Radiation imaging device, image processing device, and image processing method

Abstract

This radiation imaging device (100) comprises: a radiation imaging unit (1); and a control unit (2) having a movement vector acquisition unit (2a) that inputs a first radiation image (31) and a second radiation image (32) of a subject (90) to a trained model (20) for generating a movement vector (21) of an object (60) between a first image (30a) and a second image (30b), and acquires a movement vector of a target object (91). The control unit is configured to align, on the basis of the movement vector, the position of the target object in one of the first radiation image or the second radiation image with the position of the target object in the other of the first radiation image or the second radiation image, and to acquire a difference image (34) obtained by subtracting one of the post-alignment first radiation image or second radiation image from the other.

Description

Radiography apparatus, image processing apparatus, and image processing method

[0001] The present invention relates to a radiographic apparatus, an image processing apparatus, and an image processing method, and more particularly to a radiographic apparatus, an image processing apparatus, and an image processing method for obtaining a difference image by subtracting one of two radiographic images from the other.

[0002] Conventionally, there has been known a radiographic imaging apparatus that subtracts one of two radiographic images from the other to obtain a difference image. Such a radiographic imaging apparatus is disclosed, for example, in International Publication No. 2019 / 053935.

[0003] International Publication No. 2019 / 053935 discloses a radiographic imaging apparatus including an imaging unit, an image generation unit, a movement map generation unit, a smoothed movement map generation unit, a pixel movement processing unit, and a composite image generation unit. In the configuration disclosed in International Publication No. 2019 / 053935, the image generation unit is configured to generate a first image and a second image of the same subject captured at different times by the imaging unit. The movement map generation unit is configured to generate a movement map representing the movement direction and amount of movement of at least some of the pixels belonging to the first image. The smoothed movement map generation unit is configured to generate the smoothed movement map by suppressing high-frequency components in the spatial direction of the movement map. The pixel movement processing unit is configured to move pixels of the first image based on the smoothed movement map to generate a deformed image. The image synthesis processing unit is configured to generate a composite image by synthesizing the deformed image and the second image.

[0004] The motion map generator disclosed in WO 2019 / 053935 is configured to generate a motion map using a pixel-value-based method without using a trained model. Specifically, the motion map generator is configured to generate a motion map for a pixel of the first image based on a comparison result between a pixel value of a pixel of the second image and each of pixel values ​​of a plurality of pixels belonging to a region including the corresponding pixel in the first image.

[0005] International Publication No. 2019 / 053935

[0006] Here, as a result of intensive research by the present inventors, it has been found that, as disclosed in the above-mentioned International Publication No. 2019 / 053935, in a configuration in which the position of each pixel (object) belonging to an image is aligned using a motion vector acquired by a pixel value-based method, it is difficult to suppress the occurrence of artifacts due to object misalignment when, for example, a difference image is acquired based on two radiographic images depicting a large, slowly moving object, such as a diaphragm. Furthermore, as disclosed in the above-mentioned International Publication No. 2019 / 053935, in a configuration in which motion vectors are acquired by focusing on corresponding pixels of an object appearing in both radiographic images, it has been found that, for example, in images in which the pixel values ​​of corresponding pixels of an object appearing in both images are significantly different, it is difficult to accurately acquire motion vectors. In such cases, it becomes difficult to accurately align the position of the object appearing in the first radiographic image and the second radiographic image, and artifacts due to object misalignment occur in the difference image. Therefore, a technology capable of reducing artifacts due to object misalignment is desired.

[0007] The present invention has been made to solve the above-mentioned problems, and one object of the present invention is to provide a radiographic apparatus, an image processing apparatus, and an image processing method that are capable of reducing artifacts caused by positional deviation of an object.

