A method and system for positioning vertebrae and airway

Abstract

The present invention relates to the technical field of vertebrae and airway positioning, and in particular to a vertebrae and airway positioning method and system, the method comprising: inputting a first cervical vertebrae image into a pre-trained first semantic segmentation network model for identifying vertebrae to obtain a first output image, judging whether the first output image contains complete images of a third vertebrae and a fourth vertebrae to obtain a first judgment result, when the first judgment result is yes, inputting the first output image into a first HRnet network model, positioning the third vertebrae and the fourth vertebrae, and marking them in the first cervical vertebrae image, similarly positioning the airway, and marking it in the second cervical vertebrae image, being able to accurately locate the positions of the third vertebrae and the fourth vertebrae, and marking them in the first cervical vertebrae image, and being able to accurately locate the position of the airway, and marking it in the second cervical vertebrae image, so as to facilitate intuitive viewing, have high accuracy, and save time and effort.

Description

Background Art

[0002] With the rapid development of the economy and the substantial improvement of people's living standards, more and more people are concerned about whether their teeth have disordered occlusion or crowding, and orthodontic treatment has gradually come into the public eye. The best time for orthodontic treatment is undoubtedly during adolescence. Before orthodontic treatment, it is necessary to accurately determine the growth and development stage of the adolescent and whether there is glandular hypertrophy, so as to improve the effect and stability of correction.

[0003] Currently, the growth and development stage of adolescents is often determined through the cervical spine, and whether adolescents have adenoid hypertrophy is determined through the airway. However, when manually determining the position of vertebrae on the cervical spine and the position of the airway, there are problems such as low accuracy, time and labor consumption. Summary of the invention

[0004] The technical problem to be solved by the present invention is to provide a method and system for positioning vertebrae and airways in view of the deficiencies of the prior art.

[0005] The technical solution of a vertebra and airway positioning method of the present invention is as follows:

[0006] Acquire a first cervical vertebra image of a first user, input the first cervical vertebra image into a pre-trained first semantic segmentation network model for identifying vertebrae, obtain a first output image, determine whether the first output image contains complete images of the third vertebra and the fourth vertebra, obtain a first determination result, and when the first determination result is yes, input the first output image into a first HRnet network model, locate the third vertebra and the fourth vertebra, and mark them in the first cervical vertebra image;

[0007] Obtain a second cervical vertebra image of a second user, input the second cervical vertebra image into a pre-trained second semantic segmentation network model for identifying the airway, obtain a second output image, determine whether the second output image contains a complete image of the airway, and obtain a second judgment result. When the second judgment result is yes, input the second output image into a second HRnet network model, locate the airway, and mark it in the second cervical vertebra image.

[0008] The beneficial effects of the vertebra and airway positioning method of the present invention are as follows:

[0009] It can accurately locate the positions of the third and fourth vertebrae and mark them in the first cervical vertebra image, and can accurately locate the position of the airway and mark it in the second cervical vertebra image, which is convenient for intuitive viewing, highly accurate, and saves time and effort.

[0010] Based on the above solution, the vertebra and airway positioning method of the present invention can also be improved as follows.

[0011] Further, determining whether the first output image includes complete images of the third vertebra and the fourth vertebra to obtain a first determination result includes:

[0012] Determine whether there are any overlapping points between the first outer contour and the outer contour of the first output image to obtain a third judgment result, when the third judgment result is no, determine that the first output image includes the complete image of the third vertebra and the fourth vertebra, and the first judgment result is yes, when the third judgment result is yes, determine that the first output image does not include the complete image of the third vertebra and the fourth vertebra, and the first judgment result is no, wherein the first outer contour is: the outer contour of the area recognized by the first semantic segmentation network model in the first output image.

[0013] Further, determining whether the second output image includes a complete image of the airway to obtain a second determination result includes:

[0014] Determine whether there are any overlapping points between the second outer contour and the outer contour of the second output image to obtain a fourth judgment result, and when the fourth judgment result is no, determine that the second output image includes a complete image of the airway, and the second judgment result is yes, and when the fourth judgment result is yes, determine that the second output image does not include a complete image of the airway, and the second judgment result is no, wherein the second outer contour is: the outer contour of the area recognized by the second semantic segmentation network model in the second output image.

