Method for improving metal artifacts in spectral CT image domain

Abstract

The application provides a kind of metal artifact reduction method in spectral CT image domain, belongs to CT imaging technical field, when the target (such as patient) with metal implant is imaged, the virtual mono-energy image obtained by current equipment can still observe metal artifact, especially more obvious at low energy (low keV), the method is extracted by the non-artifact area (or low artifact area) in the virtual mono-energy image with artifact, and the base material decomposition (such as water-based and bone-based decomposition image), the relationship model of base material and artifact-free (or low artifact) image (such as high-energy virtual mono-energy image) is constructed, then the pixel (voxel) CT value corresponding to the artifact area in the aforementioned image with artifact in the aforementioned artifact-free (or low artifact) mono-energy image is brought into the aforementioned relationship model, obtain new base material decomposition, and then synthesize artifact-corrected virtual mono-energy image at any energy, the method can effectively suppress the metal artifact in the virtual mono-energy image of existing and unlimited manufacturer and model CT equipment.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of CT imaging, and particularly relates to a method for improving metal artifacts in virtual single-energy images in the spectral CT image domain. BACKGROUND

[0002] Metal artifacts are one of the main artifacts in clinical CT, which can reduce image quality, and the causes of metal artifacts include beam hardening, scattering and photon starvation. Existing metal artifact removal methods can be roughly divided into several categories. High-voltage scanning can be used in scanning mode; the classic NMAR (Normalized Metal Artifact Reduction) algorithm is used in the algorithm, the core of which is projection and interpolation, which can be optimized by iteration to obtain better images; with the wide application of deep learning in recent years, deep learning methods have also been applied to metal artifact removal; in addition, the virtual single-energy image under high energy of dual-energy CT can effectively reduce metal artifacts.

[0003] For dual-energy CT, there are several implementation methods in the world, including the dual-layer detector CT of Philips, the dual-source dual-probe CT of Siemens, the fast high-voltage switching CT of General Electric, and the 2-scan method of Canon. The energy single spectral image, i.e. the virtual single-energy image, generated by all dual-energy CTs has good metal artifact suppression effect, but in the virtual single-energy image under low energy, there are still obvious metal artifacts.

[0004] In view of this phenomenon, different manufacturers have developed corresponding metal artifact correction algorithms (MAR, Metal Artifact Reduction) for their CT products to remove metal artifacts from the virtual single-energy image again, but the core of the MAR algorithm is polynomial interpolation. If the shape of the metal is regular and the artifact is small, the correction effect of the MAR algorithm on the metal artifact will be very significant. On the contrary, for images with irregular metal shape and relatively complex artifact distribution, the MAR correction effect will be poor, and even new artifacts will be introduced, such as CT after pedicle screw placement and CBCT or CT after dental implantation. SUMMARY

[0005] To solve the defects in the prior art, the application provides a method for improving metal artifacts in the spectral CT image domain, which can effectively suppress and improve metal artifacts in virtual single-energy images of any manufacturer based on the image domain.

[0006] The application solves the above problems by using the following technical solutions:

[0007] A method for improving metal artifacts in the spectral CT image domain, the method comprising the following steps:

[0008] S1: input n virtual mono-energy images with arbitrary energy, select the mono-energy image with significant artifacts (denoted as Image_Worst);

[0009] S2: identify the artifact region and non-artifact region in the above artifact image, further obtain the artifact region mask (denoted as ImageMask1) and the non-artifact region mask (denoted as ImageMask0);

[0010] S3: select an image with no or less artifacts (denoted as Image_Best), according to the two masks obtained in S2, two new images are derived from Image_Best, one is an image composed of only the pixels or voxels of Image_Best in the region selected by ImageMask1 (denoted as Image_Best_Artifact), and the other is an image composed of only the pixels or voxels of Image_Best in the region selected by ImageMask0 (denoted as Image_Best_noArtifact);

[0011] S4: select two virtual mono-energy images with energy difference from S1 for dual-energy decomposition to obtain two or more base material maps;

