Image processing apparatus, method, and program

Abstract

Provided is an image processing device configured such that a processor: acquires a plurality of projection images generated by tomosynthesis imaging; extracts a specific structure from the plurality of projection images in order to derive a plurality of structure projection images; reconstructs the plurality of structure projection images in order to derive a structure tomographic image in each of a plurality of tomographic planes of a subject; detects at least one characteristic structure from the plurality of structure tomographic images; with the characteristic structure as a reference, and in a corresponding tomographic plane corresponding to the structure tomographic image in which the characteristic structure has been detected, corrects position shifts which are among the plurality of projection images and which are based on body movement of the subject, in order to reconstitute the plurality of projection images; and derives a corrected tomographic image in the at least one tomographic plane of the subject.

Description

Technical Field

[0001] This application claims the priority of Japanese Application No. 2022-035379 filed on March 8, 2022, the entire content of which is incorporated herein by reference. The present disclosure relates to an image processing apparatus, method, and program.

Background Art

[0002] In recent years, in a radiation imaging apparatus using radiation such as X-rays and gamma rays, in order to observe a diseased part in more detail, the radiation source is moved to irradiate the subject with radiation from a plurality of source positions for imaging, and a plurality of projection images thus obtained are added to derive a tomographic image in which a desired tomographic plane is emphasized. Tomosynthesis imaging has been proposed. In tomosynthesis imaging, according to the characteristics of the imaging apparatus and the required tomographic image, the radiation source is moved parallel to the radiation detector or moved so as to draw an arc of a circle or an ellipse, and the subject is imaged at a plurality of source positions to obtain a plurality of projection images, and these projection images are reconstructed using an inverse projection method such as a simple inverse projection method or a filtered inverse projection method to derive a tomographic image.

[0003] By deriving such tomographic images at a plurality of tomographic planes in the subject, it is possible to separate structures overlapping in the depth direction in which the tomographic planes are arranged. Therefore, it becomes possible to detect a lesion that was difficult to detect in a two-dimensional image obtained by conventional simple imaging. Note that simple imaging is an imaging method in which the subject is irradiated with radiation once to obtain a single two-dimensional image that is a transmission image of the subject.

[0004] On the other hand, tomosynthesis imaging also has a problem that the reconstructed tomographic image is blurred due to the influence of body movement of the subject caused by the time difference of imaging at each of the plurality of source positions. When the tomographic image is blurred in this way, it becomes difficult to detect lesions such as minute calcifications, which are useful for early detection of breast cancer, especially when the breast is the subject.

[0005] For this reason, methods have been proposed to correct for motion when deriving tomographic images from projection images acquired by tomosynthesis imaging. For example, International Publication No. 2020 / 067475 proposes a method in which at least one feature point is detected in the derived tomographic image, the amount of positional displacement between multiple projection images based on the subject's motion is derived using the feature point as a reference in the corresponding tomographic plane corresponding to the tomographic image in which the feature point was detected, and the amount of positional displacement is corrected to reconstruct the multiple projection images and derive a corrected tomographic image in which the effects of motion have been corrected. [Overview of the project] [Problems that the invention aims to solve]

[0006] The method described in International Publication No. 2020 / 067475 detects feature points from tomographic images, but feature points are not always detected in projection images. For example, even if feature points are present with high contrast in tomographic images, they may have low contrast in projection images, making them difficult to detect. If feature points are difficult to detect in projection images, it is not possible to accurately match feature points in tomographic images with feature points in projection images, and in that case, motion cannot be accurately corrected.

[0007] This disclosure is made in view of the above circumstances and provides an image processing apparatus, method, and program that can acquire high-quality tomographic images with accurate correction of body motion. [Means for solving the problem]

[0008] The image processing apparatus according to this disclosure comprises at least one processor, The processor is, By moving the radiation source relative to the detection surface of the detection unit and having the imaging device perform tomosynthesis imaging in which radiation is irradiated onto the subject at multiple source positions resulting from the movement of the radiation source, multiple projection images corresponding to each of the multiple source positions are obtained. By extracting specific structures from multiple projection images, multiple structural projection images are derived. By reconstructing multiple structural projection images, structural tomographic images are derived for each of the multiple tomographic planes of the subject. Detect at least one feature structure from multiple structural tomographic images, The system is configured to reconstruct multiple projection images by correcting for positional shifts between them based on the subject's motion, using the feature structure as a reference, in the corresponding tomographic plane for which a feature structure has been detected, and to derive a corrected tomographic image in at least one tomographic plane of the subject.

[0009] "Moving the radiation source relative to the detection unit" includes moving only the radiation source, moving only the detection unit, and moving both the radiation source and the detection unit.

[0010] In addition, in the image processing apparatus according to this disclosure, the specific structure may be at least one of a line structure and a point structure.

[0011] Furthermore, in the image processing apparatus according to this disclosure, the processor may extract at least one of a line structure and a point structure based on the degree of concentration of a gradient vector representing the gradient of pixel values ​​in the projected image.

[0012] Furthermore, in the image processing apparatus according to this disclosure, the processor derives the amount of positional displacement between multiple projected images based on the motion of the subject, using the characteristic structure as a reference in the corresponding tomographic plane. A corrected tomographic image may also be derived by correcting the amount of positional displacement and reconstructing multiple projection images.

[0013] Furthermore, in the image processing apparatus according to this disclosure, the processor detects multiple feature structures from multiple structural tomographic images, It is determined whether the corresponding tomographic plane for the structural tomographic image in which each of the multiple feature structures was detected is the focal plane. The method may also involve deriving the amount of displacement in the corresponding fault plane identified as the focal plane.

[0014] Furthermore, in the image processing apparatus according to this disclosure, the processor may detect points in a structural tomographic image that satisfy specific threshold conditions as feature structures.

[0015] Furthermore, in the image processing apparatus according to this disclosure, the processor updates the structural tomographic image by reconstructing the structural projection image while correcting for positional misalignment. Detect updated feature structures from updated structural tomographic images. The positional displacement amount is updated using the updated feature structure. This may involve repeatedly updating structural tomographic images, characteristic structures, and positional displacement amounts.

[0016] Furthermore, in the image processing apparatus according to this disclosure, the processor updates the structural tomographic image by reconstructing the structural projection image while correcting for positional misalignment. Based on the updated threshold conditions, updated feature structures are detected from the updated structural tomographic images. The positional displacement amount is updated using the updated feature structure. The system may repeatedly update structural tomographic images, update feature structures based on updated threshold conditions, and update positional displacement amounts.

[0017] Furthermore, in the image processing apparatus according to this disclosure, the processor projects the multiple projection images onto the corresponding tomographic plane based on the positional relationship between the source position and the detection unit at the time of acquisition for each of the multiple projection images, and derives a tomographic projection image corresponding to each of the multiple projection images. In the corresponding tomographic plane, the amount of positional displacement between multiple tomographic projection images based on the subject's motion may be derived as the amount of positional displacement between multiple projection images, using the characteristic structure as a reference.

[0018] Furthermore, in the image processing apparatus according to this disclosure, the processor may set local regions corresponding to feature structures in multiple tomographic projection images and derive the amount of positional displacement based on the local regions.

[0019] The "local region" is a region including a characteristic structure in a tomographic image or a tomographic plane projection image, and can be an arbitrarily sized region smaller than the tomographic image or the tomographic plane projection image.

[0020] Note that the local region needs to be larger than the range of movement as body movement. When the body movement is large, it may be about 2 mm. Therefore, in the case of a tomographic image or a tomographic plane projection image with a pixel size of 100 μm on each side, the local region may be, for example, a region of 50×50 pixels or 100×100 pixels around the characteristic structure.

[0021] The "region around the characteristic structure in the local region" means a region smaller than the local region and including the characteristic structure within the local region.

[0022] Also, in the image processing apparatus according to the present disclosure, the processor sets a plurality of first local regions including characteristic structures in a plurality of tomographic plane projection images, sets a second local region including a characteristic structure in a tomographic image in which the characteristic structure is detected, derives, as a provisional displacement amount, the displacement amount of each of the plurality of first local regions with respect to the second local region, and may derive the displacement amount based on the plurality of provisional displacement amounts.

[0023] Also, in the image processing apparatus according to the present disclosure, the processor may derive a provisional displacement amount based on the region around the characteristic structure in the second local region.

[0024] Also, in the image processing apparatus according to the present disclosure, the processor reconstructs a plurality of projection images excluding the target projection image corresponding to the target tomographic plane projection image for which the displacement amount is to be derived to derive a plurality of tomographic images as target tomographic images, and may derive the displacement amount for the target tomographic plane projection image using the target tomographic image.

[0025] Furthermore, in the image processing apparatus according to this disclosure, the processor may perform an image quality evaluation of the region of interest, including the feature structure in the corrected tomographic image, and determine whether the derived positional displacement is appropriate or inappropriate based on the results of the image quality evaluation.

[0026] Furthermore, in the image processing apparatus according to this disclosure, the processor derives multiple tomographic images by reconstructing multiple projection images, The image quality of the region of interest, including the characteristic structure in the tomographic image, was evaluated, and the results of the image quality evaluation of the corrected tomographic image were compared with the results of the image quality evaluation of the tomographic image. The better tomographic image may be selected as the final tomographic image.

[0027] Furthermore, in the image processing apparatus according to this disclosure, the processor derives an evaluation function for evaluating the image quality of a region of interest, including feature structures, in a corrected tomographic image. It may also be a method for deriving the positional displacement amount that optimizes the evaluation function.

[0028] Furthermore, in the image processing apparatus according to this disclosure, the subject may be a breast.

[0029] Furthermore, in the image processing apparatus according to this disclosure, the processor may change the search range for deriving the positional displacement amount according to at least one of the following: breast density, breast size, tomosynthesis imaging time, breast compression pressure during tomosynthesis imaging, and breast imaging direction.

[0030] The image processing method according to this disclosure involves moving a radiation source relative to the detection surface of a detection unit, and having the imaging device perform tomosynthesis imaging in which radiation is irradiated onto a subject at multiple source positions due to the movement of the radiation source, thereby acquiring multiple projection images corresponding to each of the multiple source positions, and By extracting specific structures from multiple projection images, multiple structural projection images are derived. By reconstructing multiple structural projection images, structural tomographic images are derived for each of the multiple tomographic planes of the subject. Detect at least one feature structure from multiple structural tomographic images, In the corresponding tomographic plane for the structural tomographic image in which a feature structure was detected, the positional displacement between multiple projection images based on the subject's motion is corrected using the feature structure as a reference to reconstruct multiple projection images, thereby deriving a corrected tomographic image for at least one tomographic plane of the subject.

