JP2008043757A - System and method for processing image data - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform a three-dimensional display processing from PSF on a tomosynthesis data set, make the processing to be more sufficiently adapted from a PSF type classification mask, and also use this processing for the framework of a three-dimensional multiple resolving power processing. <P>SOLUTION: A system (100) and a method include the processing of the data set from the classification by a point spread function using a point spread function type rule for classifying a region in the data set. Furthermore, in another viewpoint, the method includes provision of a scanning object, scanning of the object for acquiring the data set, use of the point spread function type rule for classifying the region in the data set, and processing of the data set from the classification by the point spread function. <P>COPYRIGHT: (C)2008,JPO&INPIT

Description

  The present invention generally relates to x-ray systems and methods, and more specifically, a system that uses classification of structures in a data set to guide processing path selection and reduce and eliminate blur from an image. And a method.

  Digital tomosynthesis is widely used for three-dimensional (3D) reconstruction of objects obtained from angle-limited X-ray projection imaging using a movable X-ray tube and a stationary digital detector. Digital tomosynthesis is an improved technique of conventional linear tomography that has been known since the 1930s. Similar to linear tomography, tomosynthesis also causes residual blurring of objects located outside the plane of interest. This tomographic blur is often due to the overlapping anatomy, which blurs the details in the plane of interest and limits the contrast enhancement of the projected image slice. Removing overlapping blurred anatomy improves the contrast of the in-plane structure, which limits the dynamic range of the image to the section of interest and provides frequency content similar to the object of interest in such section. This is done by removing residual structures that may have. At a basic level, the point spread function (PSF) of the tomosynthesis system determines the blur spread characteristics in the imaging volume. The PSF of the tomosynthesis system is essentially a shift variant (position dependent). However, complete blur removal is not a trivial task. This task is complicated and computationally intensive because of the PSF range, and it is not easy to eliminate the blur.

Tomosynthesis allows the retrospective formation of any number of cross-sectional images from a single pass through the x-ray tube. Projected images are acquired during x-ray scans over a limited angle and reconstructed into a complete three-dimensional volumetric image. To enhance the display quality of the reconstructed slice, the tomosynthesis data is processed by a series of algorithms after the reconfiguration. Most prior art image processing techniques perform two-dimensional (2D) processing of the reconstructed image data set without considering the third dimension and prior knowledge of the imaging geometry. This is not suitable for tomosynthesis images where there may be non-focused high frequency components that can be sharpened by the display processing algorithm. Because of these artifacts and many other artifacts, engineers, researchers and scientists are interested in alternative ways of processing such images.
US Pat. No. 5,375,175 US Pat. No. 6,058,322

  Since true imaging geometry is represented using the 3D PSF of the data set, theoretically this information can be used to improve the display processing of the reconstructed image data set. It's a nasty habit. In addition, if, for example, all points in the imaging volume in any plane can be classified based on whether they are in focus or not in focus, the processing algorithm can be more fully adapted based on the classification mask. it can.

  Therefore, it is necessary to perform PSF-based 3D display processing on the tomosynthesis data set and to more fully adapt the processing based on the PSF type classification mask. It is also assumed that this process can be used as a framework for three-dimensional multiresolution processing.

  In one aspect, a method for processing a data set based on classification by a point spread function.

  Another aspect is a method of accessing a data set, using a point spread function type rule to classify regions in the data set, and processing the data set based on the classification by the point spread function.

  In yet another aspect, providing a scan target, scanning the target to obtain a data set, using a point spread function type rule to classify regions in the data set, and based on classification by a point spread function A method of processing a data set.

  In yet another aspect, the system is operatively coupled to an X-ray source, an X-ray detector arranged to receive X-rays emitted from the source, and the source and detector. Computer. The computer is configured to process the data set based on the classification by the point spread function.

  In yet another aspect, a computer readable medium is provided that incorporates a program configured to instruct a computer to process a data set based on classification by a point spread function.

  In yet another aspect, the method includes accessing the data set, processing the data set by a multi-resolution display process, and processing the data set by a Z weight function.

  In yet another aspect, the method accesses the tomosynthesis imaging data set, reconstructs the tomosynthesis imaging data set, and processes the reconstructed tomosynthesis imaging data set by multi-resolution display processing. Is included.

  Various other features, objects, and advantages of the invention will be made apparent to those skilled in the art from the accompanying drawings and the following detailed description.