[0008] A radiographic imaging apparatus according to a first aspect of the present invention includes a radiation irradiation unit that irradiates radiation onto a subject and a radiation detection unit that detects radiation that has passed through the subject, and is equipped with a radiographic imaging unit that images the subject, a memory unit that stores a trained model that receives a first image and a second image captured at different times to generate a motion vector of an object between the first image and the second image, and generates an image at an intermediate point between the time the first image and the time the second image are captured based on the motion vector, and a control unit that inputs the first radiographic image of the subject captured by the radiographic imaging unit and the second radiographic image captured at a different time from the first radiographic image into the trained model stored in the memory unit and acquires the motion vector of the object between the first radiographic image and the second radiographic image, and is configured to align the position of the object in one of the first radiographic image and the second radiographic image with the position of the object in the other image based on the motion vector acquired by the motion vector acquisition unit, and acquire a difference image by subtracting one of the first radiographic image and the second radiographic image after alignment.

[0009] An image processing device according to a second aspect of the present invention includes an image acquisition unit that acquires a first radiographic image of a subject captured by a radiographic imaging unit including a radiation irradiation unit that irradiates the subject with radiation and a radiation detection unit that detects radiation that has passed through the subject, and a second radiographic image that is captured at a timing different from that of the first radiographic image, and stores a trained model that generates a movement vector of an object between the first image and the second image by inputting the first image and the second image that are captured at different timings, and generates an image at an intermediate point between the time when the first image was captured and the time when the second image was captured based on the movement vector. The system includes a memory unit, and a control unit having a movement vector acquisition unit that inputs the first radiographic image and the second radiographic image acquired by the image acquisition unit into a trained model stored in the memory unit and acquires a movement vector of the object between the first radiographic image and the second radiographic image, and the control unit is configured to align the position of the object in one of the first radiographic image and the second radiographic image with the position of the object in the other image based on the movement vector acquired by the movement vector acquisition unit, and to acquire a difference image by subtracting one of the first radiographic image and the second radiographic image from the other after alignment.

[0010] An image processing method according to a third aspect of the present invention includes the steps of: acquiring a first radiographic image of a subject captured by a radiographic unit including a radiation irradiation unit that irradiates the subject with radiation and a radiation detection unit that detects radiation that has passed through the subject; and acquiring a second radiographic image captured at a timing different from that of the first radiographic image; generating a movement vector of the object between the first and second images by inputting the first and second radiographic images, which have been captured at different timings, and inputting the first and second radiographic images into a trained model that generates an image at an intermediate point in time between the capture of the first image and the capture of the second image, based on the movement vector, thereby acquiring the movement vector of the object between the first and second radiographic images; and aligning a position of the object in one of the first and second radiographic images with a position of the object in the other image, based on the acquired movement vector, and acquiring a difference image by subtracting one of the aligned first and second radiographic images from the other.

[0011] In the radiographic imaging device according to a first aspect, the image processing device according to a second aspect, and the image processing method according to a third aspect of the present invention, a difference image is acquired by subtracting one of the first and second radiographic images in which the object is aligned based on a motion vector generated by inputting the first and second radiographic images into a trained model. This reduces the positional deviation of the object compared to a configuration in which a difference image is acquired by subtracting one of the first and second radiographic images in which the object is aligned based on a motion vector acquired using a pixel value-based method. As a result, artifacts due to the positional deviation of the object can be reduced in the difference image. The inventors have confirmed in experiments described below that the difference image can reduce artifacts due to the positional deviation of the object. Furthermore, the motion vector acquired using the pixel value-based method is a motion vector acquired without using a trained model based on a comparison result between the pixel value of one pixel in one of two images captured at different times and each of the pixel values ​​of multiple pixels belonging to a region including the corresponding pixel in the other image.

