CT reconstruction method and device, computer equipment, storage medium and computer program product

Abstract

The invention relates to a CT reconstruction method and device, computer equipment, a storage medium and a computer program product. The method comprises the following steps: acquiring a current reconstruction interval, and determining a reconstruction strategy according to the current reconstruction interval; under the condition that the reconstruction strategy is a first reconstruction strategy, performing CT reconstruction on the acquired projection data by adopting a coarse reconstruction mode when the projection data of preset interval duration is acquired to obtain a first CT reconstruction image, and displaying the first CT reconstruction image in a display interface in real time; performing CT reconstruction on the projection data by adopting a fine reconstruction mode to obtain a second CT reconstruction image, and storing the second CT reconstruction image; wherein the reconstruction cost of the coarse reconstruction mode is lower than that of the fine reconstruction mode. By adopting the method, the CT scanning effect and the reconstruction precision can be improved.

Description

Technical Field

[0001] The present application relates to the field of medical image processing technology, and in particular to a CT reconstruction method, apparatus, computer equipment, storage medium, and computer program product. Background Art

[0002] In the current spiral CT field, the reconstruction algorithm mainly adopts the WFBP (Weighted Filtered BackProjection) algorithm, which is an improvement based on the traditional Filtered BackProjection (FBP) algorithm.

[0003] FBP reconstructs images by back-projecting projection data onto the image plane, but this often produces artifacts and noise. The WFBP algorithm reduces noise and artifacts by applying weighted filtering. The weighting can be adjusted based on different angles or positions to adapt to image characteristics, thereby improving image quality.

[0004] However, the WFBP algorithm is computationally intensive, and the reconstruction time is generally longer than the acquisition time. That is, during the projection data acquisition process, the user cannot obtain the reconstructed image, resulting in the inability to timely know the status of the user currently undergoing CT scanning, which may lead to poor CT scanning results. Summary of the Invention

[0005] Based on this, it is necessary to provide a CT reconstruction method, apparatus, computer equipment, computer-readable storage medium and computer program product that can improve the efficiency and accuracy of CT reconstruction in order to address the above technical problems.

[0006] In a first aspect, the present application provides a CT reconstruction method applied to spiral CT, comprising:

[0007] Obtaining a current reconstruction interval, and determining a reconstruction strategy according to the current reconstruction interval;

[0008] When the reconstruction strategy is the first reconstruction strategy, each time projection data of a preset time interval is collected, CT reconstruction is performed on the collected projection data using a coarse reconstruction method to obtain a first CT reconstructed image, and the first CT reconstructed image is displayed in real time on the display interface;

[0009] Performing CT reconstruction on the projection data using a fine reconstruction method to obtain a second CT reconstructed image, and storing the second CT reconstructed image;

[0010] The reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method.

[0011] In one embodiment, determining the reconstruction strategy according to the current reconstruction interval includes:

[0012] When the current reconstruction interval is less than the interval threshold, determining the reconstruction strategy to be the first reconstruction strategy; and / or when the current reconstruction interval is greater than or equal to the interval threshold, determining the reconstruction strategy to be the second reconstruction strategy;

[0013] The first reconstruction strategy includes the coarse reconstruction method and the fine reconstruction method, and the second reconstruction strategy includes the fine reconstruction strategy.

[0014] In one embodiment, the method further comprises:

[0015] Before receiving the device preparation instruction and starting scanning to collect projection data, obtaining reconstruction parameters, pre-allocating video memory based on the reconstruction parameters, and reading the reconstruction parameters into the GPU;

[0016] The reconstruction parameters include at least one of parameter data, reconstruction layer number, horizontal motor movement starting position, horizontal motor speed, Q value weight curve, and windowing filter function.

[0017] In one embodiment, performing CT reconstruction on the collected projection data using a coarse reconstruction method to obtain a first CT reconstructed image includes:

[0018] performing filtering processing on the projection data to obtain filtered data;

[0019] Perform back-projection processing on the filtered data using a preset back-projection algorithm to obtain a back-projection image, wherein the preset back-projection algorithm includes any one of the following:

[0020] The reconstruction interval used in the back projection is a preset reconstruction interval, which is greater than the reconstruction interval used by the fine reconstruction algorithm, 360-degree interpolation is used for interpolation processing, the horizontal motor position is calculated based on the horizontal motor speed, and the layer thickness processing process is skipped.

[0021] In one embodiment, filtering the projection data to obtain filtered data includes:

[0022] The projection data are rearranged using a pre-established initial virtual coordinate system, wherein the initial virtual coordinate system has the detector illuminated by the central radiation source as the origin, the spiral line as the horizontal axis, and the horizontal displacement bed direction as the vertical axis;

[0023] The rearranged projection data is filtered according to the windowing filter kernel to obtain filtered data.

[0024] In one embodiment, rearranging the projection data using a pre-established initial virtual coordinate system includes:

[0025] Rearranging the projection data of the fan beam into the projection data of the parallel beam by rearranging the direction angle of the projection data;

[0026] According to the amount of up-sampled data and the interval of the direction angle rearrangement, the projection data of the parallel beam are subjected to three-dimensional Hermite interpolation to obtain a three-dimensional surface, and the three-dimensional surface is subjected to coordinate interpolation processing perpendicular to the initial virtual coordinate system to obtain rearranged projection data.

[0027] In one embodiment, performing coordinate interpolation processing perpendicular to the initial virtual coordinate system on the three-dimensional curved surface to obtain rearranged projection data includes:

[0028] Determining grid points on a three-dimensional surface and calculating a rotation matrix for each of the grid points;

[0029] Using the rotation matrix, the coordinates of each grid point are transformed from the initial virtual coordinate system to a target virtual coordinate system, where the horizontal axis of the target virtual coordinate system is horizontal;

[0030] Interpolation processing is performed on each of the grid points in the target virtual coordinate system to obtain rearranged projection data.

[0031] In one embodiment, performing interpolation processing on each of the grid points in the target virtual coordinate system to obtain rearranged projection data includes:

[0032] Performing interpolation processing on each of the grid points in the target virtual coordinate system to obtain an interpolation result;

[0033] The interpolation result is filtered according to the initial virtual coordinate system to obtain rearranged projection data.

[0034] In one embodiment, the method further comprises:

[0035] storing the filtered data in a video memory;

[0036] The step of performing CT reconstruction on the projection data in a fine reconstruction manner to obtain a second CT reconstructed image includes:

[0037] Acquiring filtered data from the video memory, performing back-projection processing on the filtered data to obtain a back-projection image, and extracting bone tissue data from the back-projection image;

[0038] Performing orthographic projection on the bone tissue data to obtain bone projection data;

[0039] taking the difference between the projection data and the bone projection data as bone-removed tissue data;

[0040] A second CT reconstructed image is obtained according to the bone-removed tissue data and the back-projected image.

[0041] In one embodiment, extracting bone tissue data from the back-projection image includes:

[0042] Extracting bone tissue pixels in the back-projection image, wherein the CT values ​​corresponding to the bone tissue pixels are within the range of bone tissue CT values;

[0043] For any bone tissue pixel point, determine the bone tissue segment corresponding to the bone tissue pixel point based on the CT value of the bone tissue pixel point, and determine the normalization coefficient of the bone tissue pixel point based on the upper and lower CT values ​​of the bone tissue segment, the normalization coefficients corresponding to the upper and lower CT values, and the CT value of the bone tissue pixel point, where the bone tissue CT value range is composed of multiple bone tissue segments;

[0044] Bone tissue data is obtained based on the normalization coefficient and CT value of each bone tissue pixel point.

[0045] In one embodiment, performing orthographic projection on the bone tissue data to obtain bone projection data includes:

[0046] According to each pixel point in the axial image corresponding to the bone tissue data and the ray length weight corresponding to each pixel point, the CT value of each pixel point is weightedly superimposed to obtain bone projection data;

[0047] The ray length weight is negatively correlated with the length of the ray within the voxel.

[0048] In one embodiment, obtaining a second CT reconstructed image based on the bone-removed tissue data and the back-projected image includes:

[0049] The ratio of the bone-removed tissue data to the water reference is used as the equivalent length of each projection pixel point in the bone-removed tissue data relative to water;

[0050] Performing weighted correction on the bone projection data using an artifact correction coefficient corresponding to the equivalent length to obtain corrected bone projection data;

[0051] After reconstructing the corrected bone projection data, the reconstruction result is superimposed on the back-projection image to obtain a second CT reconstructed image.

[0052] In a second aspect, the present application further provides a CT reconstruction device, comprising:

[0053] a determination module, configured to obtain a current reconstruction interval and determine a reconstruction strategy according to the current reconstruction interval;

[0054] a coarse reconstruction module, configured to, when the reconstruction strategy is the first reconstruction strategy, perform CT reconstruction on the collected projection data using a coarse reconstruction method each time projection data of a preset time interval is collected, to obtain a first CT reconstructed image, and to display the first CT reconstructed image in real time on a display interface;

[0055] a fine reconstruction module, configured to perform CT reconstruction on the projection data using a fine reconstruction method to obtain a second CT reconstructed image, and store the second CT reconstructed image;

[0056] The reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method.

[0057] In one embodiment, the determining module is further configured to:

[0058] When the current reconstruction interval is less than the interval threshold, determining the reconstruction strategy to be the first reconstruction strategy; and / or when the current reconstruction interval is greater than or equal to the interval threshold, determining the reconstruction strategy to be the second reconstruction strategy;

[0059] The first reconstruction strategy includes the coarse reconstruction method and the fine reconstruction method, and the second reconstruction strategy includes the fine reconstruction strategy.

[0060] In one embodiment, the apparatus further comprises:

[0061] A pre-reconstruction module is used to obtain reconstruction parameters before receiving the device preparation instruction and starting scanning to collect projection data, pre-allocate video memory based on the reconstruction parameters, and read the reconstruction parameters into the GPU;

[0062] The reconstruction parameters include at least one of parameter data, reconstruction layer number, horizontal motor movement starting position, horizontal motor speed, Q value weight curve, and windowing filter function.

[0063] In one embodiment, the coarse reconstruction module is further configured to:

[0064] performing filtering processing on the projection data to obtain filtered data;

[0065] Perform back-projection processing on the filtered data using a preset back-projection algorithm to obtain a back-projection image, wherein the preset back-projection algorithm includes any one of the following:

[0066] The reconstruction interval used in the back projection is a preset reconstruction interval, which is greater than the reconstruction interval used by the fine reconstruction algorithm, 360-degree interpolation is used for interpolation processing, the horizontal motor position is calculated based on the horizontal motor speed, and the layer thickness processing process is skipped.

[0067] In one embodiment, the coarse reconstruction module is further configured to:

[0068] The projection data are rearranged using a pre-established initial virtual coordinate system, wherein the initial virtual coordinate system has the detector illuminated by the central radiation source as the origin, the spiral line as the horizontal axis, and the horizontal displacement bed direction as the vertical axis;

[0069] The rearranged projection data is filtered according to the windowing filter kernel to obtain filtered data.

[0070] In one embodiment, the coarse reconstruction module is further configured to:

[0071] Rearranging the projection data of the fan beam into the projection data of the parallel beam by rearranging the direction angle of the projection data;

[0072] According to the amount of up-sampled data and the interval of the direction angle rearrangement, the projection data of the parallel beam are subjected to three-dimensional Hermite interpolation to obtain a three-dimensional surface, and the three-dimensional surface is subjected to coordinate interpolation processing perpendicular to the initial virtual coordinate system to obtain rearranged projection data.

[0073] In one embodiment, the coarse reconstruction module is further configured to:

[0074] Determining grid points on a three-dimensional surface and calculating a rotation matrix for each of the grid points;

[0075] Using the rotation matrix, the coordinates of each grid point are transformed from the initial virtual coordinate system to a target virtual coordinate system, where the horizontal axis of the target virtual coordinate system is horizontal;

[0076] Interpolation processing is performed on each of the grid points in the target virtual coordinate system to obtain rearranged projection data.

[0077] In one embodiment, the coarse reconstruction module is further configured to:

[0078] Performing interpolation processing on each of the grid points in the target virtual coordinate system to obtain an interpolation result;

[0079] The interpolation result is filtered according to the initial virtual coordinate system to obtain rearranged projection data.

[0080] In one embodiment, the apparatus further comprises:

[0081] A storage module, configured to store the filtered data in a video memory;

[0082] The fine reconstruction module is specifically used to:

[0083] Acquiring filtered data from the video memory, performing back-projection processing on the filtered data to obtain a back-projection image, and extracting bone tissue data from the back-projection image;

[0084] Performing orthographic projection on the bone tissue data to obtain bone projection data;

[0085] taking the difference between the projection data and the bone projection data as bone-removed tissue data;

[0086] A second CT reconstructed image is obtained according to the bone-removed tissue data and the back-projected image.

[0087] In one embodiment, the fine reconstruction module is specifically configured to:

[0088] Extracting bone tissue pixels in the back-projection image, wherein the CT values ​​corresponding to the bone tissue pixels are within the range of bone tissue CT values;

[0089] For any bone tissue pixel point, determine the bone tissue segment corresponding to the bone tissue pixel point based on the CT value of the bone tissue pixel point, and determine the normalization coefficient of the bone tissue pixel point based on the upper and lower CT values ​​of the bone tissue segment, the normalization coefficients corresponding to the upper and lower CT values, and the CT value of the bone tissue pixel point, where the bone tissue CT value range is composed of multiple bone tissue segments;

[0090] Bone tissue data is obtained based on the normalization coefficient and CT value of each bone tissue pixel point.

[0091] In one embodiment, the fine reconstruction module is specifically configured to:

[0092] According to each pixel point in the axial image corresponding to the bone tissue data and the ray length weight corresponding to each pixel point, the CT value of each pixel point is weightedly superimposed to obtain bone projection data;

[0093] The ray length weight is negatively correlated with the length of the ray within the voxel.

[0094] In one embodiment, the fine reconstruction module is specifically configured to:

[0095] The ratio of the bone-removed tissue data to the water reference is used as the equivalent length of each projection pixel point in the bone-removed tissue data relative to water;

[0096] Performing weighted correction on the bone projection data using an artifact correction coefficient corresponding to the equivalent length to obtain corrected bone projection data;

[0097] After reconstructing the corrected bone projection data, the reconstruction result is superimposed on the back-projection image to obtain a second CT reconstructed image.

[0098] In a third aspect, the present application further provides a computer device comprising a memory and a processor, wherein the memory stores a computer program, and the processor implements any of the above CT reconstruction methods when executing the computer program.

