An adaptive step cone beam CT limited angle reconstruction algorithm based on prior image constraint

Abstract

The application provides a prior image constraint-based adaptive step cone-beam CT limited-angle reconstruction algorithm, effectively removes limited-angle artifacts in a cone-beam CT image under a small-angle scanning range, and meets the image-guided registration task in a radiotherapy plan. The application first generates simulated projections of a planning CT according to the same scanning range, then reconstructs a planning CT limited-angle image and a cone-beam CT limited-angle image by using a FDK algorithm, then obtains a transformation matrix of the planning CT by using rigid registration, applies the transformation matrix to the planning CT to obtain a prior image, and finally reconstructs the cone-beam CT image by combining a total variation iteration with a mean square error of the cone-beam CT image and the prior image as a regularization term. The application greatly shortens the cone-beam CT scanning angle of the registration task and improves the registration efficiency.

Description

Technical Field

[0001] This invention belongs to the technical field of cone-beam CT image reconstruction, and relates to an adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints. Background Technology

[0002] CT (Computed Tomography) is a non-destructive testing method that obtains information about the internal structure of an object by acquiring its projection information from different viewpoints and combining it with image reconstruction algorithms. It has wide applications in various fields such as medicine, aviation, and industrial inspection.

[0003] Cone-beam CT imaging systems are widely used in image-guided radiotherapy, playing a very positive role in tumor treatment. Typically, cone-beam CT imaging requires a slow 360° rotation around the patient, followed by reconstruction of the guiding image using the FDK algorithm. This image is then registered with the planning CT scan for patient positioning. However, in this process, the exposed cone-beam CT system must scan very slowly, wasting significant treatment time and causing additional discomfort to the patient. Therefore, this invention attempts to reduce the scanning angle of the cone-beam CT scan to reconstruct a cone-beam CT image that meets registration requirements, thereby reducing unnecessary dose exposure and saving scanning and re-guidance time.

[0004] Traditional cone-beam CT finite-angle reconstruction typically employs iterative methods combined with total variation regularization terms to reconstruct satisfactory images. However, previous methods still struggle to reconstruct images meeting requirements within a scan range of less than 90°. This invention, tailored to specific application needs, achieves a method that can reconstruct images meeting registration requirements within a scan range of less than 90°. Summary of the Invention

[0005] The purpose of this invention is to provide an adaptive step-length cone-beam CT finite angle reconstruction algorithm based on prior image constraints, which meets the image registration requirements of image-guided radiotherapy tasks.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows: an adaptive step-length cone-beam CT finite angle reconstruction algorithm based on prior image constraints, the algorithm comprising the following steps:

[0007] Step 1: Generate a simulated projection CT scan using the same scanning method as cone-beam CT. proj First, by referring to the pre-measured HU values ​​and attenuation coefficient table, the grayscale values ​​of the planned CT scan were converted into attenuation coefficients to maintain consistency with the grayscale range of the CBCT reconstructed image. Second, to accelerate the simulation projection speed, the Siddon algorithm was used for forward projection.

[0008] Step 2: Project the simulated CT scan onto the planned CT scan. proj Reconstruction yields finite-angle CT images. limit , for CT proj Using the FDK algorithm, finite-angle CT images are obtained. limit ,

[0009] Step 3: Reconstruct the finite-angle CBCT image from the actual projection of cone-beam CT. limit The actual CBCT projection needs to be corrected for cupping artifacts, and truncation artifact correction is also required during reconstruction. Then, the CBCT is obtained using the same method as in step 2. limit The specific method for correcting cup-shaped artifacts involves measuring the ray path *l* and the corresponding projection integral *p* on the water phantom projection, and then fitting the two using a fourth-order function.

[0010] l=ap 4 +bp 3 +cp 2 +dp+e

[0011] The fitted function is then used for the actual projection to obtain the corrected projection. The specific method for truncation artifact correction involves extending the actual projection at the edges during FDK reconstruction.

