Motion artifact correction CBCT reconstruction method and system under animal unconstrained scanning

Abstract

The application discloses a motion artifact correction CBCT reconstruction method and system under animal unconstrained scanning, and comprises the following steps: acquiring unconstrained scanning data formed by multi-angle X-ray scanning of an animal in an unconstrained state; distinguishing projection changes caused by rotation of an imaging system from structural changes caused by animal motion based on the corresponding relationship of animal structures in adjacent projections and cross-angle comparable projections in a projection sequence, and determining a motion state sequence; transforming animal structures in each projection to a unified reference state according to the motion state sequence and acquisition attributes, establishing a spatial corresponding relationship between projection pixels and reconstructed space voxels, and obtaining motion-consistent projection data; back-projecting the motion-consistent projection data to a reconstruction space according to the spatial corresponding relationship, fusing projection contributions at different acquisition time sequences, and generating CBCT volume data; and outputting the CBCT volume data after motion artifact correction. The application is beneficial to improving the structural continuity and size representation reliability of animal CBCT images.

Description

Technical Field

[0001] This application relates to the field of X-ray imaging and computed tomography image reconstruction technology, and in particular to a method and system for motion artifact correction in CBCT reconstruction of animals under unconstrained scanning. Background Technology

[0002] Cone-beam computed tomography (CBCT) uses an X-ray source and a flat-panel detector to acquire multi-angle projections around the subject and obtains three-dimensional volumetric data of the subject through image reconstruction. This technique can be used for non-destructive imaging of animal bones, tissue structures, and lesion areas.

[0003] During animal scanning, to minimize the impact of positional changes on projection acquisition, immobilization devices, postural restraints, or anesthesia are typically used to keep the animal relatively still. However, these methods may alter the animal's natural physiological state and are difficult to completely suppress respiration, trunk movement, and limb movements. For animals that cannot maintain a fixed position continuously, the projection content may change significantly at different acquisition times.

[0004] Current CBCT reconstructions are typically based on the assumption that the subject remains stationary during the scanning cycle. When the animal moves, the spatial structures represented by projections from different angles become inconsistent, and projection contributions may accumulate at incorrect locations, resulting in ghosting, streaks, and blurred structural boundaries. Therefore, how to reduce projection spatial mismatch caused by movement under unconstrained animal scanning conditions is a technical problem that needs to be solved. Summary of the Invention

[0005] This application proposes a method and system for motion artifact correction in CBCT reconstruction during unconstrained animal scanning. The aim is to correct projection spatial mismatch caused by animal movement during scanning without relying on the animal remaining stationary, thereby improving the structural continuity of CBCT volumetric data. This is beneficial for improving the structural continuity and dimensional representation reliability of animal CBCT images.

[0006] In a first aspect, embodiments of this application provide a CBCT reconstruction method for motion artifact correction under unconstrained animal scanning, executed by a CBCT reconstruction device, comprising:

[0007] Acquire unconstrained scan data of an animal under unconstrained conditions through multi-angle X-ray scanning; the unconstrained scan data includes a projection sequence and acquisition attributes corresponding to each projection, the acquisition attributes including acquisition timing and imaging geometric parameters;

[0008] Based on the correspondence between animal structures in adjacent projections and cross-angle comparable projections in the projection sequence, the projection changes caused by the rotation of the imaging system and the structural changes caused by animal movement are distinguished to determine the motion state sequence; the cross-angle comparable projection is a non-adjacent projection with a structurally comparable region determined according to the imaging geometric parameters.

[0009] Based on the motion state sequence and the acquisition attributes, the animal structures in each projection are transformed to a unified reference state, and a spatial correspondence between the projection pixels and the voxels in the reconstruction space is established to obtain motion-consistent projection data. The motion-consistent projection data is back-projected into the reconstruction space according to the spatial correspondence, and the projection contributions from different acquisition time sequences are fused to generate CBCT volume data.

[0010] Output the CBCT volumetric data after motion artifact correction.

[0011] As can be seen, in this embodiment, adjacent projections and cross-angle comparable projections are used to characterize the changes in animal structure with the acquisition time sequence, and imaging geometric parameters are used to distinguish the projection changes caused by the rotation of the imaging system and the animal's movement. On this basis, the animal structure under different movement states is normalized to a unified reference state, so that the projection contributions of different acquisition time sequences can correspond to the same or consistent anatomical positions, thereby reducing the accumulation of movement inconsistencies in the reconstruction space and suppressing ghosting, streaks, and structural boundary blurring.

[0012] In one possible example, based on the correspondence between animal structures in adjacent projections and comparable projections across angles in the projection sequence, the projection changes caused by the rotation of the imaging system and the structural changes caused by animal movement are distinguished to determine the motion state sequence, including:

[0013] Based on the correspondence between animal structures in adjacent projections and cross-angle comparable projections in the projection sequence, animal outline features and internal structure features are extracted to form a projection feature sequence.

[0014] Based on the correspondence between adjacent projection features and cross-angle comparable projection features in the projection feature sequence, candidate motion observations are generated.

[0015] The continuity of the candidate motion observations is constrained by the acquisition timing, and the geometric change components corresponding to the rotation of the imaging system are eliminated according to the imaging geometric parameters to obtain the motion state sequence.

[0016] As can be seen, in this example, animal outline features can characterize overall posture and changes in external boundaries, while internal structural features can characterize positional changes in local tissues or bony structures. By combining these two types of features to form a projection feature sequence, and constraining candidate motion observations with acquisition timing and imaging geometry parameters, the possibility of misinterpreting normal perspective changes as animal movement can be reduced.

[0017] In one possible example, candidate motion observations are generated based on the correspondence between adjacent projection features and cross-angle comparable projection features in the projection feature sequence, including:

[0018] Structural matching is performed on adjacent projection features and cross-angle comparable projection features in the projection feature sequence to obtain candidate corresponding feature pairs;

[0019] Based on the displacement of the candidate corresponding feature pairs on the projection plane, the corresponding imaging geometric parameters, and the matching differences, a local motion observation carrying a matching confidence state is generated.

[0020] The local motion observations are aggregated according to spatial adjacency and the matching confidence state to generate the candidate motion observations.

[0021] As can be seen, in this example, first converting the structural matching results into local motion observations carrying confidence states, and then aggregating them according to spatial adjacency relationships, can suppress the impact of isolated mismatches on overall motion estimation and preserve the possible motion differences between different animal regions.

[0022] In one possible example, structural matching is performed on adjacent projection features and cross-angle comparable projection features in the projection feature sequence to obtain candidate corresponding feature pairs, including:

[0023] Based on the local grayscale distribution, contour morphology, and neighborhood structure of the adjacent projection features and the cross-angle comparable projection features, candidate feature description information is generated;

[0024] Based on the acquisition angle relationship and time sequence relationship corresponding to each projection, the allowed search area is determined for the candidate feature description information;

[0025] Within the allowed search area, bidirectional matching is performed on the candidate feature description information, and the bidirectional matching results are filtered using neighborhood topological consistency to obtain the candidate corresponding feature pairs.

[0026] As can be seen, in this example, by using the acquisition angle relationship and temporal relationship to limit the matching search range, and combining bidirectional matching and neighborhood topology consistency for filtering, the incorrect correspondence caused by duplicate textures, local occlusion or projection overlap can be reduced.

[0027] In one possible example, based on the acquisition angle relationship and temporal relationship corresponding to each projection, the allowed search region is determined for the candidate feature description information, including:

[0028] Based on the acquisition angle relationship and the temporal relationship, the angular interval and time interval between the source projection and the target projection to which the candidate feature description information belongs are determined, and an initial search boundary is formed;

[0029] Based on the imaging geometry parameters, the effective area of ​​the animal projection is mapped to the initial search boundary, and pixel positions that do not have structural comparability are excluded to obtain the constrained search boundary;

[0030] Based on the confirmed matching positions of the candidate feature description information in the front and back projections, a matching displacement trend is generated;

[0031] Adjust the constraint search boundary according to the matching displacement trend to determine the allowed search area.

[0032] As can be seen, in this example, the search area is simultaneously constrained by the imaging geometry, the effective projection area, and the existing matching displacement trend, and can be adjusted according to the animal's movement trend, avoiding indiscriminate searching in the entire projection and causing matching ambiguity.

[0033] In one possible example, adjusting the constraint search boundary according to the matching displacement trend to determine the allowed search area includes:

[0034] The matching displacement trend is projected onto the constraint search boundary to obtain the predicted matching region;

[0035] The range of motion uncertainty is determined based on the degree of change of the matched displacement trend with the time interval;

[0036] The predicted matching region is expanded according to the range of motion uncertainty, and the intersection of the expanded predicted matching region and the constraint search boundary is taken to obtain the allowed search region.