[0099] In a fourth aspect, the present application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program implements any of the above CT reconstruction methods when executed by a processor.

[0100] In a fifth aspect, the present application also provides a computer program product, comprising a computer program, which implements any of the above CT reconstruction methods when executed by a processor.

[0101] The above-mentioned CT reconstruction method, apparatus, computer device, storage medium and computer program product can determine a reconstruction strategy based on the current reconstruction interval. When the reconstruction strategy is the first reconstruction strategy, each time projection data of a preset interval length is collected, a coarse reconstruction method is used to perform CT reconstruction on the collected projection data to obtain a first CT reconstructed image, and the first CT reconstructed image is displayed in real time on a display interface. The projection data is then continuously CT reconstructed using a fine reconstruction method to obtain a second CT reconstructed image, and the second CT reconstructed image is stored. The reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method. By using the CT reconstruction method, apparatus, computer device, storage medium, and computer program product provided in the embodiments of the present application, a coarse reconstruction method with low reconstruction cost and less time consumption can be used to perform CT reconstruction and display the reconstructed image in real time during the acquisition process, and a fine reconstruction method with higher reconstruction cost but higher reconstruction accuracy can be used to perform high-precision CT reconstruction. High-precision CT reconstructed images are obtained and stored. This allows users to not only obtain high-precision CT reconstructed images, but also obtain corresponding reconstruction results in real time during the scanning and acquisition of projection data. Based on the reconstruction results, the status of the user currently performing the CT scan can be determined, such as whether the user has moved or whether the user's position is appropriate. Based on the user's status, corresponding judgment operations can be performed, thereby improving CT scanning effects and reconstruction accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0102] In order to more clearly illustrate the technical solutions in the embodiments of the present application or related technologies, the following briefly introduces the drawings required for use in the embodiments or related technical descriptions. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0103] Figure 1 1 is a flow chart of a CT reconstruction method according to an embodiment;

[0104] Figure 2 is a schematic diagram of an interface in an embodiment;

[0105] Figure 3 Schematic diagram of a Q value weight curve in one embodiment;

[0106] Figure 4 is a schematic diagram of a windowed filter function in one embodiment;

[0107] Figure 5a is a schematic diagram of an interface in an embodiment;

[0108] Figure 5b is a schematic diagram of an interface in an embodiment;

[0109] Figure 6 is a schematic diagram of a CT reconstruction method in one embodiment;

[0110] Figure 7 1 is a flow chart of a CT reconstruction method according to an embodiment;

[0111] Figure 8 is a schematic diagram of a CT reconstruction method in one embodiment;

[0112] Figure 9 is a schematic diagram of channel spacing in one embodiment;

[0113] Figure 10 is a schematic diagram of a three-dimensional curved surface in one embodiment;

[0114] Figure 11 is a schematic diagram of a CT reconstruction method in one embodiment;

[0115] Figure 12 FIG. 7 is a flow chart of step 702 in one embodiment;

[0116] Figure 13 1 is a flow chart of step 1206 in one embodiment;

[0117] Figure 14 1 is a flow chart of a CT reconstruction method according to an embodiment;

[0118] Figure 15 A schematic diagram of a forward projection path in one embodiment is shown;

[0119] Figure 16a The relationship between the ray and the axial image pixels in the orthographic projection is shown;

[0120] Figure 16b The relationship between the ray and the voxel diagonal within a voxel of the target reconstructed object is shown;

[0121] Figure 17 1 is a flow chart of a CT reconstruction method according to an embodiment;

[0122] Figure 18 is a flow chart of step 1408 in one embodiment;

[0123] Figure 19a is a schematic diagram of a CT reconstruction method in one embodiment;

[0124] Figure 19b is another schematic diagram of a CT reconstruction method in one embodiment;

[0125] Figure 20 1402 is a flowchart of step 1402 in one embodiment;

[0126] Figure 21 is a schematic diagram of a virtual horizontal motor position in one embodiment;

[0127] Figure 22 FIG. 2 is a flow chart of step 2004 in one embodiment;

[0128] Figure 23a is a schematic diagram of a CT reconstruction method in one embodiment;

[0129] Figure 23b is another schematic diagram of a CT reconstruction method in one embodiment;

[0130] Figure 24 is a structural block diagram of a CT reconstruction device in one embodiment;

[0131] Figure 25 is a schematic diagram of a CT reconstruction method in one embodiment;

[0132] Figure 26 is a schematic diagram of a CT reconstruction method in one embodiment;

[0133] Figure 27 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION

[0134] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.

[0135] In one embodiment, Figure 1 As shown, a CT reconstruction method is provided. This embodiment uses the method applied to a terminal as an example for illustration. It is understandable that the method can also be applied to a server, or to a system including a terminal and a server, and implemented through interaction between the terminal and the server. In this embodiment, the method includes the following steps 102 to 106, wherein:

[0136] Step 102: obtaining a current reconstruction interval and determining a reconstruction strategy according to the current reconstruction interval;

[0137] In this embodiment of the present application, if the user has not set the current reconstruction interval, the current reconstruction interval may be a preset interval. After the user sets the current reconstruction interval, the current reconstruction interval becomes the user-set interval. The reconstruction interval and the computational complexity of CT reconstruction are inversely correlated, meaning that a larger reconstruction interval results in a smaller computational complexity for CT reconstruction, and conversely, a smaller reconstruction interval results in a larger computational complexity for CT reconstruction.

[0138] In an embodiment of the present application, two reconstruction strategies can be provided: a first reconstruction strategy and a second reconstruction strategy, wherein the first reconstruction strategy includes a coarse reconstruction method and a fine reconstruction method, and the second reconstruction strategy includes a fine reconstruction method, wherein the reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method, and the reconstruction cost may include: reconstruction time, reconstruction calculation amount, reconstruction accuracy, etc.

[0139] In an exemplary embodiment, determining a reconstruction strategy according to a current reconstruction interval may include:

[0140] When the current reconstruction interval is less than the interval threshold, the reconstruction strategy is determined to be the first reconstruction strategy; or when the current reconstruction interval is greater than or equal to the interval threshold, the reconstruction strategy is determined to be the second reconstruction strategy.

[0141] That is, the interval threshold is a preset interval value. When the current reconstruction interval is less than the interval threshold, fine reconstruction takes a long time and cannot be displayed in real time. The reconstruction strategy at this time is determined to be a first reconstruction strategy including a coarse reconstruction method and a fine reconstruction method. The first CT reconstructed image is reconstructed by the coarse reconstruction method for real-time display, and the second CT reconstructed image is reconstructed by the fine reconstruction method for subsequent diagnosis and other processes. When the current reconstruction interval is greater than or equal to the interval threshold, the reconstruction time of the fine reconstruction method can meet the user's real-time display requirements, and no additional coarse reconstruction is required. Therefore, the reconstruction strategy at this time is determined to be the second reconstruction strategy including the fine reconstruction method, which can save computing resources.

[0142] For example, the CT doctor will generally select the corresponding reconstruction protocol according to the imaging part (head, chest, etc.) and the required reconstruction effect (high spatial resolution, high and low contrast resolution, etc.). These reconstruction protocols include reconstruction parameter information, such as reconstruction interval. The doctor can default to the reconstruction interval, or can adjust the reconstruction interval. For example: the reconstruction interval used in the coarse reconstruction method is the interval threshold, which is 5mm by default. If the reconstruction interval is set to be greater than or equal to 5mm, the fine reconstruction method is directly used for CT reconstruction. Its real-time display speed can also meet user needs. At this time, the reconstruction strategy is determined to be the second reconstruction strategy. If the CT doctor adjusts the reconstruction interval carried in the protocol and the reconstruction interval is less than 5mm, the reconstruction calculation amount is large at this time, and the fine reconstruction speed may not meet the requirements of real-time reconstruction. At this time, the reconstruction strategy is determined to be the first reconstruction strategy, and the real-time reconstruction results are obtained through coarse reconstruction, thereby ensuring that the doctor has a real-time display result for easy judgment.

[0143] In step 104, when the reconstruction strategy is the first reconstruction strategy, each time projection data of a preset interval is collected, the collected projection data is CT reconstructed using a coarse reconstruction method to obtain a first CT reconstructed image, and the first CT reconstructed image is displayed in real time on the display interface.

[0144] In the embodiment of the present application, when the reconstruction strategy is the first reconstruction strategy, projection data can be collected according to the preset interval length, that is, when the time length of collecting projection data reaches the preset interval length, the coarse reconstruction method is used to perform CT reconstruction on the projection data currently collected within the preset interval length to obtain a first CT reconstructed image, and the first CT reconstructed image is displayed in real time in the display interface. The display interface refers to Figure 2 It should be noted that, while reconstructing, the projection data of the next preset interval is continuously collected for the next CT reconstruction. The specific reconstruction process of the coarse reconstruction method will be described in detail in the subsequent embodiments and will not be repeated here in the embodiments of this application.

[0145] The coarse reconstruction process physically divides the original acquired projection data according to the acquisition time to ensure that the reconstruction time is less than the acquisition time, where acquisition and reconstruction are two independently running threads.

[0146] Step 106 : Perform CT reconstruction on the projection data using a fine reconstruction method to obtain a second CT reconstructed image, and store the second CT reconstructed image.

[0147] In the embodiments of the present application, a fine reconstruction method can be used to perform more accurate CT reconstruction of the projection data, thereby storing the reconstructed second CT reconstructed image in a database for subsequent access, printing, and review. The specific implementation process of the fine reconstruction method will be described in detail in the following embodiments and will not be repeated in detail in the embodiments of the present application.

[0148] In an exemplary embodiment, when the current reconstruction strategy is the second reconstruction strategy, the collected projection data can be directly CT reconstructed through fine reconstruction, and the reconstructed second CT reconstructed image can be displayed in real time on the display interface, and stored in the database for subsequent call, printing, reference, etc.

[0149] By using the CT reconstruction method provided in the embodiments of the present application, a coarse reconstruction method with low reconstruction cost and less time consumption can be used to perform CT reconstruction and display the reconstructed image in real time during the acquisition process, and a fine reconstruction method with higher reconstruction cost but higher reconstruction accuracy can be used to perform high-precision CT reconstruction. High-precision CT reconstructed images are obtained and stored, so that users can not only obtain high-precision CT reconstructed images, but also obtain corresponding reconstruction results in real time during the scanning and acquisition of projection data. Based on the reconstruction results, the status of the user currently performing the CT scan can be determined, such as whether movement has occurred or whether the user's position is appropriate. Then, corresponding judgment operations can be performed based on the user's status, thereby improving the CT scanning effect and reconstruction accuracy.

[0150] In an exemplary embodiment, the method further includes:

[0151] After receiving the device preparation instruction and before starting scanning to collect projection data, the reconstruction parameters are obtained, the video memory is pre-allocated based on the reconstruction parameters, and the reconstruction parameters are read into the GPU;

[0152] The reconstruction parameters include at least one of parameter data, the number of reconstruction slices, the horizontal motor's starting position, the horizontal motor's speed, a Q-value weighting curve, and a windowed filter function. The horizontal motor drives the bed, enabling horizontal relative motion between the patient and the CT device. During a spiral scan, the CT device rotates, and the horizontal motor drives the patient to move horizontally relative to the CT device along the axis of the rotation, achieving the spiral scan.

[0153] In an embodiment of the present application, a pre-reconstruction process may be included before adopting a reconstruction strategy for CT reconstruction. The pre-reconstruction process may include: obtaining reconstruction parameters, pre-allocating video memory based on the reconstruction parameters, and reading the reconstruction parameters into the GPU from the time when the device preparation instruction is received to the time when scanning and projection data acquisition begins.

[0154] For example, a pre-set duration preparation phase occurs between receiving a preparation instruction and starting scanning and projection data acquisition. This preparation phase is called an exposure delay phase. During this preparation phase, reconstruction parameters can be pre-acquired, video memory can be allocated based on the reconstruction parameters, and the reconstruction parameters required for the reconstruction process can be read into the GPU.

[0155] Among them, the video memory pre-opened based on the reconstruction parameters can include space for storing the following data: tomographic images after layer thickness superposition reconstruction, tomographic images after back projection, total projection data images, detector geometric position parameter index table (X, Y, Z of each detector point, in a virtual coordinate system), actual horizontal motor correction parameter index table (the position of the patient support, that is, the bed, in each frame of the scan, that is, the Z-direction value, in mm), uniform horizontal motor correction parameter index table (used for coarse reconstruction, the average calculated based on the pitch, collimation width, etc.), air data image (storing air images after dark field correction), angle data of each frame of rays relative to the central ray (in a virtual coordinate system), coordinate value distance data of each frame of rays in the longitudinal axis (in a virtual coordinate system), directional angle interpolation data (coordinate value distance data of each ray in the longitudinal axis after directional angle rearrangement), rearranged image, filter kernel data, filtered image, water uniformity correction parameter index table, water CT value correction parameter index table, bone sclerosis correction parameter index table, detector odd-even frame correction ratio parameter index table, segmented projection data image, etc.

[0156] In an embodiment of the present application, by allocating and initializing the video memory in advance, it is possible to avoid allocating and releasing the video memory during the reconstruction process (for example, real-time reconstruction of spiral scans requires multiple calls to the reconstruction module, and it is very time-consuming to allocate and release the video memory each time. The embodiment of the present application allocates the video memory in advance, which can avoid allocating and releasing the video memory each time), which can greatly reduce the reconstruction time.

[0157] The parameter data can include parameters read from the parameter file, including air correction data, detector position data, inconsistency parameters, water correction parameters, and bone sclerosis parameters. In this example, the parameter data is pre-loaded into the GPU to avoid reading from the hard disk and switching from the CPU to the GPU during the reconstruction process, significantly reducing reconstruction time and improving reconstruction efficiency.

[0158] In one example, the number of reconstruction layers may include the number of fine reconstruction layers and the number of coarse reconstruction layers, wherein the number of fine reconstruction layers may be calculated based on the reconstruction parameters, including: the number of fine reconstruction layers num_bp is calculated based on the scan length scanlenth, the collimation angle colangle, the collimation width colwidth, the extended collimation width ecol, the pitch pitch, the reconstruction interval inc, and the display layer spacing thk. Among them, the reconstruction interval of the coarse reconstruction is pre-designed, and the reconstruction interval of the fine reconstruction is selected by the user interface during shooting (or the user selects which protocol, and the protocol defaults). The collimation angle, collimation width, and extended collimation width are set before the image is acquired. The collimation angle and collimation width are determined by the machine mechanics and are related to the mechanical design. The extended collimation width is set according to the empirical value.