[0015] Furthermore, it also includes:

[0016] When the first judgment result is no, re-acquiring the cervical vertebra image of the first user as the first cervical vertebra image;

[0017] When the second judgment result is no, the cervical vertebra image of the second user is recaptured as the second cervical vertebra image.

[0018] The technical solution of a vertebra and airway positioning system of the present invention is as follows:

[0019] It includes a first positioning and marking module and a second positioning and marking module;

[0020] The first positioning and annotation module is used to: obtain a first cervical vertebra image of a first user, input the first cervical vertebra image into a pre-trained first semantic segmentation network model for identifying vertebrae, obtain a first output image, determine whether the first output image contains complete images of the third vertebra and the fourth vertebra, obtain a first judgment result, and when the first judgment result is yes, input the first output image into a first HRnet network model, locate the third vertebra and the fourth vertebra, and annotate them in the first cervical vertebra image;

[0021] The second positioning and annotation module is used to: obtain a second cervical vertebra image of a second user, input the second cervical vertebra image into a pre-trained second semantic segmentation network model for identifying the airway, obtain a second output image, determine whether the second output image contains a complete image of the airway, and obtain a second judgment result; when the second judgment result is yes, input the second output image into a second HRnet network model, locate the airway, and annotate it in the second cervical vertebra image.

[0022] The beneficial effects of a vertebra and airway positioning system of the present invention are as follows:

[0023] It can accurately locate the positions of the third and fourth vertebrae and mark them in the first cervical vertebra image, and can accurately locate the position of the airway and mark it in the second cervical vertebra image, which is convenient for intuitive viewing, highly accurate, and saves time and effort.

[0024] Based on the above solution, the vertebra and airway positioning system of the present invention can also be improved as follows.

[0025] Further, the process in which the first positioning and marking module determines whether the first output image contains complete images of the third vertebra and the fourth vertebra includes:

[0026] Determine whether there are any overlapping points between the first outer contour and the outer contour of the first output image to obtain a third judgment result, when the third judgment result is no, determine that the first output image includes the complete image of the third vertebra and the fourth vertebra, and the first judgment result is yes, when the third judgment result is yes, determine that the first output image does not include the complete image of the third vertebra and the fourth vertebra, and the first judgment result is no, wherein the first outer contour is: the outer contour of the area recognized by the first semantic segmentation network model in the first output image.

[0027] Further, determining whether the second output image includes a complete image of the airway to obtain a second determination result includes:

[0028] Determine whether there are any overlapping points between the second outer contour and the outer contour of the second output image to obtain a fourth judgment result, and when the fourth judgment result is no, determine that the second output image includes a complete image of the airway, and the second judgment result is yes, and when the fourth judgment result is yes, determine that the second output image does not include a complete image of the airway, and the second judgment result is no, wherein the second outer contour is: the outer contour of the area recognized by the second semantic segmentation network model in the second output image.

[0029] Furthermore, the first positioning and marking module is further used for: when the first judgment result is no, re-collecting the cervical vertebra image of the first user as the first cervical vertebra image;

[0030] The second positioning and marking module is further used for: when the second judgment result is no, re-collecting the cervical vertebra image of the second user as the second cervical vertebra image. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 A schematic flow chart of a method for positioning a vertebra and an airway according to an embodiment of the present invention;

[0032] Figure 2 One of the diagrams labeling the third and fourth vertebrae;

[0033] Figure 3 The second diagram is a diagram for labeling the third and fourth vertebrae;

[0034] Figure 4 The third diagram is a diagram for labeling the third and fourth vertebrae;

[0035] Figure 5 A schematic diagram for labeling the airway;

[0036] Figure 6 is one of the first output images;

[0037] Figure 7 is the second of the first output images;

[0038] Figure 8 is the third of the first output images;

[0039] Fig. 9 It is a schematic structural diagram of a vertebra and airway positioning system according to an embodiment of the present invention. DETAILED DESCRIPTION

[0040] like Figure 1 As shown, a method for positioning a vertebra and an airway according to an embodiment of the present invention comprises the following steps:

[0041] S1. Obtain a first cervical vertebra image of a first user, input the first cervical vertebra image into a pre-trained first semantic segmentation network model for identifying vertebrae, obtain a first output image, determine whether the first output image contains complete images of the third vertebrae and the fourth vertebrae, and obtain a first determination result. When the first determination result is yes, input the first output image into a first HRnet network model, locate the third vertebrae and the fourth vertebrae, and mark them in the first cervical vertebra image;

