Cervical disc segmentation method combining cervical CT sequence images with multi-view information

Abstract

The application discloses a cervical disc segmentation method of cervical CT sequence image combined with multi-view information, and mainly relates to the cervical CT image segmentation technology. The method comprises the following steps: (1) mapping the original horizontal plane human cervical CT sequence image to the sagittal plane, and performing interactive batch threshold segmentation on the cervical vertebra; (2) determining the cervical disc horizontal plane segmentation sequence image batch; (3) searching a candidate upper surface point set and a candidate lower surface point set; (4) for each candidate upper surface point, if the candidate upper surface point satisfies the cervical disc surface judgment criterion, then the cervical disc is filled along the cervical disc inclination angle; (5) for all the horizontal plane segmentation sequence image batches, steps (3) and (4) are repeated, that is, the segmentation of the cervical disc in the human cervical CT sequence image is completed. The application can perform high-precision segmentation on the cervical disc in the cervical CT sequence image, and is a key prerequisite for the three-dimensional segmentation of the cervical endplate.

Description

Technical Field

[0001] This invention relates to a method for segmenting cervical intervertebral discs in human cervical spine CT sequence images, belonging to the field of image segmentation. Background Technology

[0002] In medicine, the syndrome caused by cervical spondylosis and intervertebral disc degeneration is termed "cervical spondylosis." In recent years, the incidence of cervical spondylosis has shown a rising trend, and the affected population is increasingly younger, attracting significant attention from the medical community. Cervical intervertebral disc replacement is one of the commonly used surgical treatments for cervical spondylosis, offering advantages such as safety and effectiveness. Closely related to cervical disc replacement is the three-dimensional classification of the cervical endplate, which is crucial for guiding the 3D printing of cervical implants. Accurately segmenting the cervical intervertebral disc from cervical CT images is a key prerequisite for the three-dimensional classification of the cervical endplate.

[0003] For CT medical image segmentation, traditional image segmentation methods remain an effective approach. In 2011, researchers used a statistical level set method to segment the spine, employing the expectation-maximization algorithm for initialization and parameter estimation. In 2018, researchers achieved fully automated hip joint segmentation using specific thresholds in CT images and a watershed algorithm, accurately segmenting the femur. However, the optimal threshold needs to be adjusted based on different CT images, resulting in poor generalization of the algorithm. In 2019, researchers proposed a CT liver tumor segmentation method based on subsets, FCN, and an adaptive watershed algorithm. This method achieves high overall segmentation accuracy for liver tumors, but the processing is time-consuming. However, in CT images, the grayscale distribution of cervical intervertebral discs is similar to that of other tissues or organs in the cervical spine (excluding vertebrae), and the boundaries of cervical intervertebral discs are difficult to define in the original horizontal view. These factors increase the difficulty of segmenting cervical intervertebral discs in cervical spine CT sequences. Summary of the Invention

[0004] The purpose of this invention is to provide a method for segmenting cervical intervertebral discs in human cervical spine CT sequence images to solve the above-mentioned problems.

[0005] The present invention achieves the above objectives through the following technical solutions:

[0006] A method for cervical disc segmentation based on cervical CT sequence images with multi-view information includes the following steps:

[0007] (1) Map the original horizontal plane human cervical spine CT sequence image to the sagittal plane to obtain the corresponding sagittal plane cervical spine CT sequence image, and perform interactive batch threshold segmentation on the cervical vertebrae.

[0008] (2) In the sagittal segmentation sequence obtained in step (1), select one with good shape to determine the batch of cervical intervertebral disc horizontal segmentation sequence;

[0009] (3) In the batch of horizontal plane segmentation sequence maps obtained in step (2), search for candidate upper surface point sets and candidate lower surface point sets;

[0010] (4) For each candidate upper surface point in the candidate upper surface point set obtained in step (3), if it meets the cervical intervertebral disc surface judgment criteria, then fill the cervical intervertebral disc along the cervical intervertebral disc tilt angle.

[0011] (5) Repeat steps (3) and (4) for all batches of horizontal plane segmentation sequence images to complete the segmentation of the cervical intervertebral disc in the human cervical spine CT sequence image.

