Temporomandibular Joint Disc Bicircular Measurement Model Recognition System

Abstract

The present invention discloses a double-circle measurement model recognition system for the temporomandibular joint disc. The present invention relates to the technical field of medical recognition systems, and includes an image preprocessing and annotation module, a key point detection network module constructed based on a deep learning network, a double-circle model construction module, and a displacement parameter calculation module. The present invention proposes a double-circle geometric model, locates the center of the circle by the method of the intersection of the normal lines of three tangent points, automatically detects key points in combination with an optimized ResNet-34 deep learning network, and quantifies the displacement direction based on the included angle of the angle bisector and the intersection distance. Evaluating the severity with geometric parameters breaks through single linear measurement and realizes multi-dimensional morphological analysis. Compared with traditional methods, it improves the measurement accuracy and reduces the measurement time, providing a high-precision and automated diagnostic standard for temporomandibular joint disc displacement.

Description

Technical Field

[0001] The present invention relates to the technical field of medical identification systems, and more particularly to a double-circle measurement model identification system for the temporomandibular joint disc. Background Art

[0002] The temporomandibular joint is the joint connecting the mandible and the skull, and plays a crucial role in various functions such as chewing, speaking, breathing, and swallowing in humans. Temporomandibular joint dysfunction is a common clinical disease, manifested as joint pain, movement disorders, noise, and loss of joint function. Lateral displacement of the temporomandibular joint, both internally and externally, is one of the common symptoms in temporomandibular joint dysfunction, usually caused by improper positioning of the joint disc or damage to the joint capsule and ligaments. This displacement not only affects the chewing function of patients, but may also cause long-term pain and discomfort, affecting the quality of life of patients. Therefore, accurate measurement and diagnosis of the degree of lateral displacement of the temporomandibular joint are of great importance.

[0003] However, the current assessment of lateral displacement of the temporomandibular joint often relies on a variety of imaging techniques, such as measuring through MRI. Traditional measurement methods measure the e value, which is the internal displacement amount, and the internal displacement amount is the horizontal distance between the inner edge of the joint disc and the outermost edge of the medial condyle. The f value, which is the lateral displacement amount, and the lateral displacement amount is the horizontal distance between the outer edge of the joint disc and the outermost edge of the lateral condyle, and the e-f value, which is the internal and external difference value. The internal and external displacement characteristics of the joint disc are analyzed through these three key distance values of the e value, f value, and e-f value. However, this manual measurement method highly depends on the manual operation of imaging software. Doctors mark the edges of the joint disc and condyle according to the images. Measuring the e value and f value is not only time-consuming, but also the accuracy depends on the skills and experience of the operator. Especially in the case of unclear image edges, there may be significant differences in the measurement results between different doctors, leading to enhanced subjectivity in diagnosis and also increasing the error of repeated measurements.

[0004] In addition, traditional measurement methods only evaluate the displacement characteristics of the joint disc based on linear distances, that is, the e value, f value, and e-f value, and it is difficult to fully capture the subtle changes in the position offset of the joint disc. This method ignores the complex morphology of the joint disc in the condylar structure and cannot analyze the offset of the joint disc from more comprehensive geometric features, restricting the measurement accuracy and the effectiveness of clinical diagnosis. Therefore, it is necessary to explore a new method that can simplify the measurement process, improve the measurement accuracy, and can more intuitively reflect the displacement direction and severity. Summary of the Invention

[0005] The purpose of the present invention is to provide a double-circle measurement model identification system for the temporomandibular joint disc, which solves the problems raised in the above background art.

[0006] To achieve the above object, the present invention provides the following technical solutions: a double-circle measurement model recognition system for the temporomandibular joint disc, comprising:

[0007] An image preprocessing and annotation module, configured to obtain the MRI image of the patient's temporomandibular joint, then perform standardization processing and edge enhancement processing on the MRI image, and perform regional cropping and scaling operations on the MRI image;

[0008] A key point detection network module constructed based on a deep learning network, configured to input the preprocessed MRI image, then optimize the network structure, and perform annotation of key points, and finally perform model training;

[0009] The key points include: the center of the large circle on the condylar head, the center of the small circle on the condylar neck, the left boundary point of the articular disc, the right boundary point of the articular disc, the left boundary point of the condyle, and the right boundary point of the condyle;

[0010] A double-circle model construction module, constructing a large circle model based on the condylar head and a small circle model based on the condylar neck, and calculating the center and radius of the large circle model and the small circle model;

[0011] A displacement parameter calculation module, generating a condylar angle bisector and an articular disc angle bisector based on the key points, and performing intersection positioning based on the relationship between the condylar angle bisector and the articular disc angle bisector and the condyle and the articular disc, outputting intersection point A and intersection point B, and then measuring the included angle α of the angle bisectors and the distance between intersection point A and intersection point B to visually display the displacement condition of the articular disc.

