Three-dimensional reconstruction data generation method and device and readable storage medium

Abstract

The invention provides a three-dimensional reconstruction data generation method and device and a readable storage medium, and the method comprises the steps: obtaining two-dimensional projection data at a plurality of projection angles, executing a preset three-dimensional reconstruction algorithm based on the data, expanding a view field range through a view field expansion parameter, and covering a larger-range anatomical structure. And secondly, signal attenuation caused by missing of partial angle data is compensated in an extended view by means of a reconstruction weight parameter, so that the precision and integrity of a reconstructed image are ensured. According to the method, the traditional multiple scanning splicing and hardware transformation modes are effectively avoided, the scanning time and the radiation dosage are reduced, the speed and the real-time performance of three-dimensional image reconstruction are improved, and the method is particularly suitable for scenes with high real-time requirements such as clinical surgical navigation. In addition, the scheme is optimized through a pure algorithm means, the flexibility and stability in the imaging process are improved, and the method has remarkable technical advantages.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and particularly relates to a method, a device, and a readable storage medium for generating three-dimensional reconstruction data. Background Art

[0002] CBCT is a medical imaging technology that uses a cone X-ray beam and a flat panel detector to obtain two-dimensional projection data taken from multiple angles, and then reconstructs these projection data into a three-dimensional image through an algorithm.

[0003] In related technologies, in the technical solution for expanding the CBCT reconstruction field of view, there is a technical problem of low efficiency. Summary of the Invention

[0004] The purpose of the present invention is to overcome the above technical deficiencies, and provide a method, a device, and a readable storage medium for generating three-dimensional reconstruction data, so as to solve the technical problem of low efficiency existing in the related technology in the technical solution for expanding the CBCT reconstruction field of view.

[0005] To achieve the above technical purpose, the present invention adopts the following technical solutions: In a first aspect, the present invention provides a method for generating three-dimensional reconstruction data, including: Obtaining two-dimensional projection data at multiple projection angles for the same target; Based on the two-dimensional projection data at the multiple projection angles, performing a preset three-dimensional reconstruction algorithm to generate three-dimensional reconstruction data for the target; wherein, an extended field of view parameter and a reconstruction weight parameter corresponding to the extended field of view parameter are preset in the preset three-dimensional reconstruction algorithm; wherein, the extended field of view parameter is a field of view parameter increased on the basis of the original field of view parameter, and the original field of view parameter is a field of view parameter that can ensure that the two-dimensional projection data at all projection angles can cover the target during the three-dimensional reconstruction process; wherein, the reconstruction weight parameter is used to compensate for the signal attenuation caused by the missing data of some angles for the pixel points outside the original field of view in the extended field of view during the three-dimensional reconstruction process.

[0006] In a second aspect, the present invention provides an electronic device, including: a memory, and one or more processors communicatively connected to the memory; instructions executable by the one or more processors are stored in the memory, and when the instructions are executed by the one or more processors, the one or more processors are enabled to implement the above method.

[0007] In a third aspect, the present invention provides a computer-readable storage medium, in which a computer program is stored, and when the computer program is executed by a processor, the above method is implemented.

[0008] Beneficial effects: The 3D reconstruction data generation method provided by the present invention can effectively expand the reconstruction field of view of CBCT through algorithm optimization without relying on hardware transformation, significantly improve the image quality, and solve problems such as low efficiency and complex operation in traditional technologies. First, by acquiring two-dimensional projection data at multiple projection angles and performing a preset 3D reconstruction algorithm based on these data, the field of view range is expanded using the extended field of view parameters, thereby covering a larger range of anatomical structures. Secondly, with the help of the reconstruction weight parameters, the signal attenuation caused by the missing data at some angles is compensated in the extended field of view, thus ensuring the accuracy and integrity of the reconstructed image. This method effectively avoids the traditional methods of multiple scans and stitching and hardware transformation, reduces the scanning time, radiation dose, and improves the speed and real-time performance of 3D image reconstruction, especially suitable for scenarios with high real-time requirements such as clinical surgical navigation. In addition, this solution is optimized by pure algorithm means, improving the flexibility and stability in the imaging process, and has significant technical advantages. Description of the drawings

[0009] Figure 1 is a schematic flowchart of a method for generating 3D reconstruction data provided by an embodiment of the present invention; Figure 2 is an example diagram of the original field of view provided by an embodiment of the present invention; Figure 3 is a schematic diagram of the original field of view and the extended field of view provided by an embodiment of the present invention; Figure 4 is a schematic diagram of the extended cache space provided by an embodiment of the present invention; Figure 5 is a schematic diagram of the highlighted artifacts provided by an embodiment of the present invention; Figure 6 is a block diagram of an electronic device adopted by an embodiment of the present invention. Detailed implementation manners

[0010] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0011] Cone Beam Computed Tomography (CBCT) system is the core equipment for intraoperative three-dimensional image navigation. Its hardware architecture can include a C-arm, an X-ray tube, a flat panel detector, a data processing unit, etc. The C-arm rotates around the patient for scanning (which can be from 180 degrees to 360 degrees), collects two-dimensional projection data at multiple angles, and generates three-dimensional volume data based on a preset reconstruction algorithm. However, limited by the physical size and geometric layout of the detector, the imaging field of view of traditional CBCT is generally limited to the range of the largest inscribed circle of the detector plane. This limitation results in its inability to cover a large range of anatomical structures (for example, the pelvis, long bones, or complex trauma areas), severely restricting the surgical navigation requirements in the fields of orthopedics, oral surgery, etc.

[0012] To expand the reconstruction field of view, in related technologies, it can be achieved through the following three methods, specifically: In a related technology, multiple scans and stitching can be adopted. This method collects adjacent areas through multiple local scans, and then generates an extended field of view image through image registration and stitching. Although it can cover a larger anatomical range, its defects are extremely prominent. For example, the operation time increases significantly. It can be understood that each scan takes an additional 5 - 10 minutes, and the overall process is extended to more than 30 minutes, completely losing the intraoperative real-time performance. For example, the radiation dose accumulates. It can be understood that the radiation dose received by the patient increases linearly with the number of scans (for example, the dose of 3 scans reaches 3 times that of a single scan). Moreover, the stitching accuracy is insufficient. Mechanical rotation errors (±0.5 degrees) and patient micro-movements lead to registration misalignment, resulting in severe artifacts at the seams (such as strip-shaped highlights or dark areas), significantly reducing the image quality and possibly misleading surgical decisions.

[0013] In a related technology, the method of reducing the detector distance can be adopted. This method can expand the coverage range of a single scan by shortening the physical distance between the detector and the imaging center. However, this method requires invasive modification of the C-arm mechanical structure (for example, shortening the support arm length or adjusting the track design), thus causing multiple problems. Specifically, on the one hand, the rigidity of the modified C-arm decreases, and vibration errors are likely to occur during rotational scanning, resulting in blurred projection data. On the other hand, the detector is close to the patient's body surface, severely squeezing the operating space of the surgeon. Especially in minimally invasive surgery, the risk of collision between the instrument and the detector increases significantly.

[0014] In a related technology, the detector can be placed offset (for example, obliquely or laterally), and the extended field of view is covered by asymmetric projection. However, this method requires customized detector brackets and calibration devices, which are difficult to adapt to mainstream C-arm equipment. Moreover, the offset causes some angle projections to exceed the detector boundary, and the peripheral area is only covered at limited angles, resulting in a decrease in data integrity.

[0015] From the above analysis, it can be known that in the related art, there is an over-reliance on physical hardware to expand the field of view, and the algorithm optimization fails to be fully utilized to solve data sparsity and geometric constraints. For example, mechanical adjustment and multiple scans will significantly extend the preoperative preparation and intraoperative operation time, unable to meet the real-time requirements of emergency or complex surgeries. Another example is that the hardware-invasive transformation will damage the device stability, and the cumulative radiation dose and the compressed operation space directly threaten the safety of patients and surgeons.