[0012] FIG. 1 is a block diagram showing the configuration of a radiographic apparatus according to a first embodiment. FIG. 2 is a schematic diagram showing the configuration of a radiographic apparatus according to the first embodiment. FIG. 3 is a schematic diagram for explaining a non-contrast agent image. FIG. 4 is a schematic diagram for explaining a contrast agent image. FIG. 5 is a schematic diagram for explaining a configuration in which a trained model according to the first embodiment generates a movement vector. FIG. 6 is a schematic diagram for explaining a configuration in which a control unit according to the first embodiment acquires a difference image. FIG. 7 is a flowchart for explaining a configuration in which a control unit according to the first embodiment acquires a difference image. FIG. 8 is a block diagram showing the configuration of a radiographic apparatus according to a second embodiment. FIG. 9 is a schematic diagram for explaining a first tube voltage radiation image. FIG. 10 is a schematic diagram for explaining a second tube voltage radiation image. FIG. 11 is a schematic diagram for explaining a configuration in which a control unit according to the second embodiment acquires a difference image. FIG. 12 is a flowchart for explaining a configuration in which a control unit according to the second embodiment acquires a difference image. FIG. 13 is a schematic diagram for explaining a second difference image according to the second comparative example. FIG. 14 is a schematic diagram showing an enlarged view of an artifact in the second difference image according to the second comparative example. Fig. 10 is an enlarged schematic diagram of an artifact in the second difference image according to the second embodiment.Fig. 11 is a block diagram for explaining an image processing device according to a modified example.

[0013] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described with reference to the drawings.

[0014] First Embodiment First, the configuration of a radiation imaging apparatus 100 according to a first embodiment of the present invention will be described with reference to FIGS.

[0015] 1, the radiation imaging apparatus 100 includes a radiation imaging unit 1 and a control unit 2. The radiation imaging apparatus 100 also includes a storage unit 3, a display unit 4, and an input receiving unit 5.

[0016] The radiation imaging unit 1 includes a radiation irradiation unit 1a and a radiation detection unit 1b, and is configured to capture an image of a subject 90 (see FIG. 2). In the first embodiment, the radiation imaging unit 1 is configured to capture a moving image of the subject 90.

[0017] The radiation irradiator 1a is configured to irradiate radiation onto the subject 90. In the first embodiment, the radiation irradiator 1a is configured to irradiate X-rays as radiation. That is, the radiation irradiator 1a includes an X-ray generating device such as an X-ray tube.

[0018] The radiation detection unit 1b is configured to detect radiation that has passed through the subject 90. In the first embodiment, the radiation detection unit 1b is configured to detect X-rays that have passed through the subject 90. The radiation detection unit 1b includes, for example, an FPD (Flat Panel Detector).

[0019] The control unit 2 is configured to control each unit of the radiation imaging apparatus 100. The control unit 2 is configured with a processor such as a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), an FPGA (Field-Programmable Gate Array), or a circuit, and memories such as a ROM (Read Only Memory) and a RAM (Random Access Memory).

[0020] The control unit 2 also includes functional blocks: a motion vector acquisition unit 2a, a registration processing unit 2b, a difference processing unit 2c, and a video generation unit 2d. The motion vector acquisition unit 2a, the registration processing unit 2b, the difference processing unit 2c, and the video generation unit 2d are configured in software as functional blocks realized by the control unit 2 executing a program stored in the storage unit 3. The motion vector acquisition unit 2a, the registration processing unit 2b, the difference processing unit 2c, and the video generation unit 2d may each be configured as separate hardware with a dedicated processor (processing circuit). The functions of the motion vector acquisition unit 2a, the registration processing unit 2b, the difference processing unit 2c, and the video generation unit 2d will be described in detail below.

[0021] The storage unit 3 is configured to store various programs (not shown) executed by the control unit 2. The storage unit 3 is also configured to store a trained model 20 (described later) and a movement vector 21. The storage unit 3 includes a non-volatile storage device such as a hard disk drive (HDD) or a solid state drive (SSD).

[0022] The display unit 4 is configured to display a difference image 34 (see FIG. 6 ), which will be described later, acquired by the control unit 2. The display unit 4 is, for example, a display device such as a liquid crystal monitor or an organic EL (Electro Luminescence) monitor.

[0023] The input receiving unit 5 is configured to receive operation inputs from an operator and includes input devices such as a mouse and a keyboard.

[0024] The radiation imaging apparatus 100 is configured to capture X-ray images to be presented to an operator such as a doctor when, for example, a subject 90 is treated using a catheter 94 (see FIG. 3).

