An automatic segmentation method and system for oral and maxillofacial CT images

Abstract

This invention relates to the field of edge segmentation technology, specifically to an automatic segmentation method and system for CT images of the oral and maxillofacial region. First, local geometric texture matching is performed between the current skeletal model data and historical skeletal model data to filter out key feature points representing key textures in the CT slice images in the spatial dimension. Then, based on the temporal stability of the true edge texture, and considering the edge position conformity and temporal changes of each pixel location in various CT slice images, edge feature values ​​representing the true edge probability of each pixel location are determined. Finally, based on the distribution of edge feature values ​​and the positional distribution of key feature points, true edge lines are filtered out, resulting in higher accuracy in image segmentation of the current CT slice images corresponding to the oral and maxillofacial region based on true edge lines.

Description

Technical Field

[0001] This invention relates to the field of edge segmentation technology, specifically to an automatic segmentation method and system for oral and maxillofacial CT images. Background Technology

[0002] Automatic segmentation of oral and maxillofacial CT images is a technique that uses computers to automatically process CT images and extract structures and regions of the oral cavity and maxillofacial region. This improves the efficiency of medical image data analysis and is therefore widely used in the field of medical data mining and processing. Current techniques typically use the Canny edge detection method to directly detect edges in the corresponding CT slices of the oral and maxillofacial region, and then perform image segmentation based on the obtained edge connected components.

[0003] When performing automatic segmentation of oral and maxillofacial CT images, patients' conditions are diverse, resulting in low contrast between soft tissues or blurred structural boundaries, especially the boundary between soft tissue and bone. At the same time, false edges caused by noise interference can also affect the accuracy of image segmentation. Therefore, the accuracy of image segmentation based on edge detection results of a single CT slice image is low. Summary of the Invention

[0004] To address the low accuracy of image segmentation based on edge detection results from a single CT slice image in existing technologies, this application aims to provide an automatic segmentation method and system for oral and maxillofacial CT images. The specific technical solution adopted is as follows: The first aspect of this application provides an automatic segmentation method for CT images of the oral and maxillofacial region, including: Acquire all CT slice images of the current patient's oral and maxillofacial region, current bone model data, and historical bone model data of all historical patients' oral and maxillofacial regions; Based on the local geometric texture matching between the current bone model data and the historical bone model data, key feature points in the CT slice images are selected; based on the edge position matching of each pixel position in each CT slice image, the edge probability of each pixel position in each CT slice image is determined; based on the change of edge probability of each pixel position in each CT slice image, the edge feature value of each pixel position in the current CT slice image is determined. Image segmentation is performed on the current CT slice image corresponding to the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points.

[0005] Furthermore, the process of obtaining the key feature points includes: All initial feature points in the current bone model data are determined by corner detection method; each historical bone model data is projected onto the current bone model data to obtain the reference feature point corresponding to each initial feature point in each historical bone model data; Based on the similarity of curvature and normal vector between each initial feature point in the current skeletal model data and the corresponding reference feature point in each historical skeletal model data, determine the local matching coefficient of each initial feature point in each historical skeletal model data. The overall matching degree is determined based on the mean of the local matching coefficients of each initial feature point across all historical skeletal model data; then, feature points in the current skeletal model data are selected based on the overall matching degree. Based on the location of each selected feature point in the current skeletal model data, the key feature points of each selected feature point in each CT slice image are determined.

[0006] Furthermore, the process of obtaining the local matching coefficients includes: The difference between the curvature of each initial feature point in the current skeletal model data and the curvature of its corresponding reference feature point in each historical skeletal model data is normalized to determine the curvature deviation coefficient of each initial feature point in each historical skeletal model data. The angle between the normal vector of each initial feature point in the current skeletal model data and the normal vector of its corresponding reference feature point in each historical skeletal model data is normalized to determine the normal vector deviation coefficient of each initial feature point in each historical skeletal model data. The product of the curvature deviation coefficient and the normal vector deviation coefficient is negatively correlated to determine the local matching coefficient of each initial feature point in each historical skeletal model data.

[0007] Furthermore, the process of obtaining the selected feature points includes: Initial feature points with an overall matching degree greater than a preset matching threshold are used as the selected feature points in the current skeletal model data.