[0012] S5: compare ImageMask0 to obtain an image composed of non-artifact region pixels or voxels of the two or more base material maps;

[0013] S6: based on Image_Best_noArtifact in S3 and the image composed of non-artifact region pixels or voxels obtained in S5, construct a relationship model between Image_Best_noArtifact and the two or more base material maps;

[0014] S7: substitute Image_Best_Artifact obtained in S3 into the relationship model established in S6 to obtain the component maps of the two or more base materials on the artifact region, i.e. the corrected image;

[0015] S8: synthesize the non-artifact region image of the base material obtained in S5 and the corrected artifact region image of the base material obtained in S7, through the corrected base material image, the artifact-corrected spectral image under arbitrary energy can be synthesized.

[0016] Further, in S8, the spectral image includes but is not limited to the artifact representation in virtual monochrome image, material density image, effective atomic number image, electron density image, virtual plain scan image, iodine image, and other spectral images, and can also improve the accuracy of spectral decomposition results such as spectral curve and scatter plot which are non-image class.

[0017] Further, in S1, the energy range of the virtual mono-energy image can be, but is not limited to, 40-140 keV. After using any artifact correction algorithm to correct the image artifacts, a mask of the artifact area can be obtained, or the artifact area or non-artifact area of the image can be directly identified. The method includes, but is not limited to, traditional metal artifact removal algorithms based on interpolation reconstruction, such as MAR, fsMAR, iMAR, OMAR, sMAR, etc., and is not limited to MAR algorithms based on deep learning or directly using deep learning to identify the artifact area or non-artifact area.

[0018] Further, when extracting the artifact area, a metal threshold is set for image threshold segmentation, and the image is traversed point by point. The image CT value greater than the threshold is set to 1, and the image CT value less than the threshold is set to 0, thereby separating the metal area and the non-metal area.

[0019] Further, in S2, the threshold segmentation method is used to obtain the artifact area mask.

[0020] Further, in S3, the image without artifacts or with smaller artifacts (Image_Best) is selected from the n virtual mono-energy images in S1 or other images reconstructed in other ways, including but not limited to images reconstructed in other ways from other sources (such as different image layers, artifact correction data, etc.).

[0021] Further, in S4, a relationship model between the CT value and the water / bone basis image can be constructed, which can be extended to a relationship model between the CT value and any basis material pair of the image, and a relationship model between any other different types of spectral images.

[0022] Further, the algorithm selects (but is not limited to) 6 images, and the input mono-energy images are 50 keV, 70 keV, 80 keV, 100 keV, 120 keV, and 140 keV.

[0023] Further, the method of creating a non-artifact basis material image using two non-artifact area mono-energy images and establishing a model can be extended to using other images (both metal-containing and metal-free images) in the image sequence and establishing a model, and the two have the same effect.

[0024] Further, the method is applicable to artifact improvement of two-dimensional images, and can be directly extended and applied to artifact improvement of three-dimensional images.

[0025] Further, the method can pre-construct a model relationship between the universal single-energy image and the base material component in addition to constructing a correction model for each set of input data. In this case, one correction model can be applied to all input data (including but not limited to constructing a general model for different scanning sites, with or without contrast agent, etc.).

[0026] Further, the model method in the method includes but is not limited to polynomial fitting, deep learning (neural network), pattern recognition, etc.

[0027] Further, the model in the method can be used to optimize the artifact type, including but not limited to metal artifact, bone artifact, water hardening artifact, etc.

[0028] Further, the model in the method is not limited to spectral image modeling, and the same modeling can also be performed on non-spectral images.

[0029] The beneficial effects of the present application are:

[0030] 1. The present application develops an algorithm which can realize effective suppression and improvement of metal artifacts in virtual single-energy images of any manufacturer based on image domain, and can make the suppression effect of single-energy images on metal artifacts at low energy consistent with that at high energy.

[0031] 2. The algorithm does not introduce new artifacts, and the algorithm effect is particularly significant for complex artifact and metal morphology, which makes up for the shortcomings of traditional MAR algorithm (but not limited to MAR);

[0032] 3. The case shown in the present application establishes a model between base materials and CT values. Based on the existing results, a relationship model between base materials can also be directly established, and the two are equivalent.