[0031] The image processing program according to this disclosure moves a radiation source relative to the detection surface of the detection unit, and performs tomosynthesis imaging in which the subject is irradiated with radiation at multiple source positions due to the movement of the radiation source, thereby acquiring multiple projection images corresponding to each of the multiple source positions generated. By extracting specific structures from multiple projection images, multiple structural projection images are derived. By reconstructing multiple structural projection images, structural tomographic images are derived for each of the multiple tomographic planes of the subject. Detect at least one feature structure from multiple structural tomographic images, The computer is instructed to perform a process that includes correcting for positional shifts between multiple projection images based on the subject's motion in the corresponding tomographic plane of the structural tomographic image in which a feature structure has been detected, using the feature structure as a reference, reconstructing multiple projection images, and deriving a corrected tomographic image in at least one tomographic plane of the subject. [Effects of the Invention]

[0032] According to this disclosure, high-resolution tomographic images with accurate motion correction can be obtained. [Brief explanation of the drawing]

[0033] [Figure 1] Schematic diagram of a radiographic imaging apparatus to which the image processing apparatus according to the first embodiment of this disclosure is applied. [Figure 2] A view of the radiographic imaging device from the direction of arrow A in Figure 1. [Figure 3]This figure shows the schematic configuration of the image processing device according to the first embodiment. [Figure 4] This figure shows the functional configuration of the image processing apparatus according to the first embodiment. [Figure 5] Diagram to explain the acquisition of projected images [Figure 6] Figure showing projected image and structural projected image. [Figure 7] Diagram to explain the derivation of structural tomographic images. [Figure 8] A diagram illustrating the detection of feature structures from structural tomographic images. [Figure 9] A diagram illustrating the projection of a projected image onto a corresponding tomographic plane. [Figure 10] A diagram illustrating the interpolation of pixel values ​​in tomographic images. [Figure 11] Diagram to explain the setting of the area of ​​interest [Figure 12] A diagram showing the region of interest set in the tomographic projection image. [Figure 13] In the first embodiment, the figure shows an image within the region of interest when no body movement is occurring. [Figure 14] In the first embodiment, the figure shows an image within the region of interest when body movement occurs. [Figure 15] A diagram illustrating the scope of exploration in the area of ​​interest. [Figure 16] A diagram showing the feature structure in 3D space. [Figure 17] Diagram showing the display screen of the corrected tomographic image. [Figure 18] A flowchart illustrating the process performed in the first embodiment. [Figure 19] In the second embodiment, the figure shows an image within the region of interest when no body movement is occurring. [Figure 20] In the second embodiment, a figure showing an image within the region of interest when body movement occurs. [Figure 21] Diagram to explain the region surrounding the feature structure. [Figure 22] A schematic diagram showing the process performed in the third embodiment. [Figure 23]A flowchart illustrating the process performed in the fourth embodiment. [Figure 24] Diagram showing a warning display [Figure 25] This figure shows the functional configuration of the image processing apparatus according to the fifth embodiment. [Figure 26] Diagram to explain ripple artifacts [Figure 27] Diagram to explain the derivation of corresponding points [Figure 28] This figure shows the result of plotting the feature structure and the pixel values ​​of the corresponding points. [Figure 29] A flowchart illustrating the process performed in the fifth embodiment. [Figure 30] This figure shows the functional configuration of the image processing apparatus according to the sixth embodiment. [Figure 31] A diagram illustrating the setting of the region of interest in the sixth embodiment. [Figure 32] A flowchart illustrating the process performed in the sixth embodiment. [Figure 33] This figure shows the functional configuration of the image processing apparatus according to the seventh embodiment. [Modes for carrying out the invention]

[0034] Embodiments of this disclosure will be described below with reference to the drawings. Figure 1 is a schematic diagram of a radiographic imaging system to which an image processing device according to the first embodiment of this disclosure is applied. As shown in Figure 1, the radiographic imaging system 1 is for acquiring multiple radiographic images, i.e., multiple projection images, by photographing the breast, which is the subject, from multiple source positions in order to perform tomosynthesis imaging of the breast and derive a tomographic image. The radiographic imaging system 1 comprises an imaging device 10, a console 2, an image storage system 3, and an image processing device 4 according to this embodiment.

[0035] Figure 2 shows the imaging device in the radiographic imaging system as viewed from the direction of arrow A in Figure 1. The imaging device 10 is a mammography device that acquires multiple radiographic images, i.e., multiple projection images, by imaging the subject, the breast M, from multiple source positions in order to perform tomosynthesis imaging of the breast and derive a tomographic image.

[0036] The imaging device 10 includes an arm portion 12 connected to a base (not shown) by a rotating shaft 11. An imaging table 13 is attached to one end of the arm portion 12, and a radiation irradiation unit 14 is attached to the other end, facing the imaging table 13. The arm portion 12 is configured to rotate only at the end to which the radiation irradiation unit 14 is attached, thereby allowing the imaging table 13 to be fixed while only the radiation irradiation unit 14 is rotated. The rotation of the arm section 12 is controlled by the console 2.

[0037] The imaging table 13 is equipped with a radiation detector 15, such as a flat panel detector. The radiation detector 15 has a radiation detection surface 15A, such as X-rays. The imaging table 13 also houses a circuit board, which includes a charge amplifier that converts the charge signal read from the radiation detector 15 into a voltage signal, a correlated double sampling circuit that samples the voltage signal output from the charge amplifier, and an AD (Analog Digital) conversion unit that converts the voltage signal into a digital signal. The radiation detector 15 is just one example of a detection unit. In this embodiment, the radiation detector 15 is used as the detection unit, but it is not limited to the radiation detector 15 as long as it can detect radiation and convert it into an image.

[0038] The radiation detector 15 can repeatedly record and read out radiation images. It may use a so-called direct-type radiation detector that directly converts radiation such as X-rays into electric charge, or it may use a so-called indirect-type radiation detector that first converts radiation into visible light and then converts that visible light into an electric charge signal. Furthermore, as a method for reading out the radiation image signal, it is desirable to use a so-called TFT (Thin Film Transistor) readout method in which the radiation image signal is read out by turning a TFT switch on and off, or a so-called optical readout method in which the radiation image signal is read out by irradiating it with reading light. However, it is not limited to these, and other methods may also be used.

[0039] The radiation irradiation unit 14 houses an X-ray source 16, which is the radiation source. The timing of irradiation with X-rays from the X-ray source 16, as well as the X-ray generation conditions in the X-ray source 16, namely the selection of target and filter materials, tube voltage, and irradiation time, are controlled by the console 2.

[0040] Furthermore, the arm portion 12 is provided with a compression plate 17 positioned above the imaging table 13 to press down on and compress the breast M, a support portion 18 to support the compression plate 17, and a movement mechanism 19 to move the support portion 18 in the vertical direction as shown in Figures 1 and 2.

[0041] Console 2 has the function of controlling the imaging device 10 using imaging orders and various information obtained from a RIS (Radiology Information System) (not shown) via a wireless communication LAN (Local Area Network) or other network, as well as instructions given directly by the technician. Specifically, Console 2 causes the imaging device 10 to perform tomosynthesis imaging of the breast M, thereby acquiring multiple projection images as described later. As an example, in this embodiment, a server computer is used as Console 2.

[0042] Image storage system 3 is a system that stores image data such as radiographic images and tomographic images captured by the imaging device 10. Image storage system 3 retrieves image data from the stored image data in response to requests from console 2 and image processing device 4, etc., and transmits it to the requesting device. A specific example of image storage system 3 is PACS (Picture Archiving and Communication Systems).

[0043] Next, an image processing apparatus according to the first embodiment will be described. First, with reference to Figure 3, the hardware configuration of the image processing apparatus according to the first embodiment will be described. As shown in Figure 3, the image processing apparatus 4 is a computer such as a workstation, server computer, and personal computer, and includes a CPU (Central Processing Unit) 21 and non-volatile components. The image processing device 4 includes a storage device 23 and a memory 26 as a temporary storage area. The image processing device 4 also includes a display 24 such as a liquid crystal display, input devices 25 such as a keyboard and mouse, and a network interface 2 connected to a network (not shown). It includes 7 components: CPU 21, storage 23, display 24, input device 25, memory The RI 26 and network I / F 27 are connected to the bus 28. The CPU 21 is an example of a processor in this disclosure.

[0044] The storage 23 is implemented using an HDD (Hard Disk Drive), an SSD (Solid State Drive), and flash memory, etc. The image processing program 22 installed on the image processing device 4 is stored in the storage 23 as a storage medium. The CPU 21 reads the image processing program 22 from the storage 23, expands it into memory 26, and executes the expanded image processing program 22.

[0045] The image processing program 22 is stored in a memory device of a server computer connected to the network, or in network storage, in a state that allows external access, and is downloaded and installed on the computers comprising the image processing device 4 upon request. Alternatively, it is recorded on a recording medium such as a DVD (Digital Versatile Disc) or CD-ROM (Compact Disc Read Only Memory) and distributed, and the image processing device 4 is accessed from that recording medium. It is installed on the computers that make up the system.

[0046] Next, the functional configuration of the image processing apparatus according to the first embodiment will be described. Figure 4 is a diagram showing the functional configuration of the image processing apparatus according to the first embodiment. As shown in Figure 4, the image processing apparatus 4 comprises an image acquisition unit 31, a structure extraction unit 32, a reconstruction unit 33, a feature structure detection unit 34, a projection unit 35, a positional displacement amount derivation unit 36, and a display control unit 37. The CPU 21 functions as the image acquisition unit 31, structure extraction unit 32, reconstruction unit 33, feature structure detection unit 34, projection unit 35, positional displacement amount derivation unit 36, and display control unit 37 by executing the image processing program 22.

[0047] The image acquisition unit 31 acquires multiple projection images generated when the console 2 causes the imaging device 10 to perform tomosynthesis imaging. The image acquisition unit 31 acquires multiple projection images from the console 2 or the image storage system 3 via the network interface 18.

[0048] The image acquisition unit 31 acquires multiple projection images generated when the console 2 causes the imaging device 10 to perform tomosynthesis imaging. The image acquisition unit 31 receives network I / F from the console 2 or the image storage system 3. 27 Multiple projection images are obtained via this method.

[0049] As shown in Figure 5, the X-ray source 16 is moved to each of the source positions Si (i=1 to n), and the X-ray source 16 is driven at each source position to irradiate the breast M with X-rays. The radiation detector 15 detects the X-rays that have passed through the breast M, and projection images G1, G2, ... Gn are obtained corresponding to each of the source positions S1 to Sn. At each of the source positions S1 to Sn, the same dose of X-rays is irradiated onto the breast M.

[0050] In Figure 5, the source position Sc is the source position where the optical axis X0 of the X-rays emitted from the X-ray source 16 is perpendicular to the detection surface 15A of the radiation detector 15. Hereinafter, the source position Sc will be referred to as the reference source position Sc, and the projection image Gc obtained by irradiating the breast M with X-rays at the reference source position Sc will be referred to as the reference projection image Gc. Here, "the optical axis X0 of the X-rays is perpendicular to the detection surface 15A of the radiation detector 15" means that the optical axis X0 of the X-rays intersects the detection surface 15A of the radiation detector 15 at a 90-degree angle, but is not limited to this, and also includes cases where it intersects with a certain degree of error relative to 90 degrees. For example, even if the optical axis X0 of the X-rays intersects the detection surface 15A of the radiation detector 15 with an error of about ±3 degrees relative to 90 degrees, this is also included in "the optical axis X0 of the X-rays is perpendicular to the detection surface 15A of the radiation detector 15" in this embodiment.