[0042] The training process of the pre-trained first semantic segmentation network model for identifying vertebrae is as follows:

[0043] Collect multiple cervical vertebrae images as samples, and preprocess each cervical vertebrae image. The preprocessing operation includes cutting the collected cervical vertebrae images into a uniform size, normalizing, standardizing, randomly flipping horizontally, randomly adjusting contrast, brightness, histogram equalization and other data enhancement operations, and then training the first semantic segmentation model to obtain a trained first semantic segmentation network model for identifying vertebrae. Specifically:

[0044] The first semantic segmentation model can be specifically a U-net network model. The preprocessed cervical vertebra image is sent to the U-net network model for semantic segmentation. The U-net network model is a very popular end-to-end encoder-decoder network for semantic segmentation. The U-net network model consists of an encoder and a decoder. The encoder part uses the timm-regnety_016 network to extract features from the preprocessed cervical vertebra image, and the decoder part uses the Upsample+conv structure to fuse features of different scales. The first semantic segmentation network model is trained based on multiple cervical vertebra images to obtain a trained first semantic segmentation network model for identifying vertebrae.

[0045] Among them, the first HRnet network model structure is as follows:

[0046] HRnet starts with a high-resolution subnetwork as the first stage, gradually adds high-resolution to low-resolution subnetworks, forms more stages, and connects multi-resolution subnetworks. At the same time, multiple multi-scale fusions are performed so that each high-resolution to low-resolution representation can repeatedly receive information from other parallel representations, thereby obtaining a rich high-resolution representation. The representation information obtained in this way is not only rich but also spatially accurate, which helps to more accurately calculate the position coordinates of the first key points of the second, third, and fourth cervical vertebrae. Specifically:

[0047] Preset a plurality of first key points as marking points at the lower edge of the second cervical vertebra, the edge of the third cervical vertebra, and the edge of the fourth cervical vertebra of each cervical vertebra image as a sample, and train HRnet based on the plurality of cervical vertebra images marked with the first key points to obtain a first HRnet network model;

[0048] The specific implementation method of locating and marking the third vertebra and the fourth vertebra in the first cervical vertebra image is as follows:

[0049] Since the vertebrae to be detected, namely the third and fourth cervical vertebrae, only account for a small proportion of the original lateral film, namely the first cervical vertebrae image, in order to obtain more accurate detection results, it is necessary to crop the image containing the vertebrae. The cropping strategy is to obtain the results of the semantic segmentation module, calculate the center of mass of the joint or airway image, crop it according to the center of mass and image size, and then adjust it to a uniform size of 256×256. The cropped image results are sent to the first HRnet network model. The first HRnet network model maintains a high-resolution representation by connecting parallel representations of different resolutions and repeating multi-scale fusion. The resulting high-resolution representation is not only rich in information but also spatially accurate, so that the position coordinates of the first key point in the first cervical vertebrae image can be calculated more accurately.

[0050] That is, the first key point is identified by the first HRnet network model, and the first key point at the lower edge of the second cervical vertebra, the edge of the third cervical vertebra, and the edge of the fourth cervical vertebra is determined;

[0051] The specific method of marking is: marking the first cervical vertebra image by the first key point. Specifically, each first cervical vertebra image can be marked in the form of a circle or a dot, which is more convenient for displaying the third cervical vertebra and the fourth cervical vertebra on the first cervical vertebra image. Figure 2 and Figure 3 As shown, in addition, the first key points are connected using lines, such as Figure 4 shown.

[0052] S2. Obtain a second cervical vertebra image of a second user, input the second cervical vertebra image into a pre-trained second semantic segmentation network model for identifying the airway, obtain a second output image, determine whether the second output image contains a complete image of the airway, and obtain a second judgment result. When the second judgment result is yes, input the second output image into a second HRnet network model, locate the airway, and mark it in the second cervical vertebra image.

