A clinical data fusion method and acquisition and analysis system for temporomandibular joint disease

Abstract

This invention relates to a clinical data fusion method and acquisition and analysis system for temporomandibular joint disorder (TMD). The data fusion method is based on an unsupervised learning GCN network to learn from different types of acquired images. Through two stages of segmentation and data fusion, it achieves accurate classification results for different types of images. Based on this, diagnosis and treatment are performed on individual cases. At the same time, it combines a large database for auxiliary diagnosis and treatment. Compared with existing technologies, this invention achieves a process of feature classification from coarse to fine, and from similar to accurate, avoiding the low efficiency and misdiagnosis caused by relying solely on CT images in existing technologies. Furthermore, by combining a big data platform with clinical data, it has practical significance for improving the professional skills of medical personnel and expanding the coverage of TMD diagnosis and treatment.

Description

Technical Field

[0001] This invention relates to the field of intelligent diagnosis and treatment technology, specifically to a method for fusing clinical data on temporomandibular joint disorders and a data acquisition and analysis system. Technical Background

[0002] Temporomandibular joint disorder (TMD) has an incidence rate as high as 12%, often causing significant orofacial pain and mandibular movement disorders. It can even lead to organic damage to bilateral joints, muscles, and dentition, severely impacting patients' quality of life. In 2014, a new diagnostic system for temporomandibular joint disorder (DC / TMD) was officially published. After standardizing the examination methods, it provided specific criteria for each diagnosis and their corresponding validity. However, this system is only used to screen for risk factors of TMD and lacks a large-scale data background on TMD. It cannot provide auxiliary diagnostic results based on big data, resulting in low efficiency in the diagnosis, treatment, and rehabilitation of TMD, and even a certain percentage of misdiagnosis.

[0003] The temporomandibular joint (TMJ) department of a certain hospital has a large patient population and has achieved good clinical results through non-invasive or minimally invasive treatments. However, the treatment outcomes for these patients have not yet been fully analyzed. Given the lack of clinical evidence on temporomandibular joint disorders, the lack of systematic organization and analysis means that accumulated clinical experience cannot be effectively translated into clinical diagnostic conclusions, resulting in a low clinical treatment efficiency. Summary of the Invention

[0004] A method for fusing clinical data on temporomandibular joint disorders includes the following steps:

[0005] S11: Data Acquisition: Acquire images of the human face and joint scans of the temporomandibular joint;

[0006] The facial images include frontal and side profile images, and the joint scan images include frontal and side profile scan images;

[0007] S12: Image preprocessing: The facial image and joint scan image are imported into the image preprocessing unit for preprocessing. The preprocessing includes aligning the coordinate reference points of the facial image and joint scan image, normalizing the size of the facial image and joint scan image, and performing pixelation to obtain the front image data.

[0008] S12 also includes step S121, which is as follows:

[0009] The foreground image data is randomly grouped, with 80% of the dataset used as the training set and 20% used as the test set. The training set is used for foreground image segmentation training.

[0010] S13: Front image segmentation: Using the GCN network in conjunction with the DBscan algorithm, the front image is clustered. Skeletal tissue / facial structures that meet certain RGB true colors are clustered into pixel units. Pixel units that do not meet the preset conditions are removed, while pixel units that meet the preset conditions are retained, and finally the image segmentation block is output.

[0011] S14: Based on step S13, the image segmentation blocks in the facial image and the image segmentation blocks in the joint scan image are shared through a feature cross-layer, thereby integrating the image segmentation blocks in the facial image with the image segmentation block information in the joint scan image, and vice versa, thus achieving image data fusion.

[0012] S15: Secondary image segmentation: The image segmentation blocks of the fused facial image and the image segmentation blocks of the joint scan image are used as input to the second-level GCN network. In the second-level GCN network, the KNN algorithm is used for classification. Based on this, the previous image is classified through a classification layer to obtain the tissue structure classification result.

[0013] The second-level GCN network contains convolutional layers and pyramid pooling layers.

[0014] Each second-level GCN network is connected to the classification layer through a fully connected layer, thereby realizing the vectorization of the output image segmentation. There are multiple fully connected layers, and their output feature vectors all have a dimension of 1024.

[0015] S16: Based on step S15, the tissue structure classification results are imported into the data fusion module to achieve data fusion of human facial images and joint scan images.

