Region-of-interest reconstruction method for cone beam rotation CL

Abstract

A region-of-interest reconstruction method for cone beam rotation CL is characterized by comprising the following steps: S1, acquiring CL projection data and a corresponding CL system scanning matrix; s2, calculating a corresponding diagonal weighting matrix and a correction factor of the projection data of the region of interest according to the system scanning matrix obtained in the step S1; s3, according to the CL projection data collected in the step S1 and the correction factor designed in the step S2, establishing a weighted least square optimization model based on the region of interest (ROI); and S4, solving the weighted least square optimization model of the region of interest in the S3 by using a conjugate gradient method, and outputting a reconstructed CL image. Simulation and actual PCB experiment results show that compared with other classic reconstruction methods, the ROI-CGWLS method provided by the invention is effective in inhibiting ROI artifacts and improving image contrast.

Description

Technical Field

[0001] The invention belongs to the field of image reconstruction in the field of nondestructive testing of electronic components, and relates to a low-dose high-quality CT image restoration method for a CL reconstruction method of an area of ​​interest of an electronic component. Background Art

[0002] Computed tomography (CT) is a very powerful nondestructive inspection technique for the internal structure of objects because it can produce high-resolution reconstructions of Cone-beam rotation computed tomography (CL) is an efficient inspection technology for non-destructive inspection of high-aspect ratio objects such as printed circuit boards (PCBs) and insulated gate bipolar transistors (IGBTs). However, when scanning high-aspect ratio objects, the projection data may be truncated, resulting in region of interest (ROI) artifacts in the reconstructed image and reducing the contrast of the reconstructed image.

[0003] Under current state of the art, cone-beam rotational computed tomography (CL) can be easily achieved by tilting the scanned object or source and detector of a conventional CT system, which increases the field of view. It has been used for nondestructive testing in various fields such as fossils, paintings, printed circuit boards (PCBs), and insulated gate bipolar transistors (ICGBs). However, when the scanned object has a large aspect ratio, the projection data is truncated because it cannot be projected onto the detector as a whole. Only a small region of the entire object can be reconstructed from the projection data of CL. The truncated projection data causes rotational CL reconstruction to be a typical ill-posed problem, resulting in artifacts unique to tomography in the reconstructed image. Summary of the invention

[0004] To solve this problem, the present invention reconstructs the length of the rays within the volume and the distance between the X-ray source and the detector box position and develops a method called ROI conjugate gradient weighted least squares (ROI-CGWLS) to suppress ROI artifacts and improve the contrast of the reconstructed image in cone-beam rotation CL.

[0005] The technical solution of the present invention is: A region of interest reconstruction method for cone-beam rotation CL, The method comprises the following steps: S1: Obtain CL projection data and the corresponding CL system scanning matrix; S2: Calculate the corresponding diagonal weighting matrix and the correction factor of the projection data of the region of interest according to the system scanning matrix obtained in step S1; S3: establishing a weighted least squares optimization model based on a region of interest, i.e., ROI, according to the CL projection data collected in step S1 and the correction factor designed in step S2; S4: using the conjugate gradient method to solve the weighted least squares optimization model of the region of interest described in S3, and outputting the reconstructed CL image.

[0006] Furthermore, the parameters in the above step S1 are: cone beam rotation CL projection data p , the system scanning matrix of the imaging device under the CL scanning protocol A , the CL system scan matrix corresponding to the region of interest (ROI) A roi .