[0016] Therefore, in summary, in the related art, in the technical solution for expanding the CBCT reconstruction field of view, there is a technical problem of low efficiency.

[0017] As Figure 1 shown, this embodiment provides a method for generating three-dimensional reconstruction data. The execution subject of the method can be a server, for example, a medical image processing server. It can also be a processing device of an X-ray system, an MRI system, an ultrasound system, a CT system, or a processing device of a CBCT, etc.

[0018] The method may include: Step S12: Obtain two-dimensional projection data at multiple projection angles for the same target.

[0019] In this embodiment, the obtaining action may be that the processing device of the medical imaging device directly controls the data acquisition hardware to acquire two-dimensional projection data and receives the two-dimensional projection data transmitted by the data acquisition hardware. Specifically, the processing device of the medical imaging device may be an embedded processing module integrated in a C-arm, an X-ray machine, or a CT device (for example, an FPGA, a DSP, or a dedicated ASIC chip, etc.), and the data acquisition hardware may be an X-ray source, a flat panel detector, and a rotation control unit, etc.

[0020] In this embodiment, the obtaining action may be that the medical image processing server receives the two-dimensional projection data transmitted by the medical imaging device. The medical image processing server may be deployed locally in the hospital or in the cloud, and the medical image processing server may be connected to the medical imaging device (for example, a CBCT) through the DICOM protocol to receive the two-dimensional projection data transmitted by the CBCT.

[0021] Specifically, in a particular embodiment, the medical imaging device can be a CBCT, which can obtain two-dimensional projection data of X-rays of a certain part of the human body (such as a patient), such as the hip joint, through a C-arm. The multiple angles can be multiple angles within the angle range from the starting angle to the ending angle. The angle range from the starting angle to the ending angle can be 180 degrees plus half of the cone angle. It can be rotated about 0.5 - 1 degree each time and sampled at intervals of approximately 0.05 - 1 second to obtain two-dimensional projection data for the same target at multiple angles. The CBCT can first temporarily store the collected two-dimensional projection data in the memory of the processing device and then transmit the two-dimensional projection data to the medical image processing server in real time through the DICOM protocol.

[0022] In this embodiment, the target can be a human body or other objects. Specifically, the target can be different parts of a patient, such as the hip joint, knee joint, spine, limb bones, etc. The target can also be objects such as mechanical parts, tools, equipment, animals, and plants.

[0023] In this embodiment, the multiple projection angles can be the angle range from the starting angle to the ending angle of the medical imaging device. Specifically, for example, in a CBCT device, the scan can cover a rotation range from 180 degrees to 360 degrees. Within this angle range, the C-arm rotates around the target area, and a two-dimensional projection image is collected at each sampling angle. Throughout the projection angle range, the collected data is sampled according to a preset angle interval. The scan interval can be 0.5 to 1 degree, that is, the C-arm collects two-dimensional projection data every time it rotates 0.5 to 1 degree.

[0024] In this embodiment, the two-dimensional projection data can be represented as projection images obtained by scanning the target area by a medical imaging device (such as CBCT, CT, X-ray, etc.) at different angles.

[0025] Specifically, the two-dimensional projection data can be ordinary X-ray two-dimensional projection data. That is, by irradiating the target area with an X-ray source and using a detector behind to receive the X-rays passing through the target object, a two-dimensional image (two-dimensional projection data) is formed. The two-dimensional image shows the attenuation of the rays during the penetration of the object by the X-rays.

[0026] The two-dimensional projection data can be cone beam CT (CBCT) two-dimensional projection data. Specifically, a CBCT device uses a cone-shaped X-ray beam to scan the target area, and the two-dimensional projection data is obtained by scanning the target at multiple angles. Each two-dimensional projection data is an image taken from a different angle, and finally forms a dataset of two-dimensional projection data.

[0027] The two-dimensional projection data may be parallel beam CT two-dimensional projection data.

[0028] Step S14: Based on the two-dimensional projection data at the multiple projection angles, execute a preset three-dimensional reconstruction algorithm to generate three-dimensional reconstruction data for the target; wherein, an extended field of view parameter and a reconstruction weight parameter corresponding to the extended field of view parameter are preset in the preset three-dimensional reconstruction algorithm; wherein, the extended field of view parameter is a field of view parameter increased on the basis of the original field of view parameter, and the original field of view parameter is a field of view parameter that can ensure that the two-dimensional projection data at all projection angles can cover the target during the three-dimensional reconstruction process; wherein, the reconstruction weight parameter is used to compensate for the signal attenuation caused by the missing data of some angles for the pixel points beyond the original field of view in the extended field of view during the three-dimensional reconstruction process.

[0029] In this embodiment, the step of executing a preset three-dimensional reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles to generate three-dimensional reconstruction data for the target may be to directly input the two-dimensional projection data at the multiple projection angles into the preset three-dimensional reconstruction algorithm, so as to generate three-dimensional reconstruction data for the target.

[0030] In this embodiment, the step of executing a preset three-dimensional reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles to generate three-dimensional reconstruction data for the target may be to first execute at least one preset image processing method on the two-dimensional projection data at the multiple projection angles to improve the image quality of the two-dimensional projection data, and then input the processed two-dimensional projection data into the preset three-dimensional reconstruction algorithm for processing, so as to generate three-dimensional reconstruction data for the target.

[0031] Specifically, the preset 3D reconstruction algorithm can be the Filtered BackProjection (FBP) algorithm. Specifically, first, perform a Fourier transform on the two-dimensional projection data, then apply a RAMP filter to compensate for geometric distortion and noise. Then, back-project the filtered projection data along the ray path into the three-dimensional space and accumulate to generate 3D reconstruction data. Among them, in this filtered backprojection algorithm, the extended field of view parameter can be used to expand the reconstruction grid, and the maximum inscribed circle of the traditional field of view can be expanded to a larger radius, thereby adjusting the size of the reconstruction grid. For example, it can be expanded from 512×512 to 1024×1024. During the backprojection process, the backprojection value of each voxel can be multiplied by a reconstruction weight parameter to compensate for the signal attenuation caused by sparse data in the peripheral region. In a specific scenario example, for hip joint scanning, the traditional field of view radius can be 150 mm and can be expanded to 250 mm through the extended field of view parameter. When backprojecting, the voxel 200 mm away from the imaging center can be multiplied by the reconstruction weight parameter 1.5 to amplify the signal intensity of the voxel at this position.

[0032] The preset 3D reconstruction algorithm can also be the Feldkamp-Davis-Kress (FDK) reconstruction algorithm.