[0016] Preferably, the present invention also proposes another method for fusing clinical data on temporomandibular joint disorders, the method comprising the following steps:

[0017] S21: Data Acquisition: Acquire human facial images through a camera, acquire temporomandibular joint scan images through a CT scanner or MRI scanner, and collect basic patient information, disease type, and disease description through an intelligent consultation device;

[0018] S22: Image preprocessing: The facial image and joint scan image are imported into the image preprocessing unit for preprocessing. The preprocessing includes aligning the coordinate reference points of the facial image and joint scan image, normalizing the size of the facial image and joint scan image, and performing pixelation to obtain the front image data.

[0019] S23: Image segmentation: Based on the GCN network and combined with the DBscan algorithm, the preceding image data is segmented, clustered and registered, and classified using the softmax function to obtain the tissue structure classification results;

[0020] S24: Data Fusion: Based on step S23, the tissue structure classification results obtained from image segmentation are imported into the data fusion module to achieve data fusion of human facial images and joint scan images;

[0021] S25: Based on data fusion, provide preliminary diagnostic results;

[0022] S26: Establish an online cloud database and import the data collected in S21 into the online cloud database for data matching and storage. Data matching includes, but is not limited to, disease type matching, disease description matching, and image classification matching. Based on this, calculate the matching similarity Pm of the current patient. Based on this, feed back the auxiliary diagnosis results to step S25, and at the same time, feed back the diagnosis results of "similar cases" within the confidence interval of the matching similarity Pm to step S25.

[0023] The above Pm = w1·P1 + w2·P2 + w3·P3 + ..., where P1, P2, and P3 are the similarity percentages of disease type matching, disease description matching, and image classification matching, respectively, and w1, w2, and w3 are the corresponding weight coefficients.

[0024] S27: Medical staff will synthesize the preliminary and auxiliary diagnostic results, output the final diagnostic result, and provide corresponding treatment suggestions, including: traditional Chinese medicine treatment, integrated traditional Chinese and Western medicine treatment, laser surgery treatment, injection treatment, radiotherapy, and prosthetic interventional treatment, etc.

[0025] Based on steps S21 to S27, step S28 is also included:

[0026] S28: Track the prognosis of treatment for different types of temporomandibular joint disorders. For successfully treated cases, include them in the successful case database and update the diagnostic results in the cloud database.

[0027] In summary, the advantages of the clinical data fusion method and data acquisition and analysis system for temporomandibular joint disorders of the present invention compared with the prior art are as follows:

[0028] (1) Based on the autonomous learning function of GCN network, feature cross-fusion of different types of data is performed, and GCN network and pyramid feature pool are used for two-stage segmentation, realizing the process of feature classification from coarse to fine, from similar to accurate, avoiding the low efficiency and easy misdiagnosis caused by judging only by CT images in the existing technology.

[0029] (2) The feature classification of the present invention can not only accurately separate structures such as condyle, condylar fossa, cochlea, and facial mass, but also calculate feature parameters such as condyle height and width based on neighborhood search algorithm and the vector nature of pixel block. These parameters help to determine whether the patient belongs to temporomandibular joint disease types such as osteophyte and osteosclerosis, which greatly improves the efficiency of diagnosis and treatment and avoids the difficulty of diagnosis and treatment by visual identification.

[0030] (3) This invention combines the auxiliary diagnosis of big data with the image learning-based diagnosis of individual cases, which not only improves the diagnostic accuracy of individual patients with temporomandibular joint disease and prevents misjudgment of the condition, but also enriches the big data, provides prospective follow-up studies, and combines the big data platform with clinical data, which is of practical significance for improving the professional skills of medical staff and expanding the coverage of temporomandibular joint disease diagnosis and treatment.

[0031] (4) By tracking the prognosis of different types of temporomandibular joint disease cases, successful cases are included in the successful case database and the diagnostic results in the cloud database are updated, which helps to improve the effectiveness and accuracy of cloud database-assisted diagnosis. Attached Figure Description

[0032] Figure 1 A schematic diagram of the structure and related parameters of the human temporomandibular joint;

[0033] Figure 2 A flowchart of image autonomous learning based on GCN network;

[0034] Figure 3 Flowchart of a clinical data acquisition and analysis system for temporomandibular joint disorders;

[0035] Figure 4 This is a diagram of a clinical data collection and analysis system for temporomandibular joint disorders. Detailed Implementation

[0036] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below in conjunction with specific embodiments and corresponding drawings. Obviously, the described embodiments are only some embodiments of this invention, and not all embodiments. Based on this, all other implementation methods obtained by those skilled in the art without creative effort are within the scope of protection of this invention.

