Method for metal artifact correction of three-dimensional images

Abstract

The application discloses a kind of metal artifact correction methods of three-dimensional image, comprising: the original projection sequence image of scanning with metal implant is obtained and is carried out back projection reconstruction, and the metal region image and non-metal region image in it are obtained by segmentation to reconstructed image;The metal projection image is obtained by forward projection to the metal region image, and the corresponding metal region in original projection sequence image is linearly interpolated with metal projection image, and is carried out back projection reconstruction;The projection priori graph without metal implant is obtained by forward projection to the back projection reconstruction image of preceding interpolation, the pixel value of each point in the region where metal is located is calculated, and the original projection sequence image is interpolated accordingly, and the final image is fused after back projection reconstruction with metal region image.The application improves the definition of tissue boundary, improves the accuracy and robustness of metal artifact correction, avoids the problem of image quality reduction due to interpolation error.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of image processing, and particularly relates to a metal artifact correction method for three-dimensional images. BACKGROUND

[0002] C-arm machines are increasingly used by major hospitals due to their low radiation dose, high spatial resolution and other advantages. At the same time, they are increasingly used in the field of orthopedic surgery robots to provide three-dimensional image guidance for doctors or surgery robots.

[0003] In use, it is necessary to ensure that the C-arm provides clear and accurate three-dimensional images. However, in practice, various image artifacts may be introduced due to various reasons, resulting in poor image quality and thus misjudgment. Among them, metal artifact is a common image artifact. Since metal implants (such as metal stents, bone plates, long nails, etc.) are needed in surgery, it is inevitable to have interference of metal implants in CBCT clinical application. The most commonly used correction method for metal correction is based on linear interpolation algorithm, but when the volume of metal implants is large, the corrected value obtained by the interpolation algorithm will deviate seriously from the accurate value. SUMMARY

[0004] The present application provides a metal artifact correction method for three-dimensional images based on prior images combined with metal direction images, which significantly improves the accuracy and robustness of metal artifact correction and avoids the problem of reduced image quality caused by interpolation errors.

[0005] TECHNICAL SOLUTION

[0006] The metal artifact correction method for three-dimensional images comprises:

[0007] scanning to obtain an original projection sequence image with metal implants and performing back projection reconstruction thereon, and segmenting the reconstructed image to obtain a metal region image and a non-metal region image;

[0008] forward projecting the metal region image to obtain a metal projection image, linearly interpolating the corresponding metal region in the original projection sequence image with the metal projection image, and performing back projection reconstruction thereon;

[0009] forward projecting the back projection reconstructed image after interpolation to obtain a projection prior image without metal implants, calculating the pixel values of each point in the metal region, and interpolating the original projection sequence image according to the pixel values, and fusing the metal region image after back projection reconstruction to obtain a final image.

[0010] Calculate the pixel values of each point in the metal region:

[0011] The pixel value of the interpolation point is calculated according to the boundary points perpendicular to the extension direction of the metal implant corresponding to the interpolation point in the projection prior map.

[0012] The extension direction of the metal implant is determined by direction search of the points in the metal projection image, and the extension direction of the metal region is determined according to the boundary points.

[0013] The direction search is four-direction search or eight-direction search.

[0014] The pixel value of each point in the metal region in the projection prior map is calculated and compared with the pixel values of the corresponding two boundary points.

[0015] If the difference values of the pixel values of the two boundary points and the interpolation point are within the limited range, the pixel value of the interpolation point is calculated by linear interpolation of the pixel values of the two boundary points.

[0016] If the difference value of the pixel value of one of the boundary points and the interpolation point is within the limited range, the pixel value of the interpolation point is interpolated by selecting the pixel value of the boundary point.

[0017] If the difference values of the pixel values of the two boundary points and the interpolation point are not within the limited range, the coordinates of the current two boundary points are respectively expanded by one pixel coordinate to the two sides, and the difference values of the pixel values of the current boundary points and the interpolation point are compared again to see if they are within the limited range. If yes, the interpolation is performed according to the aforementioned interpolation method, otherwise, the pixel value of the boundary point closest to the pixel value of the interpolation point in the projection prior map is selected for interpolation.