[0033] The preset three-dimensional reconstruction algorithm can also be an iterative reconstruction algorithm. For example, it can be an ART (Algebraic Reconstruction Technique) algorithm or an SART (Simultaneous Algebraic Reconstruction Technique) algorithm. Specifically, the iterative reconstruction algorithm gradually improves the result of image reconstruction by continuously optimizing the objective function. In the iterative reconstruction algorithm, first, based on the current voxel estimate value (i.e., the currently guessed three-dimensional image), virtual projection data is generated. These projection data are obtained by simulating the X-ray passing through the object to get a theoretically projected image. Then, the difference between the actually acquired projection data and the virtual projection data generated by the forward projection is compared. This difference can be represented by calculating the projection error. Finally, the voxel value is adjusted according to the gradient direction of the error, so that the projection data gradually approaches the actually acquired data. This process is carried out in each iteration until the image error meets the preset convergence condition. In the iterative reconstruction algorithm, the extended field of view parameter can be used for the extension of the system matrix. It can be understood that the system matrix is a matrix that maps voxels to projection data. In the case of an extended field of view, the extended field of view parameter can be carried out by increasing the voxel-detector correspondence in the system matrix. That is to say, the system matrix will be adjusted to ensure that each voxel in the extended field of view can be mapped to the corresponding detector position. Even in the case of insufficient data sampling in some edge regions, the system matrix can ensure that these regions can be effectively reconstructed. When performing iterative reconstruction and when using the extended field of view parameter, the initial image (all-zero matrix) can be set to a larger size to cover the extended field of view range. In this way, the size of the initially estimated image can ensure that all regions within the extended field of view are covered, avoiding the loss of reconstruction data in the extended region during subsequent iterative processes. The reconstruction weight parameter in the iterative reconstruction algorithm is mainly used to assign different weights to voxels in different regions to compensate for the data imbalance problem caused by the extension of the field of view. In a specific implementation, the reconstruction weight parameter can be set in the objective function to perform weighting on the regularization term in the objective function. The objective function can be: In the formula, is represented as the system matrix. The system matrix is used to relate the voxels of the three-dimensional volume to the two-dimensional projection data. In each iteration, the system matrix is used to map the current voxel values to a virtual projection data set; is represented as a vector. The vector contains each voxel in the image. Each element in the vector represents the intensity of the corresponding voxel. During the iterative process, the algorithm will continuously adjust values to make the reconstructed image as close as possible to the actual projection data; is the original two-dimensional projection data obtained by scanning the object; is the error term used to minimize the error of the reconstructed image; is the balance coefficient used to control the weight between the error term and the regularization term; is the reconstructed weight parameter function which depends on the position of the voxel and increases as the distance of the voxel from the imaging center increases; is the regularization term used to represent the sum of squares of voxel intensities in image reconstruction.

[0034] The preset three-dimensional reconstruction algorithm can also be a statistical reconstruction algorithm. Specifically, it can be a reconstruction algorithm based on a statistical model (e.g., Poisson noise distribution), and this reconstruction algorithm based on the statistical model can iteratively optimize the voxel values by maximizing the likelihood function.

[0035] The preset three-dimensional reconstruction algorithm can also be a deep learning-based reconstruction algorithm. Specifically, it can be a reconstruction algorithm based on U-Net or ResNet. In the neural network, the extended field of view parameter can be a parameter of the neural network. For example, in the input layer, the two-dimensional projection data can be stitched into an extended field of view size (e.g., 1024×1024→2048×2048) as the network input. The reconstruction weight parameter can be used to weight the loss function, that is, during training, higher weights are assigned to the voxel losses in the extended area. During the pre-training process, the neural network can be trained using training data, and this training data can include paired samples of the traditional field of view and the extended field of view. During the training process, the neural network automatically compensates for the peripheral signal attenuation by learning the implicit distribution of the reconstruction weight parameter.

[0036] In this embodiment, the original field of view parameter can be represented as the maximum imaging area that CBCT can cover based on its inherent physical structure and geometric layout without any algorithm or hardware extension. Specifically, the original field of view parameter can be the largest inscribed circle of the detector plane, and this largest inscribed circle is used to ensure that the X-ray beam can completely cover this circular area at all projection angles, avoiding data truncation. It can be understood that within this original field of view, all voxels are completely penetrated by X-rays at each projection angle, and the projection data has no truncation. During back-projection integration, the data redundancy of these voxels is high, and the reconstruction result is stable and the noise is controllable.

[0037] In this embodiment, the extended field of view parameter can be expressed as the range of the imaging area that can be covered in a CBCT system by pure algorithm optimization rather than hardware modification, breaking through the limitations of the traditional physical field of view. The extended field of view parameter can be a geometric boundary of virtual extension, allowing the reconstruction algorithm to generate effective three-dimensional volume data outside the physical range of the original detector.

[0038] In a specific implementation, the extended field of view parameter can be formally defined as: In the formula, is expressed as the extension coefficient, is expressed as the original field of view parameter, that is, the radius of the largest inscribed circle.

[0039] In this embodiment, the reconstruction weight parameter can be a weight parameter used to dynamically adjust the signal contribution of voxels at different spatial positions during three-dimensional reconstruction. This reconstruction weight parameter can be used to compensate for the data sparsity caused by insufficient projection angle coverage in the peripheral area outside the extended field of view (i.e., the area outside the original field of view and within the extended field of view). This reconstruction weight parameter can quantify the data integrity of each voxel through mathematical modeling and accordingly adjust its back-projection integration or iterative update weight, so as to ensure that the signal intensity in the extended area is consistent with that in the central area (i.e., the area within the original field of view), thereby eliminating artifacts and improving image quality.

[0040] In this embodiment, the reconstruction weight parameter can be configured as a function of the distance of the voxel from the imaging center. In this function, the farther the voxel is from the imaging center, the larger the reconstruction weight parameter. In this function, voxels in the central area (i.e., the area within the original field of view) are not given weights (for example, the weight is 1), voxels outside the central area are given weights (the weight is a number greater than 1), and the farther the voxel is from the imaging center, the larger the reconstruction weight parameter.

[0041] In this embodiment, the reconstruction weight parameter can be a linear function or a non-linear function of the distance of the voxel from the imaging center. Specifically, this non-linear function can be a non-linear function that quantifies the data coverage ratio of the voxel at multiple angles. It can also be an exponential non-linear function. For example, the non-linear growth of the weight can be achieved through an exponential function. It can also be a polynomial non-linear function. It can also be designed based on the statistical distribution of projection data. It can also be a non-linear function learned from historical data through machine learning or deep learning.

[0042] The 3D reconstruction data generation method provided in this embodiment can effectively expand the reconstruction field of view of CBCT through algorithm optimization without relying on hardware modification, significantly improve the image quality, and solve problems such as low efficiency and complex operation in traditional technologies. First, by obtaining two-dimensional projection data at multiple projection angles and performing a preset 3D reconstruction algorithm based on these data, the field of view is expanded using the extended field of view parameters to cover a larger range of anatomical structures. Secondly, with the help of the reconstruction weight parameters, the signal attenuation caused by the lack of data at some angles is compensated in the extended field of view, thereby ensuring the accuracy and integrity of the reconstructed image. This method effectively avoids the traditional methods of multiple scans and stitching and hardware modification, reduces the scanning time, radiation dose, and improves the speed and real-time performance of 3D image reconstruction, especially suitable for scenarios with high real-time requirements such as clinical surgical navigation. In addition, this solution is optimized by pure algorithm means, improving the flexibility and stability during the imaging process, and has significant technical advantages.

[0043] In some embodiments, the step of performing a preset 3D reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles to generate 3D reconstruction data for the target includes: Step S142: Determine the actual imaging range that the two-dimensional projection data at each projection angle can cover in the extended field of view.

[0044] In this embodiment, the actual imaging range that the two-dimensional projection data at each projection angle can cover in the extended field of view can be determined based on a preset geometric relationship model. Specifically, a geometric relationship model of the tube (X-ray source), imaging center, and detector can be established in advance. Through this geometric relationship model, the actual imaging range that the two-dimensional projection data at each projection angle can cover in the extended field of view can be determined.

[0045] Step S144: Traverse each pixel point in the actual imaging range, and calculate the corresponding reconstruction weight parameter based on its distance from the imaging center; wherein, the reconstruction weight parameter indicates that for pixel points outside the original field of view parameters, the farther its distance from the imaging center, the greater the reconstruction weight parameter.

[0046] In this embodiment, this step S144 can be performed for the two-dimensional projection data at each projection angle to determine the reconstruction weight parameters of each pixel point in the actual imaging range of each two-dimensional projection data.

[0047] In this embodiment, the distance data corresponding to the pixel points in the actual imaging range of the two-dimensional projection data at each projection angle (that is, the distance between the pixel point and the imaging center) can be input into a preset function model related to at least this distance data to obtain the reconstruction weight parameter.

[0048] As described above, the function model can be a linear function or a non-linear function. For example, it can be a non-linear function that quantifies the data coverage ratio of voxels at multiple angles. It can also be an exponential non-linear function. For example, the non-linear growth of weights can be achieved through an exponential function. It can also be a polynomial non-linear function. It can also be based on the statistical distribution of projection data to design a non-linear function. It can also be a non-linear function learned from historical data through machine learning or deep learning.

