Tooth segmentation method and apparatus, method and apparatus for determining degree of alveolar bone resorption, and device and medium

Abstract

Provided in the present application are a tooth segmentation method and apparatus, a method and apparatus for determining the degree of alveolar bone resorption, and a device and a medium. The tooth segmentation method comprises: on the basis of morphological information of each tooth in a panoramic dental image, determining a tooth morphological map; performing tooth localization on the tooth morphological map; on the basis of an initial position of each tooth in the tooth morphological map, acquiring a region of interest of each tooth from the panoramic dental image; and segmenting the region of interest, so as to obtain a segmentation result of each tooth. The method for determining the degree of alveolar bone resorption comprises: inputting a panoramic dental image into a segmentation model, and determining a tooth contour curve, a first contour line and a second line, wherein the first contour line is a contour line that comprises coronal parts of at least two teeth, and the second contour line is a contour line of alveolar bones of the maxilla and mandible; and on the basis of distribution regions of the first contour line and the second contour line within the tooth contour curve, determining the degree of alveolar bone resorption of the teeth.

Description

Methods, apparatus, equipment and media for tooth segmentation and determination of alveolar bone resorption.

[0001] Cross-references to related applications

[0002] This application claims priority to Chinese Patent Application No. 202411546219.1, filed with the State Intellectual Property Office of the People's Republic of China on October 31, 2024, entitled "A Method and Apparatus for Determining the Degree of Alveolar Bone Resorption," the entire contents of which are incorporated herein by reference; and Chinese Patent Application No. 202411637327.X, filed with the State Intellectual Property Office of the People's Republic of China on November 14, 2024, entitled "A Method, Apparatus, Computing Device and Storage Medium for Tooth Segmentation," the entire contents of which are incorporated herein by reference. Technical Field

[0003] This application relates to the field of image processing technology, and in particular to a method, apparatus, device and medium for tooth segmentation and determining the degree of alveolar bone resorption. Background Technology

[0004] Panoramic radiographs are an important examination technique in oral examinations. They can conveniently and quickly display the teeth and surrounding anatomical structures of the entire oral cavity in a single imaging session, providing effective assistance in the diagnosis of oral diseases.

[0005] Tooth identification and segmentation in panoramic dental radiographs are crucial steps in their analysis. A common approach is manual segmentation, but this is time-consuming, inefficient, and prone to accuracy degradation due to prolonged exposure.

[0006] Alternatively, a dental panoramic image can be input into a general deep neural network model to obtain the segmentation result of each tooth. However, due to the possibility of overlapping and artifacts between adjacent teeth in the dental panoramic image, the accuracy of tooth recognition and segmentation is low.

[0007] Therefore, the problem of low tooth segmentation accuracy in panoramic dental radiographs due to overlapping and artifacts between adjacent teeth needs to be addressed. Summary of the Invention

[0008] This application provides a method, apparatus, device, and medium for tooth segmentation and determining the degree of alveolar bone resorption, which can improve the accuracy of tooth segmentation.

[0009] Firstly, embodiments of this application provide a tooth segmentation method. This method can be executed by a tooth segmentation device, which can be a terminal device or a module for a terminal device, or a server or a module for a server. This application does not limit the executing entity of this method. The method includes: determining a tooth morphology map based on the morphological information of each tooth in a panoramic oral radiograph; wherein the pixel representation information of each tooth in the tooth morphology map is higher in the central axis region than in the boundary region; performing tooth localization on the tooth morphology map to obtain the initial position of each tooth in the tooth morphology map; obtaining the region of interest of each tooth from the panoramic oral radiograph based on the initial position of each tooth in the tooth morphology map; and obtaining the segmentation result of each tooth by segmenting the region of interest of each tooth.

[0010] In the above scheme, the pixel representation information of each tooth in the central axis region of the tooth morphology image is higher than that in the boundary region, which weakens the representation information of the tooth boundary region. After the representation information of the tooth boundary region is weakened, even if there are superimposed images and artifacts between adjacent teeth in the panoramic oral radiograph, tooth instances can still be distinguished better, thus improving the accuracy of tooth segmentation.

[0011] Secondly, embodiments of this application provide a method for determining the degree of alveolar bone resorption, including:

[0012] Obtain panoramic dental radiographs;

[0013] The panoramic image of the teeth is input into the segmentation model to determine the tooth contour curve, the first contour line of all teeth, and the second contour line of all teeth; the first contour line is the outer edge line of the crown portion of the tooth including at least two teeth, and the second contour line is the outer edge line of the alveolar bone of the maxilla and mandible; the tooth contour curve is the curve formed by the outer surface edge of each tooth.

[0014] For any given tooth, the degree of alveolar bone resorption is determined based on the distribution area of ​​the first and second contour lines within the tooth's contour curve.

[0015] By determining the distribution area of ​​the first and second contour lines within the tooth contour curve of each tooth, the degree of alveolar bone resorption for each individual tooth can be determined, thus improving the accuracy of alveolar bone resorption determination.

[0016] Thirdly, embodiments of this application provide a tooth segmentation device, comprising: a morphology determination unit and a tooth segmentation unit; the morphology determination unit is used to determine a tooth morphology map based on the morphological information of each tooth in a panoramic oral radiograph; the pixel representation information of each tooth in the morphology map is higher in the central axis region than in the boundary region; the tooth segmentation unit is used to perform tooth localization on the morphology map to obtain the initial position of each tooth in the morphology map; based on the initial position of each tooth in the morphology map, to obtain the region of interest of each tooth from the panoramic oral radiograph; and to obtain the segmentation result of each tooth by segmenting the region of interest of each tooth.

[0017] Fourthly, embodiments of this application provide an apparatus for determining the degree of alveolar bone resorption, comprising:

[0018] The acquisition module is used to acquire panoramic dental radiographs.

[0019] The processing module is used to input the panoramic image of the teeth into the segmentation model to determine the tooth contour curve, the first contour line of all teeth, and the second contour line of all teeth; the first contour line is the outer edge line of the crown portion of the tooth including at least two teeth, and the second contour line is the outer edge line of the alveolar bone of the maxilla and mandible; the tooth contour curve is the curve formed by the outer surface edge of each tooth.

[0020] The judgment module is used to determine the degree of alveolar bone resorption of any tooth based on the distribution area of ​​the first contour line and the second contour line within the tooth contour curve of the tooth.

[0021] By identifying the distribution areas of the first and second contour lines within the tooth contour curve of each tooth, the degree of alveolar bone resorption for each individual tooth can be determined, thus improving the accuracy and efficiency of assessing the degree of alveolar bone resorption.

[0022] Fifthly, embodiments of this application provide a display interface that responds to one or more of the panoramic dental image, the curves of the teeth, and the degree of alveolar bone resorption of the teeth under the method described in the second aspect above.

[0023] Sixthly, embodiments of this application also provide a computing device, including:

[0024] Memory, used to store program instructions;

[0025] A processor is configured to invoke program instructions stored in the memory and execute any method of the first or second aspect described above according to the obtained program instructions.

[0026] In a seventh aspect, embodiments of this application also provide a computer-readable storage medium storing computer-readable instructions that, when read and executed by a computer, implement any of the methods in the first or second aspect described above.

[0027] Eighthly, embodiments of this application provide a computer program product including a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform any of the methods described in the first or second aspect above. Attached Figure Description

[0028] Figure 1 is a schematic flowchart of a tooth segmentation method provided in an embodiment of this application;

[0029] Figure 2 is a schematic diagram of a panoramic oral cavity structure provided in an embodiment of this application;

[0030] Figure 3 is a schematic diagram of a binary image structure provided in an embodiment of this application;

[0031] Figure 4 is a structural schematic diagram of a tooth morphology diagram provided in an embodiment of this application;

[0032] Figure 5 is a structural schematic diagram of a tooth morphology diagram provided in an embodiment of this application;

[0033] Figure 6 is a schematic flowchart of a method for determining a tooth morphology diagram according to an embodiment of this application;

[0034] Figure 7 is a structural schematic diagram of a tooth morphology diagram provided in an embodiment of this application;

[0035] Figure 8 is a flowchart illustrating a tooth segmentation method provided in an embodiment of this application;

[0036] Figure 9 is a structural schematic diagram of the overall region where a tooth is located, according to an embodiment of this application;

[0037] Figure 10 is a schematic diagram of the structure of a region of interest in a single tooth according to an embodiment of this application;

[0038] Figure 11 is a structural schematic diagram of a second model provided in an embodiment of this application;

[0039] Figure 12 is a schematic diagram of a feature enhancement module provided in an embodiment of this application;

[0040] Figure 13 is a schematic flowchart of a tooth segmentation method provided in an embodiment of this application;

[0041] Figure 14 is a structural schematic diagram of a tooth segmentation result provided in an embodiment of this application;

[0042] Figure 15 is a schematic flowchart of a method for determining the degree of alveolar bone resorption according to an embodiment of this application;