[0025] (Configuration of X-ray imaging device) As shown in Fig. 2, the radiation irradiator 1a is held on one side of the arm 6 so as to face the radiation detector 1b. Therefore, in the first embodiment, X-rays irradiated from the radiation irradiator 1a are directed toward the radiation detector 1b. In the example shown in Fig. 2, the up-down direction is defined as the Z direction, with the upward direction being the Z1 direction and the downward direction being the Z2 direction. Furthermore, two directions perpendicular to each other in a horizontal plane perpendicular to the Z direction are defined as the X direction and the Y direction. One side of the X direction is defined as the X1 direction, and the other side is defined as the X2 direction. Furthermore, one side of the Y direction is defined as the Y1 direction, and the other side is defined as the Y2 direction.

[0026] The radiation detection unit 1b is held on the other side of the arm 6 so as to face the radiation irradiator 1a. Therefore, the radiation detection unit 1b is configured to detect X-rays irradiated from the radiation irradiator 1a.

[0027] The arm 6 is configured to hold the radiation emitting unit 1 a and the radiation detecting unit 1 b. In the first embodiment, the arm 6 has an arc shape, and holds the radiation emitting unit 1 a and the radiation detecting unit 1 b at one end and the other end. The arm 6 is a so-called C-arm.

[0028] The arm driving mechanism 7 is configured to drive the arm 6. The arm 6 is configured to be able to position the radiation emission unit 1 a and the radiation detection unit 1 b in various positions by being driven by the arm driving mechanism 7.

[0029] The detector moving mechanism 8 is provided on the arm 6, and is configured to move the radiation detector 1b forward or backward in the direction of the X-ray irradiation axis 50. The detector moving mechanism 8 includes, for example, a linear motion mechanism.

[0030] The bed 9 is configured to allow the subject 90 to be placed thereon. The bed 9 includes a top board 9a and a top board moving mechanism 9b. The subject 90 is placed on the top board 9a.

[0031] 2, in the first embodiment, the arm 6 is disposed so that the tabletop 9a is located between the radiation irradiator 1a and the radiation detector 1b. Therefore, the X-rays irradiated from the radiation irradiator 1a pass through the subject 90 and are detected by the radiation detector 1b.

[0032] The top moving mechanism 9b is configured to move the top 9a in the Z direction. The top moving mechanism 9b is configured to translate the top 9a in the XY plane. The top moving mechanism 9b includes, for example, a linear motion mechanism that moves the top 9a in the X direction, a linear motion mechanism that moves the top 9a in the Y direction, and a linear motion mechanism that moves the top 9a in the Z direction.

[0033] The arm position changing mechanism 10 is configured to move the arm driving mechanism 7 and move the arm 6 together with the arm driving mechanism 7 to a desired imaging position. This allows the radiation imaging apparatus 100 to image the subject 90 from various imaging positions.

[0034] (First Radiation Image and Second Radiation Image) Next, the first radiation image 31 and the second radiation image 32 in the first embodiment will be described with reference to FIGS. 3 and 4. FIG.

[0035] The image shown in Fig. 3 is a first radiographic image 31. In the first embodiment, the first radiographic image 31 is an image (X-ray image) of the subject 90 (see Fig. 2) captured by the radiation imaging unit 1. Specifically, the first radiographic image 31 is a non-contrast image 31a captured without using a contrast agent. The non-contrast image 31a shows bones 92 and a diaphragm 93. The non-contrast image 31a also shows a catheter 94. The bones 92 include a spine 92a and ribs 92b.

[0036] The image shown in Fig. 4 is a second radiographic image 32. In the first embodiment, the second radiographic image 32 is an image (X-ray image) of the subject 90 (see Fig. 2) captured by the radiation imaging unit 1 at a timing different from that of the first radiographic image 31. Specifically, the second radiographic image 32 is a contrast agent image 32a captured using a contrast agent. A bone 92 and a diaphragm 93 are captured in the contrast agent image 32a. A catheter 94 and a blood vessel 95 are also captured in the contrast agent image 32a.

[0037] The control unit 2 (see FIG. 1) subtracts the non-contrast image 31a (see FIG. 3) from the contrast image 32a to obtain an image showing only the blood vessels 95 (a blood vessel roadmap).