[0012] In the above scheme, step (1) performs interactive batch threshold segmentation of cervical vertebrae in the sagittal plane because the sagittal plane makes it easier to distinguish the intervertebral discs and the "holes" on the cervical vertebrae.

[0013] In the above scheme, the batch of horizontal plane segmentation sequence images mentioned in step (2) can be determined by formula (1) and formula (2):

[0014]

[0015]

[0016] in The average position of each cervical vertebra. and These are the column number and row number of its average position, respectively, Z. i1 and Z i2 These then constitute each batch of cervical intervertebral discs.

[0017] In the above scheme, the candidate upper and lower surface points in step (3) are defined as follows:

[0018] For the set {p(x,y,z)|0<=x <w,0<=y<h,Z i1 <= z <Z i2 V G (x,y,z)=1,V G Each point in (x,y,z+1)=0 is called a candidate upper surface point;

[0019] For the set {p(x,y,z)|0<=x <w,0<=y<h,Z i1 <= z <Z i2 V G (x,y,z)=1,V GEach point in (x,y,z-1)=0 is called a candidate lower surface point;

[0020] Where w and h represent the width and height of the horizontal plane image, respectively, Z i1 <= z <Z i2 Then, as indicated by formula (2), V G (x,y,z) represents the value at column x and row y of the z-th image in the horizontal plane segmentation sequence diagram, where 1 represents the cervical vertebra, 2 represents the cervical intervertebral disc, and 0 represents other tissues or organs.

[0021] The cervical disc tilt angle mentioned in step (4) of the above scheme is defined as follows:

[0022]

[0023] The cervical intervertebral disc surface judgment criterion in step (4) of the above scheme is defined as follows: if a candidate upper surface point can find a candidate lower surface point in the corresponding candidate lower surface point set along the cervical intervertebral disc tilt angle shown in formula (3), then the candidate upper surface point and the candidate lower surface point are determined to be valid. Attached Figure Description

[0024] Figure 1 This is a flowchart illustrating the cervical disc segmentation method for cervical CT sequence images that incorporates multi-view information, as per the present invention.

[0025] Figure 2-1 2-2 shows the correspondence between each plane in the human cervical spine CT sequence image used in the embodiments of the present invention and the coordinate system used, as well as the segmentation results of the cervical vertebrae.

[0026] Figure 3-1 3-2 is a sagittal sequence image of the human cervical spine CT sequence used in the embodiments of the present invention, and the segmentation result of the cervical intervertebral disc.

[0027] Figure 4-1 The original image used in the embodiments of the present invention. Figure 4-2 4-3, 4-4, and 4-5 are diagrams showing the artificial markings of the cervical intervertebral discs, the maximum difference method, the double threshold segmentation method, and the segmentation results of the method of this invention. Detailed Implementation

[0028] The present invention will now be described in more detail with reference to specific embodiments and accompanying drawings. However, these embodiments are merely detailed descriptions of the implementation methods of the present invention and should not be construed as any limitation on the scope of protection of the present invention. The accompanying drawings and the embodiments described below are provided to enable those skilled in the art to understand the present invention more completely and accurately.

[0029] To make the method described in this invention easier to understand and closer to real-world applications, this embodiment uses human cervical spine CT sequence images as a dataset for testing. The image width is w, the height is h, and there are a total of k images.

[0030] Figure 1 In this study, the cervical disc segmentation method based on cervical CT sequence images with multi-view information can be divided into the following steps:

[0031] (1) The original horizontal cervical spine CT sequence image is mapped to the sagittal plane to obtain the corresponding sagittal cervical spine CT sequence image. The mapping result and the relationship between the coordinate system used in this invention and the images of each plane are as follows: Figure 2-1 As shown, interactive batch threshold segmentation of cervical vertebrae was performed on the obtained sagittal cervical CT sequence image to segment out cervical vertebrae that only contain the cervical intervertebral disc, and the corresponding sagittal segmentation sequence image was obtained.

[0032] (2) In the obtained sagittal segmentation sequence image, select one with good shape, calculate the average position of each cervical vertebra, and use the number of rows in the image where the average position of two adjacent cervical vertebrae is located to determine the batch of horizontal segmentation sequence images associated with the intervertebral disc between them. The specific calculation formula is indicated by formula (1) and formula (2).