[0043] Figure 16 is a schematic diagram of a binary masking image of teeth provided in an embodiment of this application;

[0044] Figure 17 is a schematic diagram of a binary masking image of a dental crown provided in an embodiment of this application;

[0045] Figure 18 is a schematic diagram of a binary masking image of the alveolar bone provided in an embodiment of this application;

[0046] Figure 19 is a schematic diagram of the first outline and the second outline of all teeth provided in the embodiments of this application;

[0047] Figure 20 is a schematic diagram of the distribution area provided in an embodiment of this application;

[0048] Figure 21 is a schematic diagram of the main axes of all teeth provided in the embodiments of this application;

[0049] Figure 22 is a schematic diagram of any tooth provided in an embodiment of this application;

[0050] Figure 23 is a schematic diagram of the minimum area rectangle of each tooth provided in the embodiments of this application;

[0051] Figure 24 is a schematic diagram of a tooth-splitting device provided in an embodiment of this application;

[0052] Figure 25 is a schematic diagram of a tooth-splitting device provided in an embodiment of this application;

[0053] Figure 26 is a schematic diagram of a device for determining the degree of alveolar bone resorption provided in an embodiment of this application;

[0054] Figure 27 is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0055] Figure 1 is a schematic flowchart of a tooth segmentation method provided in an embodiment of this application. This method can be executed by a tooth segmentation device, which can be a terminal device or a module for a terminal device, or a server or a module for a server. This application does not limit the subject of execution of this method.

[0056] The method includes the following steps:

[0057] Step 101: Determine the tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph.

[0058] In the tooth morphology diagram, the pixel representation information of each tooth in the central axis region is higher than that in the boundary region.

[0059] In one possible implementation, a panoramic oral radiograph is a two-dimensional image that simultaneously reflects all the teeth and alveolar bone in a patient's mouth onto a single X-ray film, displaying an overview of the entire oral cavity. For example, a panoramic oral radiograph is shown in Figure 2.

[0060] In one possible implementation, the morphological information of the tooth refers to the representation information of the tooth shape. For example, the morphological information of the tooth is the skeletal information of the tooth, and / or the distance information from any pixel point of the tooth to the nearest tooth boundary point.

[0061] Step 102: Locate the teeth in the tooth morphology diagram to obtain the initial position of each tooth in the tooth morphology diagram.

[0062] In one possible implementation, the tooth morphology map is input into a neural network model to obtain the initial position of each tooth in the tooth morphology map; wherein the neural network model includes, but is not limited to, Faster R-CNN, YOLO, and RetinaNet.

[0063] In another possible implementation, the initial position of each tooth in the tooth morphology diagram is obtained based on the boundary information of adjacent teeth in the tooth morphology diagram and the length information of each tooth in the tooth morphology diagram.

[0064] In another possible implementation, the tooth morphology map is segmented according to a region growing method to obtain the initial position of each tooth in the tooth morphology map. This application does not limit the method for determining the initial position of each tooth in the tooth morphology map.

[0065] Step 103: Based on the initial position of each tooth in the tooth morphology diagram, obtain the region of interest for each tooth from the panoramic oral radiograph.

[0066] Step 104: By segmenting the region of interest of each tooth, the segmentation result of each tooth is obtained.

[0067] In the above scheme, the pixel representation information of each tooth in the central axis region of the tooth morphology image is higher than that in the boundary region, which weakens the representation information of the tooth boundary region. After the representation information of the tooth boundary region is weakened, even if there are superimposed images and artifacts between adjacent teeth in the panoramic oral radiograph, tooth instances can still be distinguished better, thus improving the accuracy of tooth segmentation.

[0068] In one embodiment, step 101 above, determining the tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph, includes: for any tooth in the panoramic oral radiograph, determining the distance information from any pixel point of the tooth to the nearest tooth boundary point, and determining the pixel representation information of any pixel point based on the distance information, thereby obtaining the tooth morphology map; wherein, the smaller the distance from the pixel point to the nearest tooth boundary point, the weaker the pixel representation information of the pixel point; the larger the distance from the pixel point to the nearest tooth boundary point, the stronger the pixel representation information of the pixel point. This scheme can accurately and effectively determine the tooth morphology map, thereby improving the accuracy of tooth segmentation.

[0069] In one possible implementation, for any tooth in a panoramic oral radiograph, weighted distance information from any pixel of the tooth to multiple nearest boundary points of the tooth is determined, and pixel representation information of any pixel is determined based on the weighted distance information, thereby obtaining a tooth morphology map. The multiple nearest boundary points include one or more of the following: nearest upper boundary point, nearest lower boundary point, nearest left boundary point, or nearest right boundary point; specifically, weighted distance information from any pixel of the tooth to the nearest upper boundary point, nearest lower boundary point, nearest left boundary point, and nearest right boundary point of the tooth is determined, and pixel representation information of any pixel is determined based on the weighted distance information, thereby obtaining a tooth morphology map. The weighted distance information is the weighted average of the distance information from each pixel of the tooth to the multiple nearest boundary points of the tooth.

[0070] For example, suppose there is a tooth A in a panoramic oral radiograph, and tooth A has three pixel points (x1, y1), (x2, y2), and (x3, y3). The distance from (x1, y1) to the nearest tooth boundary point of tooth A is 2, the distance from (x2, y2) to the nearest tooth boundary point of tooth A is 20, and the distance from (x3, y3) to the nearest tooth boundary point of tooth A is 50. Then, the pixel representation information of (x1, y1) is determined to be 2, the pixel representation information of (x2, y2) is 20, and the pixel representation information of (x3, y3) is 50. The distance from (x1, y1) to the nearest tooth boundary point of tooth A is the smallest, and the distance from (x3, y3) to the nearest tooth boundary point of tooth A is the largest. The pixel representation information of (x3, y3) is greater than that of (x2, y2) than that of (x1, y1). The tooth morphology map obtained for tooth A is shown in Figure 4.

[0071] In one embodiment, step 101 above, determining the tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph, includes: for any tooth in the panoramic oral radiograph, determining the skeleton of the tooth, setting the pixel representation information of the pixels in the central axis region corresponding to the skeleton to a first value, and setting the pixel representation information of the pixels in the boundary region of the tooth to a second value, thereby obtaining the tooth morphology map; the boundary region is the area of ​​the tooth other than the central axis region; the first value is higher than the second value. This scheme can accurately and effectively determine the tooth morphology map, thereby improving the accuracy of tooth segmentation.

[0072] In one possible implementation, the tooth skeleton extraction algorithm includes one or more of the following: a thinning algorithm, a distance transformation algorithm, a median transformation algorithm, and a neural network algorithm. The thinning algorithm is a pixel-based skeleton extraction algorithm that iteratively thins the edges of the teeth into a skeleton structure. The distance transformation algorithm is a distance field-based skeleton extraction algorithm that calculates the distance from the tooth surface to the background to transform the tooth edges into a skeleton structure. The median transformation algorithm is a median-axis-based skeleton extraction algorithm that calculates the distance from the tooth surface to the background and the normal vector of the tooth surface to transform the changes in the teeth into a skeleton structure. The neural network algorithm is a method of extracting the tooth skeleton structure through a skeleton extraction model, where the skeleton extraction model is obtained by training the original tooth image and the extracted skeleton image. This application does not limit the tooth skeleton extraction algorithm. For example, the tooth morphology image obtained after extracting the tooth skeleton for tooth A is shown in Figure 5.

[0073] In one embodiment, step 101 above, determining the tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph, includes: for any tooth in the panoramic oral radiograph, determining the distance information from any pixel point of the tooth to the nearest tooth boundary point, and determining the pixel representation information of any pixel point based on the distance information, thereby obtaining a first morphology map; for any tooth in the panoramic oral radiograph, determining the skeleton of the tooth, and setting the pixel representation information of the pixels corresponding to the skeleton to the maximum value and the pixel representation information of the pixels not corresponding to the skeleton to the minimum value, thereby obtaining a second morphology map; performing an OR operation on the same pixel point in the first morphology map and the second morphology map to determine the tooth morphology map. This scheme effectively represents the distance from any pixel point of the tooth to the nearest tooth boundary point, but the distance information may blur sharper areas of the tooth, such as the root; while the tooth skeleton map can extract the skeleton information of the tooth, especially the finer areas at the root; therefore, combining the tooth distance information and skeleton information can better represent the morphological information of the tooth, and can also effectively distinguish tooth instances even in cases of severe tooth overlap, thus helping to improve the accuracy of tooth detection.

[0074] In one possible implementation method, the calculation method for the tooth morphology diagram is shown in formula (1). M =I S |I D ……(1)

[0075] Among them, I S This is the first morphological diagram, I D For the second morphological diagram, I M The image represents a tooth morphology diagram, and | represents the OR operation on the image.

