Multi-modal tooth root quantitative analysis method and system based on residual neural network

Abstract

The invention provides a multi-modal tooth root quantitative analysis method and system based on a residual neural network, and the method comprises the steps: obtaining multi-modal image samples of tooth roots, each multi-modal image sample comprises a pair of 5T high-field oral MRI images of the same tooth root part, and forming a tooth root 5TMRI and CBCT image data set; the constructed and trained tooth root image registration model comprises an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module; a 5TMRI image to be registered and a CBCT image to be registered are subjected to the trained tooth root image registration model, and a 5TMRI image and a CBCT image after registration are obtained; a quantitative analysis result is obtained; the periodontal health degree is judged; according to the method, higher registration precision can be realized, the tooth and periodontal health conditions are judged based on the registered image, objective quantitative evaluation of the tooth and periodontal health conditions is realized, and the accuracy of judging the tooth and periodontal health conditions is improved.

Description

Technical Field

[0001] The present invention relates to a multi-modal tooth root quantitative analysis method and system based on residual neural network, belonging to the technical field of image processing and analysis. Background Art

[0002] In dentistry, quantitative analysis of tooth roots is essential for diagnosis and treatment planning. Traditional root analysis methods mainly rely on radiographic techniques such as X-rays and CT scans, which are effective but have radiation risks and lack accuracy in analyzing root details.

[0003] 5T high-field oral magnetic resonance imaging 5TMRI and cone-beam computed tomography CBCT are two commonly used imaging technologies in medical imaging. They have unique advantages in the detection of oral and maxillofacial hard tissue and soft tissue structures respectively. 5T high-field oral magnetic resonance imaging 5TMRI uses high magnetic field strength to provide higher imaging resolution and better imaging effect for soft tissue, allowing doctors to observe tiny lesions and structures in the oral cavity, which helps in the early detection and accurate diagnosis of periodontal disease and endodontic disease.

[0004] With the development of artificial intelligence technology, especially deep learning technology, the multimodal images of tooth roots are registered and fused, that is, the image data of the two imaging technologies, 5TMRI images and CBCT images, are accurately aligned in the same coordinate system, and the tooth and periodontal conditions are quantitatively analyzed based on the results of the registration to obtain corresponding information, or a comprehensive evaluation of the oral tooth hard tissue and periodontal soft tissue can be achieved, which helps doctors to fully understand the patient's oral health status and provide more reliable support for the auxiliary diagnosis of root diseases.

[0005] Therefore, the present invention proposes a multimodal root quantitative analysis method and system based on residual neural network to address the problem of quantitative analysis of tooth roots. The method aims to utilize the image information of 5TMRI and CBCT to achieve accurate alignment and quantitative analysis of tooth roots, providing an innovative solution for auxiliary diagnosis and treatment planning of dental diseases. Summary of the invention

[0006] The purpose of the present invention is to provide a multimodal root quantitative analysis method and system based on residual neural network, which can automatically perform root image registration and perform root quantitative analysis based on the registration results, including the number of roots, the number of pulp root canals, the thickness of periodontal membrane, the height of alveolar bone and the depth of periodontal pocket, etc., to solve the problem that the accuracy of detailed root analysis in the prior art needs to be improved.

[0007] The technical solution of the present invention is:

[0008] A multimodal tooth root quantitative analysis method based on residual neural network includes the following steps:

[0009] S1. Obtain multimodal image samples of tooth roots. Each multimodal image sample includes a pair of 5T high-field oral MRI images, i.e., 5T MRI images and CBCT images, of the same tooth root site, and annotates registration transformation parameters as labels to form a tooth root 5T MRI and CBCT image dataset.

[0010] S2. Construct a tooth root image registration model, which includes an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module. The image preprocessing module preprocesses the input 5TMRI image and CBCT image to obtain the preprocessed 5TMRI image and the preprocessed CBCT image and inputs them into the residual neural network. The residual neural network predicts the registration transformation parameters required to register the 5TMRI image to the CBCT image and outputs them to the rigid registration module. The rigid registration module uses a rigid registration method to align the overall structure of the tooth root of the 5TMRI image with the CBCT image to be registered. The spatial transformation module uses a bilinear interpolation method to perform spatial transformation on the 5TMRI image to obtain the registered 5TMRI and CBCT images.

[0011] S3, using the tooth root 5TMRI and CBCT image datasets of step S1 to train the tooth root image registration model to obtain a trained tooth root image registration model;

[0012] S4, using the trained tooth root image registration model to register the 5T MRI image and the CBCT image to be registered to obtain the registered 5T MRI and CBCT images;

[0013] S5, performing quantitative analysis on the registered 5T MRI and CBCT images obtained in step S4, including analysis of the number of tooth roots, the number of pulp root canals, the thickness of the periodontal membrane, the height of the alveolar bone, and the depth of the periodontal pocket, to obtain quantitative analysis results;

[0014] S6. Based on the quantitative analysis results obtained in step S5, determine the periodontal health status.

