MRI image and CT image registration method

Abstract

The invention discloses an MRI (Magnetic Resonance Imaging) image and CT (Computed Tomography) image registration method, which comprises the following steps of: S1, acquiring an MRI original image of an affected part of a patient before an operation, and performing vertebral segment segmentation on the MRI original image to obtain an MRI vertebral segment image; s2, acquiring a CT original image of an affected part of a patient before an operation, performing vertebral segment segmentation on the CT original image to obtain a CT vertebral segment image, and performing coarse registration on the CT vertebral segment image and the MRI vertebral segment image obtained in the S1; s3, according to the coarse registration in the S2, CT target slice images corresponding to all slice images in the MRI vertebral section image are intercepted from the CT vertebral section image; and S4, performing fine registration on the MRI image and the CT image based on a two-dimensional slice texture similarity description registration method by taking the coarse registration in the S2 as a reference and combining the S1 and the S3. According to the invention, skeleton and soft tissue information can be effectively fused, and real-time navigation support is provided for spine surgery.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of image processing, in particular to a method for registering MRI image and CT image. BACKGROUND

[0002] In orthopedic surgery, CT image can clearly reconstruct the bone structure, but it is not clear for imaging of key soft tissues such as nerve root and spinal cord; MRI image can present the details of human tissues with high texture details, and plays a key role in the diagnosis of intervertebral disc herniation and spinal cord compression, etc. However, MRI image is usually obtained before surgery and cannot be updated in real time during surgery.

[0003] In order to realize accurate navigation during surgery and improve the safety and reliability of surgery, the preoperative MRI image and the preoperative CT image are registered and fused, the mapping relationship with the CT image is established, the MRI image is registered to the intraoperative scene, and the soft tissue information such as spinal cord and nerve root extracted from the MRI image is mapped to the intraoperative navigation system. SUMMARY

[0004] The present application provides a method for registering MRI image and CT image, which can realize high-precision registration of MRI image and CT image, effectively fuse bone and soft tissue information, and provide real-time navigation support for spinal surgery.

[0005] The technical scheme of the present application is as follows: S1, performing MRI sequence scanning on the patient's affected area in sagittal and coronal positions before surgery to obtain corresponding MRI original image, and performing intervertebral segment segmentation to obtain MRI intervertebral segment image; S2, obtaining CT original image of the patient's affected area before surgery, performing intervertebral segment segmentation to obtain CT intervertebral segment image, and performing coarse registration with the MRI intervertebral segment image obtained in S1; S3, according to the coarse registration of S2, cutting CT target slice image corresponding to each slice image in the MRI intervertebral segment image in the CT intervertebral segment image; S4, taking the coarse registration of S2 as a reference, combining S1 and S3, and performing fine registration on the MRI original image and the CT original image based on two-dimensional slice texture similarity description registration method.

[0006] Specifically, in S2, the coarse registration of the CT intervertebral segment image and the MRI intervertebral segment image is as follows: According to the center coordinates of the target intervertebral segment of the patient in the MRI original image and the CT original image obtained in S1, the coarse registration of the CT intervertebral segment image and the MRI intervertebral segment image is performed; The center of mass coordinate of the target vertebra in the MRI vertebra image is the center of mass coordinate of the two-dimensional profile points of all slice images in the MRI vertebra image.

[0007] Specifically, in the S3, according to the coarse registration of the S2, the slice images in the MRI vertebra image obtained by the S1 are mapped to the CT vertebra image obtained by the S2, wherein the range of the slice images in the MRI vertebra image is framed by the four corner points of the minimum circumscribed rectangle of the two-dimensional profile therein, and the CT target slice images corresponding to the slice images in the MRI vertebra image are cut in the CT vertebra image.

[0008] More specifically, on the basis of the CT target slice images obtained by the cutting, the minimum bounding box of the target vertebra therein is obtained, the center of the minimum bounding box is used as the center of expansion with a set range, and the center of expansion is used as the final CT target slice image.

[0009] Specifically, in the S4, the MRI original image and the CT original image are finely registered based on the two-dimensional slice texture similarity description registration method, and the specific steps are as follows: S41, based on the two-dimensional slice texture, the similarity of the slice images in the MRI vertebra image obtained by the S1 and the CT target slice images obtained by the S3 is calculated; S42, taking the initial transformation matrix obtained by the coarse registration of the S2 as the initial value, taking the best similarity of the slice images in the MRI vertebra image and the corresponding CT target slice images as the target, establishing a nonlinear optimization target function, and solving to obtain the optimal transformation matrix of the MRI vertebra image and the CT vertebra image, that is, the fine registration of the MRI original image and the CT original image is completed.