[0037] As can be seen in this example, the predicted matching region reflects the identified direction and amplitude of motion, while the motion uncertainty range reflects the stability of the motion trend over time. Restricting the expanded predicted matching region within the constraint search boundary balances matching coverage under rapid motion conditions with matching accuracy under geometric constraints.

[0038] In one possible example, based on the motion state sequence and the acquired attributes, the animal structures in each projection are transformed to a unified reference state, establishing a spatial correspondence between projected pixels and reconstructed spatial voxels, resulting in motion-consistent projection data, including:

[0039] Based on the motion state sequence and the acquisition attributes, the animal projection area in each projection is divided into motion coupling regions, and the overall motion component and local deformation component of each motion coupling region are determined.

[0040] The overall motion components and the local deformation components are associated with the acquisition time sequence and transformed to the unified reference state to generate a partitioned motion transformation field;

[0041] The spatial correspondence between the projected pixels and the reconstructed spatial voxels is established using the partitioned motion transformation field, and each projection is resampled to obtain the motion-consistent projection data.

[0042] As can be seen, in this example, by expressing the overall motion and local deformation separately through the motion coupling region, we can avoid treating the motion of different parts of the animal as a single rigid displacement, and enable differentiated compensation for respiratory deformation, trunk displacement and local limb movements.

[0043] In one possible example, based on the motion state sequence and the acquired attributes, the animal projection area in each projection is divided into motion coupling regions, and the overall motion component and local deformation component of each motion coupling region are determined, including:

[0044] Based on the motion state sequence and the collected attributes, the motion response at local locations is determined within the animal projection area of ​​each projection, forming a local motion response distribution;

[0045] The motion coupling region is obtained by merging adjacent local positions that are related in motion direction and whose motion amplitudes differ and satisfy the coupling condition.

[0046] The common motion component is extracted from the local motion response of each of the motion coupling regions as the overall motion component, and the spatial change component that deviates from the common motion component is taken as the local deformation component.

[0047] As can be seen, in this example, the motion coupling region is formed by the correlation between local motion responses, without relying on pre-completed precise anatomical region segmentation. Therefore, the correspondence between the motion region segmentation and the actual motion state can be maintained even when the animal's posture or projection occlusion changes.

[0048] In one possible example, after generating the CBCT volume data and before outputting it, the method further includes:

[0049] Based on the acquired attributes, the CBCT volume data is orthographically projected to obtain a verification projection sequence corresponding to the projection sequence;

[0050] Based on the spatial correspondence, the verification projection sequence is transformed to the corresponding animal movement state, and the consistency is compared with the projection sequence to obtain the projection residual distribution;

[0051] Based on the projection residual distribution, locate the residual abnormal region, and update the motion state and spatial correspondence corresponding to the residual abnormal region;

[0052] Repeat the back projection reconstruction and orthographic verification according to the updated spatial correspondence until the projection residual distribution meets the preset convergence condition, and output the updated CBCT volume data.

[0053] As can be seen, in this example, by remapping the reconstruction results to the actual acquisition state through orthographic verification and using the projection residual distribution to locate the areas with insufficient motion compensation, the motion state and spatial correspondence can be locally updated, forming a closed loop of reverse verification of motion estimation by the reconstruction results.

[0054] Secondly, embodiments of this application provide a CBCT reconstruction system for motion artifact correction during unconstrained animal scanning, comprising:

[0055] The data acquisition module is used to acquire unconstrained scan data of an animal under unconstrained conditions through multi-angle X-ray scanning; the unconstrained scan data includes a projection sequence and acquisition attributes corresponding to each projection, and the acquisition attributes include acquisition timing and imaging geometric parameters.

[0056] The motion state determination module is used to distinguish between projection changes caused by the rotation of the imaging system and structural changes caused by animal movement based on the correspondence between adjacent projections and cross-angle comparable projections in the projection sequence, and to determine the motion state sequence.

[0057] The projection uniformity module is used to transform the animal structures in each projection to a unified reference state according to the motion state sequence and the acquisition attributes, establish the spatial correspondence between the projection pixels and the reconstructed spatial voxels, and obtain motion-uniform projection data.

[0058] The volume data reconstruction module is used to backproject the motion-consistent projection data to the reconstruction space according to the spatial correspondence, and to fuse the projection contributions of different acquisition time sequences to generate motion artifact-corrected CBCT volume data.

[0059] The result output module is used to output the CBCT volume data.

[0060] The system provided in this application embodiment is based on the same inventive concept as the method described in the first aspect. It can complete the determination of motion state, projection uniformity and volume data reconstruction under unconstrained scanning conditions of animals through the coordinated execution of various functional modules, and thus has the same technical effect as the method described in the first aspect.

[0061] Thirdly, embodiments of this application provide a CBCT reconstruction device, including a processor, a memory, a communication interface, and one or more programs, the one or more programs being stored in the memory and configured to be executed by the processor, the programs including instructions for performing the steps in the first aspect of embodiments of this application.

[0062] Fourthly, embodiments of this application provide a computer storage medium storing a computer program for electronic data interchange, wherein the computer program causes a computer to perform some or all of the steps described in the first aspect of this embodiment.

[0063] As can be seen, in the embodiments of this application, compared with CBCT reconstruction methods that treat projections from different angles as the same stationary object or rely on fixation or anesthesia to suppress movement, this application first acquires the projection sequence and acquisition attributes formed by unconstrained scanning of the animal using a CBCT reconstruction device; secondly, based on the correspondence between animal structures in adjacent projections and comparable projections across angles, it distinguishes between projection changes caused by the rotation of the imaging system and structural changes caused by animal movement, and determines the motion state sequence; then, according to the motion state sequence and acquisition attributes, it transforms the animal structures in each projection to a unified reference state, establishes the spatial correspondence between projection pixels and reconstructed space voxels, and obtains motion-consistent projection data; thirdly, it backprojects the motion-consistent projection data to the reconstructed space according to the spatial correspondence, and fuses the projection contributions from different acquisition times to generate CBCT volume data; finally, it outputs CBCT volume data after motion artifact correction. By distinguishing between the regular projection changes caused by the acquisition geometry and the structural changes caused by the animal’s spontaneous movement, and by normalizing the animal’s structure under different movement states to a unified reference state before backprojection, the projection contributions of different acquisition sequences can be mapped to the same anatomical position. This reduces the accumulation of movement inconsistencies in the reconstructed space, suppresses ghosting, stripes and boundary blurring, and helps to improve the structural continuity and size representation reliability of animal CBCT images. Attached Figure Description

[0064] Figure 1 A schematic diagram of the hardware architecture of the animal unconstrained scanning and CBCT reconstruction device provided in the embodiments of this application;

[0065] Figure 2 A schematic diagram of the architecture of the motion artifact correction CBCT reconstruction software provided in the embodiments of this application;

[0066] Figure 3 This is a schematic diagram of an animal unrestrained CBCT scanning service scenario provided in an embodiment of this application;

[0067] Figure 4 A schematic flowchart illustrating the CBCT reconstruction method for motion artifact correction under unconstrained animal scanning provided in this application embodiment;

[0068] Figure 5 A schematic diagram of projection transformation decomposition and unified reference state mapping provided for embodiments of this application;

[0069] Figure 6A flowchart illustrating the motion state sequence determination process provided in this application embodiment;

[0070] Figure 7 This is a schematic diagram illustrating the generation of projection feature sequences and motion state sequences provided in an embodiment of this application.

[0071] Figure 8 A flowchart illustrating the candidate motion observation generation process provided in this application embodiment;

[0072] Figure 9 This is a schematic diagram illustrating the generation and aggregation of local motion observations provided in an embodiment of this application;

[0073] Figure 10 A flowchart illustrating the process of determining candidate corresponding feature pairs provided in this application embodiment;

[0074] Figure 11 This is a schematic diagram of bidirectional matching and neighborhood topology consistency filtering provided in an embodiment of this application;

[0075] Figure 12 A schematic diagram of the closed-loop update process based on projection residual distribution provided in an embodiment of this application;

[0076] Figure 13 This is a schematic diagram of the functional modules of the CBCT reconstruction system for motion artifact correction under unconstrained animal scanning provided in the embodiments of this application. Detailed Implementation

[0077] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. The described embodiments are used to explain this application, but are not intended to limit the scope of protection of this application. The various embodiments and their technical features can be combined with each other where there is no conflict.

[0078] In this embodiment, the unrestrained state of the animal refers to the state in which the animal is not held in a single position by a rigid fixing structure during the scanning process. The unrestrained state does not exclude the setting of a scanning chamber, safety boundaries, or non-rigid restraint structures to prevent the animal from leaving the imaging area, nor does it require the animal to move continuously during the scanning period.

[0079] Unconstrained scan data is a dataset generated from multi-angle X-ray scans of animals under unconstrained conditions. Unconstrained scan data includes at least a projection sequence and acquisition attributes, and may also include equipment calibration data, detector correction data, exposure status, and scan task identifiers. Each projection in the projection sequence can be the raw projection output from a flat panel detector, or a projection that has undergone dark field correction, gain correction, bad pixel removal, or scattering correction.