[0049] In this embodiment, the function model for reconstructing the weight parameter corresponds at least to the distance data, that is, the distance data is a variable of the function model, and the output of the function model is the reconstructed weight parameter. It can be understood that on the basis of corresponding to the distance data, the function model for reconstructing the weight parameter can also correspond to parameters such as the cone angle of the CBCT, the distance from the X-ray tube to the imaging center, and the effective width of the detector. Of course, it can be understood that these parameters are often fixed parameters.

[0050] In this embodiment, the pixel points within the actual imaging range of the two-dimensional projection data at each projection angle can be input into the above function model to obtain the reconstructed weight parameters corresponding to the pixel points. For a two-dimensional projection data, a reconstructed weight parameter matrix can be constructed. Each element in this matrix is the specific reconstructed weight parameter corresponding to each pixel. By constructing the reconstructed weight parameter matrix corresponding to the two-dimensional projection data, it is convenient for subsequent calculations.

[0051] Step S146: Based on the reconstructed weight parameters, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm, obtain three-dimensional volume data in an extended field of view.

[0052] In this embodiment, the step of obtaining three-dimensional volume data in an extended field of view based on the reconstructed weight parameters, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm may be: First, based on the reconstructed weight parameters, weight the two-dimensional projection data at each projection angle to obtain weighted two-dimensional projection data, and then input the weighted two-dimensional projection data into a preset back-projection reconstruction algorithm for three-dimensional reconstruction to obtain three-dimensional volume data in an extended field of view.

[0053] In this embodiment, the step of obtaining three-dimensional volume data in an extended field of view based on the reconstructed weight parameters, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm may be: Input the two-dimensional projection data at each projection angle into a preset back-projection reconstruction algorithm for three-dimensional reconstruction to obtain unweighted three-dimensional volume data in an extended field of view, and then weight the three-dimensional volume data in the extended field of view using the reconstruction weight parameter to obtain the three-dimensional volume data in the extended field of view.

[0054] In some embodiments, the reconstruction weight parameter can also be embedded in the back-projection reconstruction algorithm, and weighting is performed simultaneously during the back-projection of the two-dimensional projection data at each projection angle to obtain the three-dimensional volume data in the extended field of view.

[0055] In this embodiment, the preset back-projection reconstruction algorithm can be: the FBP back-projection algorithm.

[0056] In this embodiment, the preset back-projection reconstruction algorithm can be: the FDK back-projection algorithm.

[0057] In this embodiment, the preset back-projection reconstruction algorithm can be: an iterative reconstruction algorithm.

[0058] In a specific implementation, back-projection integration can be performed on the two-dimensional projection data at each angle and after RAMP filtering to generate three-dimensional volume data, and then weight normalization operation is performed on the generated projection data, for example: In the formula, and represent the projection angle range, specifically the rotation angle of the C-arm. and represent the radius range of the extended field of view. represents the reconstruction weight parameter, which is a weight function compensated according to different regions of the extended field of view. represents at angle and radius the projection data at, is the Dirac function, used to map the projection data to each pixel point in the three-dimensional space.

[0059] Step S148: Perform weight normalization processing on the three-dimensional volume data to generate three-dimensional reconstruction data for the target.

[0060] To ensure the signal uniformity of the reconstructed three-dimensional volume data, make the signal intensity in the extended field of view region consistent with that in the original field of view, and avoid image distortion caused by excessive compensation during the weighting process. Therefore, in this embodiment, weight normalization processing can be performed on the three-dimensional volume data to generate three-dimensional reconstruction data for the target.

[0061] In this embodiment, weight normalization can be performed by calculating a normalization factor.

[0062] In a specific implementation, for each voxel point in the three-dimensional volume data, the weight integral at all projection angles (i.e., the sum of the reconstruction weight parameters) can be calculated. Using this weight integral as the normalization factor, weight normalization is performed on each voxel point.

[0063] In this embodiment, weight normalization can also be performed by calculating a local normalization factor. For example, for the extended field of view region, the weighted sum of the voxels in these regions is calculated, and then compared with the original field of view region and adjusted using a scaling factor to make the signal intensities of the finally generated three-dimensional reconstruction data consistent.

[0064] By introducing the extended field of view parameter and the reconstruction weight parameter, this embodiment effectively solves the problem of data sparsity caused by insufficient projection angle coverage in the extended field of view range, especially in the region beyond the original field of view. Specifically, by calculating the distance of each pixel point from the imaging center and assigning corresponding reconstruction weight parameters, the difference in image quality between the extended region and the original field of view region can be smoothly transitioned, thus ensuring the balance and consistency of the data within the entire reconstruction region. The weighted projection data is input into the back-projection reconstruction algorithm, ensuring that the signal intensity in the extended field of view region is consistent with that in the central region, avoiding the generation of artifacts, and improving the clarity and accuracy of the image. Finally, the three-dimensional reconstruction data generated after normalization processing can provide a more detailed and accurate image of the target region, effectively improving the quality of image reconstruction. Especially when dealing with large-scale targets, the reconstruction accuracy and efficiency of the extended field of view are ensured.

[0065] As Figure 2 and Figure 3 shown, in some embodiments, the step of traversing each pixel point in the actual imaging range and calculating the corresponding reconstruction weight parameter based on its distance from the imaging center includes: Step S1442: Traverse each pixel point in the actual imaging range to determine the target pixel point; where the target pixel point is a pixel point representing beyond the original field of view parameter.

[0066] In this embodiment, the target pixel point is represented as a pixel point beyond the original field of view parameter. In this embodiment, it can be determined whether a pixel point is a target pixel point according to the distance of each pixel point from the imaging center. Specifically, if the distance of a certain pixel point from the imaging center is greater than the radius corresponding to the original field of view parameter, then this pixel point is a target pixel point.

[0067] Step S1444: Calculate the proportion of the arc length that is not irradiated or the proportion of the arc length that is irradiated in the circumference corresponding to the target pixel point; wherein, the proportion of the arc length that is not irradiated and the proportion of the arc length that is irradiated are used to characterize the degree of data loss.

[0068] In this embodiment, it is possible to calculate the proportion of the arc length that is not irradiated in the circumference corresponding to the target pixel point, or calculate the proportion of the arc length that is irradiated in the circumference corresponding to the target pixel point.

[0069] In this embodiment, the circumference corresponding to the target pixel point can be represented as a circle with the imaging center as the center and the distance from the target pixel point to the imaging center as the radius, thereby forming a circumference. In other words, the target pixel point should be located on this circumference. For example, as Figure 3 points C and B shown in

[0070] In this embodiment, the proportion of the arc length that is not irradiated in the circumference can be expressed as the ratio of the arc length that is not irradiated to the circumference. For example, as Figure 3 shown in

[0071] arc length BC is the arc length that is not irradiated. For points C and B, the proportion of the arc length that is not irradiated can be (2 * arc length BC) / 2πOC.

[0072] In this embodiment, it is possible to calculate the corresponding reconstruction weight parameter based on the proportion of the arc length that is not irradiated. The larger the proportion of the arc length that is not irradiated, the larger the reconstruction weight parameter.

[0073] Specifically, to compensate for data sparsity, the reconstruction weight parameter can be the reciprocal of the proportion that is not irradiated.

[0074] In this embodiment, it is also possible to calculate the corresponding reconstruction weight parameter based on the proportion of the arc length that is irradiated; the larger the proportion of the arc length that is irradiated, the smaller the reconstruction weight parameter.

[0075] In a specific implementation, during CBCT cone beam scanning, in order to ensure that the imaging space is uniformly scanned, the region reconstructed by traditional manufacturers is the largest inscribed circle to ensure that all regions of the imaging space can be covered by a single scan. Specifically, as Figure 3 the small circle in Figure 3 In this implementation, if the CBCT reconstruction field of view is expanded to the range of the large circle in Figure 3The position of the missing area. For this missing area, when the projection angle is the vertical angle, the projection image of this missing area can be collected. However, when the projection angle is the horizontal angle, the projection data of this missing area cannot be collected, which brings the problem of data imbalance.