[0033] 4. The decomposed base materials can be n kinds, but the case shown in the present application uses two decomposed base materials in order to simply explain the problem. BRIEF DESCRIPTION OF DRAWINGS

[0034] In order to more clearly illustrate the specific embodiments of the present application, the drawings required in the description of the specific embodiments will be briefly introduced. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without creative labor.

[0035] Figure 1 The algorithm flowchart of the present application;

[0036] Figure 2 The complete artifact image extraction process example;

[0037] Figure 3 A schematic diagram for the model construction step;

[0038] Figure 4 A virtual mono-energetic image for the phantom test result;

[0039] Figure 5 A virtual mono-energetic image for the test result of the pedicle screw placement patient data;

[0040] Figure 6 A virtual mono-energetic image for the test result of the dental implant patient data. DETAILED DESCRIPTION

[0041] In order for those skilled in the art to better understand the technical solutions of the present application, the present application will be described in detail below in conjunction with the drawings and specific embodiments.

[0042] It should be noted that the example of the method for extracting artifacts in the present application is MAR, but is not limited to MAR; in addition, the case of the present application shows the establishment of a model between the base material and the CT value, and on the existing results, a relationship model between the base materials can also be directly established, and the two are equivalent.

[0043] It should be noted that the example of the present application uses two mono-energetic images of non-artifact regions to create a water-bone image without artifacts and establish a model, which can be extended to use other images in the image sequence (both images containing metal and images not containing metal) and establish a model, and the two are equivalent.

[0044] It should be noted that the relationship model between the CT value and the water / bone base image in the example of the present application can be extended to the relationship model between the CT value and any base substance pair image, and the relationship model between any other different types of spectral images.

[0045] It should be noted that the present method is applicable to artifact improvement of two-dimensional images, and can also be directly extended and applied to artifact improvement of three-dimensional images.

[0046] It should be noted that in addition to constructing a correction model for the current data for each set of input data, the present method can also pre-construct a model relationship between the universal mono-energetic image and the base substance component, in which case one correction model can be applied to all input data (including but not limited to constructing a general model for different scanning sites, with or without contrast agent, etc. Different situations).

[0047] It should be noted that the method of establishing or optimizing the model in the present method includes but is not limited to polynomial fitting, deep learning (neural network), pattern recognition, etc.

[0048] It should be noted that the model in the method can be used for optimization of artifact types, including but not limited to metal artifacts, bone artifacts, water hardening artifacts, etc.

[0049] It should be noted that the model in the method is not limited to spectral image modeling, and the same modeling can also be performed on non-spectral images.

[0050] For the convenience of understanding, the specific meanings of the special terms used in the present application are shown in the special term correspondence table:

[0051]

[0052] As shown in Figure 1 , a dual-energy CT can generally generate n virtual single-energy images at 40-140 keV, and it is assumed that the single-energy images output by the product are 50 keV, 70 keV, 80 keV, 100 keV, 120 keV, and 140 keV, a total of 6 images. The present application proposes a method for improving metal artifacts in the spectral CT image domain, including the following steps:

[0053] S1: input n virtual single-energy images at any energy, and select a single-energy image with significant artifacts (denoted as Image_Worst);

