Method and device for generating panoramic curved surface fault sheet based on CBCT (cone beam computed tomography) image of three-dimensional segmentation

Abstract

The invention relates to a method and a device for generating a panoramic curved surface fault slice based on a CBCT image of three-dimensional segmentation, and the method comprises the steps: S1, carrying out the three-dimensional segmentation of teeth and a mandible in the CBCT image through a deep learning algorithm, and obtaining a tooth and mandible segmentation result; s2, acquiring a reference dental arch line for representing the dentition condition of the patient based on the segmentation result; s3, adjusting the dental arch line of each layer of slice containing the segmentation result based on the segmentation result and the reference dental arch line; and S4, extracting panoramic slice data from the CBCT image according to the dental arch line of each layer of slice, and generating a panoramic curved surface fault slice. According to the method, by means of the segmentation result, the oral cavity panoramic image with the target anatomical structure is accurately extracted from the CBCT image, a dentist can conveniently measure the length, the angle and the like, the adjacent relation of the dentition periphery can be conveniently observed, meanwhile, radiation absorption of secondary shooting is avoided, and the imaging accuracy is improved. And the problems of poor patient cooperation, inaccurate focusing positioning and the like during panoramic actual shooting are solved.

Description

Technical Field

[0001] The present invention relates to the technical field of oral image processing, and in particular to a method and device for generating panoramic curved surface tomograms based on three-dimensional segmented CBCT images. Background Art

[0002] Cone beam computed tomography (CBCT) can provide high-resolution three-dimensional images with low radiation doses, clearly showing the teeth, bones and soft tissues in the patient's mouth, and is an important means for dentists to check the patient's oral health. Oral panoramic curved surface tomography can directly display the positional relationship of the upper and lower jaw dentition and the surrounding oral anatomical tissues in a two-dimensional plane, and can more accurately measure parameters such as length and angle. It is an important basis for dentists to design treatment plans. Generating oral panoramic curved surface tomography based on CBCT images allows doctors to use three-dimensional images to observe the patient's oral and maxillofacial conditions from different directions, and obtain two-dimensional oral panoramic curved surface tomography without secondary shooting to perform anatomical morphological measurements.

[0003] However, due to the changes in tooth morphology caused by oral diseases, metal artifacts, similar Hu values ​​of teeth and surrounding soft tissues, and tooth tilt, the existing 3D data slicing algorithms have problems such as incomplete tooth display and unclear neural tube display. Manually adjusting the imaging range of panoramic curved tomography takes time and effort, which is difficult to implement in practical applications. Summary of the invention

[0004] The technical problem to be solved by the present invention is to design a method and device for generating panoramic curved tomograms based on three-dimensional segmented CBCT images. With the help of the segmentation results, oral panoramic images with target anatomical structures are accurately extracted from CBCT images, making it convenient for dentists to measure lengths, angles, etc., thereby solving existing technical problems.

[0005] In order to solve the above technical problems, the present invention provides a method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images, comprising the following steps:

[0006] Step S1: using a deep learning algorithm to perform three-dimensional segmentation of teeth and mandible in CBCT images to obtain segmentation results of teeth and mandible;

[0007] Step S2: based on the teeth and mandibular segmentation results, a reference dental arch line for representing the patient's dentition is obtained;

[0008] Step S3: Based on the teeth and mandibular segmentation results and the reference dental arch line, adjusting the dental arch line of each slice layer including the segmentation results;

[0009] Step S4: extracting panoramic slice data from the CBCT image according to the dental arch line of each slice layer, and generating a panoramic curved surface tomogram.

[0010] Furthermore, step S1 specifically includes the following steps:

[0011] Step S11: Collect CBCT image samples and label teeth and mandibular bones;

[0012] Step S12: performing preprocessing and using data enhancement to expand samples, wherein the preprocessing operation includes downsampling the resolution of the CBCT image and the annotation result to eliminate overly dark and overly bright values; the data enhancement method includes performing random rotation, adding Gaussian noise to simulate noise in the acquisition process, and applying a texture filter to enhance the texture features of the image;

[0013] Step S13: Divide the samples into training set, validation set and test set according to the proportion;

[0014] Step S14: construct a three-dimensional tooth and mandibular segmentation model for training; the training model is a nnUNet or VNet or 3DUNet model;

[0015] Step S15: Input the CBCT image into the constructed three-dimensional teeth and mandibular segmentation model to obtain the teeth and mandibular segmentation results.