  This document describes systems and methods useful in imaging systems such as, but not limited to, x-ray tomosynthesis systems. These systems and methods are described with reference to the drawings, wherein like reference numerals refer to the same components throughout the drawings. Such drawings are intended to be illustrative rather than restrictive and are included herein to facilitate the description of exemplary embodiments of the systems and methods of the present invention. Although described in the setting up of an X-ray tomosynthesis system, the benefits of the present invention are believed to be deceived by all imaging systems. One objective of this disclosure is to provide a classification framework that decomposes tomosynthesis data sets into multiple categories based on the nature of PSF, guides the selection of processing paths, and reduces and eliminates blur from images There is.

  Referring to the drawings, FIG. 1 illustrates an exemplary X-ray tomosynthesis imaging system 100. The imaging system 100 includes an X-ray source 102 that irradiates a structure under test 106 with X-ray photons. By way of example, the x-ray source 102 may be an x-ray tube and the structure under test 106 may be a patient, a test phantom, and / or other test inanimate object.

  The x-ray imaging system 100 also includes a detector 108 coupled to the processing circuitry. Processing circuitry (eg, CPU, microcontroller, microprocessor, custom ASIC, etc.) is coupled to the memory or data storage device and the display device. A memory or data storage device (eg, a solid state memory device, magnetic memory device, optical memory device, disk that reads instructions and / or data from a computer readable medium such as a flexible disk or other digital source such as a network or the Internet) Others including network connection devices such as storage devices, tape storage devices, flexible disk drives, hard disk drives, CD-ROM drives, DVD drives, magneto-optical disk (MOD) devices, or Ethernet (trademark) devices Any digital device, as well as one or more of the digital means under development, etc.) store the imaging data.

  The memory may also store a computer program that includes instructions that are executed by the processing circuitry to implement the operations described herein. The processing circuit forms an image for display on the device. As described in further detail herein, the images may represent various structures (eg, soft tissue, bone). The detector 108 may be, for example, a flat panel solid-state image detector, but may process conventional film images stored in digital form in memory. In one embodiment, the processing circuit executes instructions stored in firmware (not shown). Generally, the processor is programmed to perform the steps described below.

  The tomosynthesis imaging process includes a series of projected X-ray images acquired from various angles through the subject 106. Any number of discontinuous projection images are acquired over a limited angular range 104 by arcuate rotation or linear translation of the x-ray tube 102. In the embodiment shown in FIG. 1, the motion of the X-ray tube 102 is located in a plane 110 parallel to the plane 112 of the detector 108. After obtaining the projection image data set, the image slice is reconstructed using application software. Also shown in FIG. 1 is a coordinate system 114, illustrating the X, Y, and Z directions.

  Of course, the methods described herein are not limited to implementation in the system 100 and can be utilized in connection with many other types and variations of imaging systems. In one embodiment, the processing circuit is a computer programmed to perform the operations described herein, and thus the term computer as used herein is an integrated circuit referred to in the art as a computer. It refers broadly to computers, processors, microcontrollers, microcomputers, programmable logic controllers, application specific integrated circuits, and other programmable circuits. Although the methods described herein are described in the setting of a human patient, the benefits of the present invention are also discouraged for non-human imaging systems such as those typically used for small animal research. It is thought.

  As used in this document, the term element or step written in the singular and followed by a singular indefinite article is understood not to exclude the inclusion of a plurality of such elements or steps unless explicitly stated otherwise. I want. Furthermore, references to “one embodiment” of the present invention are not intended to be interpreted as excluding the existence of other embodiments that also incorporate the recited features.

  Further, the expression “reconstruct an image” used in this document does not exclude embodiments of the present invention in which data representing an image is generated but a visible image is not formed. Thus, the term “image” as used herein broadly refers to both the visible image and the data representing the visible image. However, many embodiments form (or are configured to form) at least one visible image.

  The methods described in this document are described in medical settings, such as systems typically used in industrial settings or, for example, but not limited to transportation settings such as baggage scanning systems at airports or other transportation points, etc. Even in non-medical imaging systems such as these, it is believed that the benefits of the present invention are appreciated.

  In multi-resolution processing, image representation and analysis at a resolution greater than one is performed. This provides computational simplicity and flexibility. In addition, features that are difficult to detect with one resolution may be easily found with another resolution. This multi-resolution processing technique is shown in FIG.