[0015] Further, in step S2, the input 5T MRI image and CBCT image are preprocessed to obtain a preprocessed 5T MRI image and a preprocessed CBCT image, specifically,

[0016] The 5T MRI images and CBCT images were normalized to the same setting range;

[0017] Perform data enhancement on 5TMRI images;

[0018] The 5T MRI images and CBCT images were resampled in the coronal plane using the bilinear interpolation method;

[0019] The 5T MRI image and the CBCT image are randomly cropped, and the cropped image contains the structure of the tooth root, so as to obtain the preprocessed 5T MRI image and the preprocessed CBCT image.

[0020] Furthermore, in step S2, the residual neural network includes an input layer, a convolutional layer, a pooling layer, a residual module, a global average pooling layer, a fully connected layer and an output layer.

[0021] Input layer: used to input the preprocessed 5T MRI image and preprocessed CBCT image into the convolution layer;

[0022] Convolutional layer: Use a 7x7 convolution kernel, a step size of 2, an output channel number of 64, use zero padding, and then perform batch normalization and rectified linear unit ReLU activation to obtain the convolved feature map and output it to the maximum pooling layer;

[0023] Maximum pooling layer: Use 3x3 pooling units with a stride of 2 to perform maximum pooling operations to reduce the spatial dimension of the feature map after convolution and output it to the residual block;

[0024] Residual module: contains 4 residual units connected in series, where the step size of the first residual unit is 1 and the step size of the other three residual units is 2, which is used to reduce the size of the feature map and then output it to the global average pooling layer;

[0025] Global average pooling layer: used to perform global average pooling on the input feature map, average the features in each channel, and obtain a fixed-length 5TMRI image feature vector and a fixed-length CBCT image feature vector;

[0026] Fully connected layer: used to learn the mapping relationship between the fixed-length 5TMRI image feature vector and the fixed-length CBCT image feature vector and the registration transformation parameters, and predict the registration transformation parameters;

[0027] Output layer: used to output the registration transformation parameters obtained by the fully connected layer.

[0028] Furthermore, the registration transformation parameters include a translation vector, a rotation angle, and a scaling factor.

[0029] Furthermore, in step S5, a quantitative analysis is performed based on the registered 5T MRI and CBCT images, specifically, the number of roots, the number of pulp canals, the length of the root, the thickness of the periodontal membrane, the height of the alveolar bone, and the depth of the periodontal pocket. The periodontal membrane thickness, the height of the alveolar bone, the depth of the periodontal pocket, and the health of the pulp are summed up with the set scores to obtain a total score; wherein:

[0030] ①Analyze the number of tooth roots: identify and count the tooth roots in the image;

[0031] ②Analyze the number of dental pulp root canals: identify and count the root canals in the image;

[0032] ③Analyze the root length: identify the root length in the image;

[0033] ④Analyze the periodontal membrane thickness: identify the periodontal membrane thickness in the image;

[0034] ⑤Analyze the alveolar bone height: identify the alveolar bone height in the image;

[0035] ⑥Analyze the periodontal pocket depth: identify the periodontal pocket depth in the image;

[0036] ⑦ Score the periodontal membrane thickness: 0 points for >0.1 mm and ≤0.4 mm, 1 point for >0.4 mm or <0.1 mm;

[0037] ⑧ Score the alveolar bone height: 0 points if the alveolar bone height is ≥ 2 / 3 of the root length, 1 point if the alveolar bone height is > 1 / 2 of the root length and less than 2 / 3 of the root length, 2 points if the alveolar bone height is < 1 / 2 of the root length or > 1 / 3 of the root length, and 3 points if the alveolar bone height is < 1 / 3 of the root length;

[0038] ⑨ Scoring of periodontal pocket depth: less than 3 mm for 0 points, >3 mm and ≤5 mm for 1 point, >5 mm for 2 points;

[0039] ⑩ Score the health of the dental pulp: 0 points if no abnormality is identified, 1 point if inflammation is identified, and 2 points if necrosis is identified.

[0040] Specifically, the total score = the score of periodontal membrane thickness + the score of alveolar bone height + the score of periodontal pocket depth + the score of pulp health, where:

[0041] Scores in the range of 0-2 indicate healthy periodontium;

[0042] A score in the range of 3-4 indicates mild periodontal inflammation;

[0043] Scores in the range of 5-6 indicate moderate periodontal inflammation;

[0044] A score between 7 and 8 indicates severe periodontitis.