[0007] Further, the above S2 specifically comprises: a cone-beam rotating CL scanning system with a flat-panel detector; The X-ray source and detector rotate around circles with radii SO * sinψ and OD * sinψ, respectively; SO is the distance from the X-ray source S to the rotation center O, OD is the distance from the detector center D to the rotation center O, and ψ is the tilt angle; The angle θ between the y-axis and the projection of the cone center beam on the xy plane is the projection angle of the rotating CL system, p θ (a, b) represents the measured projection values ​​at the detector position (a, b); The region of interest of the scanned object is ROI; ROI reconstruction and correction: L v L represents the length of the ray in the reconstructed body of the region of interest; sd Indicates the distance between the X-ray source and the detector unit; In the cone of cone beam rotation CL using a virtual detector located at the rotation center of the sampling unit, only the sector that intersects the detector at the linear position a is considered; The projection position of the voxel (x, y, z) on the virtual detector is (a, b), which is point H; S' is the projection point of S on the plane xy; ψ represents the angle between line segment S'O and line segment HO; β represents the angle between line segment SO and line segment HO; The angle between line segment S 'O and line segment SO is π / 2-ψ; The L in S2 v The calculation is as follows: ,in, Aroi is the CL system scanning matrix corresponding to the region of interest (ROI); N roi is the total number of voxels in the region of interest (ROI) of the object to be reconstructed; The L in S2 sd The calculation is as follows: First calculate , further calculation ,get ; The calculation of S2 L sd The angle ψ used is calculated as follows: The angle ψ is determined by the position of the virtual detector (a, b) and the projection angle θ and can be calculated as follows: when When , ψ is calculated by the following formula: ; when When ψ is calculated by the following formula: ; The correction factor in S2 The calculation is as follows: .

[0008] For each projection angle θ, the correction factor Normalization is as follows: .

[0009] Furthermore, the above S3 is specifically as follows: in order to reconstruct a high-quality stereoscopic image of the region of interest, a least squares optimization model applicable to the cone-beam rotation CL region of interest is established as follows: ,in, f ROI is the region of interest of the object to be reconstructed, W CL is the modified weighting factor in S2, W roi is a diagonal matrix with diagonal elements of , N roi is the total number of voxels in the stereo image of the region of interest of the object to be reconstructed.

[0010] The region of interest with the reconstructed object in S3 f ROI , f ROI Width ,high ,depth Determined according to the following empirical formula: , where Round() is a floor operation, D C Represents the row number of the detector unit, C is a positive integer, C D Determined by the thickness of the object.

[0011] Furthermore, the above S4 is specifically: using the conjugate gradient method to solve the weighted least squares optimization model of the region of interest described in S3 ; Use the CL system scanning matrix corresponding to the region of interest (ROI) obtained in step S1 A roi and projection data p , design the maximum number of iterations K , initialize the region of interest to be reconstructed f ROI The initial image is ; The iteration format is as follows: 1) When satisfied k ≤ K When ; 2) When not satisfied k ≤ K When , the iteration stops, and the image of the region of interest of the object to be reconstructed is output .

[0012] The beneficial effects of the present invention are as follows: the present invention introduces a weighting factor that takes into account the length of the ray in the volume of the region of interest of the reconstructed body and the distance between the X-ray source and the detector unit position, and also develops a method called ROI conjugate gradient weighted least squares (ROI-CGWLS). The proposed method can effectively suppress the region of interest artifacts and improve the contrast of the reconstructed image. The result is better than the result reconstructed by other reconstruction methods.

[0013] However, the weighting factors of the proposed method depend on the distance between the X-ray source and the detector unit position, which leads to reconstruction values ​​lower than the desired results. A more accurate estimation of the thickness of the scanned object can further improve the weighting factors of the projection data.

[0014] In the technical solution of the present invention, the size of the region of interest is determined based on experience, and a method for automatically determining the size of the region of interest will be studied in future work. In addition, a projection data preprocessing method or a regularization method with prior information can be combined to further improve the quality of the reconstructed image, such as removing beam hardening artifacts, metal artifacts, etc.

[0015] Both simulation and actual PCB experimental results show that the proposed ROI- CGWLS method is effective in suppressing ROI artifacts and improving image contrast compared with other classical reconstruction methods. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 It is a schematic diagram of computed tomography imaging of cone-beam rotation CL in the prior art solution.

[0017] Figure 2 It is a schematic diagram of computed tomography imaging of a region of interest by cone-beam rotation CL according to the technical solution of the present invention.

[0018] Figure 3 It is a schematic diagram of reconstruction and correction calculation of the region of interest for cone-beam rotation CL of the present invention.

[0019] The part of the scanned object outside the reconstruction volume also contributes to the projection data; Lv represents the length of the ray within the volume of the region of interest of the reconstruction body; Lsd represents the distance between the X-ray source and the detection unit position.