[0012] Optionally, the specific process of the image preprocessing and annotation module is as follows:

[0013] Standardization processing: unifying the brightness and contrast of the MRI image through the CLAHE algorithm, and using Gaussian filtering to eliminate the noise of the MRI image;

[0014] Edge enhancement processing: extracting and highlighting the boundaries of the condylar head, condylar neck and articular disc through the Sobel operator;

[0015] Regional cropping and scaling operation: cropping the temporomandibular joint area in the MRI image and scaling it to a unified resolution to ensure the size consistency of the input to the deep learning network.

[0016] Optionally, the optimization of the network structure in the key point detection network module constructed based on the deep learning network includes: 1×1 convolution substitution and semi-residual module;

[0017] 1×1 Convolution Substitution: In the deep learning network, 1x1 convolutional kernels are introduced in the deeper layers, i.e., closer to the output layer, while retaining the 3x3 convolutional kernels in the shallow layers to reduce the computational complexity and fully extract local detailed features to ensure effective modeling of key points;

[0018] Semi-Residual Module: To enhance the feature expression ability of the deep learning network, the network structure is designed in combination with the semi-residual module. The semi-residual module introduces learnable weighting coefficients in the traditional residual structure and dynamically adjusts the fusion ratio of the residual features and the input features through the following formula:

[0019] Where: Refers to the weighted fusion result of multi-branch features; Refers to the features after being processed by the convolutional layer; Refers to the input features;

[0020] All refer to learning parameters for adaptively adjusting the feature fusion method;

[0021] The optimized deep learning network gives full play to the advantages of shallow and deep convolutional features and enhances the feature fusion ability through the semi-residual module.

[0022] Optionally, the model training in the key point detection network module constructed based on the deep learning network is specifically as follows:

[0023] Pair the labeled MRI images with the key point coordinates to form a training dataset for model training. During the model training process, perform augmentation operations such as rotation and scaling on the images to increase data diversity and make the model adapt to different MRI image angles and qualities;

[0024] Use the mean squared error loss function to measure the error between the key point coordinates predicted by the deep learning network and the true coordinates. The calculation formula is as follows:

[0025] Where:

[0026] Loss refers to the error between the predicted key point coordinates and the true coordinates, that is, the loss value;

[0027] N Refers to the total number of samples, the total number of coordinate points in all training data;

[0028] ( x i ，y i Refers to the true coordinate value in the i-th sample; Refers to the x and yPredicted value of coordinates

[0029] Evaluate the model performance on the test set, calculate the error between the predicted key points and the true key points according to the formula, and adjust the model parameters or perform hyperparameter tuning based on the test results to improve the model's performance in key point prediction.

[0030] Optionally, the purpose of the double-circle model construction module is to determine the centers and radii of the large and small circles through the method of intersecting the normal lines at three tangent points. The step process is as follows: First, draw a large circle inside the condylar head so that it is tangent to the upper surface and the inner and outer sides of the condylar head. Then, draw a small circle at the condylar neck so that it is tangent to the inner and outer edges of the condylar neck and the lower boundary of the large circle, forming the standard geometric shape of the joint structure.

[0031] The process of constructing the large circle model based on the condylar head is as follows:

[0032] Draw the normal lines at three tangent points of the condylar structure respectively, analyze the intersection relationship of these three tangent normal lines, and calculate the coordinates and radius of the center of the circle.

[0033] If the three tangent normal lines intersect pairwise;

[0034] The intersection points of the three tangent normal lines intersecting pairwise are respectively

[0035] Construct the centroid of the triangle as the center of the circle

[0036] Position; then the radius of the large circle model based on the condylar head is to calculate the distance from the center of the large circle when the three tangent normal lines intersect pairwise to the three tangent normal lines as the radius of the circle;

[0037] If the three tangent normal lines do not intersect and are parallel; first, the tangent normal line one L1:

[0038] and the tangent normal line two L2:

[0039] are parallel. Then, obtain the intersection points of the third tangent normal line L3 with the tangent normal line one L1 and the tangent normal line two L2 respectively, and then get two effective intersection points

[0040] Then use the midpoint of these two intersection points as the substitute center of the large circle After that, combine the remaining non-parallel tangent normal line three L3, and use its tangent point

[0041] with to weightedly find the centroid together, further correct the position of the center of the circle, improve the stability of the center of the circle positioning, and obtain the center of the large circle when the three tangent normal lines do not intersect

[0042] ; Similarly, when the three tangent normal lines do not intersect, the center of the large circle , calculate the distances from the center of the circle to the three tangent normal lines and use the median method to determine the radius of the circle. For the equation of the tangent normal line Li

[0043] Center of the circle The distance from the center of the circle to the tangent normal line Li is Take the median of the distances of the three tangent normal lines as the radius.

