Scatter Correction for Computed Tomography Imaging

Abstract

The present invention is titled "Scatter Correction for Computed Tomography Imaging". The present invention provides a system and method for scatter correction of X-ray images. A partially scatter-free image acquired using a pore plate can be used to correct the scatter image of an object. The plate is positioned between the object and the radiation detector and includes grid-like pores. The original X-rays pass through these pores and the scattered X-rays can be blocked by the pore plate. The pore plate can be moved to different positions, thereby allowing the acquisition of partially scatter-free images at each position of the pore plate. A completely scatter-free image can be generated by combining the partially scatter-free images. These scatter images and scatter-free images can also be used to train a scatter correction model.

Description

[0001] Cross - reference to related applications

[0002] This application claims the benefit of U.S. Provisional Patent Application No. 63 / 090,144, filed on October 9, 2020, and entitled "Scatter Correction For Computed Tomography Imaging", the entire disclosure of which is hereby incorporated by reference. Background Art

[0003] Inspection of objects is commonly performed in the manufacturing and repair industries. Various types of inspection systems can be used during industrial inspection, such as computed tomography (CT), coordinate measuring machines (CMMs), laser profilometry, photometers, infrared, etc. For example, these inspection systems can be used to measure dimensions or identify defects in manufactured parts (e.g., turbine blades).

[0004] Each of these inspection systems has its advantages and disadvantages. Methods such as CMMs and laser profilometry can be used to measure the outer surface with high precision, but these techniques cannot measure internal features unless the object is cut open. To date, CT is the most versatile measurement / inspection system for non - destructively showing both the internal and external structures of industrial parts. Due to its ability to provide both internal and external measurements, CT - based techniques can be beneficial for processes such as reverse engineering, rapid prototyping, casting simulation and verification, tire development, first article inspection, ceramic porosity inspection, process verification, part identification, and defect detection. Summary of the Invention

[0005] However, CT - based techniques can also have certain limitations that can hinder their widespread application. For example, volume computed tomography (VCT) imaging for industrial applications (e.g., imaging of metal parts) can provide unsatisfactory images with image artifacts, which are attributed to artifacts based on radiation - matter interactions, scanner - based artifacts, reconstruction - technique - based artifacts, etc. Artifacts based on radiation - matter interactions can also include beam hardening artifacts and artifacts attributed to X - ray scatter radiation. Scatter radiation is a strong function of imaging parameters such as the object being imaged, the beam spectrum used, the geometric distance, and the surrounding medium. Generally speaking, scatter radiation in projection images is undesirable because it can reduce the projection image contrast, cause degradation or blurring of the sharp features of the object in the generated volume image, and reduce the accuracy of metrology applications and the detectability of smaller features.

[0006] Accordingly, various techniques have been developed to estimate scatter in order to reduce or eliminate it from CT images. Generally speaking, due to the various dependencies of imaging parameters, accurately estimating the scatter signal content in projection imaging can be challenging. Physics-based models are typically used to predict the scatter content in X-ray images. However, these models are time-consuming and only predict the scatter caused by the object being scanned on the premise that the material properties are known.

[0007] There are different techniques for scatter measurement and scatter correction in the acquired projection images. For example, a commonly used scatter measurement technique employs a beam blocker located between the radiation source and the object being scanned in a VCT system to measure the scatter at the corresponding position. However, most currently known techniques mainly address object scatter and involve time-consuming computer simulations.

[0008] As manufacturing tolerances become more stringent, the demand for metrology techniques for maintaining tolerances increases accordingly. The need for quality and performance testing has become an integral part of the production or manufacturing process. Therefore, in order to improve the accuracy and efficiency of CT examinations, more effective methods are needed to remove artifacts related to scattered radiation.

[0009] Embodiments of the present disclosure provide improved systems and methods for scatter correction. Additional images acquired using an anti-scatter pore plate can be used to correct the scatter image of an object. The anti-scatter pore plate may include a plurality of holes positioned on a grid. In use, the pore plate is positioned between the object being imaged and the detector. When collimated X-rays are directed from the X-ray source to the object, the original X-rays can pass through the holes, while a significant portion of the scattered X-rays can be blocked by the pore plate. The pore plate can also be configured to move between different positions. Accordingly, partially scatter-free images, herein referred to as partial scatter-free images, can be acquired at each position of the pore plate. A scatter-free image (ground truth) can be obtained by combining the partial scatter-free images. Depending on the geometry of the pore plate, the number of positions and the acquired partial scatter-free images can be varied. Generally speaking, the number of positions can be sufficient such that the holes cover the entire region of the scatter image to be scatter-corrected.

[0010] The above embodiments rely on using a pore plate with discrete holes (also referred to as beam holes) to acquire partial scatter-free images. In an alternative embodiment, a slit collimator or a beam blocker can be employed instead of the pore plate to acquire partial scatter-free images. Similar to the pore plate method, the slit collimator or the beam blocker can be positioned between the radiation source and the detector and moved. Each partial scatter-free image can be acquired at different positions of the slit collimator or the beam blocker, and the partial scatter-free images can be combined to form a scatter-free image.

[0011] Scatter-free images acquired by using any one of a pore plate, a slit collimator, or a beam blocker can also be used to facilitate scatter correction. In one aspect, the scatter-free image can be used as a reference for training a neural network to determine an algorithm for scatter correction. In another aspect, the scatter-free image can be used to determine the point spread function (PSF) of a scatter deconvolution algorithm.

[0012] In additional embodiments, a partial scatter-free image or a scatter-free image can be used to adjust a trained neural network or to adjust the parameters of a known PSF for convolution-based scatter correction.

[0013] In one embodiment, a method for scatter correction of an image of an object is provided. The method can include acquiring data representing at least one scatter image of the object by a radiation detector of an imaging system based on radiation detected passing through an imaging volume of the object. The method can also include placing a pore plate at a first position between the object and the radiation detector. The pore plate can include a plurality of holes configured to prevent scattered radiation from being detected by the radiation detector. The method can also include, when the pore plate is in the first position, acquiring data representing at least one first partial scatter-free image by the radiation detector based on radiation detected passing through the imaging volume of the object. The method can also include moving the pore plate to one or more second positions different from the first position. The holes of the first position and the one or more second positions can cover the region of the object captured in the at least one scatter image. The method can also include, when the pore plate is in the one or more second positions, acquiring data representing at least one second partial scatter-free image by the radiation detector based on radiation detected passing through the imaging volume of the object. The method can also include receiving, by an analyzer including one or more processors, the at least one first partial scatter-free image data and the second partial scatter-free image data. The method can also include generating, by the analyzer, at least one scatter-free image based on a combination of at least a portion of the at least one first partial scatter-free image data and at least a portion of the at least one second partial scatter-free image data. The method can also include updating, by the analyzer, a scatter correction model using at least a portion of the at least one scatter image and the at least one scatter-free image. The method can also include outputting, by the analyzer, the updated scatter correction model.

