CT image rapid reconstruction method and device and program product

Abstract

The invention discloses a CT image rapid reconstruction method and device and a program product, and is applied to the technical field of image reconstruction, and the method comprises the steps: obtaining two-dimensional projection data and slice parameters of a target object; wherein the slice parameters comprise positioning point coordinate data and a plane normal vector of a target slice in the target object; performing sampling processing according to the slice parameters to obtain a three-dimensional sampling point set of the target slice; and performing weighted back projection processing on the three-dimensional sampling point set according to the two-dimensional projection data to obtain a reconstructed image of the target slice. According to the method, rapid reconstruction of the slice in any direction in the CT image is realized, and invalid calculation of the non-interested area is avoided, so that the reconstruction efficiency of the CT image is effectively improved.

Description

Technical Field

[0001] This application relates to the field of image reconstruction technology, and in particular to a method, apparatus and program product for rapid reconstruction of CT images. Background Technology

[0002] X-ray computed tomography (XCT) is a novel non-destructive testing technique used to inspect and analyze the internal and external structures of industrial products. In XCT, an object is irradiated by a cone-beam X-ray beam during scanning. The beam penetrates and forms a two-dimensional projection image on a planar detector. To obtain the product's three-dimensional shape, the two-dimensional projection images from various angles need to be reconstructed into three-dimensional volume data. Currently, the Feldkamp-Davis-Kress (FDK) algorithm is a classic method for reconstructing three-dimensional volume data from two-dimensional projections under cone-beam conditions, and it is also the most widely used reconstruction algorithm in industrial computed tomography (CT). However, in practical applications, often only specific internal structures or cross-sections in specific directions are considered. The FDK algorithm typically performs voxel reconstruction of the entire three-dimensional scanning field of view or a fixed cubic region based on Cartesian coordinates. This means that the complete three-dimensional volume data must first be reconstructed before extracting slices. This operation forces a large number of non-interested regions to participate in complex weighting, filtering, and backprojection calculations, resulting in a huge waste of computing resources and a significant increase in reconstruction time, which in turn leads to low reconstruction efficiency of CT images. Summary of the Invention

[0003] This application provides a method, apparatus, and program product for rapid reconstruction of CT images, used to achieve rapid reconstruction of slices in any direction in CT images.

[0004] On the one hand, embodiments of this application provide a method for rapid reconstruction of CT images, including:

[0005] Obtain the two-dimensional projection data and slicing parameters of the target object; wherein, the slicing parameters include the coordinate data of the positioning points of the target slice in the target object and the plane normal vector;

[0006] The sampling process is performed based on the slice parameters to obtain a three-dimensional sampling point set of the target slice;

[0007] The three-dimensional sampling point set is weighted and back-projected based on the two-dimensional projection data to obtain the reconstructed image of the target slice.

[0008] Further, in one embodiment, the step of performing sampling processing based on the slice parameters to obtain a three-dimensional sampling point set of the target slice includes:

[0009] Obtain the set of edges used to describe the physical boundaries of the volume data;

[0010] Based on the slice parameters, basis vector operations are performed to obtain the local orthogonal basis vectors of the target slice;

[0011] The target slice's width, height, and pixel resolution are obtained by performing localization processing based on the local orthogonal basis vectors and the edge set.

[0012] Based on the width, height, and pixel resolution of the target slice, the target slice is sampled to obtain a three-dimensional sampling point set of the target slice.

[0013] Further, in one embodiment, obtaining the set of edges used to describe the physical boundaries of the volume data includes:

[0014] Based on the dimensions, spatial starting position, and voxel spacing of the volume data of the target object, the physical range of the volume data of the target object in the object's three-dimensional coordinate system is obtained;

[0015] Based on the physical range, determine the coordinate data of each vertex in the volume data of the target object;

[0016] The edge set is constructed based on the coordinate data of each vertex.

[0017] Further, in one embodiment, the local orthogonal basis vectors include horizontal unit basis vectors and vertical unit basis vectors; the step of performing basis vector operations based on the slice parameters to obtain the local orthogonal basis vectors of the target slice includes:

[0018] Obtain the auxiliary vector of the target slice; wherein the auxiliary vector is not parallel to the plane normal vector;

[0019] The auxiliary vector and the plane normal vector are subjected to cross product and normalization to obtain the horizontal unit basis vector;

[0020] The vertical unit basis vector is obtained by performing a cross product on the horizontal unit basis vector and the plane normal vector.

[0021] Further, in one embodiment, the step of performing positioning processing based on the local orthogonal basis vectors and the edge set to obtain the width, height, and pixel resolution of the target slice includes:

[0022] For each edge in the edge set, if the current edge intersects with the target plane, the intersection point of the current edge and the target plane is obtained and added to the cross-sectional contour point set; otherwise, the current edge is discarded. The target plane is a plane determined by the plane normal vector and the coordinate data of the positioning point.

[0023] Based on the local orthogonal basis vectors, each intersection point in the cross-sectional contour point set is projected onto the local coordinate system of the slice to obtain the coordinate data of each intersection point in the cross-sectional contour point set in the local coordinate system of the slice.

[0024] Based on the coordinate data of each intersection point in the local coordinate system of the slice, determine the extreme value range of the local coordinate system of the slice;

[0025] Based on the extreme value range and the preset slice sampling interval, the width and height of the target slice, as well as the pixel resolution of the target slice in the slice local coordinate system, are obtained.

[0026] Further, in one embodiment, the step of sampling the target slice according to its width, height, and pixel resolution to obtain a three-dimensional sampling point set of the target slice includes:

[0027] Within the extreme range of the local coordinate system of the slice, the target slice is subjected to grid sampling scanning based on the width, height and pixel resolution of the target slice;

[0028] For each sampling point obtained by the sampling scan, based on the extreme value range and the preset slice sampling interval, and combined with the pixel index of the current sampling point, the two-dimensional local coordinate data of the current sampling point are determined;

[0029] The two-dimensional local coordinate data of each sampling point are transformed from the local coordinate system of the slice to the three-dimensional coordinate system of the object to obtain the three-dimensional coordinate data of each sampling point, and then the three-dimensional sampling point set is constructed.