[0044] Optionally, the specific process of constructing the small circle model based on the condylar neck in the double - circle model construction module is as follows:

[0045] Similarly to the construction of the large circle model based on the condylar head, the construction of the small circle model based on the condylar neck is to draw normal lines at three tangent points of the condylar neck structure, analyze the mutual relationship of these three normal lines, and calculate the coordinates and radius of the center of the circle under the intersection relationship of the tangent normal lines and under the parallel relationship of the tangent normal lines respectively.

[0046] Optionally, the generation of the condylar angle bisector and the articular disc angle bisector based on the key points in the displacement parameter calculation module is specifically as follows:

[0047] Connect the center of the small circle with the medial and lateral convex points of the temporomandibular joint disc and the medial and lateral convex points of the condyle and make angle bisectors. The medial and lateral convex points of the temporomandibular joint disc correspond to the left boundary point and the right boundary point of the joint disc, and the medial and lateral convex points of the condyle correspond to the left boundary point and the right boundary point of the condyle.

[0048] Optionally, the included angle α of the angle bisectors is measured in the displacement parameter calculation module to reflect the severity of the displacement of the temporomandibular joint disc, and the displacement direction of the temporomandibular joint disc is reflected by analyzing the distance between the intersection point A and the intersection point B.

[0049] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0050] First, the present invention constructs a double - circle geometric model of the large circle of the condylar head and the small circle of the condylar neck, transforms the traditional measurement into a quantitative analysis of the geometric shape. The double - circle model can stably and efficiently determine the center of the circle and the radius through the three - tangent - normal - intersection method, and uses the included angle of the angle bisectors and the distance between the intersection points to intuitively reflect the displacement direction and severity of the articular disc. The present invention not only reduces the dependence on multiple variables in traditional measurements, but also reduces the subjective error of manual marking through a standardized model, improving the repeatability and consistency of the measurement.

[0051] Second, in view of the difficulty in locating the tangent point caused by the blurred boundary of MRI images, the present invention introduces an optimized ResNet-34 deep learning network, combines the optimization of the semi-residual module and 1x1 convolutional kernel, realizes the automatic detection of key points such as the condyle boundary, the articular disc boundary, and the centers of the two circles, and the model can efficiently extract local detail features, overcomes the dependence on the operator's experience in traditional manual marking, and solves the problem of inaccurate key point positioning caused by image blurring.

[0052] Third, the present invention connects the center of the small circle with the convex points on the inner and outer sides of the articular disc and the convex points on the inner and outer sides of the condyle, generates the angle bisector and calculates the included angle α and the intersection distance AB, quantifies the severity of the displacement with the α angle value and AB, transforms the complex anatomical structure relationship into geometric parameters, and makes the diagnostic result more intuitive and convenient for clinicians to quickly judge the condition. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Figure 1 It is the corresponding structure diagram of the temporomandibular joint disc image of the present invention;

[0054] Figure 2 It is the schematic diagram for constructing and determining the center and radius of the large circle when the normal lines of the tangent points intersect in the present invention;

[0055] Figure 3 It is the schematic diagram for constructing and determining the center and radius of the large circle when the normal lines of the tangent points do not intersect in the present invention;

[0056] Figure 4 It is the schematic diagram for determining the center and radius of the small circle when the normal lines of the tangent points intersect in the present invention;

[0057] Figure 5 It is the schematic diagram for constructing and determining the center and radius of the small circle when the normal lines of the tangent points do not intersect in the present invention;

[0058] Figure 6 It is the step flow chart of the double-circle measurement method for the temporomandibular joint disc displacement of the present invention;

[0059] Figure 7 It is the optimized network diagram of the deep learning network cooperating with the semi-residual module of the present invention;

[0060] Figure 8 It is the overall structure schematic diagram of the double-circle measurement model recognition system for the temporomandibular joint disc of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0061] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0062] Example 1. Please refer to Figures 1 to 8 , this embodiment provides a technical solution: a double-circle measurement model recognition system for the temporomandibular joint disc, including:

[0063] An image preprocessing and annotation module, which is used to obtain the MRI image of the patient's temporomandibular joint, then perform standardization processing and edge enhancement processing on the MRI image, and perform regional cropping and scaling operations on the MRI image;

[0064] A key point detection network module constructed based on a deep learning network, which is used to input the preprocessed MRI image, then optimize the network structure, annotate the key points, and finally perform model training;

[0065] The key points include: the center of the large circle on the condylar head, the center of the small circle on the condylar neck, the left boundary point of the articular disc, the right boundary point of the articular disc, the left boundary point of the condyle, and the right boundary point of the condyle;

[0066] A double-circle model construction module, which constructs a large circle model based on the condylar head and a small circle model based on the condylar neck, and calculates the center and radius of the large circle model and the small circle model;

[0067] A displacement parameter calculation module, which generates the condylar angle bisector and the articular disc angle bisector based on the key points, locates the intersection points based on the relationship between the condylar angle bisector and the articular disc angle bisector and the condyle and the articular disc, outputs the intersection points A and B, and then measures the included angle α of the angle bisectors and the distance between the intersection points A and B to visually display the displacement of the articular disc.

[0068] In this embodiment: The present invention is different from the prior art. In the traditional technology, only linear distance measurement may be relied on, ignoring the geometric relationship between the articular disc and the condyle, and unable to capture subtle morphological changes. The present invention converts the three-dimensional structures of the condylar head and neck into standardized geometric forms through a double-circle model, and combines angle bisector analysis, which can more comprehensively reflect the offset characteristics of the articular disc, such as rotation and tilt, breaking through the limitations of traditional linear measurement. In the prior art, doctors may need to manually mark the image boundary points, with low measurement efficiency and large errors. The present invention realizes the automatic detection of key points through an improved ResNet-34 deep learning network, combines a semi-residual module and a dynamic feature fusion technology, significantly improves the model's recognition ability for fuzzy boundaries, makes the measurement process automated and standardized, and reduces errors caused by manual intervention. In the prior art, the displacement may be evaluated through the e-f difference, while the present invention uses two parameters, the included angle α to reflect the displacement angle and the intersection distance AB to reflect the displacement distance, combined with positive and negative signs, to achieve multi-dimensional quantification of the displacement direction, angle, and distance, providing a richer decision-making basis for clinical diagnosis.

[0069] Please refer to Figures 1 to 8, the specific process of the image preprocessing and annotation module is as follows:

[0070] Normalization processing: The brightness and contrast of the MRI image are unified through the CLAHE algorithm, and Gaussian filtering is used to eliminate the noise in the MRI image;

[0071] Edge enhancement processing: The boundaries of the condylar head, condylar neck, and articular disc are extracted and highlighted through the Sobel operator;

[0072] Region cropping and scaling operations: The temporomandibular joint region in the MRI image is cropped and scaled to a unified resolution to ensure the dimensional consistency of the input to the deep learning network.

[0073] In this embodiment: Through this image preprocessing and annotation module, the MRI image can be efficiently processed in advance, so as to meet the use of the subsequent deep learning network, and improve the accuracy and efficiency of the overall recognition system.

[0074] Please refer to Figures 1 to 8 , the optimization of the network structure in the key point detection network module based on the deep learning network includes: 1×1 convolution substitution and semi-residual module;

[0075] 1×1 convolution substitution: In the deep learning network, 1x1 convolutional kernels are introduced in the deeper layers, that is, closer to the output layer, and the 3x3 convolutional kernels in the shallow layers are retained, so as to reduce the computational complexity while fully extracting local detail features and ensuring effective modeling of key points;

[0076] Semi-residual module: To enhance the feature expression ability of the deep learning network, the network structure is designed in combination with the semi-residual module. The semi-residual module introduces learnable weighting coefficients in the traditional residual structure, and dynamically adjusts the fusion ratio of the residual features and the input features through the following formula form:

[0077] Among them: Refers to the weighted fusion result of multi-branch features; Refers to the features after being processed by the convolutional layer; Refers to the input features;

[0078] All refer to learning parameters, which are used to adaptively adjust the feature fusion method; The optimized deep learning network ResNet-34 gives full play to the advantages of shallow and deep convolutional features, and enhances the feature fusion ability through the semi-residual module, thereby significantly improving the training efficiency and the generalization performance of the model, providing strong support for efficient and accurate key point detection.