[0038] Here, the timing at which contrast agent image 32a is captured and the timing at which non-contrast agent image 31a is captured are different from each other. Depending on the imaging unit used when capturing contrast agent image 32a, the position of object 91 may fluctuate due to breathing, heartbeat, or the like. Therefore, the position of object 91 captured in contrast agent image 32a may differ from the position of object 91 captured in non-contrast agent image 31a. When non-contrast agent image 31a is subtracted from contrast agent image 32a in a state in which the positions of object 91 captured in contrast agent image 32a and non-contrast agent image 31a are different, artifact 80 (see FIG. 8 ) occurs due to a positional shift of object 91.

[0039] Therefore, in the first embodiment, the control unit 2 is configured to align the object 91 in the non-contrast agent image 31a to the position of the object 91 in the contrast agent image 32a based on the movement vector 21 (see Figure 1) generated by the trained model 20 (see Figure 1).

[0040] (Object) In the first embodiment, the object 91 is either a bone 92 or a diaphragm 93. The object 91 may include a catheter 94.

[0041] (Trained Model and Movement Vector) Next, the trained model 20 and the movement vector 21 will be described with reference to FIG.

[0042] The trained model 20 is configured to input a first image 30a and a second image 30b captured at different times, and generate a movement vector 21 of an object 60 between the first image 30a and the second image 30b. The object 60 may be, for example, a heart, blood vessels, or bones. The trained model 20 is also configured to generate an image 30c at an intermediate point between the capture time of the first image 30a and the capture time of the second image 30b, based on the movement vector 21. Specifically, the trained model 20 is a trained model that uses an intermediate image generation AI called FILM (Frame Interpolation for Large Motion) provided by Google Research. FILM inputs two images captured at different times and performs resolution conversion processing and feature extraction processing for each image. FILM generates optical flow in the flow estimation process. FILM generates intermediate images in the fusion process. The optical flow, which is an intermediate product of generating the intermediate images, is extracted as a movement vector 21 and used to align the object 91 in the non-contrast agent image 31a.

[0043] The movement vector 21 is a vector map that indicates the movement direction and movement amount of each pixel from the first image 30a to the second image 30b, or from the second image 30b to the first image 30a.

[0044] In the first embodiment, the trained model 20 is configured to generate both a first movement vector 21 a and a second movement vector 21 b as the movement vector 21 .

[0045] The first movement vector 21a is the movement vector 21 of the object 60 between the first image 30a and the second image 30b. The second movement vector 21b is the movement vector 21 of the object 60 between the second image 30b and the first image 30a.

[0046] (Acquisition of Difference Image) Next, with reference to FIG. 6, a configuration in which the control unit 2 (see FIG. 1) according to the first embodiment acquires the difference image 34 will be described.

[0047] The movement vector acquisition unit 2a is configured to acquire a movement vector 21 (see FIG. 5) of the object 91 (see FIG. 4) between the non-contrast agent image 31a and the contrast agent image 32a.

[0048] In the first embodiment, the movement vector acquisition unit 2a is configured to acquire at least one of a first movement vector 21a and a second movement vector 21b (see Figure 4) generated by inputting a non-contrast agent image 31a and a contrast agent image 32a into a trained model 20 stored in the memory unit 3.

[0049] In the first embodiment, the movement vector acquisition unit 2a is configured to acquire the movement vector 21 of either the bone 92 or the diaphragm 93 between the non-contrast agent image 31a and the contrast agent image 32a.

[0050] 6, in the first embodiment, the movement vector acquisition unit 2a is configured to acquire a first movement vector 21a of an object 91 that has moved from a position in the non-contrast agent image 31a to the contrast agent image 32a, which is generated by inputting the non-contrast agent image 31a and the contrast agent image 32a to the trained model 20. The movement vector acquisition unit 2a outputs the acquired first movement vector 21a to the registration processing unit 2b.