[0014] In another embodiment, the scatter correction model can be a neural network model including a scatter correction algorithm. The method may further include operations performed by the analyzer including updating the scatter correction model, which is performed by operations including: applying the scatter correction algorithm to at least a portion of the at least one scatter image to generate a scatter corrected image; determining a deviation between the generated scatter corrected image and the at least one scatter-free image. These operations may further include updating the scatter correction algorithm to reduce the deviation when the deviation is greater than a predetermined deviation. These operations may further include outputting the updated scatter correction model including the updated scatter correction algorithm when the deviation is less than or equal to the predetermined deviation.

[0015] In another embodiment, the scatter correction model can be a deconvolution including a scatter edge spread function (ESF). The scatter edge spread function may further include a point spread function (PSF) configured to correct scatter within the scatter image. The method may further include updating the scatter correction model. The scatter correction model can be updated by operations including: applying the PSF to at least a portion of the at least one scatter image to generate a scatter corrected image; and determining a deviation between the scatter corrected image and the at least one scatter-free image. These operations may further include updating the parameters of the PSF to reduce the deviation and repeating the application and determination operations when the deviation is above a predetermined deviation. These operations may further include outputting the updated scatter correction model including the updated PSF parameters when the deviation is below the predetermined deviation.

[0016] In another embodiment, the detected radiation can be collimated.

[0017] In another embodiment, the at least one scatter-free image is not generated by interpolating the at least one first partial scatter-free image data or the at least one second partial scatter-free image data.

[0018] In another embodiment, the movable well plate may include at least one of one-way translation, two-way translation, or rotation.

[0019] In another embodiment, the at least one scatter image can be a set of scatter images including multiple images.

[0020] In another embodiment, the wells can be arranged in a two-dimensional grid.

[0021] In another embodiment, the wells in the first position and the at least one second position may overlap each other.

[0022] In another embodiment, the wells in the first position and the at least one second position may be spaced apart from each other by a predetermined distance.

[0023] In one embodiment, an imaging system is provided. The imaging system may include a radiation source, a movable aperture plate, and an analyzer. The radiation source may be configured to emit radiation toward an object. A radiation detector may be configured to detect the emitted radiation that has passed through an imaging volume of the object. The movable aperture plate may be positioned between the object and the radiation detector and may also include a plurality of holes configured to prevent scattered radiation from being detected by the radiation detector. The analyzer may include one or more processors. The analyzer may also be configured to receive data representing at least one scattered image of the object based on the detected radiation that has passed through the imaging volume of the object. The analyzer may also be configured to, when the aperture plate is in a first position, receive data representing at least one non-scattered image of at least a first portion of the object based on the detected radiation that has passed through the imaging volume of the object. The analyzer may also be configured to, when the aperture plate is in one or more second positions different from the first position, receive data representing at least one non-scattered image of at least a second portion of the object based on the detected radiation that has passed through the imaging volume of the object. The holes in the first position and the one or more second positions may cover regions of the object that are captured within the at least one scattered image. The analyzer may also be configured to generate at least one non-scattered image based on a combination of at least a portion of the at least one non-scattered image data of the at least a first portion and at least a portion of the at least one non-scattered image data of the at least a second portion. The analyzer may also be configured to train a scatter correction model using at least a portion of the at least one scattered image and the at least one non-scattered image. The analyzer may also be configured to output the trained scatter correction model.

[0024] In another embodiment, the scatter correction model may be a neural network model including a scatter correction algorithm. The analyzer may also be configured to train the scatter correction algorithm by: applying the scatter correction algorithm to at least a portion of the at least one scattered image to generate a scatter correction image; and determining a deviation between the generated scatter correction image and the at least one non-scattered image. When the deviation is greater than a predetermined deviation, the analyzer may also update the scatter correction algorithm to reduce the deviation. When the deviation is less than or equal to the predetermined deviation, the analyzer may also output the updated scatter correction algorithm.

[0025] In another embodiment, the scatter correction model may be a deconvolution including a scatter edge spread function (ESF). The scatter edge spread function may also include a point spread function (PSF) configured to correct scatter within the scatter image. The analyzer may also be configured to train the PSF by: applying the PSF to at least a portion of the at least one scattered image to generate a scatter correction image; and determining a deviation between the scatter correction image and the at least one non-scattered image. When the deviation is above a predetermined deviation, the analyzer may also update the parameters of the PSF to reduce the deviation and repeat the applying and determining operations. When the deviation is below the predetermined deviation, the analyzer may output the PSF parameters.

[0026] In another embodiment, the system may further include a collimator configured to collimate the emitted radiation.

[0027] In another embodiment, the analyzer is not configured to generate the at least one scatter-free image by interpolating the at least one first portion of scatter-free image data or the at least one second portion of scatter-free image data.

[0028] In another embodiment, the first position and the at least one second position may differ by at least one of unidirectional translation, bidirectional translation, or rotation.

[0029] In another embodiment, the at least one scatter image may be a set of scatter images including a plurality of images.

[0030] In another embodiment, the holes may be arranged in a two-dimensional grid.

[0031] In another embodiment, the holes at the first position of the pore plate and the at least one second position of the pore plate may overlap each other.

[0032] In another embodiment, the holes at the first position of the pore plate and the at least one second position of the pore plate may be spaced apart from each other by a predetermined distance.

[0033] In one embodiment, a method for scatter correction of an image of an object is provided. The method may include acquiring data representing a plurality of scatter images of the object by a radiation detector of an imaging system based on radiation detected as transmitted through an imaging volume of the object. The method may further include placing a pore plate between the object and the radiation detector. The pore plate may include a plurality of holes configured to prevent scattered radiation from being detected by the radiation detector. The method may further include acquiring, by the radiation detector, data representing a partial scatter-free image corresponding to each scatter image. Each partial scatter-free image may be based on radiation detected as transmitted through the imaging volume of the object when the pore plate is present. The scatter image and its corresponding partial scatter-free image may be acquired under substantially the same conditions, except for the presence or absence of the pore plate. The method may further include receiving, by an analyzer including one or more processors, the plurality of scatter image data and the corresponding partial scatter-free image data. The method may further include receiving, by the analyzer, a trained scatter correction model. The method may further include updating, by the analyzer, the trained scatter correction model based on the received plurality of scatter image data and the corresponding partial scatter-free image data to produce an updated trained scatter correction model; correcting at least one of the plurality of scatter images based on the updated trained scatter correction model; and outputting at least one corrected scatter image.