[0018] The linear interpolation is specifically:

[0019]

[0020] Where (r, c) is the coordinate of the interpolation point, p(r, c) is the image pixel value of the interpolation point; (r1, c1) and (r2, c2) are respectively the coordinates of the two selected interpolation boundary points, p1(r1, c1) and p2(r2, c2) are respectively the image pixel values of the two selected interpolation boundary points; l1 and l2 are respectively the distances from the two selected interpolation boundary points to the interpolation point,

[0021]

[0022] The projection prior map without metal implant is obtained by forward projection of the back projection reconstruction image after interpolation, including the following steps:

[0023] The interpolated back projection reconstruction image is threshold segmented, the corresponding attenuation coefficient is determined according to the threshold of different tissues, and the projection prior map without metal implant is obtained by forward projection.

[0024] The threshold segmentation is specifically:

[0025] When the pixel value of a point in the reconstructed image is less than the CT value of air, which is set as a threshold t 空气 , the attenuation coefficient value of the point is set as u 空气 .

[0026] When the pixel value of a point in the reconstructed image is less than the CT value of soft tissue, which is set as a threshold t 软组织 , the attenuation coefficient value of the point is set as u 软组织 .

[0027] The attenuation coefficient values of other points are uniformly set as u 骨头 .

[0028] Beneficial effects: The present application effectively improves the clarity of tissue boundaries, significantly improves the accuracy and robustness of metal artifact correction, avoids the problem of image quality reduction caused by interpolation error, and improves the quality of three-dimensional images. BRIEF DESCRIPTION OF DRAWINGS

[0029] Figure 1 is a flowchart of the present application;

[0030] Figure 2 is an image schematic diagram of the presence of a metal implant;

[0031] Figure 3 is Figure 2 an image schematic diagram in which the area where the metal implant is located is replaced with air;

[0032] Figure 4 is an example diagram of a projection image when a metal implant is present;

[0033] Figure 5 is an example diagram of a priori image generated by forward projection after preliminary metal artifact correction;

[0034] Figure 6 is a schematic diagram of searching in different directions for a metal region projection image;

[0035] Figure 7 is a linear interpolation schematic diagram; wherein, figures (a), (b) are four-direction search, eight-direction search, respectively;

[0036] Figure 8 is an example diagram of a slice image obtained by reconstruction without correction processing;

[0037] Figure 9 is an example diagram of a back-projection reconstructed image obtained by processing. DETAILED DESCRIPTION

[0038] The present application will be further illustrated in conjunction with the accompanying drawings and specific examples.

[0039] Figure 1 As shown in the flow chart of the present application, Figure 1 the metal artifact correction method of three-dimensional image of the present application comprises the following steps:

[0040] (1) placing the scanned object with metal implant in the center of the field of view of the C-arm machine to perform three-dimensional scanning, obtaining the original projection sequence image, as shown in Figure 2 wherein Figure 4 the specific example is shown in the example diagram;

[0041] (2) performing back-projection reconstruction on the projection sequence image obtained in step (1) to obtain the reconstructed image without metal correction;

[0042] (3) performing threshold segmentation on the reconstructed image obtained in step (2) to obtain the corresponding metal region image and non-metal region image;

[0043] (4) performing forward projection on the metal region image obtained in step (3) through the corresponding projection model to obtain the corresponding projection sequence image;

[0044] (5) performing linear interpolation on the pixel values of each point in the corresponding metal region of the original projection sequence image obtained in step (1) according to the projection sequence image obtained in step (4) to obtain the preliminarily corrected original projection sequence image;

[0045] wherein, as shown in Figure 6 the linear interpolation is specifically:

[0046] wherein, (r, c) is the coordinate of the interpolated point, p(r, c) is the image pixel value of the interpolated point; (r1, c1) and (r2, c2) are the coordinates of the two selected interpolation boundary points, respectively, and p1(r1, c1) and p2(r2, c2) are the image pixel values of the two selected interpolation boundary points, respectively; l1 and l2 are the distances from the two selected interpolation boundary points to the interpolated point,

[0047] The metal region in the original projection sequence image is replaced by linear interpolation calculation, but at this time the two interpolation boundary points for linear interpolation calculation are relatively vague, only the preliminarily corrected original projection sequence image is obtained;

[0048] (6) performing back-projection reconstruction on the preliminarily corrected original projection sequence image obtained in step (5) to obtain the preliminarily corrected reconstructed image;

[0049] (7) Threshold segmentation is performed on the reconstructed image obtained in step (6) after preliminary correction, and the reconstructed image is reprojected accordingly;

[0050] The specific threshold segmentation is as follows:

[0051]

[0052] Where f(x,y) is the CT value of a point in the reconstructed image, f′(x,y) is the attenuation coefficient value of the corresponding point, and u 空气 u 软组织 u 骨头 These are the attenuation coefficient values ​​for air, soft tissue, and bone, respectively. Since only these three tissue structures exist within the scanned object in practical applications, segmentation is performed accordingly. 空气 and t 软组织 These are the set thresholds for air and soft tissue CT values, respectively.