[0045] A multimodal tooth root quantitative analysis system of a residual neural network implementing any of the methods described above, comprising an image acquisition module, a model building module, a model training module, an image registration module, a quantitative analysis module and a result determination module,

[0046] Image acquisition module: obtain multimodal image samples of tooth roots. Each multimodal image sample includes a pair of 5T high-field oral MRI images of the same tooth root, namely, 5T MRI images and CBCT images, and annotates the registration transformation parameters as labels to form a tooth root 5T MRI and CBCT image dataset;

[0047] Model construction module: construct a tooth root image registration model, which includes an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module. The image preprocessing module preprocesses the input 5TMRI image and CBCT image to obtain the preprocessed 5TMRI image and the preprocessed CBCT image and input them into the residual neural network. The residual neural network predicts the registration transformation parameters required to register the 5TMRI image to the CBCT image and outputs them to the rigid registration module. The rigid registration module uses a rigid registration method to align the overall structure of the tooth root of the 5TMRI image with the CBCT image to be registered. The spatial transformation module uses a bilinear interpolation method to perform spatial transformation on the 5TMRI image to obtain the registered 5TMRI and CBCT images.

[0048] Model training module: Use the tooth root 5TMRI and CBCT image datasets to train the tooth root image registration model to obtain the trained tooth root image registration model;

[0049] Image registration module: The 5T MRI image to be registered and the CBCT image to be registered are used through the trained tooth root image registration model to obtain the registered 5T MRI and CBCT images;

[0050] Quantitative analysis module: Quantitative analysis is performed based on the registered 5T MRI and CBCT images, including the number of roots, number of pulp canals, root length, periodontal membrane thickness, alveolar bone height and periodontal pocket depth, to obtain quantitative analysis results;

[0051] Result determination module: Determine the periodontal health level based on the quantitative analysis results obtained by the quantitative analysis module.

[0052] The beneficial effects of the present invention are:

[0053] 1. Compared with the existing technology, this multimodal root quantitative analysis method and system based on residual neural network can achieve higher registration accuracy, determine the health status of teeth and periodontal health based on the registered images, realize objective quantitative evaluation of tooth and periodontal health status, and improve the accuracy of judgment on tooth and periodontal health status.

[0054] 2. As traditional methods may find it difficult to capture sufficient feature information, the present invention uses a residual neural network to learn high-level feature representations of images, has stronger adaptability, can better handle complex image registration problems, can effectively improve the accuracy of registration, and provide more reliable imaging support for auxiliary diagnosis of root diseases.

[0055] 3. This multimodal root quantitative analysis method and system based on residual neural network has stronger robustness. This method can learn the abstract features in the image through the residual neural network, and has a certain robustness to the noise and interference in the image. Therefore, in complex medical imaging scenes, the present invention can better maintain the stability of registration.

[0056] Fourth, this multimodal tooth root quantitative analysis method and system based on residual neural network can achieve automation and high efficiency. In view of the traditional method that requires manual selection and marking of feature points and estimation of registration transformation parameters, the residual neural network-based method of the present invention can automatically learn feature representation and predict registration transformation parameters, so it has a higher degree of automation and efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Figure 1 It is a schematic diagram of the process of a multimodal tooth root quantitative analysis method based on a residual neural network according to an embodiment of the present invention;

[0058] Figure 2 is a schematic diagram illustrating a tooth root image registration model in an embodiment;

[0059] Figure 3 is a schematic diagram for explaining a residual neural network in an embodiment;

[0060] Figure 4 is a schematic diagram for explaining tooth and periodontal setting scores in the embodiment;

[0061] Figure 5 It is a schematic diagram for explaining a multimodal tooth root quantitative analysis system based on a residual neural network according to an embodiment. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] The embodiment provides a multimodal tooth root quantitative analysis method based on residual neural network, such as Figure 1 , including the following steps:

[0064] S1. Obtain tooth root image samples. Each tooth root image sample includes a pair of 5T high-field oral MRI images of the same tooth root site, namely, a 5T MRI image and a CBCT image, and annotates the registration transformation parameters as labels to form a tooth root 5T MRI and CBCT image dataset.

[0065] In step S1, tooth root image samples are collected from the medical imaging database. Each tooth root image sample contains a pair of 5T MRI images and CBCT images, and their registration transformation parameters such as translation, rotation, and scaling are annotated. Ensure that the 5T MRI and CBCT images in the dataset correspond to the same root site of the same patient to ensure the accuracy and reliability of the registration. T2-weighted imaging is selected for MRI images. T2-weighted imaging is sensitive to moisture and is therefore suitable for evaluating the pathological state of pulp inflammation and fluid accumulation. The tooth root 5T MRI and CBCT image datasets are divided into 70% training sets, 15% validation sets, and 15% test sets for model training, tuning, and evaluation.