[0051] The registration processing unit 2b aligns the position of the object 91 in one of the non-contrast image 31a and the contrast image 32a with the position of the object 91 in the other image, based on the movement vector 21 acquired by the movement vector acquisition unit 2a. In the first embodiment, the registration processing unit 2b aligns the object 91 in the non-contrast image 31a with the object 91 in the contrast image 32a, based on the first movement vector 21a. Specifically, the registration processing unit 2b moves each pixel in the non-contrast image 31a by the movement amount and movement direction of the corresponding position of the first movement vector 21a. Note that when aligning the object 91 in the non-contrast image 31a, if a pixel does not exist at the moved position, the registration processing unit 2b interpolates the pixel value of the pixel that does not exist at the moved position, based on the pixel values ​​of surrounding pixels.

[0052] Thereafter, the registration processing unit 2b outputs the registered non-contrast agent image 33 to the subtraction processing unit 2c.

[0053] The subtraction processing unit 2c is configured to obtain a subtraction image 34 (first subtraction image 34a) obtained by subtracting one of the aligned first radiographic image 31 and the second radiographic image 32 from the other. In the first embodiment, the subtraction processing unit 2c is configured to obtain the subtraction image 34 by aligning the non-contrast agent image 31a with the contrast agent image 32a based on the first movement vector 21a and subtracting the aligned non-contrast agent image 33 from the contrast agent image 32a.

[0054] In the first embodiment, the object 91 in the non-contrast image 31a is aligned with the object 91 in the contrast image 32a using the first movement vector 21a. Therefore, even if, for example, the diaphragm 93 or the like moves due to breathing or the bone 92 moves due to body movement of the subject 90, the object 91 in the non-contrast image 31a can be aligned with high accuracy. Therefore, as shown in FIG. 6 , the first subtraction image 34a shows a blood vessel 95, but the bone 92 and the diaphragm 93 are not substantially visible. Similarly to the bone 92 and the diaphragm 93, the catheter 94 (see FIG. 4 ) is also not substantially visible. In the subtraction image 34 shown in FIG. 6 , the bone 92, the diaphragm 93, and the catheter 94 are indicated by dashed lines to indicate that the bone 92, the diaphragm 93, and the catheter 94 are not substantially visible. Note that the concept of "almost not visible" also includes cases where bones 92, diaphragm 93, and catheter 94 are visible to the extent that they do not affect the visibility of the object the user wants to check (for example, blood vessels 95).

[0055] In the first embodiment, the radiation imaging unit 1 is configured to capture moving images, and therefore the motion vector acquisition unit 2 a is configured to acquire, each time a plurality of contrast agent images 32 a are captured, a first motion vector 21 a of the object 91 that has moved from its position in the non-contrast agent image 31 a in each of the plurality of contrast agent images 32 a.

[0056] The registration processing unit 2b is configured to register the non-contrast agent image 31a with each of the plurality of contrast agent images 32a based on each of the first movement vectors 21a. The subtraction processing unit 2c is configured to obtain a plurality of subtraction images 34 by subtracting the registered non-contrast agent image 33 from each of the plurality of contrast agent images 32a.

[0057] In the first embodiment, the moving image generating unit 2d (see FIG. 1) is configured to generate a moving image based on the plurality of difference images 34 acquired by the difference processing unit 2c. The moving image generating unit 2d generates a moving image at, for example, 15 fps (frames per second). The control unit 2 then displays the moving image generated by the moving image generating unit 2d on the display unit 4 (see FIG. 1).

[0058] (Difference Image Acquisition Process) Next, with reference to FIG. 7, a process in which the control unit 2 (see FIG. 1) according to the first embodiment acquires the difference image 34 (see FIG. 6) will be described.

[0059] In step 101, the control unit 2 acquires a first radiographic image 31 (non-contrast agent image 31a) (see FIG. 3) of the subject 90 (see FIG. 2) imaged by the radiation imaging unit 1 (see FIG. 1), and a second radiographic image 32 (contrast agent image 32a) (see FIG. 3) imaged at a different time from the first radiographic image 31.

[0060] In step 102, the motion vector acquisition unit 2a (see FIG. 6) inputs the non-contrast agent image 31a and the contrast agent image 32a to the trained model 20 (see FIG. 6) and acquires the motion vector 21 (see FIG. 1) of the object 91 (see FIG. 4) between the non-contrast agent image 31a and the contrast agent image 32a. In the first embodiment, the motion vector acquisition unit 2a acquires the first motion vector 21a (see FIG. 6).