[0010] More specifically, in the S41, the mutual information is used to calculate the similarity of the slice images in the MRI vertebra image and the CT target slice images.

[0011] More specifically, in the S41, the step of normalizing the gray value of the slice images in the MRI vertebra image obtained by the S1 and the CT target slice images obtained by the S3 is further included, and the specific steps are as follows: The gray values of the slice images in the CT target slice image and the MRI vertebra image are normalized to a distribution with a mean value of 0 and a standard deviation of 1 by using mean value and standard deviation normalization.

[0012] More specifically, in the S41, a scaling factor is set, and the scaling factor is used to scale a certain slice image in a pair of slice images in a set scaling range, the similarity of the certain slice image with another slice image after each scaling is calculated, the maximum value of the similarity is obtained, and the maximum value is used as the final similarity of the pair of slice images.

[0013] Further, the setting scaling range takes 0.8~1.2.

[0014] Further, in the S42, the nonlinear optimization objective function is established as follows: ; Wherein, F min is the optimization objective function, f metric The best similarity of the slice image in the pair of CT target slice images and MRI vertebral segment images after the scale transformation is, is the gray value of the pixel point on the CT target slice image, is the gray value of the pixel point on the slice image in the MRI vertebral segment image after the transformation.

[0015] Specifically, the vertebral segment segmentation of the MRI original image and the CT original image is performed by a segmentation model trained based on a deep learning network model.

[0016] Beneficial effects: the transformation relationship between the MRI image and the CT image is established by the image features, the registration method based on the two-dimensional slice texture similarity description can solve the modal difference problem, realize the high-precision registration of the MRI image and the CT image, and effectively fuse the bone and soft tissue information, provide real-time navigation support for the spine surgery, and significantly reduce the risk of intraoperative nerve injury. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the present application, the drawings needed in the embodiment description will be briefly introduced as follows. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0018] Figure 1 It is a flow chart of the MRI image and CT image registration method of the present application.

[0019] Figure 2 It is an example diagram of the scanning mode of the MRI image and its slice image.

[0020] Figure 3 It is an example diagram of the scanning of the coronal slice position of the MRI image and its slice image.

[0021] Figure 4 It is a schematic diagram of the vertebral segment segmentation of the CT original image and the MRI original image.

[0022] Figure 5 It is an example diagram of the slice image of the MRI vertebral segment image under coarse registration.

[0023] Figure 6 An example diagram for intercepting CT target slice images corresponding to each slice image in the MRI vertebrae image in the CT vertebrae image. DETAILED DESCRIPTION

[0024] In order to make the objects, technical solutions and advantages of the present application clearer, the present application will be further described in detail below with reference to the embodiments and the accompanying drawings.

[0025] It should be noted that, unless otherwise defined, the technical terms or scientific terms used in the embodiments of the present application should be understood as the general meanings understood by those skilled in the art to which the present application belongs.

[0026] The registration method of the MRI image and the CT image of the present application is shown in Figure 1 , and includes: S1, performing MRI sequence scanning in sagittal and coronal positions on a patient's affected part to obtain corresponding MRI original images, and performing vertebrae segmentation on the MRI original images to obtain MRI vertebrae images; The sequence scanning mode of the conventional MRI original image is sagittal scanning, as shown in FIG. 2. Only sagittal slice images are obtained by sagittal scanning. Although the sagittal slice images have sufficient image texture information of the human vertebrae in the head-to-foot direction and the front-to-back direction, they can conveniently present all the vertebrae and the soft tissue near the vertebrae, but lack image texture information of the human vertebrae in the left-to-right direction. In the present application, the scanning time and the convenience of diagnosis are comprehensively considered, and on the basis of the conventional sagittal sequence scanning, the coronal sequence scanning of the patient's affected part is added, and the coronal slice images obtained are shown in Figure 3 .

[0027] The present application can enhance the three-dimensional spatial information through multi-plane scanning to support subsequent registration based on image texture.