[0066] S2, construct a tooth root image registration model, such as Figure 2 The tooth root image registration model includes an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module. The image preprocessing module preprocesses the input 5TMRI image and CBCT image to obtain the preprocessed 5TMRI image and the preprocessed CBCT image and input them into the residual neural network. The residual neural network predicts the registration transformation parameters required to register the 5TMRI image to the CBCT image and outputs them to the rigid registration module. The rigid registration module uses a rigid registration method to align the overall structure of the tooth root of the 5TMRI image with the CBCT image to be registered. The spatial transformation module uses a bilinear interpolation method to perform spatial transformation on the 5TMRI image to obtain the registered 5TMRI and CBCT images.

[0067] In step S2, the 5T MRI image and the CBCT image are preprocessed to obtain a preprocessed 5T MRI image and a preprocessed CBCT image, specifically,

[0068] 1) Normalize the 5T MRI and CBCT images to the same set range. Specifically, divide the intensity values ​​of the CBCT images by 1000 HU (Hounsfield units) and truncate the values ​​outside the range of [-1,1]. Linearly rescale the 5T MRI images and also convert them to the range of [-1,1]. This can handle different intensity values ​​generated by different imaging techniques and ensure data consistency during training;

[0069] 2) Perform data augmentation on 5T MRI images; apply random color jitter (brightness and contrast randomization: 0.2) to 5T MRI images to simulate brightness and contrast changes in real-world imaging and enhance the robustness of the model;

[0070] 3) Resample the paired images to a uniform spatial resolution of 1x1 mm in the coronal plane using bilinear interpolation, with a slice thickness of 2.5-3.5 mm. Slice the volume into 2D coronal images and delete the slices that were not segmented;

[0071] 4) Perform random cropping on the 5T MRI image and the CBCT image, adjust the image size to 1024x1024 pixels, and make the structure containing the tooth root in the cropped image clear and visible without irrelevant background or noise, so as to obtain the preprocessed 5T MRI image and the preprocessed CBCT image.

[0072] The paired 5T MRI images and CBCT images were preprocessed separately, and the preprocessing methods included image data normalization, image enhancement, image resampling, and image cropping. Such operations can reduce the differences between images, provide better input for the subsequent registration process, and improve the registration effect.

[0073] In step S2, ResNet-50 is selected as the residual neural network. The residual neural network includes an input layer, a convolutional layer, a pooling layer, a residual module, a global average pooling layer, a fully connected layer, and an output layer. Figure 3 :

[0074] Input layer: used to input the preprocessed 5T MRI image and preprocessed CBCT image into the convolution layer;

[0075] Convolution layer: Use a 7x7 convolution kernel, a step size of 2, 64 output channels, zero padding, and then perform batch normalization and rectified linear unit ReLU activation to obtain the convolved feature map and output it to the maximum pooling layer; among them, the rectified linear unit ReLU maps negative input values ​​to zero and keeps positive input values ​​unchanged. Its mathematical expression is: f(x)=max(0,x), where x is the input value.

[0076] Maximum pooling layer: Use 3x3 pooling units with a stride of 2 to perform maximum pooling operations to reduce the spatial dimension of the convolutional feature map and output it to the residual module;

[0077] Residual module: It contains 4 residual units connected in series. Each residual unit contains several convolutional layers and batch normalization layers. The step size of the first residual unit is 1, and the step size of the other three residual units is 2. It is used to reduce the size of the feature map and then output it to the global average pooling layer.

[0078] Global average pooling layer: used to perform global average pooling on the input feature map, average the features in each channel, and obtain a fixed-length 5TMRI image feature vector and a fixed-length CBCT image feature vector;

[0079] Fully connected layer: used to learn the mapping relationship between the fixed-length 5TMRI image feature vector and the fixed-length CBCT image feature vector and the registration transformation parameters, and predict the registration transformation parameters;

[0080] Output layer: used to output the registration transformation parameters obtained by the fully connected layer. It contains three nodes, corresponding to the translation vector, rotation angle and scaling factor, which are used to map the 5TMRI image to the spatial position and direction of the CBCT image. These parameters can more comprehensively describe the spatial transformation relationship between images, thereby further improving the accuracy and stability of registration.

