Metal artifact correction method, device and imaging apparatus using nonlinear projection decomposition
By using a physical nonlinear projection decomposition model to separate and subtract the contribution of metal projection, the problem of incomplete metal artifact correction in existing technologies is solved, achieving a highly efficient metal artifact correction effect while preserving the detailed information of the metal implant.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- CAPITAL NORMAL UNIVERSITY
- Filing Date
- 2023-09-04
- Publication Date
- 2026-08-04
AI Technical Summary
Existing methods for correcting metal artifacts fail to effectively consider the nonlinear characteristics of multicolor projection, resulting in incomplete deduction of metal projection contributions and insufficient artifact correction.
A physical nonlinear projection decomposition model is adopted. Through threshold segmentation, interpolation completion and rational fraction fitting methods, the contribution of metal projection is separated and deducted, and the correction models of Equation (1-1) and Equation (1-2) are established to realize the correction of metal artifacts.
It effectively removes metal artifacts while preserving details and structural information near metal implants, simplifies the calculation process, and does not rely on weighted combinations of other methods.
Smart Images
Figure CN117197270B_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of CT (Computed Tomography) image processing technology, and in particular to a method, apparatus and imaging device for correcting metal artifacts using nonlinear projection decomposition. Background Technology
[0002] Computed tomography (CT) technology has advantages such as being non-invasive and providing high-quality imaging, and has been widely used in industrial inspection and clinical diagnosis. However, when the scanned object contains metal implants, such as spinal fixation devices or high-density dental fillings, beam hardening and scattering can introduce metal artifacts into the reconstructed image, severely degrading image quality.
[0003] Existing methods for correcting metal artifacts can be mainly divided into four categories: interpolation methods, iterative methods, physical correction methods, and deep learning methods. Interpolation methods are simple to operate and time-efficient, but they cannot guarantee the consistency of interpolated data in the projection domain, easily introducing secondary artifacts into the reconstructed image and losing image detail information. Iterative methods remove metal artifacts by establishing appropriate optimization models, but these methods involve many iterations and high computational costs, making them unsuitable for practical applications. Physical correction methods typically establish suitable correction models based on the causes of metal artifact formation to remove them; these methods have low computational cost and good artifact removal results. Deep learning-based methods have attracted widespread attention due to the powerful expressive capabilities of networks, but these methods usually require a large amount of labeled training data, which is difficult to obtain in practical applications.
[0004] The search revealed that the PISC (physics-informed sinogram completed) method, based on physical processes, removes metal artifacts by establishing a linear correction model of the projection data. However, this model does not consider the nonlinear characteristics of multicolor projection, resulting in incomplete deduction of the contribution of metal projection and insufficient artifact correction. Summary of the Invention
[0005] The purpose of this invention is to provide a method, apparatus, and imaging device for correcting metal artifacts by nonlinear projection decomposition to overcome or at least mitigate at least one of the above-mentioned defects of the prior art.
[0006] To achieve the above objectives, this invention provides a method for correcting metal artifacts based on a physical nonlinear projection decomposition model, comprising:
[0007] Step 1: Obtain the original CT scan data p with artifacts, and obtain the original CT image f(X) with metal artifacts, where X represents the pixel coordinates of the reconstructed image f(X).
[0008] Step 2: Threshold segmentation of CT images with metal artifacts to separate the metal from the reconstructed image f(X), and orthographic projection of the metal region to obtain the metal projection region.
[0009] Step 3: Use interpolation to complete the metal projection area and output the interpolated corrected projection p. MAR ;
[0010] Step 4: First, correct the projection p based on the CT scan data p and the interpolation from Step 3. MAR The projection p of the metallic portion in the CT scan data is obtained using equation (1-1). Metal Then, the fitted metal projection data are obtained through the rational fraction fitting method. Finally, equation (1-2) is used to subtract the contribution of the metal projection from the CT scan data o. Obtain the corrected projection p corr :
[0011]
[0012]
[0013] In the formula, p represents CT scan data. It is an increasing function and satisfies
[0014] Step 5: Reconstruct the projection data after deducting the metal contribution, and fuse it with the metal region obtained from the threshold segmentation in Step 2 to obtain the CT image after metal artifact correction.