[0008] Furthermore, the process of obtaining the edge probability includes: Based on the edge information distribution of each CT slice image, determine the edge information parameters corresponding to each pixel position; sequentially use each CT slice image as the target slice image; use other CT slice images besides the target slice images as reference CT slice images; and sequentially use each pixel position as the target position. The angle between the normal vector of the edge point corresponding to the target location in the target slice image and the normal vector of the corresponding edge point in each reference CT slice image is positively correlated to determine the corresponding edge direction consistency; wherein, when the target location is not an edge point in the target slice image or is not an edge point in the corresponding reference CT slice image, the corresponding edge direction consistency is set to a preset edge parameter. The edge confidence level of the target location in each reference CT slice image is determined based on the overall size and relative deviation of the edge information parameters between the target slice image and each reference CT slice image, as well as the consistency of the edge direction. The mean edge confidence of the target location in all reference CT slice images is normalized to determine the edge probability of the target location in the target CT slice image.

[0009] Furthermore, the process of obtaining the edge information parameters includes: For each CT slice image, double-threshold edge detection is performed using the Canny edge detection algorithm to determine all edge information images. Among all edge information images corresponding to each CT slice image, the ratio between the number of edge information images with edge points at each pixel location and the total number of edge information images is used as the edge information parameter corresponding to each pixel location in each CT slice image.

[0010] Furthermore, the process of obtaining the edge confidence level includes: The sum of the edge information parameters corresponding to the target location in the target slice image and the corresponding edge information parameters in each reference CT slice image is used as the corresponding edge saliency. The difference between the edge information parameters corresponding to the target location in the target slice image and the edge information parameters corresponding to each reference CT slice image is negatively correlated to determine the significant consistency of the corresponding edges. The edge confidence level of the target location in each reference CT slice image is determined based on the product of the edge direction consistency, the edge saliency consistency, and the edge saliency.

[0011] Furthermore, the process of obtaining the edge feature values ​​includes: All CT slice images are arranged sequentially according to the order in which they were acquired to determine a CT slice image sequence. In the CT slice image sequence, the difference between the edge probability of each pixel position in each CT slice image and the edge probability in the previous CT slice image is negatively correlated and mapped to determine the edge change parameter of each pixel position in each CT slice image. The edge validity of each pixel position is determined based on the mean of the edge change parameter of each pixel position in all CT slice images. The edge feature value of each pixel in the current CT slice image is determined by multiplying the edge probability of each pixel location in the current CT slice image with the edge validity.

[0012] Furthermore, the process of segmenting the current CT slice image corresponding to the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points includes: The Canny edge detection algorithm is used to obtain all edge lines in the current CT slice image; the mean value of the edge feature value of all pixels on each edge line is calculated to determine the corresponding edge confidence. Among all edges lacking key feature points, all edge lines with an edge confidence score less than a preset confidence threshold are designated as pseudo-edge lines; other edge lines outside the pseudo-edge lines are designated as true edge lines in the current CT slice image; and image segmentation is performed on the current CT slice image corresponding to the oral and maxillofacial region based on the true edge lines.

[0013] Secondly, this application provides an automatic segmentation system for oral and maxillofacial CT images, the system comprising: The data acquisition and preprocessing module is used to acquire all CT slice images of the current patient's oral and maxillofacial region, current bone model data, and historical bone model data of all historical patients' oral and maxillofacial regions. The parameter determination module is used to filter out key feature points in the CT slice image based on the local geometric texture matching between the current bone model data and the historical bone model data; determine the edge probability of each pixel position in each CT slice image based on the edge position matching of each pixel position in each CT slice image; and determine the edge feature value of each pixel position in the current CT slice image based on the change of edge probability of each pixel position in each CT slice image. The image segmentation module is used to segment the current CT slice image of the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points.

[0014] Thirdly, this application provides a computer device including a memory and a processor. The memory is used to store computer program code, and the processor is used to call and run the computer program code from the memory to perform the method as described in the first aspect of this application or any embodiment of the first aspect.

[0015] Fourthly, this application provides a computer program product comprising computer program code, which, when executed, performs the method as described in the first aspect of this application or any embodiment thereof.