[0030] Further, in one embodiment, the step of performing weighted backprojection processing on the three-dimensional sampling point set based on the two-dimensional projection data to obtain the reconstructed image of the target slice includes:

[0031] The two-dimensional projection data is corrected to obtain the corrected two-dimensional projection data.

[0032] The corrected two-dimensional projection data is filtered to obtain filtered two-dimensional projection data.

[0033] The back-projection integral operation is performed on the filtered two-dimensional projection data and the three-dimensional sampling point set to obtain the reconstructed value of each sampling point in the three-dimensional sampling point set.

[0034] The reconstructed image is obtained by assembling the reconstructed values ​​of each sampling point.

[0035] Further, in one embodiment, the step of performing back-projection integration based on the filtered two-dimensional projection data and the three-dimensional sampling point set to obtain the reconstructed value of each sampling point in the three-dimensional sampling point set includes:

[0036] The three-dimensional coordinate data of each sampling point are projected from the object's three-dimensional coordinate system to the rotating coordinate system to obtain the rotating coordinate data of each sampling point;

[0037] Using a filtered back-projection algorithm and the filtered two-dimensional projection data, the rotation coordinate data of each sampling point are processed by back-projection integration to obtain the reconstructed value of each sampling point.

[0038] On the other hand, embodiments of this application provide a rapid CT image reconstruction apparatus, comprising:

[0039] The acquisition module is used to acquire two-dimensional projection data and slicing parameters of the target object; wherein, the slicing parameters include the coordinate data of the positioning points of the target slice in the target object and the plane normal vector;

[0040] The first processing module is used to perform sampling processing based on the slice parameters to obtain a three-dimensional sampling point set of the target slice;

[0041] The second processing module is used to perform weighted back projection processing on the three-dimensional sampling point set based on the two-dimensional projection data to obtain the reconstructed image of the target slice.

[0042] In another aspect, embodiments of this application provide a computer program product, including a computer program that, when executed by a processor, implements the aforementioned method for rapid reconstruction of CT images.

[0043] According to embodiments of this application, a method, apparatus, and program product for rapid reconstruction of CT images are provided. The method acquires two-dimensional projection data and slice parameters of a target object. The slice parameters include the coordinate data of positioning points and the plane normal vector of the target slice within the target object. Sampling processing is performed based on the slice parameters to obtain a three-dimensional sampling point set of the target slice. Weighted backprojection processing is then performed on the three-dimensional sampling point set based on the two-dimensional projection data to obtain a reconstructed image of the target slice. The technical solution of this application embodiment enables rapid reconstruction of slices in any direction within CT images, avoiding invalid calculations for non-interesting regions, thereby effectively improving the reconstruction efficiency of CT images. Attached Figure Description

[0044] Figure 1This is a flowchart of a rapid CT image reconstruction method provided in this application;

[0045] Figure 2 This is a flowchart of the slice geometry construction provided in this application;

[0046] Figure 3 This is an example diagram of the geometric positional relationship of cone-beam CT provided in this application;

[0047] Figure 4 This is a flowchart of FDK reconstruction provided in this application;

[0048] Figure 5 This is a diagram illustrating the specific implementation process of a rapid CT image reconstruction method provided in this application. Detailed Implementation

[0049] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0050] The present application will be further described below with reference to the accompanying drawings and specific embodiments. The described embodiments should not be considered as limitations on the present application, and all other embodiments obtained by those skilled in the art without inventive effort are within the scope of protection of the present application.

[0051] In the following description, references are made to “some embodiments,” which describe a subset of all possible embodiments. However, it is understood that “some embodiments” may be the same subset or different subsets of all possible embodiments and may be combined with each other without conflict.

[0052] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing embodiments of this application only and is not intended to limit this application.

[0053] First, the terminology used in the embodiments of this application will be explained:

[0054] (1) A voxel is the smallest unit in three-dimensional space segmentation, similar in concept to a pixel in a two-dimensional image. In CT reconstruction, a voxel represents the physical properties of a tiny cubic region inside an object.

[0055] (2) The three-dimensional scanning field of view (FOV) refers to the three-dimensional spatial area that the CT system can effectively image. It is determined by the angle of the X-ray source, the size of the detector and the scanning geometry. Objects outside this range cannot be completely reconstructed.

[0056] (3) Cone-beam CT is a CT imaging technology in which the X-ray source emits a cone-shaped beam and is used in conjunction with a two-dimensional array detector. Its characteristic is that it can acquire complete three-dimensional data of an object by rotating once.

[0057] (4) Projection data refers to the two-dimensional X-ray fluoroscopic images acquired by the detector at different rotation angles during the CT scan. These raw data are the basis for subsequent reconstruction into three-dimensional images through algorithms.

[0058] (5) Back projection, which is the core mathematical step in CT reconstruction. In layman's terms, it is to "smear" the two-dimensional projection information received by the detector back into three-dimensional space along the opposite path of X-ray propagation, and restore the internal structure of the object by accumulating the information from various angles.

[0059] (6) A slice refers to a two-dimensional image formed after an object is cut off by a plane in three-dimensional space. In traditional CT, it usually refers to a cross-section perpendicular to the axis of rotation; however, in the embodiments of this application, it specifically refers to a two-dimensional cross-section in any direction determined by an arbitrary center point and normal vector specified by the user.

[0060] (7) Volume data refers to a discretized data structure used to represent the internal information of a three-dimensional object. It consists of an array of voxels arranged regularly in three-dimensional space, and each voxel contains physical attribute values ​​corresponding to its spatial position.