[0016] Furthermore, in step S14, a three-dimensional tooth and mandibular segmentation model based on the nnUNet neural network is constructed for training, specifically including: after the training set is subjected to maximum and minimum normalization processing, it is input into the network according to a batch size of 4 to 32. The network uses DC_and_BCE_loss as a loss function to compare the gap between the segmentation result and the true label; the stochastic gradient descent (SGD) optimizer is selected for parameter optimization and the polynomial learning rate scheduler is used to adjust the learning rate, and the training rounds are set. In each round of training, the network calculates the gradient through back propagation and updates the weight parameters to gradually reduce the loss value, and saves the weight parameter model when the loss is minimum.

[0017] Furthermore, the calculation formula of DC_and_BCE_loss is:

[0018] DC_and_BCE_loss=α·Dice Loss+β·Cross_Entropy Loss,

[0019] Among them, α and β are balance coefficients, which are used to adjust the weights of the two losses in the final loss;

[0020] Among them, A is the predicted segmentation area, B is the real labeled area, |A∩B| represents the overlap between the predicted and real areas, |A| and |B| are the areas of the predicted and real areas respectively;

[0021] Among them, y is the true value, is the predicted value.

[0022] Furthermore, step S2 specifically includes the following steps:

[0023] Step S21: Perform cross-sectional maximum projection on the segmentation result;

[0024] Step S22: Binarizing the cross-sectional projection image;

[0025] Step S23: dilate, erode and thin the cross-sectional binary image to obtain a dental arch line candidate region;

[0026] Step S24: identifying the endpoints and bifurcation points in the dental arch line, and traversing all branch routes starting from the endpoints;

[0027] Step S25: Screen all branch routes to obtain the final reference dental arch line.

[0028] Furthermore, step S25 specifically includes: among all the branches obtained, if the starting and ending endpoints are the same, only the longest one is selected; if the starting and ending endpoints are different, they are saved and stored in ascending order according to the X value of the starting endpoint, and finally, the stored branches are connected as the final reference dental arch line.

[0029] Furthermore, step S3 specifically includes the following steps:

[0030] Step S31: interpolating the reference dental arch line so that the distance between the front and rear points is one pixel;

[0031] Step S32: Calculate the slicing direction, where the slicing direction is the perpendicular direction of the dental arch line tangent;

[0032] Step S33: traverse the image layer to obtain tooth label points and mandibular label points;

[0033] Step S34: Find the label points of the teeth or mandible and record the start and end positions;

[0034] Step S35: updating the dental arch line, recording the dental arch line position deviation data and slice thickness data, and obtaining the dental arch line of each slice layer containing the segmentation result.

[0035] Furthermore, step S4 specifically includes the following steps:

[0036] Step S41: pre-processing the dental arch line position deviation and slice thickness of each slice layer obtained including the segmentation result, assigning these two sets of data into a picture form, and performing a smoothing operation on the picture;

[0037] Step S42: Calculate the slicing direction, where the slicing direction is the perpendicular direction of the dental arch line tangent;

[0038] Step S43: Obtaining panoramic pixel values;

[0039] Step S44: performing post-processing of noise reduction and enhancement on the image, and finally generating a clear panoramic curved surface tomogram.

[0040] Furthermore, step S43 specifically includes: on the mth coronal plane image of the CBCT image, taking the nth point on the dental arch line of each slice layer containing the segmentation result after the update as the starting point, extending the length of the slice thickness of the point to both ends along the slice direction of the point, and accumulating the points on the line segment in the CBCT image as the pixel values ​​of the mth row and nth column in the panoramic curved surface tomogram.

[0041] The present invention also provides a device for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images, specifically comprising:

[0042] at least one processor; and

[0043] at least one memory in communication with the processor;

[0044] The memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the device to perform the aforementioned method of generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images.

[0045] Beneficial effects of the present invention:

[0046] The method of generating panoramic curved surface slices based on CBCT images of the present invention uses the segmentation results to accurately extract oral panoramic images with target anatomical structures from CBCT images, making it convenient for dentists to measure lengths, angles, etc., and to observe the adjacent relationship around the dentition. At the same time, it avoids the radiation absorption of secondary shooting, and avoids the problems of poor patient cooperation and inaccurate focus positioning during the actual panoramic shooting. Compared with the existing algorithm for performing slice calculations to obtain panoramic curved surface slices from CBCT, the present invention can effectively solve the problems of incomplete dentition display, missing temporomandibular joints, unclear neural tube display, etc. caused by incorrect positioning of the dental arch line due to the special oral tissue morphology. Since the thickness of the teeth or jaws has been obtained, data outside the target range is no longer calculated, which improves the calculation speed and reduces resource consumption. BRIEF DESCRIPTION OF THE DRAWINGS

[0047] The specific implementation of the present invention will be further explained below in conjunction with the accompanying drawings.