[0053] That is, when the CT value of a certain point in the reconstructed image is less than t 空气 When, its attenuation coefficient value is set to u 空气 ;

[0054] When the CT value of a certain point in the reconstructed image is less than t 软组织 When, its attenuation coefficient value is set to u 软组织 ;

[0055] The attenuation coefficient value at other points is uniformly set to u. 骨头 ;

[0056] By reprojecting the attenuation coefficient values ​​of different tissues obtained after the above threshold segmentation, the tissue boundaries of the areas in the image that were originally blurred due to metal artifacts can be made clearer.

[0057] (8) Forward projection is performed on the image after recalibration in step (7) to obtain a projection prior image of the metal implant-free implant, such as Figure 3 As shown, where, Figure 5 The image shown is an example of a specific instance;

[0058] (9) Perform a direction search on the projection sequence image obtained in step (4) to determine the direction of metal extension on it;

[0059] like Figure 7 As shown, if we choose to search in four directions, that is, in the top, bottom, left, and right directions of the image, we can obtain four boundary coordinate values ​​(x, y, z) in these four directions respectively. 左 ,y 左 ), (x 右 y 右 ), (x 上 y 上 ), (x下 y 下 The direction of metal extension, i.e. the approximate direction of length, is calculated based on the boundary coordinates.

[0060] Similarly, searching in eight or more directions will yield the corresponding metal extension directions;

[0061] like Figure 7 As shown in Figures (a) and (b), the direction of the arrows below indicates the direction of metal extension being searched. Figure 7 It can be seen that the boundary points along the metal direction are far from the points in the metal region. If linear interpolation is performed, the final interpolation point may be located in the non-metal region. In this case, the obtained pixel value is not applicable to the points in the metal region. Therefore, it is not suitable to use the boundary point for interpolation. Thus, the boundary points along the metal extension direction are excluded. The boundary points perpendicular to the metal direction are the most suitable for interpolation. That is, the boundary points perpendicular to the metal extension direction are selected as the boundary points for linear interpolation of the pixel values ​​of each point in the metal region.

[0062] (10) Determine the pixel values ​​of each point located within the metal region in the projected prior map;

[0063] Because the two boundaries in the vertical direction may have significant differences (e.g. Figure 8 The metal structure is flanked by bone and soft tissue on either side, therefore further assessment and selection are required, as detailed below:

[0064] Based on the projection prior map obtained in step (8), the pixel values ​​of each point located in the metal region are calculated and compared with the pixel values ​​of the two boundary points perpendicular to the metal direction in the projection prior map obtained in step (9).

[0065] If the pixel values ​​of each point in the metal region are within the specified range from the pixel values ​​of the two boundary points, then the two boundary points are used for linear interpolation calculation. The linear interpolation calculation process is based on the formula in step (2), and will not be repeated here.

[0066] If only one boundary point has a pixel value difference from the interpolated point that is within the specified range, then the pixel value of that boundary point is selected to interpolate the interpolated point (specifically, the pixel value of the interpolated point is directly replaced by the pixel value of the boundary point).

[0067] If the pixel value difference between the current two boundary points and the pixel value of the interpolated point is not within the limited range, the coordinates of the current two boundary points are extended by one pixel coordinate on both sides, and then the pixel value difference between the current boundary points and the pixel value of the interpolated point is compared again. If it is within the limited range, the interpolation is performed according to the above-mentioned interpolation method (i.e. both boundary points satisfy the condition, or only one boundary point satisfies the condition); otherwise, the pixel value of the boundary point closest to the pixel value of the interpolated point in the projection prior map is selected for direct interpolation.

[0068] wherein the pixel value difference f s The calculation method is as follows:

[0069]

[0070] wherein (r0, c0) is the coordinate of the interpolated point, p(r0, c0) is the pixel value of the interpolated point, (r k ,c k ) is the coordinate of one of the boundary points, and p(r k ,c k ) is the pixel value of the corresponding boundary point.

[0071] The limited range is generally selected as 0.03; that is, when f s is less than or equal to 0.03, it is considered that the pixel value difference between the boundary point in the projection prior map and the interpolated point is within the limited range, and the interpolated point can be interpolated by the boundary point.

