-modal datasets and multi-feature fusion network models, the problem of primary healthcare institutions having difficulty screening suspected cases of rheumatoid arthritis has been solved, enabling rapid and accurate disease prediction and timely treatment.
Patent Information
- Application Number
- CN202511314122.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-09-15
- Publication Date
- 2025-12-30
AI Technical Summary
The lack of key diagnostic tools and expertise in primary healthcare institutions makes it difficult to quickly screen suspected rheumatoid arthritis patients, thus delaying treatment.
A rheumatoid arthritis disease prediction model was constructed. By acquiring cross-modal datasets and training a multi-feature fusion network model, feature vectors from numerical detection data and image data were fused to generate a final feature vector for rapid screening of suspected patients.
It improves the testing efficiency of primary healthcare institutions, can accurately predict the risk of developing rheumatoid arthritis, and helps primary care physicians conduct timely specialized examinations and treatments, avoiding misdiagnosis and missed diagnosis.
Smart Images
Figure CN121237372A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The application belongs to the technical field of intelligent medical treatment, and particularly relates to a construction method of a rheumatoid arthritis condition prediction model and application thereof. BACKGROUND
[0002] Rheumatoid arthritis (RA) is the second largest limb disabling disease in China after cerebrovascular disease, and is characterized by chronic progression and high disability rate. RA is a multifactorial disease that irreversibly damages joints. Early diagnosis and early treatment are crucial for improving prognosis. However, the exact pathogenesis of RA is still unclear in the clinical and academic fields, and there is a lack of early diagnosis, precise classification, prognosis judgment and individualized treatment methods, resulting in that early diagnosis and early treatment are still a worldwide problem.
[0003] In the current treatment of rheumatoid arthritis, most patients suspected of having rheumatoid arthritis will initially go to primary medical institutions for treatment. However, general practitioners in primary medical institutions often lack relevant knowledge of rheumatology, and cannot accurately analyze and identify rheumatoid arthritis patients by obtaining comprehensive diagnostic data. In addition, primary medical institutions lack key diagnostic tools such as nuclear magnetic resonance imaging and X-ray films, and cannot perform various laboratory tests, so there may be delays in the treatment of rheumatoid arthritis. Therefore, primary hospitals urgently need a widely available and effective method to predict the disease condition of rheumatoid arthritis patients. SUMMARY
[0004] The purpose of the present application is to provide a construction method of a rheumatoid arthritis condition prediction model and application thereof, which solves the problem that existing primary medical institutions cannot quickly screen out suspected rheumatoid arthritis patients according to their own conditions.
[0005] To achieve the above purpose, the first technical solution of the present application is implemented as follows: a construction method of a rheumatoid arthritis condition prediction model, comprising:
[0006] acquiring a cross-modal data set;
[0007] training a multi-feature fusion network model based on the cross-modal data set.
[0008] Further, the acquiring of the cross-modal data set comprises:
[0009] acquiring an original data set composed of multiple groups of original data, each group of original data comprising corresponding numerical detection data, image data and labels;
[0010] generating a feature vector for the numerical detection data and the image data in each group of original data, respectively;
[0011] The feature vectors of numerical detection data and image data in each set of original data are merged to obtain the final feature vector, and multiple final feature vectors constitute a cross-modal dataset.
[0012] Furthermore, before generating the feature vector, the method further includes preprocessing the numerical detection data and image data in the original dataset.
[0013] Furthermore, the numerical detection data in the original data undergoes preprocessing, specifically by using a feature selection algorithm to process missing values and imbalanced data in the numerical detection data.
[0014] Furthermore, the image data in the original data is preprocessed, specifically by segmenting and resizing the ultrasound images.
[0015] Furthermore, the generation of the feature vector includes:
[0016] The grayscale image and texture image of the image data are obtained, and the first image features and the second image features are extracted using a convolutional neural network, respectively.
[0017] The numerical detection data, the first image features, and the second image features are used to generate feature vectors, respectively.