[0053] Among them, the training process of the pre-trained second semantic segmentation network model for identifying the airway is as follows:

[0054] Collect multiple cervical vertebra images as samples, and preprocess each cervical vertebra image. The preprocessing operation includes cutting the collected cervical vertebra images into a uniform size, normalizing, standardizing, randomly flipping, randomly adjusting contrast, brightness, histogram equalization and other data enhancement operations. Then, train the second semantic segmentation model to obtain a trained second semantic segmentation network model for identifying the airway. Specifically:

[0055] The second semantic segmentation model can be specifically a U-net network model. The preprocessed cervical spine image is sent to the U-net network model for semantic segmentation. The U-net network model is a very popular end-to-end encoder-decoder network for semantic segmentation. The U-net network model consists of an encoder and a decoder. The encoder part uses the timm-regnety_016 network to extract features from the preprocessed cervical spine image, and the decoder part uses the Upsample+conv structure to fuse features of different scales. During the training process, AdamW is selected as the optimizer, and the initial learning rate is 1e-4. After a certain number of rounds, it decays to 1e-5 and continues training. During the entire training process, the model with the best performance on the validation set is selected as the trained second semantic segmentation network model for identifying the airway.

[0056] Among them, the second HRnet network model structure is as follows: HRnet starts with the high-resolution subnetwork as the first stage, gradually adds high-resolution to low-resolution subnetworks, forms more stages, and connects multi-resolution subnetworks. At the same time, multiple multi-scale fusions are performed so that each high-resolution to low-resolution representation can repeatedly receive information from other parallel representations, thereby obtaining rich high-resolution representations. The representation information obtained in this way is not only rich but also spatially accurate, which helps to more accurately calculate the position coordinates of key points of the airway. Specifically:

[0057] Preset multiple second key points as marking points at the edge of the airway of each cervical spine image as a sample, train HRnet based on the multiple cervical spine images marked with the second key points, and obtain a second HRnet network model;

[0058] The specific implementation method of locating the airway and marking it in the first cervical vertebra image is as follows:

[0059] Since the airway to be detected only occupies a small proportion of the original lateral film, i.e., the first cervical vertebra image, in order to obtain more accurate detection results, the image containing the airway needs to be cropped. The purpose of cropping is to obtain the results of the semantic segmentation module, calculate the center of mass of the airway image, crop it according to the center of mass and image size, and then adjust it to a uniform size of 256×256. The cropped image is sent to the second HRnet network model. The second HRnet network model maintains a high-resolution representation by connecting parallel representations of different resolutions and repeating multi-scale fusion. The resulting high-resolution representation is not only rich in information but also spatially accurate, so that the position coordinates of the key points of the airway in the second cervical vertebra image can be more accurately calculated.

[0060] That is, the second key point is identified by the second HRnet network model to determine the second key point at the edge of the airway;

[0061] The specific method of marking is: marking the second cervical vertebra image by the second key point. Specifically, each second cervical vertebra image can be marked in the form of a circle or a dot, which is more convenient for intuitively displaying the airway on the second cervical vertebra image. Figure 5 shown.

[0062] The present invention can accurately locate the positions of the third vertebra and the fourth vertebra, and mark them in the first cervical vertebra image, and can accurately locate the position of the airway, and mark it in the second cervical vertebra image, which is convenient for intuitive viewing, has high accuracy, and saves time and effort.

[0063] Optionally, in the above technical solution, in S1, determining whether the first output image includes complete images of the third vertebra and the fourth vertebra to obtain a first determination result includes:

[0064] S10, determine whether there are overlapping points between the first outer contour and the outer contour of the first output image, and obtain a third judgment result. When the third judgment result is no, determine that the first output image includes the complete image of the third vertebra and the fourth vertebra, and the first judgment result is yes. When the third judgment result is yes, determine that the first output image does not include the complete image of the third vertebra and the fourth vertebra, and the first judgment result is no, wherein the first outer contour is: the outer contour of the area recognized by the first semantic segmentation network model in the first output image. Specifically:

[0065] Determine whether the first outer contour and the outer contour of the first output image have the same pixel points. If so, determine that there are overlapping points between the first outer contour and the outer contour of the first output image, indicating that the first output image disconnects the images of the third vertebra and the fourth vertebra, and the first output image fails to include the complete images of the third vertebra and the fourth vertebra. If not, determine that there are no overlapping points between the first outer contour and the outer contour of the first output image, indicating that the first output image does not disconnect the images of the third vertebra and the fourth vertebra, and the first output image includes the complete images of the third vertebra and the fourth vertebra. The overlapping points can be understood as overlapping pixel points.