[0076] For example, as shown in Figure 6, which is the process of generating a single tooth morphology image, merging the tooth morphology images of all teeth can yield a tooth morphology image of all teeth, as shown in Figure 7.

[0077] In one possible implementation, the panoramic oral radiograph includes not only the tooth region but also other regions to improve the accuracy of tooth segmentation. Before step 101 above, a tooth region image can be determined. The method for determining the tooth region image is shown in Figure 8 and includes the following steps:

[0078] Step 801: Determine the binary image corresponding to the panoramic oral radiograph.

[0079] Wherein, the first pixel value of the binary image corresponds to the tooth region of the panoramic oral radiograph; the second pixel value of the binary image corresponds to the non-tooth region of the panoramic oral radiograph.

[0080] In one possible implementation, the panoramic oral radiograph is input into a global tooth segmentation model to obtain a binary image corresponding to the panoramic oral radiograph, as shown in Figure 3. This application does not limit the method used to determine the binary image corresponding to the panoramic oral radiograph.

[0081] In one possible implementation, the global tooth segmentation model is a fully convolutional deep neural network. The input is a normalized original panoramic image, and the output is a binary mask image of the teeth. In the mask image, the first pixel value of 1 represents a tooth, and the second pixel value of 0 represents the background. The fully convolutional deep neural network can be a general medical image segmentation network such as UNet, Atten-UNet, UNet++, etc. Of course, other neural network models can also be used, and this application does not limit them.

[0082] Step 802: Determine the smallest region containing all teeth based on the binary image.

[0083] In one possible implementation, the vertex coordinates x1, y1, x2, y2 of the smallest rectangular bounding box containing all the tooth regions are determined based on the binary image, where (x1, y1) are the x-axis and y-axis coordinates of the upper left vertex of the rectangle, and (x2, y2) are the x-axis and y-axis coordinates of the lower right vertex of the rectangle. The smallest region containing the teeth is shown as the smaller rectangle in Figure 9. This application uses the smallest rectangular bounding box as an example, but the shape of the smallest region containing the teeth is not limited.

[0084] Step 803: Expand the minimum region to obtain the overall region where the tooth is located.

[0085] In one possible implementation, an extended rectangle is used to expand the minimum region, so that the rectangle covers part of the image surrounding the tooth. The coordinates of the upper left vertex of the expanded rectangle are (x1-margin, y1-margin), and the coordinates of the lower right vertex are (x2+margin, y2+margin), where margin is the size of the expanded region, for example, a margin of 0.2×(y2-y1); of course, the size of the expanded region can also be other values, which are not limited in this application. This scheme expands the minimum region so that the overall region containing the tooth includes both the tooth region and the background region, enabling better differentiation between the tooth region and the background region during subsequent tooth segmentation, thereby improving the accuracy of tooth segmentation. The overall region containing the tooth is shown as the larger rectangle in Figure 9.

[0086] Step 804: Cropping the entire area containing the teeth in the panoramic oral radiograph to obtain a tooth area image.

[0087] In one possible implementation, a local interest image of the entire tooth region is cropped from the original panoramic oral radiograph based on an expanded rectangular bounding box to obtain a tooth region image.

[0088] In one possible implementation, a local interest image of the entire tooth region is cropped from the binary image based on an expanded rectangular bounding box to obtain a binary image of the tooth region.

[0089] Step 805: Determine the tooth morphology map based on the morphological information of each tooth in the tooth region image.

[0090] The above method determines the tooth morphology map based on the image of the tooth region, which can eliminate the interference of non-tooth regions in the panoramic oral radiograph, and thus achieve accurate and effective determination of the tooth morphology map.

[0091] In one embodiment, the initial position of each tooth in the tooth morphology image is obtained by performing a bitwise AND operation on the same pixel in the binary image and the tooth morphology image to determine a third morphology image; and then locating the teeth in the third morphology image to obtain the initial position of each tooth in the tooth morphology image. Performing a bitwise AND operation on the same pixel in the binary image and the tooth morphology image can denoise the tooth morphology image.

[0092] Optionally, a bitwise AND operation is performed on the same pixel in the binary image of the tooth region and the tooth morphology image to determine a third morphology image.

[0093] Optionally, the method for determining the third morphological diagram is shown in formula (2). MP =I M &M……(2)

[0094] Among them, I MP This is the third morphological diagram, I. M M is a tooth morphology image, M is a binary image, and & represents the AND operation on the image.

[0095] In one embodiment, step 103 above, obtaining the region of interest (ROI) of each tooth from the panoramic oral radiograph based on the initial position of each tooth in the tooth morphology image, includes: inputting the tooth morphology image into a tooth detection and classification model to obtain the initial position of each tooth in the tooth morphology image; and obtaining the ROI of each tooth from the panoramic oral radiograph based on the initial position. Specifically, the tooth morphology image is input into a tooth detection and classification model based on a deep neural network to obtain the bounding box coordinates and tooth category of each individual tooth. Then, based on the bounding box coordinates of each tooth, a local ROI image of a single tooth region is cropped from the original panoramic oral radiograph image, and a coarse segmentation region of the single tooth region is cropped from the binary image.

[0096] In one possible implementation, a tooth detection and classification model based on a deep neural network takes as input a morphological image of all teeth and outputs the vertex coordinates x1, y1, x2, y2 of the smallest rectangle containing each individual tooth region and / or the category of each individual tooth. Here, (x1, y1) are the x-axis and y-axis coordinates of the top-left vertex of the rectangle, and (x2, y2) are the x-axis and y-axis coordinates of the bottom-right vertex of the rectangle, as shown by the smaller dashed rectangle in Figure 10. The tooth categories include maxillary incisors, maxillary canines, maxillary premolars, maxillary molars, maxillary wisdom teeth, maxillary deciduous teeth, mandibular incisors, mandibular canines, mandibular premolars, mandibular molars, mandibular wisdom teeth, mandibular deciduous teeth, and impacted teeth.

[0097] In another possible implementation, a tooth detection and classification model based on a deep neural network takes as input a morphological image of all teeth and a panoramic radiograph of the oral cavity, or a dual-channel image composed of a morphological image and a panoramic radiograph of the oral cavity, and outputs the vertex coordinates of the smallest rectangle containing each individual tooth region and / or the category of each individual tooth.

[0098] In one possible implementation, the tooth detection and classification model can employ general object detection networks such as Faster-RCNN, YOLO, and RetinaNet. This application does not limit this approach.

[0099] In one possible implementation, after obtaining the minimum bounding box for each tooth, the minimum bounding box is expanded to include a portion of the area surrounding the tooth, as shown by the larger dashed rectangle in Figure 10. The coordinates of the top-left vertex of the expanded rectangle are (x1-margin_x, y1-margin_y), and the coordinates of the bottom-right vertex are (x2+margin_x, y2+margin_y), where margin_x and margin_y are the sizes of the expanded regions along the x and y axes, respectively. For example, margin_x can be taken as 0.1×(x2-x1), and margin_y as 0.2×(y2-y1). This application does not limit the size of the expanded region.

[0100] In one embodiment, step 104 above, which involves segmenting the region of interest (ROI) of each tooth to obtain a segmentation result for each tooth, includes the following steps: inputting the ROI of each tooth into a first segmentation model to obtain a segmentation result for each tooth. This application does not limit the network structure corresponding to the first segmentation model; the first segmentation model can be UNet, Atten-UNet, UNet++, or any encoder-decoder network, etc.

[0101] In one embodiment, step 104 includes: cropping the region of interest (ROI) of a single tooth in the binary image corresponding to the panoramic oral radiograph to obtain a coarse segmentation region for each tooth; for any given tooth, inputting the ROI and the coarse segmentation region into a second segmentation model, and outputting the segmentation result of the tooth. This embodiment differs from the previous embodiment in that it simultaneously inputs both the coarse segmentation region and the ROI into the second segmentation model.

[0102] In one possible implementation, the second segmentation model includes multiple convolutional layers, wherein each convolutional layer is followed by a feature enhancement layer. The input to the feature enhancement layer is the output of the previous convolutional layer and a coarse segmentation region of the tooth. The feature enhancement layer is used to enhance tooth feature information based on the coarse segmentation region of the tooth. This scheme can employ a tooth morphology-guided feature enhancement module after each convolutional layer to better extract effective tooth feature information from the original image, thereby improving the accuracy of tooth segmentation.

[0103] In one possible implementation, the feature enhancement layer includes the following steps: performing a convolution operation on the coarsely segmented region of the tooth to obtain a first feature image; performing an addition operation on the output of the previous convolutional layer of the feature enhancement layer and the same pixel of the first feature image to obtain a second feature image; determining the attention coefficient of the region of interest of the tooth based on the activation function and the second feature image; and determining the output of the feature enhancement layer based on the attention coefficient of the region of interest of the tooth and the output of the previous convolutional layer of the feature enhancement layer.

[0104] Optionally, the output of the convolutional layer preceding the feature enhancement layer and the same pixel of the first feature image are concatenated along the channel dimension to obtain a second feature map. This application does not limit the method for determining the second feature map.