[0016] Fifthly, this application provides a computer-readable storage medium that stores computer program code, which, when executed, performs the method as described in the first aspect of this application or any embodiment thereof.

[0017] This application has the following beneficial effects: This application first performs local geometric texture matching between the current skeletal model data and historical skeletal model data to filter out key feature points representing key textures in CT slice images in the spatial dimension. Then, based on the temporal stability characteristics of real edge textures, and based on the edge position conformity and temporal changes of each pixel position in various CT slice images, it determines the edge feature value representing the real edge probability of each pixel position. Finally, based on the distribution of edge feature values ​​and the positional distribution of key feature points, it filters out real edge lines, making the image segmentation of the current CT slice image corresponding to the oral and maxillofacial region more accurate based on real edge lines. Attached Figure Description

[0018] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 A flowchart of an automatic segmentation method for oral and maxillofacial CT images provided in one embodiment of the present invention; Figure 2 This is a structural diagram of an automatic segmentation system for oral and maxillofacial CT images provided in one embodiment of the present invention; Figure 3 This is a schematic diagram of a computer device structure provided in one embodiment of the present invention. Detailed Implementation

[0020] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of an automatic segmentation method and system for oral and maxillofacial CT images proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment, and specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as implying or suggesting relative importance or implicitly indicating the number of indicated technical features. Thus, a feature defined with "first" or "second" may explicitly or implicitly include one or more of that feature.

[0021] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0022] The following description, in conjunction with the accompanying drawings, details the specific scheme of the automatic segmentation method and system for oral and maxillofacial CT images provided by this invention.

[0023] This application provides an automatic segmentation method for oral and maxillofacial CT images. Please refer to [link to relevant documentation]. Figure 1 The diagram illustrates a flowchart of an automatic segmentation method for oral and maxillofacial CT images according to an embodiment of the present invention. The method includes: Step S101: Obtain all CT slice images of the current patient's oral and maxillofacial region, current bone model data, and historical bone model data of all historical patients' oral and maxillofacial regions.

[0024] For each patient (including current and historical patients), firstly, all initial CT images corresponding to the oral and maxillofacial region of each patient are acquired. Then, based on the significant density difference between bones and group leaders, bone information will have obvious highlights in the initial CT images. The Otsu thresholding method is used to filter out the bone information regions in each initial CT image. Furthermore, considering that there is a certain degree of connectivity in the bone data in the initial CT images, the bone information regions in all initial CT images are further aligned to determine the bone model data of each patient. Thus, the current bone model data of the current patient and the historical bone model data of each historical patient are determined.

[0025] It should be noted that, in order to obtain complete skeletal model data, the initial CT images acquired in this embodiment of the invention need to completely cover the patient's oral and maxillofacial region; and the CT slice images in this embodiment of the invention are image data corresponding to one type of initial CT image at various sampling times. By analyzing all CT slice images, the structures in each type of initial CT image can be segmented more accurately; and the processing procedure for CT slice images corresponding to different types of initial CT images is the same, and will not be further elaborated hereafter. The sampling frequency can be adjusted according to the specific implementation environment; in this embodiment of the invention, it is set to be acquired once per minute.

[0026] Step S102: Based on the local geometric texture matching between the current bone model data and the historical bone model data, key feature points in the CT slice images are selected; based on the edge position matching of each pixel position in each CT slice image, the edge probability of each pixel position in each CT slice image is determined; based on the change of edge probability of each pixel position in each CT slice image, the edge feature value of each pixel position in the current CT slice image is determined.

[0027] The skeletal model data is a complete model obtained by aligning different types of CT slice images. For the current skeletal model, if a certain edge texture also has similar features at the corresponding position in other historical skeletal model data, it indicates that the geometric features of the corresponding edge texture are relatively stable and belong to a relatively critical feature. The edge information represented by the corresponding texture usually has high reliability. Therefore, this embodiment of the invention first selects key feature points in the CT slice image based on the local geometric texture matching between the current skeletal model data and the historical skeletal model data.