  FIG. 2 shows a filter processing sequence 200 for spatially selective enhancement of images. The example of FIG. 2 illustrates a two-dimensional filtering sequence (2D), but similar filtering sequences include other dimensions including a one-dimensional (1D) filtering sequence or a three-dimensional (3D) filtering sequence. Can also be achieved. This filter processing sequence includes a decomposition phase 202, an enhancement phase 204, and a resynthesis phase 206. The decomposition phase 204 includes a progressive decomposition of the image data into a series of spatial frequency images with progressively lower spatial frequencies. As used herein, the term spatial frequency is generally related to the number of individual image components within a defined image area. Due to the progressive decomposition of the image, the filtering sequence 200 forms a series of decomposed images having different spatial frequencies. Each of these images is then enhanced by applying a gain image in the enhancement phase 204. The resulting image is then recombined in the recomposition phase 206 to complete the enhanced image. Individual processing of the decomposed image at different spatial frequency levels is indicated in FIG. 2 by reference numbers 210, 212, 214, 216 and 218, where number 218 is the final level of decomposition and processing. As described above, each progressive step or level forms an intermediate image with progressively lower spatial frequencies, as shown at reference numbers 220, 222, 224, 226 and 228.

  As shown in FIG. 2, the processing block for each spatial frequency level of the decomposed phase includes a decimator 230 and an interpolator 232 represented by a low pass filter (the highest spatial frequency level in FIG. 2). Only the 210 filters are numbered). These filters progressively combine data representing adjacent pixels at the previous image level in a manner well known in the art. The filter outputs a first filtered image and a second image, the first filtered image is used at a subsequent decomposition level, and the second image is input at adder 234 for the spatial frequency level. Subtracted from the image. Thus, at the highest level 210, the original processed image data 236 is supplied to a low pass filter 230 as a decimator and an adder 234. This original image data is filtered in block 230 and block 232 and the output of block 230 is passed as input data to level 212. In this manner, the original processed image data is broken down into images with progressively lower spatial frequencies. In one embodiment, each subsequent step or level in the decomposition phase forms an image that is half the size of the input image for that level. However, other reduction ratios may be used. Further, the decomposition phase filtering may end at any desired level, but in this embodiment such filtering is performed until a single pixel (1 × 1) image is reached at the lowest level 218. .

  The filtering sequence 200 then proceeds to apply (eg, multiply) a gain image to each of the spatial frequency images in the enhancement phase 204. The gain image is indicated in FIG. 2 by reference numbers 238, 240, 242, 244 and 246, which are assigned values that emphasize certain features of the image, such as edges. In addition, some of the gain images are formed from data from relatively low spatial frequency levels. Thus, the gain image 238, denoted G1, is formed from the image 222 resulting from the decomposition phase at level 212. Similarly, a gain image 240, denoted G2, is formed from an image 224 obtained from processing at level 214. Any suitable number of levels of the filtering sequence 200 can use gain images based on such relatively low spatial frequencies. For example, in an image having an original size of 2048 × 2048 pixels, such a relatively low spatial frequency based gain image is sufficient when used at the first three levels of decomposition, such as levels 210, 212 and 214 in FIG. Edge enhancement is formed. This approach may be used at other lower levels, or these other levels may be multiplied by a gain value having other values, whether unit values or larger or smaller than unit values.

  Following application of the gain image in the enhancement phase 204, the resulting enhancement image is recombined in the recomposition phase 206. As shown in FIG. 2, each step or level of the resynthesis phase includes interpolation of output data from the previous (ie, relatively low) level at block 248 (see level 210 of FIG. 2). . The resulting image is added to the enhanced image at adder 250 at each level. The resulting enhanced and re-synthesized image data is then sent to the next (ie, relatively high) level until a re-synthesized and enhanced full-scale image 252 is obtained.

  It should be noted that for certain applications, the image data need not be completely decomposed or recombined to the lowest level. For example, if a series of relatively low spatial frequency images are applied with similar gain in phase 204, the decomposition may be stopped at this level. Gain can then be applied to the image at this end level to initiate resynthesis. Similarly, as described above, relatively high level gains are derived from the relatively low spatial frequency images obtained from the decomposition, but these gains are also recombined from the relatively low spatial frequency images. Can also be obtained based on the obtained image (ie after application of the gain image in the enhancement phase 204).