[0079] In this embodiment: The core of the present invention in constructing a double-circle model using the three-tangent normal intersection method lies in finding the three tangent points to locate the large circle, and then the three tangent points to locate the small circle. Considering that the limited clarity of the boundary of the actual MRI image makes it difficult to find the boundary tangent points and thus unable to accurately determine the center positions of the large and small circles, the present invention uses a ResNET-34 optimized deep learning network to assist in measuring the positions of key points;

[0080] Based on the relationships among the centers of the large circle, the center of the small circle, the zygomatic process boundary points, and the articular disc boundary points, the present invention uses a deep learning network, ResNET-34, to optimize the deep learning network algorithm to construct a model to identify key points on the articular disc, and then improve the positions of the centers of the large and small circles.

[0081] Based on the deep learning network, that is, ResNet-34, a key point detection network is constructed. In the residual module of ResNet, a semi-residual module is introduced, and a learnable weighting coefficient is used to dynamically adjust the fusion ratio of the residual features and the input features. The output of the model is the coordinates of multiple key points. A regression layer is added to the last layer of ResNet to directly output the coordinate values (x, y) of the key points. The goal is to predict the center coordinates of the large and small circles and the coordinates of other boundary points;

[0082] In view of the difficulty in locating tangent points caused by the blurred boundary of the MRI image, the present invention introduces an optimized ResNet-34 deep learning network, combined with the optimization of the semi-residual module and 1x1 convolutional kernels, to achieve the automatic detection of key points such as the condylar boundary, the articular disc boundary, and the centers of the two circles. By dynamically adjusting the fusion ratio of the residual features and the input features, the model can adaptively extract local detail features, overcome the dependence on the operator's experience in traditional manual marking, and solve the problem of inaccurate key point positioning caused by image blurring.

[0083] Please refer to Figures 1 to 8 , the specific model training in the key point detection network module constructed based on the deep learning network is as follows:

[0084] Pair the labeled MRI images with the key point coordinates to form a training dataset for model training. During the model training process, perform enhancement operations of rotating and scaling the images to increase data diversity and make the model adapt to different MRI image angles and qualities;

[0085] The mean square error loss function is used to measure the error between the key point coordinates predicted by the deep learning network and the true coordinates. The calculation formula is as follows:

[0086] Where:

[0087] Loss refers to the error between the predicted key point coordinates and the true coordinates, that is, the loss value;

[0088] N Refers to the total number of samples, i.e., the total number of coordinate points in all training data;

[0089] ( x i ，y i ) Refers to the true coordinate value in the i-th sample; Refers to the x and y predicted value of the coordinates of the i-th sample; Evaluate the model performance on the test set, calculate the error between the predicted key points and the true key points according to the formula, and adjust the model parameters or perform hyperparameter tuning based on the test results to improve the performance of the model in key point prediction.

[0090] Mark the positions of several key points, namely the center of the large circle, the center of the small circle, the left boundary point of the articular disc, the right boundary point of the articular disc, the left boundary point of the condyle, and the right boundary point of the condyle, on the preprocessed MRI images as the training labels of the ResNet model.

[0091] The purpose of the double-circle model construction module is to determine the centers and radii of the large and small circles by the method of intersecting the normal lines at three tangent points. The step process is as follows: First, draw a large circle inside the condyle head so that it is tangent to the upper surface and the inner and outer sides of the condyle head. Then, draw a small circle at the condyle neck so that it is tangent to the inner and outer edges of the condyle neck and the lower boundary of the large circle, forming the standard geometric shape of the joint structure.

[0092] The process of constructing the large circle model based on the condyle head is as follows:

[0093] Draw the normal lines at three tangent points of the condyle structure respectively, analyze the intersection relationship of these three normal lines, and calculate the coordinates and radius of the center of the circle;

[0094] Three tangent points:

[0095] Upper tangent point, where the highest point of the condyle head is tangent to the upper surface;

[0096] Inner tangent point: The most convex point on the inner side of the condyle head is tangent to the inner edge;

[0097] Outer tangent point: The most convex point on the outer side of the condyle head is tangent to the outer edge;

[0098] Draw three tangent normal lines based on the three tangent points;

[0099] If the three tangent normal lines intersect pairwise, as Figure 3 shown;

[0100] The intersection points of the three tangent normal lines intersecting pairwise are respectively

[0101] Construct the centroid of the triangle as the center of the circle

[0102] If it is the position, the calculation formula for the center of the circle based on the large circle model of the condylar head is as follows:

[0103] Then, the radius of the large circle model based on the condylar head is to calculate the distance from the center of the large circle when the normal lines of the three tangent points intersect pairwise to the three normal lines of the tangent points as the radius of the circle. The specific calculation process and formula are as follows: For the calculation equation of the normal line Li of the tangent point is

