SPECT pinhole collimator imaging truncation artifact filling method and related equipment

By acquiring, reconstructing, and using deep learning networks to predict data to fill in truncation artifacts in SPECT imaging, the problem of image quality degradation caused by incomplete probe coverage was solved, achieving high-quality image reconstruction and improved diagnostic accuracy.

CN121437683APending Publication Date: 2026-01-30RISHI XINHE (HEBEI) MEDICAL TECH CO LTD
View PDF 0 Cites 2 Cited by

Patent Information

Application Number
CN202511420480.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-09-30
Publication Date
2026-01-30

AI Technical Summary

Technical Problem

In CT imaging, because the SPECT probe cannot completely cover the human body, truncation artifacts occur, affecting image quality and diagnostic accuracy.

Method used

By collecting real projection data, the original image is reconstructed, high-confidence regions are identified, projection pixel mapping relationships are established, missing data is predicted using a deep learning network, and weighted reconstruction and image fusion are performed in combination with confidence allocation rules to fill in the truncated regions.

Benefits of technology

It significantly improves image quality and diagnostic accuracy, effectively eliminates truncation artifacts, and provides clearer and more reliable image evidence.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN121437683A_ABST
    Figure CN121437683A_ABST
Patent Text Reader

Abstract

The invention provides an SPECT pinhole collimator imaging truncation artifact filling method and related equipment, and relates to the technical field of CT imaging. The method comprises the following steps: reconstructing real projection data to obtain an original reconstructed image, and determining a high-confidence region in the original reconstructed image; mapping the real projection data to a projection image of an ideal detector radial distance to obtain mapping projection data; predicting projection data outside an effective area of a real acquisition detector and taking the projection data as simulation projection data; carrying out weighted reconstruction on the simulation projection data to obtain a complemented image containing a truncation region; and according to the high-confidence region, fusing the original reconstructed image and the complemented image to obtain a final image. The method aims at solving the problem of truncation artifacts caused by the fact that the probe cannot completely cover the whole human body projection image, negative effects caused by the truncation artifacts can be made up, and the effect of improving the imaging quality and the diagnosis accuracy is achieved.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of CT imaging technology, and more specifically, to a method and related equipment for filling artifacts in SPECT pinhole collimator imaging truncation. Background Technology

[0002] In CT imaging, SPECT probes are typically equipped with pinhole collimators, which allow for focusing on small organs and lesions, resulting in higher resolution images. Furthermore, the detection efficiency of a SPECT probe is inversely proportional to the distance between the probe and the body. To maximize detection efficiency, the probe should extend as far as possible to achieve sufficient clarity and contrast for organs or lesions. However, when the patient is obese and a specific organ or tumor is being scanned, the probe may not be able to capture the entire projected image of the body at certain angles to maximize detection efficiency. This results in truncation artifacts in the reconstructed tomographic images of the body area. Further, although the probe is extended as far as possible to cover the core torso to maximize the detection efficiency of the core region of interest, there are still instances where the probe may extend beyond the detection range of the arms, also leading to truncation artifacts. These truncation artifacts degrade image quality and can affect the physician's accurate diagnosis of lesion boundaries. Therefore, mitigating the negative impact of truncation artifacts has become a pressing technical problem in this field. Summary of the Invention

[0003] The purpose of this invention is to provide a method and related equipment for filling truncation artifacts in SPECT pinhole collimator imaging. This method aims to solve the problem of truncation artifacts caused by the probe not being able to completely cover the entire human body projection image in SPECT imaging. It can compensate for the negative impact of truncation artifacts and improve imaging quality and diagnostic accuracy.

[0004] In a first aspect, the present invention provides a method for filling in truncation artifacts in SPECT pinhole collimator imaging, comprising the following steps:

[0005] Collect real projection data;

[0006] The original reconstructed image is obtained by reconstructing the real projection data.

[0007] Based on the collimator configuration and collimator position when collecting real projection data, determine the high-confidence region in the original reconstructed image;

[0008] Based on the original reconstructed image, a projection pixel mapping relationship is established between the radial distance of the real detector and the radial distance of the ideal detector;

[0009] By utilizing the projection pixel mapping relationship, the real projection data is mapped onto the projection image of the radial distance of the ideal detector to obtain the mapped projection data;

[0010] Obtain the projection image of the attenuation image onto the radial distance of the ideal detector and use it as prior projection data;

[0011] By inputting the mapped projection data and prior projection data into a deep learning network, the projection data outside the effective area of ​​the real acquisition detector is predicted and used as the simulated projection data.

[0012] Based on the preset confidence allocation rules, a complete image containing the truncated region is obtained by weighted reconstruction of the simulated projection data.

[0013] Based on the high-confidence regions, the final image is obtained by fusing the original reconstructed image and the completed image.

[0014] The SPECT pinhole collimator imaging truncation artifact filling method provided by this invention can effectively solve the truncation artifact problem caused by incomplete detector coverage in SPECT imaging. It predicts missing data through a deep learning network and fuses it with high-confidence regions, which significantly improves image quality and diagnostic accuracy.

[0015] Furthermore, the steps for reconstructing the original reconstructed image from the true projection data include:

[0016] The initial image is obtained by reconstructing the actual projection data;

[0017] The initial image is preprocessed to obtain the original reconstructed image.