[0051] The structure extraction unit 32 derives multiple structural projection images by extracting specific structures from each of the multiple projection images Gi. Examples of specific structures include at least one of line structures or point structures contained in the breast M. Examples of line structures include mammary glands, spicules, and blood vessels. Examples of point structures include calcifications, intersections of multiple mammary glands, and intersections of blood vessels. Note that point structures are not limited to minute points, but also include regions with a predetermined area. Furthermore, in the following description, both line structures and point structures will be used as specific structures. In this embodiment, in order to extract line structures and point structures, the concentration of the gradient vector representing the gradient of pixel values ​​in the projection image Gi, as described in "Concentration Evaluation Method and Vector Concentration Filter," Yoshinaga et al., MEDICAL IMAGING TECHNOLOGY VOL.19 No.3, 2001, is used.

[0052] The method for extracting line structures using concentration involves using a line concentration filter to find gradient vectors on both sides of a search line along a certain direction in the projected image Gi, evaluating the degree of concentration where the gradient vectors converge, and extracting search lines with high evaluations as line structures. The method for extracting point structures using concentration involves using a point concentration filter to find gradient vectors pointing in a certain direction in the projected image Gi, evaluating the degree of concentration where the gradient vectors converge, and extracting points with high evaluations as point structures. By using the concentration of vectors, it is possible to extract point structures or line structures from the projected image Gi regardless of the contrast of the projected image Gi.

[0053] Furthermore, it is preferable to set the coefficient of the concentration filter so that line structures are approximately 1 mm in actual size, and point structures are approximately 0.5 mm in actual size. This prevents noise contained in the projected image Gi from being extracted as line structures and point structures.

[0054] Furthermore, the structure extraction unit 32 uses Harris's corner detection method and SIFT (Scale-Invariant Feature Transform), FAST (Features from Accelerated Segment Test), or Algorithms such as SURF (Speeded Up Robust Features) may be used to extract points such as edges, edge intersections, and edge corners from the projected image Gi.

[0055] Figure 6 shows the projected image and the structural projected image. As shown in Figure 6, the structural projected image SGi is obtained by extracting the linear structures contained in the projected image Gi. The structure extraction unit 32 may derive a multi-level structural projected image SGi, or it may derive a binary structural projected image SGi by binarizing the multi-level structural projected image SGi.

[0056] Furthermore, the structure extraction unit 32 may use a machine learning-based learning model to extract at least one of the line structures and point structures contained in the projected image Gi.

[0057] The reconstruction unit 33 derives structural tomographic images for each of the multiple tomographic planes of the breast M by reconstructing multiple structural projection images SGi. Furthermore, the reconstruction unit 33 derives a tomographic image that emphasizes a desired tomographic plane of the breast M by reconstructing all or part of the multiple projection images Gi while correcting for positional displacement, as will be described later. In addition, the reconstruction unit 33 derives a tomographic image that emphasizes a desired tomographic plane of the breast M by reconstructing all or part of the multiple projection images Gi without correcting for the amount of positional displacement, if necessary.

[0058] Specifically, the reconstruction unit 33 reconstructs multiple structural projection images SGi using well-known back projection methods such as simple back projection or filtered back projection, and derives multiple structural tomographic images SDj (j=1~m) for each of the multiple tomographic planes of the breast M, as shown in Figure 7. The reconstruction unit 33 also reconstructs all or part of the multiple projection images Gi to derive multiple tomographic images Dj (j=1~m) for each of the multiple tomographic planes of the breast M. At this time, a three-dimensional coordinate position in the three-dimensional space including the breast M is set, and the pixel values ​​of the corresponding pixel positions in the multiple structural projection images SGi or the multiple projection images Gi are reconstructed for the set three-dimensional coordinate position, and the pixel values ​​of those coordinate positions are calculated. Furthermore, as will be described later, when the amount of positional displacement due to the movement of the breast M during tomosynthesis imaging is derived, the reconstruction unit 33 reconstructs the multiple projection images Gi while correcting for the positional displacement, thereby deriving a corrected tomographic image with corrected body movement.

[0059] The feature structure detection unit 34 detects at least one feature structure from a plurality of structural tomography images SDj. Feature structures include point structures and line structures included in the structural tomography images SDj. Figure 8 is a diagram illustrating the detection of feature structures. In this embodiment, point structures are detected as feature structures. Furthermore, the detection of feature structures from one structural tomography image SDk among the plurality of structural tomography images SDj will be described here. As shown in Figure 8, the structural tomography image SDk includes point structures such as calcifications E1-E3 and edge intersections E4, E5 such as blood vessel intersections on the tomographic plane of the breast M from which the structural tomography image SDk was acquired.

[0060] The feature structure detection unit 34 uses a known algorithm to detect point structures such as calcifications from the structural tomography image SDk as point structures, i.e., feature structures. Alternatively, algorithms such as Harris's corner detection method, SIFT, FAST, or SURF may be used to detect points such as edges, edge intersections, and edge corners included in the structural tomography image SDk as feature structures. For example, the feature structure detection unit 34 detects the point structure E1 included in the structural tomography image SDk shown in Figure 8 as feature structure F1. Point structures all have high-luminance pixel values, i.e., small pixel values, in the structural tomography image SDj. Therefore, in any of these methods, the detected feature structures will satisfy a specific threshold condition. Specifically, point structures are points in the structural tomography image SDj whose pixel values ​​are less than or equal to a predetermined threshold Th1. If the pixel values ​​in the structural tomography image SDj are larger as the luminance increases, then the detected feature structures are points in the structural tomography image SDj whose pixel values ​​are greater than or equal to a predetermined threshold.

[0061] As mentioned above, feature structures are not limited to point structures. Even when detecting line structures as feature structures, for example, a brightness level that satisfies a specific threshold condition can be used as a criterion, or known algorithms can be used as appropriate. Furthermore, a computer-aided diagnosis (CAD) algorithm may be used to detect feature structures.

[0062] For illustrative purposes, only one feature structure F1 is detected from a single structural tomography image (SDk) in this example; however, it is preferable to detect multiple feature structures. For example, all point structures, including point structures E1-E3 and intersections E4 and E5, included in the structural tomography image (SDk) shown in Figure 8, may be detected as feature structures. A feature structure may consist of only one pixel in the structural tomography image (SDk), or it may consist of multiple pixels representing the location of the feature structure. Furthermore, for illustrative purposes, feature structures are detected only from a single structural tomography image (SDk) in this example; however, in reality, multiple feature structures are expected to be detected from each of multiple structural tomography images (SDj).

[0063] The projection unit 35 projects the multiple projection images Gi onto the corresponding tomographic plane, which is the tomographic plane corresponding to the tomographic image in which the characteristic structure F1 was detected, based on the positional relationship between the radiation source position and the radiation detector 15 at the time of imaging for each of the multiple projection images Gi. The projection unit 35 then derives a tomographic plane projection image GTi corresponding to each of the multiple projection images Gi. The derivation of the tomographic plane projection image GTi will be described below. In this embodiment, since a characteristic structure is detected in each of the multiple tomographic images Dj, the multiple projection images Gi are projected onto each of the multiple tomographic planes Tj corresponding to the multiple tomographic images Dj, thereby deriving the tomographic plane projection image GTi.

[0064] Figure 9 is a diagram illustrating the projection of a projection image onto a corresponding tomographic plane. In Figure 9, the case where one projection image Gi acquired at the source position Si is projected onto one tomographic plane Tj in the breast M is described. In this embodiment, as shown in Figure 9, the pixel values ​​of the projection image Gi located on the straight line connecting the source position Si and the pixel positions on the projection image Gi are projected onto the position where the straight line intersects with the tomographic plane Tj.

[0065] The tomographic images derived from the projected image Gi and the tomographic plane Tj consist of multiple pixels discretely arranged in two dimensions at predetermined sampling intervals, with pixels positioned at grid points corresponding to these predetermined sampling intervals. In Figure 9, short line segments perpendicular to the projected image Gi and the tomographic plane Tj indicate the pixel demarcation points. Therefore, in Figure 9, the pixel position corresponding to the grid point is the central position of the pixel demarcation points.

[0066] Here, the relationship between the coordinates of the source position at source position Si (sxi, syi, szi), the coordinates of the pixel position Pi on the projected image Gi (pxi, pyi), and the coordinates of the projection position on the tomographic plane Tj (tx, ty, tz) is expressed by the following equation (1). In this embodiment, the z-axis is set perpendicular to the detection surface 15A of the radiation detector 15, the y-axis is set parallel to the direction in which the X-ray source 16 moves on the detection surface of the radiation detector 15, and the x-axis is set perpendicular to the y-axis. pxi=(tx×szi-sxi×tz) / (szi-tz) pyi=(ty×szi-syi×tz) / (szi-tz) (1)

[0067] Therefore, by setting pxi and pyi in equation (1) as the pixel positions of the projected image Gi and solving equation (1) for tx and ty, the projection position on the tomographic plane Tj onto which the pixel values ​​of the projected image Gi are projected can be calculated. Thus, by projecting the pixel values ​​of the projected image Gi onto the calculated projection position on the tomographic plane Tj, the tomographic plane projected image GTi can be derived.

[0068] In this case, the intersection of the line connecting the source position Si and the pixel position on the projected image Gi with the tomographic plane Tj may not be the same as the pixel position on the tomographic plane Tj. For example, as shown in Figure 10, the projected position (tx, ty, tz) on the tomographic plane Tj may be located between pixel positions O1 to O4 of the tomographic image Dj on the tomographic plane Tj. In this case, the pixel value of each pixel position can be calculated by performing an interpolation operation using the pixel values ​​of the projected image at multiple projected positions surrounding each pixel position O1 to O4.

[0069] For the interpolation operation, a linear interpolation operation can be used, which weights the pixel value of the projected image at a projection position according to the distance between the pixel position and multiple projection positions surrounding it. In addition, any other method can be used, such as a nonlinear bicubic interpolation operation using the pixel values ​​of more projection positions surrounding the pixel position, and a B-spline interpolation operation. Furthermore, in addition to the interpolation operation, the pixel value of the projection position closest to the pixel position may be used as the pixel value of that pixel position. In this way, the pixel values ​​at all pixel positions on the tomographic plane Tj can be obtained for the projected image Gi. In this embodiment, for each of the multiple projected images Gi, the pixel values ​​obtained at all pixel positions on the tomographic plane Tj in this way can be obtained. A tomographic projection image GTi with prime values ​​is derived. Therefore, for a single fault plane, the number of tomographic projection images GTi is equal to the number of projection images Gi.

[0070] The positional displacement amount derivation unit 36 ​​derives the positional displacement amount between multiple tomographic projection images GTi based on the movement of the breast M during tomosynthesis imaging. First, the positional displacement amount derivation unit 36 ​​sets a local region corresponding to the feature structure F1 as the region of interest for the multiple tomographic projection images GTi. Specifically, a local region of a predetermined size centered on the coordinate position of the feature structure F1 is set as the region of interest. Figure 11 is a diagram illustrating the setting of the region of interest. In Figure 11, for illustrative purposes, it is assumed that three projection images G1 to G3 are projected onto the tomographic plane Tj to derive tomographic projection images GT1 to GT3.