[0018] Furthermore, the training of the multi-feature fusion network model specifically involves:
[0019] The cross-modal dataset is divided into a training set and a test set, and the initial network model is trained using the training set data.
[0020] The initial network model was optimized using test set data to obtain a multi-feature fusion network model.
[0021] The first technical solution of the present invention is implemented as follows: the application of a method for constructing a rheumatoid arthritis disease prediction model in the prediction of rheumatoid arthritis disease.
[0022] Furthermore, the application includes:
[0023] The system acquires the user's numerical detection data and image data, preprocesses them to generate feature vectors, and then fuses the feature vectors to obtain the final feature vector.
[0024] A rheumatoid arthritis disease prediction unit is constructed, and the prediction result of the user's rheumatoid arthritis risk is output based on the final feature vector.
[0025] Furthermore, the rheumatoid arthritis disease prediction unit is equipped with the multi-feature fusion network model.
[0026] Compared with existing technologies, the beneficial effects of this invention are as follows: This invention constructs a cross-modal dataset by acquiring clinical big data, musculoskeletal ultrasound, joint imaging, and other data from patients with a history of rheumatoid arthritis, and trains a multi-feature fusion network model based on the cross-modal dataset and verifies its performance; this multi-feature fusion network model is easy to deploy and use, and can provide a preliminary disease prediction for each user's physical condition, so as to help primary care physicians quickly screen out patients who may have rheumatoid arthritis, and the prediction results are relatively accurate. Attached Figure Description
[0027] Figure 1 This is a flowchart illustrating the construction of a rheumatoid arthritis disease prediction model in this invention;
[0028] Figure 2 This is a flowchart of the process for obtaining cross-modal datasets in this invention. Detailed Implementation
[0029] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.
[0030] In the description of this invention, it should be clarified that the terms "vertical," "lateral," "longitudinal," "front," "rear," "left," "right," "up," "down," and "horizontal," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are merely for the convenience of describing this invention. They do not imply that the device or element referred to must have a specific orientation or position, and therefore should not be construed as a limitation of this invention. In the description of this invention, it should be noted that unless otherwise explicitly specified and limited, the terms "installed," "connected," and "joined" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this invention based on the specific circumstances.
[0031] Example 1
[0032] like Figure 1 As shown in this embodiment, the method for constructing a rheumatoid arthritis disease prediction model includes:
[0033] S1, Obtain the cross-modal dataset;
[0034] S2, Construct a multi-feature fusion network model based on the cross-modal dataset.
[0035] like Figure 2As shown, in some embodiments of this application, obtaining cross-modal datasets includes:
[0036] S11. Obtain the original dataset consisting of multiple sets of original data. Each set of original data includes corresponding numerical detection data, image data, and labels, which are the detection results given by the doctor.
[0037] The numerical detection data include rheumatoid arthritis-related parameters such as red blood cell distribution width (Rf), C-reactive protein (CRP), erythrocyte sedimentation rate (ESR), uric acid (UA), hemoglobin (HB), platelets (PLT), mean platelet volume (MS), age, number of swollen joints (JS), number of tender joints (JT), and number of joints with limited range of motion (JD). The image data includes musculoskeletal ultrasound images and joint imaging data.
[0038] Specifically, it is also necessary to preprocess the numerical detection data and image data in the original dataset to improve the quality of the original data and ensure the accuracy of the multi-feature fusion network model trained based on the original data.
[0039] The numerical detection data is preprocessed by using a feature selection algorithm to remove missing values and imbalanced data, thereby reducing the number of features and improving the quality of the numerical detection data. Based on this data, the complexity of constructing a multi-feature fusion network model is reduced, and the generalization ability of the model is improved.
[0040] Image data preprocessing specifically involves image segmentation and resizing. Image segmentation helps to better understand and analyze objects and features in the image, while resizing can reduce computation while maintaining image quality, thus reducing the complexity of building a multi-feature fusion network model.