[0054] In this embodiment, as shown in Figure 3 (a), Image_Worst is the 50 keV single-energy image, which can be corrected for artifacts using any artifact correction algorithm to obtain a mask of the artifact region, or directly identify the artifact region or non-artifact region of the image. The method includes but is not limited to traditional metal artifact removal algorithms based on interpolation and reconstruction, such as MAR, fsMAR, iMAR, OMAR, sMAR, etc., and is not limited to MAR algorithms based on deep learning or directly using deep learning to identify the artifact region or non-artifact region; as shown in Figure 2The diagram illustrates one method for artifact region extraction: ① Image thresholding: Set a metal threshold, such as HU = 3000. Iterate through the image point by point; images with a CT value greater than 3000 are set to 1, and those less than 3000 are set to 0. This allows segmentation of metal and non-metal regions. ② Apply mean smoothing to the metal image. In this case, a 7×7 mean filter operator with all values ​​equal to 1 is used. A sliding window is used to multiply the smoothed image with the corresponding corresponding points in the image, resulting in a slightly enlarged smoothed metal region image. ③ Project the input image forward using a projection algorithm, simultaneously projecting the metal image forward. Based on the position of the metal projection data, determine which locations in the original image projection are metal regions. Then, perform polynomial interpolation on these regions to obtain the corrected image. ④ Reconstruct the corrected projection data to obtain a preliminary corrected image; ⑤ Smooth the image again using a 7×7 mean filter operator with all values ​​equal to 1; ⑥ Use threshold segmentation. In this embodiment, the threshold range for soft tissue is set to 800-1200. Extract the soft tissue region and calculate the average CT value of the soft tissue; ⑦ Fill the soft tissue region except for the metal region to obtain the mean image; ⑧ Perform forward projection on the mean image using a projection algorithm. The difference between the projection data of the input image and the projection data of the mean image is denoted as Pjr_diff. This difference is then subjected to polynomial interpolation, denoted as Pjr_inter; ⑨ Reconstruct the result of Pjr_inter minus Pjr_diff using a reconstruction algorithm to obtain the final artifact image.

[0055] S2: Identify the artifact regions and non-artifact regions in the above artifact image, and further obtain the artifact region mask (denoted as ImageMask1) and the non-artifact region mask (denoted as ImageMask0);

[0056] In this embodiment, the selected algorithm is the threshold segmentation method. The artifact image is segmented and calculated using a threshold, and the CT values ​​at the corresponding positions are added and reassigned to obtain the artifact region mask.

[0057] In this embodiment, a threshold of -800 is selected to extract air regions (air CT value is -1000), specifically for images with the lowest energy. Figure 3 (a)) is divided, and each point is traversed. If the value is greater than -800, it is set to 1; if the value is less than -800, it is set to 0. This will result in ( Figure 3 (d)). Using the same method, a threshold of 100 is set to extract the main artifact regions from the artifact image ( Figure 3 (c) Perform threshold segmentation to obtain Figure 3 (e), finally Figure 3 (e) and Figure 3(d) The CT values ​​at corresponding locations are added together, and points with a CT value of 2 are reset to 1. This yields the final artifact region mask, denoted as ImageMask1. Figure 3 (f)).

[0058] S3: Select an image with no artifacts or minimal artifacts (denoted as Image_Best). Based on the two masks obtained in S2, derive two new images from Image_Best: one image consisting only of pixels or voxels of Image_Best within the selection area of ​​the ImageMask1 mask (denoted as Image_Best_Artifact), and the other image consisting only of pixels or voxels of Image_Best within the selection area of ​​the ImageMask0 mask (denoted as Image_Best_noArtifact).

[0059] It should be noted that the image with no artifacts or minimal artifacts (Image_Best) can be selected from n virtual monoenergetic images of arbitrary energy, or reconstructed by other means, including but not limited to images reconstructed from other sources (e.g., different image layers, data after coarse artifact correction, etc.) or by other means.

[0060] In this embodiment, Image_Best is selected from n virtual monoenergetic images of arbitrary energy in S1; such as Figure 3 As shown in (b), an algorithm is used to automatically identify the image with the smallest artifacts among the n input monochromatic images, such as the TV (Total Variation) algorithm. In this case, the cost function of the theoretical TV function is used (as shown in formula (1), where i and j represent the pixel indices of the image, f...). i,j u represents the image pixel value at position (i,j). i,j The TV value represents the value of a pixel. By iterating through each pixel in the image, the TV value of the current point can be calculated. Finally, the TV values ​​of all points are summed to obtain the final TV value of the image. The cost function values ​​of n images are calculated sequentially. The image with the smallest cost function value is considered to have the fewest artifacts. The dot product of the corresponding positions of the mask ImageMask0 (the image with the fewest artifacts) and the mask ImageMask1 (the mask of the artifact region) yields the non-artifact region of the current image (denoted as Image_Best_noArtifact). The dot product of the image with the fewest artifacts and the mask ImageMask1 (the mask of the artifact region) yields the artifact region of the current image (denoted as Image_Best_Artifact).