[0071] As shown in Figure 11, the displacement amount derivation unit 36 ​​sets a region of interest Rf0 centered on the coordinate position of feature structure F1 in the tomographic image Dj on the fault plane Tj. Then, in each of the tomographic projection images GT1 to GT3, regions of interest R1 to R3 corresponding to region of interest Rf0 are set. The dashed lines in Figure 11 indicate the boundary between regions of interest R1 to R3 and the other regions. Therefore, on the tomographic plane Tj, the positions of region of interest Rf0 and regions of interest R1 to R3 coincide. Figure 12 shows regions of interest R1 to R3 set in the tomographic projection images GT1 to GT3. Note that the motion can be as large as 2 mm. For this reason, in the case of a tomographic image or tomographic projection image where the size of one pixel is 100 μm square, regions of interest R1 to R3 can be, for example, a 50 × 50 pixel or 100 × 100 pixel area around feature structure F1.

[0072] Furthermore, the displacement amount derivation unit 36 ​​aligns the regions of interest R1 to R3. In this case, the alignment is performed using the regions of interest set in the reference tomographic projection image as a reference. In this embodiment, the other regions of interest are aligned using the regions of interest set in the tomographic projection image (referred to as the reference tomographic projection image) for the reference projection image (referred to as Gc) acquired at the source position Sc where the optical axis X0 of the X-rays from the X-ray source 16 is orthogonal to the radiation detector 15 as a reference.

[0073] Here, we assume that the region of interest R2 shown in Figure 12 is set as the reference tomographic projection image. In this case, the positional displacement amount derivation unit 36 ​​aligns the regions of interest R1 and R3 with respect to the region of interest R2 and derives a shift vector representing the direction and amount of movement of the regions of interest R1 and R3 with respect to the region of interest R2 as the positional displacement amount. Note that alignment means determining the direction and amount of movement of the regions of interest R1 and R3 with respect to the region of interest R2 within a predetermined search range so as to maximize the correlation between the regions of interest R1 and R3 with respect to the region of interest R2. Here, normalized cross-correlation can be used as the correlation. Also, since one of the reference tomographic projection images GTi is used as the reference, the number of shift vectors will be one less than the number of tomographic projection images. For example, if the number of tomographic projection images is 15, the number of shift vectors will be 14, and if the number of tomographic projection images is 3, the number of shift vectors will be 2.

[0074] Figure 13 shows images within the three regions of interest R1 to R3 when no body movement occurs while acquiring the projected images G1 to G3. In Figure 13, the central positions of regions of interest R1 to R3, i.e., the positions P1 to P3 corresponding to feature structure F1 in the tomographic projection images GT1 to GT3, are shown, and the image F2 of feature structure F1 contained in regions of interest R1 to R3 is indicated by a large circle. As shown in Figure 13, when no body movement occurs while acquiring the projected images G1 to G3, the positions P1 to P3 corresponding to feature structure F1 and the position of the image F2 of feature structure F1 coincide in all three regions of interest R1 to R3. Therefore, the shift vectors of regions of interest R1 and R3 with respect to region R2, i.e., the positional displacement amounts, are all 0.

[0075] Figure 14 shows the images within three regions of interest R1 to R3 when body movement occurs between acquiring projected images G2 and G3. In Figure 14, since no body movement occurs between acquiring projected images G1 and G2, the positions P1 and P2 corresponding to feature structures F1 in regions of interest R1 and R2 coincide with the position of image F2 of feature structures F1 contained in regions of interest R1 and R2. Therefore, the positional displacement of region R1 relative to region R2 is 0. On the other hand, since body movement occurs between acquiring projected images G2 and G3, the position P3 corresponding to feature structures F1 in region R3 does not coincide with the position of image F2 of feature structures F1 contained in region R3. Therefore, region R3 experiences a displacement amount and direction relative to region R2, and as a result, a shift vector V10 with magnitude and direction is derived as the positional displacement amount.

[0076] Furthermore, when deriving the displacement amount, the search range used for deriving the displacement amount may be changed according to at least one of the following: breast density of breast M, breast size, tomosynthesis imaging time, compression pressure of breast M during tomosynthesis imaging, and imaging direction of breast. Figure 15 is a diagram illustrating the change in the search range. As shown in Figure 15, two types of search ranges, a small search range H1 and a large search range H2, are set as the search ranges for regions of interest R1 and R3 with respect to the reference region of interest R2.

[0077] In this context, when breast density is low, the amount of fat in the breast M increases, which tends to lead to greater body movement during imaging. Similarly, when the breast M is large, body movement tends to increase during imaging. Furthermore, the longer the tomosynthesis imaging time, the greater the body movement tends to be. Also, when the compression pressure on the breast M is low, body movement tends to increase during imaging. In addition, when the imaging direction of the breast M is MLO (Medio-Lateral Oblique), body movement tends to be greater than when it is CC (Cranio-Caudal).

[0078] Therefore, it is preferable that the positional displacement amount derivation unit 36 ​​changes the search range when deriving the positional displacement amount based on input from the input device 25 of at least one piece of information regarding breast M, such as breast density, breast M size, tomosynthesis imaging time, compression pressure of breast M during tomosynthesis imaging, and imaging direction of breast M. Specifically, if there is a tendency for body movement to be large, a larger search range H2 shown in Figure 15 should be set. Conversely, if there is a tendency for body movement to be small, a smaller search range H1 shown in Figure 15 should be set.

[0079] In the above explanation, for the sake of explanation, the positional displacement amounts of multiple tomographic projection images GTi are derived only for one feature structure F1 detected in one tomographic plane Tj. However, in reality, as shown in Figure 16, the positional displacement amount derivation unit 36 ​​derives positional displacement amounts for multiple different feature structures F (10 feature structures represented here by black circles) in the three-dimensional space within the breast M represented by multiple tomographic images Dj. As a result, for tomographic projection images corresponding to projection images acquired while body motion is occurring, positional displacement amounts for multiple different feature structures F are derived. The positional displacement amount derivation unit 36 ​​interpolates the positional displacement amounts for multiple different feature structures F to the coordinate positions in the three-dimensional space from which the tomographic image Dj is derived. As a result, the positional displacement amount derivation unit 36 ​​derives the positional displacement amounts for reconstruction for all coordinate positions in the three-dimensional space from which the tomographic image is derived, for tomographic projection images acquired while body motion is occurring.

[0080] The reconstruction unit 33 then corrects the positional displacement amount derived in this way and reconstructs the projected image Gi to derive a corrected tomographic image Dhj in which the motion has been corrected. Specifically, if the reconstruction uses the back projection method, the pixels of the projected image Gi that have positional displacement are reconstructed by correcting the positional displacement based on the derived positional displacement amount so that they are projected to the positions where the corresponding pixels of other projected images are back-projected.

[0081] Alternatively, instead of deriving the displacement amount for multiple different feature structures F, one displacement amount may be derived from multiple different feature structures F. In this case, a region of interest is set for each of the multiple different feature structures F, and the displacement amount is derived by assuming that the entire region of interest moves in the same direction by the same amount. In this case, the displacement amount should be derived in such a way that the representative value (e.g., mean, median, and maximum) of the correlation for all regions of interest among the tomographic projection images from which the displacement amount is to be derived is maximized. Here, if the signal-to-noise ratio of individual feature structures F in the tomographic projection image is not very good, the accuracy of deriving the displacement amount will be poor. However, by deriving one displacement amount from multiple different feature structures F in this way, the accuracy of deriving the displacement amount can be improved even when the signal-to-noise ratio of individual feature structures F is not very good.

[0082] Alternatively, the three-dimensional space within the breast M, represented by multiple tomographic images Dj, may be divided into multiple three-dimensional regions, and a single positional displacement amount may be derived from multiple feature structures F in each region, similar to the method described above.

[0083] Furthermore, in this embodiment, the reconstruction unit 33 derives multiple tomographic images Dj by reconstructing multiple projection images Gi without correcting for positional displacement.

[0084] The display control unit 37 displays the derived corrected tomographic image on the display 24. Figure 17 shows the display screen of the corrected tomographic image. As shown in Figure 17, the display screen 40 displays the tomographic image Dj before motion correction and the corrected tomographic image Dhj after motion correction. The tomographic image Dj is labeled "Before Correction" 41 to indicate that it has not been corrected for motion. The corrected tomographic image Dhj is labeled "After Correction" 42 to indicate that it has been corrected for motion. Note that label 41 may be applied only to the tomographic image Dj, or label 42 may be applied only to the corrected tomographic image Dhj. Of course, it is also possible to display only the corrected tomographic image Dhj. In Figure 17, structures included in the tomographic image Dj before correction are shown to be blurred with dashed lines, and structures included in the corrected tomographic image Dhj are shown to be not blurred with solid lines.

[0085] Furthermore, it is preferable that the tomographic image Dj and the corrected tomographic image Dhj display the same cross-section. It is also preferable that the tomographic planes displayed in the tomographic image Dj and the corrected tomographic image Dhj are synchronized when switching the displayed tomographic plane based on instructions from the input device 25. In addition to the tomographic image Dj and the corrected tomographic image Dhj, a projected image Gi may also be displayed.

[0086] The operator can check the success or failure of motion correction by looking at the display screen 40. However, if the motion is too large, even if reconstruction is performed while correcting the amount of positional displacement to derive a tomographic image as in this embodiment, the motion may not be corrected accurately, and motion correction may fail. In such cases, the tomographic image Dj may have higher image quality than the corrected tomographic image Dhj due to the failure of motion correction. For this reason, the input device 25 may accept instructions on whether to save the tomographic image Dj or the corrected tomographic image Dhj, and save the instructed image to the storage 23 or an external storage device.

[0087] Next, the process performed in the first embodiment will be described. Figure 18 is a flowchart showing the process performed in the first embodiment. When the input device 25 receives an instruction from the operator to start processing, the image acquisition unit 31 acquires a plurality of projection images Gi derived by the console 2 causing the imaging device 10 to perform tomosynthesis imaging of the breast M (step ST1). Then, the structure extraction unit 32 derives a plurality of structural projection images SGi by extracting specific structures from each of the plurality of projection images Gi (step ST2).

[0088] Next, the reconstruction unit 33 derives a structural tomographic image SDj for each of the multiple tomographic planes of the breast M by reconstructing the multiple structural projection images SGi (step ST3). Then, the feature structure detection unit 34 detects at least one feature structure from the multiple structural tomographic images SDj (step ST4). Furthermore, the projection unit 35 projects the multiple projection images Gi onto the corresponding tomographic planes corresponding to the tomographic images in which the feature structure F1 was detected, based on the positional relationship between the radiation source position at the time of acquisition and the radiation detector 15 for each of the multiple projection images Gi, and derives a tomographic plane projection image GTi corresponding to each of the multiple projection images Gi (step ST5).

[0089] Then, the positional displacement amount derivation unit 36 ​​derives the positional displacement amount between multiple tomographic projection images GTi (step ST6). Furthermore, the reconstruction unit 33 derives a corrected tomographic image Dhj by reconstructing the multiple projection images Gi while correcting the positional displacement (step ST7). Then, the display control unit 37 displays the corrected tomographic image Dhj on the display 24 (step ST8), and the process ends. The derived corrected tomographic image Dhj is transmitted to the image storage system 3 and stored.