[0028] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining key feature points includes: The corner detection method determines all initial feature points in the current skeletal model data; each historical skeletal model data is projected onto the current skeletal model data to obtain the reference feature point corresponding to each initial feature point in each historical skeletal model data; corner detection can quickly filter out candidate points that represent edge texture features, avoiding the waste of computing resources in a large number of irrelevant regions, making the calculation efficiency of key feature points higher.

[0029] Based on the similarity in curvature and normal vector between each initial feature point in the current skeletal model data and the corresponding reference feature point in each historical skeletal model data, the local matching coefficient of each initial feature point under each historical skeletal model data is determined; preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the local matching coefficient includes: The curvature deviation coefficient of each initial feature point in the current skeletal model data is determined by normalizing the difference between its curvature and the curvature of the corresponding reference feature point in each historical skeletal model data. The angle between the normal vector of each initial feature point in the current skeletal model data and the normal vector of the corresponding reference feature point in each historical skeletal model data is also normalized to determine the normal vector deviation coefficient of each initial feature point in each historical skeletal model data. Finally, the product of the curvature deviation coefficient and the normal vector deviation coefficient is negatively correlated to determine the local matching coefficient of each initial feature point in each historical skeletal model data.

[0030] For each initial feature point and the reference feature point in each historical bone model, the more similar their curvature magnitudes and the smaller the angle between their normal vectors, the more similar the shape features of the bone data represented by the initial feature point and the corresponding reference feature point are. This indicates that the texture features represented by the initial feature point are more stable, and the edge information of the corresponding pixel position in the corresponding CT slice image is more reliable. Normalization can reduce the influence of dimensions on the calculation results and avoid the impact of different dimensions on the calculation results.

[0031] In one specific implementation of this invention, the process of obtaining the local matching coefficients is expressed by the formula: ;in, Initial feature points in the current skeletal model data In the Local matching coefficients based on historical skeletal model data; Initial feature points in the current skeletal model data The curvature; Initial feature points in the current skeletal model data In the The curvature of the corresponding reference feature points in the historical skeletal model data; It is the absolute value symbol; It is a minimum-maximum normalization function; Initial feature points in the current skeletal model data In the Curvature deviation coefficient corresponding to each historical skeletal model data; Initial feature points in the current skeletal model data The corresponding normal vector and its th The angle between the normal vectors of the corresponding reference feature points in the historical skeletal model data; Initial feature points in the current skeletal model data In the The normal vector deviation coefficient corresponding to each historical skeletal model data; It is an exponential function with the natural constant as its base.

[0032] Further combining all historical skeletal model data, the overall matching degree is determined based on the average local matching coefficient of each initial feature point across all historical skeletal model data. A higher overall matching degree indicates more stable geometric features represented by the corresponding initial feature point. Therefore, feature points in the current skeletal model data are further selected based on the overall matching degree. Preferably, in some possible implementations of this invention, the process of obtaining the selected feature points includes: using initial feature points with an overall matching degree greater than a preset matching threshold as selected feature points in the current skeletal model data. In a specific implementation of this invention, the preset matching threshold is set to 0.7, which can be adjusted according to the specific implementation environment. Finally, based on the characteristic that skeletal model data is constructed from initial CT images, the key feature points of each selected feature point in each CT slice image are determined according to the position of each selected feature point in the current skeletal model data. Through spatial dimension analysis, the obtained key feature points can represent edge information with higher reliability, resulting in higher accuracy in image segmentation by combining the edges corresponding to the key feature points.

[0033] Furthermore, considering that real edge information is relatively stable in time, while pseudo-edges may have different edge behaviors in time, the edge probability of each pixel in each CT slice image is determined based on the edge position conformity of each pixel position in each CT slice image.

[0034] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the edge probability includes: determining the edge information parameters corresponding to the position of each pixel point based on the edge information distribution of each CT slice image; preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the edge information parameters includes: For each CT slice image, double-threshold edge detection is performed using the Canny edge detection algorithm to determine all edge information images. Among all edge information images corresponding to each CT slice image, the ratio between the number of edge information images containing an edge at each pixel location and the total number of edge information images is used as the edge information parameter for each pixel location in each CT slice image. For each pixel location, the more times it is considered an edge in all edge information images (i.e., the larger the edge information parameter), the more reliable the edge texture at that pixel location, and the higher the corresponding edge probability.