[0159] The process of determining the number of layers for fine reconstruction is shown in the following formula (1):

[0160]

[0161] Since there is a certain proportional relationship between the number of fine reconstruction layers and the number of coarse reconstruction layers, the number of coarse reconstruction layers can be further determined based on the proportional relationship. The process of determining the number of coarse reconstruction layers num_rough is shown in the following formula (II):

[0162] num_rough=(num_bp*inc-inc+5) / 5 formula (2)

[0163] Here, 5 is the reconstruction interval of 5 mm for coarse reconstruction, and may be replaced by other values ​​when other reconstruction intervals are used for coarse reconstruction.

[0164] The number of reconstructed images after fine reconstruction layer thickness processing can be further determined based on the fine reconstruction layer number, reconstruction interval, and display layer spacing, as shown in formula (3):

[0165]

[0166] In one example, the user selects the shooting area through the positioning film, and the starting point of the scan needs to be moved to the starting point selected by the user during the spiral scan. Therefore, the position can be selected based on the positioning film (the positioning film usually shows the geometric layout of the entire scanning area to help the operator and the doctor determine the specific position and range of the scan. The starting position of the scan is pre-defined on the positioning film, which is called the initial position of the positioning film. The operator will use the mouse or other tools to select an area on the positioning film. This selected area is usually to determine the region of interest (ROI) for further CT scanning. The selected position can help ensure that the scanner acquires data at the correct position and angle. The selected position also includes a starting position for the scan, which is called the initial position selected in the positioning film. For example: when scanning the chest, the positioning film may show the outline of the entire chest. The radiologist will select the lung area so that the CT scanner can focus on acquiring data from the lungs), determine the starting position of the horizontal motor movement, and move the displacement bed according to the starting position of the horizontal motor movement. The process of determining the starting position of the horizontal motor movement includes:

[0167] The initial horizontal motor position startcouphpos, the distance scoutdis from the initial position selected in the positioning piece to the initial position of the positioning piece, the collimation angle colangle, the collimation width colwidth, the extended collimation width ecol, and the pitch pitch can be obtained, and the starting position startpos of the horizontal motor movement can be calculated. Among them, the initial horizontal motor position is the mechanical position of the displacement bed, that is, the zero point position of the displacement bed. The starting position of the horizontal motor movement is the position of the horizontal motor when scanning starts from the starting position selected in the positioning piece. The specific calculation process can be referred to the following formula (4) or formula (5):

[0168]

[0169]

[0170] In the scheme corresponding to formula (4), the moving direction of the predefined horizontal motor is opposite to the V-axis of the initial virtual coordinate system, and the opposite concept is reflected by the negative value; in the scheme corresponding to formula (5), the moving direction of the predefined horizontal motor is consistent with the V-axis of the initial virtual coordinate system.

[0171] In one example, the horizontal motor speed (used for coarse reconstruction) can be calculated based on the screw pitch, which is defined as the ratio of the bed's travel distance per X-ray tube rotation to the collimation width. This reverse calculation can be used to determine the horizontal motor speed. This eliminates the need for video memory allocation and the need to retrieve and calculate reconstruction parameters from a parameter file during the reconstruction process, effectively improving reconstruction efficiency and reducing the amount of computation required. This lowers the performance requirements for the device and enhances applicability.

[0172] In one example, the Q-value weight curve can be expressed as shown in the following formula (6):

[0173]

[0174] Among them, W Q (q) is the Q value weight, q is the normalized corresponding value of the layer direction (row number), and Q is a threshold (within 1). For example, Figure 3 A schematic diagram of the Q value weight curve is shown. Assuming Q is 0.6, Figure 3 It can be seen that W Q (q) weight is the vertical axis, q is the horizontal axis, q is the normalized value according to the row direction, that is, the V direction. For example, if there are 32 rows in the V direction, then the center is the point corresponding to the 16.5th row. Figure 3 The horizontal 0 point is the position of q=0, and the 1st and 32nd rows are the positions of q=-1 and q-1.

[0175] In the embodiment of the present application, GPU texture binding can be performed in the pre-reconstruction stage, that is, the Q value weight curve will be stored in the texture memory of the GPU in advance. Texture memory is a hardware area on the graphics card, and the graphics card has its own processing for texture memory. For example: if integer data is currently stored, then for a one-dimensional texture, if the Q value weight is 0.8, the Q value weight also needs to be interpolated, and the data used for interpolation (such as 0.7 and 0.9) must be imported from the memory to the video memory before interpolation. These processes are relatively time-consuming. However, in the embodiment of the present application, texture binding of the Q value weight is performed in advance, and the Q value weight 0.8 can be directly corresponded from the Q value weight curve bound in the texture memory, which can greatly reduce the reconstruction time.

[0176] In one example, the windowing filter includes an all-pass frequency and a cutoff frequency, and the windowing filter function is shown in the following formula (VII):

[0177]

[0178] Among them, w represents the frequency component of the signal, wl represents the all-pass frequency, and wh represents the cutoff frequency.

[0179] In one example, the window function can be a sinc window, which has clearer details at the same cutoff frequency than the hanning window, which is too smooth. Take two limits, low frequency (full-pass frequency) and high frequency (cutoff frequency). When the frequency component of the projection data is less than or equal to the full-pass frequency, the window value of the corresponding windowed filter kernel can be determined to be 1; or, when the frequency component of the projection data is greater than the cutoff frequency, the window value of the corresponding windowed filter kernel can be determined to be 0; or, when the frequency component of the projection data is greater than the full-pass frequency and less than the cutoff frequency, the window value of the corresponding windowed filter kernel is

[0180] In one example, referring to Figure 4 As shown, the positions of the full-pass frequency and the cutoff frequency are shown, the horizontal axis is the sampling point (positive and negative normalized, that is, the center is the center of the image sampling width), and the vertical axis is the window function value. Figure 4 Where WL is the full-pass frequency (for example, it can take values ​​of 1, 1 / 16, 1 / 32), WH is the cutoff frequency (for example, it can take values ​​of 1, 2 / 3, 1 / 2), WL<=WH; the horizontal axis is the sampling width of the convolution (the value of 0 represents the sampling center), and the vertical axis is the normalized window width value.

[0181] In one example, according to the requirements of reconstructing the image, the full-pass frequency and the cutoff frequency can take different values ​​and be divided into multiple groups, for example, 7 groups, corresponding to different sharpening and smoothing effects. One group can be selected from the multiple groups to calculate the windowing filter kernel according to the needs, and the projection data can be subjected to windowing filtering processing.

[0182] It should be noted that in the pre-reconstruction stage, users can also customize reconstruction parameters, including customizing horizontal motor parameters and position parameters based on pitch.

[0183] In one example, after the positioning film is taken and the spiral scanning parameters are set, you can click on the device preparation interface (refer to Figure 5a As shown), enter the delayed exposure interface (refer to Figure 5b As shown, Figure 5b The left image in the middle is the positioning film, and the right image is the spiral scan image (also known as the reconstructed image). After the X-ray shooting button is pressed, the first thing is the delayed exposure, at which time the right image is completely black. After the equipment is ready, the doctor has time to check, such as to confirm whether the patient has moved, etc., and the equipment is ready to pop up. Figure 5bIn the delayed exposure interface shown, the doctor starts the delayed exposure only after clicking on the shooting button, and starts the pre-reconstruction process after the delay countdown ends. The actual pre-reconstruction process can also be executed after the ray acquisition button is pressed, or it can be completed during the equipment preparation stage. In the embodiment of the present application, there is no specific limitation on the timing of pre-reconstruction, as long as it is completed before entering the reconstruction process. In this way, the pre-reconstruction is completed before the projection data is collected, which does not take up the acquisition and reconstruction time, and the user is unaware of it, which can greatly improve the reconstruction efficiency.

[0184] In an exemplary embodiment, performing CT reconstruction on the collected projection data using a coarse reconstruction method to obtain a first CT reconstructed image includes:

[0185] Performing filtering on the projection data to obtain filtered data;

[0186] Back-projection processing is performed on the filtered data using a preset back-projection algorithm to obtain a back-projection image, where the preset back-projection algorithm includes any one of the following:

[0187] The reconstruction interval used in the back projection is a preset reconstruction interval, which is greater than the reconstruction interval used by the fine reconstruction algorithm, 360-degree interpolation is used for interpolation processing, the horizontal motor position is calculated based on the horizontal motor speed, and the layer thickness processing process is skipped.

[0188] In embodiments of the present application, during the coarse reconstruction process, the projection data can be filtered to remove noise and improve the CT reconstruction effect. For example, filtering algorithms such as Laplace filtering and Hamming window filtering can be used to filter the projection data. For example, the projection data is information about the intensity of X-rays after penetrating an object at multiple angles. First, a filter is selected, such as a Hamming window filter. The Hamming window filter can smooth the data in the frequency domain and reduce artifacts. Then, the projection data can be Fourier transformed to convert the projection data to the frequency domain. The Fourier transform can decompose the signal into different frequency components, facilitating filtering. Applying the selected filter in the frequency domain includes multiplying the Fourier transform result of the projection data by the frequency response of the filter, thereby suppressing unnecessary frequency components (such as noise and artifacts), selectively retaining low-frequency components, and removing high-frequency components. Finally, an inverse Fourier transform is performed on the filtered frequency domain data to convert it back to the time domain (or spatial domain) to obtain the filtered data (filtered projection data). Illustratively, after obtaining the filtered data, the filtered data may be further processed, such as normalized, smoothed, etc., to improve image quality and usability.

[0189] Through the above filtering process, not only the noise and artifacts are removed, but also the image contrast is improved and the edge features are enhanced, providing better quality input data for subsequent CT image reconstruction.

[0190] Back projection is the process of back-projecting each projection data into the image space according to its imaging angle. The specific operation is to back-project each projection data into the two-dimensional image along the angle at which it was collected. Exemplarily, for the filtered data at any angle, each pixel value of the filtered data can be accumulated along the angle to gradually construct a back-projection image. The filtered data of all angles are accumulated in the back-projection process to form a back-projection image, which usually contains mixed information of various tissues. In fact, weighted back-projection methods (such as normalizing the filter values) can also be used in the accumulation process to improve the back-projection quality.

[0191] Since the purpose of coarse reconstruction is only to display the approximate shooting position, there is no need for high-precision reconstruction. Therefore, a preset thicker reconstruction interval can be used for back projection, which can greatly reduce the reconstruction calculation amount and improve reconstruction efficiency. In addition, the use of 360-degree interpolation instead of 180-degree interpolation in the back projection process can reduce the weight calculation by half, further reducing the reconstruction calculation amount and improving reconstruction efficiency. The horizontal motor position corresponding to each frame can be replaced by the average value calculated based on the pitch, that is, the horizontal motor position can be calculated based on the horizontal motor speed pre-written in the GPU. There is no need to read the real-time horizontal motor position in each frame of projection data, which can greatly reduce the reconstruction time. Finally, it should be noted that the coarse reconstruction method does not have layer thickness superposition calculations, and can directly output the reconstruction results, further simplifying the reconstruction process, which can effectively improve the reconstruction efficiency and achieve the effect of real-time display of the reconstructed image.

[0192] In an exemplary embodiment, filtering the projection data to obtain filtered data may include:

[0193] The projection data are rearranged using a pre-established initial virtual coordinate system, which takes the detector irradiated by the central radiation source as the origin, the spiral line as the horizontal axis, and the horizontal displacement bed direction as the vertical axis.

[0194] The rearranged projection data is filtered according to the windowing filter kernel to obtain filtered data.

[0195] In the embodiment of the present application, the position of the detector irradiated by the central radiation source in the radiation source irradiation range (cone beam) can be first used as the origin of the UV coordinate system (initial virtual coordinate system), and then the horizontal axis (U axis) of the coordinate system is defined as the direction along the spiral line. This direction is usually a rotation trajectory around the scanning center and is used to represent the movement of the detector during the scanning process; the vertical axis (V axis) of the coordinate system is defined as a direction parallel to the horizontal displacement bed direction. This direction is used to represent the direction in which the patient moves along the bed during the scanning process. Figure 6 It should be noted that the establishment of the initial virtual coordinate system can be achieved in the pre-reconstruction process.

[0196] For example, a two-dimensional UV coordinate system is established with the center point of the detector (the point corresponding to the source is the origin. It should be noted that if the detector is not placed precisely at the center, then this center point will not coincide with the point corresponding to the source). Based on the preset sampling size, △U (the change on the U axis) and △V (the change on the V axis) in the coordinate system are calculated. Among them, the U axis is called a virtual coordinate channel. Its characteristic is that it is not a straight line, but a curved shape, which means that in this coordinate system, the values ​​of the U axis are not arranged according to a conventional linear relationship.

[0197] Each frame takes the position of the detector illuminated by the central ray of the source irradiation range (cone beam) as the origin, the spiral as the U axis, and the same direction as the horizontal displacement bed as the V axis. The UV wrapped area is an irregular image, see Figure 6 The midpoint matrix takes values ​​at equal intervals according to the wrapping distance in the V direction and at equal curve distances in the spiral direction in the U direction.

[0198] The original projection data is acquired from the CT detector. This data is typically recorded at various angles and positions, with each projection data point typically including information such as the projection angle, detector position, and corresponding radiation intensity. The acquired projection data is then rearranged using a pre-established UV coordinate system. For example, the projection data transformation parameters can be calculated, and based on the definition of the initial virtual coordinate system and the projection data transformation parameters, the position of each projection data point in the new coordinate system is calculated. Each projection data point is then transformed from the original coordinate system to the new virtual UV coordinate system. This process often requires consideration of information such as the detector's angle, position, and spiral trajectory.

[0199] In an exemplary embodiment, referring to Figure 7 As shown, rearranging the projection data using a pre-established initial virtual coordinate system may include steps 702 to 704, wherein:

[0200] Step 702, rearranging the projection data into parallel beam projection data by rearranging the projection data in terms of direction angles;

[0201] Step 704 , performing three Hermite interpolations on the parallel beam projection data according to the up-sampled data amount and the angle rearrangement interval to obtain a three-dimensional surface, and performing coordinate interpolation processing perpendicular to the initial virtual coordinate system on the three-dimensional surface to obtain rearranged projection data.