[0048] Figure 1 A flow chart of a method for generating panoramic curved tomographic slices based on CBCT images of the present invention based on three-dimensional segmentation;

[0049] Figure 2 A flow chart of the method for obtaining a reference dental arch line in the present invention;

[0050] Figure 3 A flow chart of a method for adjusting the dental arch line of each slice in the present invention;

[0051] Figure 4 The present invention is a flow chart of a method for generating panoramic curved tomographic slices based on panoramic slice data. DETAILED DESCRIPTION

[0052] Example 1

[0053] Combination Figure 1 The method of generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images in this embodiment includes the following steps:

[0054] Step S1: Use a deep learning algorithm to perform three-dimensional segmentation of teeth and mandibles in CBCT images to obtain segmentation results of teeth and mandibles.

[0055] In this embodiment, preferably, step S1 specifically includes the following steps:

[0056] Step S11: Collect CBCT image samples and label the teeth and mandibular bones.

[0057] Specifically, in this embodiment, CBCT images of a sufficient number of patients are collected, and the images cover different dental conditions, including implants, metal artifacts, etc., to ensure the diversity of the samples. Multiple dental experts use annotation tools to accurately mark the teeth and mandibles to ensure that the teeth and mandibles can be distinguished from other tissues and correctly identified and marked. In this embodiment, CBCT images of 300 patients are collected and marked by two dental experts.

[0058] Step S12: Preprocessing is performed, and the preprocessing operation includes downsampling the resolution of the CBCT image and the annotation result to eliminate the too dark and too bright values; in this embodiment, the resolution of the CBCT image and the annotation result is downsampled to 0.4, 0.3 or 0.2 spacing, and all different sizes need to be processed into the same size for easy processing. In this embodiment, preferably, the resolution of the CBCT image and the annotation result is downsampled to 0.4 spacing, so that the resource consumption during training is small. Due to the limited number of data samples, data enhancement technology is used to expand the samples. The data enhancement method includes random rotation (rotation angle of -15° to 15°), adding Gaussian noise to simulate the noise in the acquisition process, and applying texture filters to enhance the texture features of the image. These data enhancement operations help to improve the robustness of the model.

[0059] Step S13: Divide the samples into a training set, a validation set, and a test set according to a ratio. In this embodiment, the ratio of the training set, the validation set, and the test set is 7:2:1.

[0060] Step S14: construct a three-dimensional tooth and mandibular segmentation model for training, and the training model is a nnUNet or VNet or 3DUNet model.

[0061] Preferably, in the present embodiment, in step S14, a three-dimensional tooth and mandibular segmentation model based on the nnUNet neural network is constructed for training, which specifically includes: after the training set is subjected to maximum and minimum normalization processing, it is input into the network according to a batch size between 4 and 32, which is determined according to the size of the graphics card memory. Specifically, in the present embodiment, the network is input according to a batch size of 8. The network uses DC_and_BCE_loss as the loss function to compare the gap between the segmentation result and the true label; the stochastic gradient descent (SGD) optimizer is used for parameter optimization and the polynomial learning rate scheduler is used to adjust the learning rate. In the present embodiment, the number of training rounds is set to 1500 rounds. In each round of training, the network calculates the gradient through back propagation and updates the weight parameters to gradually reduce the loss value and save the weight parameter model when the loss is minimized.

[0062] Specifically, in this embodiment, the nnUNet model inherits the encoder-decoder structure of UNet. The encoder part downsamples layer by layer to extract features, and the decoder part upsamples layer by layer to restore spatial resolution. This structure helps the network effectively capture local and global information of the image. And a jump connection is used between the encoder and the decoder to pass high-resolution features to the decoder part, thereby improving the accuracy of segmentation and detail retention.

[0063] Specifically, in this embodiment, DC_and_BCE_loss is a loss function that combines Dice loss and cross-entropy loss, which can simultaneously optimize the pixel-level classification accuracy and overall structural matching of the segmentation result.