[0037] Example 1

[0038] Please investigate. Figure 1 Temporomandibular joint disorders mainly occur at the bilateral joints, especially the condylar-condylar fossa. This area is adjacent to the cochlea and the root of the upper wisdom tooth, with nerves, muscles, and connective tissues interwoven, creating a complex structure. In clinical practice, medical staff first assess the patient's condition based on facial expressions and descriptions of their symptoms using either traditional Chinese medicine or Western medicine. If the condition is deemed mild and treatable with traditional Chinese medicine or Western medicine, a prescription is directly issued. However, if the condition cannot be directly diagnosed and requires diagnostic equipment, the patient is usually advised to undergo a CT scan. Based on the CT images, medical staff determine the position of the condylar-condylar joint (e.g., condylar width, height, and the size of the intra-articular space between the condyle and fossa). However, the condylar-condylar joint is relatively small, and the scan images often include shadows from other tissues, making visual identification difficult and diagnostic efficiency low. For inexperienced physicians, misdiagnosis is common.

[0039] Based on this, the present invention proposes a method for fusing clinical data of temporomandibular joint disorders. Please examine it. Figure 2 The method includes the following steps:

[0040] S11: Data Acquisition: Acquire human facial images (including frontal and side views) via camera, and acquire temporomandibular joint scan images (including frontal and side views) via CT or MRI.

[0041] S12: Image preprocessing: The facial image and joint scan image are imported into the image preprocessing unit for preprocessing. The preprocessing includes aligning the coordinate reference points of the facial image and joint scan image, normalizing the facial image and joint scan image into a 300*300*2 normalized image, and pixelating the normalized image to obtain the front image data.

[0042] S12 also includes step S121, which is as follows:

[0043] The foreground image data is randomly grouped, with 80% of the dataset used as the training set and 20% used as the test set. The training set is used for foreground image segmentation training.

[0044] S13: Front Image Segmentation: After image preprocessing, there is still a high rate of noise in the facial images or joint scan images in the front image data. Taking advantage of the unsupervised learning characteristics of the GCN network, and combined with the DBscan algorithm, the front image is clustered. The bone tissue / facial structure that meets certain RGB true colors is clustered into pixel units. After multiple trainings of the GCN network, the pixel units that do not meet the preset conditions are removed (such as background areas and shadow areas), while the pixel units that meet the preset conditions are retained, and the final output image segmentation block is output.

[0045] The GCN network includes at least convolutional layers and pyramid pooling layers. During the DBscan clustering process, training with multiple convolutional layers can not only enhance image information but also reduce noise. Through pyramid pooling layers, feature maps of arbitrary size are first divided into multiple blocks of different sizes, and then max pooling is performed on each block. The pooled features are concatenated and output with a fixed dimension to meet the requirements of feature cross layers and secondary image segmentation.

[0046] S14: Based on step S13, the image segmentation blocks in the facial image and the image segmentation blocks in the joint scan image are shared through parameter interaction via a feature cross-layer. This integrates the image segmentation block information from the facial image with the image segmentation block information from the joint scan image, thereby achieving image data fusion, improving the accuracy of feature classification, and providing a guarantee for accurately calculating the maximum thickness of the intra-articular space, the maximum thickness of the extra-articular space, the condyle width, and the condyle height, and accurately predicting whether a cyst has formed and its location.

[0047] S15: Secondary image segmentation: The image segmentation blocks obtained in step S13 are used as input to the second-level GCN network. In the second-level GCN network, the KNN algorithm is used for classification. The convolutional layers and pyramid pooling layers in the second-level GCN network can further subdivide the interactive image segmentation blocks. On this basis, the previous image is classified through a classification layer (e.g., the Softmax function). Then, the trained classification layer extracts and accurately classifies each pixel in the image segmentation block to obtain a more complete and accurate organizational structure classification result.

[0048] The tissue classification results for the human face include mass size, facial spasm characteristics, and maxillary and mandibular misalignment characteristics. The temporomandibular joint classification results include the condyle, condylar fossa, cochlea, teeth, and the maximum distance L between the condyle and condylar fossa. 1,max (i.e., the maximum thickness of the intra-articular space, calculated based on pixel size, number, and vectors based on position coordinates using the KNN algorithm for neighborhood search), and the maximum distance L between the condyle and condylar fossa.2,max (i.e., the maximum thickness of the extra-articular space, and L) 1,max (Same algorithm), condylar height H, condylar width W, condylar thickness Th.