[0081] Residual neural network is a deep learning architecture that introduces the concept of "residual learning" to solve the problem of difficult training as the network depth increases. In traditional convolutional neural networks, the problem of gradient vanishing or exploding becomes more serious as the number of network layers increases, making deep networks difficult to train. Residual neural networks solve this problem by adding skip connections or shortcut connections, allowing gradients to flow directly to the previous layers. In the field of tooth root image registration, using residual networks as feature extractors can improve the training efficiency and performance of the network, allowing for the construction of deeper network structures while avoiding the problem of gradient vanishing or exploding.

[0082] In step S2, the rigid registration module uses an affine transformation method to rigidly transform the 5TMRI image according to the registration transformation parameters. The 5TMRI image is aligned to the CBCT image through translation, rotation and scaling operations. Among them, the translation operation is used to move the image in space, the rotation operation is used to adjust the rotation angle of the image, and the scaling operation is used to adjust the size of the image so as to better align with the target image. After rigid registration, the spatial transformation module uses a bilinear interpolation method to spatially transform the 5TMRI image to ensure the quality and accuracy of the registered image. The registered 5TMRI and CBCT images are quality evaluated to ensure the accuracy and stability of the registration. The root mean square error RMSE is used to calculate the gap between the predicted parameters and the labeled parameters to evaluate the registration error. The smaller the RMSE, the smaller the gap between the predicted parameters and the labeled parameters, the smaller the registration error, and the higher the accuracy and stability of the registration. The registered 5TMRI and CBCT images are used for medical diagnosis, surgical planning and other applications. The registered images can be further processed, such as image segmentation, feature extraction, etc., to obtain more diagnostic information.

[0083] S3. Use the tooth root 5T MRI and CBCT image datasets of step S1 to train the tooth root image registration model to obtain a trained tooth root image registration model.

[0084] In step S3, the mean square error (MSE) is used as the loss function to measure the difference between the residual neural network output and the annotated registration transformation parameters in the tooth root image registration model. The residual neural network training of the tooth root image registration model is performed using the Adam optimizer, with the learning rate set to 0.001 and the batch size set to 32.

[0085] S4. The 5T MRI image to be registered and the CBCT image to be registered are registered through the trained tooth root image registration model to obtain the registered 5T MRI and CBCT images.

[0086] In step S4, a quantitative analysis is performed based on the registered 5T MRI and CBCT images, specifically, the number of roots, the number of pulp canals, the length of the root, the thickness of the periodontal membrane, the height of the alveolar bone, and the depth of the periodontal pocket are analyzed, and the periodontal membrane thickness, the height of the alveolar bone, the depth of the periodontal pocket, and the health of the pulp are summed up with the set scores to obtain a total score; wherein:

[0087] ①Analyze the number of tooth roots: identify and count the tooth roots in the image;

[0088] ②Analyze the number of dental pulp root canals: identify and count the root canals in the image;

[0089] ③Analyze the root length: identify the root length in the image;

[0090] ④Analyze the periodontal membrane thickness: identify the periodontal membrane thickness in the image;

[0091] ⑤Analyze the alveolar bone height: identify the alveolar bone height in the image;

[0092] ⑥Analyze the periodontal pocket depth: identify the periodontal pocket depth in the image;

[0093] ⑦ Score the periodontal membrane thickness: 0 points for >0.1 mm and ≤0.4 mm, 1 point for >0.4 mm or <0.1 mm;

[0094] ⑧ Score the alveolar bone height: 0 points if the alveolar bone height is ≥ 2 / 3 of the root length, 1 point if the alveolar bone height is > 1 / 2 of the root length and less than 2 / 3 of the root length, 2 points if the alveolar bone height is < 1 / 2 of the root length or > 1 / 3 of the root length, and 3 points if the alveolar bone height is < 1 / 3 of the root length;

[0095] ⑨ Scoring of periodontal pocket depth: less than 3 mm for 0 points, >3 mm and ≤5 mm for 1 point, >5 mm for 2 points;

[0096] ⑩ Score the health of the dental pulp: 0 points if no abnormality is identified, 1 point if inflammation is identified, and 2 points if necrosis is identified.

[0097] The registered 5T MRI and CBCT images are used to evaluate the health of the dental pulp and determine whether it is in an inflammation, necrosis, or normal state:

[0098] Normal dental pulp: The signal intensity is greater than the first threshold and not greater than the second threshold, and the difference between adjacent signal intensities is less than the set threshold, that is, it displays medium to high signals, the signal is uniform, and there are no obvious bright spots or dark spots.

[0099] Pulp inflammation: When the pulp is inflamed, the water content in the pulp tissue increases due to edema and infiltration of inflammatory cells. The signal intensity is greater than the third threshold, that is, acute inflammation shows a high signal (bright), indicating fluid accumulation; the signal intensity is greater than the second threshold and not greater than the third threshold, that is, chronic inflammation, the signal is higher but not as significant as the acute stage.