[0020] Figure 4 is a geometric configuration diagram of cone-beam rotation CL using a virtual detector.

[0021] Figure 5 These are comparison diagrams of the technical effects of the present invention, wherein the last one is a reconstructed image of the present invention. DETAILED DESCRIPTION

[0022] The technical solution of the present invention is further described below in conjunction with the accompanying drawings like Figure 1 , the scanned object can be completely projected onto the detector, and a high-quality computed tomography (CT) reconstructed image can be obtained.

[0023] However, for cone-beam rotation CL, since the scanned object usually has a large planar shape, only part of the ROI is scanned, such as Figure 2 shown.

[0024] Figure 2 The upper and lower dashed lines in the figure represent the body layers ( f ), the middle rectangle represents the ROI (i.e. f ROI ), some parts of the body below the center layer are not scanned (indicated by the gray line on the lower left of the middle rectangle), and some parts of the body above the center layer are not scanned (indicated by the gray line on the upper right of the middle rectangle). f ROI Width ,high ,depth This can be determined using the following empirical formula: , where D C Represents the row number of the detector unit, C is a positive integer, C D Determined by the thickness of the object. The farther away from the center layer, the more parts are not scanned, which leads to more artifacts in the reconstructed image. The projection data of cone-beam rotational CL scanning is truncated, which leads to ROI artifacts in the reconstructed image and reduces the contrast.

[0025] ROI issues such as Figure 3 As shown, the area outside the scanned object also affects the projection data (see Figure 3 ), but are not accurately represented during the reconstruction process, resulting in ROI artifacts and reduced contrast in the reconstructed image. In order to suppress ROI artifacts and enhance the contrast of the reconstructed image, a method is proposed to exploit the prior geometric information of the scanned object, specifically, by calculating the X-ray penetration length in the object's sheet structure. In order to alleviate the limitations associated with sheet-like objects, the STL data of the scanned object is used as prior information to calculate the X-ray penetration length of the entire scanned object. Then, a binary voxel volume is created to minimize the ROI artifacts. Prior to the reconstruction process, a registration step is required, and its accuracy has a great impact on the quality of the reconstructed image.

[0026] In order to mitigate ROI artifacts and enhance the image contrast, we avoid using any information about the scanned object and instead propose a correction method based on a weighting factor for each ray. The weighting factor is determined by the length L of the ray within the reconstruction volume. V and the distance L between the radiation source and the detector sd OK, where L sd It has nothing to do with the object being scanned. The length of the ray within the creation volume is , L v This can be done by taking a vector whose elements are all 1 I It can be easily obtained by forward projection, that is, L v =1 / AI , in calculating the distance L sd Previously, it was necessary to place a virtual detector at the isocenter of the rotating sampling unit in order to make the mathematical representation of the geometric setup more intuitive.

[0027] See also Figure 4 , Figure 4 The geometry of the cone beam rotation CL using a virtual detector is shown. Within the cone, only rays in the fan-shaped region that are collinear with the detector and have a constant angle a are considered. The projection of a voxel (x, y, z) onto the virtual detector position (a, b) (i.e., point H).

[0028] S' is the projection point of S on the plane xy. The angle between line segment S'O and line segment HO is Ψ. The angle between line segment SO and line segment HO is β. The angle between line segment S'O and line segment SO is π / 2-ψ; If we obtain the distance SH between the ray source and the virtual detector unit position, then the distance L sd It can be obtained by the following formula ; According to the law of cosines and the minimum value theorem, the weighting factors are calculated as follows: .

[0029] If a ray passes through only a few voxels, then L V This indicates that most of the projections of the detector cells come from outside the reconstruction volume, resulting in some artifacts. CL to correct the projection data and vice versa. This weighting factor is incorporated into the projection data to suppress ROI artifacts and improve the contrast of the reconstructed image. sd The value is much larger than L V , which makes the value of the projection data smaller, so the dynamic range of the reconstructed image is also smaller. Therefore, at each projection angle θ, the weighting factor Perform the naturalization as follows: .

[0030] The ROI weighted least squares optimization model corresponding to the cone beam rotation CL is: ,in f ROI is the reconstructed region of interest of the object.