[0049] Furthermore, it is worth noting that each second-level GCN network is connected to the classification layer through a fully connected layer, thereby realizing the vectorization of the output image segmentation. There are multiple fully connected layers, and their output feature vectors all have a dimension of 1024.

[0050] S16: Based on step S15, the tissue structure classification results are imported into the data fusion module to achieve data fusion of human facial images and joint scan images.

[0051] When a three-dimensional model is available, the location coordinates and feature vectors of the relevant categories in the tissue structure classification results are extracted and integrated into the temporomandibular joint model for model reconstruction. The key parts in the three-dimensional model are then dimensioned, which helps medical personnel to accurately determine the type of temporomandibular joint disease and make a preliminary diagnosis.

[0052] Example 2

[0053] Based on Example 1, the present invention further proposes a method for fusing clinical data of temporomandibular joint disorders, comprising the following steps:

[0054] S21: Data Acquisition: Acquire human facial images through a camera, acquire temporomandibular joint scan images through a CT scanner or MRI scanner, and collect basic patient information, disease type, and disease description through an intelligent consultation device;

[0055] S22: Image preprocessing: The facial image and joint scan image are imported into the image preprocessing unit for preprocessing. The preprocessing includes aligning the coordinate reference points of the facial image and joint scan image, normalizing the facial image and joint scan image into a 300*300*2 normalized image, and pixelating the normalized image to obtain the front image data.

[0056] S23: Image segmentation: Based on the GCN network and combined with the DBscan algorithm, the preceding image data is segmented, clustered and registered, and classified using the softmax function to obtain the tissue structure classification results;

[0057] S24: Data Fusion: Based on step S23, the tissue structure classification results obtained from image segmentation are imported into the data fusion module to achieve data fusion of human facial images and joint scan images;

[0058] S25: Based on data fusion, provide preliminary diagnostic results;

[0059] S26: Establish an online cloud database, import the data collected in S21 into the online cloud database for data matching and storage. Data matching includes, but is not limited to, disease type matching, disease description matching, and image classification matching. On this basis, calculate the case matching similarity Pm. On this basis, feed back the auxiliary diagnosis results to step S25. At the same time, feed back the diagnosis results of "similar cases" within the confidence interval of the matching similarity Pm to step S25.

[0060] The above Pm = w1·P1 + w2·P2 + w3·P3 + ..., where P1, P2, and P3 are the similarity percentages of disease type matching, disease description matching, and image classification matching, respectively, and w1, w2, and w3 are the corresponding weight coefficients.

[0061] S27: Medical personnel, based on the preliminary and auxiliary diagnostic results, output a final diagnosis and provide corresponding treatment recommendations. These recommendations may include: traditional Chinese medicine treatment, integrated traditional Chinese and Western medicine treatment, laser surgery, injection therapy, radiotherapy, and prosthetic interventional treatment, etc.

[0062] Online cloud databases include:

[0063] (1) Input of basic patient information, including patient name, gender, age, admission examination time, etc., and numbered according to the patient's correction card number and medical record name, and stored in the cloud database.

[0064] (2) The condition type of patients with temporomandibular joint disorder is entered. The condition types include: 1) patients with temporomandibular joint disorder for the first time; 2) patients with recurrent temporomandibular joint disorder (temporomandibular joint disorder has occurred before, and the symptoms have been relieved for more than one year before it recurs); 3) patients who have received temporomandibular joint disorder treatment within one year with poor results; 4) patients who have received intraoral treatment for more than 1 hour within one month before the onset of symptoms; 5) patients who have had temporomandibular joint implantation with poor practical results.

[0065] (3) Patient's condition description: 1) Soreness / swelling in the temporomandibular joint muscles; 2) Stinging sensation during chewing movements of the temporomandibular joint; 3) Difficulty swallowing, affecting normal eating; 4) Pain in the ear canal / cochlea during chewing movements of the temporomandibular joint; 5) Twitching / spasm of the masticatory muscles.

[0066] Patient condition description entry is achieved by collecting the patient's voice or text input.

[0067] (4) Image annotation and classification: Images include facial images and bone scan images. Bone scan images can be CT images or MRT images (MRI images). Specifically, all patients' facial images are saved as .jpeg or .png files, all patients' CT images are saved as .dicom files, and all patients' MRT images are saved as .dixom files. Furthermore, all bone scan images are annotated layer by layer with labels for maximum thickness of the intra-articular space, maximum thickness of the extra-articular space, condyle width, and condyle height type. In addition, for the same patient, their facial images, CT images, and MRT images are mutually mapped and associated. When medical staff need to access them, they can click on the patient's number to export the patient's facial images, CT images, and MRT images together.