[0100] Pulp necrosis: When the signal intensity is not greater than the first threshold, it appears as a low signal (dark). Especially when the pulp is completely necrotic, the signal intensity is significantly reduced.

[0101] S5. Perform quantitative analysis on the registered 5T MRI and CBCT images obtained in step S4, including analysis of the number of roots, the number of pulp canals, the length of the roots, the thickness of the periodontal membrane, the height of the alveolar bone and the depth of the periodontal pocket, to obtain quantitative analysis results.

[0102] S6. Based on the quantitative analysis results obtained in step S5, determine the periodontal health status.

[0103] In step S6, by scoring the tooth root image, the health status of the oral tooth hard tissue and periodontal soft tissue can be quantitatively evaluated. The higher the total score, the worse the health status.

[0104] Specifically, the total score = the score of periodontal membrane thickness + the score of alveolar bone height + the score of periodontal pocket depth + the score of pulp health, where:

[0105] Scores in the range of 0-2 indicate healthy periodontium;

[0106] A score in the range of 3-4 indicates mild periodontal inflammation;

[0107] Scores in the range of 5-6 indicate moderate periodontal inflammation;

[0108] A score between 7 and 8 indicates severe periodontitis.

[0109] Compared with the existing technology, this multimodal root quantitative analysis method based on residual neural network can achieve higher registration accuracy, determine the health status of teeth and periodontal health based on the registered images, realize objective quantitative evaluation of tooth and periodontal health status, and improve the accuracy of judgment on tooth and periodontal health status.

[0110] This multimodal root quantitative analysis method based on residual neural network uses residual neural network to predict the registration transformation parameters required for 5TMRI image registration to CBCT image. It has better feature learning ability, can better capture the feature information in the image, help improve the accuracy and robustness of registration, can better optimize the entire registration process, reduce the error propagation and complexity in traditional methods, and can significantly improve the accuracy of registration. This method uses residual network as a feature extractor. By introducing the residual learning framework, the residual network allows the network to learn feature representations beyond the identity mapping, which helps to capture deeper features, including complex local structures and global context information. In view of the problem that medical images may have noise and artifacts, the residual module is composed of multiple residual units, each of which contains several layers. The layers in the residual unit learn the residual between the input and output, that is, the part that the network hopes to predict, so that it has good robustness to noise and small changes in the image, and is suitable for medical image registration.

[0111] like Figure 4 The embodiment also provides a multimodal tooth root quantitative analysis method based on residual neural network to implement any of the methods described above, including an image acquisition module, a model construction module, a model training module, an image registration module, a quantitative analysis module and a result judgment module.

[0112] Image acquisition module: obtain tooth root image samples. Each tooth root image sample includes a pair of 5T high-field oral MRI images of the same tooth root, namely, a 5T MRI image and a CBCT image, and annotates the registration transformation parameters as labels to form a tooth root 5T MRI and CBCT image dataset;

[0113] Model construction module: construct a tooth root image registration model, which includes an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module. The image preprocessing module preprocesses the input 5TMRI image and CBCT image to obtain the preprocessed 5TMRI image and the preprocessed CBCT image and input them into the residual neural network. The residual neural network predicts the registration transformation parameters required to register the 5TMRI image to the CBCT image and outputs them to the rigid registration module. The rigid registration module uses a rigid registration method to align the overall structure of the tooth root of the 5TMRI image with the CBCT image to be registered. The spatial transformation module uses a bilinear interpolation method to perform spatial transformation on the 5TMRI image to obtain the registered 5TMRI and CBCT images.

[0114] Model training module: Use the tooth root 5TMRI and CBCT image datasets to train the tooth root image registration model to obtain the trained tooth root image registration model;

[0115] Image registration module: The 5T MRI image to be registered and the CBCT image to be registered are passed through the trained root image registration model to obtain the registered 5T MRI and CBCT images.

[0116] Quantitative analysis module: Quantitative analysis is performed based on the registered 5T MRI and CBCT images, including the number of roots, the number of pulp canals, the thickness of the periodontal membrane, the height of the alveolar bone and the depth of the periodontal pocket, to obtain quantitative analysis results;

[0117] Result determination module: Determine the periodontal health level based on the quantitative analysis results obtained by the quantitative analysis module.

[0118] As traditional methods may find it difficult to capture sufficient feature information, the present invention uses a residual neural network to learn high-level feature representations of images, has stronger adaptability, can better handle complex image registration problems, and can effectively improve the accuracy of registration.