[0034] In one embodiment, the trained scatter correction model can be a trained neural network model.

[0035] In another embodiment, updating the trained scatter correction model can include: performing interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the partial scatter-free image for each pair of corresponding scatter images and partial scatter-free images; and generating an updated trained scatter correction model based on the interpolation.

[0036] In another embodiment, the trained scatter correction model can include a previously determined deconvolution point spread function (PSF) estimate.

[0037] In another embodiment, updating the trained scatter correction model can include locally parameterizing the deconvolution PSF estimate using the measurement points of the corresponding paired scatter images and partial scatter-free images; and updating the deconvolution PSF estimate based on the local parameterization.

[0038] In another embodiment, an imaging system is provided. The imaging system can include a radiation source, a radiation detector, and an analyzer. The radiation source can be configured to emit radiation toward an object. The radiation detector can be configured to detect the emitted radiation that has passed through the imaging volume of the object. The analyzer can include one or more processors. The analyzer can also be configured to receive data representing a plurality of scatter images of the object based on the detected radiation that has passed through the imaging volume of the object by the radiation source. The analyzer can also be configured to receive data representing a partial scatter-free image corresponding to each scatter image. Each partial scatter-free image can be based on the detected radiation that has passed through the imaging volume of the object when a pore plate is present. The scatter image and its corresponding partial scatter-free image can be acquired under substantially the same conditions, except for the presence or absence of the pore plate. The analyzer can also be configured to receive a trained scatter correction model, and update the trained scatter correction model based on the received plurality of scatter image data and corresponding partial scatter-free image data to produce an updated trained scatter correction model. The analyzer can also be configured to correct at least one of the plurality of scatter images based on the updated trained scatter correction model, and output at least one corrected scatter image.

[0039] In another embodiment, the trained scatter correction model can be a trained neural network model.

[0040] In another embodiment, updating the trained scatter correction model may include: performing interpolation by the analyzer between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the partial scatter-free image for each pair of corresponding scatter images and partial scatter-free images; and generating an updated trained scatter correction model based on the interpolation.

[0041] In another embodiment, the trained scatter correction model may include a previously determined deconvolution point spread function (PSF) estimate.

[0042] In another embodiment, updating the trained scatter correction model may include locally parameterizing the deconvolution PSF estimate using the measurement points of the corresponding paired scatter images and partial scatter-free images; and updating the deconvolution PSF estimate based on the local parameterization.

[0043] In another embodiment, a method for scatter correction of an image of an object is provided. The method may include acquiring data representing a plurality of scatter images of the object by a radiation detector of an imaging system based on radiation detected passing through an imaging volume of the object. The method may further include placing a pore plate between the object and the radiation detector. The pore plate may include a plurality of holes configured to prevent scattered radiation from being detected by the radiation detector. The method may further include acquiring data representing a single partial scatter-free image by the radiation detector. The single partial scatter-free image may be based on radiation detected passing through the imaging volume of the object when the pore plate is present. The method may further include receiving, by an analyzer including one or more processors, the plurality of scatter image data and the single partial scatter-free image data. The method may further include receiving, by the analyzer, a trained scatter correction model. The method may further include updating, by the analyzer, the trained scatter correction model based on the received plurality of scatter image data and the single partial scatter-free image data to produce an updated trained scatter correction model. The method may further include correcting, by the analyzer, at least one of the plurality of scatter images based on the updated trained scatter correction model, and outputting, by the analyzer, at least one corrected scatter image.

[0044] In another embodiment, the trained scatter correction model may be a trained neural network model.

[0045] In another embodiment, updating the trained scatter correction model may include: performing interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the single partial scatter-free image for each image pair including the scatter image of the plurality of scatter images and the single partial scatter-free image; and generating an updated trained scatter correction model based on the interpolation.

[0046] In another embodiment, the trained scatter correction model may include a previously determined deconvolved point spread function (PSF) estimate.

[0047] In another embodiment, updating the trained scatter correction model may include locally parameterizing the deconvolved PSF estimate using measurement points of corresponding image pairs including a scatter image of the plurality of scatter images and the single partial scatter-free image; and updating the deconvolved PSF estimate based on the local parameterization.

[0048] In another embodiment, an imaging system is provided and the imaging system may include a radiation source, a radiation detector, and an analyzer. The radiation source may be configured to emit radiation toward an object. The radiation detector may be configured to detect the emitted radiation that has passed through an imaging volume of the object. The analyzer may include one or more processors. The analyzer may also be configured to receive data representing a plurality of scatter images of the object based on detecting the radiation emitted by the radiation source that has passed through the imaging volume of the object. The analyzer may also be configured to receive data representing a single partial scatter-free image. The single partial scatter-free image may be based on detecting the radiation emitted by the radiation source that has passed through the imaging volume of the object when a pore plate is present. The analyzer may also be configured to receive a trained scatter correction model; update the trained scatter correction model based on the received plurality of scatter image data and the single partial scatter-free image data to produce an updated trained scatter correction model; correct at least one of the plurality of scatter images based on the updated trained scatter correction model; and output at least one corrected scatter image.

[0049] In another embodiment, the trained scatter correction model may be a trained neural network model.

[0050] In another embodiment, updating the trained scatter correction model may include: performing interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the single partial scatter-free image for each image pair including the scatter image of the plurality of scatter images and the single partial scatter-free image; and generating an updated trained scatter correction model based on the interpolation.

[0051] In another embodiment, the trained scatter correction model may include a previously determined deconvolved point spread function (PSF) estimate.

[0052] In another embodiment, updating the trained scatter correction model can include locally parametrically solving for the deconvolution PSF estimate using measurement points of corresponding image pairs including a scatter image from the plurality of scatter images and the single partial scatter-free image; and updating the deconvolution PSF estimate based on the local parameterization. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] These and other features will be more readily understood from the following detailed description in conjunction with the accompanying drawings, in which:

[0054] Figure 1A is a schematic top view showing an exemplary embodiment of a CT imaging system;

[0055] Figure 1B is a schematic side view showing Figure 1A a CT imaging system;

[0056] Figure 2A is a schematic diagram showing the positions of the holes of the pore plate when the pore plate is in a first position;

[0057] Figure 2B is a schematic diagram showing the positions of the holes of the pore plate when the pore plate is in a first position and a second position vertically translated relative to the first position;

[0058] Figure 2C is a schematic diagram showing the positions of the holes of the pore plate when the pore plate is in a first position and a third position horizontally translated relative to the first position;

[0059] Figure 2D is a schematic diagram showing the positions of the holes of the pore plate when the pore plate is in a first position, a second position, and a third position;