[0202] In an embodiment of the present application, the projection data is first subjected to azimuth rearrangement, and the projection data of the fan beam is rearranged into a parallel beam: the non-central rays of the fan beam are rearranged into a parallel beam pattern, so that these rays are parallel to the central ray at the azimuth angle. Exemplarily, the azimuth rearrangement may include azimuth adjustment and geometric transformation, wherein the azimuth adjustment includes adjusting the fan beam ray data at different azimuth angles to be consistent with the target parallel beam direction; the geometric transformation includes performing a geometric transformation on the data, recalculating and adjusting each ray in the fan beam to the form of a parallel beam, and the process may include rotation and scaling to match the spacing of the parallel beam.

[0203] For example, refer to Figure 8 As shown, the direction angle is first rearranged to rearrange the fan beam into a parallel beam. Figure 8 The rays parallel to the central ray at the middle direction angle (set as the starting angle 0 degrees) are rearranged to the non-central ray at the direction angle, that is, the non-central ray is parallel to the central ray, and the interval after rearrangement is smaller than the original interval, because the non-central position is arbitrary, such as the oblique ray hitting the detector at the coordinate (iu, iv), it is rotated through the rotation center to the ray parallel to the central ray at the direction angle IV ( Figure 8 The example in the figure is rotated 20.35 degrees and is closer to the center: AI <DJ。

[0204] It should be noted that the channel spacing is not equal after the direction angle is rearranged. The channel spacing is related to the installation rules of the detector. For example, assuming that the number of channels is 384, refer to Figure 9 As shown, the intervals between channels after the 384 channel interval direction angles are rearranged (the horizontal axis is the channel interval, the vertical axis is the interval distance in mm, from left to right are the interval between the 1st channel and the 0th channel, the interval between the 2nd channel and the 1st channel, ...).

[0205] After the azimuth angles are rearranged, the intervals between the rearranged channels (hereinafter referred to as the azimuth angle rearrangement intervals in the present embodiment) can be obtained. According to the amount of upsampled data and the azimuth angle rearrangement intervals, the projection data of the parallel beams are subjected to three Hermite interpolation (Hermite interpolation). The three-dimensional surface obtained after interpolation is referred to as Figure 10 Since the rearranged single-view detector data is not a standard two-dimensional matrix, but a curved matrix related to the spiral track, the cubic interpolation used in the embodiment of the present application has a better interpolation effect than the general linear interpolation.

[0206] Since the interpolation results in a three-dimensional surface, it is similar to a parallelogram in two dimensions. That is, after the Hermite interpolation in the previous step, the V axis spacing is the same, but because the U axis is not horizontal, directly using the data on the three-dimensional surface during back projection will result in inappropriate interpolation data, especially when the pitch is large. For example, refer to Figure 11 As shown (it should be noted that Figure 11 What is actually displayed in the figure is a curved surface rather than a flat surface). The points used for bilinear interpolation of point G1 in the virtual UV coordinate system are G, M, L, and H. As can be seen from the figure, point H is obviously inappropriate, and the most appropriate data points should be G, M, L, and R.

[0207] Therefore, after performing three Hermite interpolations, the three-dimensional surface obtained after the interpolation can be subjected to coordinate interpolation processing perpendicular to the initial virtual coordinate system. The purpose of the interpolation processing is to make the U axis horizontal. Among them, the virtual UV coordinate system is a two-dimensional coordinate system, in which the U axis and the V axis correspond to the horizontal and vertical directions of the texture respectively. The "vertical" in the embodiment of the present application refers to the vertical direction relative to the UV coordinate system, that is, the V-axis direction. Since the U axis in the virtual UV coordinate system is a spiral curve, that is, the U axis is curved to some extent, not horizontal. The interpolation processing perpendicular to the initial virtual coordinate system is to perform interpolation processing along the V-axis direction (vertical direction). The goal is to adjust the points on the U axis so that they are horizontally distributed along the U axis to "straighten" the U axis and make it horizontal.

[0208] In an exemplary embodiment, referring to Figure 12 As shown, in step 702, coordinate interpolation processing of the three-dimensional surface perpendicular to the initial virtual coordinate system is performed to obtain rearranged projection data, which may include steps 1202 to 1206, wherein:

[0209] Step 1202, determining the grid points on the three-dimensional surface and calculating the rotation matrix of each grid point;

[0210] Step 1204 , using a rotation matrix to transform the coordinates of each grid point from the initial virtual coordinate system to the target virtual coordinate system, where the horizontal axis of the target virtual coordinate system is horizontal;

[0211] Step 1206 , interpolating each grid point in the target virtual coordinate system to obtain rearranged projection data.

[0212] In the embodiment of the present application, after obtaining the three-dimensional surface, the UV coordinates need to be interpolated to ensure that the U axis is horizontal. For example, evenly distributed grid points are selected on the three-dimensional surface according to the requirements. Generally, they can be distributed according to the geometric characteristics of the object to ensure that the entire area within the effective range is covered. For example, the parameters u and v can be set in three-dimensional space to form a UV grid, and these grid points will be interpolated in the subsequent process.

[0213] The target direction of the U axis can be determined, and the rotation matrix R can be constructed by calculating the rotation angle θ based on the initial UV coordinates and target direction of the grid point. For example:

[0214] Furthermore, this rotation matrix is ​​applied to the UV coordinates of each grid point to adjust the direction of the U axis. For example, (u', v', 1) = R(u, v, 1), where (u, v) represents the original coordinates in the standard initial virtual coordinate system, and (u', v') represents the new coordinates after the original coordinates are converted to the target virtual coordinate system. This allows the coordinates of each grid point to be transformed from the virtual coordinate system to the virtual coordinate system horizontal to the U axis.

[0215] After the coordinate transformation, each grid point can be interpolated in the target virtual coordinate system. For example, bilinear interpolation or bicubic interpolation can be used to obtain smoother projection data.

[0216] In an exemplary embodiment, referring to Figure 13 As shown, in step 1206, interpolation processing is performed on each grid point in the target virtual coordinate system to obtain rearranged projection data, which may include the following steps 1302 to 1304, wherein:

[0217] Step 1302: interpolate each grid point in the target virtual coordinate system to obtain an interpolation result;

[0218] Step 1304 : Filter the interpolation result according to the initial virtual coordinate system to obtain rearranged projection data.

[0219] In the embodiment of the present application, each grid point can be interpolated in the target virtual coordinate system to obtain an interpolation result. The interpolation method can include methods such as bilinear interpolation or bicubic interpolation. The interpolation method is not specifically limited in the embodiment of the present application. After obtaining the interpolation result, the data within the boundary range of the original coordinate system can be saved as valid data, and the peripheral data is non-valid. For example, the boundary range of the three-dimensional surface in the initial virtual coordinate system can be determined, and the data that exceeds the boundary range in the interpolation result can be marked as invalid data to avoid affecting the result during the reconstruction process. Only the data within the boundary range is extracted as valid data to form the final rearranged projection data for subsequent CT reconstruction.

[0220] Since the initial virtual coordinate system provided in the embodiment of the present application has a spiral line as the horizontal axis and a horizontal displacement bed direction as the vertical axis, which is consistent with the scanning path, curve fitting interpolation is performed according to the U axis of the spiral line during rearrangement. After rearrangement, grid node interpolation is performed according to the UV coordinate system that approximates the horizontal U axis, which can greatly improve the accuracy of back projection.

[0221] In an exemplary embodiment, before filtering the projection data to obtain the filtered data, the method may further include the steps of: performing slope correction on the projection data, and after obtaining the filtered data, storing the filtered data in a video memory;

[0222] After obtaining the back-projection result of the coarse reconstruction process, intercept correction may be performed on the back-projection result to obtain a first CT reconstructed image.

[0223] In the related art, slope correction and intercept correction are performed during the post-reconstruction process to stretch the dynamic range of the reconstruction result so that the air reaches -1000 and the water reaches 0, that is, linear correction, and the slope and intercept are obtained to correct each pixel point. In the embodiment of the present application, the projection data is slope corrected before back projection. Exemplarily, for the projection data, the projection data can be corrected based on the air average value. The difference between the air average value under the initial angle period of the projection data and the air average value under the initial angle period of the air can be considered during the correction process. The specific process refers to the following formula (eight):

[0224]

[0225] Where Δ=I air -J air , J air Refers to the value of the initial angle of the projection data in one cycle (for example, one cycle is 360 degrees, and the projection data is at 30 degrees). airRefers to the value of the initial angle of air in one cycle (for example, one cycle is 360 degrees, and the air data is at 70 degrees), Δ represents the angle deviation, prj1 represents the projection data before correction, air j+Δ It's prj j The corresponding air data, where j represents the current angle position, is the projection data taking the data at the angle position j, and the air takes the data at the position j+Δ, prj j ' represents the projection data after correction.

[0226] After air correction, the air-corrected projection data can be further subjected to water hardening correction. Slope correction can also be performed on the water-hardening-corrected projection data. After filtering the corrected projection data, the filtered data can be stored in the video memory. This allows subsequent fine reconstruction to directly retrieve the filtered data from the video memory for back-projection, eliminating the need for repeated filtering and correction processes.

[0227] The filtered data is back-projected using a preset back-projection algorithm, and the coarsely reconstructed back-projection result is intercept-corrected to obtain the first CT reconstructed image. Because projection values ​​are a superposition of pixels, correcting the slope of the CT values ​​before reconstruction allows for equal scaling, effectively expanding the dynamic range of the projection data and improving both data accuracy and reconstruction precision.

[0228] It should be noted that when the reconstruction strategy is the second reconstruction strategy, the fine reconstruction method needs to perform operations such as filtering and correction of the projection data in the aforementioned embodiment, which will not be repeated here in the embodiment of this application, and reference can be made to the relevant description of the aforementioned embodiment.

[0229] In an exemplary embodiment, referring to Figure 14 As shown, the projection data is CT reconstructed using a fine reconstruction method to obtain a second CT reconstructed image, including the following steps 1402 to 1408, wherein:

[0230] Step 1402: Obtain filtered data from a video memory, perform back-projection processing on the filtered data to obtain a back-projection image, and extract bone tissue data from the back-projection image;

[0231] Step 1404, performing orthographic projection on the bone tissue data to obtain bone projection data;

[0232] Step 1406, taking the difference between the projection data and the bone projection data as bone-removed tissue data;

[0233] Step 1408: Obtain a second CT reconstructed image based on the bone-removed tissue data and the back-projected image.

[0234] Since the projection data has been corrected, rearranged, and filtered during the coarse reconstruction process, the filtered data corresponding to the projection data can be directly obtained from the video memory data at this time, and the filtered data can be back-projected to obtain the back-projection image, and the bone tissue data can be extracted from the back-projection image.

[0235] Back projection is the process of back-projecting each projection data into the image space according to its imaging angle. The specific operation is to back-project each projection data into the two-dimensional image along the angle at which it was collected. Exemplarily, for the filtered data at any angle, each pixel value of the filtered data can be accumulated along the angle to gradually construct a back-projection image. The filtered data of all angles are accumulated in the back-projection process to form a back-projection image, which usually contains mixed information of various tissues. In fact, weighted back-projection methods (such as normalizing the filter values) can also be used in the accumulation process to improve the back-projection quality.

[0236] The back projection algorithm used in the fine reconstruction process may include the following strategies: the reconstruction interval used in the back projection is the current reconstruction interval, 180-degree interpolation is used for interpolation processing, and the horizontal motor position of each frame is read from the horizontal motor file.

[0237] After obtaining the back-projection data, clear bone tissue data can be extracted from the back-projection image. For example, bone tissue data can be extracted from the back-projection image through the CT value range corresponding to the bone tissue. The bone tissue data extraction process will be introduced in the following embodiments of the present application, and will not be described in detail in the embodiments of the present application.

[0238] In the embodiment of the present application, when the bone tissue data is orthographically projected, the path of the orthographic projection can be regarded as a ray following the scanning path (ie, the whole follows a spiral line) at a cone angle. Figure 15 As shown, Figure 15 The image shows the forward projection path of a frame. The scanning path follows a spiral, simulating the process of forming an image in three-dimensional space. This scanning method ensures that each frame of the original image has corresponding projection data, eliminating the rearrangement step in the reconstruction process. This significantly increases data processing speed and makes the imaging process more efficient.

[0239] Furthermore, the orthographic projection process can also downsample the projection lines, which can reduce the amount of data processing. For example, by reducing the image resolution to half (or one-quarter) of the original image, this can reduce the amount of computation and improve the efficiency of orthographic projection while ensuring image quality. When downsampling to extract the original image value, in the embodiment of the present application, the x-coordinate can be simply shifted one position to the left. This allows the corresponding projection value to be quickly obtained without complex processing to achieve downsampling.

[0240] In an exemplary embodiment, in step 1404, performing orthographic projection on the bone tissue data to obtain bone projection data may include the following steps:

[0241] Based on the pixel points in the axial image corresponding to the bone tissue data and the ray length weight corresponding to each pixel point, the CT value of each pixel point is weighted and superimposed to obtain the bone projection data; among which, the ray length weight is negatively correlated with the length of the ray in the voxel.

[0242] In an embodiment of the present application, the ray length weight corresponding to the ray can be determined based on the length of the ray in the voxel, wherein the ray length weight is inversely proportional to the length of the ray in the voxel, that is, the longer the ray in the voxel, the smaller the influence of the point on the axial plane (the orthogonal weight is the largest). Determine the pixel points in the axial image corresponding to the bone tissue data through which the ray passes, and perform weighted superposition of the pixel values ​​corresponding to each pixel point and the ray length weight, so as to obtain the orthogonal projection value of each ray in the orthogonal projection, and then obtain the bone projection data. For example, Figure 16a The relationship between the ray and the axial image pixels in the orthographic projection is shown; Figure 16b The relationship between the ray and the voxel diagonal within a voxel of the target reconstructed object is shown.

[0243] In an embodiment of the present application, after obtaining the bone projection data, the difference between the projection data and the bone projection data can be determined. This difference represents the signal of other tissues (such as soft tissue or liquid) other than bone tissue at the pixel position. The difference is used as bone-removed tissue data, and the bone-removed tissue data is an image of other tissues other than bone tissue.

[0244] In the embodiment of the present application, the bone projection data can be corrected by removing the bone tissue data, and the corrected bone projection data can be used for reconstruction to obtain a new bone reconstruction image. The new bone reconstruction image can be superimposed with the back-projection image to obtain a CT reconstructed image.

[0245] The CT reconstruction method provided in the embodiment of the present application can filter the projection data to obtain filtered data, and perform back projection processing on the filtered data. After obtaining the back projection image, bone tissue data can be extracted from the back projection image. The bone tissue data is forward projected to obtain bone projection data, and the difference between the projection data and the bone projection data is used as bone-removed tissue data, and then a CT reconstructed image is obtained based on the bone-removed tissue data and the back projection image. Using the CT reconstruction method provided in the embodiment of the present application, in the post-reconstruction bone sclerosis correction process, there is no need to perform a forward projection on the original data. The reconstructed projection data can be directly used, which can reduce the amount of calculation and improve the efficiency of CT reconstruction.