  In one embodiment, a gain based on a relatively low spatial frequency is derived as follows. Following the decomposition of the original filtered image data, the processing circuit applies an algorithm that can be expressed as:

G _i (x, y) = max (1.0, E _i * S (O _j (x, y)))
Where G _i (x, y) is the value of each pixel in the gain image at level i to be determined, E _i is a user-defined edge intensity value greater than 1.0, and S is A spatial sensitivity function that can be set by the user, where O _j (x, y) is the value of each pixel of image j at a relatively low spatial frequency level. Therefore, it should be noted that when the calculated value obtained is smaller than the unit value, the above algorithm effectively holds the unit value in the derived gain image. This algorithm also allows the user to effectively define the desired edge strength value, making it easier to increase or decrease the relative sharpness between the edge and the recombined and enhanced image. . Similarly, a spatial sensitivity function can be selected by the user to adjust the amount of influence exerted by the relatively low spatial frequency image values in the gain calculation. The spatial sensitivity function includes a linear function, a piecewise linear function, a sigmoid function, and the like. In the present embodiment, the value S obtained from the function S may be set anywhere between 0 and 1. Further, the edge intensity value or the spatial sensitivity function, or both, may be unique for a particular spatial frequency level (ie unique for the individual levels 210, 212, and 214 in FIG. 2). Further, the specific level from which the gain image based on the spatial frequency is derived may be a parameter that can be configured by the user. Thus, the user can select a spatial frequency based enhancement for 1, 2, 3 or even more levels depending on the desired enhancement and the frequency level at which useful information or noise is expected. .

  Within the multi-resolution framework, a set of tomosynthesis reconstructed slices is considered as one of the inputs. Another input is a PSF type classification mask corresponding to a large number of regions such that there is one classification type for each region. As an example, there are the following four classification types. (1) in focus, (2) non-focus, (3) background, and (4) low frequency, and these classification types are described in US patent application Ser. No. 11 / 463,845. . This mask was numerically encoded to identify distinct regions. That is, (1) m = 1 for in-focus data, (2) m = 2 for non-focus data, (3) m = 3 for background data, and (4) m = 4 for low-frequency data. Where m represents the value of each pixel of the mask. These inputs are assumed to have the same resolution using up-sampling (re-synthesis) or down-sampling (decomposition) methods. Once each input has the same resolution, the data can be processed using several different methods.

  A new methodology is presented for 3D processing of tomosynthesis data based on PSF. There are several forms that this system can take while balancing the efficiency and accuracy of computation. The following are examples of several methodologies.

  The method 300 for processing an imaging data set uses conventional two-dimensional multi-resolution display processing and a one-dimensional (1D) Z weighting function 306 derived from the PSF along the midline of the three-dimensional PSF. FIG. 3 is a flow diagram illustrating the method 300. The method 300 includes the following steps. That is, (1) Step 302 for reconstructing an image slice from the acquired projection image, (2) In order to obtain a displayed image slice, conventional two-dimensional multi-resolution display processing is performed on each image slice. Performing 304 (not utilizing PSF classification regions); (3) determining a PSF-based one-dimensional Z-weighted profile along the PSF midline at each point of the displayed image slice data; And (4) using a PSF-based one-dimensional Z-weighted profile for the one-dimensional Z-weighting process along the profile direction (not utilizing the PSF classification region) to obtain an enhanced final image 308, where step 306 The one-dimensional Z weighting process may be a sharpening operation that reduces blurring.

  The method 400 for processing an imaging data set uses a conventional two-dimensional multi-resolution display process, a PSF type classification mask 406, and a one-dimensional Z weighting function 408 derived from the shape of the PSF and three-dimensional PSF. The one-dimensional Z weighting process 408 uses the PSF classification areas 410, 412, 414, and 416. FIG. 4 is a flow diagram illustrating the method 400. The method 400 includes the following steps. (1) Step 402 for reconstructing an image slice from the acquired projection image, (2) In order to obtain a displayed image slice, conventional two-dimensional multi-resolution display processing is performed on each image slice. Performing 404 (not utilizing PSF classification regions), (3) determining a PSF-based one-dimensional Z-weighted profile along the PSF midline at each point of the displayed image slice data; And (4) adaptively using a PSF-based one-dimensional Z-weighted profile for the one-dimensional Z-weighting process 408 along the profile direction based on the pre-classification type 406 to obtain an enhanced final image 418 (PSF classification Step 408) (using regions 410, 412, 414, 416). The one-dimensional Z weighting process 408 may be a sharpening operation that reduces blur. The strength of sharpening can vary based on the classification type.