[0060] Figure 2E is a schematic diagram showing the positions of the holes of the pore plate when the pore plate is in a plurality of positions covering substantially the entire desired field of view;

[0061] Figure 3A is a CT image with scatter correction using a single position of the pore plate;

[0062] Figure 3B is a CT image with scatter correction using four positions of the pore plate;

[0063] Figure 4 is a schematic block diagram showing a computational technique that uses scatter images and scatter-free images as inputs for developing a scatter correction algorithm or deconvolution;

[0064] Figure 5 is a flowchart showing an exemplary embodiment of a method for scatter correction of CT images;

[0065] Figure 6 is a schematic diagram showing a large number of measurement points for adjustment given by a pore plate;

[0066] Figure 7 is a flowchart showing an exemplary embodiment of a method for scatter correction, in which a previously trained neural network or a previously determined deconvolution PSF is adjusted based on corresponding image pairs in a set of scatter images and a set of partially scatter-free images;

[0067] Figure 8 is a flowchart showing an exemplary embodiment of another method for scatter correction, in which a previously trained neural network or a previously determined deconvolution PSF is adjusted based on corresponding image pairs in a set of scatter images and a single scatter-free image;

[0068] Figure 9 is a flowchart showing an exemplary embodiment of another method for scatter correction, in which a previously trained neural network or a previously determined deconvolution PSF is adjusted based on an undefined number of sets of scatter images of different object types, where a single partially scatter-free image or a set of partially scatter-free images is employed;

[0069] Figure 10A is a schematic diagram showing a top view of an exemplary embodiment of a CT imaging system; and

[0070] Figure 10B is showing Figure 10A a schematic diagram of a side view of the CT imaging system.

[0071] It should be noted that the drawings are not necessarily drawn to scale. The drawings are only intended to depict typical aspects of the subject matter disclosed herein and should not be considered as limiting the scope of the disclosure. Detailed Description

[0072] When performing X-ray examinations, X-rays can pass through a target object and be detected by a detector to generate an image. Some X-rays can scatter from their initial trajectories, which introduces artifacts that reduce image contrast. Techniques for scatter correction have been developed, but each technique has problems. In one example, Monte Carlo methods have been developed to model scatter. However, such Monte Carlo methods may be limited to either having no experimental data (pure Monte Carlo methods) and / or may rely primarily on interpolation (supplementing limited experimental data through interpolation). Additionally, not all aspects of scatter can be simulated or interpolated with sufficient accuracy. The lack of accuracy can lead to artifacts being introduced in the scatter-corrected image. Thus, the scatter estimates produced by these Monte Carlo methods can deviate significantly from the actual scatter. Accordingly, improved systems and methods for scatter correction in X-ray examinations (e.g., computed tomography) are provided. The scatter-free image and the scatter image can be used as inputs to computational techniques such as deep learning (e.g., neural networks) or deconvolution models. The scatter-free image can be generated based on a plurality of partial scatter-free images acquired using a pore plate that includes holes. Each partial scatter-free image is acquired at a different location using the pore plate, and a portion of the scattered X-rays is blocked. Sufficient partial scatter-free images can be acquired such that the combination of the hole locations of the pore plate at different locations covers the entire scatter image. The combination of the partial scatter-free images gives the scatter-free image. The scatter-free image generated in this manner without interpolation is measured in a smaller collimation region, resulting in a high-precision scatter estimate.

[0073] Embodiments of the present disclosure generally relate to scatter correction for computed tomography (CT) imaging to achieve improved image quality. Such imaging techniques can be used in a variety of imaging environments, such as medical imaging, industrial metrology and inspection, security screening, baggage or package inspection, and the like. Additionally, such imaging techniques can be used in a variety of imaging systems, such as CT systems, tomosynthesis systems, X-ray imaging systems, and the like. Although this discussion provides examples of its implementation in an industrial inspection environment with improved measurement and inspection accuracy with respect to a CT system, those of ordinary skill in the art will readily recognize that the application of these techniques in other environments and other systems is well within the scope of the present technology.

[0074] Figures 1A to 1B An imaging system 200 configured to produce high-resolution images is shown. The imaging system 200 can be a volumetric computed tomography (VCT) system that is designed to acquire image data and process the image data for display and analysis. As shown, the imaging system 200 can include a radiation source 202, such as an X-ray source 202. A collimator 205 can be positioned adjacent to the radiation source 202 to collimate the radiation 204 emitted by the radiation source 202 and to adjust the size and shape of the emitted radiation 204 emitted by the radiation source 202.

[0075] A radiation beam 204 of 204 beams can be projected towards a detector array 206 which is placed on the opposite side of a radiation source 202 with respect to an object 208 to be imaged. The object 208 can be any object suitable for X-ray inspection (e.g., a turbine blade). The radiation beam 204 can enter an imaging volume in which the object 208 is to be imaged. A portion of the radiation 204 passes through or around the object 208 and impinges on the detector array 206. The detector array 206 can generally be formed as a two-dimensional array of detection elements. The data collected by the detector array 206 can be output to an analyzer 207 including one or more processors.

[0076] If desired, the object 208, the radiation source 202, and the detector array 206 can be shifted relative to each other, thereby allowing projection data to be acquired at various perspectives with respect to the object 208. For example, the object 208 can be positioned on a table such as a turntable so that the object 208 can rotate about a rotation axis 210. In some embodiments, the data collected from the detector array 206 can be preprocessed (e.g., by the analyzer 207) to condition the data to represent a line integral of the attenuation coefficient of the scanned object 208. The processed data or projections can then be reconstructed (e.g., by the analyzer 207 or another computing device) to formulate a volume image of the scanned region, as discussed in more detail in U.S. Patent No. 9,804,106, which is incorporated herein by reference in its entirety.

[0077] The imaging system 200 can employ various scatter mitigation and / or correction techniques to improve image quality and resolution. For example, an anti-scatter pore plate 212 can be employed to reject scatter radiation generated by the object 208 as well as scatter radiation generated by the background. To further improve in terms of resolution and image quality, the pore plate 212 can be moved between a plurality of positions, which will be discussed in more detail below. By moving the pore plate 212 between the plurality of positions, smaller structures on the object 208 can be identified and artifacts can be better avoided.

[0078] As discussed in more detail in U.S. Patent No. 9,804,106, the pore plate 212 can include a plurality of sub-centimeter-sized holes 48 drilled in the plate, such as circular holes. The holes 48 can be positioned on a two-dimensional grid. Embodiments of the holes 48 have any geometry, such as circular, rectangular, or hexagonal, etc. In some embodiments, the circular holes 48 can have a diameter of about 1 millimeter to 2 millimeters and be spaced apart from each other by about 5 millimeters (center to center).