[0246] In an exemplary embodiment, referring to Figure 17 As shown, extracting bone tissue data from the back-projection image may include the following steps 1702 to 1706, wherein:

[0247] Step 1702: extract bone tissue pixels in the back-projection image, where the CT values ​​corresponding to the bone tissue pixels are within the bone tissue CT value range;

[0248] Step 1704: For any bone tissue pixel, the bone tissue segment corresponding to the bone tissue pixel is determined based on the CT value of the bone tissue pixel. Furthermore, the normalization coefficient of the bone tissue pixel is determined based on the upper and lower CT values ​​of the bone tissue segment, the normalization coefficients corresponding to the upper and lower CT values, and the CT value of the bone tissue pixel. The bone tissue CT value range is composed of multiple bone tissue segments.

[0249] Step 1706: Obtain bone tissue data based on the normalization coefficient and CT value of each bone tissue pixel.

[0250] In an embodiment of the present application, a bone tissue CT value range is pre-set, and the bone tissue CT value range can be composed of multiple bone tissue segments. For example, the bone tissue CT value range is hu1~hu2, which includes three bone tissue segments: the first segment hu1~hua (e.g., 850~1150), which is a CT value between brain tissue and bone tissue, the second segment hua~hub (e.g., 1150~2075), which is a CT value range of bone tissue, and the third segment hub~hu2 (e.g., 2075~3000), which is a CT value between bone tissue and high-density body. The upper and lower CT values ​​of each segment have a preset normalization coefficient, for example, hu1 corresponds to s1 (e.g., 0), hua corresponds to s2 (e.g., 0.4), hub corresponds to s3 (e.g., 0.7), and hu2 corresponds to s4 (e.g., 1).

[0251] For each pixel point in the back-projection image, the pixel point whose corresponding CT value is within the range of bone tissue CT value can be regarded as a bone tissue pixel point. For example, when the CT value hui of pixel point i is between hu1 and hu2, the pixel point i can be determined as a bone tissue pixel point.

[0252] After obtaining each bone tissue pixel point, the CT value of each bone tissue pixel point can be obtained, and the bone tissue segment corresponding to each bone tissue pixel point can be determined based on the CT value of each bone tissue pixel point. Based on the upper and lower CT values ​​of the bone tissue segment corresponding to each bone tissue pixel point, the normalization coefficient corresponding to the upper and lower CT values, and the CT value of the bone tissue pixel point, the normalization coefficient of each bone tissue pixel point can be determined. For example, the process of determining the normalization coefficient can refer to the following formula (9):

[0253]

[0254] Among them, s represents the normalization coefficient corresponding to the bone tissue pixel, s i is the normalization coefficient corresponding to the lower limit CT value of the bone tissue segment where the bone tissue pixel point is located, s j is the normalization coefficient corresponding to the upper limit CT value of the bone tissue segment where the bone tissue pixel point is located, hu is the CT value of the bone tissue image point, hu i is the lower limit CT value of the bone tissue segment where the bone tissue pixel is located, hu j It is the upper limit CT value of the bone tissue segment where the bone tissue pixel point is located.

[0255] After obtaining the normalization coefficient corresponding to the bone tissue pixel point, the CT value of each bone tissue pixel point is multiplied by the normalization coefficient corresponding to each bone tissue pixel point, which is the segmented and extracted bone tissue data.

[0256] By adopting the CT reconstruction method provided in the embodiment of the present application, the accuracy of bone tissue data extraction can be improved by setting multiple bone tissue segments, and then bone sclerosis correction can be performed based on the extracted bone tissue data, which can further improve the CT reconstruction accuracy.

[0257] In an exemplary embodiment, referring to Figure 18 As shown, in step 1408, a second CT reconstructed image is obtained based on the bone-removed tissue data and the back-projected image, including the following steps 1802 to 1806, wherein:

[0258] Step 1802: The ratio of the bone-removed tissue data to the water reference is used as the equivalent length of each projected pixel point in the bone-removed tissue data relative to the water;

[0259] Step 1804 , performing weighted correction on the bone projection data using the artifact correction coefficient corresponding to the equivalent length to obtain corrected bone projection data;

[0260] Step 1806 : After reconstructing the corrected bone projection data, the reconstruction result is superimposed on the back-projection image to obtain a second CT reconstructed image.

[0261] In the embodiment of the present application, the bone-removed tissue data is first divided by the water reference, and this ratio is used as the equivalent length of each projection pixel in the bone-removed tissue data relative to water. Based on the equivalent length, the corresponding artifact correction coefficient is then searched in a pre-set coefficient table, and the polynomial weighted correction is performed on the bone projection data using the artifact correction coefficient to obtain the corrected bone projection data. The correction process can be described by referring to the following formula (10):

[0262]

[0263] Where prj represents the bone projection data before correction, prj′ represents the bone projection data after correction, N is the number of polynomials, α i is the artifact correction coefficient.

[0264] After obtaining the corrected bone projection data, reconstruction can be performed based on the corrected bone projection data to obtain a reconstruction result. The specific reconstruction process is subject to the selected reconstruction algorithm. The reconstruction result is superimposed with the back-projection image to obtain a second CT reconstructed image.

[0265] Since the scanning path of the bone projection data in the embodiment of the present application is a rearranged path, the efficiency of bone sclerosis correction can be greatly improved because there is no need for further rearrangement.

[0266] In the post-reconstruction process, the reconstruction result and the back-projection image can be superimposed to obtain a superimposed result, and the superimposed result can be intercept-corrected to obtain a corresponding correction result. The correction result can be further subjected to layer thickness processing to obtain a second CT reconstructed image.

[0267] The process of layer thickness processing includes: layer thickness superposition calculation, obtaining effective layer thickness according to reconstruction interval, display layer spacing and preset layer thickness, layer thickness superposition calculation, and reference Figure 19a and Figure 19b As shown, schematic diagrams are shown without layer thickness processing and with layer thickness processing respectively.

[0268] In the present application, refer to Figure 20 As shown, in step 1402, back-projection processing is performed on the filtered data to obtain a back-projection image, which may include the following steps 2002 to 2004, wherein:

[0269] Step 2002: determining an extension range according to the extension collimation width, the pitch, and the reconstruction interval, and shifting the horizontal motor starting position left and right based on the extension range to obtain a virtual horizontal motor position;

[0270] Step 2004 : performing interpolation processing based on the virtual horizontal motor position and the filtered data to obtain extended projection data, and performing weighted back projection processing based on the extended projection data and the filtered data to obtain a back projection image.

[0271] In the embodiment of the present application, during the CT reconstruction process, the projection data can also be extended in the Z direction, including setting a virtual horizontal motor position. For example, the extension range can be determined according to the extension collimation width, pitch and reconstruction interval, and then the horizontal motor starting position is shifted left and right respectively based on the extension range to obtain the virtual horizontal motor position. Figure 21As shown, the virtual horizontal motor position (i.e., the position of the spiral sweep extending upward and downward) is shown. The process of determining the extension range can refer to the following formula (XI):

[0272]

[0273] Where HarfOffset represents the extension range, ecol represents the extension collimation width, inc represents the reconstruction interval, and p represents the pitch.

[0274] In the embodiment of the present application, interpolation can be performed based on the virtual motor horizontal motor position and projection data to obtain extended projection data. CT weighted back projection is performed on the extended projection data using the original projection data and the extended projection data, thereby increasing the number of reconstruction frames and reducing artifacts caused by incomplete projections.

[0275] Exemplarily, each pixel point and each rearranged view in the XY plane of the target coordinate system (each axial diagram) is traversed to determine the corresponding starting and ending positions of the horizontal motor under the rearranged view. The starting and ending positions are shifted left and right according to the previously determined extension range, and the upper and lower boundaries of the effective Z plane under the rearranged view can be determined. The pixel points on XY are converted to the initial virtual coordinate system to determine the UV coordinate value, and the channel position is determined according to the U coordinate value, that is, the position of the horizontal channel for collecting rearranged data (only valid horizontal channels are calculated); according to the V-direction position of the ray, the V coordinate value determined by converting the pixel points on XY to the initial virtual coordinate system and the V position (i.e., Z position) of the desired image plane, the row number is determined, that is, the position of the longitudinal channel for collecting rearranged data (only valid longitudinal channels are calculated). According to the position of the valid longitudinal channel for collecting rearranged data, the weight W is calculated according to the aforementioned formula (six) Q (q).

[0276] Iterate the rearranged data corresponding to the corresponding XYZ values ​​and multiply them by the corresponding weight W Q (q). The weights in the same direction are weighted averaged, that is, 180-degree interpolation is used for interpolation reconstruction. In the embodiment of the present application, an extended image can be obtained by extension, and the weight of the position of the extended image can be adjusted (the virtual horizontal motor parameters are set, and the weight of the effective weight W value is obtained by extension), so as to increase the proportion of rays and reduce the artifacts of incomplete projection. Finally, the superimposed axial image is multiplied by the angle weight, which is the back-projection image obtained after back-projection.

[0277] In an exemplary embodiment, referring to Figure 22 As shown, in step 2004, weighted back projection processing is performed based on the extended projection data and the filtered data to obtain a back projection image, including the following steps 2202 to 2206, wherein:

[0278] Step 2202: performing weighted back-projection processing based on the extended projection data and the filtered data to obtain an initial back-projection image;

[0279] Step 2204: For any pixel in the initial back-projection image, calculate the distance between the vertical coordinate value of the pixel and the position of the horizontal motor corresponding to the central ray passing through the pixel. If the distance is less than a preset threshold, the pixel is determined to be valid.

[0280] Step 2206 , performing weighted summation processing on the rearranged data corresponding to the valid pixel points according to the Q value weight curve to obtain a back-projection image.

[0281] In an embodiment of the present application, after weighted back projection processing based on the extended projection data and filtered data, an initial back projection image can be obtained. During the reconstruction process, effective reconstruction points can be screened from the initial back projection image. For example, the distance between the V coordinate value (coordinate value on the vertical axis) of each coordinate point (reconstruction point) and the position of the horizontal motor of the center ray of the ray passing through it can be calculated. If this distance is greater than or equal to the preset threshold value, the pixel point is considered invalid, otherwise the pixel point is considered valid.

[0282] Among them, the V coordinate value = frame number * inc + startPos, wherein the frame number indicates which layer the reconstruction point is in, that is, the frame number, inc indicates the layer interval, that is, the distance between the two frames, startPos indicates the distance of the starting reconstruction (that is, the starting position of the horizontal motor movement), which indicates the distance from the 0th image to the initial horizontal motor position; the position of the horizontal motor of the center ray of the ray passing through the reconstruction point is the position value corresponding to the frame where the center ray is located read from the horizontal motor parameters, representing the distance from the frame to the initial horizontal motor position; the difference between the V coordinate value of the reconstruction point and the position of the horizontal motor of the center ray of the ray passing through it is the distance between the V coordinate value of the reconstruction point and the position of the horizontal motor of the center ray of the ray passing through it.

[0283] In fact, referring to Figure 23a As shown in the figure, the reconstructed target field of view is a sphere (or an ellipsoid in one example) for each frame ray. If the reconstructed point is inside the sphere, it means that the reconstructed point is valid and can participate in the reconstruction. If the reconstructed point is outside the sphere, it can be considered that the reconstructed point is too far from the source and is an edge ray. Due to errors and other reasons, the accuracy of the reconstructed point is poor, so the reconstructed point can be considered invalid and will not be included in the superposition calculation (including not considering it when calculating the weight). Figure 23b As shown in the figure, the same reconstruction point on the two spheres is a valid point for one ray and is included in the reconstruction, but is an invalid point for the other ray and is not included in the reconstruction. This is equivalent to an effective 3D simulated collimator, which can effectively improve the image's cone angle artifacts and windmill artifacts.

[0284] The CT reconstruction method provided in the embodiments of the present application achieves real-time reconstruction (reconstruction while capturing images) through multi-segment reconstruction. By segmenting the reconstruction process, low-precision coarse reconstruction is used for real-time reconstruction and display, while high-precision fine reconstruction is used to save the reconstruction results. This ensures both the requirements for real-time acquisition and reconstruction and the ultimate accuracy. Furthermore, a pre-reconstruction stage is added to complete some reconstruction content before acquisition. This allows real-time reconstruction to be achieved even on low-end computers, without affecting the doctor's real-time supervision of the imaging process. For example, this allows the doctor to detect any sudden movements or other unexpected events during the imaging process and to check whether the scanning position is normal in real time.

[0285] A judgment can be made before coarse reconstruction: based on the operator's selected parameters, the need for coarse reconstruction is determined. When the reconstruction computational complexity is low, fine reconstruction is directly initiated. Only when the reconstruction computational complexity is high does a coarse and fine reconstruction combination take place. Coarse reconstruction also includes calibration, meaning scanning, acquisition, calibration, reconstruction, and display are all completed in real time.

[0286] The virtual coordinate system is established based on the scanned spiral curve. The entire coordinate system is an approximately rectangular surface, with the U axis as the curve axis and the V axis as the direction of bed movement. In the entire reconstruction process, the coordinate system can be regarded as two: one is the initial UV virtual coordinate system during rearrangement, that is, during rearrangement, curve fitting interpolation is performed according to the U axis (not the final U axis, the U axis of the previous UV virtual coordinate system, that is, the axis of the spiral line), not linear interpolation. The other is the final UV virtual coordinate system interpolated after rearrangement. The retaken points are interpolated according to the final UV virtual (U axis horizontal) coordinate system. That is, the slanted points are made to go straight, which improves the accuracy of the back projection.

[0287] Although the back projection of coarse reconstruction and the back projection of fine reconstruction use the same basic method, the back projection process of coarse reconstruction requires less computation, which can greatly speed up the back projection efficiency. Furthermore, the coarse reconstruction algorithm can be adjusted based on the acquisition time. For example, if the acquisition time is set to 1-2 seconds, and the acquisition time is much longer than the time for one cycle of coarse reconstruction, the coarse reconstruction can be divided into multiple cycles, and the coarse reconstruction can be completed entirely within the acquisition time. If the acquisition time is subsequently increased, such as the fastest spiral CT in the industry at 0.28 seconds, the first six steps of coarse reconstruction can be divided into shorter frames, such as half a cycle or even just one frame, and only a full cycle is performed during back projection. This can also ensure that coarse reconstruction and acquisition are performed simultaneously and do not take up separate time.