[0031] The conjugate gradient method is used to solve the region of interest weighted least squares optimization model (abbreviated as ROI-CGWLS). The ROI-CGWLS method is as follows: .

Claims

1. A method for region of interest reconstruction based on cone-beam rotation CL. It is characterized in that The method comprises the following steps: S1: Obtain CL projection data and the corresponding CL system scanning matrix; S2: Calculate the corresponding diagonal weighting matrix and the correction factor of the projection data of the region of interest according to the system scanning matrix obtained in step S1; S3: establishing a weighted least squares optimization model based on a region of interest, i.e., ROI, according to the CL projection data collected in step S1 and the correction factor designed in step S2; S4: using the conjugate gradient method to solve the weighted least squares optimization model of the region of interest described in S3, and outputting the reconstructed CL image.

2. The reconstruction method according to claim 1, characterized in that: The parameters in step S1 are: cone beam rotation CL projection data p , System scanning matrix of imaging device under CL scanning protocol A , the CL system scan matrix corresponding to the region of interest (ROI) A roi .

3. The reconstruction method according to claim 1, characterized in that: The S2 is specifically: a cone-beam rotating CL scanning system with a flat-panel detector; The X-ray source and detector rotate around circles with radii SO * sinψ and OD * sinψ, respectively; SO is the distance from the X-ray source S to the rotation center O, OD is the distance from the detector center D to the rotation center O, and ψ is the tilt angle; The angle θ between the y-axis and the projection of the cone center beam on the xy plane is the projection angle of the rotating CL system, p θ (a, b) represents the measured projection values ​​at the detector position (a, b); The region of interest of the scanned object is ROI; ROI reconstruction and correction: L v L represents the length of the ray in the reconstructed body of the region of interest; sd Indicates the distance between the X-ray source and the detector unit; In the cone of cone beam rotation CL using a virtual detector located at the rotation center of the sampling unit, only the sector that intersects the detector at the linear position a is considered; The projection position of the voxel (x, y, z) on the virtual detector is (a, b), which is point H; S' is the projection point of S on the plane xy; ψ represents the angle between line segment S'O and line segment HO; β represents the angle between line segment SO and line segment HO; The angle between line segment S 'O and line segment SO is π / 2-ψ; The L in S2 v The calculation is as follows: ,in, A roi is the CL system scanning matrix corresponding to the region of interest (ROI); N roi is the total number of voxels in the region of interest (ROI) of the object to be reconstructed; The L in S2 sd The calculation is as follows: First calculate, , further calculation ,get ; The calculation of S2 L sd The angle ψ used is calculated as follows: The angle ψ is determined by the position of the virtual detector (a, b) and the projection angle θ, and can be calculated as follows: when When ψ is calculated by the following formula: ; when When ψ is calculated by the following formula: ; The correction factor in S2 The calculation is as follows: ， For each projection angle θ, the correction factor Normalization is as follows: 。 4. The reconstruction method according to claim 1, characterized in that: The S3 is specifically as follows: in order to reconstruct a high-quality stereoscopic image of the region of interest, a least squares optimization model applicable to the cone-beam rotation CL region of interest is established as follows: in, f ROI is the region of interest of the object to be reconstructed, W CL is the modified weighting factor in S2, W roi is a diagonal matrix with diagonal elements of , N roi is the total number of voxels in the stereo image of the region of interest of the object to be reconstructed, The region of interest with the reconstructed object in S3 f ROI , f ROI Width ,high ,depth Determined according to the following empirical formula: ， Round() is a round-down operation. C Represents the row number of the detector unit, C is a positive integer, C D Determined by the thickness of the object.

5. The reconstruction method according to claim 1, characterized in that: The specific step S4 is: using the conjugate gradient method to solve the weighted least squares optimization model of the region of interest in S3 ; Use the CL system scanning matrix corresponding to the region of interest (ROI) obtained in step S1 A roi and projection data p , design the maximum number of iterations K , initialize the region of interest to be reconstructed f ROI The initial value image of is, ; The iteration format is as follows: 1) When satisfied k ≤ K When ； 2) When not satisfied k ≤ K When , the iteration stops, and the image of the region of interest of the object to be reconstructed is output .

Images