[0079] As described above, the pore plate 212 can be moved between a plurality of positions in order to improve the resolution and quality of the generated image. As Figure 2AAs shown, the aperture plate 212 can be initially placed in the first position 211. After acquiring the first grid image, the aperture plate 212 can be repositioned to the second positions 213, 214 and the second grid image can be acquired. In one embodiment, the aperture plate 212 can move unidirectionally. For example, as Figure 2B shown, the aperture plate 212 can move vertically from the first position 211 to the second position 213 and the second image can be collected, or as Figure 2C shown, the aperture plate 212 can move horizontally from the first position 211 to the third position 214 and the second image can be collected. In another embodiment, the aperture plate 212 can move bidirectionally. For example, the aperture plate 212 can move vertically and horizontally, as Figure 2D shown. In this embodiment, the aperture plate 212 can be placed in the first position 211 and the first image can be acquired, it can be moved to the second position 213 and the second image can be acquired, it can be moved to the third position 214 and the third image can be acquired, and it can be moved to the fourth position 215 and the fourth image can be acquired. In another embodiment, the aperture plate 212 can rotate relative to the object 208. Images can be acquired at each position 211, 213, 214, 215 of the aperture plate 212. As Figure 2E shown, the process of repeatedly moving the aperture plate 212 and acquiring images can be performed until the hole positions cover the entire desired area.

[0080] In another embodiment, the resolution of the image can also be improved by moving the object 208 in front of the grid of the aperture plate 212. Similar to repositioning the aperture plate 212 described above, in this embodiment, the sample can move, for example, unidirectionally, bidirectionally, or rotationally. By repositioning the holes 48 of the aperture plate 212 relative to the object 208, the resolution of the image can be improved. The aperture plate 212 and / or the object 208 can be repositioned manually or automatically.

[0081] The resolution of the image can be determined by the number of positions where the aperture plate 212 and / or the object 208 can be placed. As the number of positions increases, the resolution of the image also improves. For example, using the bidirectional movement of the aperture plate 212, positioning the aperture plate 212 in four positions improves the image resolution by a factor of two (2). In another example, positioning the aperture plate 212 in sixteen positions improves the image resolution by a factor of four (4). Figures 3A to 3B Improving the image resolution by repositioning the aperture plate 212 is shown.

[0082] Figure 3A An image 216 of the object 208 generated with the aperture plate 212 placed in a single position is shown. Figure 3BAn image 219 of object 208 is shown, where the pore plate 212 is placed in four positions during data collection. As shown in the first position 217 and the second position 218 indicated in these figures, additional positions of the pore plate 212 result in Figure 3B images in which, compared to Figure 3A , additional details are visible at each of the first position 217 and the second position 218.

[0083] The scatter-free image can be used in combination with a computational method for scatter correction. This combination can produce better results and perform scatter correction in X-ray tomography with less scanning effort. Examples of computational methods can include deep learning (e.g., neural networks) or deconvolution methods.

[0084] As Figure 4 shown, a set of scatter images 402 and a set of scatter-free images can be input into a model 406 (e.g., a neural network or a scatter edge spread function (ESF)). In the case of a neural network, one or more algorithms for scatter correction can be received or generated and applied to the set of scatter images 402. In the case of deconvolution, using the scatter ESF, a point spread function (PSF) for deconvolution can be received or generated and applied to the set of scatter images 402. In either case, the deviation between the scatter-corrected image set and the scatter-free image set 404 is determined and used as feedback 410 for the correction algorithm. This process repeats itself, updating the algorithm or parameters at each iteration until the deviation between the scatter-corrected image set and the scatter-free image set 404 is less than a predetermined deviation. Subsequently, the determined algorithm or PSF is output.

[0085] Figure 5 is a flowchart showing an exemplary embodiment of a method 500 for scatter correction of an image of an object such as object 208 by employing advanced scatter measurement and correction techniques on an imaging system 200. As shown, method 500 includes operations 502 to 512. It should be understood that this method is merely exemplary and the selected operations can be changed, added, removed, and / or rearranged as needed.

[0086] In operation 502, at least one scatter image of object 208 can be acquired. In certain embodiments, multiple scatter images (e.g., a set of scatter images) can be acquired.

[0087] In operation 504, the pore plate 212 is placed at a first position between the object 208 and the detector array 206. As described above, the pore plate 212 may include a plurality of pores 211 and may be configured to prevent scattered radiation (e.g., X-rays scattered from the object 208 and / or the background) from being detected by the detector array 206. When the pore plate 212 is in the first position, at least one first partial scatter-free image (e.g., a single partial scatter-free image or a set of partial scatter-free images including multiple partial scatter-free images) may be acquired.

[0088] In operation 506, the pore plate 212 may be moved to one or more other positions (e.g., one or more second positions) different from the first position. For example, the movement of the pore plate 212 may include a separate translation, a separate rotation, or a combination of translation and rotation. The translation may include movement in at least one direction (e.g., a horizontal direction, a vertical direction, or a combination thereof). The pores may overlap or be spaced apart by a predetermined distance between the first position and the second position. When the pore plate 212 is positioned at a selected position in the second position (e.g., at least a portion and up to all of the second position), a second partial scatter-free image or a set of partial scatter-free images may be acquired. Operation 506 may be repeated to generate as many partial scatter-free images / image sets as desired such that the pores substantially cover the entire scatter image.

[0089] In operation 510, the acquired partial scatter-free images may be received and combined by an analyzer to generate a scatter-free image / image set. In some embodiments, at least a portion of the partial scatter-free images is employed.

[0090] In operation 512, the scatter image / image set and the scatter-free image / image set are employed for training a deep learning algorithm or for PSF estimation of a deconvolution algorithm.

[0091] It should be understood that changes to the X-ray inspection system may change the images acquired by the X-ray inspection system. Such changes may include, but are not limited to, the X-ray detector, the X-ray source, the target being inspected, and the environment. X-ray detector changes may include one or more of replacing one X-ray detector with another and changes in detection capabilities due to aging. X-ray source changes may include any change in the X-ray spectrum emitted by the X-ray source. Such changes may occur due to replacement of components of the X-ray source (e.g., filters, tubes, etc.) or replacement of the entire X-ray source with another X-ray source. Environmental changes may include any change in the X-ray scattering behavior due to the environment surrounding the X-ray source.

[0092] Due to these changes, a neural network algorithm that has been trained prior to such changes may introduce errors in the output scatter-corrected image. Similarly, a deconvolution PSF estimate determined prior to such changes may introduce errors in the output scatter-corrected image.