[0105] This application does not limit the network structure corresponding to the second segmentation model. The second segmentation model can be UNet, Atten-UNet, UNet++, or any encoder-decoder network. The second segmentation model will be explained using an encoder-decoder architecture as an example.

[0106] In one possible implementation, the second segmentation model is shown in Figure 11. The second segmentation model includes an encoder module and a decoder module. The encoder module includes multiple convolutional modules, each followed by a feature enhancement module, and each feature enhancement module is followed by a pooling layer. The decoder module includes multiple deconvolutional layers and convolutional modules.

[0107] Optionally, the feature enhancement module can also be set in the decoder module. For example, the feature enhancement module can be set between the deconvolution layer and the convolution module. This application does not limit this.

[0108] In one possible implementation, the network architecture corresponding to the feature enhancement module is shown in Figure 12. Here, X represents the output of the previous convolutional layer, and M represents the coarse segmentation region of the tooth, as shown by the dashed box in Figure 12. In existing technologies, the input to the next network layer is often the output of the previous network layer; that is, the gate signal of the attention module usually comes from the feature map of a deep layer in the network. In this application, the input to the next network layer, in addition to receiving the output of the previous network layer, also receives an image with tooth morphology features. This means that a feature enhancement module based on tooth morphology guidance is used to better extract effective tooth feature information from the original image, thereby improving the accuracy of tooth segmentation.

[0109] In one possible implementation, the feature enhancement layer includes the following steps:

[0110] a) Apply a convolution operation with a kernel size of 1×1 to the input feature image X to obtain the feature image X1.

[0111] b) Apply a convolution operation with a kernel size of 3×3, a normalization operation, a ReLU activation function, and a convolution operation with a kernel size of 1×1 to the coarse segmentation region M of the input teeth in sequence to obtain the feature map M1.

[0112] c) Perform pixel-wise addition on X1 and M1 to obtain feature map X2.

[0113] d) Apply the ReLU activation function, a 1×1 convolution operation, and the Sigmoid activation function sequentially to the feature map X2 to obtain the attention coefficient image A based on tooth morphology.

[0114] e) Apply pixel-based multiplication to the attention coefficient image A and the input feature image X to obtain the feature image X. A For feature image X and feature image X A The enhanced feature image X is obtained by applying pixel-based addition operations. E .

[0115] In one embodiment, step 104 includes: cropping individual teeth from the tooth morphology image to obtain the morphological region of each tooth; for any tooth, inputting the region of interest / coarse segmentation region of the tooth and the morphological region of the tooth into a third segmentation model, and outputting the segmentation result of the tooth. This application does not limit the network structure corresponding to the third segmentation model, which can be UNet, Atten-UNet, UNet++, or any encoder-decoder network, etc.

[0116] Optionally, the region of interest of the tooth, the coarse segmentation region of the tooth, and the morphological region of the tooth are input into the third segmentation model to output the segmentation result of the tooth.

[0117] Optionally, the region of interest and the morphological region of the tooth are input into a third segmentation model to output the segmentation result of the tooth.

[0118] In one embodiment, the segmentation result of each tooth is obtained by the following method, as shown in Figure 13, which includes the following steps:

[0119] Step 1301: Obtain a panoramic oral radiograph of the patient and input the panoramic oral radiograph into the first model to obtain a binary image.

[0120] In one possible implementation, the first model can be a general medical image segmentation network such as UNet, Atten-UNet, UNet++, etc. This application does not limit this.

[0121] Step 1302: Determine the overall region where the patient's teeth are located based on the binary image.

[0122] Step 1303: Based on the overall area where the teeth are located, crop the entire teeth in the panoramic oral radiograph to obtain an image of the patient's tooth area.

[0123] Step 1304: Input the image of the tooth region into the second model to obtain a tooth morphology image.

[0124] The second model is trained based on the distance information from any pixel of each tooth to the nearest tooth boundary point, as well as the skeleton information of each tooth.

[0125] In one possible implementation, the second model can be a general medical image segmentation network such as UNet, Atten-UNet, UNet++, etc. This application does not limit this.

[0126] Step 1305: Input the tooth morphology image into the third model to obtain the initial position information of each tooth in the tooth region map.

[0127] The third model is used to predict the positional information of each tooth in the tooth morphology diagram.

[0128] In one possible implementation, the third model can be a general object detection network such as Faster-RCNN, YOLO, RetinaNet, etc. This application does not limit this.

[0129] Step 1306: Based on the initial position information, crop each tooth in the tooth region image to obtain the region of interest for each tooth in the tooth region image.

[0130] Step 1307: Based on the initial position information, each tooth in the binary image is cropped to obtain a coarse segmentation region for each tooth in the binary image.

[0131] Step 1308: For any tooth, input the region of interest and the coarse segmentation region of each tooth into the fourth model to determine the segmentation result of each tooth in the tooth region image.

[0132] The fourth model includes multiple convolutional layers, with each convolutional layer followed by a feature enhancement layer. The input to the feature enhancement layer is the output of the previous convolutional layer and a coarse segmentation region of the tooth. The feature enhancement layer is used to enhance the tooth feature information based on the coarse segmentation region of the tooth.

[0133] In one possible implementation, the fourth model can be an encoder-decoder combination. This application does not limit this approach.

[0134] In one embodiment, the segmentation results of each tooth in a panoramic radiograph of the patient's mouth are displayed, as shown in Figure 14.

[0135] One possible implementation involves displaying a panoramic radiograph of the patient's mouth, along with the segmentation results of each tooth within the panoramic radiograph.

[0136] In one possible implementation, the segmentation results of the corresponding tooth type and / or tooth number are displayed based on the tooth type and / or tooth number in the patient's panoramic oral radiograph.

[0137] One possible implementation involves displaying at least two images from the patient's panoramic oral radiograph, binary image, tooth morphology map, or final tooth segmentation result, and showing the comparison effect of the different images.

[0138] The above method can help doctors and patients intuitively understand the segmentation results of each tooth.

[0139] Panoramic radiographs are an important examination technique in oral examinations. Through a single imaging process, panoramic radiographs can conveniently and quickly display the teeth and surrounding anatomical structures throughout the oral cavity, providing effective assistance in the diagnosis of oral diseases.

[0140] The degree of alveolar bone resorption in panoramic radiographs is an important reference for developing orthodontic treatment plans. Current panoramic radiograph methods for grading alveolar bone resorption have low accuracy and are insufficient to meet the needs of dentists.

[0141] Figure 15 shows a flowchart of a method for determining the degree of alveolar bone resorption according to an embodiment of this application, which specifically includes the following steps:

[0142] Step S1501: Obtain a panoramic dental radiograph.

[0143] Specifically, a digital panoramic dental X-ray device can be used to take panoramic images of the patient's teeth. In this application, the shooting angle of the panoramic dental images is not limited. Panoramic dental images at a specific angle can be taken according to the actual diagnostic needs.

[0144] Step S1502: Input the panoramic image of the teeth into the segmentation model to determine the tooth contour curve, the first contour line and the second contour line; the first contour line is the outer edge line of the crown portion of the tooth including at least two teeth, and the second contour line is the outer edge line of the alveolar bone of the maxilla and mandible; the tooth contour curve is the curve formed by the outer surface edge of each tooth.

[0145] Specifically, after normalizing the panoramic dental images, they are input into the segmentation model to obtain binary mask images of each tooth (as shown in Figure 16), binary mask images of the crowns of all teeth and the alveolar bone of all teeth (as shown in Figure 17). In these binary mask images, 1 represents the segmented target object and 0 represents the background. The edge lines of these mask images are then obtained to obtain the tooth contour curve of each tooth (as shown in Figure 18), the first contour line of all teeth, and the second contour line of all teeth (as shown in Figure 19).

[0146] One possible implementation is to use two segmentation models: a tooth segmentation model and a crown and alveolar bone segmentation model. That is, the panoramic images of the teeth are input into the tooth segmentation model and the crown and alveolar bone segmentation model respectively to obtain a binary mask image of each tooth, a binary mask image of the crown of all teeth, and a binary mask image of the alveolar bone.

[0147] Optionally, the tooth segmentation model can adopt, but is not limited to, the following implementation methods: (1) Using a general instance segmentation network, such as the deep learning model Mask R-CNN (Mask Region-based Convolutional Neural Network). (2) Using a two-stage network: the first stage is a tooth localization network, such as using object detection algorithms such as Faster-RCNN, YOLO, RetinaNet, etc., and the second stage is a tooth semantic segmentation network, such as using fully convolutional neural networks UNet, Atten-UNet, UNet++, etc. (3) Using a multi-channel output semantic segmentation network, where each output channel is a binary mask image of a tooth. The semantic segmentation network can adopt a general segmentation network, such as fully convolutional neural networks UNet, Atten-UNet, UNet++, etc.