[0288] To help those skilled in the art better understand the embodiments of the present disclosure, the following describes the embodiments of the present application through specific examples. In this example, the CT reconstruction method includes four stages: pre-reconstruction, coarse reconstruction, fine reconstruction, and post-reconstruction. Specifically:

[0289] 1. In the pre-reconstruction stage, the process includes calculating the number of reconstruction layers, calculating the starting position of the horizontal motor according to the positioning frame selection position, allocating video memory, reading the data of each parameter file in advance, establishing a virtual coordinate system, customizing missing parameters, texture binding of weights, and obtaining the window filter kernel.

[0290] Pre-reconstruction can be performed before data acquisition, eliminating the need for acquisition and reconstruction. Because storage space must be allocated in advance (which takes time), the number of layers to be reconstructed must be known. The reconstruction interval and layer thickness are both used during reconstruction, and they differ between coarse and fine reconstruction. The number of layers for coarse and fine reconstruction can be calculated using the aforementioned formulas (1) to (3), which will not be detailed in this example.

[0291] Calculating the horizontal motor's starting position based on the position selected on the positioning sheet involves the user selecting the capture area using the positioning sheet. During spiral scanning, the starting point of the scan must be moved to the user-selected starting point. The scanning distance is the distance specified in the positioning sheet. Therefore, only Formula (4) or Formula (5) is required to calculate the horizontal motor's starting position.

[0292] During the process of allocating video memory, the video memory is allocated and initialized in advance to avoid allocating and releasing video memory during the reconstruction process, which greatly improves the reconstruction time. Specifically, the opened video memory space includes space for storing the following data: tomographic images after layer thickness superposition reconstruction; tomographic images after back projection; total projection data images; detector geometric position parameter index table (X, Y, Z of each detector point, in a virtual coordinate system); actual horizontal motor correction parameter index table (the position of the patient support, that is, the bed, in each frame of the scan, in mm value in the Z direction); uniform horizontal motor correction parameter index table (used for coarse reconstruction, average calculated based on pitch, collimation width, etc.); air data image (storing air images after dark field correction); angle data of each frame of rays relative to the central ray (in the UV coordinate system); coordinate value distance data of each frame of rays in the V direction (in the UV coordinate system); directional angle interpolation data (coordinate value distance data of each ray in the V direction after directional angle rearrangement); rearranged image; filter kernel data; filtered image; water uniformity correction parameter index table; water CT value correction parameter index table; bone sclerosis correction parameter index table; detector odd-even frame correction ratio parameter index table; segmented projection data image, etc.

[0293] The data of each parameter file is read in advance, including air correction data, detector position data, inconsistency parameters, water correction parameters, bone sclerosis parameters, etc. The parameter data is read into the GPU in advance to avoid reading the hard disk and switching from the CPU to the GPU during the reconstruction process, which greatly increases the reconstruction time.

[0294] In the process of establishing the virtual coordinate system, the center point of the detector (the point corresponding to the source, such as the detector is offset and not the center point) is used as the origin to establish a two-dimensional UV coordinate system. According to the preset sampling size, △U and △V are obtained to obtain the size of the virtual coordinate U channel. The U axis of the virtual coordinate system is not a linear axis, but a curved axis. Under each frame of the image, the position of the detector irradiated by the central ray of the source irradiation range (cone beam) is used as the origin, the spiral is used as the U axis, and the V axis is in the same direction as the horizontal displacement bed. The UV wrapped area is an irregular image, see Figure 6 The midpoint matrix takes values ​​at equal intervals according to the wrapping distance in the V direction and at equal curve distances in the spiral direction in the U direction.

[0295] Like oral scan registration and fusion, CT also requires coordinate system transformation, though its coordinate systems are simpler, consisting of two: the target coordinate system, which can be considered the world coordinate system and is the x, y, and z coordinates of the reconstructed image. The other is the acquisition geometric coordinate system, which is the coordinate system of the detector at a fixed position within a frame of image. The length and width directions u and v of the detector are the x and y coordinates, and the direction perpendicular to the detector from the source focal point is z. This coordinate system can be called a virtual coordinate system. Backprojection involves transforming the virtual coordinate system to the target coordinate system. The backprojection process involves superimposing the target position in each virtual coordinate system, so this transformation is involved. One purpose of reordering is to facilitate backprojection interpolation. Another reordering process is also an interpolation process. For example, if 384 points were originally reordered to 512 points, this is equivalent to densifying the data, which improves image quality.

[0296] Three-dimensional geometric position display reference after spiral CT rearrangement Figure 10 , after spiral CT rearrangement, the detector UV direction plane display reference Figure 25 As shown, taking the plane display as an example, the U axis of the coordinate system is the horizontal axis (but it is actually a curved surface, the curved surface where the spiral is located, or the spiral is attached to the outer surface of the cylinder. The U axis is the curved axis of the outer surface of the cylinder (the curved axis is perpendicular to the central axis of the cylinder), and the V axis of the coordinate system is the vertical axis (parallel to the central axis of the cylinder). The advantage of this is that the interpolation change is accurate, that is, the points rearranged to the detector are not straight lines with completely equal intervals, but points on an oblique surface. This position interpolation is more realistic than the point interpolation position change of an approximate square dot matrix.

[0297] Customization of missing parameters includes: customizing horizontal motor parameters based on pitch, position parameters (that is, the mm value of the horizontal motor encoder for each frame of image, which has different values ​​for each frame of image captured), etc. This is used for coarse reconstruction, that is, coarse reconstruction takes approximate horizontal motor parameters.

[0298] During the texture binding process of the weight, the specific calculation formula of the weight is shown in formula (6).

[0299] When calculating the window filter kernel, a sinc window is used as the window function. Compared to a hanning window with the same cutoff frequency, this window function produces clearer details, while the hanning window is overly smooth. Two thresholds are set: low and high frequencies. When the window value is below the low frequency, the threshold is 1, meaning the threshold is fully open, and this is called the all-pass frequency. When the window value is above the high frequency, the threshold is 0, meaning the high-frequency threshold is the cutoff frequency. Depending on the reconstructed image requirements, the all-pass frequency and cutoff frequency can be set to different values, divided into seven groups, each corresponding to different sharpening and smoothing effects.

[0300] The position of the full-pass frequency and the cutoff frequency, the horizontal axis is the sampling point (positive and negative normalization, that is, the center is the center of the image sampling width), and the vertical axis is the window function value. Figure 4 As shown, WL is the full-pass frequency (with values ​​of 1, 1 / 16, and 1 / 32), WH is the cutoff frequency (with values ​​of 1, 2 / 3, and 1 / 2), and WL <= WH. The horizontal axis is the convolution sampling width (the value of 0 represents the sampling center), and the vertical axis is the normalized window width value. The formula for the windowed filter kernel is shown in Formula (7). In this way, the filter kernel is prepared in the preprocessing, that is, before acquisition. The texture binding above is also prepared before acquisition, and does not occupy the acquisition and reconstruction time.

[0301] 2. During the coarse reconstruction phase, to display reconstruction results in real time and facilitate physician assessment of acquisition status, a coarse reconstruction is performed. The reconstruction interval, display slice spacing, and slice thickness are all fixed at 5 mm. This process includes logarithmic calculation, water hardening correction, CT value correction, realignment, FFT, fast back projection, and HU value calculation.

[0302] Coarse reconstruction physically splits the original acquisition based on acquisition time, ensuring that reconstruction time is less than acquisition time (acquisition and reconstruction are two independent threads). Horizontal motor speed (bed stepping speed) is calculated based on the average of user-set parameters such as pitch, rather than real-time speed (this eliminates the need to read horizontal motor files, resulting in faster reconstruction).

[0303] During logarithmic calculation, the initial angle of the air correction parameter shooting and the initial angle of the collected data are considered, and the offset value is added to make the angles correspond. The specific process is shown in formula (8), and this example will not be repeated here.

[0304] And considering that the ray is not always stable after it is turned on, for example, it shows a slow attenuation trend. Although the amplitude is not large and meets the requirements of the source stability, it still has a certain impact on the image brightness level correction. Therefore, the influence factor Δ is added, which is the difference between the air average value under the prj initial angle period and the air average value under the air initial angle period.

[0305] In the water hardening correction process, polynomial fitting correction is adopted, and the coefficients of the polynomial are determined based on the water models taken multiple times.

[0306] In the CT value correction process, the advantage of placing the slope correction before the back projection is that the dynamic range of prj can be enlarged and the data accuracy can be improved (when the same float type). Generally, the CT value correction is placed after reconstruction, and the dynamic range of the reconstruction result is stretched so that the air reaches -1000 and the water reaches 0, that is, linear correction, and the slope and intercept are obtained to correct each pixel point. In the embodiment of the present disclosure, it is divided into two steps. Before reconstruction, the slope part k of the CT value is corrected first. Because the projection value is a superposition of pixel points, it can be equally scaled, and the implementation of the slope method effectively expands the dynamic range of the data and improves the reconstruction accuracy. After the reconstruction is completed, the intercept part b=1000 is processed. The correction of the CT value is all linear stretching.

[0307] The CT value correction formula is as follows:

[0308] In the rearrangement process, first: Direction angle rearrangement: fan beam is rearranged into parallel beam. With this direction angle IV (refer to Figure 8 , the direction angle is just above, set as the starting angle 0 degrees) and the rays parallel to the central ray are rearranged to the non-central ray of the direction angle, that is, the non-central ray is parallel to the central ray, and the interval after rearrangement is smaller than the original interval, because the non-central position is arbitrary, such as hitting the coordinate (iu, iv) on the detector obliquely. The ray rotates through the rotation center to the ray parallel to the central ray under the direction angle IV ( Figure 8 The example shown is rotated 20.35 degrees) closer to the center: AI <DJ。

[0309] After the direction angles are rearranged, the channel spacing is not equal. The 384 channel spacings are as follows: Figure 9 The distance between channels after the azimuth angle is rearranged (the horizontal axis is the channel distance, the vertical axis is the distance in mm, from left to right it is the distance between channel 1 and channel 0, the distance between channel 2 and channel 1, etc.).

[0310] Next, cubic Hermite interpolation is performed based on the amount of upsampled data and the azimuth reordering interval. (Because the reordered single-view detector data is not a standard two-dimensional matrix but a curved matrix related to the spiral orbit, cubic spline interpolation is more effective than standard linear interpolation.) After azimuth reordering, the detector orientation forms a three-dimensional curved surface, which approximates a parallelogram in two dimensions. This means that after the previous Hermite interpolation step, the V-axis interval remains the same, but because the U-axis is non-horizontal, directly using this data during backprojection will result in inappropriate interpolation data. This is particularly noticeable at low pitches. Figure 11This is an approximation (because the actual surface is curved, not flat). The bilinear interpolation point for point G1 in this coordinate system is GMLH. Point H is clearly inappropriate; GMLR is preferred. Therefore, an additional step is added here to interpolate the vertical UV coordinates to level the U axis. This means setting a horizontal grid based on the sampling density. Data within the boundaries of the original coordinate system is considered valid, while data outside the boundaries is considered invalid.

[0311] In the FFT process, first, a power-of-2 zero-padding extension is performed; second, an FFT transformation is performed; third, frequency domain filtering is performed (using a pre-reconstruction filter kernel, refer to the aforementioned embodiment); then, an inverse FFT transformation is performed; and finally, multiplication by a coefficient △ is performed.

[0312] In the assignment stage, the results of the projection processing are stored in the overall pre-backprojection data, so that these processings do not need to be considered during fine reconstruction, and only backprojection is required.

[0313] During fast backprojection, since the purpose of coarse reconstruction is only to display the approximate shooting position, fast backprojection is used in this example, and image accuracy is not required to be high. Fast backprojection is achieved by using a thicker reconstruction interval for backprojection, using 360-degree interpolation instead of 180-degree interpolation to reduce the weight calculation by half, and replacing the horizontal motor position corresponding to each frame with a uniform average value calculated based on the pitch. That is, the intervals are equal, and the frame number is simply multiplied by the horizontal interval. There is no slice thickness overlay calculation, and the display result is directly displayed (the slice thickness is equal to the reconstruction interval).

[0314] During the HU value correction process, refer to the aforementioned CT value correction formula and perform a subtraction process of 1000, or perform a window level addition process of 1000 during display.

[0315] 3. During the fine reconstruction phase, the slice spacing and slice thickness are re-assigned according to the pre-set values ​​in the user interface. (The coarse reconstruction was set to 5mm. If the slice thickness in the user agreement is relatively small and the calculation is relatively large, the fine reconstruction will be based on the user settings.) The specific process of fine reconstruction includes:

[0316] Read the horizontal motor parameters. Each frame of the image capture corresponds to the physical position of the horizontal displacement table. Coarse reconstruction approximates the horizontal motor parameters, based on the average pitch interval. However, the actual horizontal displacement of each frame is not equal. Fine reconstruction is based on the actual position and is more accurate.

[0317] Reconstruct the extension of the Z-direction field of view, expand the reconstruction range, and improve the dose utilization rate. The extended half-width range is calculated using formula (11). For example, the position of the upper and lower extension of the spiral scan is referenced Figure 21By adjusting the weights at the location of the extended image (setting the virtual horizontal motor parameters and weighting the effective weight W obtained by the extension), the ray ratio is increased and artifacts of incomplete projections are reduced. The weights required for the extension are calculated by taking into account the extended horizontal motor, such as the weight W at the -10 position, and adding it to the denominator weight of the superimposed projection.

[0318] The horizontal position corresponding to the maximum reconstruction point to the center line of the superposition ray is determined based on the pitch and collimation width. This process is equivalent to the reconstructed target field of view being a sphere for each frame of ray (of course it may be an ellipsoid). That is, if the reconstruction point is inside the sphere, it is effectively reconstructed. If the reconstruction point is outside the sphere, it is invalid and does not enter the superposition calculation (of course, it is not considered when calculating the weight). It is equivalent to an effective three-dimensional simulated collimator. This method can effectively improve the cone angle artifact and windmill artifact of the image. The main reason for this is that the reconstruction point is too far from the source and the edge ray itself is of low quality due to error, so it does not enter the superposition calculation. Figure 23b For the same reconstruction point of the two spheres, one ray is a valid point and enters the reconstruction, while the other ray is an invalid point and does not enter the reconstruction.

