Methods, systems, and media for noise reduction in computed tomography images

1. A method for reducing metal artifacts in a kilovoltage computed tomography (KVCT) image, comprising:

receiving a first computed tomography (CT) image and a second CT image wherein the first CT image is a KVCT image, and wherein the second CT image is a megavoltage computed tomography (MVCT) image;

generating, by a processing device, a prior image based on the second CT image by:

registering the first CT image and the second CT image to generate a registered image; and

reducing noise artifacts in the registered image to generate the prior image:

generating a noise image based on the first CT image, wherein the noise image corresponds to metal artifacts in the first CT image; and

reconstructing, by the processing device, the first CT image based at least in part on the prior image and the noise image.

2. The method of claim 1 , wherein registering the first CT image and the second CT image comprises making pixels of the first CT image match with pixels of the second CT image.

3. The method of claim 1 , wherein reducing the noise artifacts in the registered image comprises performing a normalized metal artifact reduction (NMAR) algorithm.

4. The method of claim 1 , wherein generating the noise image comprises performing image segmentation on the first CT image.

5. The method of claim 1 , wherein the noise image is a metal image.

6. The method of claim 1 , wherein reconstructing the first CT image comprises:

generating a first sinogram of the prior image, a second sinogram of the noise image, and a third sinogram of the first CT image;

generating a corrected sinogram of the first CT image based at least in part on the first sinogram, the second sinogram, and the third sinogram; and

reconstructing the first CT image based on the corrected sinogram.

7. The method of claim 6 , wherein generating the second sinogram and the third sinogram comprises performing forward projection.

8. The method of claim 6 , further comprising: constructing an optimization model based on the first sinogram, wherein the corrected sinogram of the first CT image is generated based on the optimization model, and the optimization model comprises a variational model via L 1 sparse regularization in a gradient domain.

9. The method of claim 6 , wherein reconstructing the first CT image comprises performing backward projection on the corrected sinogram.

10. A system for reducing metal artifacts in a kilovoltage computed tomography (KVCT) image, comprising:

at least one processor to:

receive a first computed tomography (CT) image and a second CT image wherein the first CT image is a KVCT image, and wherein the second CT image is a megavoltage computed tomography (MVCT) image;

generate a prior image based on the second CT image by:

registering the first CT image and the second CT image to generate a registered image; and

reducing noise artifacts in the registered image to generate the prior image;

generate a noise image based on the first CT image, wherein the noise image corresponds to metal artifacts in the first CT image; and

reconstruct the first CT image based at least in part on the prior image and the noise image.

11. The system of claim 10 , wherein, to register the first CT image and the second CT image, the at least one processor is further to make pixels of the first CT image match with pixels of the second CT image.

12. The system of claim 10 , wherein, to reduce the noise artifacts in the registered image, the at least one processor is further to perform a normalized metal artifact reduction (NMAR) algorithm.

13. The system of claim 10 , wherein, to generate the noise image, the at least one processor is further to perform image segmentation on the first CT image.

14. The system of claim 10 , wherein to reconstruct the first CT image, the at least one processor is further to:

generate a first sinogram of the prior image, a second sinogram of the noise image, and a third sinogram of the first CT image;

generate a corrected sinogram of the first CT image based at least in part on the first sinogram, the second sinogram, and the third sinogram; and

reconstruct the first CT image based on the corrected sinogram.

15. The system of claim 14 , wherein the at least one processor is further to:

construct an optimization model based on the first sonogram, wherein the optimization model comprises a variational model via L 1 sparse regularization in a gradient domain; and

generate the corrected sinogram of the first CT image based on the optimization model.

16. A non-transitory machine-readable storage medium including instructions that, when accessed by at least one processor, cause the at least one processor to:

receive a first computed tomography (CT) image and a second CT image wherein the first CT image is a KVCT image, and wherein the second CT image is a megavoltage computed tomography (MVCT) image;

generate a prior image based on the second CT image by:

registering the first CT image and the second CT image to generate a registered image; and

reducing noise artifacts in the registered image to generate the prior image;

generate a noise image based on the first CT image, wherein the noise image corresponds to metal artifacts in the first CT image; and

reconstruct the first CT image based at least in part on the prior image and the noise image.