[0076] To solve the above problems, a correction factor (reconstruction weight parameter) is introduced to the reconstructed image, which represents the weight coefficient introduced due to the different amounts of data from the imaging center in the enlarged reconstruction space. One scanning method is 180 degrees plus a cone angle. The whole scanning ensures that the scanned object is uniformly irradiated at all angles. At the same time, it also ensures that the unirradiated areas are uniformly "missed". Based on this theory, at a projection angle, by calculating the ratio of the irradiated arc length at different radii to the circumference of the same radius, the reduction value of the irradiation intensity at this radius can be obtained. Specifically, Figure 3 In [the figure], the larger radius of the ring is R, that is, OB = R, and OS represents the distance between the X-ray tube and the rotation center (imaging center), denoted as D. Half of the fan angle is denoted as , that is Figure 3 the angle shown as OSA in [the figure]. OA is perpendicular to the outermost ray of the detector. Among them, points B and C are the intersections of this ray and the target ring.

[0077] If the included angle between OA and OC is , then: ; At this time, the length of the short side BC of the arc is: The circumference of the ring is .

[0078] Therefore, for the pixels at the ring with a radius of R, the unirradiated ratio is: This formula shows that when R increases, the angular range of the target pixel irradiated by the cone beam decreases, resulting in gradually decreasing.

[0079] Therefore, it can be understood that the weaker the pixel irradiation, the sparser the data. To compensate for its importance in reconstruction, a higher weight should be given. Therefore, the actual weight used for reconstruction should be: In the formula, when ≤ , that is, within the traditional small field of view, 1, 1. When ≥ , < 1, then > 1, the farther away, the greater the weight. It is monotonically increasing within the extended area, showing a correction trend that the farther away from the center, the higher the weight; represents the fixed distance from the tube (X-ray source) to the imaging center, represents the semi-divergence angle of the fan-beam X-ray.

[0080] In this embodiment, first, for each target pixel point in the extended field of view, a circle is constructed with the imaging center as the center and the distance from the pixel point to the center as the radius. By calculating the proportion of the arc length not covered by X-rays on this circle (i.e., the proportion of the unirradiated arc segment in the entire circle), it directly reflects the sparsity caused by the lack of partial projection angle data at this position; second, based on this proportion, the reconstruction weight parameter is dynamically adjusted (the larger the unirradiated proportion, the higher the weight), so that the signal compensation in the peripheral area strictly matches the physical projection characteristics. Compared with the traditional distance attenuation model, it can eliminate the error introduced by the uniform assumption and improve the contrast and signal-to-noise ratio at the edge of the extended field of view; finally, this method does not need to rely on hardware parameter calibration or multiple scans, and can achieve weight optimization only through the geometric analysis of single-projection data, ensuring the feasibility of intraoperative real-time reconstruction while reducing the computational complexity.

[0081] In some embodiments, the step of performing a preset three-dimensional reconstruction algorithm on the two-dimensional projection data under the multiple projection angles to generate three-dimensional reconstruction data for the target includes: Inputting the two-dimensional projection data under the multiple projection angles into a preset coverage angle number model to obtain the number of times each pixel point in the extended field of view is covered by rays at all projection angles; wherein, the number of times covered by rays is used to characterize the degree of data loss, and the fewer the number of coverage times, the higher the degree of data loss.

[0082] In this embodiment, the coverage angle number model can be a geometric statistical model constructed in advance according to geometric data such as the distance from the tube to the imaging center, the size of the flat panel detector, and the cone angle. This model dynamically evaluates the degree of data loss by quantifying the number of times each pixel point in the extended field of view is covered by X-rays at all projection angles. The more the number of coverage times, the higher the data redundancy, the stronger the signal stability, and the lower the weight compensation requirement; the fewer the number of coverage times, the more significant the data sparsity, and higher weights are required to compensate for signal attenuation.

[0083] Therefore, the two-dimensional projection data at each projection angle can be input into a preset coverage angle quantity model. When each pixel point within the extended field of view is irradiated under this projection angle, it can be understood that there are some pixel points within the extended field of view that are irradiated, and there are also some remaining pixel points that are not irradiated (it can be understood that for the two-dimensional projection data at a certain projection angle, there are two possibilities for the pixel points within its extended field of view: being irradiated and not being irradiated). Therefore, by inputting all the two-dimensional projection data into the preset coverage angle quantity model for statistical calculation, the number of times each pixel point within the extended field of view is covered by the ray at all projection angles can be obtained.

[0084] Based on the number of times covered by the ray, calculate the corresponding reconstruction weight parameter; wherein, the smaller the number of times covered by the ray, the larger the reconstruction weight parameter.

[0085] In this embodiment, a corresponding function model can be pre-constructed based on the number of times covered by the ray. This function model can be a linear function model or a non-linear function model.

[0086] Therefore, in this embodiment, the step of calculating the corresponding reconstruction weight parameter based on the number of times covered by the ray can be to input the number of times covered by the ray into a preset function model related to the number of times covered by the ray for calculation to obtain the reconstruction weight parameter.

[0087] In a possible implementation, the function model related to the number of times covered by the ray can be the maximum coverage number (which can be the value in the central region, that is, the value within the traditional field of view) / (the coverage number corresponding to the current pixel point + smoothing factor).

[0088] Based on the reconstruction weight parameter, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm, three-dimensional volume data within the extended field of view is obtained.

[0089] In this embodiment, the step of obtaining three-dimensional volume data within the extended field of view based on the reconstruction weight parameter, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm can be: First, based on the reconstruction weight parameter, weight the two-dimensional projection data at each projection angle to obtain weighted two-dimensional projection data, and then input the weighted two-dimensional projection data into the preset back-projection reconstruction algorithm for three-dimensional reconstruction to obtain three-dimensional volume data within the extended field of view.

[0090] In this embodiment, the step of obtaining three-dimensional volume data within the extended field of view based on the reconstruction weight parameter, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm can be: Input the two-dimensional projection data at each projection angle into a preset back-projection reconstruction algorithm for three-dimensional reconstruction to obtain unweighted three-dimensional volume data in an extended field of view, and then weight the three-dimensional volume data in the extended field of view using the reconstruction weight parameter to obtain three-dimensional volume data in the extended field of view.

[0091] In some embodiments, the reconstruction weight parameter can also be embedded in the back-projection reconstruction algorithm, and weighting is performed simultaneously during the back-projection of the two-dimensional projection data at each projection angle to obtain three-dimensional volume data in an extended field of view.

[0092] In this embodiment, the preset back-projection reconstruction algorithm can be the FBP back-projection algorithm.

[0093] In this embodiment, the preset back-projection reconstruction algorithm can be the FDK back-projection algorithm.

[0094] Perform weight normalization processing on the three-dimensional volume data to generate three-dimensional reconstruction data for the target.

[0095] In this embodiment, the degree of data loss is measured by counting the number of times each pixel is covered by rays at all projection angles. Its beneficial effects are as follows: First, the coverage count directly quantifies data redundancy, avoiding parameter dependencies and computational overhead of complex geometric models; second, based on the weight calculation using the coverage count (the smaller the coverage count, the larger the weight), signal attenuation in the peripheral region can be compensated more accurately; finally, this method eliminates the bias introduced by the weight through normalization processing, ensuring the overall signal-to-noise ratio of the extended field of view, and without the need to predefine the cone angle or tube distance, significantly enhancing the generality and clinical deployment flexibility of the algorithm.