[0104] The distance from the center of the large circle to the normal line Li of the tangent point is , then The calculation formula of is as follows:

[0105] Where:

[0106] Refers to the center of the circle to the distance of the normal line Li of the tangent point; Both refer to the equation coefficients of the normal line Li of the tangent point. In the calculation formula of the normal line Li of the tangent point, Both refer to the direction coefficients of the normal line Li of the tangent point, which are used to determine the slope of the normal line Li of the tangent point, while Refers to the intercept of the normal line Li of the tangent point, which is used to determine the position of the normal line Li of the tangent point;

[0107] Furthermore, take the median value of the distances of the three normal lines of the tangent points as the radius, and the calculation formula is as follows:

[0108] Where:

[0109] R refers to the radius of the large circle model when the three normal lines of the tangent points intersect pairwise;

[0110] Median refers to taking the middle value in ( d1, d2, d3 );

[0111] If the three normal lines of the tangent points do not intersect and are parallel;

[0112] First, the normal line of the tangent point one L1:

[0113] and the normal line of the tangent point two L2:

[0114] are parallel. Then, obtain the intersections of the third normal line of the tangent point L3 with the normal line of the tangent point one L1 and the normal line of the tangent point two L2 respectively, and then obtain two effective intersections

[0115] The calculation formula is as follows:

[0116] Then, use the midpoint of these two intersection points as the center of the substitute large circle.

[0117] The calculation formula is as follows: After that, combine the remaining non - parallel tangent normal line L3, and use its tangent point and

[0118] to calculate the centroid by weighted average together, further correct the position of the center of the circle, improve the stability of the center - of - circle positioning, and obtain the center of the large circle when the three tangent normal lines do not intersect. , and the calculation formula is as follows:

[0119] Similarly, according to the center of the large circle when the three tangent normal lines do not intersect , calculate the distances from the center of the circle to the three tangent normal lines and use the median method to determine the radius of the circle. For the equation of the tangent normal line Li the center of the circle the distance to the tangent normal line Li is :

[0120] Take the median of the distances of the three tangent normal lines as the radius, and the calculation formula is as follows; .

[0121] In this embodiment: construct a double - circle measurement model of the temporomandibular joint disc in the MRI image: First, draw a circle inside the condylar head, which is tangent to the upper surface and the inner and outer sides of the condylar head. Then, draw a small circle on the condylar neck, which is tangent to the inner and outer edges of the condylar neck and the lower boundary of the large circle, forming a standard geometric shape of the joint structure.

[0122] In the present invention, there are certain difficulties in the actual implementation process for the large circle and the small circle to intersect with the three tangent points on the condylar head and the neck respectively. Therefore, the present invention introduces the method of intersecting the three tangent normal lines to determine the centers and radii of the large and small circles.

[0123] In the present invention, find the three tangent points where the large circle is tangent to the condylar head, mark the normal lines respectively, analyze the mutual relationship between the three normal lines, calculate the position of the center of the circle forming the large - circle model according to the formula, and calculate the distances from the center of the circle to the three tangent normal lines, and take the median as the radius to form the large circle inside the condylar head.

[0124] Based on the construction of the large - circle model, similarly determine the center and radius of the small circle. Find the two tangent points where the small circle is tangent to the condylar neck and the tangent point where the small circle is tangent to the large circle, draw the normal lines respectively, analyze the mutual relationship between the three normal lines, calculate the position of the center of the circle forming the small - circle model according to the formula, and calculate the distances from the center of the circle to the three tangent normal lines, and take the median as the radius to form the small circle on the condylar neck.

[0125] Please refer to Figures 1 to 8, the specific process of constructing the small circle model based on the condylar neck in the double circle model construction module is as follows:

[0126] Similarly, for the construction of the large circle model based on the condylar head, in this construction of the small circle model based on the condylar neck, normal lines are drawn at three tangent points of the condylar neck structure, the mutual relationship of these three normal lines is analyzed, and the coordinates and radii of the center of the circle are calculated respectively under the condition of the intersection of the normal lines at the tangent points and the condition of the parallel relationship of the normal lines at the tangent points.

[0127] In this embodiment: through the double circle model construction module, the medial displacement and lateral displacement of the temporomandibular joint disc are analyzed and judged. According to the positional relationship of the key points, the displacement direction is represented by positive and negative signs. The medial displacement of the joint disc is recorded as -, and the lateral displacement is recorded as +. The medial and lateral offset conditions of the joint disc are marked, and the severity of the displacement is reflected by measuring geometric parameters such as the included angle and distance.