[0018] It can effectively eliminate or significantly reduce image artifacts and noise caused by gamma photon attenuation and scattering effects.

[0019] Further preprocessing includes attenuation correction and scattering correction.

[0020] Furthermore, based on the original reconstructed image, the steps for establishing the projected pixel mapping relationship between the real detector radial distance and the ideal detector radial distance include:

[0021] The first projected image is obtained by projecting the original reconstructed image according to the radial distance of the real detector;

[0022] Based on the original reconstructed image, calculate the radial distance of an ideal detector capable of completely scanning the entire human body surface;

[0023] A second projected image is obtained by projecting the original reconstructed image along the radial distance of an ideal detector;

[0024] Based on the first and second projected images, the projected pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector is obtained.

[0025] It can establish the projected pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector in a more accurate and systematic way.

[0026] Furthermore, the attenuation image is an attenuation coefficient distribution map generated using CT scan data.

[0027] Furthermore, the confidence assignment rules include assigning higher weights to mapped projected data compared to simulated projected data.

[0028] Furthermore, based on the high-confidence regions, the steps to obtain the final image by fusing the original reconstructed image and the completed image include:

[0029] Based on the high-confidence regions, the original reconstructed image and the completed image are fused using a mask interpolation method to obtain the final image.

[0030] Using mask interpolation for image fusion can significantly improve the quality and reliability of the final image.

[0031] Secondly, the present invention provides a device for filling artifacts in SPECT pinhole collimator imaging truncation, comprising:

[0032] The acquisition module is used to acquire real projection data;

[0033] The reconstruction module is used to reconstruct the real projection data to obtain the original reconstructed image;

[0034] The determination module is used to determine the high-confidence region in the original reconstructed image based on the collimator configuration and collimator position when acquiring real projection data;

[0035] A module is established to create a projected pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector based on the original reconstructed image;

[0036] The mapping module is used to map real projection data onto the projection image of the radial distance of the ideal detector using the projection pixel mapping relationship, so as to obtain mapped projection data;

[0037] The acquisition module is used to acquire the projection image of the attenuation image on the radial distance of the ideal detector and use it as prior projection data;

[0038] The prediction module is used to predict projection data outside the effective area of ​​the real acquisition detector and use it as simulated projection data by inputting the mapped projection data and prior projection data into the deep learning network.

[0039] The completion module is used to obtain a completed image containing the truncated region by performing weighted reconstruction on the simulated projection data based on a preset confidence allocation rule.

[0040] The fusion module is used to obtain the final image by fusing the original reconstructed image and the completed image based on high-confidence regions.

[0041] The SPECT pinhole collimator imaging truncation artifact filling device provided by the present invention can make full use of the accurate information of the high confidence region in the original image, and effectively fill the truncation region, so as to obtain a high-quality, artifact-free image.

[0042] Thirdly, the present invention provides an electronic device including a processor and a memory, the memory storing computer-readable instructions, which, when executed by the processor, perform the steps of the SPECT pinhole collimator imaging truncation artifact filling method provided in the first aspect above.

[0043] Fourthly, the present invention provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, performs the steps of the SPEC T pinhole collimator imaging truncation artifact filling method provided in the first aspect above.

[0044] As described above, the SPECT pinhole collimator imaging truncation artifact filling method provided by this invention obtains the original image by acquiring real projection data and reconstructing it, and then determines the high-confidence region based on the collimator configuration and position. Based on this, a projection pixel mapping relationship is established between the real detector and the ideal detector, mapping the real projection data onto the radial distance of the ideal detector. Combining the attenuated image as prior projection data, a deep learning network predicts the projection data outside the real acquisition area. Finally, the simulated projection data is weighted and reconstructed based on confidence allocation rules to obtain the completed image, which is then fused with the original reconstructed image in the high-confidence region to obtain the final image. This method effectively solves the truncation artifact problem caused by incomplete detector coverage in SPECT imaging in existing technologies, significantly improving the quality of the reconstructed image and diagnostic accuracy. By introducing a deep learning network for data prediction, the limitations of traditional methods in handling complex truncation artifacts are overcome, achieving intelligent completion of missing data. Meanwhile, the identification of high-confidence regions and the weighted reconstruction strategy ensured accuracy in the areas covered by real data, avoided the introduction of artifacts, and made the final image both complete and reliable, providing clearer and more accurate imaging evidence for clinical diagnosis.

[0045] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing embodiments of the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings. Attached Figure Description

[0046] Figure 1 This is a flowchart of a SPECT pinhole collimator imaging truncation artifact filling method provided in an embodiment of the present invention.

[0047] Figure 2 This is an image of real projection data in an embodiment of the present invention.

[0048] Figure 3 This is a schematic diagram of a first projected image in an embodiment of the present invention.

[0049] Figure 4 This is a schematic diagram of a second projected image in an embodiment of the present invention.

[0050] Figure 5 This is an image of mapped projection data in an embodiment of the present invention.

[0051] Figure 6 This is an image of prior projection data in an embodiment of the present invention.

[0052] Figure 7 This is an image of simulated projection data predicted using mapped projection data and prior projection data in an embodiment of the present invention.