[0080] Acquisition timing is used to characterize the order and timing of acquisition of each projection. Imaging geometry parameters are used to characterize the geometric relationship between the X-ray source, detector, and reconstruction space, and may include acquisition angle, source-to-detector distance, source-to-rotation center distance, detector pixel size, detector orientation, and rotation center position.

[0081] Adjacent projections are projections in a projection sequence that are acquired at adjacent times or whose time intervals meet the condition of being adjacent. Cross-angle comparable projections are projections in which comparable structural regions can be determined based on imaging geometry parameters among non-adjacent projections. Comparable structural regions can be areas of the same animal structure that overlap after geometric mapping, or they can be local areas with stable outlines, bony edges, or tissue grayscale structures.

[0082] A motion state sequence is a collection of animal motion representation data arranged chronologically according to the acquisition time. Motion states can include at least one of the following: overall displacement, overall rotation, local displacement, local deformation, direction of motion, amplitude of motion, and confidence state. A unified reference state can be the animal state at the start of the scan, the animal state at a time when the amplitude of motion is relatively small, the animal state at a time in the middle of the scan, or a representative state determined by multiple projections.

[0083] Spatial correspondence is used to represent the mapping relationship between projected pixels and reconstructed spatial voxels after motion compensation. Motion-consistent projection data can be represented as a resampled projected image, or as association data formed by projected pixel values ​​and motion compensation mapping indices, so that the projection contribution can be written into the corresponding reconstructed spatial location during backprojection.

[0084] In one possible example, see Figure 1 The animal unconstrained scanning and CBCT reconstruction device 100 may include a CBCT scanning device 110, a CBCT reconstruction server 120, a data storage device 130, an image workstation 140, and a communication network 150.

[0085] The CBCT scanning device 110 may include an X-ray source 111, a flat panel detector 112, a rotating gantry 113, a scanning chamber 114, and an acquisition controller 115. The X-ray source 111 and the flat panel detector 112 are located on opposite sides of the rotating gantry 113. The scanning chamber 114, situated within the imaging area formed between the X-ray source 111 and the flat panel detector 112, is used to accommodate an unrestrained animal. The scanning chamber 114 may be made of a material with low X-ray attenuation and defines a safe range of movement for the animal without requiring it to maintain a single position.

[0086] The acquisition controller 115 controls the rotating gantry 113, X-ray source 111, and flat panel detector 112 to perform multi-angle projection acquisition, and records the acquisition timing and imaging geometry parameters for each projection. The acquisition controller 115 can also record the exposure status, rotating gantry position, and scanning task identifier.

[0087] The CBCT reconstruction server 120 may include a processor 121, a memory 122, a communication interface 123, and a computing accelerator 124. The processor 121 performs motion state determination, projection homogenization, back-projection reconstruction, and orthographic verification. The memory 122 stores the reconstruction program, motion state sequences, spatial correspondences, and intermediate data. The communication interface 123 receives unconstrained scan data output from the CBCT scanning device 110 via a communication network 150. The computing accelerator 124 can perform feature matching, projection resampling, and voxel back-projection in parallel.

[0088] Data storage device 130 is used to store projection sequences, acquisition attributes, device calibration data, motion state sequences, projection residual distribution, and CBCT volume data. Image workstation 140 is used to submit reconstruction tasks, display volume data slices or 3D reconstruction views, and can display residual abnormal areas and motion confidence states.

[0089] In other embodiments, the CBCT reconstruction server 120 can be integrated with the acquisition controller 115 into the same device, or it can be implemented by multiple computing nodes working together. As long as the corresponding device can execute the data processing procedure described in this application, it can be used as an implementation form of the CBCT reconstruction device.

[0090] In one possible example, see Figure 2 The motion artifact correction CBCT reconstruction software 200 may include a data access layer 210, a temporal and geometric management layer 220, a projection preprocessing layer 230, a projection feature processing layer 240, a motion state processing layer 250, a projection homogenization layer 260, a volume data reconstruction layer 270, an orthographic verification layer 280, and a result service layer 290.

[0091] The data access layer 210 is used to receive projection sequences, acquisition timing, and imaging geometry parameters, and to associate data belonging to the same scanning task. The timing and geometry management layer 220 is used to verify the acquisition order of the projections and maintain the geometric relationship between the X-ray source, the flat panel detector, and the reconstruction space.

[0092] The projection preprocessing layer 230 is used to perform dark field correction, gain correction, bad pixel processing, effective region identification, and projection intensity normalization. The projection feature processing layer 240 is used to extract animal contour features and internal structural features, and to establish candidate correspondences between adjacent projection features and cross-angle comparable projection features.

[0093] The motion state processing layer 250 is used to generate candidate motion observations, eliminate geometric change components corresponding to the rotation of the imaging system, and generate a motion state sequence according to the acquisition time sequence. The projection uniformity layer 260 is used to determine a unified reference state, generate motion transformation relationships, establish spatial correspondence between projected pixels and reconstructed spatial voxels, and form motion-consistent projection data.

[0094] The volume data reconstruction layer 270 is used to backproject and fuse motion-consistent projection data according to spatial correspondence. The orthographic verification layer 280 is used to generate a verification projection sequence from the CBCT volume data, determine the projection residual distribution, and update the motion state and spatial correspondence corresponding to the residual abnormal regions. The result service layer 290 is used to save and output the motion artifact-corrected CBCT volume data.

[0095] Each software layer can be implemented by the processor 121 in the CBCT reconstruction server 120 executing program instructions, or it can be deployed on different computing devices according to the data processing load. The data transmitted between software layers can carry scan task identifiers, projection identifiers, and processing versions to maintain the consistency of data source and processing sequence.

[0096] In one possible example, see Figure 3 In unconstrained CBCT scanning scenarios for animals, animal 320 can be a laboratory animal of suitable size for the scanning chamber 310 or an animal requiring imaging examination. This embodiment uses rabbits as an example; animal 320 is located within the scanning chamber 310 and is not rigidly fixed. A rotating gantry 350 drives the X-ray source 330 and the flat panel detector 340 to perform multi-angle acquisition around the scanning chamber 310.

[0097] At the first data acquisition moment, animal 320 is in the first posture, and the flat panel detector 340 forms the first moment projection 361. At the second data acquisition moment, animal 320 undergoes local deformation of the chest and abdomen due to respiration, forming the second moment projection 362. At the third data acquisition moment, animal 320 undergoes trunk displacement or limb movement, forming the third moment projection 363. The motion state marker 370 is used to indicate the overall displacement and local deformation between different data acquisition moments.

[0098] The CBCT reconstruction server determines the motion state sequence based on the first-time projection 361, the second-time projection 362, the third-time projection 363, and other angular projections, transforms the animal structures in each projection to a unified reference state, and generates a reconstruction result 380. The reconstruction result 380 may include axial slices, coronal slices, sagittal slices, and three-dimensional volumetric views.

[0099] In this business scenario, respiratory deformation, trunk displacement, and limb movements are only used to illustrate different types of animal movement. In actual processing, it is not required to predetermine the type of movement; instead, the corresponding movement state can be determined based on the structural changes between projected features.

[0100] In one possible example, see Figure 4 The motion artifact correction CBCT reconstruction method under unconstrained animal scanning can be achieved through the following steps S101 to S105, and can be performed by an unconstrained animal scanning and CBCT reconstruction device.

[0101] S101. Acquire unconstrained scan data of an animal under unconstrained conditions through multi-angle X-ray scanning; the unconstrained scan data includes a projection sequence and acquisition attributes corresponding to each projection, the acquisition attributes including acquisition timing and imaging geometric parameters.

[0102] The CBCT reconstruction equipment can receive projection sequences according to the scan task identifier and associate each projection with the corresponding acquisition timing and imaging geometric parameters. For projections lacking acquisition attributes, they can be recovered based on the attribute records of adjacent projections or the scan trajectory saved by the acquisition controller; projections that cannot be recovered can be marked as not eligible for motion estimation, but their participation in subsequent reconstruction can still be determined based on validity criteria.

[0103] The projection sequence can be normalized in intensity and the effective region extracted before entering the motion state determination process. The effective region is the area in the detector that actually receives the animal's projection contribution, excluding collimator-occluded areas, invalid detector edges, and obvious saturation areas. The preprocessed projection still retains the original projection identifier so that the motion state, spatial correspondence, and reconstruction contribution can be traced back to the corresponding acquisition time.

[0104] S102. Based on the correspondence between animal structures of adjacent projections and cross-angle comparable projections in the projection sequence, distinguish between projection changes caused by the rotation of the imaging system and structural changes caused by animal movement, and determine the motion state sequence; the cross-angle comparable projection is a non-adjacent projection with a structurally comparable region determined according to the imaging geometric parameters.