[0064] Dice loss is an indicator to measure the similarity between two samples. It can handle the problem of class imbalance and is defined as follows:

[0065]

[0066] Among them, A is the predicted segmentation area, B is the real labeled area, |A∩B| represents the overlapping part of the predicted and real areas, and |A| and |B| are the areas of the predicted and real areas respectively.

[0067] Cross entropy loss is used to measure the difference between the predicted probability distribution and the true label. For each pixel, the cross entropy loss calculates the negative log-likelihood of the predicted probability distribution and the true label, and the formula is:

[0068]

[0069] Among them, y is the true value, is the predicted value.

[0070] The DC_and_BCE_loss loss combines the above two loss functions to simultaneously optimize the pixel-level classification accuracy and overall structural matching of the segmentation results.

[0071] Preferably, in this embodiment, the calculation formula of DC_and_BCE_loss is:

[0072] DC_and_BCE_loss=α·Dice Loss+β·Cross_Entropy Loss,

[0073] Among them, α and β are balance coefficients, which are used to adjust the weights of the two losses in the final loss.

[0074] Specifically, in this embodiment, the recommended range of α is between 0.7 and 1, and the recommended range of β is between 0.1 and 1. If the categories of the data set are relatively balanced, the value of β can be larger.

[0075] Step S15: Input the CBCT image into the constructed three-dimensional teeth and mandibular segmentation model to obtain the teeth and mandibular segmentation results.

[0076] Step S2: Based on the teeth and mandibular segmentation results, a reference dental arch line is obtained to represent the patient's dentition.

[0077] In this embodiment, preferably, Figure 2 As shown, step S2 specifically includes the following steps:

[0078] Step S21: Perform cross-sectional maximum projection on the segmentation result.

[0079] Specifically, in this embodiment, the projection rule is that if the pixel value on the projection path is consistent with the label value of the mandible and the teeth, the point on the projection image is set to 255.

[0080] Step S22: Binarize the cross-sectional projection image.

[0081] Step S23: dilate, erode and thin the cross-sectional binary image to obtain a dental arch line candidate region.

[0082] Step S24: Identify the endpoints and bifurcation points in the dental arch line, and traverse all branch routes starting from the endpoints.

[0083] Specifically, in this embodiment, each point on the dental arch line is traversed, and the points in its eight-connected domain that are also dental arch lines are stored. If only one point is stored in the eight-connected domain, the point is an endpoint; if three or more points are stored in the eight-connected domain, the point is defined as a bifurcation point.

[0084] Specifically, in this embodiment, starting from the leftmost endpoint, walk down along the neighborhood points stored in the eight-connected domain, and iterate continuously. If a bifurcation point is encountered, walk down from different neighborhood directions respectively; if an endpoint is encountered, the traversal of the current path ends. Repeat the above process until all endpoints have been traversed.

[0085] Step S25: Screen all branch routes to obtain the final reference dental arch line.

[0086] In this embodiment, step S25 specifically includes: among all the branches obtained, if the starting and ending endpoints are the same, only the longest one is selected; if the starting and ending endpoints are different, they are saved and stored in ascending order of the X value of the starting endpoint, and finally, the stored branches are connected as the final reference dental arch line. The reference dental arch line is sampled at intervals to obtain the reference dental arch line control points.

[0087] Step S3: Based on the teeth and mandibular segmentation results and the reference dental arch line, adjust the dental arch line of each slice layer containing the segmentation results.

[0088] In this embodiment, preferably, Figure 3 As shown, step S3 specifically includes the following steps:

[0089] Step S31: interpolate the reference dental arch line so that the distance between the front and rear points is one pixel.

[0090] Step S32: Calculate the slice direction, where the tangent direction is the perpendicular direction of the dental arch line tangent.

[0091] Specifically, in this embodiment, when a panoramic image is acquired in a slicing manner, a slicing direction needs to be set. The vertical direction of the dental arch line tangent is set as the slicing direction, and the calculation formula is as follows:

[0092] archDir i =normalize(arch i+1 -arch i );

[0093] rayDir i =Cross(zAxis,archDir i );

[0094] Among them, archDir i is the tangent direction of the i-th point on the dental arch line, normalize(·) is the normalization operation, arch i is the i-th point on the dental arch line, rayDir i is the slicing direction, Cross(·) is the cross product operation, and zAxis is the unit vector in the positive direction of the Z axis.