[0053] Figure 8 These are images before and after completion in an embodiment of the present invention.

[0054] Figure 9 This is a schematic diagram of a SPECT pinhole collimator imaging truncation artifact filling device provided in an embodiment of the present invention.

[0055] Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention.

[0056] Label Explanation:

[0057] 100. Acquisition Module; 200. Reconstruction Module; 300. Determination Module; 400. Establishment Module; 500. Mapping Module; 600. Acquisition Module; 700. Prediction Module; 800. Completion Module; 900. Fusion Module; 13. Electronic Equipment; 1301. Processor; 1302. Memory; 1303. Communication Bus. Detailed Implementation

[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the claimed invention, but merely to illustrate selected embodiments of the invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without inventive effort are within the scope of protection of the present invention.

[0059] It should be noted that similar reference numerals and letters in the following figures indicate similar items; therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures. Furthermore, in the description of this invention, terms such as "first," "second," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0060] Reference Appendix Figure 1 Appendix Figure 2 Appendix Figure 3 Appendix Figure 4 Appendix Figure 5 Appendix Figure 6 Appendix Figure 7 and attached Figure 8 This invention provides a method for filling in truncation artifacts in SPECT pinhole collimator imaging, comprising the following steps:

[0061] Collect real projection data;

[0062] The original reconstructed image is obtained by reconstructing the real projection data.

[0063] Based on the collimator configuration and position when collecting real projection data, the high confidence region in the original reconstructed image is determined (by the collimator configuration (aperture, angle) and position, it is confirmed which voxels in the three-dimensional space are contributed by each angle detector, so as to obtain a mask of the three-dimensional space voxels that can be seen by all detectors or by some detectors. Based on this mask, the intervals that will not have truncation artifacts and have the highest confidence can be determined).

[0064] Based on the original reconstructed image, a projection pixel mapping relationship is established between the radial distance of the real detector and the radial distance of the ideal detector;

[0065] By utilizing the projection pixel mapping relationship, the real projection data is mapped onto the projection image of the radial distance of the ideal detector to obtain the mapped projection data;

[0066] The projection image of the attenuation image onto the radial distance of the ideal detector is acquired and used as prior projection data; the attenuation image is an attenuation coefficient distribution map generated using CT scan data, which is used to correct the Compton scattering and photoelectric absorption effects that occur when gamma photons pass through human tissues of different densities and compositions.

[0067] By inputting the mapped projection data and prior projection data into a deep learning network, the projection data outside the effective area of ​​the real acquisition detector is predicted and used as the simulated projection data.

[0068] Based on the preset confidence allocation rules, a complete image containing the truncated region is obtained by weighted reconstruction of the simulated projection data. The confidence allocation rules include assigning higher weights to the mapped projection data compared to the simulated projection data (so that the voxels in the three-dimensional space are contributed more by the pixels of the real detector, rather than by the average contribution of the pixels of the real detector and the ideal detector).

[0069] Based on the high-confidence regions, the final image is obtained by fusing the original reconstructed image and the completed image.

[0070] "True projection data" refers to the raw data collected by the SPECT system through the pinhole collimator during the actual scanning process. This data may be truncated due to the limitations of the detector's coverage area.

[0071] "Original reconstructed image" refers to the image obtained after preliminary reconstruction of the real projection data, which may contain truncation artifacts.

[0072] "High confidence region" refers to the region in the original reconstructed image that, determined by the collimator configuration (aperture, angle) and position, will not exhibit truncation artifacts and has the highest confidence.

[0073] "Actual detector radial distance" refers to the radial distance between the actual SPECT detector and the imaging center during the scanning process.

[0074] "Ideal detector radial distance" refers to the radial distance between the detector and the imaging center, assuming the detector can completely scan the entire human body surface.

[0075] "Projected pixel mapping relationship" refers to a correspondence between projected pixels on the radial distance of a real detector and projected pixels on the radial distance of an ideal detector.

[0076] "Mapped projection data" refers to projection data that converts real projection data to the radial distance of an ideal detector based on the mapping relationship between projection pixels.

[0077] "Attenuation image" refers to an attenuation coefficient distribution map generated using CT scan data, used to correct for Compton scattering and photoelectric absorption effects that occur when gamma photons pass through human tissues of different densities and compositions.

[0078] "Prior projection data" refers to the projection of the attenuated image onto the radial distance of the ideal detector, serving as auxiliary information for prediction by the deep learning network.

[0079] "Simulated projection data" refers to the projection data outside the effective area of ​​the actual acquisition detector predicted by the deep learning network, which is used to fill in the truncated area.

[0080] "Confidence assignment rules" refer to the rules used to determine the weights of different data sources (mapped projection data and simulated projection data) when performing weighted reconstruction of simulated projection data.

[0081] "Complete image" refers to an image obtained through weighted reconstruction that includes information to fill in truncated regions.

[0082] The "final image" refers to a high-quality image without truncation artifacts obtained by fusing the original reconstructed image with the completed image.

[0083] The implementation environment of this application typically includes a SPECT imaging system, a CT scanning system, and computer equipment for data processing and image reconstruction.