[0105] Among these, the acquisition time between adjacent projections is relatively short, providing information on motion continuity, but it also includes normal projection changes caused by the rotation of the gantry. Cross-angle comparable projections can observe changes in the same or related animal structures over a longer time span, and are used to help identify continuous movements that are difficult to distinguish between adjacent projections.

[0106] CBCT reconstruction equipment can predict the geometrical positional changes of stationary animal structures in the projection plane as the acquisition angle changes based on imaging geometry parameters, and compare the actual structural changes with the predicted geometrical changes. The portions that match the predicted geometrical changes can be attributed to imaging system rotation, while the portions that continuously deviate from the predicted geometrical changes can be used to create observations of animal movement.

[0107] Each motion state in the motion state sequence can correspond to a single projection or a time slice formed by multiple consecutive projections. When the reliable features of a certain acquisition moment are insufficient, the continuity of the motion states before and after can be used to determine the motion state at that acquisition moment, and a lower confidence state can be reserved for that motion state for subsequent residual verification.

[0108] S103. Based on the motion state sequence and the acquisition attributes, transform the animal structures in each projection to a unified reference state, establish the spatial correspondence between the projection pixels and the reconstructed spatial voxels, and obtain motion-consistent projection data.

[0109] The unified reference state is used to provide a common representation benchmark for animal structures at different acquisition times. CBCT reconstruction equipment can select acquisition times with less overall animal movement, more available structural features, or higher projection quality as reference times, or it can generate representative reference states based on multiple movement states.

[0110] For regions dominated by overall displacement, rigid transformation or affine transformation can be used to transform the animal structure to a unified reference state. For regions with respiratory deformation or local limb movements, local displacement fields or zoned motion transformation fields can be used. The transformation results of different regions can be made continuous at the region boundaries according to the motion confidence state.

[0111] Spatial correspondences can record the projection positions of reconstructed spatial voxels in the reference state at each acquisition time, and can also record the reconstructed spatial positions of each projected pixel after motion compensation. Motion-consistent projection data can be generated through explicit resampling or by calling spatial correspondences during backprojection, thus avoiding redundant interpolation.

[0112] S104. The motion-consistent projection data is back-projected to the reconstruction space according to the spatial correspondence, and the projection contributions from different acquisition times are fused to generate CBCT volume data.

[0113] Among them, CBCT reconstruction equipment can perform weighting, filtering, and backprojection on motion-consistent projection data. For projection areas with high confidence, normal reconstruction weights can be used; for areas with low confidence or local occlusion, the corresponding projection contribution can be reduced, or projections with structural comparability from other acquisition times can be used to supplement reconstruction information.

[0114] When fusing projection contributions from different acquisition times, each projection contribution is written to the voxel position under a unified reference state according to spatial correspondence. For multiple projection contributions received from the same voxel, they can be fused by combining acquisition angle coverage, motion confidence state, and projection validity to reduce the impact of a single anomalous projection on CBCT volumetric data.

[0115] S105. Output the CBCT volume data after motion artifact correction.

[0116] CBCT volumetric data can be saved as voxel data files, or converted into continuous slices, 3D surface models, or visualized volumetric data. The output can include scan task identifiers, reference status identifiers, motion compensation versions, and reconstruction parameters to support result traceability.

[0117] The technical advantages of this embodiment are as follows: First, the projection sequence is associated with the acquisition attributes; second, the motion state sequence is determined based on adjacent projections and cross-angle comparable projections; then, the animal structures in each projection are transformed to a unified reference state; finally, back-projection fusion is completed according to spatial correspondence; and finally, CBCT volumetric data is output. Because the structural changes caused by animal movement are compensated before back-projection, the projection contributions at different acquisition times can be written into mutually consistent spatial positions, thereby reducing motion artifacts and maintaining the continuity of animal structures.

[0118] In one possible example, see Figure 5 In the schematic diagram of projection change decomposition and unified reference state mapping, the original projection sequence 510 is entered into the adjacent projection comparison group 520 and the cross-angle comparable projection group 530 according to the acquisition time sequence. The adjacent projection comparison group 520 is used to extract structural displacement within a short time range, and the cross-angle comparable projection group 530 is used to verify structural changes within a longer time span.

[0119] The CBCT reconstruction equipment determines the geometric change component 540 from the projection changes based on the imaging geometric parameters, and defines the structural changes that remain after subtracting or compensating for the geometric change component 540 as the animal motion component 550. The animal motion component 550 is combined according to the acquisition time sequence to form a motion state sequence 560.

[0120] The unified reference state 570 provides the target state for the motion state sequence 560. The CBCT reconstruction device forms a spatial correspondence 580 based on the transformation relationship between each motion state and the unified reference state 570, and uses the spatial correspondence 580 to fuse the projection contributions at different acquisition times into corrected volume data 590.

[0121] This example avoids directly interpreting all structural displacements in the projection plane as animal motion by processing the geometric change component 540 and the animal motion component 550 separately; and connects motion estimation and volume data reconstruction by unifying the reference state 570 and the spatial correspondence 580, so that the output of the motion state sequence 560 can directly participate in the subsequent back projection processing.

[0122] In one possible example, see Figure 6 The process of determining the motion state sequence in the aforementioned step S102 can be achieved through the following steps S201 to S203.

[0123] S201. Based on the correspondence between animal structures of adjacent projections and cross-angle comparable projections in the projection sequence, extract animal outline features and internal structural features to form a projection feature sequence.

[0124] Animal contour features can include the outer boundary of the animal's projection area, the location of boundary curvature, contour protrusions, and contour direction. Internal structural features can include bony edges, tissue grayscale transition areas, stable texture locations, and local structural centers. Both contour features and internal structural features are associated with the acquisition timing and imaging geometry parameters of the corresponding projection.

[0125] The projection feature sequence can save each projection feature according to the order of projection acquisition, or it can be divided into different spatial branches according to the animal's projection region. For features that are not visible in some projections or are occluded by overlapping structures, the missing state can be retained instead of forcing a correspondence using low-reliability information.

[0126] S202. Based on the correspondence between adjacent projection features and cross-angle comparable projection features in the projection feature sequence, generate candidate motion observations.

[0127] Candidate motion observations are used to represent structural changes that may be caused by animal movement before subtracting normal imaging geometric changes. A candidate motion observation may include source projection identifier, target projection identifier, local location, projection plane displacement, matching difference, and confidence state.

[0128] The correspondence between adjacent projection features is used to generate short-term motion observations, while the correspondence between comparable projection features across angles is used to generate cross-time verification observations. When two types of observations point to the same direction of motion, the confidence state of the corresponding candidate motion observation can be improved; when two types of observations conflict with each other, the conflicting state can be retained and handed over to subsequent temporal continuity processing.

[0129] S203. The continuity of the candidate motion observation is constrained by the acquisition timing, and the geometric change components corresponding to the rotation of the imaging system are eliminated according to the imaging geometric parameters to obtain the motion state sequence.

[0130] The continuity constraint ensures that the direction and amplitude of motion between adjacent acquisition times cannot undergo abrupt changes lacking structural evidence. The geometric change component corresponding to the rotation of the imaging system can be determined based on the expected projection position of the stationary structure at the corresponding acquisition angle. After removing the geometric change component from candidate motion observations, state changes related to the animal's own motion can be obtained.

[0131] When there is a short-term gap in candidate motion observations, a supplementary motion state can be determined based on the previous motion state, the next motion state, and the motion states of adjacent regions. When the gap in time exceeds the continuity maintenance condition, the existing motion trend is not directly continued, but the corresponding state is marked as a state to be verified.

[0132] Specifically, bidirectional temporal verification can be performed on the motion state sequence. Forward verification propagates confirmed motion states according to the order of projection acquisition, while reverse verification propagates confirmed motion states in the reverse order. States supported only by a single direction and conflicting with cross-angle verification observations can have their confidence level reduced. This implementation addresses the problem of unidirectional motion trend accumulation caused by local occlusion, enabling reliable structural information appearing in the later stages of scanning to back-verify the motion state in the earlier stages.

[0133] The technical advantage of this embodiment is that the projection feature sequence simultaneously preserves the animal's external contour and internal structural information; candidate motion observations are simultaneously verified by adjacent projections and cross-angle comparable projections; and are further constrained by the acquisition timing and imaging geometric parameters. Therefore, a motion state sequence corresponding to each acquisition time can be formed from the projection sequence without external motion sensors.

[0134] In one possible example, see Figure 7 In the schematic diagram of the generation of projection feature sequence and motion state sequence, adjacent projection group 710 and cross-angle projection group 720 provide projection inputs with different time spans. The CBCT reconstruction device extracts animal contour features 730 and internal structure features 740 from the above projection inputs and writes them into the projection feature sequence 750 according to the projection identifier and acquisition time sequence.