[0015] Furthermore, the fitted metal projection data are obtained through a rational fraction fitting method. The calculation formula is equation (1-3):
[0016]
[0017] In the formula, c1, c2, c3, and c4 are all fitting coefficients, and L represents the length of the ray passing through the metal region.
[0018] Furthermore, c1, c2, c3, and c4 are all based on the measured value L. j and p j_Metal We obtain the following using equation (1-4):
[0019]
[0020] In the formula, L j Let p be the length of the j-th ray passing through the metallic region. j_Metal The contribution of the metallic region on the j-th ray, j = 1, ..., N.
[0021] Furthermore, the interpolation method in step 3 specifically includes:
[0022] Step 31: Based on the CT image f(X), use a threshold segmentation method to determine the regions Ω of metal, bone, soft tissue, and air. metal Ω bone Ω soft Ω air Then, using equations (2-1) and (2-2), we can construct binary images f respectively. binary (X) and bone prior image f bone (X), finally for f binary (X) and f bone (X) are projected onto P respectively to obtain P binary and p bone :
[0023]
[0024]
[0025] Step 32, according to p binary and p bone First, normalize the CT scan data p using equation (2-3), then process p... Norm The corresponding metal region Ω metal Interpolation is performed to complete the data and obtain the repaired projection data.
[0026]
[0027] Step 33, use equation (2-4) to repair the projection data. Perform denormalization and then convert the bone projection data p bone Add back interpolation to repair projection data Obtain interpolation correction projection p MAR :
[0028]
[0029] Further, in step 2, the threshold segmentation method provided by equation (3-1) is used to obtain the metal region χ(X) of the reconstructed image f(X). The region where f(X) is greater than or equal to T is the metal region, and χ(X) takes a value of 1, while other regions take a value of 0.
[0030]
[0031] The present invention also provides a metal artifact correction device based on a physical nonlinear projection decomposition model, comprising:
[0032] The CT projection data acquisition unit is used to acquire the original CT scan data p with artifacts and obtain the original CT image f(X) with metal artifacts, where X represents the pixel coordinates of the reconstructed image f(X).
[0033] The metal region segmentation unit is used for threshold segmentation of CT images with metal artifacts, to segment the metal from the reconstructed image f(X), and to obtain the metal projection region by orthographic projection of the metal region.
[0034] The interpolation completion unit is used to complete the metal projection area using interpolation methods and outputs the interpolated corrected projection p. MAR ;
[0035] The projection decomposition and correction unit is used to first correct the projection p based on the CT scan data p and the interpolation in step 3. MAR The projection p of the metallic portion in the CT scan data is obtained using equation (1-1). Metal Then, the fitted metal projection data are obtained through the rational fraction fitting method. Finally, using equation (1-2), the contribution of the metal projection is subtracted from the CT scan data p to obtain the corrected projection p. corr :
[0036]
[0037]
[0038] In the formula, p represents CT scan data. It is an increasing function and satisfies
[0039] The CT image reconstruction unit is used to reconstruct the projection data after deducting the contribution of metal and fuse the metal regions to obtain a CT image after correcting for metal artifacts.
[0040] Furthermore, the projection decomposition correction unit includes a coefficient fitting subunit, which obtains the fitted metal projection data through a rational fraction fitting method. The calculation formula is equation (1-3):
[0041]
[0042] In the formula, c1, c2, c3, and c4 are all fitting coefficients, and L represents the length of the ray passing through the metal region.
[0043] Furthermore, c1, c2, x3, and c4 are all based on the measured value L. j and p j_Metal We obtain the following using equation (1-4):
[0044]
[0045] In the formula, L j Let p be the length of the j-th ray passing through the metallic region. j_Metal The contribution of the metallic region on the j-th ray, j = 1, ..., N.
