A C-arm CT geometric correction method based on nonlinear registration model

Abstract

The present invention discloses a C-arm CT geometric artifact correction method based on a nonlinear registration model, comprising the following steps: p According to the original design geometry of the CT device (the geometric parameters remain consistent at each angle), the reconstructed image I with geometric artifacts is obtained. p ; The artifact-free prior information of the scanned object is registered to I by simple rigid registration p Position; Project the registered prior information according to the original design geometry of the CT device (the geometric parameters remain consistent at each angle) to obtain the ideal projection S of the CT device i ; The original jittered projection S p With the ideal projection S of prior information i Bring in a specially designed nonlinear 2D-2D registration model to fit the projection matrix at each viewpoint; use the predicted projection matrix to act on the original projection S p Compared with the prior art, the geometric correction parameter estimation performed by the present invention is quite close to the gold standard, with lower computational complexity. After geometric correction, the CT image quality is higher, the display effect is better, and the secondary artifacts are reduced.

Description

Technical Field

[0001] The present invention belongs to the technical field of cone-beam C-arm CT geometric correction, and in particular relates to a cone-beam C-arm CT geometric correction method of a nonlinear registration model. Background Art

[0002] In medical imaging, cone-beam CT is a modern, universal 3D fluoroscopic scanning device. It uses the classic FDK reconstruction algorithm to reconstruct scan projections based on the CT device's original design geometry. However, in clinical C-arm CT systems, these systems use an open gantry and lack a slip ring structure, which exacerbates geometric perturbations and causes significant geometric artifacts in the reconstructed images.

[0003] Among geometric correction methods, the use of special steel ball phantoms is a mature and effective solution. While achieving high correction accuracy, this accuracy is limited by the phantom manufacturing process, resulting in high production costs. Furthermore, geometric changes over time require regular re-correction. Another type of correction solution, purely software-based optimization, can address the aforementioned issues, but the correction accuracy is low and the computational complexity is often high, exceeding the scope of clinical application. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned prior art and provide a C-arm CT geometric correction method based on a nonlinear registration model.

[0005] In order to achieve the above technical objectives, the technical solution adopted by the present invention is:

[0006] A C-arm CT geometric correction method using a nonlinear registration model is characterized by comprising the following steps:

[0007] Step S1: Projection S of the original jitter p According to the original design geometry of the CT device (the geometric parameters remain consistent at each angle), the reconstructed image I with geometric artifacts is obtained. p ;

[0008] Step S2: The artifact-free prior information of the scanned object is registered to I by simple rigid registration. p Location;

[0009] Step S3: Project the registered prior information according to the original design geometry of the CT device (the geometric parameters remain consistent at each angle) to obtain the ideal projection S of the CT device. i ;

[0010] Step S4: The original jittered projection S p With the ideal projection S of prior information iA specially designed nonlinear 2D-2D registration model is introduced to fit the projection matrix at each viewing angle;

[0011] Step S5: Use the predicted projection matrix to act on the original projection S p Complete the geometric correction.

[0012] To optimize the above technical solutions, specific measures taken also include:

[0013] Furthermore, a specially designed nonlinear 2D-2D registration model is used in step 4:

[0014]

[0015]

[0016] in:

[0017]

[0018]

[0019]

[0020]

[0021]

[0022] and:

[0023]

[0024] (u ′ ,v ′ ) is the actual projection position of the detector, (u, v) is the ideal projection position when the detector is not offset, (Δu, Δv) is the offset position error of the detector; SDD is the distance from the light source to the center of the detector, and SID is the distance from the light source to the center of the scanned object; in a1-a7, is the three-dimensional direction vector along the xyz direction of the three-dimensional coordinate system, and β is the corresponding detector rotation angle;

[0025] is the detector horizontal and vertical unit direction vector offset error; δx d,o ,δx s are the detector center offset vector and the light source offset vector respectively; is the magnification, and · is the vector inner product.

[0026] Furthermore, in step 4, the nonlinear 2D-2D registration model is used as the fitting framework to ideally project the prior information S i Projection S to the original jitterp Fitting is done on the original jittered projection matrix S to obtain the coefficients a1-a7 composed of geometric parameters. A1-a7 are synthesized into the projection matrix under this perspective. When the projection matrices at all positions are fitted, they are combined into the final geometric correction matrix. p Complete geometric correction.

[0027] Furthermore, step 4 is specifically as follows:

[0028] The original jittered projection S p With the ideal projection S of prior information i Introducing a specially designed nonlinear 2D-2D registration model, the ideal projection S of the prior information i Use bilinear interpolation sampling to obtain a new projection under the estimated geometric parameters to make it as close as possible to the original jittered projection S p When the registration optimization stops, the estimated coefficients a1-a7 are combined to form the projection matrix for this viewpoint. After the projection matrices for all positions are fitted, they are combined to form the final geometric correction matrix. It is important to note that different geometric parameter offsets have different effects on the geometric correction capabilities of the final reconstruction. Therefore, optimizing the coefficients with greater influence can further improve the geometric correction effect of this method.

[0029] Beneficial effects of the present invention:

[0030] Compared with existing technologies, this method analyzes the transformation relationship between the geometrically offset projection and the original design geometry of the CT device to derive a 2D-2D nonlinear transformation described by geometric parameters. This eliminates the 3D-2D computational overhead required for re-projection in previous software-based geometric correction solutions, significantly reducing computational complexity. Furthermore, the geometric correction parameter estimation performed by this method closely matches the gold standard. After geometric correction, CT images are of higher quality, better display quality, and fewer secondary artifacts. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] Figure 1 It is a schematic flow diagram of the present invention;

[0032] Figure 2 Schematic diagram of the coordinate system and direction vectors used in the present invention;

[0033] Figure 3 is a reconstruction effect diagram of geometric correction by the method of the present invention;

[0034] Figure 4 It is a line graph of the geometric correction parameters of the method of the present invention, the solid line is the gold standard, and the dotted line is the result of the method. DETAILED DESCRIPTION

[0035] The embodiments of the present invention are described in further detail below with reference to the accompanying drawings.