[0068] (5) Diagnostic result classification: 1) Bone hyperplasia; 2) Bone sclerosis; 3) Bone fracture; 4) Cyst; 5) Influence of ear disease; 6) Other causes; The diagnostic result classification corresponds to the specific medical records, and the cloud database can provide the number of cases under a certain diagnostic result.

[0069] (6) Successful Case Database: The successful case database is associated with the above-mentioned diagnostic result classification. It contains details of successfully treated cases. In addition to basic patient information, disease type, disease description, and image annotation classification, the case details also include auxiliary treatment suggestions automatically provided by the cloud database and final treatment suggestions from medical staff. The successful case database helps medical staff access relevant cases in step S27. Even inexperienced physicians can easily operate it, avoiding misdiagnosis and accumulating experience.

[0070] Example 3

[0071] Please investigate. Figure 4 A clinical data acquisition and analysis system for temporomandibular joint disorders includes a data acquisition unit, a server, a cloud database, and a terminal diagnostic device (i.e., a PC). The data acquisition unit is electrically connected to the server via a communication network, and the server is electrically connected to the cloud database and the terminal diagnostic device via the communication network. Thus, the transmission and sharing of clinical data on temporomandibular joint disorders are realized through the server.

[0072] The data acquisition unit includes, but is not limited to, an intelligent consultation device (used to collect patients' voice data and text input data), a camera (used to collect patients' facial data to analyze the misalignment of the temporomandibular joint and whether there is tumor formation), and a PC or MRI machine (mainly used to scan and collect images of the patient's temporomandibular joint bone).

[0073] The cloud database includes at least patient medical record data, scan image library data, facial image library data, and a patient type database. Except for patient medical record data, the other three types of data can be categorized by patient type. Patient medical record data, scan image library data, facial image library data, and the patient type database can interact with each other.

[0074] The scope of cloud database data collection can be that of this hospital, a city, or even an entire province. When the cloud database covers a city or an entire province, any member hospital or other member medical service institution can access the data.

[0075] The cloud database stores a large amount of clinical data and medical records of temporomandibular joint disorders in a formatted manner, which has certain guiding significance for the diagnosis and treatment of individual cases. When the terminal diagnostic and treatment device needs to access the analysis and judgment results in the cloud database, it sends a request signal to the cloud database through the server. After receiving the request signal, the cloud database sends a "reference diagnosis result" to the terminal diagnostic and treatment device through the server and provides reference diagnosis and treatment suggestions based on big data analysis.

[0076] To ensure data quality and improve the accuracy of cloud database analysis, the intelligent consultation device in the data acquisition unit can be presented as a mobile app. Patients with relevant medical records can install the intelligent consultation app on their phones. This app wirelessly connects to the terminal diagnostic equipment, allowing medical staff to conduct regular follow-ups (e.g., every 3 or 6 months) by consulting with patients through the app. The data from these consultations is stored on a PC, and medical staff upload the results in a specific format to update patient records and patient type data in the cloud database. For incomplete or lost-to-follow-up cases, the discrepancies between the auxiliary diagnostic results in the cloud database and the final diagnosis made by medical staff should be checked, and the cloud database should be updated online in a timely manner to improve the effectiveness of cloud database-assisted diagnosis.

Claims

1. A method for fusing clinical data on temporomandibular joint disorders, characterized in that: The method includes the following steps: S11: Data Acquisition: Acquire images of the human face and joint scans of the temporomandibular joint; S12: Image preprocessing: The facial image and joint scan image are imported into the image preprocessing unit for preprocessing. The preprocessing includes aligning the coordinate reference points of the facial image and joint scan image, normalizing the size of the facial image and joint scan image, and performing pixelation to obtain the front image data. S12 also includes step S121, which is as follows: The foreground image data is randomly grouped, with 80% of the dataset used as the training set and 20% used as the test set. The training set is used for foreground image segmentation training. S13: Front image segmentation: Using the GCN network in conjunction with the DBscan algorithm, the front image is clustered. Skeletal tissue / facial structures that meet certain RGB true colors are clustered into pixel units. Pixel units that do not meet the preset conditions are removed, while pixel units that meet the preset conditions are retained, and finally the image segmentation block is output. S14: Based on step S13, the image segmentation blocks in the facial image and the image segmentation blocks in the joint scan image are shared through a feature cross-layer, so that the image segmentation blocks in the facial image are integrated with the image segmentation block information in the joint scan image, and vice versa.