[0046] Furthermore, the interpolation completion unit specifically includes:
[0047] The orthographic projection data acquisition subunit is used to determine the regions Ω of metal, bone, soft tissue, and air based on the CT image f(X) from the CT scan data using a threshold segmentation method. metal Ω bone Ω soft Ω air Then, using equations (2-1) and (2-2), we can construct binary images f respectively. binary (X) and bone prior image f bone (X), finally for f binary (X) and f bone (X) are obtained by orthographic projection of each. binary and p bone :
[0048]
[0049]
[0050] Projection-normalized subunit, which is used to determine p binary and p bone First, normalize the CT scan data using equation (2-3), then normalize the p... Norm The corresponding metal region Ω metal Interpolation is performed to complete the data and obtain the repaired projection data.
[0051]
[0052] The projection denormalization subunit is used to repair the projection data using equation (2-4). Perform denormalization and then convert the bone projection data p bone Add back interpolation to repair projection data Obtain interpolation correction projection p MAR :
[0053]
[0054] The present invention also provides an imaging device, comprising:
[0055] An X-ray source;
[0056] An X-ray detector;
[0057] A mechanical control system;
[0058] A data workstation;
[0059] Multiple applications;
[0060] And one or more programs, wherein one or more programs are stored in the workstation memory, and when one or more programs are executed by an application on the workstation, cause the imaging device to perform the above-mentioned CT image metal artifact physical nonlinear projection decomposition model correction method or correction device.
[0061] The present invention, by adopting the above technical solutions, has the following advantages: 1. The present invention fully considers the nonlinear projection process of the metal region, establishes a physical nonlinear projection decomposition model, deducts the contribution of metal projection in CT scan data, and integrates useful information from CT scan data into the correction process. 2. The present invention can better preserve the details and structural information near the metal implant while removing metal artifacts. 3. The physical nonlinear projection decomposition model established by the present invention does not require weighted combination with other methods, making it simpler and more efficient. Attached Figure Description
[0062] Figure 1 The image shown is the original CT image with metal artifacts obtained in step 1 of this embodiment of the invention.
[0063] Figure 2 , Figure 3 , Figure 4 These are schematic diagrams of the interpolation method, the linear physical model, and the correction results obtained by this invention. Detailed Implementation
[0064] The present invention will now be described in detail with reference to the accompanying drawings and embodiments.
[0065] The metal artifact correction method based on a physically nonlinear projection decomposition model provided in this embodiment includes:
[0066] Step 1: Obtain the raw CT scan data p with artifacts to obtain the raw CT image f(X) with metal artifacts, as shown below. Figure 1 As shown, X represents the pixel coordinates of the reconstructed image f(X). For example, methods for obtaining CT scan data using equation (4-1) include:
[0067] p(L)=-ln∫S(E)e -∫μ(x，E)dl dE (4-1)
[0068] In the formula, L is the ray path, S(E) is the normalized energy spectrum, and μ(x,E) is the linear attenuation coefficient of the scanned object at energy E. The obtained CT scan data is reconstructed using the filtered back projection algorithm (FBP) to obtain the original CT image with metal artifacts.
[0069] Step 2: Threshold segmentation of CT images with metal artifacts to separate the metal from the reconstructed image f(X), and orthographic projection of the metal region to obtain the metal projection region.
[0070] In one embodiment, step 2 uses the threshold segmentation method provided by equation (3-1) to obtain the metallic region χ(x) of the reconstructed image f(X). The region where f(X) is greater than or equal to T is the metallic region, and χ(X) takes a value of 1, while other regions take a value of 0.
[0071]
[0072] In the formula, the specific value of T can be determined based on the difference in attenuation coefficients between the metal region and the non-metal region. Any threshold that can completely divide the metal region is acceptable.
[0073] Of course, step 2 can also use existing methods such as human-computer interaction segmentation and unsupervised segmentation to obtain the metal projection area.
[0074] Step 3: Use interpolation to complete the metal projection area and output the interpolated corrected projection p. MAR .
[0075] In one embodiment, the interpolation method in step 3 specifically includes:
[0076] Step 31: Based on the reconstructed image f(X), a thresholding segmentation method is used to determine the regions Ω of metal, bone, soft tissue, and air. metal Ω bone Ω soft Ω air Then, using equations (2-1) and (2-2), we can construct binary f respectively. binary (X) and bone prior image f bone (X), finally for f binary (X) and f bone (X) are obtained by orthographic projection of each. binary and p bone :
[0077]
[0078]
[0079] Step 32, according to p binary and p boneFirst, normalize the CT scan data using equation (2-3), then normalize the p... Norm The corresponding metal region Ω metal Interpolation is performed to complete the data and obtain the repaired projection data.