[0090] In tomosynthesis imaging, multiple scans are performed. To reduce radiation exposure, the amount of radiation irradiated in each scan is low, resulting in a noisy projected image Gi. Consequently, structures such as calcifications within the breast M may have reduced contrast depending on how the structures overlap, and may be obscured by noise in the projected image Gi. Therefore, when feature structures are detected from tomographic images derived by reconstructing multiple projected images Gi, it may become difficult to accurately match the feature structures detected from the tomographic images with the structures corresponding to the feature structures contained in the projected images. As a result, it may be difficult to accurately correct the positional displacement of the projected image Gi using the feature structures.

[0091] In the first embodiment, a structural projection image SGi is derived by extracting specific structures such as line structures and point structures from the projection image, a structural tomographic image SDj is derived by reconstructing the structural projection image SGi, and feature structures are detected from the structural tomographic image SDj. Here, since the structural tomographic image SDj is derived from the structural projection image SGi, it is guaranteed that structures corresponding to the feature structures detected from the structural tomographic image SDj are included in the structural projection image SGi and further in the projection image Gi. Therefore, according to the first embodiment, the amount of positional displacement between multiple projection images Gi can be appropriately derived using the detected feature structures, and as a result, according to this embodiment, a high-quality corrected tomographic image Dhj with reduced effects of body motion can be obtained.

[0092] Furthermore, in the first embodiment, characteristic structures are detected from multiple structural tomographic images SDj, rather than from the projection image Gi or the tomographic projection image GTi. Here, the structural tomographic image SDj includes only structures contained in the corresponding tomographic plane Tj. Therefore, structures in other tomographic planes, such as those contained in the projection image Gi, are not included in the structural tomographic image SDj. Consequently, according to the first embodiment, characteristic structures can be detected with high accuracy without being affected by structures in other tomographic planes. Thus, the amount of positional displacement between multiple projection images Gi can be appropriately derived, and as a result, according to this embodiment, a high-quality corrected tomographic image Dhj with reduced effects of body motion can be obtained.

[0093] Next, a second embodiment of the present disclosure will be described. The configuration of the image processing apparatus according to the second embodiment is the same as that of the image processing apparatus according to the first embodiment shown in Figure 4, and only the processing performed is different, so a detailed explanation of the apparatus will be omitted here. In the first embodiment described above, the amount of positional displacement is derived between tomographic projection images GTi. In the second embodiment, a region of interest Rf0 centered on the coordinate position of feature structure F1 is set in the structural tomographic image SDj, and the amount of positional displacement of the region of interest Ri set in the tomographic projection image GTi with respect to the set region of interest Rf0 is derived as a provisional amount of positional displacement. The difference from the first embodiment is that the amount of positional displacement between multiple tomographic projection images GTi is derived based on the derived provisional amount of positional displacement. Note that the region of interest Ri set in multiple tomographic projection images GTi corresponds to the first local region, and the region of interest Rf0 set in the structural tomographic image SDj corresponds to the second local region.

[0094] Figure 19 is a diagram illustrating the derivation of the displacement amount in the second embodiment. Note that the regions of interest Rf0 and R1-R3 in Figure 19 are the same as the regions of interest Rf0 and R1-R3 shown in Figure 11 and above. In the second embodiment, the displacement amount derivation unit 36 ​​first uses the region of interest Rf0 set in the structural tomographic image SDj as a reference and derives the displacement amounts of regions R1-R3 set in the tomographic projection images GTi (GT1-GT3 in Figure 19 (none of which are shown)) relative to the region of interest Rf0 as provisional displacement amounts. If no body movement occurs while acquiring the projection images G1-G3, the positions P1-P3 corresponding to the feature structure F1 and the position F2 of the image of the feature structure F1 coincide in all three regions of interest R1-R3. For this reason, the shift vectors of regions R1-R3 relative to the region of interest Rf0 (hereinafter referred to as Vf1, Vf2, Vf3), i.e., the provisional displacement amounts, are all 0.

[0095] Figure 20 shows the images within the three regions of interest R1 to R3 when body movement occurs between acquiring projected images G2 and G3. In Figure 20, since no body movement occurs between acquiring projected images G1 and G2, the positions P1 and P2 corresponding to feature structures F1 in regions of interest R1 and R2 coincide with the position of image F2 of feature structures F1 contained in regions of interest R1 and R2. Therefore, the positional displacement of regions of interest R1 and R2 relative to region of interest Rf0 is 0. On the other hand, since body movement occurs between acquiring projected images G2 and G3, the position P3 corresponding to feature structures F1 in region of interest R3 does not coincide with the position of image F2 of feature structures F1 contained in region of interest R3. Therefore, region R3 experiences both a displacement and a direction of displacement relative to region of interest Rf0. Therefore, the shift vectors Vf1 and Vf2 of regions of interest R1 and R2 relative to region of interest Rf0, i.e., the hypothetical positional displacement amounts, are 0, but the shift vector Vf3 of region of interest R3 relative to region of interest Rf0, i.e., the hypothetical positional displacement amount, will have a value.

[0096] In the second embodiment, the positional displacement amount derivation unit 36 ​​derives the positional displacement amount between tomographic projection images GTi based on a provisional positional displacement amount. Specifically, similar to the first embodiment, the positional displacement amount is derived using the projection image acquired at the reference source position Sc, where the optical axis X0 of the X-rays from the X-ray source 16 is orthogonal to the radiation detector 15, as a reference. Here, assuming that projection image G2 is the reference tomographic projection image, the positional displacement amount derivation unit 36 ​​derives the positional displacement amount between tomographic projection image GT1 and tomographic projection image GT2 from the difference value Vf1-Vf2 of the shift vectors Vf1 and Vf2 of regions of interest R1 and R2 with respect to region of interest Rf0. Furthermore, the positional displacement amount derivation unit 36 ​​derives the positional displacement amount between tomographic projection image GT3 and tomographic projection image GT2 from the difference value Vf3-Vf2 of the shift vectors Vf3 and Vf2 of regions of interest R3 and R2 with respect to region of interest Rf0.

[0097] Thus, in the second embodiment, a provisional displacement amount is derived between the region of interest Rf0 set in the structural tomographic image SDj and the regions of interest R1 to R3 set in the tomographic projection image GTi, and the displacement amount between the tomographic projection images GTi is derived based on the provisional displacement amount. Here, since the region of interest Rf0 is set in the structural tomographic image SDj, unlike the projection image Gi, it contains only structures on the fault plane from which the structural tomographic image SDj was acquired. Therefore, according to the second embodiment, the influence of structures included in fault planes other than the fault plane in which the characteristic structure is set is reduced, and the displacement amount is derived. Therefore, according to the second embodiment, the influence of structures in other fault planes is further reduced, and the displacement amount between multiple projection images Gi can be derived with high accuracy. As a result, according to the second embodiment, a high-resolution corrected tomographic image Dhj with reduced effects of motion can be obtained.

[0098] In the second embodiment, as in the first embodiment, the search range for deriving the displacement amount may be changed according to at least one of the following: the glandular density of the breast M, the size of the breast M, the imaging time for tomosynthesis imaging, the compression pressure of the breast M during tomosynthesis imaging, and the imaging direction of the breast M.

[0099] Furthermore, in the second embodiment, the shift vectors Vf1 to Vf3 of regions of interest R1 to R3 with respect to region of interest Rf0 are derived as provisional positional displacement amounts. However, as shown in Figure 21, a surrounding region Ra0 smaller than the region of interest Rf0 may be set around the feature structure F1 in the region of interest Rf0, and the shift vectors may be derived based on the surrounding region Ra0. In this case, the shift vectors may be derived using only the surrounding region Ra0. Also, when deriving the correlation between regions of interest R1 to R3, the surrounding region Ra0 may be given a greater weight than the regions other than the surrounding region Ra0 in the regions of interest R1 to R3.

[0100] Furthermore, in the second embodiment described above, the region of interest Rf0 is set in the structural tomographic image SDj, but the structural tomographic image to be derived may be different for each tomographic projection image GTi from which the provisional displacement amount is derived. Specifically, it is preferable to derive the structural tomographic image excluding the target projection image corresponding to the target tomographic projection image from which the provisional displacement amount is to be derived. This will be described below as the third embodiment.

[0101] Figure 22 schematically shows the process performed in the third embodiment. Here, we will explain the case in which, in the tomographic plane Tj of the breast M, projection image G1 is used as the target projection image and the tomographic plane projection image GT1 is used as the target tomographic plane projection image, and a provisional positional displacement amount for projection image G1 is derived. In this case, the reconstruction unit 33 reconstructs structural projection images SG2 to SG15 other than the structural projection image SG1 derived from the target projection image G1 in the tomographic plane Tj to derive a structural tomographic image (denoted as SDj_1). Then, the feature structure detection unit 34 detects feature structures from the structural tomography image SDj_1, the projection unit 35 derives tomographic projection images GT1 to GT15 from projection images G1 to G15, and the positional displacement amount derivement unit 36 ​​sets a region of interest Rf0_1 in the structural tomography image SDj_1 and derives the shift vector Vf1 of the region of interest R1 set in the tomographic projection image GT1 relative to the region of interest Rf0_1 as a provisional positional displacement amount.

[0102] When deriving a provisional displacement amount for projection image G2, the reconstruction unit 33 reconstructs structural projection images SG1, SG3 to SG15 (excluding structural projection image SG2 derived from projection image G2) to derive a structural tomographic image (denoted as SDj_2). Then, the feature structure detection unit 34 detects feature structures from the structural tomographic image SDj_2, the projection unit 35 derives tomographic plane projection images GT1 to GT15 from projection images G1 to G15, and the displacement amount deriving unit 36 ​​sets a region of interest Rf0_2 in the structural tomographic image SDj_2 and projects the tomographic plane onto the region of interest Rf0_2. The shift vector Vf2 of the region of interest R2 set in the shadow image GT2 is derived as a provisional positional displacement.

[0103] Then, the target tomographic projection image is sequentially changed to derive a provisional displacement amount for all tomographic projection images GTi, and from the provisional displacement amount, the displacement amount between tomographic projection images GTi is derived in the same manner as in the second embodiment described above.

[0104] Thus, according to the third embodiment, a structural tomographic image unaffected by the target projection image can be obtained. This allows for the derivation of a provisional displacement amount. Therefore, the provisional displacement amount can be derived with greater accuracy, and as a result, the displacement amount can be derived with greater accuracy.

[0105] In the third embodiment, when reconstructing the structural tomography image excluding the structural projection image for the target projection image, the structural tomography image may be derived by subtracting the corresponding pixel value Gp of the structural projection image SGi derived from the target projection image Gi from the pixel value Dp of each pixel of the structural tomography image SDj derived by reconstructing all structural projection images SGi, as shown in equation (2) below, and then multiplying the subtracted pixel value by n / (n-1). Although the method in equation (2) is a simple method, it can reduce the amount of computation required to derive the structural tomography image excluding the structural projection image for the target projection image, and thus the processing for deriving the provisional positional displacement amount can be performed at high speed. Structural tomographic image excluding the target projection image = (Dp - Gp) × n / (n-1) (2)

[0106] Next, a fourth embodiment will be described. Note that the configuration of the image processing apparatus in the fourth embodiment is the same as that of the image processing apparatus in the first embodiment shown in Figure 4, and only the processing performed is different, so a detailed explanation of the apparatus will be omitted here. The fourth embodiment differs from the first embodiment in that the structural tomographic image SDj is updated by reconstructing the structural projection image SGi while correcting the amount of positional displacement, the updated feature structure is detected from the updated structural tomographic image using the updated threshold, the amount of positional displacement is updated using the updated feature structure, and the updating of the structural tomographic image, detection of the updated feature structure using the updated threshold, and updating of the amount of positional displacement are repeated until the amount of positional displacement converges.