[0028] In the present application, the first vertebrae segmentation model pre-trained can be used to perform vertebrae segmentation on the MRI original images obtained in S1. Specifically, the first vertebrae segmentation model pre-trained can use a network model based on deep learning.

[0029] In the present embodiment, the MRI original images are segmented, and specifically: The MRI vertebral images of the vertebral parts in the sagittal and coronal directions of patients of different ages and different case characteristics are acquired before surgery, each vertebral level in the vertebrae is labeled and numbered by an authoritative doctor, the labeled MRI vertebral images are taken as training data, input into a network model based on deep learning for training, and a first vertebral level segmentation model is obtained, and the MRI original images of the target vertebrae of the current patient to be operated are input into the first vertebral level segmentation model, so that a plurality of independent single vertebral level images with numbers can be segmented and obtained, and the target vertebral level images of the current patient are extracted, that is, the MRI vertebral level images are obtained.

[0030] In the present application, the MRI image layer thickness is relatively large, and the number of single vertebral level sagittal slices is about 10, so the complete vertebral level surface cannot be reconstructed, and the images obtained by vertebral level segmentation are actually a plurality of two-dimensional contour images of single vertebral levels, as shown in Figure 4 (b) in the middle, that is, the schematic diagram of 10 sagittal slice images and 1 coronal slice image of a certain vertebral level, a total of 11 two-dimensional contour images.

[0031] S2, the CT original images of the patient's affected part are acquired before surgery, and the vertebral level segmentation is performed to obtain CT vertebral level images, which are roughly registered with the MRI vertebral level images obtained in S1; In the present application, the vertebral level segmentation of the CT original images of the patient's affected part can be performed by using a pre-trained second vertebral level segmentation model. Specifically, the pre-trained second vertebral level segmentation model can adopt a network model based on deep learning.

[0032] In the present embodiment, it is also necessary to acquire CT vertebral images of patients of different ages and different case characteristics, each vertebral level in the vertebrae is labeled and numbered by an authoritative doctor, the labeled CT vertebral images are taken as training data, input into a network model based on deep learning for training, and a second vertebral level segmentation model is obtained, and the CT images of the target vertebrae of the current patient obtained by preoperative scanning are input into the second vertebral level segmentation model, so that a plurality of independent single vertebral level images with numbers can be segmented and obtained, and the target vertebral level images of the current patient are extracted, that is, the CT vertebral level images are obtained. In the present application, the layer thickness during CT scanning is generally sub-millimeter, so the scanned vertebral level surface information is rich, and the complete bone surface contour of the vertebral level can be obtained, as shown in Figure 4 (a) in the middle.

[0033] In the present application, since the segmented MRI vertebral level images are a plurality of two-dimensional contour images of the vertebral level, and the CT vertebral level images are three-dimensional images, they cannot be registered in a similar three-dimensional point cloud registration manner, so a two-dimensional slice texture similarity description registration method is adopted to roughly register the MRI vertebral level images obtained in S1 and the CT vertebral level images obtained in the present step, and this method belongs to the category of nonlinear iterative optimization, so the rough registration of the MRI vertebral level images and the CT vertebral level images needs to be completed first.

[0034] Specifically, the center coordinates of the target vertebra of the patient in the MRI vertebra image and the CT vertebra image are obtained according to S1, and then the CT vertebra image and the MRI vertebra image are roughly registered.

[0035] In the embodiment, the image coordinate system corresponding to the CT original image (hereinafter referred to as the CT original image coordinate system) and the image coordinate system corresponding to the CT vertebra image (hereinafter referred to as the CT vertebra image coordinate system) are both body position coordinate systems of the human body. In the embodiment, the body position coordinate system of the human body takes the RAS body position coordinate system as an example, that is, the Z axis is the head-to-foot direction of the human body, the Y axis is the front-to-back direction of the human body, and the X axis is the left-to-right direction of the human body. Then, for any point S on the target vertebra of the patient, the following relationship is obtained: ; Among them, and are the coordinates of point S in the CT vertebra image coordinate system and the MRI vertebra image coordinate system (hereinafter referred to as the MRI vertebra image coordinate system) respectively, is the transformation matrix of the CT vertebra image coordinate system to the CT original image coordinate system, is the coordinate of point S transformed from the CT vertebra image coordinate system to the CT original image coordinate system; is the transformation matrix of the MRI vertebra image coordinate system to the MRI original image coordinate system (hereinafter referred to as the MRI original image coordinate system), is the coordinate of point S transformed from the MRI vertebra image coordinate system to the MRI original image coordinate system; is the transformation matrix of the MRI vertebra image coordinate system to the CT vertebra image coordinate system.

