CN106875376A - The construction method and lumbar vertebrae method for registering of lumbar vertebrae registration prior model - Google Patents

Abstract

The invention relates to a construction method of a lumbar registration prior model and a lumbar registration method. The present invention obtains projection image samples by performing DRR projection on a large number of lumbar spine CT samples, and extracts feature points from the projection image samples, thereby constructing a priori model of lumbar spine registration. When performing lumbar spine registration, first obtain the patient's lumbar spine CT image before the operation, and use the high correlation between the shape model and projection parameters to establish a pose model; after obtaining the shape parameters of the lumbar spine through X-ray images during the operation , the projection parameters can be obtained directly through the attitude model, so as to complete the registration. It avoids searching for a projection image with the best matching degree in a large number of projection images in the traditional method, so that the present invention can efficiently perform 3D/2D registration of the lumbar spine, and meet the high real-time requirements in the operation while ensuring accuracy.

Description

Construction method of lumbar registration prior model and lumbar registration method

technical field

The invention relates to the field of image processing, in particular to a method for constructing a lumbar registration prior model and a lumbar registration method.

Background technique

During spine surgery, doctors need image guidance so as to be able to know the position and posture of the spine and the specific position of surgical implants (such as screws). However, because intraoperative computerized tomography (Computed Tomography, CT) equipment is expensive, only a few hospitals are equipped, so preoperative CT and intraoperative X-ray (X-ray) images are often used to assist doctors in surgery. Intraoperative X-ray can provide two-dimensional images. To obtain three-dimensional information, intraoperative X-ray and preoperative CT need to be registered. Therefore, 3D/2D image registration is a key issue in image-guided surgery.

The current 3D/2D registration generally projects 3D data onto a 2D plane, then performs 2D/2D registration, and finally finds an optimal projection parameter in the projection parameter space so that the projection image and the target image can be optimally registered. . However, due to the high space complexity of the projection parameters, this search strategy-based 3D/2D registration method is computationally intensive. Some processing methods require more manual intervention, which also leads to low efficiency.

Contents of the invention

In order to solve the above problems in the prior art, the present invention proposes a method for constructing a priori model of lumbar registration and a lumbar registration method, which greatly improves the efficiency of registration while ensuring accuracy, and achieves real-time registration during surgery. standard requirements.

The present invention proposes a method for constructing a lumbar registration prior model and a lumbar registration method, comprising the following steps:

Step A1, performing digitally reconstructed radiograph (Digitally Reconstructed Radiograph, DRR) projection on a preset number of lumbar spine CT image samples to obtain projected image samples, and extracting feature points from the projected image samples;

Step A2, establishing a statistical shape model according to the feature points extracted in step A1;

Step A3, establishing a statistical grayscale model;

In step A4, the parameters of the shape model and the parameters of the gray model are concatenated to establish a joint model.

Preferably, step A2 is specifically:

Use the Platts analysis method to map the extracted feature points to a common two-dimensional coordinate system, so that the center of gravity of each projected image sample feature point coincides with the coordinate origin, and eliminate translation, scaling, and rotation on different projected image sample feature points. Influence; use principal component analysis (PCA) to obtain the orthonormal basis P _s of the shape model for the feature points mapped to the common two-dimensional coordinate system, then the feature point shape model of each projected image sample is:

in, is the average shape, the average shape is the average position of each feature point, and b _s is the shape model parameter.

Preferably, step A3 is specifically:

Normalize the shape of each projected image sample so that its feature points are deformed to the average shape; sample the shape-standardized image in the area covered by its shape model, and normalize the sampling points to make the gray mean value is 0, and the variance is 1; the apparent model Pg is obtained by principal component analysis, then the grayscale model _g of the feature points of each projected image sample is:

in, is the average gray level, P _g is the orthonormal basis of the gray model, and b _g is the parameter of the gray model.

Preferably, step A4 is specifically:

Step A41, connect the shape model parameter b _s and the gray scale model parameter b _g in series,

Among them, W _s is a diagonal matrix, which is used to balance the dimensions of the shape model and gray model parameters;

Step A42, utilize principal component analysis method to obtain joint model to the shape model of series connection and gray scale model parameter:

Since the mean value of the shape model parameters and the gray model parameters is 0, the mean value is 0, then there are:

b=Qc,

Among them, c is the parameter of the joint model, Q is the orthogonal basis of the joint model, Q _s is the orthogonal basis of the shape model, and Q _g is the orthogonal basis of the gray model;

Step A43, use joint model parameter c to express shape x and grayscale g:

in, is the average shape, P _s is the orthonormal basis of the shape model, is the average gray level, and P _g is the orthonormal basis of the gray level model.