[0107] Figure 23 is a flowchart showing the process performed in the fourth embodiment. Note that in Figure 23, the processes from steps ST11 to ST16 are the same as the processes from steps ST1 to ST6 shown in Figure 18, so a detailed explanation is omitted here. When the positional displacement amount is derived in step ST16, the positional displacement amount derivation unit 36 ​​determines whether the positional displacement amount has converged (step ST17). The determination of whether the positional displacement amount has converged can be made by determining whether the positional displacement amount derived for each tomographic projection image GTi is less than or equal to a predetermined threshold Th10. The threshold Th10 should be set to a value such that it can be said that there is no effect of body motion on the tomographic image even without further correction of the positional displacement amount. Alternatively, the determination of whether the positional displacement amount has converged can be made by determining whether a representative value such as the average value of the positional displacement amounts derived for multiple tomographic projection images GTi is less than or equal to the threshold Th10.

[0108] If step ST17 is rejected, the reconstruction unit 33 updates the structural tomography image by reconstructing multiple structural projection images SGi while correcting the positional displacement (step ST18). Then, it returns to the processing of step ST14 and performs the processing of steps ST14 to ST17. In this case, during the processing of step ST14, the feature structure detection unit 34 updates the threshold Th1 used when detecting the feature structure in the first processing of step ST14. In this embodiment, since the pixel value of the structural tomography image SDj becomes smaller as the brightness increases, the updated feature structure is detected using the updated threshold Th2, which is smaller than the threshold Th1 used in the first processing of step ST14.

[0109] Furthermore, in step ST15, the projection unit 35 projects the multiple projection images Gi onto the corresponding tomographic planes corresponding to the tomographic images in which the updated feature structure F1 was detected, based on the positional relationship between the source position at the time of acquisition and the radiation detector 15 for each of the multiple projection images Gi, thereby deriving updated tomographic plane projection images GTi corresponding to each of the multiple projection images Gi. Furthermore, in step ST16, the positional displacement amount derivation unit 36 ​​derives the updated positional displacement amount between the multiple updated tomographic plane projection images GTi.

[0110] Furthermore, if the pixel values ​​of the structural tomographic image SDj are larger when the brightness is high, the updated feature structure is detected using an updated threshold Th2, which is larger than the threshold Th1 used in the first step ST14.

[0111] If step ST17 is rejected, steps ST18 and ST14 to ST16 are repeated until step ST17 is confirmed. During this time, the feature structure detection unit 34 detects feature structures from the updated structural tomography image SDj using the updated threshold.

[0112] If step ST17 is confirmed, the reconstruction unit 33 derives a corrected tomographic image Dhj by reconstructing multiple projection images Gi while correcting the updated positional displacement (step ST19). Then, the display control unit 37 displays the corrected tomographic image Dhj on the display 24 (step ST20), and the process ends. The derived corrected tomographic image Dhj is transmitted to the image storage system 3 and stored.

[0113] Thus, in the fourth embodiment, the structural tomographic image SDj is updated by reconstructing the structural projection image SGi while correcting the amount of positional displacement, updated feature structures are detected from the updated structural tomographic image using the updated threshold, the amount of positional displacement is updated using the updated feature structures, and the updating of the structural tomographic image, detection of updated feature structures using the updated threshold, and updating of the amount of positional displacement are repeated until the amount of positional displacement converges. As a result, positional displacement caused by body movement can be removed more effectively, and as a result, higher quality tomographic images can be obtained.

[0114] In addition, in the second and third embodiments described above, the process of updating the positional displacement amount may be repeated until the positional displacement amount converges, similar to the fourth embodiment.

[0115] Furthermore, in the fourth embodiment described above, the structural tomography image is updated while updating the threshold, the updated feature structure is detected using the updated threshold, and the positional displacement is updated repeatedly until the positional displacement amount converges, but the embodiment is not limited to this. The structural tomography image may be updated, the updated feature structure is detected, and the positional displacement is updated repeatedly without updating the threshold.

[0116] Furthermore, in the fourth embodiment described above, the process of updating the positional displacement amount is repeated until the positional displacement amount converges, but this is not limited to this. The process of updating the positional displacement amount may be repeated a predetermined number of times.

[0117] Furthermore, in each of the above embodiments, the positional displacement amount derived by the positional displacement amount derivation unit 36 ​​may be compared with a predetermined threshold, and the tomographic image may be reconstructed while correcting the positional displacement amount only if the positional displacement amount exceeds the threshold. The threshold should be set to a value such that the tomographic image is not affected by body motion even without correcting the positional displacement amount. In this case, as shown in Figure 24, a warning display 45 may be displayed on the display 24 to notify that the body motion has exceeded the threshold. The operator can indicate whether or not to perform body motion correction by selecting YES or NO in the warning display 45.

[0118] Furthermore, in each of the above embodiments, in order to facilitate the derivation of the displacement amount and the provisional displacement amount, a region of interest is set in the structural tomographic image SDj and the tomographic projection image GTi, and the direction and amount of movement of the region of interest are derived as a shift vector, i.e., the displacement amount and the provisional displacement amount, but the embodiment is not limited to this. The displacement amount may also be derived without setting a region of interest.

[0119] Next, a fifth embodiment of the present disclosure will be described. Figure 25 shows the functional configuration of the image processing apparatus according to the fifth embodiment. In Figure 25, components similar to those in Figure 4 are given the same reference numerals as in Figure 4, and detailed explanations are omitted here. The image processing apparatus 4A according to the fifth embodiment further includes a focal plane determination unit 38 that determines whether the corresponding tomographic plane corresponding to the structural tomographic image in which each of the multiple feature structures F is detected is a focal plane, and the positional displacement amount derivation unit 36 ​​derives the positional displacement amount at the corresponding tomographic plane determined to be a focal plane, which is different from the first embodiment. Although the processing according to the fifth embodiment is also applicable to the second to fourth embodiments, only the case where it is applied to the first embodiment will be described here.

[0120] In tomosynthesis imaging, tomographic images acquired by tomosynthesis show that structures are not reflected in tomographic images that contain them. This is called a ripple artifact. Figure 26 is a diagram illustrating ripple artifacts. As shown in Figure 26, if a structure 48 is included in tomographic image D3, then ripple artifacts of structure 48 will be included in the tomographic images corresponding to the upper and lower tomographic planes of tomographic image D3. The ripple artifact becomes more widespread and blurred the further it is from the tomographic plane containing structure 48. The extent of the ripple artifact corresponds to the extent of movement of the X-ray source 16. Ripple artifacts also occur in the structural tomographic image SDj.

[0121] Here, if the feature structure F detected by the feature structure detection unit 34 from the structural tomographic image SDj of the corresponding tomographic plane is a ripple artifact, the feature structure F will be blurred and spread over a wide area. Therefore, if such a feature structure F is used, it will not be possible to accurately derive the amount of positional displacement.

[0122] Therefore, in the fifth embodiment, the focal plane determination unit 38 determines whether the corresponding tomographic plane for the structural tomographic image SDj in which the feature structure F is detected is a focal plane. For the corresponding tomographic plane determined to be a focal plane, the projection unit 35 derives a tomographic plane projection image GTi, and the positional displacement amount derivation unit 36 ​​derives the positional displacement amount. Specifically, the positional displacement amount is derived using the feature structure detected from the structural tomographic image of the corresponding tomographic plane determined to be a focal plane. The determination of whether or not it is a focal plane will be explained below.

[0123] The focal plane discrimination unit 38 derives corresponding points for the feature structure detected by the feature structure detection unit 34 in multiple structural tomography images SDj. Figure 27 is a diagram illustrating the derivation of the corresponding points. As shown in Figure 27, if a feature structure F3 is detected in a certain structural tomography image SDk, the positional displacement amount derivation unit 36 ​​derives corresponding points C1, C2, C3, C4… in multiple structural tomography images located in the thickness direction of structural tomography image SDk that correspond to the feature structure F3. In the following explanation, the reference symbol for the corresponding points will be C. The derivation of the corresponding points C can be done by aligning the region of interest containing the feature structure F3 with structural tomography images other than structural tomography image SDk. The focal plane discrimination unit 38 then plots the pixel values ​​of the feature structure F3 and the corresponding points C in the order in which the tomography planes are arranged. Figure 28 is a diagram showing the result of plotting the pixel values ​​of the feature structure and the corresponding points. As shown in Figure 28, the feature structure and the corresponding pixel values ​​change due to the effect of ripple artifacts, resulting in a minimum value in the feature structure. Here, if feature structure F3 is at the focal plane, it is not blurred and has high brightness, i.e., a small pixel value. On the other hand, if feature structure F3 is not at the focal plane, it is a ripple artifact, so the pixel value is blurred and the pixel value is larger than the minimum value.

[0124] Therefore, the focal plane discrimination unit 38 determines that the tomographic plane where the feature structure F3 was detected is the focal plane if, in the result of plotting the feature structure F3 and the pixel values ​​of the corresponding points, the position of the tomographic plane where the feature structure F3 was detected is the position P0 shown in Figure 28, which has the minimum pixel value. On the other hand, if the tomographic plane where the feature structure F3 was detected is at a position P1, as shown in Figure 28, where the pixel value is not the minimum, then the corresponding tomographic plane where the feature structure F3 was detected is determined not to be the focal plane.

[0125] The projection unit 35 derives a tomographic projection image GTi in the same manner as in each of the embodiments described above, but only for the corresponding tomographic plane that has been determined to be the focal plane. The positional displacement amount derives the positional displacement amount of the tomographic projection image GTi in the corresponding tomographic plane that has been determined to be the focal plane. That is, the positional displacement amount derives the positional displacement amount of the tomographic projection image GTi using the characteristic structure detected in the corresponding tomographic plane that has been determined to be the focal plane.

[0126] Next, the process performed in the fifth embodiment will be described. Figure 29 is a flowchart showing the process performed in the fifth embodiment. Note that the processes in steps ST21 to ST24 in Figure 29 are the same as the processes in steps ST1 to ST4 in Figure 18, so a detailed explanation is omitted here. In the fifth embodiment, it is assumed that multiple feature structures have been detected.

[0127] When the feature structure detection unit 34 detects multiple feature structures, the focal plane determination unit 38 determines whether the corresponding tomographic plane that corresponds to the structural tomographic image in which each of the multiple feature structures detected by the feature structure detection unit 34 is a focal plane (focal plane determination; step ST25). Then, the projection unit 35 derives a tomographic plane projection image GTi for the corresponding tomographic plane that has been determined to be a focal plane (step ST26), and the positional displacement amount derivation unit 36 ​​derives the positional displacement amount using the feature structures detected in the structural tomographic image of the corresponding tomographic plane that has been determined to be a focal plane (step ST27).