[0035] Each CT slice image is sequentially used as the target slice image; other CT slice images besides the target slice image are used as reference CT slice images; each pixel position is sequentially used as the target position; the angle between the normal vector of the edge point corresponding to the target position in the target slice image and the normal vector of the edge point corresponding to the target position in each reference CT slice image is positively correlated and mapped to determine the corresponding edge direction consistency; wherein, when the target position is not an edge point in the target slice image or is not an edge point in the corresponding reference CT slice image, the corresponding edge direction consistency is set to a preset edge parameter. For the same pixel position, if it exhibits the same edge morphology in CT slice images of different time series, it indicates that the stability of its corresponding edge information is higher; therefore, the greater the edge direction consistency, the higher the temporal stability of the edge texture represented by the target position, the higher the reliability of its corresponding edge texture, and the greater the edge probability. In a specific implementation of this invention, the preset edge parameter is set to 0, which can be adjusted according to the specific implementation environment, and will not be further elaborated here.

[0036] Based on the overall size and relative deviation of the edge information parameters between the target location in the target slice image and each reference CT slice image, as well as the consistency of the edge direction, the edge confidence level of the target location in each reference CT slice image is determined; preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining the edge confidence level includes: The sum of the edge information parameters corresponding to the target location in the target slice image and the corresponding edge information parameters in each reference CT slice image is used as the corresponding edge saliency. Since the edge information parameters only represent the reliability of the edge texture of each pixel location in the same CT slice image, for each pixel location, if it also has a large edge information parameter in other reference CT slice images, that is, the larger the edge saliency, then it means that the edge texture corresponding to that pixel location is more saliency in time, the possibility of it being a false edge is lower, and the reliability of the edge texture is higher.

[0037] By negatively correlating the differences between the edge information parameters corresponding to the target location in the target slice image and the corresponding edge information parameters in each reference CT slice image, the significant consistency of the edges is determined. A greater significant consistency indicates stronger temporal stability of the edge information, making it less consistent with the unstable characteristics of pseudo-edges, and thus increasing the reliability of the corresponding edge texture.

[0038] Furthermore, by combining the characteristics of edge texture credibility represented by edge direction consistency, edge saliency consistency, and edge saliency, the edge credibility of the target location under each reference CT slice image is determined based on the product of edge direction consistency, edge saliency consistency, and edge saliency. This ensures that the higher the edge credibility, the more realistic the edge texture represented by the target location under the reference of the corresponding CT slice image. Therefore, by synthesizing all reference CT slice images, the mean of the edge credibility of the target location under all reference CT slice images is normalized to determine the edge probability of the target location in the target CT slice image. This results in higher accuracy in representing the edge information of each pixel location based on the edge probability.

[0039] In one specific implementation of this invention, the process of obtaining the marginal probability is expressed by the following formula: ;in, For target location In the target slice image Marginal probabilities in; Slice the image for the target The number of corresponding reference CT slice images; For target location In the target slice image The corresponding edge point normal vector in the middle and the first edge point normal vector in the middle are the same as the first edge point normal vector in the middle. The angle between the normal vectors of corresponding edge points in a reference CT slice image; For target location In the target slice image The corresponding number The edge direction consistency under a reference CT slice image is positively correlated through this method, so that the value of edge direction consistency is always greater than or equal to 1, that is, to prevent the value of 0 from affecting the calculation process. For target location In the target slice image The corresponding edge information parameters; For target location In the target slice image The corresponding number Edge information parameters corresponding to a reference CT slice image; For target location In the target slice image The corresponding number Edge saliency under a reference CT slice image; For target location In the target slice image The corresponding number Significant edge uniformity under reference CT slice images; For target location In the target slice image The corresponding number Edge confidence level under a reference CT slice image.

[0040] Furthermore, considering that true edges do not change significantly over time, and their corresponding edge probabilities typically exhibit small temporal variations across all CT slice images, while false edges, due to their unstable nature, often show significant changes in edge probabilities, we further determine the edge feature value of each pixel in the current CT slice image based on the edge probability changes of each pixel location across various CT slice images. This ensures that the larger the edge feature value, the higher the reliability that the edge information at the corresponding pixel location belongs to a true edge.