[0012] Step 4: Rigidly register the finite-angle CT image and the finite-angle CBCT image to obtain the transformation matrix for the planned CT. Because the timing of the planned CT and the CBCT taken during radiotherapy are different, and the patient's posture is different, there is a certain degree of positioning error between the two. Therefore, the CBCT... limit As a reference image, CT limit As a floating image, minimize CT through rigid registration. limit and CBCT limit The mean square error within the field of view is used to obtain the transformation matrix T, i.e.:

[0013] min(CBCT limit -T·CT limit ) 2

[0014] Step 5: Perform a rigid transformation on the planned CT scan to obtain the prior CT image;

[0015] Step 6: Using prior CT, iteratively reconstruct the actual projection of cone-beam CT to obtain the cone-beam CT image (CBCT). rec This invention reconstructs cone-beam computed tomography (CBCT) images. rec Projection and actual projection CBCT proj The mean square error is used as the data fidelity term to measure the CBCT. rec With CT pTotal variation, CBCT rec Total variation and CBCT rec With CT p The mean squared error is used as the data regularization term to construct the objective function F, i.e.:

[0016]

[0017] Where A is the projection matrix, α controls the ratio of the two total variational matrixes, and k is the CBCT matrix. rec With CT p The mean squared error regularization coefficient is given by $\frac{ ...

[0018]

[0019] According to the convergence analysis based on the Lipschitz descent lemma, the Lipschitz constant L of the objective function F is:

[0020]

[0021] Where μ is the estimation factor. Using the continuity definition of the Lipschitz constant, the Lipschitz constant for the k-th round...

[0022] This invention calculates the Lipschitz constant of the objective function F during the iterative process to control the step size of each iteration, employs gradient descent to minimize the function F, and combines this with convex set projection to obtain the final cone-beam CT image. This invention is applicable to tasks with prior images that have similar or identical modalities. The effectiveness of this invention can be evaluated from multiple perspectives using methods such as peak signal-to-noise ratio and structural similarity.

[0023] Compared with existing technologies, the beneficial effects of this invention are as follows: This technical solution effectively utilizes the planned CT images in radiotherapy scenarios while avoiding the influence of planned CT images on cone-beam CT images, thus ensuring the accuracy of cone-beam CT images; in addition, the scanning range required for imaging in this invention is much smaller than that of existing technologies, further improving its practicality; ultimately, this invention meets the need for image registration requirements for small-angle cone-beam CT reconstruction in image-guided radiotherapy scenarios, providing an effective method for improving the quality of small-angle cone-beam CT reconstruction and reducing imaging conditions. Attached Figure Description

[0024] Figure 1 This is a schematic diagram of the process of the present invention;

[0025] Figure 2 This is a schematic diagram of the iterative process;

[0026] Figure 3 This is a schematic diagram of the imaging scanning method. Detailed Implementation

[0027] The present invention will be further illustrated below with reference to the accompanying drawings and specific embodiments. It should be understood that these embodiments are for illustrative purposes only and are not intended to limit the scope of the invention. After reading this invention, any modifications of the invention in various equivalent forms by those skilled in the art will fall within the scope defined by the appended claims.

[0028] Example 1: See Figure 1 An adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints.

[0029] The specific steps are as follows:

[0030] Step 1: Generate a simulated projection CT scan using the same scanning method as cone-beam CT. proj First, by referring to the pre-measured HU values ​​and attenuation coefficient table, the grayscale values ​​of the planned CT scan were converted into attenuation coefficients to maintain consistency with the grayscale range of the CBCT reconstructed image. Second, to accelerate the simulated projection speed, the Siddon algorithm was used for forward projection. The planned CT voxel size was 0.5*0.5*1mm, with a resolution of 512*512*225. The projected pixel size was 0.615*0.615mm, with a resolution of 704*704. The scanning range was -40° to +40°. The flat panel detector size was 433*433mm. The distance from the isocenter to the emission source was 784mm, and the distance from the isocenter to the flat panel was 553mm.

[0031] Step 2: Project the simulated CT scan onto the planned CT scan. proj Reconstruction yields finite-angle CT images. limit For CT scans proj Using the FDK algorithm, finite-angle CT images are obtained. limit

[0032] Step 3: Reconstruct the finite-angle CBCT image from the actual projection of cone-beam CT. limit The actual CBCT projection needs to be corrected for cupping artifacts, and truncation artifact correction is also required during reconstruction. Then, the CBCT is obtained using the same method as in step 2. limit The specific method for truncation artifact correction is to extend the actual projection by repeating edge values ​​during FDK reconstruction.

[0033] Step 4: Perform rigid registration on the finite-angle CT image and the finite-angle CBCT image to obtain the transformation matrix of the planned CT.

[0034] Step 5: Perform a rigid transformation on the planned CT to obtain the prior CT image.

[0035] Step 6: Using prior CT, iteratively reconstruct the actual projection of cone-beam CT to obtain the cone-beam CT image (CBCT). rec .