[0041] S12. Generate feature vectors for the numerical detection data and image data in each set of original data;
[0042] Specifically, grayscale images and texture images of the image data in each set of original data are obtained, and first image features and second image features are extracted from the grayscale images and texture images respectively;
[0043] For each set of original data, generate feature vectors from the numerical detection data, the first image feature, and the second image feature.
[0044] S13. Use weighted summation and other methods to fuse multiple feature vectors in each set of original data to obtain the final feature vector. Normalize the multiple final feature vectors. The normalized data constitutes a cross-modal dataset.
[0045] This invention acquires raw data such as clinical big data, musculoskeletal ultrasound, and joint imaging from patients with a history of rheumatoid arthritis. After preprocessing the raw data, feature vectors are extracted and used to construct a cross-modal dataset. Based on this dataset, a multi-feature fusion network model is built. This model is easy to deploy and use, and can provide a preliminary disease prediction for each user's physical condition. This helps primary care physicians quickly screen for patients who may have rheumatoid arthritis, and the prediction results are relatively accurate. Based on this prediction, primary care physicians can promptly conduct specialized examinations for suspected rheumatoid arthritis patients and promptly refer them for specialized treatment after diagnosis. This improves the detection efficiency of primary healthcare institutions and avoids situations where rheumatoid arthritis patients are diagnosed and treated in a timely manner.
[0046] In some embodiments of this application, the construction of the multi-feature fusion network model described in S2 specifically refers to:
[0047] The cross-modal dataset is divided into training and test sets in a ratio of 7:3, 4:1, or 5:1.
[0048] Using random forest as the network model, we set the initial hyperparameters of random forest, such as the number of decision trees, maximum depth, and feature selection method. We then train the network model using the training set data to obtain the initial network model, which learns the relationship between the features and labels of each group of data in the training set.
[0049] The feature is the final feature vector generated based on the numerical detection data and image data in the original data, and the label is the detection result given by the doctor based on the patient's detection data, i.e. whether the patient has rheumatoid arthritis.
[0050] The final feature vector of the test set is input into the initial network model, which then provides a prediction result. The prediction result is compared with the corresponding label of the test set data to determine the accuracy of the initial network model. Based on the comparison result, the hyperparameters of the initial network model are adjusted. When the accuracy is low, the number of decision trees and the maximum depth are increased, and when overfitting occurs, the maximum depth is decreased, thereby optimizing and obtaining a multi-feature fusion network model.
[0051] The method for constructing a rheumatoid arthritis disease prediction model provided in this embodiment first preprocesses the original data to improve the quality of the original data and ensure the prediction accuracy of the multi-feature fusion network model trained on it. This embodiment also extracts feature vectors from numerical detection data and image data, and fuses the feature vectors to obtain the final feature vector. Based on the final feature vector, a multi-feature fusion network model that can accurately predict the user's rheumatoid arthritis is obtained through training and optimization.
[0052] The multi-feature fusion network model constructed in this embodiment can be easily deployed on a controller with storage and computing functions, enabling mobile healthcare, improving ease of use and prediction accuracy, and laying the foundation for the research and application of an intelligent system for the prevention and monitoring of autoimmune diseases.
[0053] Example 2
[0054] This embodiment applies the rheumatoid arthritis disease prediction model constructed in Example 1 to the prediction of rheumatoid arthritis disease, specifically as follows:
[0055] The system acquires the user's numerical detection data and image data, preprocesses them to generate feature vectors, and then fuses the feature vectors to obtain the final feature vector.
[0056] A rheumatoid arthritis (RA) disease prediction unit is constructed, which includes the aforementioned multi-feature fusion network model. Based on the final feature vector, the unit outputs the user's RA disease risk prediction result.