  The method 500 for processing an imaging data set uses a PSF type classification mask 504 for both a conventional two-dimensional multi-resolution display process 506 and a one-dimensional Z weighting process along the PSF midline profile. The conventional two-dimensional multi-resolution display processing 506 uses the PSF classification areas 510, 512, 514, and 516. The one-dimensional Z weighting process 508 uses the PSF classification regions 520, 522, 524, and 526. FIG. 5 is a flowchart illustrating the method 500. The method 500 includes the following steps. That is, (1) Step 502 for reconstructing an image slice from the acquired projection image, and (2) Conventional image processing for each image slice based on the pre-classification type 504 to obtain a displayed image slice. Step 506 for adaptively performing a two-dimensional multi-resolution display process (utilizing PSF classification regions 510, 512, 514, 516) (Note that this adaptive process is another conventional technique that the conventional method may already have. (Note that this is done in addition to the adaptive processing), (3) determining a one-dimensional Z-weighted profile based on the PSF along the midline of the PSF at each point of the displayed image slice data. And (4) a one-dimensional Z-weighting process along the profile direction based on the pre-classification type 504 to obtain an enhanced final image 518 For (utilizing PSF classified regions 520, 522, 524, 526) using a one-dimensional Z weighted profile adaptively based on PSF in a step 508. The one-dimensional Z weighting process may be a sharpening operation that reduces blur. The strength of sharpening can vary based on the classification type. Different processing algorithms can be applied to different areas of the mask. For example, the in-focus structure can be sharpened to reduce the effects of non-focus details. The resulting data set can be later sampled to achieve the desired resolution.

  A method 600 for processing an imaging data set uses a PSF type classification mask 604 with a three-dimensional multi-resolution display process 606. The three-dimensional multi-resolution display process 606 uses the PSF classification areas 610, 612, 614, and 616. FIG. 6 is a flow diagram illustrating the method 600. The method 600 includes the following steps. That is, based on the pre-classification type 604 to obtain (1) a step 602 to reconstruct an image slice from the acquired projection image, and (2) a displayed image slice 606 and an enhanced final image 608. Step 606 for adaptively performing a 3D multi-resolution display process for each image slice (using PSF classification regions 610, 612, 614, 616), where different regions in the encoded mask Different processing algorithms may be applied.

  A method 700 for processing a tomosynthesis imaging data set uses a conventional three-dimensional multi-resolution display process 704 (not using a PSF classification region). FIG. 7 is a flow diagram illustrating a method 700. The method 700 includes the following steps. (1) Step 702 for reconstructing a tomosynthesis image slice from the acquired projection image, and (2) conventional three-dimensional multi-resolution display for each image slice to obtain an enhanced final image 706. Step 704 of executing the process (not using the PSF classification area).

  The method 800 for processing an imaging data set uses a PSF type classification mask 804 for a conventional two-dimensional multi-resolution display process 806 and a one-dimensional Z-weighting process 808 along the PSF midline profile. The conventional two-dimensional multi-resolution display processing uses PSF classification regions 810, 812, 814, and 816. The one-dimensional Z weighting process 808 does not use the PSF classification region. FIG. 8 is a flow diagram illustrating a method 800. The method 800 includes the following steps. That is, (1) Step 802 for reconstructing an image slice from the acquired projection image, and (2) Conventional image processing for each image slice based on the pre-classification type 804 to obtain a displayed image slice. Step 806 (uses PSF classification regions 810, 812, 814, 816) to adaptively execute the two-dimensional multi-resolution display process (Note that this adaptive process is another conventional technique that the conventional method may already have. (Note that this is done in addition to the adaptive processing), (3) determining a one-dimensional Z-weighted profile based on the PSF along the midline of the PSF at each point of the displayed image slice data. Step, and (4) one-dimensional Z based on PSF for one-dimensional Z weighting processing along the profile direction to obtain an enhanced final image 818 Heavy profiles used (not using PSF classification region) is a step 808, wherein the one-dimensional Z weighting process may be a sharpening operation to reduce blurring.

  That is, many different forms can be created based on the general framework described in this document. These forms arise because processing in one or more dimensions can be performed adaptively or non-adaptively. In addition to adapting conventional image processing, the present invention discloses a new PSF type classification that adapts display processing.

  One technical effect is that the methods and systems described herein provide the ability to classify structures (ie, regions) based on PSF, so that it is different for each different class as desired. Data can be selectively processed. This improved ability to process the imaging data leads to relatively little artifacts and / or blur in the final reconstructed image.

  The example of the embodiment has been described in detail above. These assemblies and methods are not limited to the specific embodiments described herein, and each assembly and / or method component is utilized independently and separately from the other components described herein. You can do it.