[0061] The embodiments of this application will be further explained and described below with reference to the accompanying drawings and specific examples.

[0062] X-ray computed tomography (XCT) is a novel non-destructive testing (NDT) technique used to inspect and analyze the internal and external structures of industrial products. Its principle is similar to traditional XCT systems, utilizing X-rays to project images onto an object from different directions, reconstructing the product's three-dimensional shape. Because XCT scanning can reveal the internal structure of a product without damaging it, it has been widely adopted for product quality control and has gained significant attention in industries such as aerospace, automotive, food packaging, and additive manufacturing.

[0063] Specifically, in XCT inspection technology, an object is irradiated by a cone beam of X-rays during the scanning process. After the rays penetrate, they form a two-dimensional projection image on a planar detector. In order to obtain the three-dimensional shape of the product, it is necessary to reconstruct the three-dimensional volume data from the two-dimensional projection images from various angles.

[0064] Currently, filtered backprojection algorithms are a classic method for reconstructing 3D volume data from 2D projection under cone-beam projection conditions, and are also the most widely used reconstruction algorithms in industrial CT, such as the FDK algorithm. The overall process of the FDK algorithm can be divided into three main steps. The first step is weighted correction, which involves geometrically weighting the original projection data to compensate for the non-uniformity of cone-beam projection at different angles and positions. The second step is convolution filtering, which involves performing row-by-row one-dimensional filtering on the weighted and corrected projection data to enhance edges and improve the spatial resolution of the reconstruction. The third step is weighted backprojection, which involves weighted backprojecting the filtered projection data back into 3D space along the X-ray path to reconstruct the voxel gray values. The weight function of the backprojection depends on the distance from the reconstruction point to the focal point. Through the above steps, the FDK algorithm can directly reconstruct the entire 3D volume data, thereby obtaining a 3D voxel model of the object's internal structure.

[0065] However, in practical applications, the focus is often only on specific internal structures or cross-sections in a specific direction. The FDK algorithm, on the other hand, typically performs voxel reconstruction of the entire 3D scanning field of view or a fixed cubic region based on Cartesian coordinates. This means that the complete 3D volume data must first be reconstructed before extracting slices. This operation forces a large number of non-interest-bearing regions to participate in complex weighting, filtering, and backprojection calculations, resulting in a significant waste of computational resources and a substantial increase in reconstruction time, ultimately leading to low reconstruction efficiency of CT images.

[0066] In view of this, embodiments of this application provide a method, apparatus and program product for rapid reconstruction of CT images. Based on the FDK algorithm, it improves the back-projection geometric model and can perform weighted back-projection on a user-defined slice area point set, thereby realizing rapid reconstruction of slice images in any direction.

[0067] First, the implementation steps of a rapid CT image reconstruction method provided in the embodiments of this application will be described in detail below.

[0068] This application provides a method for rapid reconstruction of CT images, which can be applied to terminals, servers, or software running on either terminal or server. Terminals can be tablets, laptops, desktop computers, etc., but are not limited to these. Servers can be independent physical servers, server clusters or distributed systems composed of multiple physical servers, or cloud servers providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Furthermore, a server can be a node server in a blockchain network, but is not limited to these. Blockchain is a new application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms.

[0069] Reference Figure 1 , Figure 1 This is a flowchart of a rapid reconstruction method for CT images provided in this application, which may include the following steps S101-S103.

[0070] S101, obtain the two-dimensional projection data and slicing parameters of the target object; wherein, the slicing parameters include the coordinate data of the positioning points of the target slice in the target object and the plane normal vector.

[0071] It should be noted that the target object refers to the object to be reconstructed, and the target slice refers to the two-dimensional slice region in the target object to be reconstructed. This region is specified in advance by the user object.

[0072] In this step, the cone-beam CT scan projection data of the target object is obtained as the two-dimensional projection data of the target object, and the slicing parameters set by the user object for the target slice are obtained. These slicing parameters may include, but are not limited to, the location point coordinate data and plane normal vector of the target slice.

[0073] S102, perform sampling processing based on the slice parameters to obtain the three-dimensional sampling point set of the target slice.

[0074] It should be noted that the set of three-dimensional sampling points may include, but is not limited to, multiple sampling points and the coordinate data of each sampling point in the object's three-dimensional coordinate system. Here, the object's three-dimensional coordinate system refers to the three-dimensional coordinate system in which the target object resides.

[0075] In this step, sampling is performed based on the slice parameters to reverse-calculate the specific coordinate point set of the target slice in the object's three-dimensional coordinate system, i.e., the three-dimensional sampling point set.

[0076] S103, perform weighted back projection processing on the three-dimensional sampling point set based on the two-dimensional projection data to obtain the reconstructed image of the target slice.

[0077] In this step, after obtaining the three-dimensional sampling point set, the three-dimensional sampling point set is weighted back-projected based on the two-dimensional projection data and the FDK algorithm to obtain the reconstructed image of the target slice, thereby realizing the rapid reconstruction of CT images for a specific slice area.

[0078] Therefore, this application embodiment changes the traditional FDK algorithm's approach of first reconstructing complete 3D volume data and then extracting slices. By directly constructing a sampling point set of the target slice in 3D space, it achieves on-demand reconstruction of arbitrary slices. Specifically, this application embodiment abandons the traditional method of post-reconstruction processing. By inputting the center coordinates and normal vector of the slice, it directly reconstructs the spatial geometric description of the slice plane, i.e., the 3D sampling point set. Based on this, in the back-projection stage, it directly calculates the projection position of the pixels on the slice plane onto the detector, thereby skipping the process of generating 3D volume data and directly outputting slice images in any direction. Thus, this application embodiment reduces the reconstruction computation from the 3D volume level to the 2D slice level by locking a specific slice region point set. This on-demand reconstruction method simplifies the data processing flow, avoids invalid calculations for non-interesting regions, significantly reduces computational complexity and memory usage, and enables rapid reconstruction of slices in any direction in high-resolution CT images with ordinary hardware configurations, thereby improving the reconstruction efficiency of CT images.