[0093] Thus, to address the effects of changes in the X-ray inspection system, the previously determined neural network algorithm can be retrained or the deconvolution PSF estimate can be updated. That is, it is not necessary to completely regenerate the neural network training or the deconvolution PSF estimate. Advantageously, retraining the previously determined neural network algorithm or updating the previously determined PSF estimate can require significantly less training data and time compared to generating from scratch.

[0094] As discussed in more detail below, retraining the neural network algorithm or updating the deconvolution PSF estimate can be performed in a variety of ways. A single image or an entire set of images can be employed. The adjustment can be effective for one data set and one sample type or can also be effective for many data sets with different sample types. The choice of training data depends on whether the neural network or the PSF being evaluated was created for general calibration or for a more specialized scenario.

[0095] Figure 6 FIG. is a schematic view of a perforated plate 212 showing a large number of different possible measurement points that can be used to acquire partial scatter-free images for retraining a neural network or adjusting the PSF estimate of a deconvolution algorithm.

[0096] Figure 7 FIG. is a flow chart of an exemplary embodiment of a method 700 for retraining a neural network or adjusting the PSF estimate of a deconvolution algorithm. As shown, method 700 includes operations 702 to 712. It should be understood that this method is merely exemplary and the selected operations can be changed, added, removed, and / or rearranged as needed.

[0097] In operation 702, a plurality of scatter images (e.g., a set of scatter images) of an object 208 can be acquired.

[0098] In operation 704, the perforated plate 212 is placed between the object 208 and the detector array 206, and partial scatter-free images corresponding to each of the acquired scatter images can be acquired by the radiation detector array 206. That is, the partial scatter-free images corresponding to the scatter images can be acquired under the same conditions or substantially the same conditions (e.g., within the device tolerances) as the acquisition of the scatter images, except for the presence or absence of the perforated plate 212 during the acquisition of the partial scatter-free images.

[0099] The acquired scatter image data and partial scatter-free image data can also be received by an analyzer 207. In some embodiments, the analyzer 207 can directly receive the acquired scatter image data and partial scatter-free image data from the radiation detector array 206. In other embodiments, the analyzer 207 can receive the acquired scatter image data and partial scatter-free image data from another source (e.g., a memory device).

[0100] In operation 706, the analyzer 207 may use the set of scattered images and the set of partially scatter-free images to retrain (adjust or update) a previously trained scatter correction model (e.g., a previously trained scatter correction algorithm of a neural network or a previously determined deconvolution PSF estimate). For a scatter correction model employing a neural network, retraining may include performing interpolation between the output of the neural network trained as previously for a given scattered image and the corresponding acquired partially scatter-free image. Thus, the analyzer 207 may determine an interpolation for each corresponding scattered image and partially scatter-free image pair. The neural network may use the corresponding interpolation to generate an adjusted scatter correction algorithm. For a scatter correction model employing a deconvolution method, the deconvolution function may be locally parameterized using each measurement point (e.g., each corresponding scattered image and partially scatter-free image pair), rather than generally parameterized.

[0101] In operation 710, the analyzer 207 may use the updated scatter correction model (e.g., a neural network algorithm or a deconvolution PSF estimate) to correct at least one of the plurality of scattered images.

[0102] In operation 712, the analyzer 207 may output at least one of the corrected scattered images. For example, the at least one corrected scattered image may be output to a memory device and / or a display device for viewing.

[0103] Figure 8 A flowchart of an exemplary embodiment of another method 800 that may be performed by the analyzer 207 for retraining (adjusting or updating) a neural network or a PSF estimate of a deconvolution algorithm. As shown, method 800 includes operations 802 to 812. It should be understood that this method is merely exemplary and the selected operations may be changed, added, removed, and / or rearranged as needed.

[0104] In operation 802, the radiation detector array 206 may acquire a plurality of scattered images (e.g., a set of scattered images) of the object 208.

[0105] In operation 804, the pore plate 212 is placed between the object 208 and the detector array 206, and the radiation detector array 206 may acquire a single partially scatter-free image.

[0106] The acquired scattered image data and partially scatter-free image data may also be received by the analyzer 207. In some embodiments, the analyzer 207 may directly receive the acquired scattered image data and partially scatter-free image data from the radiation detector array 206. In other embodiments, the analyzer 207 may receive the acquired scattered image data and partially scatter-free image data from another source (e.g., a memory device).

[0107] In operation 806, the multiple scattered image data and the single partial scatter-free image data can be used to retrain (e.g., adjust or update) a previously trained scatter correction model (e.g., a previously trained scatter correction algorithm of a neural network or a previously determined deconvolution PSF estimate). For a scatter correction model using a neural network, retraining can include performing interpolation between the output of a neural network previously trained for a given scattered image and the single acquired partial scatter-free image. Thus, the analyzer 207 can determine the interpolation for each pair of scattered image and single partial scatter-free image. The neural network can use the corresponding interpolation to generate an adjusted scatter correction algorithm. For a scatter correction model using a deconvolution method, the deconvolution function can be locally parameterized for each measurement point (e.g., each pair of scattered image and single partial scatter-free image), rather than generally parameterized.

[0108] In operation 810, the analyzer 207 can use the updated scatter correction model (e.g., a neural network algorithm or a deconvolution PSF estimate) to correct at least one of the multiple scattered images.

[0109] In operation 812, the analyzer 207 can output at least one of the corrected scattered images. For example, the at least one corrected scattered image can be output to a memory device and / or a display device for viewing.

[0110] In the case of acquiring a single partial scatter-free image for adjustment, the X-ray inspection system can optionally omit the ability to move the aperture plate 212, as Figures 10A to 10B shown.

[0111] Whether to use method 700 or method 800 to adjust the neural network or deconvolution ESF can be selected according to the use case and the required accuracy of scatter correction. For example, in a case where each inspected object 208 is the same, the single partial scatter-free image method of method 800 may be sufficient. The measurement can be performed periodically to adjust the neural network or deconvolution ESF to account for component aging.

[0112] Figure 9 A flowchart of an exemplary embodiment of another method 900 for retraining (adjusting) the PSF estimate of a neural network or deconvolution algorithm. As shown, method 900 includes operations 902 to 912. It should be understood that this method is merely exemplary and the selected operations can be changed, added, removed, and / or rearranged as needed.

[0113] In operation 902, a scattered image or a set of scattered images of the object 208 is acquired.

[0114] In operation 904, the pore plate 212 is placed between the object 208 and the detector array 206, and a single partial scatter-free image or a set of partial scatter-free images is acquired. As described above, when a set of partial scatter-free images is acquired, each image in the scatter-free image set can correspond to a corresponding image in the scatter image set.