[0080]
[0081] In the formula, the projection data is repaired. It uses the non-metallic region to interpolate the value of the metallic region to obtain a more consistent value for the metallic projection region.
[0082] Step 33, repair the projection data Perform denormalization and then convert the bone projection data p bone Add back interpolation to repair projection data Obtain interpolation correction projection p MAR :
[0083]
[0084] Step 4: First, correct the projection p based on the CT scan data p and the interpolation from Step 3. MAR The projection p of the metallic portion in the CT scan data is obtained using equation (1-1). Metal Then, the fitted metal projection data are obtained through the rational fraction fitting method. Finally, using equation (1-2), the contribution of the metal projection is subtracted from the CT scan data p to obtain the corrected projection p. corr :
[0085]
[0086]
[0087] In the formula, p represents CT scan data. It is an increasing function and satisfies
[0088] In one embodiment, fitted metal projection data is obtained using a rational fraction fitting method. The calculation formula is equation (1-3):
[0089]
[0090] In the formula, c1, c2, c3, and c4 are all fitting coefficients, and L represents the length of the ray passing through the metal region.
[0091] This embodiment uses a quadratic fitting method. Those skilled in the art can also use other fitting methods (e.g., 3 or 4 times) to fit and obtain c1, c2, c3, c4.
[0092] In one embodiment, c1, c2, c3, and c4 are all based on the measured value L. j and p j_Metal We obtain the following using equation (1-4):
[0093]
[0094] In the formula, L j Let p be the length of the j-th ray passing through the metallic region. j_Metal The contribution of the metallic region on the j-th ray, j = 1, ..., N.
[0095] Step 5: Reconstruct the projection data after deducting the metal contribution, and fuse it with the metal region obtained from the threshold segmentation in Step 2 to obtain the CT image after metal artifact correction.
[0096] In one embodiment, a hardening correction method can also be used to deduct the contribution of metal projection in metal projection data in order to extract all information including the projection of non-metallic parts in the metal region (trajectory) of the projection data.
[0097] This invention also provides a metal artifact correction device based on a physical nonlinear projection decomposition model, which includes a CT projection data acquisition unit, a metal region segmentation unit, an interpolation and completion unit, a projection decomposition correction unit, and a CT image reconstruction unit, wherein:
[0098] The CT projection data acquisition unit is used to acquire the original CT scan data p with artifacts and obtain the reconstructed image f(X), where X represents the pixel coordinates of the reconstructed image f(X). Preferably, the CT projection data acquisition unit uses equation (2-1) to acquire the CT scan data, and uses the filtered back projection algorithm (FBP) to reconstruct the obtained CT scan data to obtain the original CT image with metal artifacts.
[0099] The metal region segmentation unit is used for threshold segmentation of CT images with metal artifacts, separating the metal from the reconstructed image f(X), and orthographically projecting the metal region to obtain the metal projection region. Preferably, the metal region segmentation unit specifically includes a threshold segmentation subunit, which is used to obtain the metal region χ(X) of the CT image f(X) according to the segmentation threshold T and the threshold segmentation method provided by equation (3-1). The region where f(X) is greater than or equal to T is the metal region, which takes a value of 1 in the metal region χ(X) and a value of 0 in other regions, and orthographically projects the metal region χ(X) to obtain the metal projection region.
[0100] The interpolation completion unit is used to complete the metal projection area using interpolation methods, and outputs the interpolated corrected projection p. MARPreferably, the interpolation completion unit specifically includes an orthographic projection data acquisition subunit, a projection normalization subunit, and a projection denormalization subunit, wherein: the orthographic projection data acquisition subunit is used to determine the metal, bone, soft tissue, and air regions Ω based on the CT image f(X) of the CT scan data using a threshold segmentation method. metal Ω bone Ω soft Ω air Then, using equations (2-1) and (2-2), the metal projection region f is constructed respectively. binary (X) and bone prior image f bone (X), finally for f binary (X) and f bone (X) are obtained by orthographic projection of each. binary and p bone The projection normalization sub-unit is used to determine the projection based on p. binary and p bone First, normalize the CT scan data using equation (2-3), then normalize the p... Norm The corresponding metal region Ω metal Interpolation is performed to complete the data and obtain the repaired projection data. The projection denormalization subunit is used to repair projection data. Perform denormalization and then convert the bone projection data p bone Add back interpolation to repair projection data Obtain interpolation correction projection p MAR .