[0041] Preferably, in some possible implementations of the embodiments of the present invention, the process of obtaining edge feature values ​​includes: All CT slice images are arranged sequentially according to the order in which they were acquired to determine the CT slice image sequence. Within this sequence, the difference between the edge probability of each pixel location in each CT slice image and the edge probability in the previous CT slice image is negatively correlated to determine the edge variation parameter of each pixel location in each CT slice image. The edge validity of each pixel location is determined based on the mean of the edge variation parameters across all CT slice images. A higher edge validity indicates more stable edge probability across all CT slice images, suggesting a higher probability of representing a true edge texture. It should be noted that the edge variation parameter of each pixel location in the first CT slice image is set to the mean of the edge variation parameters in the other CT slice images. It should also be noted that the difference in this embodiment represents the absolute value of the difference, which will not be further elaborated here.

[0042] Furthermore, considering that the edge probability of pixels without edges is relatively stable (approaching 0), in order to avoid affecting the calculation of edge feature values, the edge probability of the current slice image is further combined with the edge probability of each pixel position in the current CT slice image. The edge feature value of each pixel position in the current CT slice image is determined by multiplying the edge probability of each pixel position in the current CT slice image with the edge validity. This makes the edge feature value larger, and the higher the probability that the edge information of the corresponding pixel position belongs to the real edge texture.

[0043] In one specific implementation of this invention, the process of obtaining edge feature values ​​is expressed by the following formula: ;in, The pixel position in the current CT slice image Edge feature values; pixel position Edge probabilities in the current CT slice image; Pixel positions in a CT slice image sequence In the The difference between the edge probability in a CT slice image and the edge probability in the previous CT slice image; pixel position Edge validity; pixel position In the Edge variation parameters in a CT slice image.

[0044] Step S103: Perform image segmentation on the current CT slice image corresponding to the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points.

[0045] Furthermore, based on the characteristics of key feature points representing edge information in the spatial dimension and edge feature values ​​representing the credibility of real edges, image segmentation is performed on the current CT slice image corresponding to the oral and maxillofacial region according to the distribution of edge feature values ​​and the location distribution of key feature points.

[0046] In one specific implementation of this invention, all edge lines in the current CT slice image are obtained using the Canny edge detection algorithm; the mean value of the edge feature values ​​at all pixel locations on each edge line is calculated to determine the corresponding edge confidence. It should be noted that when dividing the edge lines, nodes with bifurcations are used as interval points for segmentation to determine all edge lines.

[0047] For each edge line, a higher edge confidence score indicates a higher reliability and a greater likelihood that it is a true edge line. When key feature points exist on an edge line, these points represent edge textures with high spatial confidence; therefore, the corresponding edge lines are usually not pseudo-edges. Therefore, among all edges without key feature points, all edge lines with an edge confidence score less than a preset confidence threshold are designated as pseudo-edge lines. Other edge lines outside the pseudo-edge lines are designated as true edge lines in the current CT slice image. Image segmentation is then performed on the current CT slice image corresponding to the oral and maxillofacial region based on the true edge lines.

[0048] In summary, an automatic segmentation method for oral and maxillofacial CT images first performs local geometric texture matching between the current skeletal model data and historical skeletal model data to filter out key feature points representing key textures in the CT slice images in the spatial dimension. Then, based on the temporal stability characteristics of the real edge texture, and based on the edge position conformity and temporal changes of each pixel position in various CT slice images, the edge feature value representing the real edge probability of each pixel position is determined. Finally, based on the distribution of edge feature values ​​and the positional distribution of key feature points, the real edge lines are filtered out, resulting in higher accuracy in image segmentation of the current CT slice images corresponding to the oral and maxillofacial region based on the real edge lines.

[0049] This application also provides an automatic segmentation system for oral and maxillofacial CT images. Please refer to [link to relevant documentation]. Figure 2 The diagram shows a structural diagram of an automatic segmentation system for oral and maxillofacial CT images provided by an embodiment of the present invention. The system includes: a data acquisition and preprocessing module 201, a parameter determination module 202, and an image segmentation module 203.