[0135] Corresponding features in the projection feature sequence 750 form candidate motion observations 760. These candidate motion observations 760 are first subjected to temporal continuity constraints 770, and then compared with normal projection changes determined based on imaging geometric parameters, resulting in geometric change elimination results 780. The geometric change elimination results 780 are combined according to the acquisition time sequence to form a motion state sequence 790.

[0136] This example makes the projected feature sequence 750 an intermediate data carrier between the projected data and the motion state sequence 790, which can support motion estimation and also trace the feature evidence of the corresponding motion state during subsequent residual verification.

[0137] In one possible example, see Figure 8 The process of generating candidate motion observations in the aforementioned step S202 can be achieved through the following steps S301 to S303.

[0138] S301. Perform structural matching on adjacent projection features and cross-angle comparable projection features in the projection feature sequence to obtain candidate corresponding feature pairs.

[0139] Structural matching is used to determine whether local features in different projections originate from the same or have a stable correspondence with animal structures. Adjacent projection features and cross-angle comparable projection features are matched separately, so that the two types of projection relationships retain their respective matching sources, which facilitates the setting of different confidence conditions based on the time span and acquisition angle.

[0140] Candidate correspondence feature pairs include source projection features, target projection features, and candidate correspondences between them. A source projection feature can correspond to multiple target projection features in the initial stage, and subsequent filtering is performed through matching differences, bidirectional matching, and neighborhood topology consistency.

[0141] S302. Based on the displacement of the candidate corresponding feature pairs on the projection plane, the corresponding imaging geometric parameters, and the matching differences, generate local motion observations carrying matching confidence states.

[0142] The displacement of the projection plane is used to represent the positional change of the candidate corresponding feature pair; the imaging geometry parameters are used to determine the expected component of the positional change caused by the change in the acquisition angle; and the matching difference is used to characterize the degree of consistency between the two features in terms of grayscale, contour and neighborhood structure.

[0143] Matching confidence states can include confirmed states, candidate states, and excluded states. A confirmed state indicates that the candidate corresponding feature pair simultaneously satisfies the conditions of structural similarity, geometric interpretability, and bidirectional matching; a candidate state indicates that there is some supporting evidence but spatial or temporal verification is still needed; and an excluded state indicates that the correspondence clearly conflicts with the geometric conditions or neighborhood structure.

[0144] Specifically, the matching confidence state can be associated with and saved as the source of evidence. For example, contour matching evidence, internal structure matching evidence, cross-angle verification evidence, and temporal continuity evidence can each form a confidence basis. When a certain type of evidence is missing due to occlusion, other independent evidence can maintain the candidate state instead of directly excluding the local motion observation. This implementation addresses the problem of instability of single projection features and helps improve the persistence of motion observations in partially occluded scenarios.

[0145] S303. Aggregate the local motion observations according to the spatial adjacency relationship and the matching confidence state to generate the candidate motion observations.

[0146] Spatial adjacency can be determined based on the distance, contour connectivity, or internal structural relationships of local motion observations within the animal's projection area. Local motion observations with similar directions, compatible amplitudes, and spatial proximity can be aggregated into the same candidate motion observation.

[0147] During the aggregation process, local motion observations of confirmed states can serve as the aggregation core. Local motion observations of candidate states are added to the corresponding aggregation results when the neighborhood consistency condition is met. Local motion observations of excluded states do not participate in the aggregation. The aggregation results can preserve the spatial distribution of local motion observations for subsequent formation of overall motion and local deformation.

[0148] The technical advantage of this embodiment is that candidate corresponding feature pairs are not directly converted into single motion results, but rather local motion observations carrying matching confidence states are first formed, and then aggregation is completed based on spatial adjacency relationships. This reduces the impact of isolated mismatches and local low-quality projections on candidate motion observations, and preserves local difference information for partitioned motion compensation.

[0149] In one possible example, see Figure 9 In the schematic diagram of local motion observation generation and aggregation, the source projection feature 910 and the target projection feature 920 are structurally matched to form candidate corresponding feature pairs 930. The CBCT reconstruction equipment extracts the projection plane displacement 940 from the candidate corresponding feature pairs 930 and combines it with imaging geometric constraints 950 and matching differences 960 to form local motion observations 970.

[0150] Local motion observations 970 can carry location, direction of motion, amplitude of motion, and matching confidence state. Spatial adjacency relationships 980 connect local motion observations 970 located in adjacent contour regions, adjacent internal structure regions, or within a preset neighborhood. Local motion observations 970 that meet the aggregation conditions form candidate motion observations 990.

[0151] This example makes the projection plane displacement 940, imaging geometric constraints 950, and matching differences 960 traceable evidence of local motion observation 970. When a candidate motion observation 990 is determined to be anomaly in subsequent processing, the candidate corresponding feature pair 930 causing the anomaly can be located without having to re-search all projection features.

[0152] In one possible example, see Figure 10 The process of determining candidate corresponding feature pairs in the aforementioned step S301 can be implemented through the following steps S401 to S403.

[0153] S401. Generate candidate feature description information based on the local grayscale distribution, contour shape, and neighborhood structure of the adjacent projection features and the cross-angle comparable projection features.

[0154] Local grayscale distribution can characterize the brightness levels, grayscale variation direction, and tissue projection differences around a feature. Contour morphology can characterize contour direction, curvature trend, protrusion position, and boundary continuity. Neighborhood structure can characterize the relative position and connectivity between the current feature and other surrounding features.

[0155] Candidate feature descriptions can be formed by one or more of the above information. For animal outer contour features, the representation ratio of contour morphology and neighborhood structure can be increased; for internal bony structures or stable tissue structures, the representation ratio of local grayscale distribution and structural edges can be increased.

[0156] S402. Based on the acquisition angle relationship and time sequence relationship corresponding to each projection, determine the allowed search area for the candidate feature description information.

[0157] Among them, the acquisition angle relationship is used to limit the possible imaging positions of the stationary animal structure in different projections, and the temporal relationship is used to limit the possible range of motion of the animal structure within the corresponding time interval. The allowed search region is the set of positions in the target projection that are allowed to match the candidate feature description information of the source projection.

[0158] For projections with short time intervals and stable motion trends, a relatively concentrated allowable search area can be set around the predicted location. For cross-angle comparable projections, a structurally comparable region can be determined first based on imaging geometry parameters, and then an allowable search area can be set within this structurally comparable region.

[0159] S403. Perform bidirectional matching on the candidate feature description information within the allowed search area, and use neighborhood topological consistency to filter the bidirectional matching results to obtain the candidate corresponding feature pairs.

[0160] Bidirectional matching includes forward matching from source projection features to target projection features, and backward matching from target projection features back to source projection features. When forward and backward matching point to corresponding features, a bidirectional consistent result can be formed.

[0161] Neighborhood topology consistency is used to determine whether neighboring features around a matching feature maintain compatible relative orientations, relative distances, and connection orders. Even if the descriptive information of individual features is similar, if the spatial relationships of their surrounding features clearly conflict, the corresponding bidirectional matching result can be excluded.

[0162] Specifically, local topological units can be constructed around candidate corresponding feature pairs. A local topological unit can include the current candidate corresponding feature pair and multiple neighboring candidate corresponding feature pairs, comparing the relative arrangement direction, adjacency order, and local morphological changes between the source and target projection sides. When the local topological unit as a whole satisfies the continuous deformation condition, the candidate state feature pairs within it are retained; when only a single feature is similar but the local topological unit undergoes irregular flipping, that feature pair is excluded. This implementation can address the local mismatch problem caused by repetitive grayscale textures, thus improving the structural reliability of candidate corresponding feature pairs.

[0163] The technical advantages of this embodiment are as follows: it allows the search region to first constrain the matching space, then performs bidirectional matching to mutually verify the feature correspondences, and further utilizes the surrounding structure to eliminate isolated false matches based on neighborhood topological consistency. This process progressively narrows the candidate range, improving the stability of structural matching between adjacent projections and comparable projections across angles.

[0164] In one possible example, see Figure 11 In the schematic diagram of bidirectional matching and neighborhood topology consistency screening, the CBCT reconstruction device selects candidate feature description information 1130 from the source projection 1110, and determines the allowed search area 1140 in the target projection 1120 according to the acquisition angle relationship and temporal relationship between the source projection 1110 and the target projection 1120.

[0165] Candidate feature description information 1130 forms a positive matching result 1150 within the allowed search region 1140. A reverse search is performed starting from the target feature pointed to by the positive matching result 1150, forming a reverse matching result 1160. When the positive matching result 1150 and the reverse matching result 1160 point to each other, a bidirectional consistent result 1170 is generated.

[0166] The CBCT reconstruction equipment further compares the relative positions and connection order of features surrounding the bidirectional consistency result 1170 to form a neighborhood topology verification result 1180. The bidirectional consistency result 1170 that passes the verification is identified as a candidate corresponding feature pair 1190.