[0033] (3) In the obtained batch of horizontal plane segmentation sequence maps, search for the candidate upper surface point set and the candidate lower surface point set. The definitions of the candidate upper and lower surface points have been given above.

[0034] (4) For each candidate upper surface point in the obtained candidate upper surface point set, search for a corresponding candidate lower surface point in the candidate lower surface point set along the cervical intervertebral disc tilt angle shown in formula (3). If it exists, then all pixels between the two belong to the segmented target cervical intervertebral disc.

[0035] (5) Repeat steps (3) and (4) for all batches of horizontal plane segmentation sequence images to complete the segmentation of the cervical intervertebral disc in the human cervical spine CT sequence image.

[0036] Figure 3-1 This is one of the sagittal images corresponding to a true horizontal cervical spine CT scan. Figure 3-2 This is the result of segmenting the cervical intervertebral disc using the method described in this invention. Combined with... Figure 2-2 As can be seen, the segmentation results of the cervical intervertebral disc are largely consistent with people's subjective perception, and the boundary definition of the cervical intervertebral disc is also relatively accurate.

[0037] To illustrate the advantages of the method of the present invention, it is compared with the traditional dual threshold segmentation method and the maximum variance method.

[0038] Figure 4-1 The original image used in the embodiments of the present invention. Figure 4-2 4-3, 4-4, and 4-5 are diagrams showing the artificial markings of the cervical intervertebral discs, the maximum difference method, the double threshold segmentation method, and the segmentation results of the method of this invention.

[0039] Table 1 Evaluation Indicators for Segmentation Results (Unit: %)

[0040]

[0041] Experimental results show that, compared with traditional dual-threshold segmentation methods and maximum variance methods, the method of this invention can achieve higher accuracy in segmenting cervical intervertebral disc targets in cervical spine CT sequence images.

Claims

1. A method for cervical disc segmentation based on cervical CT sequence images combined with multi-view information, characterized in that... Includes the following steps: (1) Map the original horizontal plane cervical spine CT sequence image to the sagittal plane to obtain the corresponding sagittal plane cervical spine CT sequence image, and perform interactive batch threshold segmentation on the cervical vertebrae. (2) In the obtained sagittal segmentation sequence image, select one with good morphology, calculate the average position of each cervical vertebra, and use the row number of the average position of two adjacent cervical vertebrae in the image to determine the batch of horizontal segmentation sequence images associated with the intervertebral discs between them. The formula is expressed as follows: in The average position of each cervical vertebra. and These are the column number and row number of its average position, respectively. and This constitutes each batch of cervical intervertebral discs; (3) In the obtained batch of horizontal plane segmentation sequence maps, search for candidate upper surface point sets and candidate lower surface point sets; (4) For each candidate upper surface point in the obtained candidate upper surface point set, if the candidate upper surface point can find a corresponding candidate lower surface point in the corresponding candidate lower surface point set along the cervical intervertebral disc tilt angle, the upper surface point and the lower surface point are deemed to be valid, and the candidate upper surface point is said to satisfy the decision criterion. Then, the cervical intervertebral disc is filled along the cervical intervertebral disc tilt angle, and the pixels between the "straight lines" formed by the valid upper and lower surface points are the cervical intervertebral disc. (5) Repeat steps (3) and (4) for all batches of horizontal plane segmentation sequence images to complete the segmentation of the cervical intervertebral disc in the human cervical spine CT sequence image.

2. The method for cervical disc segmentation based on cervical CT sequence images incorporating multi-view information according to claim 1, characterized in that... The definition of candidate upper and lower surface points in step (3) is as follows: For sets Each point in the matrix is ​​called a candidate upper surface point; For sets Each point in the matrix is ​​called a candidate lower surface point; in and These represent the width and height of the horizontal plane image, respectively. Then, as indicated by formula (2), This represents the first segment in the horizontal plane segmentation sequence diagram. open List The value at the row, where 1 represents a cervical vertebra and 0 represents other tissues or organs.

3. The method for cervical disc segmentation based on cervical CT sequence images combined with multi-view information according to claim 1, characterized in that... The definition of the cervical intervertebral disc tilt angle mentioned in step (4): in , , , The definition is based on formula (1).

Images