[0036] In the embodiment, the coordinate system directions of the CT vertebra image coordinate system and the MRI vertebra image coordinate system are the same, and the origins are the center coordinates of the corresponding vertebra images, and the directions are consistent with the directions of the RAS coordinate system. Then, for the point S of the same vertebra of the patient, the coordinates of the point S in the CT vertebra image coordinate system and the MRI vertebra image coordinate system are the same, that is, and are equal, and the above relationship can be simplified as follows: ; That is: ; Through the above steps of segmenting the CT original image and extracting the CT vertebra image of the target vertebra of the patient, the complete bone surface profile of the target vertebra of the patient is obtained, and the center of gravity of all bone surface points of the target vertebra of the current patient is As the centroid of CT vertebral segment image, all slice images of the target vertebral segment of the patient have been obtained by segmenting the MRI original image in step S1, which are 11 slice images in this embodiment, i.e. the two-dimensional contour points corresponding to all slice images of the target vertebral segment of the patient are obtained, and the centroid of the two-dimensional contour points corresponding to all slice images of the target vertebral segment of the patient is taken as the centroid of the MRI vertebral segment image, i.e. According to the above transformation relationship, the initial transformation matrix of the MRI vertebral segment image and the CT vertebral segment image can be calculated , and the coarse registration of the two is completed, as follows: ; Because the body position information of the patient cannot be completely horizontal or vertical when CT and MRI scanning, and the imaging principles of CT and MRI are different, the sizes of the vertebral segments reconstructed by the two devices are also different, so the initial transformation matrix of the MRI vertebral segment image and the CT vertebral segment image calculated by the above formula is not accurate, but it is sufficient as the initial value for subsequent fine registration, which can ensure that the MRI vertebral segment image after coarse registration is near the corresponding CT vertebral segment image.

[0037] S3, according to the coarse registration of S2, the CT target slice images corresponding to each slice image in the MRI vertebral segment image are cut in the CT vertebral segment image; In this application, according to the coarse registration of S2, each slice image in the MRI vertebral segment image obtained in S1 can be mapped to the CT vertebral segment image obtained in S2, wherein the range of each slice image in the MRI vertebral segment image can be framed by the four corner points of the minimum circumscribed rectangle of the two-dimensional contour therein, and accordingly the CT target slice images corresponding to each slice image in the MRI vertebral segment image are cut in the CT vertebral segment image.

[0038] In this embodiment, as above, in the MRI vertebral segment image coordinate system, the coordinates of the four corner points of the range of each slice image constituting the target vertebral segment are respectively (i represents the i-th slice image, and the value range is 1-11), as shown in Figure 5 According to the coarse registration of S2, their coordinates in the CT vertebral segment image coordinate system are obtained, which are respectively , and further the positions of each slice image in the MRI vertebral segment image in the CT vertebral segment image coordinate system are obtained, as follows: ; Therefore, the CT target slice images corresponding to each slice image in the MRI vertebral segment image are cut in the CT vertebral segment image, which can be used for subsequent similarity calculation, as shown in the left image of Figure 6 .

[0039] In the present application, on the basis of the aforementioned CT target slice image obtained by intercepting, the minimum bounding box of the target vertebral bone surface profile is obtained, and the center of the minimum bounding box is used to expand the set range to obtain the final CT target slice image, as shown in the right part of Fig. 6.

[0040] S4, based on the coarse registration of S2, combining each slice image in the MRI vertebral image obtained by S1 and the CT target slice image obtained by S3, based on the two-dimensional slice texture similarity description registration method, the MRI original image and the CT original image are precisely registered; Specifically includes: S41, based on two-dimensional slice texture, the similarity of each slice image in the MRI vertebral image obtained by S1 and the CT target slice image obtained by S3 is calculated. In the present application, the mutual information can be used to calculate the similarity of each slice image in the MRI vertebral image and the CT target slice image.