[0167] This example sets the similarity of descriptive information, reasonable search location, consistency of bidirectional correspondence, and compatibility of neighborhood structure as step-by-step filtering conditions, which can clearly limit the formation process of candidate corresponding feature pairs 1190 without increasing the complexity of feature vocabulary.

[0168] In one possible example, as a further implementation of the aforementioned step S402, the process of determining the allowed search area for candidate feature description information based on the acquisition angle relationship and temporal relationship corresponding to each projection can be achieved through the following steps S501 to S504.

[0169] S501. Based on the acquisition angle relationship and the temporal relationship, determine the angle interval and time interval between the source projection and the target projection to which the candidate feature description information belongs, and form an initial search boundary.

[0170] The angular interval characterizes the angular change of the rotating frame between the source projection and target projection acquisition times, while the time interval characterizes the time range within which the animal may move. The initial search boundary can be determined jointly based on the expected projection range of the stationary structure at the corresponding angular interval and the allowable movement range of the animal structure at the corresponding time interval.

[0171] The initial search boundary can be represented by a rectangular region, a contour-like region, or a region extending along the expected projection trajectory. The initial search boundary is used to exclude target projection regions that are clearly not positionally related to the source projection features.

[0172] S502. Based on the imaging geometric parameters, the effective area of ​​the animal projection is mapped to the initial search boundary, and pixel positions that do not have structural comparability are excluded to obtain the constrained search boundary.

[0173] The effective area of ​​animal projection is the region in the target projection that contains effective animal projection contributions and meets the conditions for detector usability. CBCT reconstruction equipment can map the local structural range in the source projection to the target projection based on imaging geometry parameters, and then overlap the mapped range with the initial search boundary and the effective area of ​​animal projection.

[0174] Pixel locations lacking structural comparability can include those outside the effective area of ​​the animal projection, those obscured by the collimator, those exhibiting significant saturation, and those whose ray correspondence with the source projection structure cannot be established based on imaging geometry. These locations are excluded to form the constraint search boundary.

[0175] S503. Generate a matching displacement trend based on the confirmed matching positions of the candidate feature description information in the front and back projections.

[0176] Among them, the confirmed matching position is the position of the candidate corresponding feature pair with a confirmed matching confidence state in the preceding and following projections. The matching displacement trend can characterize the main movement direction, displacement change, and stability of the current local structure during continuous acquisition time.

[0177] When the current candidate feature description information lacks sufficient historical matching positions, the confirmed matching positions of adjacent features within the same motion coupling region can be used to generate a regional matching displacement trend, and the confidence state of the trend can be reduced.

[0178] S504. Adjust the constraint search boundary according to the matching displacement trend to determine the allowed search area.

[0179] The CBCT reconstruction equipment can move the center position of the constraint search boundary along the matching displacement trend and adjust the boundary range according to the stability of the trend. When the matching displacement trend is stable, the search range perpendicular to the direction of movement can be narrowed; when the matching displacement trend changes significantly, the search range can be expanded without exceeding the constraint search boundary.

[0180] Specifically, the constraint search boundary can be divided into a core search region and a supplementary search region. The core search region is set around the predicted position indicated by the matching displacement trend, and structural matching is performed first; when no matching result that meets the confidence condition is formed in the core search region, the process then proceeds to the supplementary search region. The supplementary search region is still constrained by imaging geometry and the effective area of ​​animal projection. This implementation can address the problem of motion amplitude changing during scanning, reduce invalid matches in the stable motion phase, and retain the necessary search coverage in the abrupt motion phase.

[0181] The technical advantage of this embodiment is that the initial search boundary is sequentially constrained by the effective area of ​​the animal projection, imaging geometric parameters, and matching displacement trends, ultimately forming an allowable search area. Therefore, the search range can dynamically change with the acquisition angle, time interval, and animal movement trends, reducing matching ambiguities caused by global searches.

[0182] In one possible example, as a further implementation of the aforementioned step S504, the process of adjusting the constraint search boundary according to the matching displacement trend and determining the allowed search area can be implemented through the following steps S601 to S603.

[0183] S601. Project the matching displacement trend onto the constraint search boundary to obtain the predicted matching region.

[0184] In this system, the CBCT reconstruction equipment can determine the prediction center position of candidate features in the target projection based on the matching displacement trend, and form a prediction matching region around the prediction center position. The extension direction of the prediction matching region can be consistent with the main direction of the matching displacement trend.

[0185] When the matching displacement trend originates from the same local feature, the predicted matching region can be set around the predicted position of that local feature; when the matching displacement trend originates from the motion coupling region, the predicted center position can be corrected according to the relative position of the current feature in the motion coupling region.

[0186] S602. Determine the motion uncertainty range based on the degree of change of the matching displacement trend with the time interval.

[0187] The degree of change can be determined based on the changes in displacement direction and amplitude between multiple consecutive confirmed matching positions. The more stable the motion trend, the more concentrated the range of motion uncertainty can be; when the direction or amplitude of motion changes, the range of motion uncertainty can be expanded accordingly.

[0188] The range of motion uncertainty can also be determined in conjunction with the matching confidence state. When there are few confirmed matching locations used to generate the matching displacement trend, or when there is occlusion in the local structure, the range of motion uncertainty can be expanded to avoid missing effective matches due to an overly narrow prediction region.

[0189] S603. Expand the predicted matching region according to the motion uncertainty range, and take the intersection of the expanded predicted matching region and the constraint search boundary to obtain the allowed search region.

[0190] The predicted matching region can be expanded along the main direction of motion and in the direction intersecting the main direction of motion. The expanded predicted matching region is not directly used as the allowed search region, but rather its intersection with the constraint search boundary is taken, thereby excluding positions that exceed the geometrically comparable range or the effective area of ​​the animal projection.

[0191] The technical advantages of this embodiment are as follows: the predicted matching region utilizes existing motion trends to centrally match positions; the motion uncertainty range provides necessary range expansion based on the degree of trend change; and the constrained search boundary limits the upper limit of the expansion. These three elements work together to balance matching recall under motion prediction bias and matching reliability under imaging geometric constraints.

[0192] In one possible example, as a further implementation of the aforementioned step S103, the process of obtaining motion-consistent projection data based on the motion state sequence and the acquired attributes can be implemented through the following steps S701 to S703.

[0193] S701. Based on the motion state sequence and the acquisition attributes, divide the animal projection area in each projection into motion coupling regions, and determine the overall motion component and local deformation component of each motion coupling region.

[0194] Among them, the motion coupling region is a projection area where the motion responses between local locations are correlated, and it does not necessarily have to be completely consistent with the pre-defined anatomical divisions. For example, when the animal trunk undergoes overall displacement, the trunk outline and internal bony structures can form a motion coupling region; when the chest and abdomen undergo local deformation with respiration, a motion coupling region can be formed that is related to the overall displacement of the trunk but has different deformation characteristics.

[0195] The overall motion component is used to characterize the movement or rotation shared by multiple local locations within the motion coupling region. The local deformation component is used to characterize the deviation of each local location from the common motion component, such as the expansion of the chest and abdomen contour, local tissue compression, or the bending change of a limb relative to the trunk.

[0196] S702. Associate the overall motion component and the local deformation component with the acquisition timing and transform them to the unified reference state to generate a partitioned motion transformation field.

[0197] The partitioned motion transformation field describes the transformation relationship between the motion coupling region and the unified reference state at each acquisition time. The overall motion component can determine the transformation of the main body of the region, while the local deformation component can further correct the position within the region.

[0198] A continuous transition region can be set between adjacent motion coupling regions. Within the continuous transition region, different transformation results can be fused based on the confidence state of local motion observations and the adjacency relationship to each motion coupling region, so as to reduce spatial discontinuities at the partition boundaries.

[0199] S703. Establish the spatial correspondence between the projected pixels and the reconstructed spatial voxels using the partitioned motion transformation field, and resample each projection to obtain the motion-consistent projection data.

[0200] In this system, CBCT reconstruction equipment can start from the reconstructed spatial voxels under a unified reference state, determine the motion position of the voxel at the corresponding acquisition time based on the partitioned motion transformation field, and then determine its sampling position in the projection plane based on the imaging geometry parameters. Thus, a spatial correspondence can be established from the reconstructed spatial voxels to the projected pixels.

[0201] Projection resampling can employ neighbor-to-neighbor sampling, linear interpolation, or interpolation that preserves structural edges. For the boundaries of motion-coupled regions, resampling weights can be determined by combining the transformation confidence states of adjacent regions. The resampled projection is expressed according to a unified reference state, forming motion-consistent projection data.

[0202] The technical advantage of this embodiment is that by processing overall motion and local deformation separately through motion coupling regions, it avoids using a single rigid transformation to process the entire animal structure. The partitioned motion transformation field associates the motion state of each region with the acquisition time sequence and further transforms it into a spatial correspondence, enabling different types of animal motion to obtain coordinated compensation in the same reconstruction process.