[0036] This invention reconstructs cone-beam computed tomography (CBCT) images. rec Projection and actual projection CBCT proj The mean square error is used as the data fidelity term to measure the CBCT. rec With CT p Total variation, CBCT rec Total variation and CBCT rec With CT p The mean squared error is used as the data regularization term to construct the objective function F, i.e.:

[0037]

[0038] Where A is the projection matrix, α controls the ratio of the two total variations and is set to 0.8, and k is the CBCT. rec With CT p The mean squared error regularization coefficient is set to 0.1, and iter is the iteration number, set to 100. The total variation (TV) is expressed as:

[0039]

[0040] According to the convergence analysis based on the Lipschitz descent lemma, the Lipschitz constant L of the objective function F is:

[0041]

[0042] Where μ is the estimation factor, set to 0.8.

[0043] Effectiveness evaluation:

[0044] Table 1

[0045] method PSNR (dB) SSIM RMSE(HU) FDK (Existing Method 1) 24.71 0.837 242.25 SIRT (Existing Method 2) 26.65 0.873 191.51 ASDPOCS (Existing Method 3) 26.64 0.862 191.70 PICCS (Existing Method 4) 28.12 0.891 161.69 This invention 29.45 0.912 139.03

[0046] This invention presents a deep learning-based algorithm for removing dynamic blur from natural images. It effectively utilizes planning CT images in radiotherapy scenarios while avoiding the influence of planning CT images on cone-beam CT images, thus ensuring the accuracy of cone-beam CT images. Furthermore, the scanning range required for imaging with this invention is significantly smaller than existing methods, further improving its practicality. Table 1 shows a comparison of the performance metrics of this invention with existing methods on abdominal images. The results demonstrate that this invention significantly improves commonly used metrics such as PSNR and SSIM.

[0047] It should be noted that the above embodiments are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Equivalent substitutions or replacements made based on the above technical solutions are all within the scope of protection of the present invention.

Claims

1. An adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints, characterized in that, The algorithm includes the following steps: Step 1: Generate a simulated projection of the planned CT scan using the same scanning method as cone-beam CT. ; Step 2: Project the simulated CT scan. Reconstruction yields finite-angle CT images ; Step 3: Reconstruct the finite-angle CBCT image from the actual projection of the cone-beam CT. ; Step 4: Perform rigid registration on the finite-angle CT image and the finite-angle CBCT image to obtain the transformation matrix for the planned CT scan; therefore, As a reference image, As a floating image, it is minimized through rigid registration. and Mean square error within the field of view; Step 5: Perform a rigid transformation on the planned CT scan to obtain prior CT images. :Will As a reference image, Perform rigid transformation to obtain prior CT images , Step 6: Using prior CT, iteratively reconstruct the actual projection of cone-beam CT to obtain cone-beam CT images. ; Cone-beam CT reconstructed images Projection and actual projection The mean square error is used as a data fidelity term. and The mean squared error is used as the data regularization term. and Total variation of error Total variational construction of the objective function , Right now: in Let be the projection matrix. Control the ratio of the two total variations for and The coefficient of the mean square error regularization term, This represents the number of iterations. In step 6, prior CT is used to iteratively reconstruct the actual projection of cone-beam CT to obtain cone-beam CT images. The total variation is specifically expressed as: 。 2. The adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints according to claim 1, characterized in that, Step 1: Generate a simulated projection of the planned CT using the same scanning method as cone-beam CT. First, convert the grayscale value of the planned CT into an attenuation coefficient by referring to the pre-measured HU value and attenuation coefficient table, so that it is consistent with the grayscale range of the CBCT reconstructed image. Second, in order to speed up the simulated projection, the Siddon algorithm is used for forward projection.

3. The adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints according to claim 1, characterized in that, Step 2: Project the simulated CT scan. Reconstruction yields finite-angle CT images :right Using the FDK algorithm, finite-angle CT images are obtained. .

4. The adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints according to claim 1, characterized in that, Step 3: Reconstruct the finite-angle CBCT image from the actual projection of the cone-beam CT. The CBCT actual projection needs to be corrected for cupping artifacts, and truncation artifact correction is also required during reconstruction. Then, the same method as in step 2 is used to obtain the desired result. .

5. The adaptive step-size cone-beam CT finite angle reconstruction algorithm based on prior image constraints according to claim 1, characterized in that, Step 6: Using prior CT, iteratively reconstruct the actual projection of cone-beam CT to obtain cone-beam CT images. By calculating the Lipschitz constant of the objective function F during the iteration process, controlling the step size of each iteration, and using the gradient descent method to minimize the function F, the final cone-beam CT image is obtained.

Images