[0084] The SPECT pinhole collimator imaging truncation artifact filling method proposed in this application is based on a series of steps to effectively identify, predict and fill truncation artifacts caused by insufficient detector coverage.

[0085] First, true projection data needs to be acquired. This can be done through a routine scan using a SPECT system. For example, during a focused scan of a patient's heart or thyroid gland, the SPECT detector rotates around the patient, collecting gamma photon signals in the process to form a series of two-dimensional projection images. This raw acquired data is the true projection data.

[0086] Next, the true projection data is reconstructed to obtain the original reconstructed image. The reconstruction process can employ various existing techniques, such as iterative reconstruction algorithms (e.g., OSEM) or filtered back-projection algorithms. For example, the acquired true projection data can be input into reconstruction software, and by setting the appropriate reconstruction parameters, a three-dimensional original reconstructed image can be generated.

[0087] Subsequently, based on the collimator configuration and position during the acquisition of real projection data, high-confidence regions in the original reconstructed image are determined. Specifically, based on the collimator aperture, angle, and the spatial position of the detector during the scanning process, it can be calculated which voxels in three-dimensional space can be completely covered by all detector angles, or at least partially covered with high confidence by some detector angles. For example, through geometric projection analysis, a three-dimensional mask can be generated that identifies regions far from the detector edges and less affected by truncation; these regions are the high-confidence regions.

[0088] Based on this, a projection pixel mapping relationship is established between the radial distance of the real detector and the radial distance of the ideal detector, using the original reconstructed image. For example, the original reconstructed image can first be projected along the radial distance of the real detector to obtain a first projected image. Then, based on the size and scanning range of the original reconstructed image, the radial distance of an ideal detector capable of completely covering the entire human body surface is calculated. Next, the original reconstructed image is projected along this radial distance of the ideal detector to obtain a second projected image. Finally, by comparing the first and second projected images, a pixel correspondence between them is established, thus obtaining the projection pixel mapping relationship.

[0089] By utilizing the projection pixel mapping relationship, the real projection data is mapped onto the projection image of the ideal detector's radial distance, thus obtaining the mapped projection data. For example, based on the previously established mapping relationship, each pixel in the real projection data can be interpolated or resampled to adjust its position to the position corresponding to the radial distance of the ideal detector, thereby generating a projection image with complete coverage of the ideal detector's radial distance.

[0090] Simultaneously, the projection image of the attenuation image onto the radial distance of the ideal detector is acquired and used as prior projection data. The attenuation image is typically an attenuation coefficient distribution map generated from CT scan data, used to correct for Compton scattering and photoelectric absorption effects that occur when gamma photons pass through human tissues of different densities and compositions. For example, the attenuation image obtained from a CT scan can be forward-projected to obtain its projection image onto the radial distance of the ideal detector.

[0091] Then, by inputting the mapped projection data and prior projection data into a deep learning network, the network predicts projection data outside the effective region of the real acquisition detector and uses this as simulated projection data. The deep learning network can be a convolutional neural network (CNN), trained to learn and predict missing projection data from partial projection data and prior information. For example, by using the mapped projection data and prior projection data as input, the network extracts and learns features through multiple convolutional, pooling, and fully connected layers, ultimately outputting simulated projection data outside the effective region of the real acquisition detector to fill in the truncated region.

[0092] Next, based on a preset confidence allocation rule, a weighted reconstruction is performed on the simulated projection data to obtain a completed image containing the truncated region. The confidence allocation rule can be simply set so that mapped projection data is given a higher weight than simulated projection data. For example, during the reconstruction process, regions contributed by mapped projection data are given a higher weight, while regions contributed by simulated projection data are given a lower weight. This ensures that the voxels in 3D space are contributed more by pixels from the real detector, rather than by an average contribution from pixels from both the real and ideal detectors.

[0093] Finally, based on the high-confidence regions, the original reconstructed image and the completed image are fused to obtain the final image. For example, the high-confidence regions can be used as masks to directly retain the pixel values ​​of the high-confidence regions in the original reconstructed image, while the pixel values ​​of the corresponding truncated regions in the completed image are filled or interpolated, thus obtaining a final image that both retains the high-confidence information of the real data and fills in the truncated regions.

[0094] The SPECT pinhole collimator imaging truncation artifact filling method proposed in this application works by first acquiring real projection data and performing preliminary reconstruction to obtain an original reconstructed image that may contain truncation artifacts. To address the truncation problem, this application introduces the concept of "ideal detector radial distance" and establishes a projection pixel mapping relationship between the real detector and the ideal detector. This mapping relationship allows the real projection data to be "extended" to the full range covered by the ideal detector, thus providing a foundation for subsequent truncation filling.

[0095] After acquiring the mapped projection data and prior projection data (generated from CT attenuation images), this application ingeniously utilizes a deep learning network to predict projection data outside the effective area of ​​the real acquisition detector. By learning patterns in a large amount of data, the deep learning network can infer the missing projection data from limited real data and auxiliary information, thereby effectively simulating the real situation of the truncated region. This step is one of the core innovations of this application, overcoming the limitations of traditional methods in dealing with complex truncated artifacts.