[0101] The projection decomposition and correction unit is used to first correct the projection p based on the CT scan data p and the interpolation in step 3. MAR The projection p of the metallic portion in the CT scan data is obtained using equation (1-1). Metal Then, the fitted metal projection data are obtained through the rational fraction fitting method. Finally, using equation (1-2), the contribution of the metal projection is subtracted from the CT scan data p to obtain the corrected projection p. corr .
[0102] The CT image reconstruction unit is used to reconstruct the projection data after deducting the contribution of metal and fuse the metal regions to obtain a CT image after correcting for metal artifacts.
[0103] Figure 2 , Figure 3 , Figure 4 These are schematic diagrams comparing the interpolation method, the linear physical model, and the correction results obtained by this invention. Figure 2 The display shows that the dark band area between the two metal parts is still present, and the bone information in the lower left corner of the second metal part is lost and has not been recovered. Figure 3 Display: The dark area between the two metal parts is relatively... Figure 2 The corresponding part showed some improvement. The bone information in the lower left corner of the second metal was recovered, but artifacts still existed around the second metal. Figure 4 Display: The dark area between the two metal parts is relatively... Figure 3 The artifacts were further reduced, and while the bone information in the lower left corner of the second metal was restored, there were basically no artifacts around the metal.
[0104] This invention also provides an imaging device, which includes:
[0105] An X-ray source;
[0106] An X-ray detector;
[0107] A mechanical control system;
[0108] A data workstation;
[0109] Multiple applications;
[0110] And one or more programs, wherein one or more programs are stored in the workstation memory, and when one or more programs are executed by an application on the workstation, cause the imaging device to perform the CT image metal artifact physical nonlinear projection decomposition model correction method or correction device as described in the above embodiments.
[0111] This invention can be applied to the correction of metal artifacts in medical and industrial X-ray CT imaging equipment, without relying on the geometry and internal structure of the scanned sample, and without requiring other auxiliary hardware.
[0112] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them. Those skilled in the art should understand that modifications can be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for correcting metal artifacts based on a physical nonlinear projection decomposition model, characterized in that, include: Step 1: Acquire raw CT scan data with artifacts. Obtain raw CT images with metal artifacts. , Represents the reconstructed image Pixel coordinates; Step 2: Threshold segmentation of CT images with metal artifacts to remove metal from the reconstructed image. The metal region is segmented and projected onto the metal region to obtain the metal projection region. Step 3: Use interpolation to complete the metal projection area and output the interpolated corrected projection. ; Step 4, first based on CT scan data and the interpolation correction projection in step 3 The projection of the metal part in the CT scan data is obtained using equation (1-1). Then, the fitted metal projection data are obtained through the rational fraction fitting method. Finally, equation (1-2) is used to extract data from CT scans. The contribution of the metal projection portion is deducted. , to obtain the corrected projection : （1-1） （1-2） In the formula, Represents CT scan data, It is an increasing function and satisfies ; Step 5: Reconstruct the projection data after deducting the metal contribution, and fuse it with the metal region obtained from the threshold segmentation in Step 2 to obtain the CT image after metal artifact correction. The interpolation methods in step 3 specifically include: Step 31, based on the reconstructed image Threshold segmentation method was used to determine the regions of metal, bone, soft tissue and air. Then, binary images are constructed using equations (2-1) and (2-2) respectively. and bone prior images Finally, for and Obtained by orthographic projection respectively and : （2-1） （2-2） Step 32, according to and First, use equation (2-3) to process the CT scan data. Perform normalization, and then... Corresponding metal area Interpolation is performed to complete the data and obtain the repaired projection data. : （2-3） Step 33, use equation (2-4) to repair the projection data. Perform denormalization and then convert the bone projection data. Add back interpolation to repair projection data To obtain interpolated corrected projection : （2-4）。 2. The metal artifact correction method based on a physical nonlinear projection decomposition model as described in claim 1, characterized in that, The fitted metal projection data were obtained using a rational fraction fitting method. The calculation formula is equation (1-3): （1-3） In the formula, All are fitting coefficients. This represents the length of the metal region through which the ray passes.