[0095] Step S33: traverse the image layer to obtain tooth label points and mandibular label points.

[0096] Specifically, in this embodiment, starting from the current layer of the teeth and mandibular segmentation results, traversal is performed along the direction of the generated reference dental arch line vertical line, and pixels on the vertical line are checked layer by layer to obtain possible teeth label points and mandibular label points.

[0097] Step S34: Find the label points of the teeth or mandible and record the starting and ending positions.

[0098] Specifically, in the present embodiment, during the traversal process, it is determined whether each point passed by has a label of "teeth" or "mandible". The first point and the last point that meet the conditions are found, and the positions and corresponding labels of the points are recorded.

[0099] Step S35: Update the dental arch line. Traverse each layer of the image, calculate the distance between the midpoint of the starting and ending points of the found label points and the reference dental arch line, record it as a set of dental arch line position deviation data, and then calculate the distance between the starting and ending points of the found label points, record it as a set of slice thickness data, and obtain the dental arch line of each slice layer containing the segmentation result.

[0100] Step S4: extracting panoramic slice data from the CBCT image according to the dental arch line of each slice layer, and generating a panoramic curved surface tomogram.

[0101] In this embodiment, preferably, Figure 4 As shown, step S4 specifically includes the following steps:

[0102] Step S41: pre-processing the dental arch line position deviation and slice thickness of each slice layer obtained including the segmentation result, because the data of the dental arch line position deviation and slice thickness correspond one-to-one to the points on the panoramic curved surface tomography, and the length of the panoramic curved surface tomography is determined by the number of points on the dental arch line, and the width of the panoramic curved surface tomography is determined by the number of layers of the CBCT image, so the two sets of data of the dental arch line position deviation and slice thickness can also be assigned to the form of a picture, and the picture is smoothed, so as to eliminate abnormal points in the data.

[0103] Step S42: Calculate the slicing direction, which is the perpendicular direction of the dental arch line tangent.

[0104] Step S43: Obtain panoramic pixel values.

[0105] Preferably, in this embodiment, step S43 specifically includes: on the m-th coronal plane image of the CBCT image, taking the n-th point on the dental arch line of each slice layer containing the segmentation result after the update as the starting point, extending the length of the slice thickness of the point to both ends along the slice direction of the point, and accumulating the points on the line segment in the CBCT image as the pixel values ​​of the m-th row and the n-th column in the panoramic curved surface tomogram.

[0106] Step S44: performing post-processing of noise reduction and enhancement on the image, and finally generating a clear panoramic curved surface tomogram.

[0107] Example 2

[0108] This embodiment provides a device for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images, specifically comprising:

[0109] at least one processor; and

[0110] at least one memory in communication with the processor;

[0111] The memory stores instructions executable by the processor, and the instructions are executed by the processor to enable the device to perform the method of generating panoramic curved tomograms based on three-dimensional segmented CBCT images in Example 1.

[0112] Many specific details are described in the above description to facilitate a full understanding of the present invention. However, the above description is only a preferred embodiment of the present invention. The present invention can be implemented in many other ways different from those described herein, so the present invention is not limited to the specific implementation disclosed above. At the same time, any person familiar with the art can make many possible changes and modifications to the technical solution of the present invention using the methods and technical contents disclosed above without departing from the scope of the technical solution of the present invention, or modify it into an equivalent embodiment of equivalent changes. Any simple modification, equivalent change and modification made to the above embodiments according to the technical essence of the present invention without departing from the content of the technical solution of the present invention still falls within the scope of protection of the technical solution of the present invention.

Claims

1. A method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images, characterized in that: The steps include: Step S1: using a deep learning algorithm to perform three-dimensional segmentation of teeth and mandible in CBCT images to obtain segmentation results of teeth and mandible; Step S2: based on the teeth and mandibular segmentation results, a reference dental arch line for representing the patient's dentition is obtained; Step S3: Based on the teeth and mandibular segmentation results and the reference dental arch line, adjusting the dental arch line of each slice layer including the segmentation results; Step S4: extracting panoramic slice data from the CBCT image according to the dental arch line of each slice layer, and generating a panoramic curved surface tomogram.

2. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 1, characterized in that: Step S1 specifically includes the following steps: Step S11: Collect CBCT image samples and label teeth and mandibular bones; Step S12: performing preprocessing and using data enhancement to expand samples, wherein the preprocessing operation includes downsampling the resolution of the CBCT image and the annotation result to eliminate overly dark and overly bright values; the data enhancement method includes performing random rotation, adding Gaussian noise to simulate noise in the acquisition process, and applying a texture filter to enhance the texture features of the image; Step S13: Divide the samples into training set, validation set and test set according to the proportion; Step S14: constructing a three-dimensional teeth and mandibular segmentation model for training; Step S15: Input the CBCT image into the constructed three-dimensional teeth and mandibular segmentation model to obtain the teeth and mandibular segmentation results.

3. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 2, characterized in that: In step S14, a three-dimensional tooth and mandibular segmentation model based on the nnUNet neural network is constructed for training, specifically including: after the training set is normalized to the maximum and minimum, it is input into the network according to a batch size of 4 to 32; the network uses DC_and_BCE_loss as a loss function to compare the gap between the segmentation result and the true label; the stochastic gradient descent optimizer is used for parameter optimization and the polynomial learning rate scheduler is used to adjust the learning rate, and the training rounds are set. In each round of training, the network calculates the gradient through back propagation and updates the weight parameters to gradually reduce the loss value, and saves the weight parameter model when the loss is minimum.

4. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 3, characterized in that: The calculation formula of DC_and_BCE_loss is: DC_and_BCE_loss=α·Dice Loss+β·Cross_Entropy Loss, Among them, α and β are balance coefficients, which are used to adjust the weights of the two losses in the final loss; Among them, A is the predicted segmentation area, B is the real labeled area, |A∩B| represents the overlap between the predicted and real areas, |A| and |B| are the areas of the predicted and real areas respectively; Among them, y is the true value, is the predicted value.

5. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 1, characterized in that: Step S2 specifically includes the following steps: Step S21: Perform cross-sectional maximum projection on the segmentation result; Step S22: Binarizing the cross-sectional projection image; Step S23: dilate, erode and thin the cross-sectional binary image to obtain a dental arch line candidate region; Step S24: identifying the endpoints and bifurcation points in the dental arch line, and traversing all branch routes starting from the endpoints; Step S25: Screen all branch routes to obtain the final reference dental arch line.

6. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 5, characterized in that: Step S25 specifically includes: among all the branches obtained, if the starting and ending endpoints are the same, only the longest one is selected; if the starting and ending endpoints are different, they are saved and stored in ascending order according to the X value of the starting endpoint, and finally, the stored branches are connected as the final reference dental arch line.

7. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 1, characterized in that: Step S3 specifically includes the following steps: Step S31: interpolating the reference dental arch line so that the distance between the front and rear points is one pixel; Step S32: Calculate the slicing direction, where the slicing direction is the perpendicular direction of the dental arch line tangent; Step S33: traverse the image layer to obtain tooth label points and mandibular label points; Step S34: Find the label points of the teeth or mandible and record the start and end positions; Step S35: updating the dental arch line, recording the dental arch line position deviation data and slice thickness data, and obtaining the dental arch line of each slice layer containing the segmentation result.

8. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 1, characterized in that: Step S4 specifically includes the following steps: Step S41: pre-processing the dental arch line position deviation and slice thickness of each slice layer obtained including the segmentation result, assigning these two sets of data into a picture form, and performing a smoothing operation on the picture; Step S42: Calculate the slicing direction, where the slicing direction is the perpendicular direction of the dental arch line tangent; Step S43: Obtaining panoramic pixel values; Step S44: performing post-processing of noise reduction and enhancement on the image, and finally generating a clear panoramic curved surface tomogram.

9. The method for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images according to claim 1, characterized in that: Step S43 specifically includes: on the m-th coronal plane image of the CBCT image, taking the n-th point on the dental arch line of each slice layer containing the segmentation result after the update as the starting point, extending the length of the slice thickness of the point to both ends along the slice direction of the point, and accumulating the points on the line segment in the CBCT image as the pixel values ​​of the m-th row and the n-th column in the panoramic curved surface tomogram.

10. A device for generating panoramic curved tomographic slices based on three-dimensional segmented CBCT images, characterized in that: include: at least one processor; as well as at least one memory in communication with the processor; The memory stores instructions executable by the processor, and the instructions are executed by the processor so that the device executes the method for generating panoramic curved tomograms based on three-dimensional segmented CBCT images as described in any one of claims 1 to 9.

Images