[0061]

[0062] S4: Select two virtual mono-energy images with energy difference from S1 to perform dual-energy decomposition, to obtain two or more base material maps;

[0063] S5: Obtain the image composed of non-artifact region pixels or voxels of the two or more base material maps by comparing with ImageMask0;

[0064] It should be noted that the decomposed base material can be n, but the embodiment of the present application uses two decomposed base materials for the purpose of simple description, the first base material is water-based, and the second base material is bone-based. The water-based material and the bone-based material density image are multiplied by the mask of the non-artifact region to obtain the water-based material map and the bone-based material map without artifacts.

[0065] The water-based map (denoted as imgWater) and the bone-based map (denoted as imgBone) are obtained by performing dual-energy decomposition on the image (Image_Best) with the least artifacts and the image (Image_Worst) with the lowest energy identified in S3. The dual-energy decomposition process is as follows (formula (2)), wherein E1, E2 represent two energies, and Image_Best and Image_Worst correspond in turn, ρ w represents the density of water, m w (E) represents the mass attenuation coefficient value of water material at energy E, m b (E) represents the mass attenuation coefficient of bone material at energy E. The mass attenuation coefficient values of water at different energies and the mass attenuation coefficient values of bone material at different energies are known and can be queried on a public website. Thus, the water-based map and the bone-based map can be solved from the equation; finally, the water-based map and the bone-based map are multiplied by the mask ImageMask0 of the non-artifact region to obtain the water-based map and the bone-based map without artifacts (denoted as imgWater_noArtifact and imgBone_noArtifact);

[0066]

[0067] S6: Based on the Image_Best_noArtifact image in S3 and the image composed of non-artifact region pixels or voxels obtained in S5, a relationship model between the Image_Best_noArtifact image and the two or more base material maps is constructed;

[0068] Based on the CT value of the image without artifacts (Image_Best_noArtifact) and the water-based and bone-based images without artifacts (imgWater_noArtifact and imgBone_noArtifact), each pixel point of Image_Best_noArtifact, imgWater_noArtifact, and imgBone_noArtifact is sequentially traversed, the pixel values at the same position are taken out, and two groups of data pairs can be formed, which are CT value-water-based image value and CT value-bone-based image value; that is, a plurality of scatter points are formed, which correspond to Figure 4 to Figure 6 The scatter point distribution of (g) and (h) is obtained, and finally a polynomial fitting is performed on the scatter points to obtain a red model curve. Thus, the relationship model between the CT value and the water / bone-based image is constructed.

[0069] It should be noted that the algorithm is used to construct a model of the current data to be processed for a certain data, and therefore the model form is not unique.

[0070] S7: The Image_Best_Artifact image obtained in S3 is substituted into the relationship model established in S6 to obtain a component image on the artifact region of two or more base materials, that is, a corrected image;

[0071] The artifact region image obtained in S3, that is, Image_Best_Artifact, is traversed for each pixel point in the image to obtain a CT value, which is substituted into the red curve fitting formula established in step 5 to obtain a water-based image value and a bone-based image value corresponding to each CT value (denoted as imgWater_ArtifactCorrect and imgBone_ArtifactCorrect). Finally, imgWater_ArtifactCorrect+imgWater_noArtifact is denoted as imgWater_Correct, and imgBone_ArtifactCorrect+imgBone_noArtifact is denoted as imgBone_Correct.

[0072] S8: The base material images (imgWater_Correct and imgBone_Correct) obtained in S7 can be used to synthesize a spectral image corrected for artifacts at any energy;

[0073] It should be noted that the spectral image includes but is not limited to the artifact performance in the spectral images such as virtual monochrome image, material density image, effective atomic number image, electron density image, virtual plain scan image, iodine image, and other spectral images, and can also improve the accuracy of spectral decomposition results of non-image types such as spectral curve and scatter plot.