[0057] Specifically, users deploy data acquisition units in locations with limited testing equipment and insufficient medical personnel, such as community health centers, elderly care facilities, private hospitals, and secondary hospitals. These units collect numerical and image data of the users to be predicted, including clinical big data, musculoskeletal ultrasound, and joint imaging. The data is then preprocessed to eliminate the impact of missing values and imbalanced distribution of numerical data on the prediction results. Image data is also segmented and resized.
[0058] The grayscale and texture images of the preprocessed image data are then acquired, and the first and second image features are extracted respectively. Feature vectors are generated from the preprocessed numerical detection data, the first image features, and the second image features. The three feature vectors are fused to obtain the final feature vector. The final feature vector is then input into the rheumatoid arthritis disease prediction unit to obtain the prediction result for the user to be predicted. Based on this prediction result, primary care physicians can quickly screen out patients who may have rheumatoid arthritis, so as to conduct further special examinations for suspected patients and avoid delaying the treatment of patients.
[0059] This invention can be widely used and effectively screen for suspected rheumatoid arthritis patients, enabling primary care physicians to conduct targeted examinations and treatments, avoiding misdiagnosis and missed diagnosis. It also provides accurate treatment recommendations based on the user's disease risk prediction results, such as dietary supplementation, use of appropriate medications, or referrals, enabling precise treatment for patients and preventing the further development of rheumatoid arthritis.
[0060] The above description is merely a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for constructing a disease prediction model for rheumatoid arthritis, characterized in that, The application relates to a method for constructing a rheumatoid arthritis disease condition prediction model. The method comprises the following steps: acquiring a cross-modal data set; 2. The method of claim 1, wherein the method is for constructing a rheumatoid arthritis disease condition prediction model. training a multi-feature fusion network model based on the cross-modal data set. The acquiring of the cross-modal data set comprises the following steps: acquiring an original data set composed of multiple groups of original data, each group of original data comprising corresponding numerical detection data, image data and labels; generating feature vectors of the numerical detection data and the image data in each group of original data respectively; 3. The method of claim 2, wherein the method is for constructing a rheumatoid arthritis disease condition prediction model. fusing the feature vectors of the numerical detection data and the image data in each group of original data to obtain final feature vectors, and the multiple final feature vectors constitute the cross-modal data set.
4. The method of claim 3, wherein the method is characterized by, Before the generating of the feature vectors, the numerical detection data and the image data in the original data set are preprocessed.
5. The method of claim 3, wherein the method is characterized by: The numerical detection data in the original data is preprocessed, specifically by processing missing values and unbalanced data of the numerical detection data through a feature selection algorithm.
6. The method according to any one of claims 2-5, wherein, The image data in the original data is preprocessed, specifically by performing image segmentation and size readjustment on the ultrasonic image. The generating of the feature vectors comprises the following steps: acquiring a grayscale image and a texture image of the image data, and extracting first image features and second image features by using a convolutional neural network respectively; 7. The method of claim 1, wherein the method is for constructing a rheumatoid arthritis disease condition prediction model. generating feature vectors of the numerical detection data, the first image features and the second image features respectively. The training of the multi-feature fusion network model comprises the following steps: dividing the cross-modal data set into a training set and a test set, and training an initial network model by using the training set data; optimizing the initial network model by using the test set data to obtain the multi-feature fusion network model.
9. The use of the method of claim 8 for predicting the condition of rheumatoid arthritis, characterized in that, 8. The application of the method for constructing the rheumatoid arthritis disease condition prediction model according to any one of claims 1-7 in rheumatoid arthritis disease condition prediction. The application comprises the following steps: acquiring numerical detection data and image data of a user to be predicted, preprocessing the data and generating feature vectors, fusing the feature vectors to obtain final feature vectors; 10. The use of the method of claim 9 for predicting the condition of rheumatoid arthritis, characterized in that, constructing a rheumatoid arthritis disease condition prediction unit, and outputting a rheumatoid arthritis disease risk prediction result based on the final feature vectors. The rheumatoid arthritis disease condition prediction unit is provided with the multi-feature fusion network model.