[0050] The data acquisition and preprocessing module 201 is used to acquire all CT slice images of the current patient's oral and maxillofacial region, current bone model data, and historical bone model data of all historical patients' oral and maxillofacial regions. The parameter determination module 202 is used to filter out key feature points in CT slice images based on the local geometric texture matching between the current bone model data and historical bone model data; determine the edge probability of each pixel position in each CT slice image based on the edge position matching of each pixel position in each CT slice image; and determine the edge feature value of each pixel position in the current CT slice image based on the edge probability change of each pixel position in each CT slice image. The image segmentation module 203 is used to segment the current CT slice image corresponding to the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points.

[0051] It should be noted that the system provided in the above embodiments is only an example of the division of the above functional modules. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the automatic segmentation system for oral and maxillofacial CT images and the automatic segmentation method for oral and maxillofacial CT images provided in the above embodiments belong to the same concept, and their specific implementation process is detailed in the method embodiments, which will not be repeated here.

[0052] This application also provides a computer device; please refer to [link / reference]. Figure 3 The diagram illustrates a computer device structure according to an embodiment of the present invention. The computer device includes a memory 301, a processor 302, and a computer program 303 stored in the memory 301 and running on the processor 302. When the processor 302 executes the computer program 303, the computer device can execute any of the aforementioned automatic segmentation methods for oral and maxillofacial CT images.

[0053] This application also provides a computer program product that, when run on a computer device, enables the computer device to execute any of the aforementioned automatic segmentation methods for oral and maxillofacial CT images.

[0054] This application also provides a computer-readable storage medium storing computer program code. When the computer program code is run on a computer device, the computer device can execute any of the aforementioned automatic segmentation methods for oral and maxillofacial CT images.

[0055] In the embodiments provided in this application, it should be understood that the computer device, computer program product and computer-readable storage medium provided are all used to perform the corresponding methods provided above, and therefore the beneficial effects they can achieve can be referred to the beneficial effects of the methods provided above, which will not be repeated here.

[0056] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0057] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

Claims

1. An automatic segmentation method for oral and maxillofacial CT images, characterized in that, The method includes: Acquire all CT slice images of the current patient's oral and maxillofacial region, current bone model data, and historical bone model data of all historical patients' oral and maxillofacial regions; Based on the local geometric texture matching between the current bone model data and the historical bone model data, key feature points in the CT slice images are selected; based on the edge position matching of each pixel position in each CT slice image, the edge probability of each pixel position in each CT slice image is determined; based on the change of edge probability of each pixel position in each CT slice image, the edge feature value of each pixel position in the current CT slice image is determined. Image segmentation is performed on the current CT slice image corresponding to the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points; The process of obtaining the key feature points includes: All initial feature points in the current bone model data are determined by corner detection method; each historical bone model data is projected onto the current bone model data to obtain the reference feature point corresponding to each initial feature point in each historical bone model data; Based on the similarity of curvature and normal vector between each initial feature point in the current skeletal model data and the corresponding reference feature point in each historical skeletal model data, determine the local matching coefficient of each initial feature point in each historical skeletal model data. The overall matching degree is determined based on the mean of the local matching coefficients of each initial feature point across all historical skeletal model data; then, feature points in the current skeletal model data are selected based on the overall matching degree. Based on the position of each selected feature point in the current skeletal model data, determine the key feature point of each selected feature point in each CT slice image; The process of obtaining the local matching coefficients includes: The curvature deviation coefficient of each initial feature point in the current bone model data is normalized to determine the difference between the curvature of its corresponding reference feature point in each historical bone model data. The angle between the normal vector of each initial feature point in the current bone model data and the normal vector of its corresponding reference feature point in each historical bone model data is normalized to determine the normal vector deviation coefficient of each initial feature point in each historical bone model data. The product of the curvature deviation coefficient and the normal vector deviation coefficient is negatively correlated to determine the local matching coefficient of each initial feature point in each historical bone model data. The process of obtaining the marginal probability includes: Based on the edge information distribution of each CT slice image, determine the edge information parameters corresponding to each pixel position; sequentially use each CT slice image as the target slice image; use other CT slice images besides the target slice images as reference CT slice images; and sequentially use each pixel position as the target position. The angle between the normal vector of the edge point corresponding to the target location in the target slice image and the normal vector of the corresponding edge point in each reference CT slice image is positively correlated to determine the corresponding edge direction consistency; wherein, when the target location is not an edge point in the target slice image or is not an edge point in the corresponding reference CT slice image, the corresponding edge direction consistency is set to a preset edge parameter. The edge confidence level of the target location in each reference CT slice image is determined based on the overall size and relative deviation of the edge information parameters between the target slice image and each reference CT slice image, as well as the consistency of the edge direction. The mean edge confidence of the target location in all reference CT slice images is normalized to determine the edge probability of the target location in the target CT slice image. The process of obtaining the edge information parameters includes: For each CT slice image, double-threshold edge detection is performed using the Canny edge detection algorithm to determine all edge information images. Among all edge information images corresponding to each CT slice image, the ratio between the number of edge information images with edge points at each pixel location and the total number of edge information images is used as the edge information parameter corresponding to each pixel location in each CT slice image.