[0074] In this embodiment, based on the results imgWater_Correct and imgBone_Correct of the sixth step, the artifact-corrected virtual mono-energy images at any energy can be obtained by dual-energy decomposition (formula (3)) again.

[0075]

[0076] The above algorithm verification is verified. Among them, the red arrow represents the artifact existing in the image itself, the yellow arrow represents the artifact introduced by the correction algorithm (two kinds, product MAR algorithm and the algorithm proposed in the present application) relative to the original image, and the yellow box represents the change of image structure caused by the correction algorithm relative to the original image. The results are shown in Figure 4

[0077] Verification example 1: a self-developed CT performance phantom, the phantom has a diameter of 200 mm, and two titanium cylindrical phantoms with a diameter of 10 mm are embedded in the inside, which is scanned by a dual-energy CT of a certain international well-known manufacturer.

[0078] As shown in Figure 5 , the first row of results is the virtual mono-energy image at different energies without MAR correction provided by the manufacturer's software. It can be seen that in the mono-energy image at low energy, such as below 70 keV, there is obvious artifact between the two metals, and the artifact in the mono-energy image at high energy is better corrected; the second row of results is the result of MAR correction by the manufacturer's software, and it can be seen that the artifact of the overall image is suppressed to a certain extent, but a lot of new artifacts are introduced, and the correction effect of the 50 keV image is also not ideal; the third row is the correction result of the algorithm described in the present application based on the first row of images, and the overall correction effect is better than that of the manufacturer's MAR algorithm, and no new artifact is introduced, the artifact suppression effect of the algorithm is significant, which indicates the effectiveness of the algorithm; the fourth row is the enlarged image of the 70 keV image in the first three rows of results. It can be seen that the proposed algorithm effectively improves the artifact in the region indicated by the red arrow without introducing new artifacts.

[0079] Verification example 2: a postoperative CT of a pedicle screw placement surgery patient, which is scanned by a dual-energy CT of a certain international well-known manufacturer.

[0080] As shown in Figure 4 , similar to Figure 6 ​, the first row is the original single-energy image of the manufacturer, the second row is the single-energy image after the manufacturer MAR, the third row is the correction result based on the first row image by the current algorithm, and the fourth row is the enlarged image of the 70keV image in the results of the first three rows; the product MAR correction result obviously introduces dark band artifacts (as indicated by the yellow arrow), the metal around is unclear, and causes the change of the tissue structure (as indicated by the yellow box), while the current algorithm correction effectively improves the artifacts of the images at all energies, and the metal boundary is clearer.

[0081] Verification example 3: a postoperative CT of a patient with an oral prosthesis implant, scanned by a certain internationally renowned dual-energy CT.

[0082] As shown in ​ , similarly, the first row is the original single-energy image of the manufacturer, the second row is the single-energy image after the manufacturer MAR, the third row is the correction result based on the first row image by the current algorithm, and the fourth row is the enlarged image of the 70keV image in the results of the first three rows; the product MAR correction result, although it improves the artifacts in the original image, also introduces a lot of new artifacts, and the current correction algorithm is significantly better than the product MAR result, without introducing new artifacts, and effectively improving the artifacts in the low keV image.

[0083] The above has carried out the detailed description to the application of the present application through the embodiment, but the content described is only the preferred embodiment of the present application, and cannot be considered for limiting the implementation range of the present application. Any equivalent change and improvement made according to the scope of the present application should still belong to the patent coverage range of the present application.