[0036] Example: Figure 1 As shown, the present invention is a C-arm CT geometric correction method based on a nonlinear registration model, comprising the following steps:

[0037] Step S1: Projection S of the original jitter p According to the original design geometry of the CT device (the geometric parameters remain consistent at each angle), the reconstructed image I with geometric artifacts is obtained. p ;

[0038] Reconstruction I p The purpose of

[15] is to ensure that the artifact-free prior information of the subsequent scanned object can be in the same coordinate space as the jittered projection, which facilitates the parameter extraction of the nonlinear registration model.

[0039] Step S2: The artifact-free prior information of the scanned object is registered to I by simple rigid registration. p Location;

[0040] Rigid registration is used to unify the coordinate system of the prior information and the jittered projection. The accuracy of rigid registration has a relatively small impact on the subsequent nonlinear 2D-2D registration model. In other words, the nonlinear registration model can still accurately perform registration within a certain range of precision errors, showing high robustness.

[0041] Step S3: Project the registered prior information according to the original design geometry of the CT device (the geometric parameters remain consistent at each angle) to obtain the ideal projection S of the CT device. i ;

[0042] Step S4: The original jittered projection S p With the ideal projection S of prior information i A specially designed nonlinear 2D-2D registration model is introduced to fit the projection matrix at each viewing angle;

[0043] Furthermore, a specially designed nonlinear 2D-2D registration model is used in step 4:

[0044]

[0045]

[0046] in:

[0047]

[0048]

[0049]

[0050]

[0051]

[0052] and:

[0053]

[0054] (u ′ ,v ′ ) is the actual projection position of the detector, (u, v) is the ideal projection position when the detector is not offset, (Δu, Δv) is the offset position error of the detector; SDD is the distance from the light source to the center of the detector, and SID is the distance from the light source to the center of the scanned object; in a1-a7, is the three-dimensional direction vector along the xyz direction of the three-dimensional coordinate system, and β is the corresponding detector rotation angle; is the detector horizontal and vertical unit direction vector offset error; δx d,o ,δx s are the detector center offset vector and the light source offset vector respectively; is the magnification, and · is the vector inner product.

[0055] The above nonlinear 2D-2D registration model is simplified by the following relationship between geometric parameters and direction vectors:

[0056]

[0057]

[0058]

[0059]

[0060] By reducing the small amount, a nonlinear 2D-2D registration model is obtained.

[0061] Furthermore, in step 4, the nonlinear 2D-2D registration model is used as the fitting framework to ideally project the prior information S i Projection S to the original jitter p Fitting is performed on the image to obtain coefficients a1-a7 composed of geometric parameters, which are then combined to form the projection matrix at this perspective. It should be noted that since different geometric parameter offsets have different effects on the geometric correction capabilities of the final reconstruction, optimizing the coefficients with greater influence can further improve the geometric correction effect of this method.

[0062] Step S5: Use the predicted projection matrix to act on the original projection S p Complete the geometric correction.

[0063] like Figure 3 、 4 The figure shows the processing results of cone-beam C-arm CT data using the method of the present invention. It can be clearly seen from the figure that the geometric correction performed by this method is quite close to the gold standard. Since no additional projection is required, the computational complexity is reduced. After completing the geometric correction, the CT image quality is higher, the display effect is better, and the secondary artifacts are reduced.

[0064] The above are merely preferred embodiments of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions based on the principles of the present invention are within the scope of protection of the present invention. It should be noted that for those skilled in the art, various improvements and modifications that do not depart from the principles of the present invention should be considered within the scope of protection of the present invention.

Claims

1. A C-arm CT geometric correction method based on a nonlinear registration model, characterized in that: The method comprises the following steps: Step S1: Projection S of the original jitter p According to the original design of the CT device, the geometric parameters are kept consistent at each angle, and the reconstructed image I with geometric artifacts is obtained. p ; Step S2: The artifact-free prior information of the scanned object is registered to I by simple rigid registration. p Location; Step S3: Project the registered prior information according to the original design geometry of the CT device. The geometric parameters remain consistent at each angle to obtain the ideal projection S of the prior information. i ; Step S4: The original jittered projection S p With the ideal projection S of prior information i A specially designed nonlinear 2D-2D registration model is introduced to fit the projection matrix at each viewing angle; Step S5: Use the predicted projection matrix to act on the original jittered projection S p On, complete the geometric correction; Wherein, in step S4, a specially designed nonlinear 2D-2D registration model is: in: and: (u ′ ,v ′ ) is the actual projection position of the detector, (u, v) is the ideal projection position when the detector is not offset, (Δu, Δv) is the offset position error of the detector; SDD is the distance from the light source to the center of the detector, and SID is the distance from the light source to the center of the scanned object; in a1-a7, is the three-dimensional direction vector along the xyz direction of the three-dimensional coordinate system, and β is the corresponding detector rotation angle; is the detector horizontal and vertical unit direction vector offset error; δx d,o ,δx s are the detector center offset vector and the light source offset vector respectively; is the magnification, · is the vector inner product; In step S4, a nonlinear 2D-2D registration model is used as a fitting framework to ideally project the prior information S i Projection S to the original jitter p Fitting is done on the original jittered projection matrix S to obtain the coefficients a1-a7 composed of geometric parameters. A1-a7 are synthesized into the projection matrix under this perspective. When the projection matrices at all positions are fitted, they are combined into the final geometric correction matrix. p Complete geometric correction.

Images