Preferably, when performing shape normalization on each projected image sample, a triangle deformation algorithm is used.

The present invention simultaneously proposes a lumbar registration method, comprising the following steps:

Step B1, obtaining the target lumbar CT image before the operation, performing DRR projection on the CT image to generate a preset number of DRR images;

Step B2, for each DRR image, segment the target lumbar spine and extract shape parameters according to the preset initial position and lumbar registration prior model.

Step B3, establishing an attitude model according to the projection parameters of the DRR image and its corresponding shape parameters;

Step B4, acquiring an X-ray image during the operation, on the X-ray image, according to the lumbar registration prior model, segmenting the target lumbar spine, and extracting shape parameters;

In step B5, according to the shape parameters of the target lumbar spine obtained on the X-ray image and the attitude model obtained in step B3, the corresponding projection parameters are obtained, thereby completing the registration.

Preferably, the establishment of the posture model described in step B3 is specifically:

Using the linear model R=MB, the attitude model is obtained:

M=RB ^T (BB ^T ) ⁻¹ .

Among them, R is the projection parameter, B is the shape parameter, and M is the pose model.

Preferably, step B5 is specifically:

According to the shape parameter obtained in the attitude model and step B4, calculate the corresponding projection parameters of the X-ray image:

R _X-ray = MB _X-ray ,

Where B _X-ray is the shape parameter of the X-ray image extracted in step B4, and M is the pose model.

Preferably, in step B2 and step B4, the AAM algorithm is used to segment the target lumbar spine and extract shape parameters.

The invention establishes a two-dimensional statistical model of the lumbar spine, and uses the high correlation between the shape model and projection parameters to learn a posture model. After the shape parameters of the lumbar spine are obtained during the operation, the projection parameters can be obtained directly through the attitude model, thereby avoiding searching for a projection image with the best matching degree in a large number of projection images in the traditional method, so that the present invention can be carried out efficiently The 3D/2D registration of the lumbar spine meets the high real-time requirements of the operation while ensuring the accuracy.

Description of drawings

FIG. 1 is a schematic flow chart of a method for constructing a priori model for lumbar registration in the present embodiment;

Fig. 2 is a schematic diagram of extracting feature points from DRR projection images in the present embodiment;

Fig. 3 is a schematic flowchart of the lumbar registration method in this embodiment.

detailed description

Preferred embodiments of the present invention are described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the present invention, and are not intended to limit the protection scope of the present invention.

In this embodiment, on a PC configured as Intel(R) Core(TM) i5-2400CPU@3.10GHz, 4G memory, the algorithm platform Matlab R2014b is run.

The present invention proposes a method for constructing a lumbar registration prior model and a lumbar registration method, as shown in Figure 1, comprising the following steps:

Step A1, performing DRR projection on a preset number of lumbar spine CT image samples to obtain projected image samples, and extracting feature points from the projected image samples;

Step A2, establishing a statistical shape model according to the feature points extracted in step A1;

Step A3, establishing a statistical grayscale model;

In step A4, the parameters of the shape model and the parameters of the gray model are concatenated to establish a joint model.

In this embodiment, when establishing the statistical shape model, the model on the two-dimensional image is selected, because the feature points of the two-dimensional image are easier to mark and acquire. As shown in Figure 2, 93 feature points are extracted from each DRR image. The gray-scale depression (as shown in sub-figure c), the pedicle close to the physiological structure (as shown in sub-figures d and e), and the left and right undercut angles of the vertebral body (as shown in sub-figures f and g).

In this embodiment, step A2 is specifically:

Use the Platts analysis method to map the extracted feature points to a common two-dimensional coordinate system, so that the center of gravity of each projected image sample feature point coincides with the coordinate origin, and eliminate translation, scaling, and rotation on different projected image sample feature points. Influence; the orthonormal basis P _s of the shape model is obtained by using the principal component analysis method for the feature points mapped to the common two-dimensional coordinate system, then the feature point shape model of each projected image sample is shown in the formula (1):

in, is the average shape, the average shape is the average position of each feature point, and b _s is the shape model parameter.

In this embodiment, step A3 is specifically:

Normalize the shape of each projected image sample so that its feature points are deformed to the average shape; sample the shape-standardized image in the area covered by its shape model, and normalize the sampling points to make the gray mean value is 0, and the variance is 1; using the principal component analysis method to obtain the apparent model P _g , then the feature point gray model g of each projected image sample, as shown in formula (2):

in, is the average gray level, P _g is the orthonormal basis of the gray model, and b _g is the parameter of the gray model.