[0096] Subsequently, during the reconstruction process, this application introduces a pre-defined confidence allocation rule to perform weighted reconstruction of the simulated projection data. This rule ensures that data collected by the real detector (mapped projection data) has a higher weight in the reconstruction, while simulated projection data predicted by the deep learning network serves as a supplement, participating in the reconstruction with a lower weight. This weighting strategy guarantees the accuracy of the reconstructed image within the area covered by the real data, while effectively filling in truncated regions and avoiding errors that may arise from over-reliance on predicted data.

[0097] Finally, the original reconstructed image and the completed image are fused to obtain the final image. The fusion process utilizes previously identified high-confidence regions to ensure that accurate portions of the original reconstructed image are preserved, while truncated regions are filled in with information from the completed image. The entire process forms a closed loop, with each step closely linked and working together from data acquisition to final image generation, aiming to provide a high-quality SPECT image free of truncation artifacts.

[0098] The SPECT pinhole collimator imaging truncation artifact filling method proposed in this application has significant advantages and innovations compared to existing technologies. Traditional methods for handling truncation artifacts in SPECT imaging often rely on simple interpolation or extrapolation techniques. These methods often lack accuracy and robustness when dealing with complex human structures and irregular truncation regions, easily introducing new artifacts or causing image distortion. For example, some image domain-based filling methods may fail to fully utilize information from the projection domain, resulting in unsatisfactory filling effects.

[0099] The core innovation of this application lies in introducing a deep learning network to predict projection data outside the effective area of ​​the real acquisition detector, and combining this with a confidence allocation rule for weighted reconstruction. Specifically, by establishing a projection pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector, this application can effectively extend the limited real projection data to a wider range of ideal detectors, providing a more comprehensive input for the prediction of the deep learning network. Leveraging its powerful feature learning and pattern recognition capabilities, the deep learning network can learn complex projection patterns from the mapped projection data and prior projection data, thereby accurately predicting the missing data in the truncated region.

[0100] Furthermore, the confidence allocation rule introduced in this application during the reconstruction process ensures the dominance of the real-world acquired data in the final image, while simulated projection data serves as an effective supplement. This strategy avoids the uncertainty that may arise from over-reliance on predicted data, thereby guaranteeing the accuracy and reliability of the final image. By fusing the original reconstructed image with the completed image, this application can fully utilize the accurate information from high-confidence regions in the original image while effectively filling in truncated regions, ultimately obtaining a high-quality, artifact-free image.

[0101] In summary, this application provides a comprehensive and efficient scheme for filling truncation artifacts in SPECT pinhole collimator imaging by combining techniques such as projected pixel mapping, deep learning prediction, and weighted reconstruction. This scheme not only significantly improves image quality and provides more accurate evidence for clinical diagnosis, but also demonstrates significant progress in its technical approach, opening up new avenues for technological development in the field of SPECT imaging.

[0102] In some of the embodiments described above in this application, the original reconstructed image is obtained by reconstructing the real projection data. However, in the actual SPECT imaging process, gamma photons are attenuated and scattered when passing through human tissue. If these physical effects are not corrected, artifacts and noise will appear in the reconstructed image, thereby affecting the image quality and the accuracy of subsequent processing.

[0103] In some embodiments, the step of reconstructing the original reconstructed image from the real projection data includes:

[0104] The initial image is obtained by reconstructing the actual projection data;

[0105] The initial image is preprocessed to obtain the original reconstructed image; the preprocessing includes attenuation correction and scattering correction.

[0106] Specifically, reconstructing the true projection data to obtain the initial image involves using traditional SPECT image reconstruction algorithms, such as iterative reconstruction algorithms (e.g., OSEM, MLEM) or analytical reconstruction algorithms (e.g., FBP), to convert the acquired true projection data into an initial image in three-dimensional space. This initial image may still contain artifacts and noise caused by the attenuation and scattering of gamma photons in tissue. Preprocessing the initial image to obtain the original reconstructed image aims to optimize it, improving its quality and accuracy. This preprocessing specifically includes attenuation correction and scattering correction. Attenuation correction compensates for the attenuation effect of gamma photons passing through human tissues of different densities and compositions, ensuring that the distribution of radioactive tracers in the image more accurately reflects their true concentration. Scattering correction eliminates or reduces image blurring and contrast reduction caused by Compton scattering of gamma photons in tissue, thereby improving image clarity and quantitative accuracy. Through these preprocessing steps, noise and artifacts in the image can be significantly reduced, making the original reconstructed image closer to the true radioactive distribution.

[0107] The proposed solution effectively addresses the artifacts and noise issues present in traditional reconstructed images by introducing preprocessing steps—attenuation correction and scattering correction—after obtaining the initial image. Specifically, attenuation correction compensates for the energy loss of gamma photons in the tissue, enabling the signal received by the detector to more accurately reflect the true distribution of the tracer and avoiding image intensity distortion caused by attenuation. Simultaneously, scattering correction identifies and removes the contribution of scattered photons to the detector signal, eliminating image blurring and contrast reduction caused by scattering, thereby improving image clarity and quantitative accuracy. It is precisely the introduction of these correction steps that significantly improves the quality of the original reconstructed image, providing more reliable basic data for subsequent truncation artifact filling.