[0115] In operation 906, the set of scatter images and the single partial scatter-free image or the set of scatter-free images can be used to retrain (adjust) a previously trained neural network or a previously determined deconvolution PSF estimate. The adjustment performed using the set of partial scatter-free images can be performed as discussed above in operation 706. The adjustment performed using the single partial scatter-free image can be performed as discussed above in operation 806.

[0116] In operation 910, the adjusted neural network algorithm or deconvolution PSF estimate can be used to correct the set of scatter images.

[0117] In operation 912, a set of scatter images or a set of scatter images of the same type of object 208 or a different type of object 208 can be acquired. The scatter correction can be performed using the adjusted neural network algorithm or deconvolution PSF estimate. Operation 912 can be repeated as long as the correction quality is sufficient to meet the desired result without re-adjustment.

[0118] As a non-limiting example, exemplary technical effects of the methods, systems, and devices described herein include improved scatter correction of X-ray images. A scatter-free image can be generated based on a plurality of partial scatter-free images acquired using a pore plate positioned at different locations. The scatter-free image can be combined with computational methods (e.g., deep learning, deconvolution, etc.) for scatter correction. This combination can produce better results and perform scatter correction in X-ray tomography with less scanning effort.

[0119] Certain exemplary embodiments are described to provide a comprehensive understanding of the principles of the structure, function, manufacture, and use of the systems, devices, and methods disclosed herein. One or more examples of these embodiments are shown in the drawings. Those skilled in the art will understand that the systems, devices, and methods specifically described and shown in the drawings are non-limiting exemplary embodiments, and the scope of the present invention is defined only by the claims. Features shown or described in connection with one exemplary embodiment can be combined with features of other embodiments. Such modifications and variations are intended to be included within the scope of the present invention. Additionally, in the present disclosure, components with similar names in the embodiments generally have similar features, and thus, not every feature of each similarly named component may be fully elaborated within a specific embodiment.

[0120] The subject matter described herein can be implemented in analog electronic circuits, digital electronic circuits, and / or computer software, firmware, or hardware (including the structural devices and their structural equivalents disclosed in this specification), or a combination thereof. The subject matter described herein can be implemented as one or more computer program products, tangibly embodied in an information carrier (e.g., embodied in a machine-readable storage device) or embodied in a propagated signal, for use by one or more computer programs executed by a data processing apparatus (e.g., a programmable processor, a computer, or multiple computers) or to control the operation of the data processing apparatus. A computer program (also referred to as a program, software, software application, or code) can be written in any form of programming language (including compiled or interpreted languages), and it can be deployed in any form, including as a stand-alone program or as a module, component, subroutine, or other unit suitable for use in a computing environment. A computer program does not necessarily correspond to a file. The program can be stored in a part of a file that holds other programs or data, stored in a single file dedicated to the program being considered, or stored in multiple cooperating files (e.g., files that store one or more modules, subroutines, or portions of code). A computer program can be deployed to be executed on one computer or on multiple computers, which are located at one site or distributed across multiple sites and interconnected by a communication network.

[0121] The processes and logical flows described in this specification, including the method steps of the subject matter described herein, can be executed by one or more programmable processors executing one or more computer programs to perform the functions of the subject matter described herein by operating on input data and generating output. The processes and logical flows can also be executed by dedicated logic circuitry (e.g., FPGA (Field Programmable Gate Array) or ASIC (Application Specific Integrated Circuit)), and the apparatus of the subject matter described herein can be implemented as dedicated logic circuitry (e.g., FPGA (Field Programmable Gate Array) or ASIC (Application Specific Integrated Circuit)).

[0122] By way of example, processors suitable for executing a computer program include both general and special purpose microprocessors, as well as any one or more processors of any kind of digital computer. In general, a processor will receive instructions and data from a read only memory or a random access memory or both. The essential elements of a computer are a processor for executing instructions and one or more memory devices for storing instructions and data. In general, a computer will also include one or more mass storage devices for storing data (e.g., magnetic disks, magneto-optical disks, or optical disks), or operatively coupled to receive data from and / or transfer data to one or more mass storage devices for storing data (e.g., magnetic disks, magneto-optical disks, or optical disks). Information carriers suitable for embodying computer program instructions and data include all forms of non-volatile memory, including by way of example semiconductor memory devices (e.g., EPROM, EEPROM, and flash memory devices); magnetic disks (e.g., internal hard disks or removable disks); magneto-optical disks; and optical disks (e.g., CD and DVD disks). The processor and the memory can be supplemented by, or incorporated in, special purpose logic circuitry.

[0123] For purposes of providing an interaction with a user, the subject matter described herein may be implemented on a computer having a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user and a keyboard and a pointing device (e.g., a mouse or a trackball) by which the user can provide input to the computer. Other kinds of devices can also be used to provide for interaction with the user. For example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback), and input received from the user can be in any form, including acoustic, speech, or tactile input.

[0124] The techniques described herein may be implemented using one or more modules. As used herein, the term "module" refers to computing software, firmware, hardware, and / or various combinations thereof. However, at a minimum, a module should not be construed as software that is not implemented in hardware, firmware, or recorded on a non-transitory processor-readable storage medium (i.e., a module is not software per se). In fact, a "module" will be construed to always include at least some physical non-transitory hardware, such as a processor or a portion of a computer. Two different modules may share the same physical hardware (e.g., two different modules may use the same processor and network interface). The modules described herein may be combined, integrated, separated, and / or replicated to support various applications. Additionally, instead of or in addition to the functions performed at a particular module, the functions described herein as being performed at a particular module may be performed at one or more other modules and / or by one or more other devices. Further, modules may be implemented locally or remotely relative to each other across multiple devices and / or other components. Additionally, a module may be moved from one device and added to another device, and / or may be included in both devices.

[0125] The subject matter described herein may be implemented in a computing system that includes backend components (e.g., data servers), middleware components (e.g., application servers), or frontend components (e.g., client computers having a graphical user interface or a web browser through which a user may interact with an implementation of the subject matter described herein), or any combination of such backend components, middleware components, and frontend components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include local area networks ("LANs") and wide area networks ("WANs"), such as the Internet.

[0126] As used throughout this specification and the claims, approximating language may be used to modify any quantitative representation that could vary without resulting in a change in the basic function associated therewith. Accordingly, values modified by terms such as "about," "approximately," and "substantially" are not to be limited to the exact values specified. In at least some instances, the approximating language may correspond to the precision of the instrument used to measure the value. Herein, as well as throughout the specification and the claims, range limitations may be combined and / or interchanged, and such ranges are identified and include all the subranges contained therein unless the context or language indicates otherwise.