[0128] Furthermore, the reconstruction unit 33 derives a corrected tomographic image Dhj by reconstructing multiple projection images Gi while correcting the positional displacement (step ST28). Then, the display control unit 37 displays the corrected tomographic image Dhj on the display 24 (step ST29), and the process ends. The derived corrected tomographic image Dhj is transmitted to the image storage system 3 and stored.

[0129] Thus, in the fifth embodiment, the amount of displacement is derived in the corresponding tomographic plane that is determined to be the focal plane. Therefore, the amount of displacement can be derived with high accuracy without being affected by ripple artifacts, and as a result, a corrected tomographic image Dhj with accurately corrected displacement can be derived.

[0130] In the fifth embodiment described above, the determination of whether a corresponding tomographic plane is a focal plane is made using the plot results of the feature structure and the pixel values ​​of the corresponding points, but the determination of whether or not it is a focal plane is not limited to this. The difference in contrast with surrounding pixels is greater for feature structures than for ripple artifacts. For this reason, the contrast between the feature structure and the surrounding pixels at the corresponding points can be derived, and if the contrast for the feature structure is the maximum, it can be determined that the corresponding tomographic plane where the feature structure was detected is a focal plane. In addition, the pixel values ​​at the positions corresponding to the feature structure in the projected image show little variation between projected images if the feature structure is on the focal plane, but if the feature structure is not on the focal plane, it may represent a structure other than the structure corresponding to the feature structure on the projected image, resulting in greater variation between projected images. For this reason, the variance of the pixel values ​​corresponding to the feature structure between projected images Gi can be derived, and if the variance is less than or equal to a predetermined threshold, it can be determined that the corresponding tomographic plane where the feature structure was detected is a focal plane. Furthermore, the focal plane discrimination unit 38 may be equipped with a discriminator that has been machine-trained to output a discrimination result indicating whether or not the corresponding fault plane from which the feature structure was detected is a focal plane, when the feature structure and the pixel values ​​of its surroundings are input.

[0131] Next, a sixth embodiment of the present disclosure will be described. Figure 30 shows the functional configuration of the image processing apparatus according to the sixth embodiment. In Figure 30, components similar to those in Figure 4 are given the same reference numerals as in Figure 4, and detailed explanations are omitted here. The image processing apparatus 4B according to the sixth embodiment differs from the first embodiment in that it further includes a positional displacement amount determination unit 39 that performs image quality evaluation of the region of interest, including the feature structure in the corrected tomographic image Dhj, and determines whether the derived positional displacement amount is appropriate or inappropriate based on the image quality evaluation result. Although the processing according to the sixth embodiment is also applicable to the second to fifth embodiments, only the case where it is applied to the first embodiment will be described here.

[0132] The positional displacement determination unit 39 sets regions of interest Rh1 and Rh2 centered on the coordinate positions of multiple (in this case, two) feature structures F4 and F5 included in the corrected tomographic image Dhj, as shown in Figure 31, in order to evaluate image quality. Then, it derives a high-frequency image by extracting high-frequency components in each of the regions of interest Rh1 and Rh2. The extraction of high-frequency components can be performed, for example, by filtering with a Laplacian filter to derive a second-order derivative image, but is not limited to this. Furthermore, the positional displacement determination unit 39 derives the magnitude of the high-frequency components in the regions of interest Rh1 and Rh2. The magnitude of the high-frequency components can be derived by the sum of the squares of the pixel values ​​of the high-frequency image, but is not limited to this. Finally, the positional displacement determination unit 39 derives the sum of the magnitudes of the high-frequency components for all regions of interest Rh1 and Rh2.

[0133] Here, if the displacement amount is appropriately derived and the displacement correction is performed appropriately, the corrected tomographic image Dhj will have less image blur and more high-frequency components. On the other hand, if the derived displacement amount is not appropriate and the displacement correction is inappropriate, the corrected tomographic image Dhj will have more image blur and fewer high-frequency components. For this reason, in the sixth embodiment, the displacement amount determination unit 39 evaluates image quality based on the magnitude of the high-frequency components. That is, the displacement amount determination unit 39 determines whether the sum of the magnitudes of the high-frequency components for all regions of interest Rh1 and Rh2 derived as described above is greater than or equal to a predetermined threshold Th20. If the sum is greater than or equal to the threshold Th20, the displacement amount determination unit 39 determines that the displacement amount is appropriate, and if the sum is less than the threshold Th20, the displacement amount determination unit 39 determines that the displacement amount is inappropriate. If the positional displacement determination unit 39 determines that the positional displacement is inappropriate, the reconstruction unit 33 reconstructs the multiple projection images Gi without correcting the positional displacement to derive a tomographic image Dj. Then, the display control unit 37 displays the uncorrected tomographic image Dj on the display 24 instead of the corrected tomographic image Dhj. In this case, the uncorrected tomographic image Dj is transmitted to the external storage device instead of the corrected tomographic image Dhj.

[0134] Next, the process performed in the sixth embodiment will be described. Figure 32 is a flowchart of the process performed in the sixth embodiment. Note that the processes in steps ST31 to ST37 in Figure 32 are the same as the processes in steps ST1 to ST7 in Figure 18, so a detailed explanation is omitted here. When the reconstruction unit 33 derives the corrected tomographic image Dhj, the positional displacement amount determination unit 39 evaluates the image quality of the region of interest including the feature structure in the corrected tomographic image Dhj, and determines whether the derived positional displacement amount is appropriate based on the image quality evaluation result (step ST38).

[0135] If the amount of misalignment is appropriate, the display control unit 37 displays the corrected tomographic image Dhj on the display 24 (step ST39) and terminates the process. The derived corrected tomographic image Dhj is then transmitted to the image storage system 3 and stored. On the other hand, if the amount of misalignment is inappropriate, the reconstruction unit 33 reconstructs multiple projection images Gi without correcting the amount of misalignment to derive a tomographic image Dj (step ST40). Then, the display control unit 37 displays the tomographic image Dj on the display 24 (step ST41) and terminates the process. In this case, The tomographic image Dj is transmitted to and stored in the image storage system 3.

[0136] Here, when the displacement amount is derived by the displacement amount derivation unit 36, it may not be possible to derive an appropriate displacement amount due to the influence of structures other than the characteristic structure. In the sixth embodiment, the image quality of the corrected tomographic image Dhj is evaluated, and it is determined whether the displacement amount is appropriate or inappropriate based on the image quality evaluation result. Therefore, it is possible to appropriately determine whether the derived displacement amount is appropriate or inappropriate. Furthermore, if it is determined that the displacement amount is inappropriate, the uncorrected tomographic image Dj is displayed or saved, thereby reducing the possibility of an incorrect diagnosis being made based on a corrected tomographic image Dhj derived based on an inappropriate displacement amount.

[0137] In the sixth embodiment described above, image quality evaluation is performed based on the magnitude of the high-frequency components of the region of interest set in the corrected tomographic image Dhj, but the system is not limited to this. The reconstruction unit 33 may derive multiple tomographic images Dj by reconstructing multiple projection images Gi without performing positional displacement correction, and the positional displacement amount determination unit 39 may further evaluate the image quality of the region of interest, including the feature structure in the tomographic image Dj, compare the image quality evaluation result for the corrected tomographic image Dhj with the image quality evaluation result for the tomographic image Dj, and determine the tomographic image with the higher image quality evaluation as the final tomographic image. Here, the final tomographic image is the tomographic image that is displayed on the display 24 or transmitted to an external device and saved.

[0138] In addition, in the fifth and sixth embodiments described above, the amount of positional displacement may be repeatedly updated, similar to the fourth embodiment.

[0139] Furthermore, in the fifth and sixth embodiments described above, the positional displacement amount derived by the positional displacement amount derivation unit 36 ​​may be compared with a predetermined threshold value, and only if the positional displacement amount exceeds the threshold value, the tomographic image may be reconstructed while correcting the positional displacement amount.

[0140] Next, a seventh embodiment of the present disclosure will be described. Figure 33 shows the functional configuration of the image processing apparatus according to the seventh embodiment. In Figure 33, components similar to those in Figure 4 are given the same reference numerals as in Figure 4, and detailed explanations are omitted here. The image processing apparatus 4C according to the seventh embodiment includes an evaluation function derivation unit 50 that derives an evaluation function for evaluating the image quality of a region of interest including feature structures in a corrected tomographic image Dhj, and differs from the first embodiment in that the positional displacement amount derivation unit 36 ​​derives a positional displacement amount that optimizes the evaluation function. Although the processing according to the seventh embodiment is also applicable to the second to fifth embodiments, only the case where it is applied to the first embodiment will be described here.

[0141] In the seventh embodiment, the evaluation function derivation unit 50 derives a high-frequency image for the region of interest corresponding to the feature structure F set by the position displacement amount derivation unit 36 ​​for the tomographic projection image GTi. The high-frequency image can be derived by performing filtering processing using a Laplacian filter or the like to derive a second-order derivative image, similar to the position displacement amount determination unit 39 in the sixth embodiment. Let qkl be the pixel values ​​of the high-frequency image within the derived region of interest. k represents the k-th projection image, and l represents the number of pixels in the region of interest.

[0142] Here, let Wk be the transformation matrix for correcting the displacement, and let θk be the transformation parameter in the transformation matrix. The transformation parameter θk corresponds to the displacement. In this case, the image quality evaluation value in the region of interest corresponding to the feature structure F in the corrected tomographic image Dhj can be considered as the sum of the sizes of the high-frequency images for the region of interest after displacement correction in each of the projected images Gi. By deriving the transformation parameter θk, i.e., the displacement, such that this sum is maximized, a corrected tomographic image Dhj with appropriately corrected displacement can be obtained. This can be derived.

[0143] Therefore, the evaluation function derivation unit 50 derives the evaluation function shown in equation (3) below. The evaluation function Ec shown in equation (3) is an evaluation function Ec that finds the transformation parameter θk that minimizes the value in the parentheses on the right side with a minus sign added in order to maximize the above addition result. Note that the evaluation function shown in equation (3) has multiple local solutions. Therefore, constraints are placed on the range and mean value of the transformation parameter θk. For example, a constraint is placed so that the mean of the transformation parameter θk for all projected images is 0. More specifically, if the transformation parameter θk is a translation vector representing translation, a constraint is placed so that the mean value of the translation vector for all projected images Gi is 0. Then, in the seventh embodiment, the positional displacement amount derivation unit 36 ​​derives the transformation parameter θk that minimizes the evaluation function Ec shown in equation (3) below, i.e., the positional displacement amount.

[0144]

number

[0145] Thus, in the seventh embodiment, an evaluation function derivation unit 50 is provided for deriving an evaluation function for evaluating the image quality of a region of interest, including feature structures, in the corrected tomographic image Dhj, and a positional displacement amount derivation unit 36 ​​derives a positional displacement amount that optimizes the evaluation function. Therefore, the possibility of making an incorrect diagnosis based on a corrected tomographic image Dhj derived based on an inappropriate positional displacement amount can be reduced.