[0079] In some implementations, in order to achieve local reconstruction of the target slice, it is first necessary to determine the target slice in the object's three-dimensional coordinate system. The three-dimensional sampling point set below Therefore, this embodiment provides a method for rapidly generating target slices in the object's three-dimensional coordinate system. The sampling point set method uses a parameterized form of "location point + direction vector" to represent the position and orientation of the target slice in any space. Based on this, the effective sampling area on the plane where the target slice is located is adaptively calculated and the corresponding three-dimensional sampling point set is generated based on the geometric boundary of the volume data of the target object, thereby improving the accuracy and sampling efficiency of the sampling points of the target slice.

[0080] Specifically, refer to Figure 2 In step S102 above, sampling processing is performed based on the slice parameters to obtain a three-dimensional sampling point set of the target slice, which may include the following steps S201-S204.

[0081] S201, Obtain the set of edges used to describe the physical boundaries of the volume data.

[0082] In this step, an edge set is constructed, which includes multiple edges that collectively describe the physical boundaries of the target object's volume data.

[0083] Furthermore, in some embodiments, in step S201 above, obtaining the edge set used to describe the physical boundary of the volume data includes: obtaining the physical range of the volume data of the target object in the three-dimensional coordinate system of the object based on the dimension, spatial starting position and voxel spacing of the volume data of the target object; determining the coordinate data of each vertex in the volume data of the target object based on the physical range; and constructing the edge set based on the coordinate data of each vertex.

[0084] Here, firstly, based on the dimensions of the target object's volume data... The voxel spacing and the spatial starting position of the volume data determine the volume data in the object's three-dimensional coordinate system. The physical extent of the target slice is determined. Then, based on this physical extent, the coordinates of the eight vertices in the volumetric data cube are calculated. Finally, based on the coordinates (i.e., indices) of each vertex, a set of 12 edges describing the boundary of the volumetric data cube is constructed. This process establishes the geometric constraints of the target slice in subsequent calculations.

[0085] S202, perform basis vector operations based on the slice parameters to obtain the local orthogonal basis vectors of the target slice.

[0086] In this step, based on the location point coordinate data and plane normal vector of the target slice, a local orthogonal coordinate system is constructed on the plane where the target slice is located, and then the basis vectors orthogonal to the plane where the target slice is located are determined, namely the local orthogonal basis vectors.

[0087] Furthermore, in some embodiments, the aforementioned local orthogonal basis vectors may include horizontal unit basis vectors and vertical unit basis vectors; in step S202, the basis vector operation processing based on the slice parameters to obtain the local orthogonal basis vectors of the target slice includes: obtaining auxiliary vectors of the target slice; wherein, the auxiliary vectors and the plane normal vectors are not parallel to each other; performing cross product and normalization processing on the auxiliary vectors and the plane normal vectors to obtain horizontal unit basis vectors; and performing cross product processing on the horizontal unit basis vectors and the plane normal vectors to obtain vertical unit basis vectors.

[0088] Here, the slicing parameters may include, but are not limited to, the coordinate data of the target slice's positioning points. and plane normal vector To establish a two-dimensional parametric coordinate system on the plane containing the target slice. It is necessary to construct local orthogonal basis vectors for the target slice.

[0089] First, choose a plane normal vector. Non-parallel auxiliary vectors It should be understood that the auxiliary vector can be any vector that is not parallel to the plane's normal vector. The vector.

[0090] Then, through the plane normal vector and auxiliary vector The cross product result is normalized, and the horizontal unit basis vector of the plane where the target slice is located is calculated, i.e., the horizontal unit basis vector, as shown in the following formula (1):

[0091] (1);

[0092] In equation (1), This represents the horizontal unit basis vector.

[0093] Finally, by using the plane normal vector and horizontal unit basis vectors Perform a cross product to calculate the unit basis vector in the perpendicular direction of the plane containing the target slice, i.e., the perpendicular unit basis vector, as shown in the following formula (2):

[0094] (2);

[0095] In equation (2), This represents the vertical unit basis vector.

[0096] The above process yields the orthogonal local basis vectors on the plane containing the target slice. .

[0097] S203, based on the local orthogonal basis vectors and edge set, performs positioning processing to obtain the width, height and pixel resolution of the target slice.

[0098] In this step, based on the local orthogonal basis vectors obtained in the previous steps, the geometric intersection of the plane containing the target slice and the preset volume data bounding box is calculated. By solving for the intersection points of the plane containing the target slice and each edge in the edge set obtained in the previous steps, the effective physical range of the target slice in the local coordinate system can be further determined, namely the width and height of the target slice, and the pixel resolution of the target slice can also be further determined.

[0099] Further, in some embodiments, step S203 above, which involves performing positioning processing based on local orthogonal basis vectors and edge sets to obtain the width, height, and pixel resolution of the target slice, includes: for each edge in the edge set, if the current edge intersects with the target plane, then the intersection point between the current edge and the target plane is obtained and added to the cross-sectional contour point set; otherwise, the current edge is discarded; wherein, the target plane is a plane determined by the plane normal vector and the coordinate data of the positioning points; based on the local orthogonal basis vectors, each intersection point in the cross-sectional contour point set is projected onto the slice local coordinate system to obtain the coordinate data of each intersection point in the cross-sectional contour point set in the slice local coordinate system; based on the coordinate data of each intersection point in the slice local coordinate system, the extreme value range of the slice local coordinate system is determined; based on the extreme value range and a preset slice sampling interval, the width and height of the target slice, as well as the pixel resolution of the target slice in the slice local coordinate system, are obtained.