[0127] Based on the above-described embodiments, those skilled in the art will appreciate other features and advantages of the present invention. Accordingly, except as indicated by the appended claims, this application is not limited by what has been specifically shown and described herein. All publications and references cited herein are hereby expressly incorporated by reference in their entirety.

Claims

1. A method for scatter correction of an image of an object, comprising: acquiring data representing a plurality of scatter images of the object by a radiation detector of an imaging system based on radiation detected that has transmitted through an imaging volume of the object; placing a pore plate between the object and the radiation detector, the pore plate including a plurality of pores configured to prevent scattered radiation from being detected by the radiation detector; acquiring, by the radiation detector, data representing a partial scatter-free image corresponding to each scatter image, wherein each partial scatter-free image is based on radiation detected that has transmitted through the imaging volume of the object when the pore plate is present, and wherein the scatter image and its corresponding partial scatter-free image are acquired under substantially the same conditions, except for the presence or absence of the pore plate; receiving, by an analyzer including one or more processors, the plurality of scatter image data and the corresponding partial scatter-free image data; receiving, by the analyzer, a trained scatter correction model, wherein the trained scatter correction model is a trained neural network model; updating, by the analyzer, the trained scatter correction model based on the received plurality of scatter image data and the corresponding partial scatter-free image data to produce an updated trained scatter correction model; correcting, by the analyzer, at least one of the plurality of scatter images based on the updated trained scatter correction model; and outputting, by the analyzer, at least one corrected scatter image, wherein updating the trained scatter correction model includes: performing interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the partial scatter-free image for each pair of corresponding scatter image and partial scatter-free image; and generating the updated trained scatter correction model based on the interpolation.

2. The method according to claim 1, wherein the trained scatter correction model includes a previously determined deconvolution point spread function (PSF) estimate.

3. The method according to claim 2, wherein updating the trained scatter correction model includes: locally parameterizing the deconvolution PSF estimate using measurement points of corresponding pairs of the corresponding scatter image and partial scatter-free image; and updating the deconvolution PSF estimate based on the local parameterization.

4. An imaging system, comprising: a radiation source configured to emit radiation toward an object; a radiation detector configured to detect the emitted radiation that has transmitted through an imaging volume of the object; and an analyzer including one or more processors and configured to: receive data representing a plurality of scatter images of the object based on radiation detected that has transmitted through the imaging volume of the object by the radiation source; Receiving data representing a partial scatter-free image corresponding to each scatter image, where each partial scatter-free image is based on radiation detected as transmitted through the imaging volume of the object when the pore plate is present, and where the scatter image and its corresponding partial scatter-free image are acquired under substantially the same conditions, except for the presence or absence of the pore plate; Receiving a trained scatter correction model, where the trained scatter correction model is a trained neural network model; Updating the trained scatter correction model based on the received multiple scatter image data and corresponding partial scatter-free image data to produce an updated trained scatter correction model; Correcting at least one of the multiple scatter images based on the updated trained scatter correction model; And Outputting at least one corrected scatter image, where updating the trained scatter correction model includes: Performing interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the partial scatter-free image for each pair of corresponding scatter image and partial scatter-free image; And Generating the updated trained scatter correction model based on the interpolation.

5. The imaging system according to claim 4, wherein the trained scatter correction model includes a previously determined deconvolution point spread function (PSF) estimate.

6. The imaging system according to claim 5, wherein updating the trained scatter correction model includes: Locally parameterizing the deconvolution PSF estimate using the measurement points of the corresponding paired scatter image and partial scatter-free image; And Updating the deconvolution PSF estimate based on the local parameterization.

7. A method for scatter correction of an image of an object, comprising: Acquiring, by a radiation detector of an imaging system, data representing multiple scatter images of the object based on radiation detected as transmitted through the imaging volume of the object; Placing a pore plate between the object and the radiation detector, the pore plate including a plurality of holes configured to prevent scattered radiation from being detected by the radiation detector; Acquiring, by the radiation detector, data representing a single partial scatter-free image, where the single partial scatter-free image is based on radiation detected as transmitted through the imaging volume of the object when the pore plate is present; Receiving, by an analyzer including one or more processors, the multiple scatter image data and the single partial scatter-free image data; Receiving, by the analyzer, a trained scatter correction model, where the trained scatter correction model is a trained neural network model; Updating, by the analyzer, the trained scatter correction model based on the received multiple scatter image data and the single partial scatter-free image data to produce an updated trained scatter correction model; Correcting, by the analyzer, at least one of the multiple scatter images based on the updated trained scatter correction model; And Outputting, by the analyzer, at least one corrected scatter image, Wherein updating the trained scatter correction model includes: For each image pair including a scatter image among the plurality of scatter images and the single partial scatter-free image, perform interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the single partial scatter-free image; And Generate the updated trained scatter correction model based on the interpolation.

8. The method according to claim 7, wherein the trained scatter correction model includes a previously determined deconvolution point spread function (PSF) estimate.

9. The method according to claim 8, wherein updating the trained scatter correction model includes: Locally parameterize the deconvolution PSF estimate using measurement points of corresponding image pairs including a scatter image among the plurality of scatter images and the single partial scatter-free image; and Update the deconvolution PSF estimate based on the local parameterization.

10. An imaging system, comprising: A radiation source configured to emit radiation toward an object; A radiation detector configured to detect the emitted radiation that has passed through an imaging volume of the object; And An analyzer including one or more processors and configured to: Receive data representing a plurality of scatter images of the object based on detecting the radiation transmitted through the imaging volume of the object by the radiation source; Receive data representing a single partial scatter-free image, wherein the single partial scatter-free image is based on detecting the radiation transmitted through the imaging volume of the object when a pore plate is present; Receive a trained scatter correction model, wherein the trained scatter correction model is a trained neural network model; Update the trained scatter correction model based on the received plurality of scatter image data and the single partial scatter-free image data to produce an updated trained scatter correction model; Correct at least one scatter image among the plurality of scatter images based on the updated trained scatter correction model; And Output at least one corrected scatter image, Wherein updating the trained scatter correction model includes: For each image pair including a scatter image among the plurality of scatter images and the single partial scatter-free image, perform interpolation between the output of the trained neural network model for the scatter image and the output of the trained neural network model for the single partial scatter-free image; And Generate the updated trained scatter correction model based on the interpolation.

11. The imaging system according to claim 10, wherein the trained scatter correction model includes a previously determined deconvolution point spread function (PSF) estimate.

12. The imaging system according to claim 11, wherein updating the trained scatter correction model includes: Locally parameterize the deconvolution PSF estimate using measurement points of corresponding image pairs including a scatter image among the plurality of scatter images and the single partial scatter-free image; and Update the deconvolution PSF estimate based on the local parameterization.

Images