[0146] In each of the above embodiments, a region of interest is set in the tomographic image Dj and the tomographic projection image GTi to facilitate the derivation of the displacement amount and the provisional displacement amount, and the direction and amount of movement of the region of interest are derived as a shift vector, i.e., the displacement amount and the provisional displacement amount. However, the embodiment is not limited to this. The displacement amount may be derived without setting a region of interest.

[0147] Furthermore, in each of the above embodiments, the projection unit 35 derives the tomographic projection image GTi, and the positional displacement amount derives the positional displacement amount between the tomographic projection images GTi by the positional displacement amount derives by the positional displacement amount derives by the positional displacement amount derives by the positional displacement amount derives by the positional displacement amount derives by the positional displacement amount derives by the positional relationship of the projection images Gi in the corresponding tomographic plane corresponding to the tomographic image in which the feature structure F was detected.

[0148] Furthermore, although the subject is the breast M in each of the above embodiments, it is not limited to this, and any part of the human body, such as the chest or abdomen, may be used as the subject.

[0149] Furthermore, in each of the above embodiments, the hardware structure of the Processing Unit that performs various processes such as the image acquisition unit 31, structure extraction unit 32, reconstruction unit 33, feature structure detection unit 34, projection unit 35, positional displacement amount derivation unit 36, display control unit 37, focal plane discrimination unit 38, positional displacement amount determination unit 39, and evaluation function derivation unit 50 can be any of the following types of processors. As mentioned above, the above types of processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as a Programmable Logic Device (PLD), which is a processor whose circuit configuration can be changed after manufacturing, such as an FPGA (Field Programmable Gate Array), and a dedicated electrical circuit, which is a processor with a circuit configuration specifically designed to perform a particular process, such as an ASIC (Application Specific Integrated Circuit).

[0150] A single processing unit may consist of one of these various processors, or it may consist of a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs or a combination of a CPU and an FPGA). Alternatively, multiple processing units may be composed of a single processor.

[0151] Examples of configuring multiple processing units with a single processor include, firstly, a configuration where one or more CPUs and software combine to form a single processor, as exemplified by client and server computers, and this processor functions as multiple processing units. Secondly, a configuration using a processor that realizes the functions of the entire system, including multiple processing units, on a single IC (Integrated Circuit) chip, as exemplified by System-on-a-Chip (SoC). Thus, various processing units are configured, in terms of hardware structure, using one or more of the above-mentioned processors.

[0152] Furthermore, the hardware structure of these various processors can more specifically utilize electrical circuits (Circuitry) that combine circuit elements such as semiconductor elements. [Explanation of symbols]

[0153] 1. Radiation imaging device 2 Console 3. Image storage system 4,4A,4B,4C Image Processing Device 10. Photography Department 11 Rotation axis 12 Arm section 13 Shooting platform 14. Radiation irradiation area 15. Radiation detector 15A Detection surface 16 X-ray source 17 Compression plate 18 Support part 19 Moving mechanism 21 CPU 22 Image Processing Programs 23 Storage 24 displays 25 Input Devices 26 memory 27 Network Interface 28 buses 31 Image acquisition unit 32 Structure extraction part 33 Reconstruction part 34 Feature Structure Detection Unit 35 Projection section 36 Positional displacement amount derivation section 37 Display Control Unit 38 Focal plane discrimination section 39 Positional displacement amount determination unit 40 display screen Labels 41, 42 45 Warning display 48 Structures 50. Derivation of the Evaluation Function C1~C4 Corresponding Points DJ tomography DHJ-corrected tomographic images F1, F3, F4, F5 Feature Structure Image of F2 feature structure Gi (i=1~n) Projection Image Gc reference projection image GTi(i=1~n) Fault plane projection image H1, H2 Search Range M Breast O1~O4 Pixel positions on the tomographic plane P1~P3 Pixel positions of the projected image Areas of interest: Rf0, R1-R3, Rh1, Rh2 Ra0 surrounding area Si(i=1~n) Source position Sc reference source position SDj,SDk structural fault images SGi (i=1~n) Structural Projection Image Tj fault plane V10, Vf3 Shift Vector X0 optical axis

Claims

1. Equipped with at least one processor, The aforementioned processor, The radiation source is moved relative to the detection surface of the detection unit, and the imaging device is made to perform tomosynthesis imaging, irradiating the subject with radiation at multiple source positions resulting from the movement of the radiation source, thereby acquiring multiple projection images corresponding to each of the multiple source positions. A specific structure is extracted from the aforementioned multiple projection images to derive multiple structural projection images. The plurality of structural projection images are reconstructed to derive structural tomographic images for each of the plurality of tomographic planes of the subject, From the aforementioned multiple structural tomographic images, at least one characteristic structure is detected. In the corresponding tomographic plane where the feature structure is detected, the multiple projection images are reconstructed by correcting the positional displacement between them based on the motion of the subject, using the feature structure as a reference, and a corrected tomographic image is derived in at least one tomographic plane of the subject. An image processing device configured in such a way.

2. The image processing apparatus according to claim 1, wherein the specified structure is at least one of a line structure and a point structure.

3. The processor extracts at least one of the line structure and the point structure based on the degree of concentration of the gradient vector representing the gradient of pixel values ​​in the projected image. The image processing apparatus according to claim 2, configured as described above.

4. The processor derives the amount of positional displacement between the plurality of projection images based on the motion of the subject, using the characteristic structure as a reference in the corresponding tomographic plane. The corrected tomographic image is derived by correcting the aforementioned positional displacement and reconstructing the multiple projection images. An image processing apparatus according to any one of claims 1 to 3, configured as described above.

5. The processor detects multiple feature structures from the multiple structural tomographic images, It is determined whether the corresponding tomographic plane that corresponds to the structural tomographic image in which each of the aforementioned multiple feature structures is detected is the focal plane. The amount of displacement is derived in the corresponding fault plane that has been identified as the focal plane. The image processing apparatus according to claim 4, configured as described above.

6. The processor detects points in the structural tomographic image that satisfy specific threshold conditions as feature structures. The image processing apparatus according to claim 5, configured as described above.

7. The processor updates the structural tomographic image by reconstructing the structural projection image while correcting the positional misalignment. The updated feature structures are detected from the updated structural tomographic image. The amount of positional displacement is updated using the updated feature structure. The process of updating the structural tomographic image, updating the characteristic structure, and updating the positional displacement amount is repeated. The image processing apparatus according to claim 6, configured as described above.

8. The processor updates the structural tomographic image by reconstructing the structural projection image while correcting the positional misalignment. Based on the updated threshold conditions, the updated feature structures are detected from the updated structural tomographic image. The amount of positional displacement is updated using the updated feature structure. The process of updating the structural tomographic image, updating the feature structure based on the updated threshold conditions, and updating the positional displacement amount is repeated. The image processing apparatus according to claim 6, configured as described above.

9. The processor projects the plurality of projection images onto the corresponding tomographic plane based on the positional relationship between the source position and the detection unit at the time of acquisition for each of the plurality of projection images, and derives a tomographic projection image corresponding to each of the plurality of projection images. In the corresponding tomographic plane, the amount of positional displacement between the multiple tomographic projection images based on the motion of the subject is derived as the amount of positional displacement between the multiple projection images, using the characteristic structure as a reference. The image processing apparatus according to claim 4, configured as described above.

10. The processor sets local regions corresponding to the feature structures in the plurality of tomographic projection images and derives the amount of positional displacement based on the local regions. The image processing apparatus according to claim 9, configured as described above.

11. The processor sets a plurality of first local regions including the feature structure in the plurality of tomographic projection images, sets a second local region including the feature structure in the tomographic image in which the feature structure is detected, derives the positional displacement amounts of the plurality of first local regions relative to the second local region as provisional positional displacement amounts, and derives the positional displacement amount based on the plurality of provisional positional displacement amounts. The image processing apparatus according to claim 9, configured as described above.

12. The processor derives the provisional displacement amount based on the region surrounding the feature structure in the second local region. The image processing apparatus according to claim 11, configured as described above.

13. The processor reconstructs the plurality of projection images, excluding the target projection image corresponding to the target tomographic plane projection image from which the positional displacement amount is to be derived, and derives the plurality of tomographic images as the target tomographic image. The amount of positional displacement for the target tomographic projection image is derived using the target tomographic image. The image processing apparatus according to claim 11, configured as described above.

14. The processor performs an image quality evaluation of the region of interest, including the feature structure, in the corrected tomographic image, and determines whether the derived positional displacement is appropriate or inappropriate based on the results of the image quality evaluation. The image processing apparatus according to claim 4, configured as described above.

15. The processor derives multiple tomographic images by reconstructing the multiple projection images, The image quality of the region of interest, including the characteristic structure in the tomographic image, is evaluated. The results of the image quality evaluation for the corrected tomographic image are compared with the results of the image quality evaluation for the original tomographic image. The tomographic image with the better image quality evaluation result is determined to be the final tomographic image. The image processing apparatus according to claim 14, configured as described above.

16. The processor derives an evaluation function for evaluating the image quality of the region of interest, including the feature structure, in the corrected tomographic image. The positional displacement amount that optimizes the evaluation function is derived. The image processing apparatus according to claim 4, configured as described above.

17. The image processing apparatus according to any one of claims 1 to 3, wherein the subject is a breast.

18. The processor changes the search range for deriving the positional displacement amount according to at least one of the following: breast density, breast size, tomosynthesis imaging time, breast compression pressure during tomosynthesis imaging, and breast imaging direction. The image processing apparatus according to claim 17, configured as described above.

19. The radiation source is moved relative to the detection surface of the detection unit, and the imaging device is made to perform tomosynthesis imaging, irradiating the subject with radiation at multiple source positions resulting from the movement of the radiation source, thereby acquiring multiple projection images corresponding to each of the multiple source positions. A specific structure is extracted from the aforementioned multiple projection images to derive multiple structural projection images. The plurality of structural projection images are reconstructed to derive structural tomographic images for each of the plurality of tomographic planes of the subject, From the aforementioned multiple structural tomographic images, at least one characteristic structure is detected. In the corresponding tomographic plane where the feature structure is detected, the multiple projection images are reconstructed by correcting the positional displacement between them based on the motion of the subject, using the feature structure as a reference, and a corrected tomographic image is derived in at least one tomographic plane of the subject. An image processing method that includes the following.

20. The radiation source is moved relative to the detection surface of the detection unit, and the imaging device is made to perform tomosynthesis imaging, irradiating the subject with radiation at multiple source positions resulting from the movement of the radiation source, thereby acquiring multiple projection images corresponding to each of the multiple source positions. A specific structure is extracted from the aforementioned multiple projection images to derive multiple structural projection images. The plurality of structural projection images are reconstructed to derive structural tomographic images for each of the plurality of tomographic planes of the subject, From the aforementioned multiple structural tomographic images, at least one characteristic structure is detected. In the corresponding tomographic plane corresponding to the structural tomographic image in which the characteristic structure was detected, the multiple projection images are reconstructed by correcting the positional displacement between them based on the motion of the subject, using the characteristic structure as a reference, and a corrected tomographic image is derived in at least one tomographic plane of the subject. An image processing program that causes a computer to perform a process that includes [specific processing steps].

Images