[0148] Optionally, the crown and alveolar bone segmentation model can be implemented in, but is not limited to, the following ways: (1) using a multi-class semantic segmentation network. (2) using a multi-label semantic segmentation network. (3) using two separate semantic segmentation networks, the first network for alveolar bone segmentation and the second network for crown segmentation. The semantic segmentation network can be a general segmentation network, such as a fully convolutional neural network UNet, Atten-UNet, UNet++, etc.

[0149] One possible implementation is that the segmentation model is determined based on the process of the tooth segmentation method described above.

[0150] One possible implementation is that the segmentation model can be divided into three parts: a tooth segmentation model, a crown segmentation model, and an alveolar bone segmentation model. That is, the panoramic images of the teeth are input into the tooth segmentation model, the crown segmentation model, and the alveolar bone segmentation model respectively to obtain a binary mask image of each tooth, a binary mask image of the crown of all teeth, and a binary mask image of the alveolar bone.

[0151] The following is an embodiment provided by this application. The instance segmentation network Mask R-CNN is selected as the tooth segmentation model, the semantic segmentation network UNet is selected as the crown segmentation model, and the semantic segmentation network UNet++ is selected as the alveolar bone segmentation model. The panoramic images of the teeth are input into the tooth segmentation model, the crown segmentation model, and the alveolar bone segmentation model, respectively, to obtain binary mask images of each tooth, binary mask images of the crowns of all teeth, and binary mask images of the alveolar bones of the maxilla and mandible. The edge lines of these mask images are obtained, and the tooth contour curve of each tooth, the outer edge line of the crown portion of all teeth (first contour line), and the outer edge line of the alveolar bone of the maxilla and mandible (second contour line) are obtained.

[0152] Based on the tooth contour curve, first contour line and second contour line obtained in step S1502 above, step S1503 is used to make a judgment. For any tooth, the degree of alveolar bone resorption is determined based on the distribution area of ​​the first contour line and the second contour line within the tooth contour curve.

[0153] By identifying the distribution areas of the first and second contour lines within the tooth contour curve of each tooth, the degree of alveolar bone resorption for each individual tooth is determined, improving the accuracy and efficiency of assessing alveolar bone resorption. This application provides a fully automated process for determining alveolar bone resorption, requiring no manual interaction. For example, on a computer equipped with an Intel Core i7-7800X (3.50GHz) CPU and an NVIDIA 1080Ti GPU, the average recognition time is approximately 3 seconds, representing a significant improvement in recognition efficiency.

[0154] Step S1503 above includes: determining the distribution area of ​​the first contour line and the second contour line within the tooth contour curve; determining the degree of alveolar bone resorption of the tooth based on the area of ​​the distribution area and / or the length of the reference line within the distribution area.

[0155] One possible implementation involves the first and second contour lines forming several distribution regions within the contour line of any tooth, as shown in Figure 20. These distribution regions exhibit irregular shapes, and the degree of alveolar bone resorption can be determined based on the proportional relationship between their areas. For example, the degree of alveolar bone resorption can be determined by the proportional relationship between the areas of distribution region 2 and distribution region 3. When the ratio of the area of ​​distribution region 2 to the area of ​​distribution region 3 is less than a certain value, it indicates that the degree of alveolar bone resorption is relatively mild; if the ratio of the area of ​​distribution region 2 to the area of ​​distribution region 3 is greater than a set threshold, it indicates that the alveolar bone resorption is relatively severe. This application does not limit the specific calculation method for the area of ​​the distribution regions; mathematical calculations can be performed on the area of ​​the distribution regions according to actual usage needs.

[0156] Another method for determining the degree of alveolar bone resorption is to establish reference lines on each tooth, such as the main axis of each tooth or a line connecting any two points in the distribution area. The degree of alveolar bone resorption is then determined by setting the length of these reference lines within the distribution area.

[0157] Optionally, the degree of alveolar bone resorption can be determined based on the length of the reference line within the distribution area. This can be achieved in various ways, with the following four examples illustrating the methods:

[0158] Method 1: If the reference line is set as the main axis of each tooth, the degree of alveolar bone resorption is determined based on the length of the tooth's main axis within the distribution area.

[0159] Specifically, by observing tooth morphology, determining reference planes (occlusal plane and midline plane), and determining the direction of the principal axes (root and crown directions), the principal axes of each tooth are drawn and adjusted using tools; or the principal component analysis algorithm is used to automatically calculate the principal axes of each tooth, resulting in a schematic diagram of the principal axes of all teeth as shown in Figure 21. In any tooth schematic diagram as shown in Figure 22, the principal axis intersects the first and second contour lines within the tooth contour curve, producing two intersection points, C and A. The degree of alveolar bone resorption can be determined based on the length between C and A. A length threshold can also be set; if the length between C and A is less than the threshold, the degree of alveolar bone resorption is determined to be mild; if the length between C and A is greater than the threshold, the degree of alveolar bone resorption is determined to be severe. Furthermore, length ranges can be set, with different length ranges corresponding to different resorption levels.

[0160] In another embodiment, the minimum area rectangle of each tooth can be obtained based on the tooth contour curve shown in Figure 18. Figure 23 shows the minimum area rectangle of each tooth. In the tooth diagram shown in Figure 22, the main axis of the tooth intersects with the minimum area rectangle, producing two intersection points, M and R. The degree of alveolar bone resorption for each tooth can be determined based on the ratio between the lengths of the CA and CR segments on the main axis, or the ratio between the lengths of the CA and AR segments. Determining the degree of alveolar bone resorption based on the length of the tooth's main axis within its distribution area can be achieved using methods other than those described above, which are not specifically limited in this application.

[0161] Method 2: Determine the first length of the tooth contour curve in the mesial direction within the distribution area, and determine the degree of alveolar bone resorption based on the first length.

[0162] Specifically, as shown in Figure 22, based on the length of the curve segment between C1 and A1 on the tooth contour curve, this length is recorded as the first length. A first length threshold can be set. If the length of the curve segment between C1 and A1 is less than the first length threshold, the degree of alveolar bone resorption is determined to be mild; if the length of the curve segment between C1 and A1 is greater than the first length threshold, the degree of alveolar bone resorption is determined to be severe. Alternatively, the degree of alveolar bone resorption can be determined based on the proportional relationship between the length of the curve segment between C1 and A1 and the length of the curve segment between C1 and R. Furthermore, a first length range can be set, with different first length ranges corresponding to different resorption levels. Other calculation methods besides the above are also included, but are not specifically limited in this application.

[0163] In another implementation, based on the first intersection point C1 between the tooth contour curve and the first contour line in the mesial direction, and the second intersection point A1 between the tooth contour curve and the second contour line in the mesial direction, the distance C1A1 between the first and second intersection points is used as the first length, where C1A1 is the length of alveolar bone resorption in the mesial direction. The length of the straight line segment between C1A1 is recorded as the first length. A length threshold can be set; if the length of the straight line segment between C1A1 is less than the length threshold, the degree of alveolar bone resorption is determined to be mild; if the length of the straight line segment between C1A1 is greater than the first length threshold, the degree of alveolar bone resorption is determined to be severe. Alternatively, the degree of alveolar bone resorption can be determined based on the proportional relationship between the length of the straight line segment between C1A1 and the length of the straight line segment between C1R.

[0164] Method 3: Determine the second length of the tooth contour curve in the distal direction within the distribution area, and determine the degree of alveolar bone resorption based on the second length.

[0165] As shown in Figure 22, the mesial direction of the tooth contour curve intersects with the first and second contour lines, creating two intersection points, C2 and A2. The length of the curve segment between C2 and A2 on the tooth contour curve can be designated as the second length. Two second length thresholds can be set. If the length of the curve segment between C2 and A2 is less than the second length threshold, the alveolar bone resorption is determined to be mild; if the length is greater than the second length threshold, the alveolar bone resorption is determined to be severe. Alternatively, the degree of alveolar bone resorption can be determined based on the proportional relationship between the length of the curve segment between C2 and A2 and the length of the curve segment between C1 and R. Furthermore, second length intervals can be set, with different intervals corresponding to different resorption levels. Other calculation methods besides those described above are also included, but are not specifically limited in this application.

[0166] In another implementation, the third intersection point C2 between the tooth contour curve and the first contour line in the distal direction, and the fourth intersection point A2 between the tooth contour curve and the second contour line in the distal direction are determined. The distance C2A2 between the third intersection point C2 and the fourth intersection point A2 is taken as the second length, and C2A2 is the length of alveolar bone resorption in the distal direction. The length of the straight line segment between C2A2 is recorded as the first length. A length threshold can be set. If the length of the straight line segment between C2A2 is less than the length threshold, the degree of alveolar bone resorption is determined to be mild; if the length of the straight line segment between C2A2 is greater than the first length threshold, the degree of alveolar bone resorption is determined to be severe. Alternatively, the degree of alveolar bone resorption can be determined based on the proportional relationship between the length of the straight line segment between C2A2 and the length of the straight line segment between C2R.