[0066] In another embodiment, since the position of the vertebra in the captured cervical vertebra image is generally on the left, it is also possible to determine whether the leftmost side of the first outer contour and the leftmost side of the outer contour of the first output image have the same pixel point, and to determine whether the bottom side of the first outer contour and the bottom side of the outer contour of the first output image have the same pixel point, then:

[0067] 1) When the leftmost side of the first outer contour does not have the same pixel point as the leftmost side of the outer contour of the first output image, and when it is determined that the bottom side of the first outer contour does not have the same pixel point as the bottom side of the outer contour of the first output image, it is determined that the first outer contour and the outer contour of the first output image do not have any overlapping points;

[0068] 2) When the leftmost side of the first outer contour has the same pixel point as the leftmost side of the outer contour of the first output image, and / or when it is determined that the bottom side of the first outer contour has the same pixel point as the bottom side of the outer contour of the first output image, it is determined that the first outer contour and the outer contour of the first output image have an overlapping point.

[0069] like Figure 6 As shown, the outer contour of the first output image is Figure 6 The area identified by the first semantic segmentation network model in the first output image is Figure 6 The first outer contour includes the outer contours of the two white areas, the upper white area is the third vertebra, and the lower white area is the fourth vertebra. Through the above method, it can be determined that the first outer contour and the outer contour of the first output image do not have the same pixel points, indicating that the first output image contains complete images of the third vertebra and the fourth vertebra.

[0070] like Figure 7 As shown, the outer contour of the first output image is Figure 7 The area identified by the first semantic segmentation network model in the first output image is Figure 7The first outer contour includes the outer contours of the two white areas, the upper white area is the third vertebra, and the lower white area is the fourth vertebra. Through the above method, it can be determined that the first outer contour and the outer contour of the first output image have the same pixel points, specifically, the right side of the outer contour of the lower white area has the same pixel points as the outer contour of the first output image, indicating that the first output image does not contain a complete image of the third vertebra and the fourth vertebra, which is specifically reflected in: the first output image does not include a complete image of the fourth vertebra;

[0071] like Figure 8 As shown, the outer contour of the first output image is Figure 8 The area identified by the first semantic segmentation network model in the first output image is Figure 8 The first outer contour includes the outer contours of the two white areas, the upper white area is the third vertebra, and the lower white area is the fourth vertebra. Through the above method, it can be determined that the first outer contour and the outer contour of the first output image have the same pixel points, specifically, the lower side of the outer contour of the lower white area and the outer contour of the first output image have the same pixel points, indicating that the first output image does not contain a complete image of the third vertebra and the fourth vertebra, which is specifically reflected in: the first output image does not include a complete image of the fourth vertebra.

[0072] When the first judgment result is no, the cervical vertebra image of the first user is recaptured as the first cervical vertebra image, and then the process returns to S1 until the third vertebra and the fourth vertebra are located and marked in the first cervical vertebra image.

[0073] Optionally, in the above technical solution, in S2, determining whether the second output image includes a complete image of the airway to obtain a second determination result includes:

[0074] S20, determining whether there are overlapping points between the second outer contour and the outer contour of the second output image, and obtaining a fourth judgment result. When the fourth judgment result is no, determining that the second output image includes a complete image of the airway, the second judgment result is yes. When the fourth judgment result is yes, determining that the second output image does not include a complete image of the airway, the second judgment result is no, wherein the second outer contour is: the outer contour of the area identified by the second semantic segmentation network model in the second output image, specifically:

[0075] Determine whether the second outer contour and the outer contour of the second output image have the same pixel points. If so, determine that there are overlapping points between the second outer contour and the outer contour of the second output image, indicating that the first output image disconnects the image of the airway, and the second output image fails to contain a complete image of the airway. If not, determine that there are no overlapping points between the second outer contour and the outer contour of the second output image, indicating that the second output image does not disconnect the image of the airway, and the second output image contains a complete image of the airway.

[0076] When the second judgment result is no, the cervical vertebra image of the second user is recaptured as the second cervical vertebra image, and then the process returns to S1 until the airway is located and marked in the second cervical vertebra image.

[0077] In the above embodiments, although the steps are numbered S1, S2, etc., these are only specific embodiments given in the present application. Those skilled in the art may adjust the execution order of S1, S2, etc. according to actual conditions, which is also within the scope of protection of the present invention. It can be understood that in some embodiments, some or all of the above embodiments may be included.