[0041] In the present application, because the principles of CT imaging and nuclear magnetic imaging are different, the gray values of the slice images of the CT target slice image and the slice images of the MRI vertebral image are not in the same gray space, the gray value of the slice of the CT target slice image is generally CT value, and the value range is 16 bits, while the gray value of each slice image of the MRI vertebral image is usually a relative value, and the specific gray value meaning depends on the imaging sequence, so it is necessary to normalize them before similarity calculation, and the normalization adopts mean and standard deviation normalization, that is, the gray values of the slices of the CT target slice image and each slice image in the MRI vertebral image are normalized to a distribution with a mean value of 0 and a standard deviation of 1, and the specific formula is as follows: ; Wherein, I is the gray value of any point on each slice image in the MRI vertebral image / CT target slice image, I norm is the normalized gray value of the point, and μ and σ are the mean and standard deviation of the gray value of the corresponding slice image.

[0042] Because the imaging principles of CT and nuclear magnetic are different, CT generates images by measuring the attenuation degree of X-rays passing through human tissues, while MRI is an imaging technology based on nuclear magnetic resonance principle, which generates images by measuring the signal of hydrogen atomic nucleus in human tissues in a strong magnetic field. Although the generated images basically conform to the true vertebral profile, the spatial scales of the slice images of the two are not consistent, which will lead to inaccurate similarity of the CT target slice image and the MRI vertebral slice image calculated subsequently.

[0043] Therefore, in the application, when calculating the similarity, a scaling factor η is additionally added, the selected slice image is scaled in a set scaling range, the similarity of the slice image after each scaling with another slice image is calculated, and the maximum value of the similarity is extracted as the optimal similarity of the pair of slice images.

[0044] Specifically, the scaling range can be 0.8-1.2.

[0045] In the application, the selected slice image is each slice image in the MRI vertebral image, and the specific scaling manner is that the center of each slice image in the MRI vertebral image is taken as a scaling center, and center scaling is performed based on the scaling factor η.

[0046] S42, taking the initial transformation matrix obtained by the coarse registration of S2 as an initial value, taking the optimal similarity of the slice image in the MRI vertebral image and the corresponding CT target slice image as a target, establishing a nonlinear optimization target function, and solving to obtain the optimal transformation matrix of the MRI vertebral image and the CT vertebral image, that is, the transformation matrix of the MRI original image and the CT original image; In the application, the nonlinear optimization target function is established as follows: ; Wherein, F min is an optimization target function, f metric is the optimal similarity of a pair of CT target slice images and the slice image in the MRI vertebral image after scale transformation, is the gray value of a pixel point on the CT target slice image, is the gray value of a pixel point on the slice image in the MRI vertebral image after transformation.

[0047] The optimization of the above optimization target function can be optimized by using a mature nonlinear optimization algorithm, and finally the optimal transformation matrix of the MRI vertebral image and the CT vertebral image can be obtained , that is, the fine registration of the MRI original image and the CT original image is completed.

[0048] In the application, the transformation relationship of the MRI image and the CT image is established by image features, after registration, the rich nerve root, spinal cord canal and other soft tissue texture information in the MRI image can be displayed in the navigation system in real time, and the doctor can be assisted to diagnose the vertebral related diseases more accurately and quickly.

[0049] The application fully considers the modal difference between the MRI image and the CT image during registration, describes the registration method based on two-dimensional slice texture similarity, combines the iterative optimization of the scaling factor, realizes the high-precision registration of the MRI image and the CT image, can effectively fuse the bone and soft tissue information, provides real-time navigation support for the spine surgery, and significantly reduces the risk of nerve injury in the operation.

[0050] Those skilled in the art should understand that the above discussion of any of the embodiments is only exemplary and is not intended to imply that the scope (including claims) of the present application is limited to these examples; the technical features in the above embodiments or different embodiments can also be combined, the steps can be implemented in any order, and there are many other changes of different aspects of the embodiments of the present application as described above, which are not provided in details for the sake of brevity.

[0051] The embodiments of the present application are intended to cover all such alternatives, modifications, and variations as fall within the broad scope of the appended claims. Accordingly, any one of the above-mentioned alternatives, modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the embodiments of the present application shall be included in the scope of protection of the present application.