[0096] In some embodiments, the step of performing a preset three-dimensional reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles to generate three-dimensional reconstruction data for the target includes: Step S1402: Perform a preset first image processing method and / or a preset second image processing method on the two-dimensional projection data at the multiple projection angles; wherein, the first image processing method represents performing image processing on the two-dimensional projection data in the dimension of the image domain to improve image quality; the second image processing method represents performing image processing on the two-dimensional projection data in the dimension of the frequency domain to improve image quality.

[0097] Step S1404: Input the two-dimensional projection data processed by the first image processing method and / or the preset second image processing method into a preset three-dimensional reconstruction algorithm to generate three-dimensional reconstruction data for the target.

[0098] In this embodiment, before performing 3D reconstruction on the two-dimensional projection data at the multiple projection angles, the two-dimensional projection data at the multiple projection angles can be first processed by a first image processing method, and the two-dimensional projection data processed by the first image processing method is input into a preset 3D reconstruction algorithm to generate 3D reconstruction data for the target.

[0099] In this embodiment, before performing 3D reconstruction on the two-dimensional projection data at the multiple projection angles, the two-dimensional projection data at the multiple projection angles can be first processed by a first image processing method, and the two-dimensional projection data processed by the second image processing method is input into a preset 3D reconstruction algorithm to generate 3D reconstruction data for the target.

[0100] In this embodiment, before performing 3D reconstruction on the two-dimensional projection data at the multiple projection angles, the two-dimensional projection data at the multiple projection angles can be sequentially processed by the first image processing method and the second image processing method, and the two-dimensional projection data processed by the first image processing method and the second image processing method is input into a preset 3D reconstruction algorithm to generate 3D reconstruction data for the target.

[0101] In this embodiment, the first image processing method may be image smoothing processing, image noise reduction processing, image contrast enhancement processing, histogram enhancement processing, edge enhancement sharpening processing, or data normalization processing.

[0102] In this embodiment, the second image processing method may be RAMP filtering processing, or may be low-pass or high-pass filtering processing, etc.

[0103] This embodiment significantly improves the quality and efficiency of 3D reconstruction images by performing collaborative preprocessing (noise reduction, contrast enhancement, and frequency domain filtering) on two-dimensional projection data in the image domain and the frequency domain: Image domain processing directly optimizes the signal-to-noise ratio and detail visibility of the original data, and frequency domain processing corrects geometric distortion and suppresses high-frequency noise. The combination of the two can improve the spatial resolution and signal-to-noise ratio of the reconstructed 3D image.

[0104] In some embodiments, the preset first image processing method includes: image smoothing processing, image noise reduction processing, image contrast enhancement processing, histogram enhancement processing, edge enhancement sharpening processing, or data normalization processing.

[0105] In this embodiment, the image smoothing processing may be mean filtering processing or Gaussian filtering processing.

[0106] In this embodiment, the image noise reduction processing may be non-local mean noise reduction processing or wavelet threshold noise reduction processing.

[0107] In this embodiment, the contrast enhancement processing of the image can be linear stretching enhancement or adaptive contrast enhancement.

[0108] In this embodiment, the edge enhancement and sharpening processing can be edge enhancement and sharpening processing based on the Laplacian operator.

[0109] In this embodiment, the data normalization processing can be geometric normalization processing, which is used to correct the intensity attenuation of the X-ray beam and the non-uniformity of the detector response.

[0110] In this embodiment, by performing multi-dimensional preprocessing such as image smoothing processing, noise reduction processing, contrast enhancement, histogram enhancement, edge sharpening, and data normalization on the two-dimensional projection data in the image domain, the quality and consistency of the original two-dimensional projection data are significantly optimized. The smoothing and noise reduction processing can effectively suppress quantum noise and detector noise, thereby improving the signal-to-noise ratio. The contrast enhancement and histogram equalization expand the gray-scale dynamic range, thereby improving the detail visibility of low-contrast anatomical structures. The edge sharpening processing can strengthen the tissue boundary and improve the spatial resolution. The data normalization eliminates the differences in device parameters, ensures the geometric consistency of the multi-angle projection data, and reduces the reconstruction artifacts. These processes work together to provide high signal-to-noise ratio, high-contrast, and geometrically consistent input data for subsequent three-dimensional reconstruction.

[0111] In some cases, when processing a large amount of two-dimensional projection data, especially in application scenarios that require efficient processing of real-time data (such as surgical navigation), the real-time requirement becomes particularly critical. Surgical navigation systems usually need to immediately display the three-dimensional reconstruction data to provide accurate real-time operation guidance for doctors. If the processing is too slow or the delay is too high, it may cause doctors to be unable to obtain the required imaging information in a timely manner during the operation, affecting the accuracy of decision-making and operation, and even may lead to the interruption of the operation or unnecessary risks. For example, in complex surgeries, quickly obtaining information such as bone structures and tumor positions is crucial for precise cutting. If the processing delay is too long, it may cause doctors to miss the best operation opportunity and even be unable to respond to emergencies in a timely manner.

[0112] In some image processing methods, operations such as Fourier transform and frequency domain filtering are often bottlenecks in the processing process because these operations involve complex mathematical transformations on each two-dimensional projection data. Especially for the Fourier transform, it is necessary to transform the image from the spatial domain to the frequency domain, perform frequency domain filtering, and then inverse transform it back to the spatial domain. This process usually significantly increases the computational complexity because each two-dimensional projection data requires multiple mathematical operations, including Fourier transform, filtering, and inverse transform. For real-time processing of a large amount of data, performing these operations one by one will result in significant time delays, unable to meet the system's requirements for fast processing and real-time updates. Therefore, in these real-time applications, excessive delays may directly affect the operation efficiency of doctors and surgical safety.

[0113] This time-consuming problem becomes even more serious, especially when processing data at multiple projection angles. The data at each projection angle needs to independently perform Fourier transform and frequency domain filtering. If each step needs to be processed separately, the overall processing time will increase exponentially, further slowing down the system's response speed and reducing real-time performance. Therefore, how to accelerate the Fourier transform and frequency domain processing and reduce the computational delay is a technical problem that needs to be solved.

[0114] In view of the above technical problems, in some embodiments, the step of performing a preset second image processing method on the two-dimensional projection data at the multiple projection angles includes: Step S14022: Store two two-dimensional projection data that are before and after in the time series into two preset processing spaces in the buffer respectively.

[0115] In this embodiment, the two two-dimensional projection data that are before and after in the time series may be two two-dimensional projection data that are sequential in time during the process of collecting the two-dimensional projection data. Specifically, during the scanning process, the C-arm device will collect two-dimensional projection data of the target area one by one at a certain angular interval (for example, 0.5 degrees to 1 degree). These data are collected sequentially in time, that is, the first projection data is the earliest collected, followed by the second one, and so on. Therefore, the two two-dimensional projection data that are before and after in the time series refer to two adjacent projection data that are collected immediately one after another in time, and their collection order is relative in the entire scanning process.

[0116] In this embodiment, the processing space may be represented as a buffer area preset in the computer system for processing two-dimensional projection data. This processing space may be a dedicated storage area in the memory, used to temporarily store the image data to be processed or being processed. In order to efficiently perform image processing operations such as Fourier transform, frequency domain filtering, and Fourier inverse transform, the two-dimensional projection data can be quickly accessed and stored in these buffer spaces.

[0117] Specifically, one of the two processing spaces can be used to store the real part of the data, and the other processing space can be used to store the imaginary part. Such a design is to support parallel processing, enabling the simultaneous processing of two two-dimensional projection data during the Fourier transform and inverse transform processes, thereby improving the efficiency of data processing.

[0118] Step S14024: Simultaneously perform Fourier transform on the two two-dimensional projection data in the two processing spaces to obtain the frequency-domain data corresponding to the two two-dimensional projection data.

[0119] Step S14026: Simultaneously perform a preset frequency-domain filtering process on the two frequency-domain data to obtain two frequency-domain data after the frequency-domain filtering process.

[0120] In this embodiment, the preset frequency-domain filtering can be low-pass filtering.