Claims

1. A method for metal artifact reduction in spectral CT image domain, characterized in that: The method comprises the following steps: S1: input n virtual single-energy images of any energy, select a single-energy image with significant artifacts, denoted as Image_Worst; S2: identify the artifact region and non-artifact region in the Image_Worst, further obtain an artifact region mask and a non-artifact region mask, the artifact region mask is denoted as ImageMask1, and the non-artifact region mask is denoted as ImageMask0; S3: select an image without or with less artifacts, denoted as Image_Best, and obtain two new images from Image_Best according to the two masks obtained in S2, one is an image only containing the pixels or voxels of Image_Best in the region selected by ImageMask1, denoted as Image_Best_Artifact, and the other is an image only containing the pixels or voxels of Image_Best in the region selected by ImageMask0, denoted as Image_Best_noArtifact; S4: perform dual-energy decomposition on two virtual single-energy images with energy difference selected from S1 to obtain two or more base material images; S5: compare the ImageMask0 to obtain an image composed of the non-artifact region pixels or voxels of the two or more base material images; S6: based on the Image_Best_noArtifact image in S3 and the image composed of the non-artifact region pixels or voxels obtained in S5, a relationship model between the Image_Best_noArtifact image and the two or more base material images is constructed; S7: the Image_Best_Artifact image obtained in S3 is substituted into the relationship model established in S6 to obtain the component images of the two or more base materials on the artifact region, that is, the corrected image; S8: the non-artifact region image of the base material obtained in S5 and the corrected artifact region image of the base material obtained in S7 are synthesized to obtain a pseudo-artifact-corrected spectral image under any energy through the corrected base material image.

2. The method of claim 1, wherein the method is performed in the spectral CT image domain. In S8, the spectral image includes the pseudo-artifact representation in the virtual monochrome image, the material density image, the effective atomic number image, the electron density image, the virtual plain scan image and the iodine spectral image.

3. The method of claim 1, wherein the method is performed in the spectral CT image domain. In S1, the energy range of the virtual single-energy image is 40-140 keV, and the mask of the artifact region is obtained after the image is corrected for artifacts using any artifact correction algorithm, or the artifact region or non-artifact region of the image is directly identified.

4. The method of claim 3, wherein the method is performed in the spectral CT image domain. When the artifact region is extracted, an image threshold segmentation is set with a metal threshold, and the image is traversed point by point, the image CT value greater than the threshold is set to 1, and the image CT value less than the threshold is set to 0, thereby segmenting the metal region and the non-metal region.

5. The method of claim 1, wherein: In S2, a threshold segmentation method is used to obtain the artifact region mask.

6. The method of claim 1, wherein: In S3, the selection of the image without or with less artifacts is derived from the n virtual single-energy images of any energy in S1 or other images reconstructed in other ways.

7. The method of claim 1, wherein the method is performed in the spectral CT image domain. In S4, a relationship model between CT value and water / bone basis material map is constructed, which is generalized to a relationship model between CT value and any basis material pair, and a relationship model between different types of spectral images.

8. The method of improving metal artifacts in spectral CT image domain according to any one of claims 1-7, characterized in that: In the algorithm, 6 images are selected, and the input single-energy images are 50 keV, 70 keV, 80 keV, 100 keV, 120 keV and 140 keV.

9. The method of improving metal artifacts in spectral CT image domain according to any one of claims 1-7, characterized in that: The method of creating a non-artifact basis material image using two single-energy images in a non-artifact region and establishing a model is generalized to using other images in the image sequence and establishing a model.

10. The method of improving metal artifacts in spectral CT image domain according to any one of claims 1-7, characterized in that: The method is suitable for artifact improvement of two-dimensional images or artifact improvement of three-dimensional images.

11. The method of improving metal artifacts in spectral CT image domain according to any one of claims 1-7, characterized in that: The method constructs a correction model of the current data for each set of input data, or pre-constructs a general model relationship between single-energy images and basis material components.

12. The method of claim 1-7, wherein the method is a method of metal artifact reduction in spectral CT image domain. In the method, the model is established or optimized by a method including polynomial fitting, deep learning and pattern recognition.

13. The method of improving metal artifacts in spectral CT image domain according to any one of claims 1-7, characterized in that: In the method, the model is used for optimization of artifact types including metal artifact, bone artifact and water hardening artifact.

14. The method of claim 1-7, wherein the method is a method of metal artifact reduction in spectral CT image domain. In the method, the model is used for spectral image modeling or non-spectral image modeling.

Images