Claims

1. A method of registering MRI images with CT images, the method comprising: The application relates to a method for preoperative registration of MRI and CT images of a patient, comprising the following steps: S1, performing MRI sequence scanning on a patient's affected part in a sagittal direction and a coronal direction to obtain corresponding MRI original images, and performing vertebral segment segmentation on the MRI original images to obtain MRI vertebral segment images; S2, obtaining CT original images of the patient's affected part, performing vertebral segment segmentation on the CT original images to obtain CT vertebral segment images, and performing coarse registration on the CT vertebral segment images and the MRI vertebral segment images obtained in S1; S3, according to the coarse registration in S2, cutting CT target slice images corresponding to each slice image in the MRI vertebral segment images from the CT vertebral segment images; S4, performing fine registration on the MRI original images and the CT original images based on a two-dimensional slice texture similarity description registration method, taking the coarse registration in S2 as a reference, and combining S1 and S3.

2. The method of registering MRI and CT images of claim 1, wherein, In S2, the coarse registration of the CT vertebral segment images and the MRI vertebral segment images is performed as follows: According to the center coordinates of the target vertebral segment of the patient in the MRI original images and the CT original images obtained in S1, coarse registration is performed on the CT vertebral segment images and the MRI vertebral segment images; Wherein, the center coordinates of the target vertebral segment in the MRI vertebral segment images are the center coordinates of the two-dimensional contour points of all slice images in the MRI vertebral segment images.

3. The method of registering MRI and CT images of claim 1, wherein, In S3, according to the coarse registration in S2, each slice image in the MRI vertebral segment images obtained in S1 is mapped to the CT vertebral segment images obtained in S2, wherein the range of each slice image in the MRI vertebral segment images is framed by the four corner points of the minimum circumscribed rectangle of the two-dimensional contour therein, and the CT target slice images corresponding to each slice image in the MRI vertebral segment images are cut from the CT vertebral segment images.

4. The method of registering MRI and CT images of claim 3, wherein, On the basis of the cut CT target slice images, the minimum bounding box of the target vertebral segment in the CT target slice images is obtained, the center of the minimum bounding box is expanded by a set range, and the expanded center is taken as the final CT target slice image.

5. The method of registering MRI and CT images of claim 1, wherein, In S4, the fine registration of the MRI original images and the CT original images is performed based on a two-dimensional slice texture similarity description registration method, and the fine registration is performed as follows: S41, based on two-dimensional slice texture, the similarity of each slice image in the MRI vertebral segment images obtained in S1 and the CT target slice images obtained in S3 is calculated; S42, taking the initial transformation matrix obtained in the coarse registration in S2 as an initial value, taking the best similarity of the slice images in the MRI vertebral segment images and the corresponding CT target slice images as a target, establishing a nonlinear optimization target function, and solving the optimal transformation matrix of the MRI vertebral segment images and the CT vertebral segment images, that is, completing the fine registration of the MRI original images and the CT original images.

6. The method of registering MRI and CT images of claim 5, wherein, In S41, mutual information is used to calculate the similarity of each slice image in the MRI vertebral segment images and the CT target slice images.

7. The method of registering MRI and CT images of claim 5, wherein, In S41, the step of normalizing the gray value of each slice image in the MRI vertebral segment images obtained in S1 and the CT target slice images obtained in S3 is further included, and the step is as follows: The gray value of each slice image in the CT target slice images and the MRI vertebral segment images is normalized to a distribution with a mean value of 0 and a standard deviation of 1 by using mean value and standard deviation normalization.

8. The method of registering MRI and CT images of claim 5, wherein, In the S41, a scaling factor is set, and each slice image in a certain pair of slice images is scaled in a set scaling range, the similarity of each slice image after scaling each time with another slice image is calculated, and the maximum value of the similarity is obtained as the final similarity of the pair of slice images.

9. The method of registering MRI and CT images of claim 8, wherein, The set scaling range is 0.8-1.

2.

10. The method of registering MRI and CT images according to claim 5 or 8, wherein, In the S42, a nonlinear optimization target function is established as follows: ； where F min is the optimization objective function, f metric The best similarity of the slice images in the pair of CT target slice images and MRI vertebra slice images after the scale transformation is is the gray value of the pixel point on the CT target slice image, is the gray value of the pixel point on the slice image in the MRI vertebra slice image after the transformation.

11. The method of registering MRI and CT images of claim 1, wherein, The segmentation of the vertebral segments of the MRI original image and the CT original image is performed through a segmentation model of a network model based on deep learning.

Images