[0203] In one possible example, as a further implementation of the aforementioned step S701, the process of dividing the motion coupling region and determining the overall motion components and local deformation components can be achieved through the following steps S801 to S803.

[0204] S801. Based on the motion state sequence and the acquisition attributes, determine the motion response at local locations within the animal projection area of ​​each projection to form a local motion response distribution.

[0205] Local locations can correspond to animal outline features, internal structural features, or local regions formed by adjacent features. Local motion responses can include motion direction, motion amplitude, change sequence, and motion confidence state.

[0206] Local motion response distributions are used to represent the motion state of each local location within the animal's projection area. For locations lacking reliable features, constrained supplementation can be made based on the motion responses of adjacent local locations, while retaining the corresponding supplementary identifiers and confidence states.

[0207] S802. Merge adjacent local positions with related motion directions and different motion amplitudes that satisfy the coupling condition to obtain the motion coupling region.

[0208] Among these, motion direction correlation refers to the fact that the main motion directions of adjacent local locations are the same, similar, or have a continuous changing relationship. Motion amplitude differences satisfying the coupling condition means that the amplitude changes between adjacent local locations can be explained by the overall motion or continuous local deformation.

[0209] The merging process can begin with the local motion response of the confirmed state and gradually absorb adjacent local positions that satisfy the coupling conditions. When the motion directions between two regions are significantly separated, or when the changes in motion amplitude cannot be explained by continuous deformation, they can be retained as regions with different motion couplings.

[0210] S803. Extract the common motion component from the local motion response of each of the motion coupling regions as the overall motion component, and take the spatial change component that deviates from the common motion component as the local deformation component.

[0211] The common motion component can be determined based on the consistent motion trend of multiple reliable local locations within the motion coupling region. The spatial variation component deviating from the common motion component can be determined based on the directional and amplitude differences between each local location and the common motion trend.

[0212] Local deformation components are preserved in their spatial distribution within the motion coupling region, rather than being aggregated into a single deviation value. Therefore, the partitioned motion transformation field can express continuous expansion, contraction, bending, or local displacement within the same region.

[0213] The technical advantage of this embodiment is that the motion coupling region is formed based on the actual local motion response, and the overall motion component and the local deformation component are obtained from the same local motion response distribution decomposition, so that the two types of motion components have a consistent data source. Therefore, the compensation error caused by the inconsistency between the preset anatomical partition and the actual motion region can be reduced.

[0214] In one possible example, see Figure 12 After generating CBCT volume data in step S104 and before outputting in step S105, a closed-loop update based on the projection residual distribution can be performed through the following steps S901 to S904.

[0215] S901. Based on the acquired attributes, perform orthographic projection on the CBCT volume data to obtain a verification projection sequence corresponding to the projection sequence.

[0216] In this process, orthographic projection uses the acquisition angle, source-to-detector distance, source-to-rotation center distance, and detector attitude corresponding to the actual projection acquisition to generate verification projections for each acquisition moment from the CBCT volume data. The verification projection sequence maintains the same projection identifier and acquisition order as the original projection sequence.

[0217] Figure 12 The closed-loop update process 1200 shown may include candidate CBCT volume data 1210, verification projection sequence 1220, actual projection sequence 1230, projection residual distribution 1240, residual abnormal region 1250, motion state update result 1260, spatial correspondence update result 1270, repeated reconstruction result 1280, and convergence judgment result 1290. The candidate CBCT volume data 1210 is orthographically projected to form the verification projection sequence 1220.

[0218] S902. Based on the spatial correspondence, the verification projection sequence is transformed to the corresponding animal movement state, and a consistency comparison is performed with the projection sequence to obtain the projection residual distribution.

[0219] The verification projection sequence 1220 initially corresponds to a unified reference state. The CBCT reconstruction device transforms each verification projection to the animal's motion state corresponding to the actual acquisition time based on the spatial correspondence, and then compares it with the corresponding projection in the actual projection sequence 1230.

[0220] Consistency comparison can be based on at least one of the following: differences in projection intensity, differences in structural edge positions, and local gradient changes, to form a projection residual distribution 1240. The projection residual distribution 1240 preserves the residual position, residual direction, residual degree, and corresponding acquisition time.

[0221] S903. Locate the abnormal residual regions based on the projection residual distribution, and update the motion state and spatial correspondence corresponding to the abnormal residual regions.

[0222] The residual abnormal region 1250 can be a location where the residual degree continuously exceeds the normal fluctuation range, where the structural edge is continuously misaligned, or where the residual appears repeatedly in multiple comparable projections. Isolated detector noise or local differences that only appear in a single projection can be given lower processing priority.

[0223] The CBCT reconstruction equipment uses the projection markers corresponding to the residual abnormal region 1250 to retrospectively generate candidate corresponding feature pairs and local motion observations for the corresponding motion states. Based on the residual direction and structural misalignment, the corresponding motion states are corrected, resulting in motion state update result 1260, and spatial correspondence update result 1270 is generated accordingly.

[0224] The update process can be applied only to the residual abnormal region 1250 and its adjacent transition region, while keeping the motion state and spatial correspondence of other regions that have met the consistency conditions unchanged, thereby avoiding the impact of local updates on the already stable reconstruction region.

[0225] S904. Repeat the back projection reconstruction and orthographic verification according to the updated spatial correspondence until the projection residual distribution meets the preset convergence condition, and output the updated CBCT volume data.

[0226] In this process, the CBCT reconstruction equipment regenerates motion-consistent projection data for the corresponding region based on the spatial correspondence update result 1270, forming a repeat reconstruction result 1280. The repeat reconstruction result 1280 is then verified by orthographic projection to form a new projection residual distribution.

[0227] The convergence judgment result 1290 can be determined based on the overall change of the residuals tending to stabilize, the range of the residual abnormal region meeting the output conditions, or the number of repetitions reaching a preset upper limit. When the convergence conditions are met, the repeated reconstruction result 1280 is determined as the updated CBCT volume data; when not met, the motion state and spatial correspondence corresponding to the residual abnormal region are updated.

[0228] The technical advantages of this embodiment are as follows: orthographic verification converts candidate CBCT volume data back into data that can be directly compared with the actual acquired projection, and the projection residual distribution further locates the reconstruction inconsistencies to specific acquisition times and local regions. By updating only the motion states and spatial correspondences corresponding to the residual abnormal regions, a closed loop can be formed between motion estimation, backprojection reconstruction, and projection verification, improving the consistency between motion compensation results and actual projection data.

[0229] In one possible example, see Figure 13The CBCT reconstruction system 1300 for motion artifact correction under unconstrained animal scanning may include a data acquisition module 1310, a motion state determination module 1320, a projection uniformity module 1330, a volume data reconstruction module 1340, and a result output module 1350.

[0230] Data acquisition module 1310 is used to acquire unconstrained scan data of an animal formed by multi-angle X-ray scanning in an unconstrained state; the unconstrained scan data includes a projection sequence and acquisition attributes corresponding to each projection, the acquisition attributes include acquisition timing and imaging geometric parameters;

[0231] The motion state determination module 1320 is used to distinguish between projection changes caused by the rotation of the imaging system and structural changes caused by animal movement based on the correspondence between adjacent projections and cross-angle comparable projections in the projection sequence, and to determine the motion state sequence.

[0232] The projection unification module 1330 is used to transform the animal structure in each projection to a unified reference state according to the motion state sequence and the acquisition attributes, establish the spatial correspondence between the projection pixels and the reconstructed spatial voxels, and obtain motion-consistent projection data.

[0233] The volume data reconstruction module 1340 is used to backproject the motion-consistent projection data to the reconstruction space according to the spatial correspondence, and fuse the projection contributions of different acquisition time sequences to generate motion artifact-corrected CBCT volume data.

[0234] The result output module 1350 is used to output the CBCT body data.

[0235] The data acquisition module 1310 can receive projection sequences and acquisition attributes from the CBCT scanning device or data storage device, and complete data association according to the scanning task identifier. The motion state determination module 1320 may include a projection feature processing unit, a candidate motion observation generation unit, and a motion state sequence generation unit.

[0236] The projection uniformity module 1330 may include a reference state determination unit, a motion coupling region processing unit, a partitioned motion transformation field generation unit, and a spatial correspondence generation unit. The volume data reconstruction module 1340 may include a projection resampling unit, a back projection unit, a projection contribution fusion unit, and an orthographic projection verification unit. The result output module 1350 can output CBCT volume data, slice images, 3D views, and reconstruction process records to the image workstation.

[0237] Each module in System 1300 can be implemented by one or more processors executing program instructions, or by multiple computing nodes communicating with each other. Module division is used to illustrate functional relationships and does not imply that each module must adopt an independent hardware structure.