[0108] By employing the aforementioned technical solution, image artifacts and noise caused by gamma photon attenuation and scattering effects can be effectively eliminated or significantly reduced during the reconstruction of the original reconstructed image from real projection data. This not only improves the quality and accuracy of the original reconstructed image, making it more realistically reflect the distribution of radioactive tracers in the body, but also provides cleaner and more reliable input data for subsequent steps such as high-confidence region determination, projection pixel mapping, and deep learning prediction, thereby ultimately improving the overall accuracy and reliability of SPECT pinhole collimator imaging truncation artifact filling.

[0109] In some embodiments, the step of establishing a projected pixel mapping relationship between the actual detector radial distance and the ideal detector radial distance based on the original reconstructed image includes:

[0110] The first projected image is obtained by projecting the original reconstructed image according to the radial distance of the real detector;

[0111] Based on the original reconstructed image, calculate the radial distance of an ideal detector capable of completely scanning the entire human body surface;

[0112] A second projected image is obtained by projecting the original reconstructed image along the radial distance of an ideal detector;

[0113] Based on the first and second projected images, the projected pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector is obtained.

[0114] Specifically, in the acquisition of the first projected image, the original reconstructed image is treated as a three-dimensional voxel distribution, which is projected according to the radial distance of the real detector. The radial distance of the real detector refers to the distance from each point on the detector surface to its rotation center or imaging center axis when the actual projection data is acquired. Through this projection, the projection intensity distribution of the original reconstructed image at various radial positions under the geometry of the real detector can be simulated, thereby obtaining the first projected image.

[0115] The calculation of the radial distance of an ideal detector capable of completely scanning the entire human body surface aims to determine the size and location of a detector that could theoretically completely cover the imaged object (e.g., the human body surface). The radial distance of the ideal detector can be understood as the radial range corresponding to a detector capable of completely capturing all gamma photon information without truncation artifacts. This distance can be calculated based on the size of the original reconstructed image, the estimated size of the imaged object, and the geometric parameters of the imaging system.

[0116] Furthermore, during the acquisition of the second projected image, the original reconstructed image is projected according to the radial distance of the ideal detector calculated above. This simulates the projection intensity distribution of the original reconstructed image under the ideal detector geometry, thus obtaining the second projected image.

[0117] Therefore, based on the first and second projected images, a projection pixel mapping relationship can be established between the radial distance of the real detector and the radial distance of the ideal detector. This mapping relationship is essentially a geometric transformation or interpolation relationship, describing how a projected pixel at a certain radial position on the real detector corresponds to the corresponding radial position on the ideal detector. For example, this mapping can be established by comparing the feature points, edges, or overall intensity distribution of the two projected images, using image registration, interpolation algorithms, or geometric transformation models. Its purpose is to provide a foundation for the subsequent accurate conversion of the real projection data to the ideal detector space.

[0118] The proposed method first projects the original reconstructed image onto both the radial distance of the real detector and the radial distance of the ideal detector, resulting in two projected images under different geometric conditions. By comparing and analyzing these two projected images, the correspondence between the radial distances of the real and ideal detectors can be accurately established. This method avoids errors that may arise from directly estimating or assuming a mapping relationship. Instead, it provides a data-driven and geometrically consistent mapping establishment method based on the reconstructed image information and by simulating projections under different detector geometries.

[0119] The above technical solution enables a more precise and systematic establishment of the projected pixel mapping relationship between the actual detector radial distance and the ideal detector radial distance. This significantly improves the accuracy and reliability of the mapping, providing a solid foundation for subsequently mapping the actual projection data onto the projected image of the ideal detector radial distance. This effectively reduces artifacts or information distortion caused by inaccurate mapping, thereby enhancing the quality and diagnostic value of the final image.

[0120] In some embodiments, the step of obtaining the final image by fusing the original reconstructed image and the completed image based on the high-confidence region includes:

[0121] Based on the high-confidence regions, the original reconstructed image and the completed image are fused using a mask interpolation method to obtain the final image.

[0122] Specifically, the mask interpolation method uses a mask to control the weight distribution between the original reconstructed image and the completed image during the fusion process, and achieves a smooth transition through an interpolation algorithm. The mask can be generated based on high-confidence regions; for example, a high-confidence region can be defined as the part of the mask with higher weight, while the region to be completed can be defined as the part with lower weight. The interpolation algorithm is then used to blend the pixel values ​​of the original reconstructed image and the completed image according to the weight ratio within the transition region defined by the mask, thereby avoiding abrupt changes and artifacts that may occur from direct stitching.

[0123] The proposed solution employs a mask interpolation method for image fusion, effectively addressing potential image quality degradation, discontinuities, or artifacts in the fusion region between the original reconstructed and completed images. Specifically, a high-confidence region is used to generate a smooth mask. This mask has higher weights in areas of high confidence in the original reconstructed image and lower weights in areas of significant contribution from the completed image, or vice versa. Through an interpolation algorithm, pixel values ​​from the original reconstructed and completed images are smoothly blended within the transition region defined by the mask, avoiding boundary effects caused by hard stitching. This gradual fusion method ensures a natural transition between different regions in the final image, thereby improving overall image consistency and visual quality.