[0100] Here, it will be determined by the plane normal vector. and positioning point coordinate data The defined infinite plane is defined as the target plane, and each edge in the edge set constructed in step S201 above is traversed. For each edge, it is determined whether the current edge intersects with the target plane. If so, the intersection point between the current edge and the target plane is obtained and added to the cross-sectional profile point set. Otherwise, discard the current edge.

[0101] Next, the locally orthogonal basis vectors obtained in the aforementioned steps are used... Set the cross-sectional profile points Each intersection point is projected onto the local coordinate system of the slice, as shown in formula (3) below:

[0102] , (3);

[0103] In equation (3), Indicates the intersection point in the local coordinate system of the slice. Axis coordinate values; Indicates the intersection point in the local coordinate system of the slice. Axis coordinate values.

[0104] Then, based on the coordinate data of each intersection point in the local coordinate system of the slice, the coordinates of all intersection points in the local coordinate system of the slice are calculated. Minimum value on the axis With the maximum value Thus, the local coordinate system of the slice is obtained. Extreme range of the axis Simultaneously, based on the coordinate data of each intersection point in the local coordinate system of the slice, the coordinates of all intersection points in the local coordinate system of the slice are... Minimum value on the axis With the maximum value Thus, the local coordinate system of the slice is obtained. Extreme range of the axis .

[0105] Finally, based on the aforementioned extreme value range and the preset slice sampling interval, the physical width and physical height of the target slice, as well as the required pixel resolution of the target slice in the slice local coordinate system, are determined. As shown in the following formula (4):

[0106] , (4);

[0107] In equation (4), Indicates the width of the target slice; Indicates the height of the target slice.

[0108] S204. Based on the width, height, and pixel resolution of the target slice, perform sampling processing on the target slice to obtain a three-dimensional sampling point set of the target slice.

[0109] In this step, within the defined effective physical range, the target slice is divided into grids according to its pixel resolution, and each two-dimensional pixel point (i.e., sampling point) of the target slice is mapped to the object's three-dimensional coordinate system, thereby generating a three-dimensional sampling point set of the target slice.

[0110] Furthermore, in some embodiments, in step S204 above, sampling processing is performed on the target slice according to the width, height, and pixel resolution to obtain a three-dimensional sampling point set of the target slice. This includes: performing grid sampling scanning on the target slice according to the width, height, and pixel resolution within the extreme value range of the slice's local coordinate system; for each sampling point obtained by the sampling scan, determining the two-dimensional local coordinate data of the current sampling point based on the extreme value range and a preset slice sampling interval, combined with the pixel index of the current sampling point; transforming the two-dimensional local coordinate data of each sampling point from the slice's local coordinate system to the object's three-dimensional coordinate system to obtain the three-dimensional coordinate data of each sampling point, thereby constructing a three-dimensional sampling point set.

[0111] Here, within the boundary range of the local coordinate system of the slice. Within this range, the target slice is divided into grids and scanned according to its pixel resolution in the local coordinate system of the slice. For each sampling point obtained from the sampling scan, we have:

[0112] First, based on the extreme value range and the preset slice sampling interval, combined with the pixel index of the current sampling point... The two-dimensional local coordinate data of the current sampling point are determined as shown in the following formula (5):

[0113] , , , (5);

[0114] In equation (5), Indicates the slice sampling interval.

[0115] Then, the two-dimensional local coordinate data of the current sampling point is transformed from the slice local coordinate system to the object three-dimensional coordinate system to obtain the final three-dimensional sampling coordinates of the current sampling point, i.e., the three-dimensional coordinate data, as shown in the following formula (6):

[0116] (6);

[0117] In equation (6), Indicates pixel index is The three-dimensional coordinate data of the sampling points.

[0118] By traversing all the sampling points obtained from the sampling scan, the complete target slice region in the object's three-dimensional coordinate system can be obtained. The three-dimensional sampling point set below .

[0119] In some implementations, after obtaining the 3D sampling point set, FDK reconstruction is performed directly on the point set. Unlike traditional reconstruction methods that traverse the entire voxel space, this implementation only reconstructs the aforementioned point set.

[0120] Specifically, refer to Figure 3 and Figure 4 , Figure 3 The image shows the geometric positional relationships of cone-beam CT. For the coordinate system of the virtual detector projection data, Let be the coordinate system where the ray source is located, i.e., the rotating coordinate system. The object's spatial coordinate system is the object's three-dimensional coordinate system. In step S103 above, the three-dimensional sampling point set is subjected to weighted back projection processing based on the two-dimensional projection data to obtain the reconstructed image of the target slice, including the following steps S301-S304.

[0121] S301, Correct the two-dimensional projection data to obtain the corrected two-dimensional projection data.

[0122] In this step, since the cone-beam X-rays are divergent, the original projection data needs to be geometrically corrected to obtain the corrected two-dimensional projection data, as shown in the following formula (7):

[0123] (7);

[0124] In equation (7), Represents the original two-dimensional projection data; This represents the corrected two-dimensional projection data; The coordinate data representing the coordinate system of the virtual detector projection data; This represents the distance from the source point to the center of rotation.

[0125] S302, the corrected two-dimensional projection data is filtered to obtain filtered two-dimensional projection data.

[0126] In this step, the corrected two-dimensional projection data is subjected to one-dimensional ramp filtering row by row to obtain the filtered two-dimensional projection data, as shown in the following formula (8):

[0127] (8);

[0128] In equation (8), This represents the filtered two-dimensional projection data; Represents the convolution function; This represents a one-dimensional convolution operation.

[0129] S303, perform back projection integration based on the filtered two-dimensional projection data and the three-dimensional sampling point set to obtain the reconstructed values ​​of each sampling point in the three-dimensional sampling point set.