[0167] Method 4: Determine the degree of alveolar bone resorption of the tooth based on the first and second lengths.

[0168] In one embodiment, the degree of alveolar bone resorption of a tooth can be determined based on the longest of the first and second lengths obtained through methods two and three. For example, if the first length C1A1 is 1 mm and the second length C2A2 is 2.5 mm, the degree of alveolar bone resorption of the tooth can be determined based on the second length C2A2.

[0169] In another implementation, the degree of alveolar bone resorption of the tooth is determined based on the average length of the first length and the second length. For example, if the first length C1A1 is 1 mm and the second length C2A2 is 2.5 mm, the average length is calculated to be 1.75 mm, and the degree of alveolar bone resorption of the tooth is determined based on the average length.

[0170] By determining the distribution area of ​​the first and second contour lines within the tooth contour curve of each tooth, the degree of alveolar bone resorption for each individual tooth can be determined, thus improving the accuracy of alveolar bone resorption determination.

[0171] In another implementation, the root length of the tooth is determined based on the tooth's main axis and a first contour line; a first ratio of the first length to the root length and a second ratio of the second length to the root length are determined; and the degree of alveolar bone resorption is determined based on the largest of the first and second ratios.

[0172] Specifically, as shown in Figure 22, a minimum area rectangle is determined based on the tooth outline. Two intersection points are determined between the tooth's main axis and the minimum area rectangle: the intersection point in the crown direction is denoted as M, and the intersection point in the root direction is denoted as R. The root length CR is obtained based on the intersection point C of the main axis and the first outline. The first ratio of the first length C1A1 to the root length CR is calculated and denoted as R1. The first ratio is the alveolar bone resorption ratio in the mesial direction, C1A1 / CR. The second ratio of the second length C2A2 to the root length CR is calculated and denoted as R2. The second ratio is the alveolar bone resorption ratio in the distal direction, C2A2 / CR. The larger ratio between the first and second ratios is taken as the alveolar bone resorption ratio of the tooth.

[0173] Optionally, the largest ratio between the first ratio and the second ratio is compared with each preset standard; wherein each preset standard corresponds to a different alveolar bone resorption grade; and the alveolar bone resorption grade of the tooth is determined based on the comparison results.

[0174] Specifically, the maximum value of the first ratio of mesial alveolar bone resorption to the second ratio of distal alveolar bone resorption is denoted as RMax. RMax is then used to classify alveolar bone resorption according to preset rules. For example, the preset rules can classify the degree of alveolar bone resorption into four levels: no resorption, mild resorption, moderate resorption, and severe resorption. If RMax < 1 / 6, it is considered no resorption; if 1 / 6 ≤ RMax < 1 / 2, it is considered mild resorption; if 1 / 2 ≤ RMax < 2 / 3, it is considered moderate resorption; and if RMax ≥ 2 / 3, it is considered severe resorption.

[0175] The degree of alveolar bone resorption is graded for each individual tooth, which improves the accuracy of alveolar bone resorption grading.

[0176] Based on the same technical concept, Figure 24 exemplarily illustrates a tooth segmentation device 2400 provided in an embodiment of this application. As shown in Figure 24, it includes: a morphology determination unit 2401 and a tooth segmentation unit 2402; the morphology determination unit 2401 is used to determine a tooth morphology map based on the morphological information of each tooth in a panoramic oral radiograph; the pixel representation information of each tooth in the morphology map is higher in the central axis region than in the boundary region; the tooth segmentation unit 2402 is used to perform tooth localization on the morphology map to obtain the initial position of each tooth in the morphology map; based on the initial position of each tooth in the morphology map, the region of interest of each tooth is obtained from the panoramic oral radiograph; and the segmentation result of each tooth is obtained by segmenting the region of interest of each tooth.

[0177] In one possible implementation, the morphology determination unit 2401 is used to determine the distance information from any pixel point of any tooth to the nearest tooth boundary point for any tooth in a panoramic oral radiograph, and to determine the pixel representation information of any pixel point based on the distance information, thereby obtaining a tooth morphology map; wherein, the smaller the distance from the pixel point to the nearest tooth boundary point, the weaker the pixel representation information of the pixel point; the larger the distance from the pixel point to the nearest tooth boundary point, the stronger the pixel representation information of the pixel point.

[0178] In one possible implementation, the morphology determination unit 2401 is used to determine the skeleton of any tooth in a panoramic oral radiograph, and set the pixel representation information of the pixels in the central axis region corresponding to the skeleton to a first value, and set the pixel representation information of the pixels in the boundary region of the tooth to a second value, thereby obtaining a tooth morphology map; the boundary region is the region of the tooth other than the central axis region; the first value is higher than the second value.

[0179] In one possible implementation, the morphology determination unit 2401 is used to determine the distance information from any pixel point of any tooth to the nearest tooth boundary point for any tooth in a panoramic oral radiograph, and determine the pixel representation information of any pixel point based on the distance information to obtain a first morphology image; for any tooth in the panoramic oral radiograph, determine the skeleton of the tooth, and set the pixel representation information of the pixels corresponding to the skeleton to the maximum value and the pixel representation information of the pixels not corresponding to the skeleton to the minimum value to obtain a second morphology image; and perform an OR operation on the same pixel point in the first morphology image and the second morphology image to determine a tooth morphology image.

[0180] In one possible implementation, the tooth segmentation unit 2402 is used to determine the overall region where the teeth are located from the panoramic oral radiograph; cropping the overall region where the teeth are located in the panoramic oral radiograph to obtain a tooth region image; and the morphology determination unit 2401 is used to determine a tooth morphology image based on the morphological information of each tooth in the tooth region image.

[0181] In one possible implementation, the tooth segmentation unit 2402 is used to determine a binary image corresponding to the panoramic oral radiograph; the first pixel value of the binary image corresponds to the tooth region of the panoramic oral radiograph; the second pixel value of the binary image corresponds to the non-tooth region of the panoramic oral radiograph; based on the binary image, a minimum region containing all teeth is determined; and the minimum region is expanded to obtain the overall region containing the teeth.

[0182] In one possible implementation, the morphology determination unit 2401 is used to perform an AND operation on the same pixel point of the binary image and the tooth morphology image to determine a third morphology image; and to perform tooth localization on the third morphology image to obtain the initial position of each tooth in the tooth morphology image.

[0183] In one possible implementation, the tooth segmentation unit 2402 is used to input the region of interest of each tooth into the first segmentation model to obtain the segmentation result of each tooth.

[0184] In one possible implementation, the tooth segmentation unit 2402 is used to crop the region of interest of a single tooth in the binary image corresponding to the panoramic oral radiograph to obtain a coarse segmentation region for each tooth; for any tooth, the region of interest and the coarse segmentation region of the tooth are input into a second segmentation model to output the segmentation result of the tooth.

[0185] In one possible implementation, the second segmentation model includes multiple convolutional layers, wherein each convolutional layer is followed by a feature enhancement layer; the input to the feature enhancement layer is the output of the previous convolutional layer and a coarse segmentation region of the tooth; the feature enhancement layer is used to enhance tooth feature information based on the coarse segmentation region of the tooth.

[0186] In one possible implementation, the tooth segmentation unit 2402 is used to perform a convolution operation on the coarsely segmented region of the tooth to obtain a first feature image; add the output of the previous convolutional layer of the feature enhancement layer to the same pixel of the first feature image to obtain a second feature image; determine the attention coefficient of the region of interest of the tooth according to the activation function and the second feature image; and determine the output of the feature enhancement layer according to the attention coefficient of the region of interest of the tooth and the output of the previous convolutional layer of the feature enhancement layer.

[0187] In one possible implementation, the tooth segmentation unit 2402 is used to crop individual teeth from the tooth morphology image to obtain the morphological region of each tooth; for any tooth, the region of interest / coarse segmentation region of the tooth and the morphological region of the tooth are input into the third segmentation model, and the segmentation result of the tooth is output.

[0188] In one possible implementation, the device further includes a display unit 2403 for displaying the segmentation results of each tooth in the patient's panoramic oral radiograph.

[0189] Based on the same technical concept, this application provides a tooth segmentation device 2500, which can be, for example, a computing device. As shown in FIG25, a tooth segmentation device 2500 includes at least one processor 2501 and a memory 2502 connected to the at least one processor. This application does not limit the specific connection medium between the processor 2501 and the memory 2502; FIG25 shows an example where the processor 2501 and the memory 2502 are connected via a bus. The bus can be divided into an address bus, a data bus, a control bus, etc.

[0190] In this embodiment of the application, the memory 2502 stores instructions that can be executed by at least one processor 2501. By executing the instructions stored in the memory 2502, at least one processor 2501 can execute the above-described tooth segmentation method.