In this embodiment, step A4 is specifically:

Step A41, connect the shape model parameter b _s and the gray scale model parameter b _g in series, as shown in formula (3):

Among them, W _s is a diagonal matrix, which is used to balance the dimensions of the shape model and gray model parameters;

Step A42, using the principal component analysis method to obtain the joint model for the parameters of the concatenated shape model and gray model, as shown in formula (4):

Since the mean value of the shape model parameters and the gray model parameters is 0, the mean value is 0, as shown in formula (5):

b=Qc (5)

Among them, Q is the orthogonal basis of the joint model, and the Q value is shown in formula (6):

Q _s is the orthogonal basis of the shape model, Q _g is the orthogonal basis of the gray model, and c is the parameter of the joint model;

Step A43, use the joint model parameter c to express the shape x and the gray level g, as shown in formulas (7) and (8) respectively:

in, is the average shape, P _s is the orthonormal basis of the shape model, is the average gray level, and P _g is the orthonormal basis of the gray level model.

In this embodiment, when performing shape standardization on each projected image sample, a triangle deformation algorithm is used.

The invention also proposes a lumbar spine registration method. The method acquires the lumbar spine CT image of the patient to be operated before the operation, processes the CT image, and provides information for intraoperative segmentation and registration. In this way, more calculation work is performed before the operation, so that segmentation and registration can be performed in a more efficient manner during the operation, so as to meet the real-time requirements.

As shown in Figure 3, the following steps are included:

Step B1, obtaining the target lumbar CT image before the operation, performing DRR projection on the CT image to generate a preset number of DRR images;

Step B2, for each DRR image, segment the target lumbar spine and extract shape parameters according to the preset initial position and lumbar registration prior model.

Step B3, establishing an attitude model according to the projection parameters of the DRR image and its corresponding shape parameters;

Step B4, acquiring an X-ray image during the operation, on the X-ray image, according to the lumbar registration prior model, segmenting the target lumbar spine, and extracting shape parameters;

In step B5, according to the shape parameters of the target lumbar spine obtained on the X-ray image and the attitude model obtained in step B3, the corresponding projection parameters are obtained, thereby completing the registration.

In the present embodiment, the establishment of the posture model described in step B3 is specifically:

Use the linear model R=MB to obtain the attitude model, as shown in formula (9):

M = RB ^T (BB ^T ) ^-1 (9)

Among them, R is the projection parameter, B is the shape parameter, and M is the pose model.

In this embodiment, step B5 is specifically:

According to the shape parameter obtained in the attitude model and step B4, calculate the projection parameter corresponding to the X-ray image, and this projection parameter is the required registration result; the calculation method is as shown in formula (10):

R _X-ray = MB _X-ray (10)

Where B _X-ray is the shape parameter of the X-ray image extracted in step B4, and M is the attitude model.

In this embodiment, in step B2 and step B4, the AAM (Active Appearance Model) algorithm is used to segment the target lumbar spine and extract shape parameters.

Those skilled in the art should be able to realize that the method steps described in conjunction with the embodiments disclosed herein can be implemented by electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the possibility of electronic hardware and software For interchangeability, in the above description, the composition and steps of each example have been generally described according to their functions. Whether these functions are performed by electronic hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art may implement the described functionality using different methods for each particular application, but such implementation should not be considered as exceeding the scope of the present invention.

So far, the technical solutions of the present invention have been described in conjunction with the preferred embodiments shown in the accompanying drawings, but those skilled in the art will easily understand that the protection scope of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to relevant technical features, and the technical solutions after these changes or substitutions will all fall within the protection scope of the present invention.

Claims (9)