[0119] In this multimodal tooth root quantitative analysis method and system based on residual neural network, in the tooth root image registration model, after preprocessing the 5T MRI image and CBCT image, the residual module is used in the residual neural network to extract features and then the fully connected layer performs parameter prediction. The residual module is used as a feature extractor, which can extract and represent features from the input 5T MRI image and CBCT image, and can more directly use the original image for registration, which can reduce the complexity of intermediate steps and data conversion.

[0120] This multimodal root quantitative analysis method and system based on residual neural network has stronger robustness. The method can learn abstract features in the image through the residual neural network and has a certain robustness to noise and interference in the image. Therefore, in complex medical imaging scenes, the present invention can better maintain the stability of registration.

[0121] This multimodal tooth root quantitative analysis method and system based on residual neural network can achieve automation and high efficiency. In view of the traditional method that requires manual selection and marking of feature points and estimation of registration transformation parameters, the present invention can automatically learn feature representation and predict registration transformation parameters, thus having a higher degree of automation and efficiency.

[0122] Those skilled in the art should understand that those skilled in the art can implement variations by combining the prior art and the above embodiments, which will not be described in detail here. Such variations do not affect the essential content of the present invention, and will not be described in detail here.

[0123] The above are only embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be thought of by a person skilled in the art within the technical scope disclosed by the present invention without creative work should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope defined in the claims.

Claims

1. A multimodal tooth root quantitative analysis method based on residual neural network, characterized by: The following steps are included: S1. Obtain multimodal image samples of tooth roots. Each multimodal image sample includes a pair of 5T high-field oral MRI images, i.e., 5T MRI images and CBCT images, of the same tooth root site, and annotates registration transformation parameters as labels to form a tooth root 5T MRI and CBCT image dataset. S2. Construct a tooth root image registration model, which includes an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module. The image preprocessing module preprocesses the input 5TMRI image and CBCT image to obtain the preprocessed 5TMRI image and the preprocessed CBCT image and inputs them into the residual neural network. The residual neural network predicts the registration transformation parameters required to register the 5TMRI image to the CBCT image and outputs them to the rigid registration module. The rigid registration module uses a rigid registration method to align the overall structure of the tooth root of the 5TMRI image with the CBCT image to be registered. The spatial transformation module uses a bilinear interpolation method to perform spatial transformation on the 5TMRI image to obtain the registered 5TMRI and CBCT images. S3, using the tooth root 5TMRI and CBCT image datasets of step S1 to train the tooth root image registration model to obtain a trained tooth root image registration model; S4, using the trained tooth root image registration model to register the 5T MRI image and the CBCT image to be registered to obtain the registered 5T MRI and CBCT images; S5, performing quantitative analysis on the registered 5T MRI and CBCT images obtained in step S4, including analysis of the number of tooth roots, the number of pulp root canals, the thickness of the periodontal membrane, the height of the alveolar bone, and the depth of the periodontal pocket, to obtain quantitative analysis results; S6. Based on the quantitative analysis results obtained in step S5, determine the periodontal health status.

2. The multimodal tooth root quantitative analysis method based on residual neural network according to claim 1, characterized in that: In step S2, the input 5T MRI image and CBCT image are preprocessed to obtain a preprocessed 5T MRI image and a preprocessed CBCT image, specifically, The 5T MRI images and CBCT images were normalized to the same setting range; Perform data enhancement on 5T MRI images; The 5T MRI images and CBCT images were resampled in the coronal plane using the bilinear interpolation method; The 5T MRI image and the CBCT image are randomly cropped, and the cropped image contains the structure of the tooth root, so as to obtain the preprocessed 5T MRI image and the preprocessed CBCT image.

3. The multimodal tooth root quantitative analysis method based on residual neural network according to claim 1, characterized in that: In step S2, the residual neural network includes an input layer, a convolutional layer, a pooling layer, a residual module, a global average pooling layer, a fully connected layer and an output layer. Input layer: used to input the preprocessed 5T MRI image and preprocessed CBCT image into the convolution layer; Convolutional layer: Use a 7x7 convolution kernel, a step size of 2, an output channel number of 64, use zero padding, and then perform batch normalization and rectified linear unit ReLU activation to obtain the convolved feature map and output it to the maximum pooling layer; Maximum pooling layer: Use 3x3 pooling units with a stride of 2 to perform maximum pooling operations to reduce the spatial dimension of the feature map after convolution and output it to the residual block; Residual module: contains 4 residual units connected in series, where the step size of the first residual unit is 1 and the step size of the other three residual units is 2, which is used to reduce the size of the feature map and then output it to the global average pooling layer; Global average pooling layer: used to perform global average pooling on the input feature map, average the features in each channel, and obtain a fixed-length 5TMRI image feature vector and a fixed-length CBCT image feature vector; Fully connected layer: used to learn the mapping relationship between the fixed-length 5TMRI image feature vector and the fixed-length CBCT image feature vector and the registration transformation parameters, and predict the registration transformation parameters; Output layer: used to output the registration transformation parameters obtained by the fully connected layer.