[0072] (11) Interpolating each point in the metal region in the original projection sequence image obtained in step (1) according to step (10) to obtain a corrected original projection sequence image, and performing back projection reconstruction to obtain a corrected metal-free reconstruction image;

[0073] (12) Fusing the metal region image obtained in step (3) and the corrected metal-free reconstruction image obtained in step (11) to obtain a projection sequence image after correction (i.e. removal of metal artifacts) of the metal, as shown in Figure 9 .

[0074] The present application solves the problem that the reconstructed three-dimensional CT image has serious metal artifacts caused by metal implants, which leads to the reduction of CT image quality and the reliability of CT image as the basis for medical diagnosis. Because there is a big difference between the interpolated metal point and the interpolated point, the pixel value of the interpolated metal point deviates far from its accurate value, thus reducing the clarity of the tissue boundary of the reconstructed three-dimensional CT image. The present application can effectively judge the similarity between the interpolated point and the interpolated metal point through the prior graph data generated by coarse correction, thereby significantly improving the accuracy of interpolation, effectively improving the clarity of the tissue boundary, and improving the quality of the three-dimensional CT image.

[0075] The above describes the preferred embodiments of the present application, but the present application is not limited to the specific details in the above embodiments, and various equivalent transformations (such as quantity, shape, position, etc.) can be made to the technical solutions of the present application within the technical concept range of the present application, and these equivalent transformations all belong to the protection range of the present application.

Claims

1. A method of metal artifact correction for three-dimensional imagery, the method comprising: The application comprises the following steps: Obtaining the original projection sequence image with metal implant by scanning and carrying out back-projection reconstruction, and segmenting the reconstructed image to obtain the metal region image and non-metal region image; Carrying out forward projection on the metal region image to obtain the metal projection image, carrying out linear interpolation on the corresponding metal region in the original projection sequence image with the metal projection image, and carrying out back-projection reconstruction; Carrying out threshold segmentation on the back-projection reconstructed image after the aforementioned interpolation, determining the corresponding attenuation coefficient according to the threshold of different tissues, re-projecting, and carrying out forward projection to obtain the projection prior image without metal implant, calculating the difference value of the pixel values of two boundary points and the interpolated point corresponding to the interpolated point in the projection prior image located in the metal region, judging whether the difference value is within the limited range, selecting the pixel value of the corresponding boundary point for interpolation to obtain the pixel value of the interpolated point, and carrying out interpolation on the original projection sequence image and carrying out back-projection reconstruction to obtain the final image by fusing the metal region image.

2. The method of claim 1, wherein: The extension direction of the metal implant is determined by searching the direction of the point in the metal projection image, and the extension direction of the metal region is determined according to the boundary point of the metal region.

3. The method of claim 2, wherein: The direction search is four-direction search or eight-direction search.

4. The method of claim 1, wherein: The pixel value of each point in the projection prior image located in the metal region is compared with the pixel values of the corresponding two boundary points; If the difference values of the pixel values of the two boundary points and the interpolated point are within the limited range, the pixel value of the interpolated point is calculated by linear interpolation with the pixel values of the two boundary points; If the difference value of the pixel value of one boundary point and the interpolated point is within the limited range, the pixel value of the interpolated point is interpolated with the pixel value of the boundary point; If the difference values of the pixel values of the two boundary points and the interpolated point are not within the limited range, the coordinates of the two boundary points are respectively expanded by one pixel coordinate to the left and right, and the difference values of the pixel values of the current boundary points and the interpolated point are compared again to see whether they are within the limited range, if yes, the interpolation is carried out according to the aforementioned interpolation method, otherwise, the pixel value of the interpolated point is interpolated with the pixel value of the boundary point in the projection prior image closest to the pixel value of the interpolated point.

5. The method of claim 1 or 4, wherein: The linear interpolation is specifically: ； wherein (r, c) is the coordinate of the interpolated point, p(r, c) is the image pixel value of the interpolated point; (r1, c1) and (r2, c2) are the coordinates of the two selected interpolation boundary points, respectively, p1(r1, c1) and p2(r2, c2) are the image pixel values of the two selected interpolation boundary points, respectively; l1 and l2 are the distances from the two selected interpolation boundary points to the interpolated point, respectively, , .

6. The method of metal artifact correction for three-dimensional imagery of claim 1, wherein: The threshold segmentation is specifically: When the pixel value of a point in the reconstructed image is less than the CT value of air, which is set as a threshold t 空气 , the attenuation coefficient value of the point is set as u 空气 ​ When the pixel value of a point in the reconstructed image is less than the CT value of the soft tissue, which is set as a threshold t 软组织 , the attenuation coefficient value of the point is set as u 软组织 . The attenuation coefficient values of the other points are uniformly set to u 骨头 .

Images