[0121] In this embodiment, the preset frequency-domain filtering can be high-pass filtering.

[0122] In this embodiment, the preset frequency-domain filtering can be band-pass filtering.

[0123] In this embodiment, the preset frequency-domain filtering can be RAMP filtering. It can be understood that the RAMP filter corrects geometric distortions in the image and enhances the resolution by applying different gains to each frequency component in the frequency domain. This filtering method can enhance the details of the image, especially at the edges of the image, and improve its contrast.

[0124] In this embodiment, the preset frequency-domain filtering can be mean filtering and Gaussian filtering.

[0125] Step S14028: Simultaneously perform inverse Fourier transform on the two frequency-domain data after the frequency-domain filtering process to obtain two two-dimensional projection data after the frequency-domain filtering process.

[0126] This embodiment processes multiple two-dimensional projection data by using the temporary cache space of a computer. One cache space is used to store the real part, and the other cache space is used to store the imaginary part, so as to achieve parallel processing during the Fourier transform and inverse Fourier transform processes. In this way, two X-ray images can be processed simultaneously at the same time, thus significantly improving the processing speed. This method is particularly suitable for scenarios with high real-time requirements. For example, in surgical navigation systems, in these scenarios, it is crucial to quickly generate high-quality three-dimensional images. Since the processing speed doubles each time, the reconstruction results can be provided more quickly, ensuring that during emergency or complex surgical procedures, doctors can obtain accurate three-dimensional images in real time, improving the efficiency of diagnosis and treatment. At the same time, this method reduces the calculation time, optimizes the utilization of system resources, and improves the response speed and real-time performance of the overall image processing.

[0127] In some cases, during the execution of the fast Fourier transform process, when the scale of the two-dimensional projection data does not meet the preset requirements (for example, it cannot reach a power of 2), it needs to be extended. Otherwise, during the Fourier transform, since the data does not conform to the standard size, the calculation efficiency will be reduced, resulting in an increase in processing time and a decrease in the real-time performance of the system. In addition, when expanding the data scale, if it is only expanded by filling zeros on both sides of the data, such a method may cause the generation of artifacts. Especially when expanding the field of view parameters, these artifacts usually appear at the edges of the image (as Figure 5 shown), but if the expanded field of view parameter increases, the artifacts may shift to the central area of the image, affecting the quality of the final reconstructed image. The appearance of artifacts may mislead clinical diagnosis and affect the doctor's judgment.

[0128] For the above technical problems, as Figure 4 and Figure 5 shown, in some embodiments, the processing space is an extended cache space representing a preset scale in terms of data scale; the step of respectively storing two two-dimensional projection data that are before and after in the time series into two preset processing spaces in the cache includes: Step S140222: Determine whether the scale of the two two-dimensional projection data meets the preset scale in at least one data dimension; wherein, the preset scale is greater than the data scale of the two-dimensional projection data.

[0129] In this embodiment, it can be determined whether the scale in one or two data dimensions meets the preset scale. This data dimension can be represented as the data dimension of the X-axis or the Y-axis of the two-dimensional projection data.

[0130] In this embodiment, the preset scale can be a power of 2. Specifically, for Fourier transform, especially the Fast Fourier Transform (FFT) algorithm, the optimal computational efficiency can be achieved at a scale that is a power of 2. By expanding the data to a power of 2, the FFT algorithm can make full use of the binary structure, making the calculation process more efficient and fast.

[0131] Step S140224: When the preset scale is not satisfied, store the two two-dimensional projection data respectively in the middle of two preset extended buffer spaces in the buffer.

[0132] In this embodiment, the extended buffer space can be an extended buffer space with a preset scale in at least one data dimension. Specifically, this extended buffer space can be an extended buffer space with a preset scale in the data dimension of the X-axis.

[0133] In this embodiment, storing the two-dimensional projection data in the middle of the extended buffer space means placing the original two-dimensional projection data at the central position of the buffer space and padding on both sides of the data. For example, if the X-axis size of the extended buffer space is 1024 and the X-axis size of the original projection data is 512, then the original data will be stored in the middle of the 1024 dimension, and the remaining space (512 pixels) will be evenly distributed on both sides of the data, forming data expansion. This storage method ensures the symmetry of the data and ensures that the extended area can be effectively utilized when performing Fourier transform, thereby improving the computational efficiency and reducing errors and artifacts in the transformation process.

[0134] Step S140226: For the two two-dimensional projection data, in at least one dimension, obtain the projection data at a preset distance on both sides of the two-dimensional projection data in this dimension.

[0135] In this embodiment, the projection data at a preset distance on both sides of the two-dimensional projection data in this dimension can be obtained in one or two data dimensions.

[0136] It can be understood that if only one data dimension is expanded, obtain the projection data at a preset distance on both sides of the data dimension that needs to be expanded.

[0137] In this embodiment, the preset distance can be a preset number of pixels, such as 10 pixels, 20 pixels, etc.

[0138] Step S140228: Based on the projection data at the preset distance on both sides and a preset virtual data generation algorithm, generate the virtual projection data on the corresponding side to fill the extended buffer space.

[0139] In this embodiment, the step of generating virtual projection data for the corresponding side based on the projection data on both sides with a preset distance and a preset virtual data generation algorithm to fill the extended cache space may be as follows: The projection data of one row within the preset distance on one side (for example, the dimension is the X-axis dimension) can be input into the preset virtual data generation algorithm for data generation to generate the virtual projection data of one row on this side. In the same way, by traversing all the rows of projection data within the preset distance on this side, all the rows of virtual projection data on this side can be generated. The virtual data generation algorithm can be the RANCS algorithm.

[0140] In this embodiment, it is also possible to generate the virtual projection data of one row on one side based on the projection data of multiple rows within the preset distance on this side.

[0141] In this embodiment, it is also possible to generate the virtual projection data of one row on one side based on all the rows of projection data within the preset distance on this side.

[0142] In this embodiment, it is also possible to generate the virtual projection data of multiple rows on one side based on the projection data of multiple rows within the preset distance on this side.

[0143] In this embodiment, it is also possible to generate all the rows of virtual projection data on one side based on all the rows of projection data within the preset distance on this side.

[0144] In this embodiment, the virtual data generation algorithm can be the RANCS algorithm.

[0145] In this embodiment, the virtual data generation algorithm can be a linear interpolation algorithm.

[0146] In this embodiment, the virtual data generation algorithm can be a polynomial interpolation algorithm.

[0147] In this embodiment, the virtual data generation algorithm can be a virtual data generation algorithm based on deep learning. For example, a deep learning model can be used to learn the training data to infer the mapping relationship of the missing data from the existing projection data. Through a large amount of training data, the deep learning model can learn the complex relationships between the data and accurately generate the missing data.

[0148] In this embodiment, the virtual data generation algorithm can also be a virtual data generation algorithm based on a pre-constructed physical model.

[0149] In this embodiment, on the one hand, by scaling up the two-dimensional projection data to a preset scale (for example, a power of 2), it is ensured that the data meets the requirements of the fast Fourier transform. The fast Fourier transform often needs to expand the data to a size that is a power of 2 during the processing to optimize the calculation efficiency. By expanding the data to the preset scale when the data dimension does not meet it, the processing speed of the Fourier transform can be increased, avoiding the additional computational overhead caused by inconsistent data scales, thereby greatly improving the overall data processing efficiency. On the other hand, when expanding the cache space, by filling virtual projection data on both sides instead of simply filling with 0, it can effectively avoid the highlighted circular artifacts caused by filling with 0 in related methods. In some back-projection or three-dimensional reconstruction algorithms, such artifacts usually appear at the edges of the image and do not affect the user's diagnosis. However, if the field of view parameter is expanded, the artifacts may move to the middle area of the image, affecting the overall quality and diagnostic results of the image. Therefore, by filling virtual projection data to compensate for the expanded area, the continuity and accuracy of the reconstructed image can be maintained, preventing artifact interference, and ensuring that the quality of the final three-dimensional reconstructed image is clearer and more reliable. Especially in the case of expanding the field of view, more accurate imaging results can be provided.