[0078] like Fig. 9 As shown, a vertebra and airway positioning system 200 according to an embodiment of the present invention includes a first positioning marking module 210 and a second positioning marking module 220;

[0079] The first positioning and labeling module 210 is used to: obtain a first cervical vertebra image of a first user, input the cervical vertebra image into a pre-trained first semantic segmentation network model for identifying vertebrae, obtain a first output image, determine whether the first output image contains complete images of the third vertebrae and the fourth vertebrae, obtain a first determination result, and when the first determination result is yes, input the first output image into the first HRnet network model, locate the third vertebrae and the fourth vertebrae, and label them in the first cervical vertebra image;

[0080] The second positioning and annotation module 220 is used to: obtain a second cervical spine image of a second user, input the cervical spine image into a pre-trained second semantic segmentation network model for identifying the airway, obtain a second output image, determine whether the second output image contains a complete image of the airway, and obtain a second judgment result. When the second judgment result is yes, input the second output image into the second HRnet network model, locate the airway, and annotate it in the second cervical spine image.

[0081] It can accurately locate the positions of the third and fourth vertebrae and mark them in the first cervical vertebra image, and can accurately locate the position of the airway and mark it in the second cervical vertebra image, which is convenient for intuitive viewing, highly accurate, and saves time and effort.

[0082] Optionally, in the above technical solution, the process of the first positioning and marking module 210 determining whether the first output image contains complete images of the third vertebra and the fourth vertebra includes:

[0083] Determine whether there are any overlapping points between the first outer contour and the outer contour of the first output image, and obtain a third judgment result. When the third judgment result is no, determine that the first output image includes a complete image of the third vertebra and the fourth vertebra, and the first judgment result is yes. When the third judgment result is yes, determine that the first output image does not include a complete image of the third vertebra and the fourth vertebra, and the first judgment result is no, wherein the first outer contour is: the outer contour of the area recognized by the first semantic segmentation network model in the first output image.

[0084] Optionally, in the above technical solution, determining whether the second output image 220 includes a complete image of the airway to obtain a second determination result includes:

[0085] Determine whether there are any overlapping points between the second outer contour and the outer contour of the second output image to obtain a fourth judgment result. When the fourth judgment result is no, it is determined that the second output image includes a complete image of the airway, and the second judgment result is yes. When the fourth judgment result is yes, it is determined that the second output image does not include a complete image of the airway, and the second judgment result is no, wherein the second outer contour is: the outer contour of the area identified by the second semantic segmentation network model in the second output image.

[0086] Optionally, in the above technical solution, the first positioning and marking module 210 is further used to: when the first judgment result is no, re-collect the cervical vertebra image of the first user as the first cervical vertebra image;

[0087] The second positioning and marking module 220 is further used for: when the second judgment result is no, re-collecting the cervical vertebra image of the second user as the second cervical vertebra image.

[0088] The above parameters and steps of each unit module implementing corresponding functions in the vertebra and airway positioning system 200 of the present invention can refer to the parameters and steps in the embodiment of the vertebra and airway positioning method above, and will not be repeated here.

[0089] Those skilled in the art will appreciate that the present invention may be implemented as a system, method or computer program product.

[0090] Therefore, the present disclosure may be specifically implemented in the following forms, namely: it may be completely hardware, it may be completely software (including firmware, resident software, microcode, etc.), or it may be a combination of hardware and software, generally referred to herein as a "circuit", "module" or "system". In addition, in some embodiments, the present invention may also be implemented in the form of a computer program product in one or more computer-readable media, and the computer-readable medium contains computer-readable program code.

[0091] Any combination of one or more computer-readable media can be used. Computer-readable media can be computer-readable signal media or computer-readable storage media. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices or devices, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: electrical connections with one or more wires, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the above. In this document, computer-readable storage media can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0092] Although the embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and are not to be construed as limitations of the present invention. A person skilled in the art may change, modify, replace and vary the above embodiments within the scope of the present invention.