[0130] In this step, the three-dimensional sampling point set obtained in the previous steps is used. The filtered two-dimensional projection data is then substituted into the back projection integral formula of FDK to perform back projection integral operation, thereby obtaining the reconstructed values ​​of each sampling point in the three-dimensional sampling point set.

[0131] Furthermore, in some embodiments, in step S303 above, the back-projection integration operation is performed based on the filtered two-dimensional projection data and the three-dimensional sampling point set to obtain the reconstructed value of each sampling point in the three-dimensional sampling point set, including: projecting the three-dimensional coordinate data of each sampling point from the object's three-dimensional coordinate system to a rotating coordinate system to obtain the rotating coordinate data of each sampling point; and using the filtered back-projection algorithm and the filtered two-dimensional projection data, performing back-projection integration operation on the rotating coordinate data of each sampling point to obtain the reconstructed value of each sampling point.

[0132] Here, for the three-dimensional sampling point set For each sampling point in the image, there are: for each fixed projection angle... , the three-dimensional coordinate data of the current sampling point By projecting the object's 3D coordinate system onto the rotated coordinate system, we obtain the coordinate data of the current sampling point in the rotated coordinate system, i.e., the rotated coordinate data. As shown in the following formula (9):

[0133] , (9).

[0134] Subsequently, the rotation coordinate data of the current sampling point and the filtered two-dimensional projection data are substituted into the back projection integral formula of the FDK algorithm to obtain the reconstructed value of the current sampling point, as shown in the following formula (10):

[0135] (10);

[0136] In equation (10), This represents the coordinate data of the current sampling point in the rotating coordinate system, i.e., the reconstructed value, which is a grayscale value.

[0137] S304. Based on the reconstructed values ​​of each sampling point, the image is assembled to obtain the reconstructed image.

[0138] In this step, the reconstructed values ​​of each sampling point are assembled according to the two-dimensional pixel arrangement order of the target slice, thereby outputting the reconstructed image of the target slice. This process does not require the generation of complete three-dimensional volume data.

[0139] The method of this application embodiment will be explained in principle below using an application scenario. (Refer to...) Figure 5 In this application scenario, XCT technology is used to reconstruct a slice of an object, referred to as the target object and the slice as the target slice. The reconstruction process is specifically divided into the geometric construction of the slice and the FDK reconstruction process based on sparse point sets. The geometric reconstruction of the slice refers to determining the position and orientation of the slice by inputting point coordinates and normal vectors by the user, then limiting the size of the slice by a 3D volume frame, and finally generating a set of coordinate points for the slice region. The FDK reconstruction based on sparse point sets refers to weighting and correcting the projection data, performing convolution filtering, then backprojecting the generated point set of the specific slice region, and finally outputting the slice image.

[0140] The specific reconstruction process is as follows:

[0141] 1) Obtain the cone-beam CT scan projection data of the target object as the two-dimensional projection data of the target object, and obtain the slice parameters set by the user object for the target slice. The slice parameters may include, but are not limited to, the location point coordinate data and plane normal vector of the target slice.

[0142] 2) Construct a 3D sampling point set for the target slice:

[0143] 2.1) Constructing the Physical Boundary of Volume Data: First, based on the dimensions, voxel spacing, and spatial starting position of the volume data of the target object, determine the physical extent of the volume data in the object's three-dimensional coordinate system. Then, based on the physical extent, calculate the coordinate data of the eight vertices in the volume data cube. Finally, based on the coordinate data (i.e., the index) of each vertex, construct a set of 12 edges to describe the boundary of the volume data cube.

[0144] 2.2) Constructing a local planar coordinate system: First, select an auxiliary vector that is not parallel to the plane normal vector. Then, calculate the horizontal unit basis vector of the plane containing the target slice by normalizing the cross product of the plane normal vector and the auxiliary vector, as shown in formula (1) above. Finally, calculate the vertical unit basis vector of the plane containing the target slice by cross product of the plane normal vector and the horizontal unit basis vector, as shown in formula (2) above. Through the above process, the orthogonal local basis vectors on the plane containing the target slice can be obtained.

[0145] 2.3) Calculate the geometric section between the slice plane and the physical boundary: First, define the infinite plane determined by the plane normal vector and the coordinate data of the positioning point as the target plane, and traverse each edge in the edge set. For each edge, determine whether the current edge intersects with the target plane. If so, obtain the intersection point between the current edge and the target plane and add it to the cross-sectional contour point set; otherwise, discard the current edge. Next, using the local orthogonal basis vector, project each intersection point in the cross-sectional contour point set onto the slice local coordinate system, as shown in the above formula (3). Afterward, based on the coordinate data of each intersection point in the slice local coordinate system, calculate the coordinates of all intersection points in the slice local coordinate system. The minimum and maximum values ​​on the axis are used to obtain the local coordinate system of the slice. The extreme range of the axes. Simultaneously, based on the coordinate data of each intersection point in the local coordinate system of the slice, the coordinates of all intersection points in the local coordinate system of the slice... The minimum and maximum values ​​on the axis are used to obtain the local coordinate system of the slice. The extreme range of the axis. Finally, based on the above extreme range and the preset slice sampling interval, the physical width and physical height of the target slice, as well as the pixel resolution required by the target slice in the slice local coordinate system, are determined, as shown in the above formula (4).