[0191] The processor 2501 serves as the control center of the tooth segmentation device 2500. It can connect to various parts of a computer device via various interfaces and lines, and performs resource settings by running or executing instructions stored in the memory 2502 and calling data stored in the memory 2502. Optionally, the processor 2501 may include one or more determining units. The processor 2501 may integrate an application processor and a modem processor, wherein the application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 2501. In some embodiments, the processor 2501 and the memory 2502 may be implemented on the same chip; in some embodiments, they may also be implemented on separate chips.

[0192] Processor 2501 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0193] Memory 2502, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 2502 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage, magnetic disk, optical disk, etc. Memory 2502 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 2502 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0194] This application also provides a computer-readable storage medium storing a computer-executable program for causing a computer to perform a tooth segmentation method listed in any of the above embodiments.

[0195] This application provides a computer program product, including a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform a tooth segmentation method listed in any of the above embodiments.

[0196] As shown in Figure 26, an apparatus 2600 for determining the degree of alveolar bone resorption according to an embodiment of this application includes:

[0197] Module 2601 is used to acquire panoramic dental radiographs;

[0198] Processing module 2602 is used to input the panoramic image of the teeth into a segmentation model to determine the tooth contour curve, the first contour line of all teeth, and the second contour line of all teeth; the first contour line includes the outer edge line of the crown portion of at least two teeth, and the second contour line is the outer edge line of the alveolar bone of the maxilla and mandible; the tooth contour curve is the curve formed by the outer surface edge of each tooth.

[0199] The judgment module 2603 is used to determine the degree of alveolar bone resorption of any tooth based on the distribution area of ​​the first contour line and the second contour line within the tooth contour curve of the tooth.

[0200] By identifying the distribution areas of the first and second contour lines within the tooth contour curve of each tooth, the degree of alveolar bone resorption for each individual tooth can be determined, thus improving the accuracy and efficiency of assessing the degree of alveolar bone resorption.

[0201] Optionally, the determination module 2603 is specifically used for:

[0202] Determining the degree of alveolar bone resorption of the tooth based on the distribution area of ​​the first contour line and the second contour line within the tooth contour curve includes:

[0203] Determine the distribution areas of the first contour line and the second contour line within the tooth contour curve of the tooth.

[0204] The degree of alveolar bone resorption of the tooth is determined based on the area of ​​the distribution region and / or the length of the reference line within the distribution region.

[0205] Optionally, the determination module 2603 is specifically used for:

[0206] Determining the degree of alveolar bone resorption of the tooth based on the length of a set reference line within the distribution area includes at least one of the following:

[0207] The degree of alveolar bone resorption of the tooth is determined based on the first length;

[0208] The degree of alveolar bone resorption of the tooth is determined based on the second length;

[0209] The degree of alveolar bone resorption of the tooth is determined based on the first length and the second length;

[0210] The first length is the length of the tooth contour curve in the mesial direction within the distribution area, and the second length is the length of the tooth contour curve in the distal direction within the distribution area.

[0211] Optionally, the determination module 2603 is specifically used for:

[0212] Determining the first length of the tooth contour curve in the mesial direction within the distribution area includes:

[0213] Determine the first intersection point between the tooth contour curve and the first contour line in the mesial direction, and the second intersection point between the tooth contour curve and the second contour line in the mesial direction;

[0214] The distance between the first intersection point and the second intersection point is taken as the first length;

[0215] Determining the second length of the tooth contour curve in the distal direction within the distribution area includes:

[0216] Determine the third intersection point of the tooth contour curve and the first contour line in the distal direction, and the fourth intersection point of the tooth contour curve and the second contour line in the distal direction.

[0217] The distance between the third intersection point and the fourth intersection point is taken as the second length.

[0218] Optionally, the determination module 2603 is specifically used for:

[0219] Determining the degree of alveolar bone resorption of the tooth based on the first length and the second length includes at least one of the following:

[0220] The degree of alveolar bone resorption of the tooth is determined based on the longest of the first and second lengths.

[0221] The degree of alveolar bone resorption of the tooth is determined based on the average length of the first length and the second length.

[0222] Optionally, the determination module 2603 is specifically used for:

[0223] Determining the degree of alveolar bone resorption of the tooth based on the longest of the first and second lengths includes:

[0224] The root length of the tooth is determined based on the tooth's principal axis and the first contour line.

[0225] Determine a first ratio of the first length to the root length and a second ratio of the second length to the root length;

[0226] The degree of alveolar bone resorption of the tooth is determined based on the larger of the first ratio and the second ratio.

[0227] Optionally, the determination module 2603 is specifically used for:

[0228] Determining the degree of alveolar bone resorption of the tooth based on the larger of the first ratio and the second ratio includes:

[0229] The largest ratio between the first ratio and the second ratio is compared with each preset standard; wherein each preset standard corresponds to a different alveolar bone resorption grade.

[0230] Based on the comparison results, the alveolar bone resorption grade of the teeth was determined.

[0231] Based on the same technical concept, this application provides a computer device, as shown in FIG27, including at least one processor 2701 and a memory 2702 connected to at least one processor. This application does not limit the specific connection medium between the processor 2701 and the memory 2702; FIG27 shows an example where the processor 2701 and the memory 2702 are connected via a bus. The bus can be divided into address bus, data bus, control bus, etc.

[0232] In this embodiment of the application, the memory 2702 stores instructions that can be executed by at least one processor 2701. By executing the instructions stored in the memory 2702, at least one processor 2701 can perform the steps of the method described above for determining the degree of alveolar bone resorption.

[0233] The processor 2701 is the control center of the computer device. It can connect to various parts of the computer device using various interfaces and lines. By running or executing instructions stored in the memory 2702 and calling data stored in the memory 2702, it determines the degree of alveolar bone resorption of each tooth based on the distribution area of ​​the first and second contour lines within the tooth contour curve. Optionally, the processor 2701 may include one or more processing units. The processor 2701 may integrate an application processor and a modem processor. The application processor mainly handles the operating system, user interface, and applications, while the modem processor mainly handles wireless communication. It is understood that the modem processor may not be integrated into the processor 2701. In some embodiments, the processor 2701 and the memory 2702 may be implemented on the same chip; in some embodiments, they may also be implemented on separate chips.

[0234] Processor 2701 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0235] Memory 2702, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 2702 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic memory, magnetic disk, optical disk, etc. Memory 2702 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures that can be accessed by a computer, but is not limited thereto. In the embodiments of this application, memory 2702 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0236] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the method for determining the degree of alveolar bone resorption described above.

[0237] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0238] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to this application. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in one or more blocks of the flowchart illustrations and / or one or more blocks of the block diagrams.

[0239] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means that implement the functions specified in one or more flowcharts and / or one or more block diagrams.

[0240] These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer-implemented process, such that the instructions, which execute on the computer or other programmable apparatus, provide steps for implementing the functions specified in one or more flowcharts and / or one or more block diagrams.

[0241] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method for tooth segmentation, characterized in that, include: Based on the morphological information of each tooth in the panoramic radiograph, a tooth morphology diagram is determined. In the tooth morphology diagram, the pixel representation information of each tooth in the central axis region is higher than that in the boundary region; The tooth morphology diagram is used to locate the teeth, and the initial position of each tooth in the tooth morphology diagram is obtained. Based on the initial position of each tooth in the tooth morphology diagram, the region of interest for each tooth is obtained from the panoramic oral radiograph. The segmentation result of each tooth is obtained by segmenting the region of interest of each tooth.

2. The method as described in claim 1, characterized in that, The process of determining a tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph includes: For any tooth in the panoramic oral radiograph, the distance information from any pixel of the tooth to the nearest tooth boundary point is determined, and the pixel representation information of any pixel is determined based on the distance information, thereby obtaining the tooth morphology map; wherein, the smaller the distance from the pixel to the nearest tooth boundary point, the weaker the pixel representation information of the pixel; the larger the distance from the pixel to the nearest tooth boundary point, the stronger the pixel representation information of the pixel.

3. The method as described in claim 1, characterized in that, The process of determining a tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph includes: For any tooth in the panoramic oral radiograph, the skeleton of the tooth is determined, and the pixel representation information of the pixels in the central axis region corresponding to the skeleton is set to a first value, and the pixel representation information of the pixels in the boundary region of the tooth is set to a second value, thereby obtaining the tooth morphology map; the boundary region is the region of the tooth other than the central axis region; the first value is higher than the second value.

4. The method as described in claim 1, characterized in that, The process of determining a tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph includes: For any tooth in the panoramic oral radiograph, the distance information from any pixel point of the tooth to the nearest tooth boundary point is determined, and the pixel representation information of any pixel point is determined based on the distance information, thereby obtaining a first morphological image; For any tooth in the panoramic oral radiograph, the skeleton of the tooth is determined, and the pixel representation information of the pixels corresponding to the skeleton is set to the maximum value, while the pixel representation information of the pixels not corresponding to the skeleton is set to the minimum value, thereby obtaining the second morphological image. The tooth morphology image is determined by performing an OR operation on the same pixel in the first morphology image and the second morphology image.