[0319] During the back-projection process, each pixel of the target coordinate system XY plane (each axial map) and each acquisition rearrangement perspective is traversed to determine the corresponding starting and ending positions of the horizontal motor under the rearrangement perspective, and the upper and lower boundaries of the effective Z plane under the acquisition rearrangement perspective are determined according to the semi-extension range. The pixel points on XY are converted to the virtual coordinate system to determine the UV coordinate value, and the channel position is determined according to the U coordinate value, that is, the position of the horizontal channel of the acquisition rearrangement data (only the effective horizontal channel is calculated); according to the V-direction position of the ray, the V coordinate value determined by converting the pixel points on XY to the virtual coordinate system and the V position (i.e., Z position) of the desired image plane, the row number is determined, that is, the position of the longitudinal channel of the acquisition rearrangement data (only the effective longitudinal channel is calculated). And according to the position of the effective longitudinal channel of the acquisition rearrangement data, the Q value weight W is calculated according to the following formula (six): Q (q).

[0320] Iterate the rearranged data corresponding to the corresponding XYZ values ​​and multiply them by the corresponding weights. The weights in the same direction are weighted averaged, that is, 180-degree interpolation is performed (compared to the 360-degree interpolation of coarse reconstruction, 180-degree interpolation has more Figure 26 Finally, the superimposed axial image is multiplied by the angle weight to obtain the back-projected image.

[0321] The post-reconstruction stage includes bone sclerosis correction, CT value correction and layer thickness processing.

[0322] Bone sclerosis correction is actually the optimization of high-density body artifacts. It is generally divided into two methods, one is pre-processing and the other is post-processing. Pre-processing generally requires modeling based on a preset model. Since the bone density and morphological steps of each patient are different, there will be certain deviations and the application is limited. Post-processing is also roughly divided into two types. One is the orthographic projection reconstruction method of the remaining tissues after removing the bone tissue, and the other is the orthographic projection reconstruction method of the bone tissue after extracting the bone tissue. In the embodiment of the present disclosure, first, the bone tissue segmentation adopts segmented correction segmentation, and secondly, the orthographic projection is not a projection according to the rotation of the axial scan, but a projection method of the spiral scan. The steps are as follows:

[0323] First, segment the bone tissue. (Note that the CT value -1000 correction has not yet been performed at this step, so these values ​​are 1000 higher than normal.) Bone images are segmented using a segmented approach. Because the CT values ​​of bone images vary slightly across each axial layer, each part, and each individual, accurate segmentation using a single threshold is difficult. Therefore, a two-step approach is employed. First, the bone tissue is extracted, defined as the portion with CT values ​​between hu1 and hu2 (e.g., 850 to 3000). This portion is intended to encompass as many conditions as possible, with a relatively wide range. The CT values ​​of the remaining portions are set to 0, meaning that all bone areas are included in hu1 to hu2. The CT values ​​between hu1 and hu2 are then segmented. Here, we divide them into three segments, but multiple segments are possible. The first segment, hu1 to hua (e.g., 850 to 1150), has CT values ​​between brain and bone tissue. The second segment, hua to hub (e.g., 1150 to 2075), covers the CT value range for bone tissue. The third segment, hub to hu2 (e.g., 2075 to 3000), covers the CT value range between bone tissue and high-density volumes. By presetting normalization coefficients for each node, for example, hu1 corresponds to s1 (e.g., 0), hua corresponds to s2 (e.g., 0.4), hub corresponds to s3 (e.g., 0.7), and hu2 corresponds to s4 (e.g., 1), the coefficient s is then calculated for each point in the segmented bone tissue portion, hu1 to hu2. Si is the low-end coefficient value of the segment in which the point lies, sj is the high-end coefficient value of the segment in which the point lies, hu is the CT value of the point, hui is the low-end CT value of the segment in which the point lies, and huj is the high-end CT value of the segment in which the point lies. The CT value of the bone tissue portion is multiplied by this coefficient to obtain the final segmented bone tissue.

[0324] Secondly, perform orthographic projection on the bone tissue. First, the projection path is as follows: Figure 15As shown, it can be regarded as a parallel beam of rays with a cone angle. Its scanning path follows a spiral line as a whole. The advantage of this is that a pair of original images have projections of this position (equivalent to the result of rearrangement of the projection map), which can be directly matched. Secondly, this omits the step of rearrangement in the reconstruction, thereby increasing the speed. Thirdly, the positive projection of this position is downsampled to obtain the projection line according to half of the rearrangement (the sampling can be reduced according to the demand for the amount of calculation), which can reduce the amount of calculation, and there is no need to process the projection of the original image. It is only necessary to shift the x-coordinate one position to the left when taking the value of the original image.

[0325] Iterate based on the ray passing through the axial image pixels and the ray length weight to obtain the total value of each ray in the orthographic projection. The ray length weight is inversely proportional to the ray length, that is, the longer the ray in the voxel, the smaller the influence of the point on the axial surface (the orthographic weight is the largest). That is, the weight is AG / IJ (such as Figure 16b , where AG is the most diagonal length of the voxel and IJ is the path length of one of the rays passing through the voxel).

[0326] Next, we correct for bone sclerosis parameters. We perform a weighted correction on the bone projection image. First, we subtract the orthographic projection from the original image projection (of course, this involves a left shift, i.e., division by 2). The result is the projection of the tissue without the bone tissue. This projection is then divided by the water reference to obtain the equivalent length of each projection pixel relative to water. This equivalent length is used to determine the artifact correction coefficient, which is then used to correct the orthographic projection.

[0327] Finally, the corrected bone projection image is reconstructed (there's no need to rearrange the previous steps in the reconstruction process, as the projection image's scanning path is always the rearranged path) to obtain a new bone reconstruction image. The original image and the new bone reconstruction image are superimposed to obtain the final corrected image.

[0328] During the CT value correction process, a 1000 offset correction is performed, which means that the values ​​of all pixels are subtracted by 1000.

[0329] During the layer thickness processing, layer thickness superposition calculation is performed. The effective layer thickness is obtained according to the reconstruction interval, display layer spacing and preset layer thickness, and layer thickness superposition calculation is performed. Figure 19a and Figure 19b The data is stored in the specified memory (the space allocated in advance during pre-reconstruction).

[0330] Repeat the above steps until all the collected data are processed. The time for one acquisition cycle is set to 1 second or 2 seconds according to the user, and the time for one rough reconstruction cycle does not exceed 0.2s (ordinary computer configuration, high-end computer configuration is much less than 0.2a). The reconstruction time is much shorter than the acquisition time, so the axial reconstruction image of the current shooting position can be displayed almost in real time, thereby achieving real-time reconstruction display, which is convenient for doctors to understand the shooting situation in time. If there is a problem with the shooting position, it can be canceled in time without having to confirm it after the shooting is completed.

[0331] The overall reconstruction process includes coarse reconstruction, fine reconstruction, and post-reconstruction. The coarse reconstruction, fine reconstruction, and post-reconstruction are displayed simultaneously. The following steps are included:

[0332] After pressing the ray collection button, the first step is delayed exposure. For the delayed exposure interface, refer to Figure 5a As shown, the left picture is the positioning film picture, and the right picture is the spiral scan picture. At this time, the right picture is completely black.

[0333] Pre-reconstruction can be completed before acquisition, either during this delay or through simultaneous multi-threading. This takes less than 4 seconds. If the delay is too short, it's recommended to do it beforehand to ensure it doesn't interfere with subsequent simultaneous acquisition and reconstruction. In this example, it's actually done after clicking the device ready button. After device preparation, some doctors need to check, such as ensuring the patient hasn't moved. The delayed exposure interface appears when the device is ready, and the delay begins when you click Capture.

[0334] After capturing the topography and setting the spiral scan parameters, you can click Device Preparation. Pre-reconstruction is included in this process, meaning it's completed before acquisition. This doesn't take up acquisition and reconstruction time, and the user won't notice it.

[0335] The results are displayed in the image on the right after one scan, indicating simultaneous acquisition and coarse reconstruction. In this example, coarse reconstruction is set after one scan. Depending on the acquisition time setting, if the acquisition time is too long, you can shorten it accordingly (e.g., half a scan). In this example, acquisition takes 1-2 seconds per scan, and coarse reconstruction takes only 0.1+ seconds, which is more than sufficient.

[0336] The number of frames collected per revolution depends on the computer's configuration. This example sets the frame count to less than 2000 (1440 frames per revolution). Higher resolutions are possible on higher resolutions, while lower resolutions can be used to ensure real-time performance. This can be considered as segmenting the image to achieve simultaneous reconstruction. A callback function informs the system software of the number of reconstructed images. During coarse reconstruction, the callback function returns the total number of axial images in the coarse reconstruction and the number of axial images currently reconstructed. Because coarse reconstruction also takes acquisition time into account (reconstruction time is negligible relative to acquisition), the number of reconstructed axial images is displayed based on a 2-second acquisition time. For example, if two new images are reconstructed, one image is displayed per 1 second. The coarse reconstruction interval is set to -1 because the acquisition time is the primary time, not the reconstruction time (coarse reconstruction truncates the reconstruction time to be less than the acquisition time).

[0337] During the post-reconstruction process, the post-reconstruction module includes bone sclerosis, CT value intercept correction, and slice thickness calculation. The post-reconstruction time is synchronized with the fine reconstruction callback. This means that after the fine reconstruction interval is displayed, the post-reconstruction is complete. Therefore, it is recommended to refresh the displayed data after the fine reconstruction progress bar is fully completed.

[0338] Using the spiral CT reconstruction method provided by the embodiment of the present disclosure, in order to achieve real-time reconstruction (reconstruction while acquiring), spiral CT currently generally adopts methods such as using a faster processor (high-end reconstruction computer), uploading reconstruction to a cloud server, or a new reconstruction algorithm. In this example, the reconstruction process is segmented, low-precision coarse reconstruction is used for display during acquisition, and high-precision fine reconstruction is used to save the reconstruction results. This ensures the needs of real-time acquisition and reconstruction, and the final accuracy. In addition, a pre-reconstruction stage is added, and part of the reconstruction content is completed before acquisition. The spiral CT reconstruction method provided in this example can achieve real-time reconstruction even with a low-end computer, and does not affect the doctor's real-time supervision of the shooting process. Since there may be emergencies such as patient movement during the shooting process, it is necessary to detect in real time whether the scanning position is normal.

[0339] Before coarse reconstruction, a judgment is made, allowing the user to enter the coarse reconstruction step. Depending on the operator's selected parameters, the need for coarse reconstruction is determined. If the reconstruction computational complexity is minimal, fine reconstruction is performed directly. Coarse reconstruction also includes calibration, meaning scanning, acquisition, calibration, reconstruction, and display are all completed in real time.

[0340] Setting the UV virtual coordinate system: The entire coordinate system is an approximately rectangular surface, with the U axis as the curve axis. This coordinate system can be considered as two systems. One is the pre-UV virtual coordinate system, where curve fitting interpolation is performed along the U axis (not the final U axis, which is the axis of the helix) during rearrangement, not linear interpolation. The other is the final UV virtual coordinate system, where the re-shot points are interpolated according to the final UV coordinate system. This means that slanted points are made straight, improving the accuracy of the back projection.

[0341] Rapid back-projection of coarse reconstruction and back-projection of fine reconstruction significantly improve speed. The coarse reconstruction algorithm can be adjusted based on acquisition time. For example, if the acquisition time is 1-2 seconds, which is significantly longer than the time required for one coarse reconstruction cycle, the coarse reconstruction can be split into multiple cycles, allowing it to be completed entirely within the acquisition time. If acquisition time is subsequently increased, such as the industry's fastest spiral CT at 0.28 seconds, the first six steps of coarse reconstruction can be split into shorter frames, such as half a cycle or even a single frame, with back-projection performed only for a full cycle. This ensures that coarse reconstruction and acquisition are performed simultaneously, without requiring separate time constraints.

[0342] The spiral CT reconstruction provided in this example effectively reduces cone beam artifacts and also improves windmill artifacts. Collimation is usually added by hardware, but here software simulation is used to increase collimation to achieve the same effect of removing cone beam artifacts.

[0343] It should be understood that, although the various steps in the flowcharts involved in the various embodiments described above are displayed in sequence according to the instructions of the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and these steps can be executed in other orders. Moreover, at least a portion of the steps in the flowcharts involved in the various embodiments described above can include multiple steps or multiple stages, and these steps or stages are not necessarily executed and completed at the same time, but can be executed at different times, and the execution order of these steps or stages is not necessarily to be carried out in sequence, but can be executed in turn or alternately with other steps or at least a portion of steps or stages in other steps.

[0344] Based on the same inventive concept, embodiments of the present application further provide a CT reconstruction device for implementing the aforementioned CT reconstruction method. The solution provided by this device is similar to the solution described in the aforementioned method. Therefore, the specific limitations of one or more CT reconstruction device embodiments provided below can be found in the above-described limitations of the CT reconstruction method and will not be further elaborated here.

[0345] In an exemplary embodiment, Figure 24As shown, a CT reconstruction device is provided, including: a determination module 2402, a coarse reconstruction module 2404 and a fine reconstruction module 2406, wherein:

[0346] A determination module 2402 is configured to obtain a current reconstruction interval and determine a reconstruction strategy based on the current reconstruction interval;

[0347] a coarse reconstruction module 2404 configured to, when the reconstruction strategy is the first reconstruction strategy, perform CT reconstruction on the collected projection data using a coarse reconstruction method each time projection data of a preset time interval is collected, to obtain a first CT reconstructed image, and to display the first CT reconstructed image in real time on a display interface;

[0348] a fine reconstruction module 2406, configured to perform CT reconstruction on the projection data using a fine reconstruction method to obtain a second CT reconstructed image, and store the second CT reconstructed image;

[0349] The reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method.

[0350] By using the CT reconstruction device provided in the embodiments of the present application, a coarse reconstruction method with low reconstruction cost and less time consumption can be used to perform CT reconstruction and display the reconstructed image in real time during the acquisition process, and a fine reconstruction method with higher reconstruction cost but higher reconstruction accuracy can be used to perform high-precision CT reconstruction. The high-precision CT reconstructed image is obtained and stored, so that the user can not only obtain a high-precision CT reconstructed image, but also obtain the corresponding reconstruction results in real time during the scanning and acquisition of projection data. Based on the reconstruction results, the status of the user currently performing the CT scan can be determined, such as whether the user has moved or whether the user's position is appropriate. Then, corresponding judgment operations can be performed based on the user's status, thereby improving the CT scanning effect and reconstruction accuracy.