[0146] 2.4) Generate a spatial sampling point set within the target slice area: Within the boundary range of the slice local coordinate system, the target slice is divided into grids and scanned according to the pixel resolution in the slice local coordinate system. For each sampling point obtained by the sampling scan, we have: First, based on the extreme value range and the preset slice sampling interval, combined with the pixel index of the current sampling point, the two-dimensional local coordinate data of the current sampling point is determined, as shown in the above formula (5). Then, the two-dimensional local coordinate data of the current sampling point is transformed from the slice local coordinate system to the object three-dimensional coordinate system, thereby obtaining the final three-dimensional sampling coordinates of the current sampling point, that is, the three-dimensional coordinate data, as shown in the above formula (6). By traversing all the sampling points obtained by the sampling scan, the complete three-dimensional sampling point set of the target slice area in the object three-dimensional coordinate system can be obtained.

[0147] 3) FDK reconstruction based on sparse point sets:

[0148] 3.1) Geometric weighted correction of projection data: Since the cone beam X-rays are divergent, the original projection data needs to be geometrically corrected to obtain the corrected two-dimensional projection data, as shown in the above formula (7).

[0149] 3.2) Convolution filtering of projection data: Perform one-dimensional ramp filtering on each row of the corrected two-dimensional projection data to obtain the filtered two-dimensional projection data, as shown in the above formula (8).

[0150] 3.3) Weighted back projection of the target point set: For each sampling point in the three-dimensional sampling point set, we have: For each fixed projection angle, the three-dimensional coordinate data of the current sampling point is projected from the object's three-dimensional coordinate system to the rotated coordinate system to obtain the coordinate data of the current sampling point in the rotated coordinate system, i.e., the rotated coordinate data, as shown in the above formula (9). Subsequently, the rotated coordinate data of the current sampling point and the filtered two-dimensional projection data are substituted into the back projection integral formula of the FDK algorithm to obtain the reconstructed value of the current sampling point, as shown in the above formula (10).

[0151] 3.4) The reconstructed values ​​of each sampling point are assembled according to the two-dimensional pixel arrangement order of the target slice to output the reconstructed image of the target slice. This process does not require the generation of complete three-dimensional volume data.

[0152] Through the above process, this embodiment of the application directly utilizes weighted back projection to perform directional back projection on the pre-selected slice sampling point set. Compared to reconstructing the entire cube data, this method avoids redundant calculations for non-slice regions (invalid data), thereby significantly improving the reconstruction efficiency of slices at arbitrary angles while ensuring mathematical accuracy.

[0153] Furthermore, embodiments of this application also provide a rapid CT image reconstruction device, comprising:

[0154] The acquisition module is used to acquire the two-dimensional projection data and slicing parameters of the target object; wherein, the slicing parameters include the coordinate data of the positioning points of the target slice in the target object and the plane normal vector;

[0155] The first processing module is used to perform sampling processing based on the slice parameters to obtain a three-dimensional sampling point set of the target slice;

[0156] The second processing module is used to perform weighted back projection processing on the three-dimensional sampling point set based on the two-dimensional projection data to obtain the reconstructed image of the target slice.

[0157] The content of the above method embodiments is applicable to the device embodiments. The specific functions implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0158] Finally, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the above-described method for rapid reconstruction of CT images.

[0159] The content of the above method embodiments is applicable to the embodiments of this program product. The specific functions implemented by the embodiments of this program product are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0160] In summary, on the one hand, the embodiments of this application provide an adaptive reconstruction region determination method based on the size of the three-dimensional scanning field of view or volume data. Before constructing the three-dimensional sampling point set, the geometric intersection of the target slice and the preset three-dimensional reconstruction scene (usually a cube bounding box) is calculated. Then, by solving the intersection points of the target plane (the plane determined by the plane normal vector and the coordinate data of the positioning point) and the edge of the bounding box, these three-dimensional intersection points are projected onto the local coordinate system of the slice, thereby determining the physical width and height of the target slice.

[0161] On the other hand, this application provides a slice space coordinate construction and mapping mechanism based on normal vectors. It constructs local orthogonal basis vectors based on the plane normal vector and positioning point coordinate data of the target slice, and uses the above formula (6) to establish a one-to-one mapping relationship between the pixel index of each sampling point in the target slice and the three-dimensional coordinate system of the object. In this way, during the back projection process, it is only necessary to project each sampling point obtained by the above mapping onto the virtual detector projection data coordinate system for interpolation sampling to obtain the corresponding voxel value, and finally obtain the reconstructed image of the target slice.

[0162] On the other hand, embodiments of this application provide a direct reconstruction method for slices in any direction. After the projection data has undergone weighted correction and convolution filtering, it does not reconstruct the complete three-dimensional volume data. Instead, it directly constructs a spatial point set of the target slice plane, i.e., a three-dimensional sampling point set, based on user-defined slice parameters (location points and normal vectors). Weighted back projection calculation is then performed only on the coordinate points in this point set to directly obtain the reconstructed image of the target slice.

[0163] The core of this application's embodiments lies in changing the computational domain of the traditional FDK algorithm. The traditional FDK method is voxel-driven and must rely on a Cartesian network covering the entire 3D space; while the embodiments of this application are slice-driven. By establishing a direct mapping relationship between slice pixel coordinates and object 3D space coordinates, the calculation is performed only on the discrete point set of the slice plane during the backprojection stage. This method skips the intermediate process of generating 3D volume data and achieves end-to-end output from projection data to arbitrary slice images.

[0164] Furthermore, traditional FDK methods require voxel reconstruction of the entire 3D scanning field of view, with computational complexity increasing cubically with volume. In contrast, the embodiments of this application reduce the computational load of reconstruction to a two-dimensional planar level by locking onto specific slice regions of interest to the user, thereby increasing reconstruction speed several times over under the same hardware conditions.

[0165] In summary, the embodiments of this application realize the rapid reconstruction of slices in arbitrary directions in CT images, avoiding invalid calculations for non-interested regions, thereby effectively improving the reconstruction efficiency of CT images.

[0166] Although embodiments of this application have been shown and described, those skilled in the art will understand that various changes, modifications, substitutions and variations can be made to these embodiments without departing from the principles and spirit of this application, the scope of which is defined by the claims and their equivalents.