3. The metal artifact correction method based on a physical nonlinear projection decomposition model as described in claim 2, characterized in that, All are based on measured values and We obtain the following using equation (1-4): （1-4） In the formula, For the first The length of the ray passing through the metal region, For the first The contribution of the metallic region on the ray, .
4. The method for correcting metal artifacts based on a physically nonlinear projection decomposition model as described in claim 1, characterized in that, Step 2 uses the threshold segmentation method provided by equation (3-1) to obtain the reconstructed image. metal area , Greater than or equal to The region is a metallic region, in The value in the middle area is 1, and the value in other areas is 0: （3-1）。 5. A metal artifact correction device based on a physical nonlinear projection decomposition model, characterized in that, include: The CT projection data acquisition unit is used to acquire raw CT scan data with artifacts. Obtain raw CT images with metal artifacts. , Represents the reconstructed image Pixel coordinates; The metal region segmentation unit is used for thresholding segmentation of CT images with metal artifacts, removing metal from the reconstructed image. The metal region is segmented and projected onto the metal region to obtain the metal projection region. The interpolation completion unit is used to complete the metal projection area using interpolation methods and output the interpolated corrected projection. ; The projection decomposition correction unit is used to first determine the projection decomposition correction based on the CT scan data. and the interpolation correction projection in step 3 The projection of the metal part in the CT scan data is obtained using equation (1-1). Then, the fitted metal projection data are obtained through the rational fraction fitting method. Finally, equation (1-2) is used to extract data from CT scans. The corrected projection is obtained by subtracting the contribution of the metal projection portion. : （1-1） （1-2） In the formula, Represents CT scan data, It is an increasing function and satisfies ; The CT image reconstruction unit is used to reconstruct the projection data after deducting the contribution of metal, and to fuse the metal regions to obtain a CT image corrected for metal artifacts. The interpolation completion unit specifically includes: The orthographic projection data acquisition subunit is used to acquire CT images based on CT scan data. Threshold segmentation method was used to determine the regions of metal, bone, soft tissue and air. Then, binary images are constructed using equations (2-1) and (2-2) respectively. and bone prior images Finally, for and Obtained by orthographic projection respectively and : （2-1） （2-2） Projection normalization subunit, which is used according to and First, normalize the CT scan data using equation (2-3), then... Corresponding metal area Interpolation is performed to complete the data and obtain the repaired projection data. : （2-3） The projection denormalization subunit is used to repair the projection data using equation (2-4). Perform denormalization and then convert the bone projection data. Add back interpolation to repair projection data To obtain interpolated corrected projection : （2-4）。 6. The metal artifact correction device based on a physical nonlinear projection decomposition model as described in claim 5, characterized in that, The projection decomposition and correction unit includes a coefficient fitting subunit, which obtains the fitted metal projection data through a rational fraction fitting method. The calculation formula is equation (1-3): （1-3） In the formula, All are fitting coefficients. This represents the length of the metal region through which the ray passes.
7. The metal artifact correction device based on a physical nonlinear projection decomposition model as described in claim 6, characterized in that, All are based on measured values and We obtain the following using equation (1-4): （1-4） In the formula, For the first The length of the ray passing through the metal region, For the first The contribution of the metallic region on the ray, .
8. An imaging device, characterized in that, include: An X-ray source; An X-ray detector; A mechanical control system; A data workstation; Multiple applications; And one or more programs, wherein one or more programs are stored in the workstation memory, and when one or more programs are executed by an application on the workstation, cause the imaging device to perform the CT image metal artifact physical nonlinear projection decomposition model correction method or correction device as claimed in any one of claims 1-7.