[0238] In this embodiment, the data acquisition module 1310, motion state determination module 1320, projection unification module 1330, volume data reconstruction module 1340, and result output module 1350 form the functional chain of the present application. The output of the previous module serves as the input of the next module, and there is no isolated data that is detached from subsequent processing, enabling the system 1300 to complete the complete processing from unconstrained scan data input to motion artifact correction CBCT volume data output.

[0239] This application also provides a computer storage medium storing a computer program for electronic data interchange, which causes a computer to perform some or all of the steps of any of the methods described in the above method embodiments, wherein the computer includes an electronic device.

[0240] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0241] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0242] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of the units described above is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0243] The units described above as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0244] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0245] If the aforementioned integrated units are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage device (CMD). Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, electronic device, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned memory includes various media capable of storing program code, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0246] Those skilled in the art will understand that all or part of the steps in the various methods of the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage device, which may include: a flash drive, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk, etc.

[0247] The embodiments of this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of this application. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of this application. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A method for motion artifact correction and CBCT reconstruction in unconstrained animal scanning, characterized in that, Performed by CBCT reconstruction equipment, including: Acquire unconstrained scan data of an animal under unconstrained conditions through multi-angle X-ray scanning; the unconstrained scan data includes a projection sequence and acquisition attributes corresponding to each projection, the acquisition attributes including acquisition timing and imaging geometric parameters; Based on the correspondence between animal structures in adjacent projections and cross-angle comparable projections in the projection sequence, the projection changes caused by the rotation of the imaging system and the structural changes caused by animal movement are distinguished to determine the motion state sequence; the cross-angle comparable projection is a non-adjacent projection with a structurally comparable region determined according to the imaging geometric parameters. A two-way time-series verification is performed on the motion state sequence. The two-way time-series verification includes: forward verification propagating the confirmed motion state in the order of projection acquisition, and reverse verification propagating the confirmed motion state in the opposite order, thereby reducing the confidence level of states that are supported only by a single direction and conflict with cross-angle verification observations. Based on the motion state sequence and the acquisition attributes, the animal structures in each projection are transformed to a unified reference state, and a spatial correspondence between the projection pixels and the voxels in the reconstruction space is established to obtain motion-consistent projection data. The motion-consistent projection data is back-projected into the reconstruction space according to the spatial correspondence, and the projection contributions from different acquisition time sequences are fused to generate CBCT volume data. Output the CBCT volumetric data after motion artifact correction; After generating CBCT volumetric data and before outputting it, the method further includes: performing orthographic projection on the CBCT volumetric data based on the acquired attributes to obtain a verification projection sequence corresponding to the projection sequence; transforming the verification projection sequence to the corresponding animal motion state based on the spatial correspondence, and comparing its consistency with the projection sequence to obtain a projection residual distribution; locating residual abnormal regions based on the projection residual distribution, and backtracking to form candidate corresponding feature pairs and local motion observations for the corresponding motion state based on the projection identifiers corresponding to the residual abnormal regions, correcting the corresponding motion state based on the residual direction and structural misalignment, and updating the spatial correspondence accordingly; wherein, the update process only applies to the residual abnormal regions and their adjacent transition regions, while keeping the motion state and spatial correspondence of other regions that have met the consistency conditions unchanged; repeating backprojection reconstruction and orthographic verification according to the updated spatial correspondence until the projection residual distribution meets the preset convergence condition, and outputting the updated CBCT volumetric data.

2. The method according to claim 1, characterized in that, The method of distinguishing between projection changes caused by imaging system rotation and structural changes caused by animal movement based on the correspondence between adjacent projections and comparable projections across angles in the projection sequence, and determining the motion state sequence, includes: Based on the animal structure correspondence between adjacent projections and cross-angle comparable projections in the projection sequence, animal outline features and internal structure features are extracted to form a projection feature sequence. Based on the correspondence between adjacent projection features and cross-angle comparable projection features in the projection feature sequence, candidate motion observations are generated. The continuity of the candidate motion observations is constrained by the acquisition timing, and the geometric change components corresponding to the rotation of the imaging system are eliminated according to the imaging geometric parameters to obtain the motion state sequence.

3. The method according to claim 2, characterized in that, The generation of candidate motion observations based on the correspondence between adjacent projection features and cross-angle comparable projection features in the projection feature sequence includes: Structural matching is performed on adjacent projection features and cross-angle comparable projection features in the projection feature sequence to obtain candidate corresponding feature pairs; Based on the displacement of the candidate corresponding feature pairs on the projection plane, the corresponding imaging geometric parameters, and the matching differences, a local motion observation carrying a matching confidence state is generated. The local motion observations are aggregated according to spatial adjacency and the matching confidence state to generate the candidate motion observations.

4. The method according to claim 3, characterized in that, The step of performing structural matching on adjacent projection features and cross-angle comparable projection features in the projection feature sequence to obtain candidate corresponding feature pairs includes: Based on the local grayscale distribution, contour morphology, and neighborhood structure of the adjacent projection features and the cross-angle comparable projection features, candidate feature description information is generated; Based on the acquisition angle relationship and time sequence relationship corresponding to each projection, the allowed search area is determined for the candidate feature description information; Within the allowed search area, bidirectional matching is performed on the candidate feature description information, and the bidirectional matching results are filtered using neighborhood topology consistency to obtain the candidate corresponding feature pairs.

5. The method according to claim 4, characterized in that, The step of determining the allowed search region for the candidate feature description information based on the acquisition angle relationship and temporal relationship corresponding to each projection includes: Based on the acquisition angle relationship and the temporal relationship, the angular interval and time interval between the source projection and the target projection to which the candidate feature description information belongs are determined, and an initial search boundary is formed; Based on the imaging geometry parameters, the effective area of ​​the animal projection is mapped to the initial search boundary, and pixel positions that do not have structural comparability are excluded to obtain the constrained search boundary; Based on the confirmed matching positions of the candidate feature description information in the front and back projections, a matching displacement trend is generated; Adjust the constraint search boundary according to the matching displacement trend to determine the allowed search area.

6. The method according to claim 5, characterized in that, The step of adjusting the constraint search boundary according to the matching displacement trend and determining the allowed search area includes: The matching displacement trend is projected onto the constraint search boundary to obtain the predicted matching region; The range of motion uncertainty is determined based on the degree of change of the matched displacement trend with the time interval; The predicted matching region is expanded according to the range of motion uncertainty, and the intersection of the expanded predicted matching region and the constraint search boundary is taken to obtain the allowed search region.

7. The method according to any one of claims 1-6, characterized in that, The process involves transforming the animal structures in each projection to a unified reference state based on the motion state sequence and the acquired attributes, establishing a spatial correspondence between projected pixels and reconstructed spatial voxels, and obtaining motion-consistent projection data, including: Based on the motion state sequence and the acquisition attributes, the animal projection area in each projection is divided into motion coupling regions, and the overall motion component and local deformation component of each motion coupling region are determined. The overall motion components and the local deformation components are associated with the acquisition time sequence and transformed to the unified reference state to generate a partitioned motion transformation field; The spatial correspondence between the projected pixels and the reconstructed spatial voxels is established using the partitioned motion transformation field, and each projection is resampled to obtain the motion-consistent projection data.

8. The method according to claim 7, characterized in that, Based on the motion state sequence and the acquired attributes, the animal projection area in each projection is divided into motion coupling regions, and the overall motion component and local deformation component of each motion coupling region are determined, including: Based on the motion state sequence and the acquired attributes, the motion response at local locations is determined within the animal projection area of ​​each projection, forming a local motion response distribution; The motion coupling region is obtained by merging adjacent local positions that are related in motion direction and whose motion amplitudes differ and satisfy the coupling condition. The common motion component is extracted from the local motion response of each of the motion coupling regions as the overall motion component, and the spatial change component that deviates from the common motion component is taken as the local deformation component.

9. A CBCT reconstruction system for motion artifact correction under unconstrained animal scanning, used to implement the CBCT reconstruction method for motion artifact correction under unconstrained animal scanning as described in any one of claims 1 to 8, characterized in that, include: The data acquisition module is used to acquire unconstrained scan data of an animal under unconstrained conditions through multi-angle X-ray scanning; the unconstrained scan data includes a projection sequence and acquisition attributes corresponding to each projection, and the acquisition attributes include acquisition timing and imaging geometric parameters. The motion state determination module is used to distinguish between projection changes caused by the rotation of the imaging system and structural changes caused by animal movement based on the correspondence between adjacent projections and cross-angle comparable projections in the projection sequence, and to determine the motion state sequence. The projection uniformity module is used to transform the animal structures in each projection to a unified reference state according to the motion state sequence and the acquisition attributes, establish the spatial correspondence between the projection pixels and the reconstructed spatial voxels, and obtain motion-uniform projection data. The volume data reconstruction module is used to backproject the motion-consistent projection data to the reconstruction space according to the spatial correspondence, and to fuse the projection contributions of different acquisition time sequences to generate motion artifact-corrected CBCT volume data. The result output module is used to output the CBCT volume data.

Images