[0167] The above is a detailed description of the preferred embodiments of this application, but this application is not limited to the embodiments described. Those skilled in the art can make various equivalent modifications or substitutions without departing from the spirit of this application, and these equivalent modifications or substitutions are all included within the scope defined by the claims of this application.

Claims

1. A method for rapid reconstruction of CT images, characterized in that, include: Obtain the two-dimensional projection data and slicing parameters of the target object; wherein, the slicing parameters include the coordinate data of the positioning points of the target slice in the target object and the plane normal vector; The sampling process is performed based on the slice parameters to obtain a three-dimensional sampling point set of the target slice; The three-dimensional sampling point set is weighted and back-projected based on the two-dimensional projection data to obtain the reconstructed image of the target slice.

2. The method according to claim 1, characterized in that, The step of sampling based on the slice parameters to obtain the three-dimensional sampling point set of the target slice includes: Obtain the set of edges used to describe the physical boundaries of the volume data; Based on the slice parameters, basis vector operations are performed to obtain the local orthogonal basis vectors of the target slice; The target slice's width, height, and pixel resolution are obtained by performing localization processing based on the local orthogonal basis vectors and the edge set. Based on the width, height, and pixel resolution of the target slice, the target slice is sampled to obtain a three-dimensional sampling point set of the target slice.

3. The method according to claim 2, characterized in that, The step of obtaining the set of edges used to describe the physical boundaries of the volume data includes: Based on the dimensions, spatial starting position, and voxel spacing of the volume data of the target object, the physical range of the volume data of the target object in the object's three-dimensional coordinate system is obtained; Based on the physical range, determine the coordinate data of each vertex in the volume data of the target object; The edge set is constructed based on the coordinate data of each vertex.

4. The method according to claim 2, characterized in that, The local orthogonal basis vectors include horizontal unit basis vectors and vertical unit basis vectors; the step of performing basis vector operations based on the slice parameters to obtain the local orthogonal basis vectors of the target slice includes: Obtain the auxiliary vector of the target slice; wherein the auxiliary vector is not parallel to the plane normal vector; The auxiliary vector and the plane normal vector are subjected to cross product and normalization to obtain the horizontal unit basis vector; The vertical unit basis vector is obtained by performing a cross product on the horizontal unit basis vector and the plane normal vector.

5. The method according to claim 2, characterized in that, The step of performing positioning processing based on the local orthogonal basis vectors and the edge set to obtain the width, height, and pixel resolution of the target slice includes: For each edge in the edge set, if the current edge intersects with the target plane, the intersection point of the current edge and the target plane is obtained and added to the cross-sectional contour point set; otherwise, the current edge is discarded. The target plane is a plane determined by the plane normal vector and the coordinate data of the positioning point. Based on the local orthogonal basis vectors, each intersection point in the cross-sectional contour point set is projected onto the local coordinate system of the slice to obtain the coordinate data of each intersection point in the cross-sectional contour point set in the local coordinate system of the slice. Based on the coordinate data of each intersection point in the local coordinate system of the slice, determine the extreme value range of the local coordinate system of the slice; Based on the extreme value range and the preset slice sampling interval, the width and height of the target slice, as well as the pixel resolution of the target slice in the slice local coordinate system, are obtained.

6. The method according to claim 2, characterized in that, The step of sampling the target slice based on its width, height, and pixel resolution to obtain a three-dimensional sampling point set of the target slice includes: Within the extreme range of the local coordinate system of the slice, the target slice is subjected to grid sampling scanning based on the width, height and pixel resolution of the target slice; For each sampling point obtained by the sampling scan, based on the extreme value range and the preset slice sampling interval, and combined with the pixel index of the current sampling point, the two-dimensional local coordinate data of the current sampling point are determined; The two-dimensional local coordinate data of each sampling point are transformed from the local coordinate system of the slice to the three-dimensional coordinate system of the object to obtain the three-dimensional coordinate data of each sampling point, and then the three-dimensional sampling point set is constructed.

7. The method according to claim 1, characterized in that, The step of performing weighted backprojection processing on the three-dimensional sampling point set based on the two-dimensional projection data to obtain the reconstructed image of the target slice includes: The two-dimensional projection data is corrected to obtain the corrected two-dimensional projection data. The corrected two-dimensional projection data is filtered to obtain filtered two-dimensional projection data. The back-projection integral operation is performed on the filtered two-dimensional projection data and the three-dimensional sampling point set to obtain the reconstructed value of each sampling point in the three-dimensional sampling point set. The reconstructed image is obtained by assembling the reconstructed values ​​of each sampling point.

8. The method according to claim 7, characterized in that, The step of performing back-projection integration based on the filtered two-dimensional projection data and the three-dimensional sampling point set to obtain the reconstructed value of each sampling point in the three-dimensional sampling point set includes: The three-dimensional coordinate data of each sampling point are projected from the object's three-dimensional coordinate system to the rotating coordinate system to obtain the rotating coordinate data of each sampling point; Using a filtered back-projection algorithm and the filtered two-dimensional projection data, the rotation coordinate data of each sampling point are processed by back-projection integration to obtain the reconstructed value of each sampling point.

9. A rapid reconstruction device for CT images, characterized in that, include: The acquisition module is used to acquire two-dimensional projection data and slicing parameters of the target object; wherein, the slicing parameters include the coordinate data of the positioning points of the target slice in the target object and the plane normal vector; The first processing module is used to perform sampling processing based on the slice parameters to obtain a three-dimensional sampling point set of the target slice; The second processing module is used to perform weighted back projection processing on the three-dimensional sampling point set based on the two-dimensional projection data to obtain the reconstructed image of the target slice.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by the processor, it implements the rapid CT image reconstruction method according to any one of claims 1-8.

Images