4. The multimodal tooth root quantitative analysis method based on residual neural network according to claim 3, characterized in that: The registration transformation parameters include translation vector, rotation angle and scaling factor.

5. The multimodal tooth root quantitative analysis method based on residual neural network according to any one of claims 1 to 3, characterized in that: In step S5, a quantitative analysis is performed based on the registered 5T MRI and CBCT images, specifically, the number of roots, the number of pulp canals, the length of the root, the thickness of the periodontal membrane, the height of the alveolar bone, and the depth of the periodontal pocket are analyzed, and the total score is obtained by summing the set scores for the thickness of the periodontal membrane, the height of the alveolar bone, the depth of the periodontal pocket, and the health of the pulp; wherein: ①Analyze the number of tooth roots: identify and count the tooth roots in the image; ②Analyze the number of dental pulp root canals: identify and count the root canals in the image; ③Analyze the root length: identify the root length in the image; ④Analyze the periodontal membrane thickness: identify the periodontal membrane thickness in the image; ⑤Analyze the alveolar bone height: identify the alveolar bone height in the image; ⑥Analyze the periodontal pocket depth: identify the periodontal pocket depth in the image; ⑦ Score the periodontal membrane thickness: 0 points for >0.1 mm and ≤0.4 mm, 1 point for >0.4 mm or <0.1 mm; ⑧ Score the alveolar bone height: 0 points if the alveolar bone height is ≥ 2 / 3 of the root length, 1 point if the alveolar bone height is > 1 / 2 of the root length and less than 2 / 3 of the root length, 2 points if the alveolar bone height is < 1 / 2 of the root length or > 1 / 3 of the root length, and 3 points if the alveolar bone height is < 1 / 3 of the root length; ⑨ Scoring of periodontal pocket depth: less than 3 mm for 0 points, >3 mm and ≤5 mm for 1 point, >5 mm for 2 points; ⑩ Score the health of the dental pulp: 0 points if no abnormality is identified, 1 point if inflammation is identified, and 2 points if necrosis is identified.

6. The multimodal tooth root quantitative analysis method based on residual neural network according to claim 5, characterized in that: Total score = periodontal membrane thickness score + alveolar bone height score + periodontal pocket depth score + pulp health score, where: Scores in the range of 0-2 indicate healthy periodontium; A score in the range of 3-4 indicates mild periodontal inflammation; Scores in the range of 5-6 indicate moderate periodontal inflammation; A score between 7 and 8 indicates severe periodontitis.

7. A multimodal tooth root quantitative analysis system based on residual neural network for implementing the method according to any one of claims 1 to 6, characterized in that: It includes image acquisition module, model building module, model training module, image registration module, quantitative analysis module and result judgment module. Image acquisition module: obtain multimodal image samples of tooth roots. Each multimodal image sample includes a pair of 5T high-field oral MRI images of the same tooth root, namely, 5T MRI images and CBCT images, and annotates the registration transformation parameters as labels to form a tooth root 5T MRI and CBCT image dataset; Model construction module: construct a tooth root image registration model, which includes an image preprocessing module, a residual neural network, a rigid registration module and a spatial transformation module. The image preprocessing module preprocesses the input 5TMRI image and CBCT image to obtain the preprocessed 5TMRI image and the preprocessed CBCT image and input them into the residual neural network. The residual neural network predicts the registration transformation parameters required to register the 5TMRI image to the CBCT image and outputs them to the rigid registration module. The rigid registration module uses a rigid registration method to align the overall structure of the tooth root of the 5TMRI image with the CBCT image to be registered. The spatial transformation module uses a bilinear interpolation method to perform spatial transformation on the 5TMRI image to obtain the registered 5TMRI and CBCT images. Model training module: Use the tooth root 5TMRI and CBCT image datasets to train the tooth root image registration model to obtain the trained tooth root image registration model; Image registration module: The 5T MRI image to be registered and the CBCT image to be registered are used through the trained tooth root image registration model to obtain the registered 5T MRI and CBCT images; Quantitative analysis module: Quantitative analysis is performed based on the registered 5T MRI and CBCT images, including the number of roots, number of pulp canals, root length, periodontal membrane thickness, alveolar bone height and periodontal pocket depth, to obtain quantitative analysis results; Result determination module: Determine the periodontal health level based on the quantitative analysis results obtained by the quantitative analysis module.

Images