[0124] By employing the above-described technical solution and utilizing mask interpolation for image fusion, the quality and reliability of the final image can be significantly improved. This method avoids boundary artifacts and discontinuities that may result from simple stitching or averaging fusion, ensuring that high-confidence regions of the original reconstructed image retain their original accuracy. Simultaneously, smooth transitions are achieved in truncated regions through image completion, resulting in a more visually natural and coherent final image and improving the diagnostic accuracy of lesion areas.

[0125] Please refer to Figure 2 , Figure 2 This invention provides a SPECT pinhole collimator imaging truncation artifact filling device in some embodiments. The SPECT pinhole collimator imaging truncation artifact filling device is integrated into a back-end control device in the form of a computer program, comprising:

[0126] Acquisition module 100 is used to acquire real projection data;

[0127] The reconstruction module 200 is used to reconstruct the real projection data to obtain the original reconstructed image;

[0128] The determination module 300 is used to determine the high-confidence region in the original reconstructed image based on the collimator configuration and collimator position when acquiring real projection data;

[0129] Module 400 is established to create a projected pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector based on the original reconstructed image.

[0130] The mapping module 500 is used to map real projection data onto the projection image of the radial distance of the ideal detector using the projection pixel mapping relationship, so as to obtain mapped projection data.

[0131] The acquisition module 600 is used to acquire the projection image of the attenuation image on the radial distance of the ideal detector and use it as prior projection data.

[0132] The prediction module 700 is used to predict projection data outside the effective area of ​​the real acquisition detector and use it as simulated projection data by inputting the mapped projection data and prior projection data into the deep learning network.

[0133] The completion module 800 is used to obtain a completed image containing the truncated region by performing weighted reconstruction on the simulated projection data based on a preset confidence allocation rule.

[0134] The fusion module 900 is used to obtain the final image by fusing the original reconstructed image and the completed image based on the high-confidence region.

[0135] In some embodiments, the reconstruction module 200 performs the following when reconstructing the real projection data to obtain the original reconstructed image:

[0136] The initial image is obtained by reconstructing the actual projection data;

[0137] The initial image is preprocessed to obtain the original reconstructed image.

[0138] In some embodiments, the establishment module 400 is performed when establishing a projected pixel mapping relationship between the true detector radial distance and the ideal detector radial distance based on the original reconstructed image:

[0139] The first projected image is obtained by projecting the original reconstructed image according to the radial distance of the real detector;

[0140] Based on the original reconstructed image, calculate the radial distance of an ideal detector capable of completely scanning the entire human body surface;

[0141] A second projected image is obtained by projecting the original reconstructed image along the radial distance of an ideal detector;

[0142] Based on the first and second projected images, the projected pixel mapping relationship between the radial distance of the real detector and the radial distance of the ideal detector is obtained.

[0143] In some embodiments, the fusion module 900 performs the following when obtaining the final image by fusing the original reconstructed image and the completed image based on high-confidence regions:

[0144] Based on the high-confidence regions, the original reconstructed image and the completed image are fused using a mask interpolation method to obtain the final image.

[0145] Please refer to Figure 3 , Figure 3This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. The present invention provides an electronic device 13, including: a processor 1301 and a memory 1302. The processor 1301 and the memory 1302 are interconnected and communicate with each other via a communication bus 1303 and / or other forms of connection mechanism (not shown). The memory 1302 stores computer-readable instructions executable by the processor 1301. When the electronic device is running, the processor 1301 executes the computer-readable instructions to perform the SPECT pinhole collimator imaging truncation artifact filling method in any optional implementation of the above embodiments, to achieve the following functions: acquiring real projection data; reconstructing the real projection data to obtain the original reconstructed image; and according to the collimator configuration and collimator position when acquiring the real projection data... The process involves: identifying high-confidence regions in the original reconstructed image; establishing a projection pixel mapping relationship between the real detector radial distance and the ideal detector radial distance based on the original reconstructed image; mapping the real projection data onto the projection image of the ideal detector radial distance using the projection pixel mapping relationship to obtain mapped projection data; acquiring the projection image of the attenuated image on the ideal detector radial distance and using it as prior projection data; inputting the mapped projection data and prior projection data into a deep learning network to predict projection data outside the effective area of ​​the real acquisition detector and using it as simulated projection data; based on a preset confidence allocation rule, performing weighted reconstruction on the simulated projection data to obtain a completed image containing truncated regions; and finally, fusing the original reconstructed image and the completed image according to the high-confidence regions to obtain the final image.

[0146] This invention provides a computer-readable storage medium storing a computer program. When the computer program is executed by a processor, it performs the SPECT pinhole collimator imaging truncation artifact filling method in any optional implementation of the above embodiments to achieve the following functions: acquiring real projection data; reconstructing the real projection data to obtain an original reconstructed image; determining a high-confidence region in the original reconstructed image based on the collimator configuration and collimator position when acquiring the real projection data; establishing a projection pixel mapping relationship between the real detector radial distance and the ideal detector radial distance based on the original reconstructed image; mapping the real projection data onto the projection image of the ideal detector radial distance using the projection pixel mapping relationship to obtain mapped projection data; acquiring the projection image of the attenuated image on the ideal detector radial distance and using it as prior projection data; predicting projection data outside the effective area of ​​the real acquisition detector by inputting the mapped projection data and prior projection data into a deep learning network and using it as simulated projection data; obtaining a completed image containing the truncation region by weighted reconstruction of the simulated projection data based on a preset confidence allocation rule; and obtaining the final image by fusing the original reconstructed image and the completed image based on the high-confidence region.