  Although the present invention has been described with reference to various embodiments, those skilled in the art will recognize that several substitutions, modifications, and omissions may be made to the embodiments without departing from the spirit of the invention. Accordingly, the foregoing description is intended to be illustrative only and is not intended to limit the scope of the invention as recited in the claims. Further, the reference numerals in the claims corresponding to the reference numerals in the drawings are merely used for easier understanding of the present invention, and are not intended to narrow the scope of the present invention. Absent. The matters described in the claims of the present application are incorporated into the specification and become a part of the description items of the specification.

1 illustrates an exemplary X-ray imaging system. FIG. It is a figure which shows the image decomposition series which divides | segments an image into the decomposition image referred by a desired frequency band. 5 is a flow diagram illustrating an exemplary embodiment for processing an imaging data set. 3 is a flow diagram illustrating another exemplary embodiment for processing an imaging data set. 6 is a flow diagram illustrating yet another exemplary embodiment for processing an imaging data set. 6 is a flow diagram illustrating yet another exemplary embodiment for processing an imaging data set. 6 is a flow diagram illustrating yet another exemplary embodiment for processing an imaging data set. 5 is a flow diagram illustrating an exemplary embodiment for processing an imaging data set.

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Imaging system 102 X-ray source 104 Limited angle range 106 Test structure 108 Detector 110 X-ray source plane 112 Detector plane 114 X, Y, Z coordinate system 200 Filtering series 202 Decomposition phase 204 Enhancement phase 206 Re Composite Phase 210, 212, 214, 216, 218 Spatial Frequency Level 220, 222, 224, 226, 228 Intermediate Image 230 Decimator 232 Interpolator 234 Adder 236 Original Processed Image Data 238, 240, 242, 244, 246 Gain Image 248 Interpolator 250 Adder 252 Recombined and enhanced image 300 Method of processing image data set 302 Reconstructing image slice 304 Conventional two-dimensional multi-resolution processing 306 One-dimensional Z-weighted processing 308 Enhanced final Image 400 Image data 402 Reconstruct image slices 404 Conventional two-dimensional multi-resolution processing 406 PSF type classification mask 408 Adaptive PSF based one-dimensional Z weighting process 410, 412, 414, 416 PSF classification region 418 Enhanced final image 500 Method of processing image data set 502 Reconstruct image slice 504 PSF type classification mask 506 Adaptive PSF based 2D multi-resolution processing 508 Adaptive PSF based 1D Z weighting processing 510 512, 514, 516, 520, 522, 524, 526 PSF classification region 518 Enhanced final image 600 Method of processing image data set 602 Reconstruct image slice 604 PSF type classification mask 606 Based on adaptive PSF Three-dimensional multi-resolution processing 608 Toned final image 610, 612, 614, 616 PSF classification region 700 Method of processing image data set 702 Reconstruct image slice 704 Conventional three-dimensional multi-resolution processing 706 Enhanced final image 800 Process image data set Method 802 Reconstruct image slice 804 PSF type classification mask 806 Two-dimensional multi-resolution processing based on adaptive PSF 808 One-dimensional Z weight processing 810, 812, 814, 816 PSF classification region 818 Enhanced final image

Claims (10)

An X-ray source (102);
An X-ray detector (108) arranged to receive X-rays emitted from the source (102);
A computer coupled in operation to the source (102) and the detector (108) and configured to process a data set based on classification by a point spread function;
A system (100) comprising:   The system (100) of claim 1, wherein the data set is a tomosynthesis data set.   The system (100) of claim 1, wherein the classification is a classification into a class including a background class, a focused class, a non-focused class, and a low frequency class.   The system (100) of claim 1, wherein the classification is a classification into a class consisting of a background class, a focused class, a non-focused class, and a low frequency class.   The system (100) of claim 2, wherein the classification is a classification into a class including a background class, a focused class, a non-focused class, and a low frequency class.   The system (100) of claim 1, wherein the computer provides a platform for three-dimensional adaptive deblurring of tomosynthesis data sets.   The system (100) of claim 1, wherein the computer is configured to provide adaptive processing with a point spread function (PSF) of tomosynthesis images.   The system (100) of claim 1, wherein the processing encompasses all forms of multi-resolution processing of tomosynthesis images.   A method comprising the step of processing a data set based on classification by a point spread function. Accessing the data set;
Using point spread functional rules to classify regions in the data set;
Processing the data set based on classification by the point spread function;
With a method.

Images