[0128] This geometric method through the double circle model construction module not only reduces the dependence on multiple variables in traditional measurements, but also reduces the subjective error of manual operations through a standardized model, significantly improving the repeatability and consistency of measurements.

[0129] Embodiment 2: On the basis of the above embodiment:

[0130] Please refer to Figures 1 to 8 , the generation of the condylar angle bisector and the joint disc angle bisector based on the key points in the displacement parameter calculation module is specifically as follows:

[0131] Connect the center of the small circle with the medial and lateral convex points of the temporomandibular joint disc and the medial and lateral convex points of the condyle and make angle bisectors. The medial and lateral convex points of the temporomandibular joint disc correspond to the left boundary point and the right boundary point of the joint disc, and the medial and lateral convex points of the condyle correspond to the left boundary point and the right boundary point of the condyle;

[0132] In the displacement parameter calculation module, the included angle α of the angle bisectors is measured to reflect the severity of the displacement of the temporomandibular joint disc, and the displacement direction of the temporomandibular joint disc is reflected by analyzing the distance between the intersection point A and the intersection point B.

[0133] In this embodiment: The specific illustration is Figure 7 the fifth step in. Specifically, the included angle α of the angle bisectors can be calculated using the cosine theorem, and the distance between the intersection point A and the intersection point B can be calculated using the Euclidean distance calculation formula. Thus, the included angle α of the angle bisectors and the distance between the intersection point A and the intersection point B can be quickly output. Through this displacement parameter calculation module of the present invention, the complex anatomical structure relationship can be transformed into geometric parameters, making the diagnostic result more intuitive and facilitating clinicians to quickly judge the condition.

[0134] Although embodiments of the present invention have been shown and described, those of ordinary skill in the art will appreciate that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. The double-circle measurement model recognition system for the temporomandibular joint disc is characterized by: include: The image preprocessing and annotation module is used to obtain the MRI image of the patient's temporomandibular joint, and then perform standardization and edge enhancement on the MRI image, and perform regional cropping and scaling operations on the MRI image; A key point detection network module is built based on a deep learning network to input preprocessed MRI images, then optimize the network structure, annotate key points, and finally train the model; The key points include: the center of the large circle on the condylar head, the center of the small circle on the condylar neck, the left boundary point of the articular disc, the right boundary point of the articular disc, the left boundary point of the condyle, and the right boundary point of the condyle; A double circle model building module is used to build a large circle model based on the condylar head and a small circle model based on the condylar neck, and calculate the center and radius of the large circle model and the small circle model; A displacement parameter calculation module generates a condylar angle bisector and an articular disc angle bisector based on the key points, locates the intersection based on the relationship between the condylar angle bisector and the articular disc angle bisector and the condyle and the articular disc, outputs intersection A and intersection B, and then measures the angle α of the angle bisector and the distance between intersection A and intersection B to intuitively display the displacement of the articular disc; The specific process of the image preprocessing and annotation module is as follows: Standardization processing: The brightness and contrast of MRI images are unified through the CLAHE algorithm, and Gaussian filtering is used to eliminate the noise of MRI images; Edge enhancement processing: The boundaries of the condylar head, condylar neck and articular disc are extracted and highlighted by using the Sobel operator; Region cropping and scaling operations: The temporomandibular joint region in the MRI image is cropped and scaled to a uniform resolution to ensure the size consistency of the input deep learning network; The optimization of the network structure in the key point detection network module based on the deep learning network includes: 1×1 convolution replacement and semi-residual module; 1×1 convolution replacement: In the deep learning network, 1x1 convolution kernels are introduced in the deeper layers, that is, near the output layer, and the 3x3 convolution kernels in the shallow layer are retained to reduce the computational complexity while fully extracting local detail features to ensure effective modeling of key points; Semi-residual module: To enhance the feature expression capability of deep learning networks, the network structure is designed in combination with the semi-residual module. The semi-residual module introduces a learnable weighting coefficient into the traditional residual structure, and dynamically adjusts the fusion ratio of residual features and input features through the following formula: OutPut = α*F(x) + β*x; in: OutPut refers to the weighted fusion result of multi-branch features; F(x) refers to the features after processing by the convolution layer; x refers to the input feature; α and β both refer to learning parameters, which are used to adaptively adjust the feature fusion method; The optimized deep learning network gives full play to the advantages of shallow and deep convolutional features, and enhances the feature fusion capability through the semi-residual module.