Claims

1. A method for positioning vertebrae and airways, characterized in that: include: Acquire a first cervical vertebra image of a first user, input the first cervical vertebra image into a pre-trained first semantic segmentation network model for identifying vertebrae, obtain a first output image, determine whether the first output image contains complete images of the third vertebra and the fourth vertebra, obtain a first determination result, and when the first determination result is yes, input the first output image into a first HRnet network model, locate the third vertebra and the fourth vertebra, and mark them in the first cervical vertebra image; Obtain a second cervical vertebra image of a second user, input the second cervical vertebra image into a pre-trained second semantic segmentation network model for identifying the airway, obtain a second output image, determine whether the second output image contains a complete image of the airway, and obtain a second judgment result. When the second judgment result is yes, input the second output image into a second HRnet network model, locate the airway, and mark it in the second cervical vertebra image; Determining whether the first output image includes complete images of the third vertebra and the fourth vertebra to obtain a first determination result includes: Determine whether there is an overlap between the first outer contour and the outer contour of the first output image, and obtain a third judgment result; when the third judgment result is no, determine that the first output image includes a complete image of the third vertebra and the fourth vertebra, and the first judgment result is yes; when the third judgment result is yes, determine that the first output image does not include a complete image of the third vertebra and the fourth vertebra, and the first judgment result is no, wherein the first outer contour is: the outer contour of the area recognized by the first semantic segmentation network model in the first output image; Determining whether the second output image includes a complete image of the airway to obtain a second determination result includes: Determine whether there are any overlapping points between the second outer contour and the outer contour of the second output image to obtain a fourth judgment result, and when the fourth judgment result is no, determine that the second output image includes a complete image of the airway, and the second judgment result is yes, and when the fourth judgment result is yes, determine that the second output image does not include a complete image of the airway, and the second judgment result is no, wherein the second outer contour is: the outer contour of the area recognized by the second semantic segmentation network model in the second output image.

2. A method for positioning vertebrae and airways according to claim 1, characterized in that: Also includes: When the first judgment result is no, re-acquiring the cervical vertebra image of the first user as the first cervical vertebra image; When the second judgment result is no, the cervical vertebra image of the second user is recaptured as the second cervical vertebra image.

3. A vertebra and airway positioning system, characterized in that: It includes a first positioning and marking module and a second positioning and marking module; The first positioning and annotation module is used to: obtain a first cervical vertebra image of a first user, input the first cervical vertebra image into a pre-trained first semantic segmentation network model for identifying vertebrae, obtain a first output image, determine whether the first output image contains complete images of the third vertebra and the fourth vertebra, obtain a first judgment result, and when the first judgment result is yes, input the first output image into a first HRnet network model, locate the third vertebra and the fourth vertebra, and annotate them in the first cervical vertebra image; The second positioning and annotation module is used to: obtain a second cervical vertebra image of a second user, input the second cervical vertebra image into a pre-trained second semantic segmentation network model for identifying airways, obtain a second output image, determine whether the second output image contains a complete image of the airway, and obtain a second judgment result. When the second judgment result is yes, input the second output image into a second HRnet network model, locate the airway, and annotate it in the second cervical vertebra image; The process of the first positioning and marking module determining whether the first output image contains complete images of the third vertebra and the fourth vertebra includes: Determine whether there is an overlap between the first outer contour and the outer contour of the first output image, and obtain a third judgment result; when the third judgment result is no, determine that the first output image includes a complete image of the third vertebra and the fourth vertebra, and the first judgment result is yes; when the third judgment result is yes, determine that the first output image does not include a complete image of the third vertebra and the fourth vertebra, and the first judgment result is no, wherein the first outer contour is: the outer contour of the area recognized by the first semantic segmentation network model in the first output image; Determining whether the second output image includes a complete image of the airway to obtain a second determination result includes: Determine whether there are any overlapping points between the second outer contour and the outer contour of the second output image to obtain a fourth judgment result, and when the fourth judgment result is no, determine that the second output image includes a complete image of the airway, and the second judgment result is yes, and when the fourth judgment result is yes, determine that the second output image does not include a complete image of the airway, and the second judgment result is no, wherein the second outer contour is: the outer contour of the area recognized by the second semantic segmentation network model in the second output image.

4. A vertebra and airway positioning system according to claim 3, characterized in that: The first positioning and marking module is further used for: when the first judgment result is no, re-collecting the cervical vertebra image of the first user as the first cervical vertebra image; The second positioning and marking module is further used for: when the second judgment result is no, re-collecting the cervical vertebra image of the second user as the second cervical vertebra image.

Images