2. The automatic segmentation method for oral and maxillofacial CT images according to claim 1, characterized in that, The process of obtaining the selected feature points includes: Initial feature points with an overall matching degree greater than a preset matching threshold are used as the selected feature points in the current skeletal model data.

3. The automatic segmentation method for oral and maxillofacial CT images according to claim 1, characterized in that, The process of obtaining the edge credibility includes: The sum of the edge information parameters corresponding to the target location in the target slice image and the corresponding edge information parameters in each reference CT slice image is used as the corresponding edge saliency. The difference between the edge information parameters corresponding to the target location in the target slice image and the edge information parameters corresponding to each reference CT slice image is negatively correlated to determine the significant consistency of the corresponding edges. The edge confidence level of the target location in each reference CT slice image is determined based on the product of the edge direction consistency, the edge saliency consistency, and the edge saliency.

4. The automatic segmentation method for oral and maxillofacial CT images according to claim 1, characterized in that, The process of obtaining the edge feature values ​​includes: All CT slice images are arranged sequentially according to the order in which they were acquired to determine a CT slice image sequence. In the CT slice image sequence, the difference between the edge probability of each pixel position in each CT slice image and the edge probability in the previous CT slice image is negatively correlated and mapped to determine the edge change parameter of each pixel position in each CT slice image. The edge validity of each pixel position is determined based on the mean of the edge change parameter of each pixel position in all CT slice images. The edge feature value of each pixel in the current CT slice image is determined by multiplying the edge probability of each pixel location in the current CT slice image with the edge validity.

5. The automatic segmentation method for oral and maxillofacial CT images according to claim 1, characterized in that, The process of segmenting the current CT slice image of the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points includes: The Canny edge detection algorithm is used to obtain all edge lines in the current CT slice image; the mean value of the edge feature value of all pixels on each edge line is calculated to determine the corresponding edge confidence. Among all edges lacking key feature points, all edge lines with an edge confidence score less than a preset confidence threshold are designated as pseudo-edge lines; other edge lines outside the pseudo-edge lines are designated as true edge lines in the current CT slice image; and image segmentation is performed on the current CT slice image corresponding to the oral and maxillofacial region based on the true edge lines.

6. An automatic segmentation system for oral and maxillofacial CT images, characterized in that, The system is used to implement the automatic segmentation method for oral and maxillofacial CT images as described in any one of claims 1 to 5, the system comprising: The data acquisition and preprocessing module is used to acquire all CT slice images of the current patient's oral and maxillofacial region, current bone model data, and historical bone model data of all historical patients' oral and maxillofacial regions. The parameter determination module is used to filter out key feature points in the CT slice image based on the local geometric texture matching between the current bone model data and the historical bone model data; determine the edge probability of each pixel position in each CT slice image based on the edge position matching of each pixel position in each CT slice image; and determine the edge feature value of each pixel position in the current CT slice image based on the change of edge probability of each pixel position in each CT slice image. The image segmentation module is used to segment the current CT slice image of the oral and maxillofacial region based on the distribution of edge feature values ​​and the location distribution of key feature points.

Images