[0150] According to an embodiment of the present invention, an electronic device is provided. Please refer to Figure 6 . The electronic device in this embodiment may include one or more of the following components: a processor, a network interface, a memory, a non-volatile memory, and one or more application programs. One or more application programs may be stored in the non-volatile memory and configured to be executed by one or more processors. One or more programs are configured to execute the method described in the foregoing method embodiments.

[0151] According to an embodiment of the present invention, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a computer, the computer executes the method described in any of the above embodiments.

[0152] According to an embodiment of the present invention, a computer program product containing instructions is further provided. When the instructions are executed by a computer, the computer executes a method in any of the above embodiments.

[0153] It should be noted that the terms "first", "second", etc. in the description, claims and the above-mentioned drawings of this application are used to distinguish similar objects, and do not necessarily have to be used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged under appropriate circumstances so that the embodiments of the present application described here can be implemented in an order other than those illustrated or described here. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion. For example, a process, method, system, product or device that comprises a series of steps or units does not necessarily have to be limited to those steps or units clearly listed, but may include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0154] Optionally, the specific examples in this embodiment may refer to the examples described in the above embodiments, and will not be elaborated here.

[0155] The serial numbers of the above embodiments of the present application are only for description and do not represent the advantages or disadvantages of the embodiments.

[0156] In the above embodiments of the present application, the descriptions of the various embodiments have their own emphases. For the parts not detailed in a certain embodiment, reference may be made to the relevant descriptions of other embodiments.

[0157] The above are only the preferred embodiments of the present application. It should be noted that for those of ordinary skill in the art, without departing from the principle of the present application, several improvements and refinements can be made, and these improvements and refinements should also be regarded as the protection scope of the present application.

Claims

1. A method for generating three-dimensional reconstruction data, characterized in that Including: Obtaining two-dimensional projection data at multiple projection angles for the same target; Based on the two-dimensional projection data at the multiple projection angles, performing a preset three-dimensional reconstruction algorithm to generate three-dimensional reconstruction data for the target; wherein, an extended field of view parameter and a reconstruction weight parameter corresponding to the extended field of view parameter are preset in the preset three-dimensional reconstruction algorithm; wherein, the extended field of view parameter is a field of view parameter increased on the basis of the original field of view parameter, and the original field of view parameter is a field of view parameter that can ensure that the two-dimensional projection data at all projection angles can cover the target during the three-dimensional reconstruction process; wherein, the reconstruction weight parameter is used to compensate for the signal attenuation caused by the missing data of the pixel points beyond the original field of view in the extended field of view during the three-dimensional reconstruction process.

2. The method according to claim 1, wherein The step of performing a preset three-dimensional reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles to generate three-dimensional reconstruction data for the target includes: Determining the actual imaging range that the two-dimensional projection data at each projection angle can cover in the extended field of view; Traversing each pixel point in the actual imaging range, and calculating the corresponding reconstruction weight parameter based on its distance from the imaging center; wherein, the reconstruction weight parameter indicates that for the pixel points beyond the original field of view parameter, the farther its distance from the imaging center, the larger the reconstruction weight parameter; Based on the reconstruction weight parameter, the two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm, obtaining three-dimensional volume data in the extended field of view; Performing weight normalization processing on the three-dimensional volume data to generate three-dimensional reconstruction data for the target.

3. The method according to claim 2, characterized in that The step of traversing each pixel point in the actual imaging range and calculating the corresponding reconstruction weight parameter based on its distance from the imaging center includes: Traversing each pixel point in the actual imaging range to determine the target pixel point; wherein, the target pixel point is a pixel point that represents beyond the original field of view parameter; Calculating the proportion of the arc length not irradiated or the proportion of the arc length irradiated in the circumference corresponding to the target pixel point; wherein, the proportion of the arc length not irradiated and the proportion of the arc length irradiated are used to characterize the degree of data loss; Based on the proportion of the arc length not irradiated or the proportion of the arc length irradiated, calculating the corresponding reconstruction weight parameter; wherein, the larger the proportion of the arc length not irradiated, the larger the reconstruction weight parameter, and the larger the proportion of the arc length irradiated, the smaller the reconstruction weight parameter.

4. The method according to claim 1, wherein The step of performing a preset three-dimensional reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles to generate three-dimensional reconstruction data for the target includes: Inputting the two-dimensional projection data at the multiple projection angles into a preset coverage angle number model to obtain the number of times each pixel point in the extended field of view is covered by rays at all projection angles; wherein, the number of times covered by rays is used to characterize the degree of data loss, and the fewer the number of coverage times, the higher the degree of data loss; Based on the number of times covered by rays, calculating the corresponding reconstruction weight parameter; wherein, the smaller the number of times covered by rays, the larger the reconstruction weight parameter; Based on the reconstructed weight parameters, two-dimensional projection data at each projection angle, and a preset back-projection reconstruction algorithm, three-dimensional volume data in an extended field of view is obtained; Perform weight normalization processing on the three-dimensional volume data to generate three-dimensional reconstruction data for the target.

5. The method according to claim 1, wherein The step of generating three-dimensional reconstruction data for the target by performing a preset three-dimensional reconstruction algorithm based on the two-dimensional projection data at the multiple projection angles includes: Perform a preset first image processing method and / or a preset second image processing method on the two-dimensional projection data at the multiple projection angles; wherein, the first image processing method represents performing image processing on the two-dimensional projection data in the dimension of the image domain to improve the image quality; the second image processing method represents performing image processing on the two-dimensional projection data in the dimension of the frequency domain to improve the image quality; Input the two-dimensional projection data processed by the first image processing method and / or the preset second image processing method into a preset three-dimensional reconstruction algorithm to generate three-dimensional reconstruction data for the target.

6. The method according to claim 5, wherein The preset first image processing method includes: image smoothing processing, image noise reduction processing, image contrast enhancement processing, histogram enhancement processing, edge enhancement and sharpening processing, or data normalization processing.

7. The method according to claim 5, characterized in that The step of performing the preset second image processing method on the two-dimensional projection data at the multiple projection angles includes: Store two two-dimensional projection data that are before and after in the time series into two preset processing spaces in the cache respectively; Perform Fourier transform on the two two-dimensional projection data in the two processing spaces simultaneously to obtain frequency domain data corresponding to the two two-dimensional projection data; Perform a preset frequency domain filtering process on the two frequency domain data simultaneously to obtain two frequency domain data after the frequency domain filtering process; Perform inverse Fourier transform on the two frequency domain data after the frequency domain filtering process simultaneously to obtain two two-dimensional projection data after the frequency domain filtering process.

8. The method according to claim 7, wherein The processing space represents an extended cache space with a preset scale in the data scale; The step of storing two two-dimensional projection data that are before and after in the time series into two preset processing spaces in the cache respectively includes: Judge whether the scale of the two two-dimensional projection data in at least one data dimension meets a preset scale; wherein, the preset scale is greater than the data scale of the two-dimensional projection data; In the case of not meeting the preset scale, store the two two-dimensional projection data into the middle of two preset extended cache spaces in the cache respectively; For the two two-dimensional projection data, at least in one dimension, obtain projection data at a preset distance on both sides of the two-dimensional projection data in this dimension; Based on the projection data at the preset distance on both sides and a preset virtual data generation algorithm, generate virtual projection data on the corresponding side to fill the extended cache space.

9. An electronic device, characterized in that, Includes: A memory, and one or more processors communicatively connected to the memory; Instructions executable by the one or more processors are stored in the memory, and the one or more processors execute the instructions to cause the one or more processors to implement the method according to any one of claims 1 to 8.

10. A computer-readable storage medium, characterized in that, A computer program is stored in the readable storage medium, and when the computer program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images