2. The temporomandibular joint disc double circle measurement model recognition system according to claim 1, characterized in that: The model training in the key point detection network module based on the deep learning network is specifically as follows: Pair the annotated MRI images with the key point coordinates to form a training data set for model training. During the model training process, the images are rotated and scaled to increase data diversity and adapt the model to different MRI image angles and qualities. The mean square error loss function is used to measure the error between the key point coordinates predicted by the deep learning network and the actual coordinates. The calculation formula is as follows: in: Loss refers to the error between the predicted key point coordinates and the actual coordinates, that is, the loss value; N refers to the total number of samples, the total number of coordinate points in all training data; (x i ,y i ) refers to the true coordinate value in the i-th sample; Refers to the predicted values ​​of the x and y coordinates of the i-th sample; Evaluate the model performance on the test set, calculate the error between the predicted key points and the actual key points according to the formula, and adjust the model parameters or perform hyperparameter tuning based on the test results to improve the model's performance in key point prediction.

3. The temporomandibular joint disc double circle measurement model recognition system according to claim 2, characterized in that: The purpose of the double circle model building module is to determine the center and radius of the large and small circles by the three-tangent point normal intersection method. The steps are as follows: first, a large circle is drawn inside the condylar head, so that it is tangent to the upper surface and the inner and outer sides of the condylar head, and then a small circle is drawn inside the condylar neck, so that it is tangent to the inner and outer edges of the condylar neck and the lower boundary of the large circle, so as to form a standard geometric shape of the joint structure; The process of constructing the great circle model based on the condylar head is as follows: Draw normal lines at the three tangent points of the condylar structure, analyze the intersection relationship of the three tangent point normal lines, and calculate the coordinates of the center of the circle and the radius; If the three tangent point normals intersect each other; The intersection points of the three tangent normal lines are A(x1,y1), B(x2,y2) and C(x3,y3). The centroid of the triangle is used as the center of the circle G(x g ,y g )Location; Then the radius of the great circle model based on the condylar head is calculated by calculating the center of the great circle G(x g ,y g ) to the normals of the three tangent points as the circle radius; If the three tangent point normals do not intersect, they are parallel; First, the tangent point normal line 1 L1: a1x+b1y+c1=0 is parallel to the tangent point normal line 2 L2: a2x+b2y+c2=0. Then, the intersection of the third tangent point normal line L3 with the tangent point normal line 1 L1 and the tangent point normal line 2 L2 is obtained, and then two valid intersection points P are obtained. 13 (x 13 ,y 13 ) and P 23 (x 23 ,y 23 ); Then use the midpoint of these two intersection points as the center of the replacement large circle G'(x g ,y g ); Then combine the remaining non-parallel tangent point normal line L3 and connect its tangent point P3(x3,y3) with P 13 (x 13 ,y 13 ) and P 23 (x 23 ,y 23 ) together with the weighted center of gravity, further correct the position of the center of the circle, improve the stability of the center of the circle positioning, and obtain the center of the large circle G(x g ,y g ); Similarly, according to the three tangent point normals do not intersect the center of the great circle G (x g ,y g ), calculate the distance from the center of the circle to the three tangent point normals and use the median method to determine the radius of the circle. For the equation of the tangent point normal Li a i x+b i y+c i =0, center G(x g ,y g ) to the tangent point normal line Li is d i : Take the median of the normal distances of the three tangent points as the radius.

4. The temporomandibular joint disc double circle measurement model recognition system according to claim 3, characterized in that: The specific process of constructing the small circle model based on the condylar neck in the double circle model construction module is as follows: Similarly, based on the large circle model of the condylar head, the small circle model based on the condylar neck is constructed by drawing normals at the three tangent points of the condylar neck structure, analyzing the relationship between the three normals, and calculating the coordinates and radius of the center of the circle when the tangent point normals intersect and when the tangent point normals are parallel.

5. The temporomandibular joint disc double circle measurement model recognition system according to claim 4, characterized in that: The generation of the condylar angle bisector and the articular disc angle bisector based on the key points in the displacement parameter calculation module is specifically as follows: Connect the center of the small circle with the inner and outer convex points of the temporomandibular joint disc and the inner and outer convex points of the condyle and draw an angle bisector, the inner and outer convex points of the temporomandibular joint disc correspond to the left boundary point of the joint disc and the right boundary point of the joint disc, and the inner and outer convex points of the condyle correspond to the left boundary point of the condyle and the right boundary point of the condyle.

6. The temporomandibular joint disc double circle measurement model recognition system according to claim 5, characterized in that: The angle α of the angle bisector measured in the displacement parameter calculation module is used to reflect the severity of the displacement of the temporomandibular joint disc, and the distance between the intersection point A and the intersection point B is analyzed to reflect the displacement direction of the temporomandibular joint disc.

Images