2. The method for fusing clinical data of temporomandibular joint disorders according to claim 1, characterized in that: Following S14, steps S15 to S16 are also included: S15: Secondary image segmentation: The image segmentation blocks of the fused facial image and the image segmentation blocks of the joint scan image are used as input to the second-level GCN network. In the second-level GCN network, the KNN algorithm is used for classification. Based on this, the previous image is classified through a classification layer to obtain the tissue structure classification result. S16: Based on step S15, the tissue structure classification results are imported into the data fusion module to achieve data fusion of human facial images and joint scan images.

3. The method for fusing clinical data of temporomandibular joint disorders according to claim 1, characterized in that: The facial images include frontal and side profile images, and the joint scan images include frontal and side profile scan images.

4. The method for fusing clinical data of temporomandibular joint disorders according to claim 2, characterized in that: The second-level GCN network includes convolutional layers and pyramid pooling layers. Each second-level GCN network is connected to the classification layer through a fully connected layer, thereby realizing the vectorization of the output image segmentation. There are multiple fully connected layers, and their output feature vectors all have a dimension of 1024.

5. The method for fusing clinical data of temporomandibular joint disorders according to claim 2, characterized in that: The tissue classification results for the human face include mass size, facial spasm characteristics, and maxillary and mandibular misalignment characteristics. The temporomandibular joint classification results include the condyle, condylar fossa, cochlea, teeth, and the maximum distance L between the condyle and condylar fossa. 1,max The maximum distance L between the condyle and the condylar fossa 2,max Condylar height H, condylar width W, condylar thickness Th.

6. The method for fusing clinical data of temporomandibular joint disorders according to claim 2, characterized in that: L 1,max L 2,max All results are derived from neighborhood search using the KNN algorithm, based on pixel size, number of pixels, and vector calculations based on position coordinates.

7. A method for fusing clinical data on temporomandibular joint disorders, characterized in that: The method includes the following steps: S21: Data Acquisition: Acquire human facial images through a camera, acquire temporomandibular joint scan images through a CT scanner or MRI scanner, and collect basic patient information, disease type, and disease description through an intelligent consultation device; S22: Image preprocessing: The facial image and joint scan image are imported into the image preprocessing unit for preprocessing. The preprocessing includes aligning the coordinate reference points of the facial image and joint scan image, normalizing the size of the facial image and joint scan image, and performing pixelation to obtain the front image data. S23: Image segmentation: Based on the GCN network and combined with the DBscan algorithm, the preceding image data is segmented, clustered and registered, and classified using the softmax function to obtain the tissue structure classification results; S24: Data Fusion: Based on step S23, the tissue structure classification results obtained from image segmentation are imported into the data fusion module to achieve data fusion of human facial images and joint scan images; S25: Based on data fusion, provide preliminary diagnostic results; S26: Establish an online cloud database, import the data collected in S21 into the online cloud database for data matching and storage, calculate the matching similarity Pm of the current patient, and also feed back the diagnosis results of "similar cases" within the confidence interval of the matching similarity Pm to step S25.

8. The method for fusing clinical data of temporomandibular joint disorders according to claim 7, characterized in that: The method also includes the calculation of matching similarity Pm, Pm = w1·P1 + w2·P2 + w3·P3 + ..., where P1, P2, and P3 are respectively the similarity percentages of disease type matching, disease description matching, and image classification matching, and w1, w2, and w3 are the corresponding weight coefficients.

9. A method for fusing clinical data of temporomandibular joint disorders according to claim 7, characterized in that: Step S26 is followed by steps S27 and S28: S27: Medical personnel will synthesize the preliminary and auxiliary diagnostic results, output the final diagnostic result, and provide corresponding treatment suggestions, including but not limited to traditional Chinese medicine treatment, integrated traditional Chinese and Western medicine treatment, laser surgery treatment, injection treatment, radiotherapy, and prosthesis interventional treatment; S28: Track the prognosis of treatment for different types of temporomandibular joint disorders. For successfully treated cases, include them in the successful case database and update the diagnostic results in the cloud database.

10. A clinical data acquisition and analysis system for temporomandibular joint disorders, used in the clinical data fusion method for temporomandibular joint disorders as described in any one of claims 7-9, characterized in that: The analysis system includes a data acquisition unit, a server, a cloud database, and end-point diagnostic and treatment equipment. The data acquisition unit is electrically connected to the server via a communication network, and the server is electrically connected to the cloud database and the end-point diagnostic and treatment equipment via the communication network.

Images