[0147] The computer-readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0148] In the embodiments provided by this invention, it should be understood that the disclosed apparatus and method can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the displayed or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0149] Furthermore, the units described 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.

[0150] Furthermore, the functional modules in the various embodiments of the present invention can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.

[0151] In this document, relational terms such as first and second are used only to distinguish one entity or operation from another entity or operation, without necessarily requiring or implying any such actual relationship or order between these entities or operations.

[0152] The above description is merely an embodiment of the present invention and is not intended to limit the scope of protection of the present invention. For those skilled in the art, the present invention can have various modifications and variations. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method of SPECT pinhole collimator imaging truncation artifact filling, comprising: The method comprises the following steps: acquiring real projection data; reconstructing the real projection data to obtain an original reconstructed image; determining a high-confidence region in the original reconstructed image according to a collimator configuration and a collimator position when the real projection data is acquired; establishing a projection pixel mapping relationship between a real detector radial distance and an ideal detector radial distance based on the original reconstructed image; mapping the real projection data to a projection image of the ideal detector radial distance by using the projection pixel mapping relationship to obtain mapped projection data; obtaining a projection image of an attenuation image on the ideal detector radial distance and taking the projection image as prior projection data; predicting projection data outside a real acquisition detector effective region as simulated projection data by inputting the mapped projection data and the prior projection data into a deep learning network; obtaining a completed image containing a truncated region by weighted reconstruction of the simulated projection data based on a preset confidence distribution rule; obtaining a final image by fusing the original reconstructed image and the completed image according to the high-confidence region.

2. The SPECT pinhole collimator imaging truncation artifact filling method of claim 1, wherein, The step of reconstructing the real projection data to obtain the original reconstructed image comprises: reconstructing the real projection data to obtain an initial image; preprocessing the initial image to obtain the original reconstructed image.

3. The SPECT pinhole collimator imaging truncation artifact filling method of claim 2, wherein, The preprocessing comprises attenuation correction and scatter correction.

4. The SPECT pinhole collimator imaging truncation artifact filling method of claim 1, wherein, The step of establishing the projection pixel mapping relationship between the real detector radial distance and the ideal detector radial distance based on the original reconstructed image comprises: projecting the original reconstructed image according to the real detector radial distance to obtain a first projection image; calculating a radial distance of the ideal detector capable of completely scanning the surface of the whole human body according to the original reconstructed image; projecting the original reconstructed image according to the radial distance of the ideal detector to obtain a second projection image; obtaining the projection pixel mapping relationship between the real detector radial distance and the ideal detector radial distance based on the first projection image and the second projection image.

5. The SPECT pinhole collimator imaging truncation artifact filling method of claim 1, wherein, The attenuation image is an attenuation coefficient distribution map generated by using CT scanning data.

6. The SPECT pinhole collimator imaging truncation artifact filling method of claim 1, wherein, The confidence distribution rule comprises that the mapped projection data is given a higher weight than the simulated projection data.

7. The SPECT pinhole collimator imaging truncation artifact filling method of claim 1, wherein, The step of obtaining the final image by fusing the original reconstructed image and the completed image according to the high-confidence region comprises: obtaining the final image by fusing the original reconstructed image and the completed image by a mask interpolation method according to the high-confidence region.

8. A SPECT pinhole collimator imaging truncation artifact filling device, characterized by, The method comprises: an acquisition module, configured to acquire real projection data; a reconstruction module, configured to reconstruct the real projection data to obtain an original reconstructed image; a determination module, configured to determine a high-confidence region in the original reconstructed image according to a collimator configuration and a collimator position when the real projection data is acquired; an establishment module, configured to establish a projection pixel mapping relationship between a real detector radial distance and an ideal detector radial distance based on the original reconstructed image; a mapping module, configured to map the real projection data to a projection image of the ideal detector radial distance by using the projection pixel mapping relationship to obtain mapped projection data; an acquisition module, configured to obtain a projection image of an attenuation image on the ideal detector radial distance and take the projection image as prior projection data; The prediction module is configured to predict projection data outside the effective area of the real acquisition detector as simulated projection data by inputting the mapping projection data and the prior projection data into a deep learning network. The completion module is configured to obtain a completed image containing a truncated area by weighting reconstruction on the simulated projection data based on a preset confidence distribution rule. The fusion module is configured to obtain a final image by fusing the original reconstructed image and the completed image according to a high-confidence area.

9. An electronic device, comprising: The computer program is executed by the processor to run the steps of the SPECT pinhole collimator imaging truncated artifact filling method according to any one of claims 1-7.

10. A computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program is executed by the processor to run the steps of the SPECT pinhole collimator imaging truncated artifact filling method according to any one of claims 1-7.

Citation Information

Cited By

  • Image completion method and device, computer equipment and storage medium

    CN121235948A

  • Scanning image reconstruction method adaptive to incomplete data and scanning imaging system

    CN121746250A