1. A method for building a lumbar registration prior model and a lumbar registration method, characterized in that, comprising the following steps: Step A1, performing DRR projection on a preset number of lumbar spine CT image samples to obtain projected image samples, and extracting feature points from the projected image samples; Step A2, establishing a statistical shape model according to the feature points extracted in step A1; Step A3, establishing a statistical grayscale model; In step A4, the parameters of the shape model and the parameters of the gray model are concatenated to establish a joint model. 2. The method according to claim 1, wherein step A2 is specifically: Use the Platts analysis method to map the extracted feature points to a common two-dimensional coordinate system, so that the center of gravity of each projected image sample feature point coincides with the coordinate origin, and eliminate translation, scaling, and rotation on different projected image sample feature points. Influence; the orthonormal basis P _s of the shape model is obtained by principal component analysis for the feature points mapped to the common two-dimensional coordinate system, then the feature point shape model of each projected image sample is: $\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; ,</span>$ in, is the average shape, the average shape is the average position of each feature point, and b _s is the shape model parameter. 3. The method according to claim 2, wherein step A3 is specifically: Normalize the shape of each projected image sample so that its feature points are deformed to the average shape; sample the shape-standardized image in the area covered by its shape model, and normalize the sampling points to make the gray mean value is 0, and the variance is 1; the apparent model Pg is obtained by principal component analysis, then the grayscale model _g of the feature points of each projected image sample is: $\stackrel{<span\; class="notranslate">\; ^</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; ,</span>$ in, is the average gray level, P _g is the orthonormal basis of the gray model, and b _g is the parameter of the gray model. 4. The method according to claim 3, wherein step A4 is specifically: Step A41, connect the shape model parameter b _s and the gray scale model parameter b _g in series, $\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}\\ {\mathrm{<span\; class="notranslate">\; b</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}\end{array}\right)<span\; class="notranslate">\; =</span>\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; )</span>\\ {\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}^{\mathrm{<span\; class="notranslate">\; T</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; -</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; )</span>\end{array}\right)<span\; class="notranslate">\; ,</span>$ Among them, W _s is a diagonal matrix, which is used to balance the dimensions of the shape model and gray model parameters; Step A42, utilize principal component analysis method to obtain joint model to the shape model of series connection and gray scale model parameter: $\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; b</span>}}<span\; class="notranslate">\; +</span>\mathrm{<span\; class="notranslate">\; Q</span>}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>$ Since the mean value of the shape model parameters and the gray model parameters is 0, the mean value is 0, then there are: b=Qc, $\mathrm{<span\; class="notranslate">\; Q</span>}<span\; class="notranslate">\; =</span>\left(\begin{array}{c}{\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}\\ {\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}\end{array}\right)<span\; class="notranslate">\; ,</span>$ Among them, c is the parameter of the joint model, Q is the orthogonal basis of the joint model, Q _s is the orthogonal basis of the shape model, and Q _g is the orthogonal basis of the gray model; Step A43, use joint model parameter c to express shape x and gray level g: $\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; x</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; W</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}{\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>$ $\mathrm{<span\; class="notranslate">\; g</span>}<span\; class="notranslate">\; =</span>\stackrel{<span\; class="notranslate">\; \&OverBar;</span>}{\mathrm{<span\; class="notranslate">\; g</span>}}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; P</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}{\mathrm{<span\; class="notranslate">\; Q</span>}}_{\mathrm{<span\; class="notranslate">\; g</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; ,</span>$ in, is the average shape, P _s is the orthonormal basis of the shape model, is the average gray level, and P _g is the orthonormal basis of the gray level model. 5. The method according to claim 3, wherein a triangle deformation algorithm is used when performing shape normalization on each projected image sample. 6. A lumbar registration method, characterized in that, comprising the following steps: Step B1, obtaining the target lumbar CT image before the operation, performing DRR projection on the CT image to generate a preset number of DRR images; Step B2, for each DRR image, segment the target lumbar spine and extract shape parameters according to the preset initial position and the lumbar registration prior model constructed by any one of the methods in claims 1 to 5; Step B3, establishing an attitude model according to the projection parameters of the DRR image and its corresponding shape parameters; Step B4, acquiring an X-ray image during the operation, on the X-ray image, based on the lumbar registration prior model described in step B2, segmenting the target lumbar spine and extracting shape parameters; In step B5, according to the shape parameters of the target lumbar spine obtained on the X-ray image and the attitude model obtained in step B3, the corresponding projection parameters are obtained, thereby completing the registration. 7. method according to claim 6, is characterized in that, the described attitude model of step B3 is set up, specifically: Using the linear model R=MB, the attitude model is obtained: M=RB ^T (BB ^T ) ^-1 , Among them, R is the projection parameter, B is the shape parameter, and M is the pose model. 8. The method according to claim 7, wherein step B5 is specifically: According to the shape parameter obtained in the attitude model and step B4, calculate the corresponding projection parameters of the X-ray image: R _X-ray = MB _X-ray ; Wherein, B _X-ray is the shape parameter of the X-ray image extracted in step B4, and M is the pose model. 9. The method according to claim 6, characterized in that, in step B2 and step B4, the target lumbar vertebra is segmented and the shape parameters are extracted using the AAM algorithm.