[0351] In one embodiment, the determining module 2402 is further configured to:

[0352] When the current reconstruction interval is less than the interval threshold, determining the reconstruction strategy to be the first reconstruction strategy; and / or when the current reconstruction interval is greater than or equal to the interval threshold, determining the reconstruction strategy to be the second reconstruction strategy;

[0353] The first reconstruction strategy includes the coarse reconstruction method and the fine reconstruction method, and the second reconstruction strategy includes the fine reconstruction strategy.

[0354] In one embodiment, the apparatus further comprises:

[0355] A pre-reconstruction module is used to obtain reconstruction parameters before receiving the device preparation instruction and starting scanning to collect projection data, pre-allocate video memory based on the reconstruction parameters, and read the reconstruction parameters into the GPU;

[0356] The reconstruction parameters include at least one of parameter data, reconstruction layer number, horizontal motor movement starting position, horizontal motor speed, Q value weight curve, and windowing filter function.

[0357] In one embodiment, the coarse reconstruction module 2404 is further configured to:

[0358] performing filtering processing on the projection data to obtain filtered data;

[0359] Perform back-projection processing on the filtered data using a preset back-projection algorithm to obtain a back-projection image, wherein the preset back-projection algorithm includes any one of the following:

[0360] The reconstruction interval used in the back projection is a preset reconstruction interval, which is greater than the reconstruction interval used by the fine reconstruction algorithm, 360-degree interpolation is used for interpolation processing, the horizontal motor position is calculated based on the horizontal motor speed, and the layer thickness processing process is skipped.

[0361] In one embodiment, the coarse reconstruction module 2404 is further configured to:

[0362] The projection data are rearranged using a pre-established initial virtual coordinate system, wherein the initial virtual coordinate system has the detector illuminated by the central radiation source as the origin, the spiral line as the horizontal axis, and the horizontal displacement bed direction as the vertical axis;

[0363] The rearranged projection data is filtered according to the windowing filter kernel to obtain filtered data.

[0364] In one embodiment, the coarse reconstruction module 2404 is further configured to:

[0365] Rearranging the projection data of the fan beam into the projection data of the parallel beam by rearranging the direction angle of the projection data;

[0366] According to the amount of up-sampled data and the interval of the direction angle rearrangement, the projection data of the parallel beam are subjected to three-dimensional Hermite interpolation to obtain a three-dimensional surface, and the three-dimensional surface is subjected to coordinate interpolation processing perpendicular to the initial virtual coordinate system to obtain rearranged projection data.

[0367] In one embodiment, the coarse reconstruction module 2404 is further configured to:

[0368] Determining grid points on a three-dimensional surface and calculating a rotation matrix for each of the grid points;

[0369] Using the rotation matrix, the coordinates of each grid point are transformed from the initial virtual coordinate system to a target virtual coordinate system, where the horizontal axis of the target virtual coordinate system is horizontal;

[0370] Interpolation processing is performed on each of the grid points in the target virtual coordinate system to obtain rearranged projection data.

[0371] In one embodiment, the coarse reconstruction module 2404 is further configured to:

[0372] Performing interpolation processing on each of the grid points in the target virtual coordinate system to obtain an interpolation result;

[0373] The interpolation result is filtered according to the initial virtual coordinate system to obtain rearranged projection data.

[0374] In one embodiment, the apparatus further comprises:

[0375] A storage module, configured to store the filtered data in a video memory;

[0376] The fine reconstruction module 2406 is specifically configured to:

[0377] Acquiring filtered data from the video memory, performing back-projection processing on the filtered data to obtain a back-projection image, and extracting bone tissue data from the back-projection image;

[0378] Performing orthographic projection on the bone tissue data to obtain bone projection data;

[0379] taking the difference between the projection data and the bone projection data as bone-removed tissue data;

[0380] A second CT reconstructed image is obtained according to the bone-removed tissue data and the back-projected image.

[0381] In one embodiment, the fine reconstruction module 2406 is specifically configured to:

[0382] Extracting bone tissue pixels in the back-projection image, wherein the CT values ​​corresponding to the bone tissue pixels are within the range of bone tissue CT values;

[0383] For any bone tissue pixel point, determine the bone tissue segment corresponding to the bone tissue pixel point based on the CT value of the bone tissue pixel point, and determine the normalization coefficient of the bone tissue pixel point based on the upper and lower CT values ​​of the bone tissue segment, the normalization coefficients corresponding to the upper and lower CT values, and the CT value of the bone tissue pixel point, where the bone tissue CT value range is composed of multiple bone tissue segments;

[0384] Bone tissue data is obtained based on the normalization coefficient and CT value of each bone tissue pixel point.

[0385] In one embodiment, the fine reconstruction module 2406 is specifically configured to:

[0386] According to each pixel point in the axial image corresponding to the bone tissue data and the ray length weight corresponding to each pixel point, the CT value of each pixel point is weightedly superimposed to obtain bone projection data;

[0387] The ray length weight is negatively correlated with the length of the ray within the voxel.

[0388] In one embodiment, the fine reconstruction module 2406 is specifically configured to:

[0389] The ratio of the bone-removed tissue data to the water reference is used as the equivalent length of each projection pixel point in the bone-removed tissue data relative to water;

[0390] Performing weighted correction on the bone projection data using an artifact correction coefficient corresponding to the equivalent length to obtain corrected bone projection data;

[0391] After reconstructing the corrected bone projection data, the reconstruction result is superimposed on the back-projection image to obtain a second CT reconstructed image.

[0392] Each module in the aforementioned CT reconstruction device may be implemented in whole or in part through software, hardware, or a combination thereof. Each module may be embedded in or independent of a processor in a computer device in the form of hardware, or may be stored in a memory in the computer device in the form of software, so that the processor can call and execute the corresponding operations of each module.

[0393] In an exemplary embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as shown in FIG. Figure 27As shown. The computer device includes a processor, a memory, an input / output interface, a communication interface, a display unit and an input device. The processor, the memory and the input / output interface are connected via a system bus, and the communication interface, the display unit and the input device are connected to the system bus via the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, a mobile cellular network, NFC (near field communication) or other technologies. When the computer program is executed by the processor, a CT reconstruction method is implemented. The display unit of the computer device is used to form a visually visible picture, which can be a display screen, a projection device or a virtual reality imaging device. The display screen can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the computer device casing, or an external keyboard, touchpad or mouse.

[0394] Those skilled in the art will understand that Figure 27 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.

[0395] In one embodiment, a computer device is further provided, including a memory and a processor. The memory stores a computer program, and the processor implements the steps in the above method embodiments when executing the computer program.

[0396] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps in the above-mentioned method embodiments are implemented.

[0397] In one embodiment, a computer program product is provided, including a computer program, which implements the steps in the above method embodiments when executed by a processor.

[0398] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0399] Those skilled in the art will understand that all or part of the processes in the above-mentioned embodiment methods can be implemented by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, database or other media used in the embodiments provided in this application may include at least one of non-volatile and volatile memory. Non-volatile memory may include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory may include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM). The database involved in the various embodiments provided herein may include at least one of a relational database and a non-relational database. Non-relational databases may include, but are not limited to, distributed databases based on blockchains. The processor involved in the various embodiments provided herein may be, but are not limited to, a general-purpose processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic unit, a data processing logic unit based on quantum computing, and the like.

[0400] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0401] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present application. It should be noted that a person of ordinary skill in the art may make various modifications and improvements without departing from the spirit of the present application, and these modifications and improvements fall within the scope of protection of the present application. Therefore, the scope of protection of the present application shall be determined by the appended claims.

Claims

1. A CT reconstruction method, applied to spiral CT, characterized in that: The method comprises: Obtaining a current reconstruction interval, and determining a reconstruction strategy according to the current reconstruction interval; When the reconstruction strategy is the first reconstruction strategy, each time projection data of a preset time interval is collected, a coarse reconstruction method is used to perform CT reconstruction on the collected projection data to obtain a first CT reconstructed image, and the first CT reconstructed image is displayed in real time on the display interface; Performing CT reconstruction on the projection data by adopting a fine reconstruction method to obtain a second CT reconstructed image, and storing the second CT reconstructed image; The reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method.

2. The method according to claim 1, characterized in that The determining of the reconstruction strategy according to the current reconstruction interval includes: In the case where the current reconstruction interval is less than the interval threshold, determining the reconstruction strategy to be the first reconstruction strategy; and / or, in the case where the current reconstruction interval is greater than or equal to the interval threshold, determining the reconstruction strategy to be the second reconstruction strategy; The first reconstruction strategy includes the coarse reconstruction method and the fine reconstruction method, and the second reconstruction strategy includes the fine reconstruction strategy.

3. The method according to claim 1 or 2, characterized in that: The method also includes: After receiving the device preparation instruction and before starting scanning to collect projection data, obtaining reconstruction parameters, pre-allocating video memory based on the reconstruction parameters, and reading the reconstruction parameters into the GPU; The reconstruction parameters include at least one of parameter data, reconstruction layer number, horizontal motor movement start position, horizontal motor speed, Q value weight curve, and windowing filter function.

4. The method according to claim 3, characterized in that: The method of performing CT reconstruction on the collected projection data in a coarse reconstruction manner to obtain a first CT reconstructed image includes: Performing filtering on the projection data to obtain filtered data; The filtered data is back-projected using a preset back-projection algorithm to obtain a back-projection image, wherein the preset back-projection algorithm includes any one of the following: The reconstruction interval used by the back projection is a preset reconstruction interval, which is greater than the reconstruction interval used by the fine reconstruction algorithm, 360-degree interpolation is used for interpolation processing, the horizontal motor position is calculated based on the horizontal motor speed, and the layer thickness processing process is skipped.

5. The method according to claim 4, characterized in that The filtering of the projection data to obtain filtered data includes: The projection data are rearranged using a pre-established initial virtual coordinate system, wherein the initial virtual coordinate system takes the detector irradiated by the central radiation source as the origin, the spiral line as the horizontal axis, and the horizontal displacement bed direction as the vertical axis; The rearranged projection data is filtered according to the windowing filter kernel to obtain filtered data.

6. The method according to claim 5, characterized in that The method of rearranging the projection data by using a pre-established initial virtual coordinate system includes: Rearranging the projection data of the fan beam into the projection data of the parallel beam by rearranging the direction angle of the projection data; According to the up-sampled data amount and the interval of the direction angle rearrangement, the projection data of the parallel beam is subjected to three-fold Hermite interpolation to obtain a three-dimensional surface, and the three-dimensional surface is subjected to coordinate interpolation processing perpendicular to the initial virtual coordinate system to obtain rearranged projection data.

7. The method according to claim 6, characterized in that The performing coordinate interpolation processing perpendicular to the initial virtual coordinate system on the three-dimensional curved surface to obtain rearranged projection data includes: Determine the grid points in the three-dimensional surface and calculate the rotation matrix of each of the grid points; The rotation matrix is ​​used to transform the coordinates of each of the grid points from the initial virtual coordinate system to a target virtual coordinate system, wherein the horizontal axis of the target virtual coordinate system is horizontal; Interpolation processing is performed on each of the grid points in the target virtual coordinate system to obtain rearranged projection data.

8. The method according to claim 7, characterized in that The interpolation processing is performed on each of the grid points in the target virtual coordinate system to obtain rearranged projection data, including: Performing interpolation processing on each of the grid points in the target virtual coordinate system to obtain an interpolation result; The interpolation result is filtered according to the initial virtual coordinate system to obtain rearranged projection data.

9. The method according to any one of claims 4 to 8, characterized in that The method further comprises: storing the filtered data in a video memory; The step of performing CT reconstruction on the projection data in a fine reconstruction manner to obtain a second CT reconstructed image includes: Acquire filter data from the video memory, perform back-projection processing on the filter data to obtain a back-projection image, and extract bone tissue data from the back-projection image; Performing orthographic projection on the bone tissue data to obtain bone projection data; Taking the difference between the projection data and the bone projection data as bone removal tissue data; A second CT reconstructed image is obtained according to the bone removal tissue data and the back-projection image.

10. The method according to claim 9, characterized in that The step of extracting bone tissue data from the back-projection image comprises: Extracting bone tissue pixel points in the back-projection image, wherein the CT values ​​corresponding to the bone tissue pixel points are within the range of bone tissue CT values; For any bone tissue pixel point, determine the bone tissue segment corresponding to the bone tissue pixel point according to the CT value of the bone tissue pixel point, and determine the normalization coefficient of the bone tissue pixel point according to the upper limit CT value and the lower limit CT value of the bone tissue segment, the normalization coefficients corresponding to the upper limit CT value and the lower limit CT value, and the CT value of the bone tissue pixel point, wherein the bone tissue CT value range is composed of a plurality of the bone tissue segments; Bone tissue data is obtained according to the normalization coefficient and CT value of each bone tissue pixel point.

11. The method according to claim 9, characterized in that The step of performing orthographic projection on the bone tissue data to obtain bone projection data includes: According to each pixel point in the axial image corresponding to the ray passing through the bone tissue data and the ray length weight corresponding to each pixel point, the CT value of each pixel point is weightedly superimposed to obtain the bone projection data; The ray length weight is negatively correlated with the length of the ray in the voxel.

12. The method according to claim 9, characterized in that The step of obtaining a second CT reconstructed image according to the bone removal tissue data and the back-projection image comprises: The ratio of the bone-removed tissue data to the water reference is used as the equivalent length of each projection pixel point in the bone-removed tissue data relative to water; Performing weighted correction on the bone projection data using the artifact correction coefficient corresponding to the equivalent length to obtain corrected bone projection data; After the corrected bone projection data is reconstructed, the reconstruction result is superimposed on the back-projection image to obtain a second CT reconstructed image.

13. A CT reconstruction device, applied to spiral CT, characterized in that: The device comprises: A determination module, configured to obtain a current reconstruction interval and determine a reconstruction strategy according to the current reconstruction interval; a coarse reconstruction module, configured to, when the reconstruction strategy is the first reconstruction strategy, perform CT reconstruction on the collected projection data by using a coarse reconstruction method each time projection data of a preset time interval is collected to obtain a first CT reconstructed image, and display the first CT reconstructed image in real time on a display interface; A fine reconstruction module, used for performing CT reconstruction on the projection data in a fine reconstruction manner to obtain a second CT reconstructed image, and storing the second CT reconstructed image; The reconstruction cost of the coarse reconstruction method is lower than the reconstruction cost of the fine reconstruction method.

14. A computer device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the method according to any one of claims 1 to 12 are implemented.

15. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

16. A computer program product comprising a computer program, characterized in that When the computer program is executed by a processor, the steps of the method according to any one of claims 1 to 12 are implemented.

Images