5. The method according to any one of claims 1 to 4, characterized in that, Before finalizing the tooth morphology diagram, the following steps are also required: Determine the overall area where the teeth are located from the panoramic oral radiograph; The entire area containing the teeth in the panoramic oral radiograph is cropped to obtain a tooth region image; Based on the morphological information of each tooth in the panoramic dental radiograph, a tooth morphology diagram is determined, including: Based on the morphological information of each tooth in the tooth region image, a tooth morphology map is determined.

6. The method as described in claim 5, characterized in that, Determining the overall region where the teeth are located from the panoramic oral radiograph includes: Determine the binary image corresponding to the panoramic oral radiograph; the first pixel value of the binary image corresponds to the tooth region of the panoramic oral radiograph; the second pixel value of the binary image corresponds to the non-tooth region of the panoramic oral radiograph. Based on the binary image, determine the smallest region containing all the teeth; The minimum region is expanded to obtain the entire region where the tooth is located.

7. The method as described in claim 6, characterized in that, The tooth morphology map is used to locate the teeth, obtaining the initial position of each tooth in the map, including: Perform an AND operation on the same pixel in the binary image and the tooth morphology image to determine the third morphology image; The teeth are located in the third morphological diagram to obtain the initial position of each tooth in the tooth morphological diagram.

8. The method according to any one of claims 1 to 4, characterized in that, By segmenting the region of interest for each tooth, the segmentation results for each tooth are obtained, including: The region of interest for each tooth is input into the first segmentation model to obtain the segmentation result for each tooth.

9. The method according to any one of claims 1 to 4, characterized in that, By segmenting the region of interest for each tooth, the segmentation results for each tooth are obtained, including: The region of interest of a single tooth in the binary image corresponding to the panoramic oral radiograph is cropped to obtain a coarse segmentation region for each tooth; For any given tooth, the region of interest and the coarse segmentation region of the tooth are input into the second segmentation model, and the segmentation result of the tooth is output.

10. The method as described in claim 9, characterized in that, The second segmentation model includes multiple convolutional layers, wherein each convolutional layer is followed by a feature enhancement layer; the input of the feature enhancement layer is the output of the previous convolutional layer and the coarse segmentation region of the tooth; the feature enhancement layer is used to enhance the tooth feature information based on the coarse segmentation region of the tooth.

11. The method as described in claim 10, characterized in that, The feature enhancement layer includes the following steps: The coarsely segmented region of the tooth is subjected to a convolution operation to obtain the first feature image; The output of the convolutional layer preceding the feature enhancement layer is added to the same pixel in the first feature image to obtain the second feature map. Based on the activation function and the second feature image, the attention coefficient of the region of interest of the tooth is determined; The output of the feature enhancement layer is determined based on the attention coefficient of the region of interest of the tooth and the output of the convolutional layer preceding the feature enhancement layer.

12. The method according to any one of claims 1 to 4, characterized in that, By segmenting the region of interest for each tooth, the segmentation result for each tooth is obtained, including: The individual teeth in the tooth morphology diagram are cropped to obtain the morphological region of each tooth; For any given tooth, the region of interest, the coarse segmentation region, and the morphological region of the tooth are input into the third segmentation model, and the segmentation result of the tooth is output.

13. The method according to any one of claims 1 to 12, characterized in that, The method further includes: This shows the segmentation results of each tooth in the patient's panoramic oral radiograph.

14. A method for determining the degree of alveolar bone resorption, characterized in that, include: Obtain panoramic dental radiographs; The panoramic image of the teeth is input into the segmentation model to determine the tooth contour curve, the first contour line, and the second contour line; the first contour line is the outer edge line of the crown portion of at least two teeth, and the second contour line is the outer edge line of the alveolar bone of the maxilla and mandible; the tooth contour curve is the curve formed by the outer surface edge of each tooth. For a single tooth, the degree of alveolar bone resorption is determined based on the distribution area of ​​the first and second contour lines within the tooth contour curve of the tooth.

15. The method as described in claim 14, characterized in that, Determining the degree of alveolar bone resorption of the tooth based on the distribution area of ​​the first contour line and the second contour line within the tooth contour curve includes: Determine the distribution areas of the first contour line and the second contour line within the tooth contour curve of the tooth. The degree of alveolar bone resorption of the tooth is determined based on the area of ​​the distribution region and / or the length of the reference line within the distribution region.

16. The method as described in claim 15, characterized in that, Determining the degree of alveolar bone resorption of the tooth based on the length of a set reference line within the distribution area includes at least one of the following: The degree of alveolar bone resorption of the tooth is determined based on the length of the tooth's principal axis within the distribution area. The degree of alveolar bone resorption of the tooth is determined based on the first length; The degree of alveolar bone resorption of the tooth is determined based on the second length; The degree of alveolar bone resorption of the tooth is determined based on the first length and the second length; The first length is the length of the tooth contour curve in the mesial direction within the distribution area, and the second length is the length of the tooth contour curve in the distal direction within the distribution area.

17. The method as described in claim 16, characterized in that, Determining the first length of the tooth contour curve in the mesial direction within the distribution area includes: Determine the first intersection point between the tooth contour curve and the first contour line in the mesial direction, and the second intersection point between the tooth contour curve and the second contour line in the mesial direction; The distance between the first intersection point and the second intersection point is taken as the first length; Determining the second length of the tooth contour curve in the distal direction within the distribution area includes: Determine the third intersection point of the tooth contour curve and the first contour line in the distal direction, and the fourth intersection point of the tooth contour curve and the second contour line in the distal direction. The distance between the third intersection point and the fourth intersection point is taken as the second length.

18. The method as described in claim 16, characterized in that, Determining the degree of alveolar bone resorption of the tooth based on the first length and the second length includes at least one of the following: The degree of alveolar bone resorption of the tooth is determined based on the longest of the first and second lengths. The degree of alveolar bone resorption of the tooth is determined based on the average length of the first length and the second length.

19. The method as described in claim 18, characterized in that, Determining the degree of alveolar bone resorption of the tooth based on the longest of the first and second lengths includes: The root length of the tooth is determined based on the tooth's principal axis and the first contour line. Determine a first ratio of the first length to the root length and a second ratio of the second length to the root length; The degree of alveolar bone resorption of the tooth is determined based on the larger of the first ratio and the second ratio.

20. The method as described in claim 19, characterized in that, Determining the degree of alveolar bone resorption of the tooth based on the larger of the first ratio and the second ratio includes: The largest ratio between the first ratio and the second ratio is compared with each preset standard; wherein each preset standard corresponds to a different alveolar bone resorption grade. Based on the comparison results, the alveolar bone resorption grade of the teeth was determined.

21. The method as described in claim 14, characterized in that, The segmentation model is determined based on the method described in any one of claims 1 to 13.

22. A device for determining the degree of alveolar bone resorption, characterized in that, include: The acquisition module is used to acquire panoramic dental radiographs. The processing module is used to input the panoramic image of the teeth into the segmentation model to determine the tooth contour curve, the first contour line of all teeth, and the second contour line of all teeth; the first contour line is the outer edge line of the crown portion of the tooth including at least two teeth, and the second contour line is the outer edge line of the alveolar bone of the maxilla and mandible; the tooth contour curve is the curve formed by the outer surface edge of each tooth. The judgment module is used to determine the degree of alveolar bone resorption of any tooth based on the distribution area of ​​the first contour line and the second contour line within the tooth contour curve of the tooth.

23. A display interface, characterized in that, The display interface responds to one or more of the panoramic radiograph of the teeth, the curves of the teeth, and the degree of alveolar bone resorption of the teeth under any of the methods described in claims 14 to 21.

24. A tooth-splitting device, characterized in that, Includes morphological defining units and tooth segmentation units: The morphology determination unit is used to determine a tooth morphology map based on the morphological information of each tooth in the panoramic oral radiograph; the pixel representation information of each tooth in the tooth morphology map is higher in the central axis region than in the boundary region. The tooth segmentation unit is used to locate the teeth in the tooth morphology diagram to obtain the initial position of each tooth in the tooth morphology diagram. Based on the initial position of each tooth in the tooth morphology map, the region of interest for each tooth is obtained from the panoramic oral radiograph; by segmenting the region of interest for each tooth, the segmentation result of each tooth is obtained.

25. A computing device, characterized in that, include: Memory, used to store program instructions; A processor is configured to invoke program instructions stored in the memory and execute the method as described in any one of claims 1 to 13 or 14 to 21 according to the obtained program instructions.

26. A computer-readable storage medium, characterized in that, Includes computer-readable instructions that, when read and executed by a computer, cause the method as described in any one of claims 1 to 13 or 14 to 21 to be implemented.

27. A computer program product, characterized in that, It includes a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the method according to any one of claims 1 to 13 or 14 to 21.

Images