[0061] In step 103, the registration processing unit 2b aligns the position of the object 91 in one of the non-contrast image 31a and the contrast image 32a with the position of the object 91 in the other image, based on the movement vector 21 acquired by the movement vector acquisition unit 2a. In the first embodiment, the registration processing unit 2b aligns the position of the object 91 in the non-contrast image 31a with the position of the object 91 in the contrast image 32a, based on the first movement vector 21a.

[0062] In step 103, the subtraction processor 2c subtracts the registered non-contrast agent image 31a from the registered non-contrast agent image 32a to obtain a subtraction image 34. In the first embodiment, the subtraction processor 2c subtracts the registered non-contrast agent image 33 from the registered contrast agent image 32a to obtain a first subtraction image 34a. The process then ends.

[0063] In the first embodiment, the control unit 2 repeats the processes of steps 101 to 103 to acquire a plurality of difference images 34. Then, the moving image generation unit 2d (see FIG. 1) generates a moving image including the acquired plurality of difference images 34. Thereafter, the moving image generation unit 2d displays the generated moving image on the display unit 4 (see FIG. 1).

[0064] (Effects of First Embodiment) In the first embodiment, the following effects can be obtained.

[0065] In the first embodiment, as described above, the radiation imaging apparatus 100 includes a radiation irradiating unit 1 a that irradiates radiation onto the subject 90 and a radiation detecting unit 1 b that detects radiation that has passed through the subject 90, a radiation imaging unit 1 that images the subject 90, a storage unit 3 that stores a trained model 20 that receives a first image 30 a and a second image 30 b that have been captured at different times to generate a movement vector 21 of the object 60 between the first image 30 a and the second image 30 b, and generates an image 30 c at an intermediate point between the time at which the first image 30 a and the second image 30 b were captured based on the movement vector 21, and a first radiation image 31 of the subject 90 captured by the radiation imaging unit 1 and a second image 30 c at a point in time different from the first radiation image 31. and a control unit 2 having a movement vector acquisition unit 2a that inputs the first radiographic image 31 and a second radiographic image 32 captured at a timing specified by the movement vector acquisition unit 2a into a trained model 20 stored in a memory unit 3, and acquires a movement vector 21 of the object 91 between the first radiographic image 31 and the second radiographic image 32. The control unit 2 (alignment processing unit 2b and difference processing unit 2c) is configured to align the position of the object 91 in one of the first radiographic image 31 and the second radiographic image 32 with the position of the object 91 in the other image based on the movement vector 21 acquired by the movement vector acquisition unit 2a, and to acquire a difference image 34 obtained by subtracting one of the first radiographic image 31 and the second radiographic image 32 from the other after alignment.

[0066] As a result, a difference image 34 is acquired by subtracting one of the first radiographic image 31 and the second radiographic image 32, in which the object 91 has been aligned, based on the movement vector 21 generated by the trained model 20, thereby reducing the positional deviation of the object 91 compared to a configuration in which a difference image is acquired by subtracting one of the first radiographic image and the second radiographic image, in which the object 91 has been aligned, based on a movement vector acquired using a pixel value-based method. As a result, artifacts 80 caused by the positional deviation of the object 91 can be reduced in the difference image 34. The fact that artifacts 80 caused by the positional deviation of the object 91 can be reduced in the difference image 34 has been confirmed in experiments conducted by the present inventors, which will be described later.

[0067] Furthermore, in the first embodiment, as described above, the image processing method includes steps of acquiring a first radiographic image 31 of the subject 90 captured by a radiation imaging unit 1 including a radiation irradiation unit 1 a that irradiates radiation onto the subject 90 and a radiation detection unit 1 b that detects radiation that has passed through the subject 90, and a second radiographic image 32 that has been captured at a timing different from that of the first radiographic image 31; and inputting the first radiographic image 31 and the second radiographic image 32 together with a first image 30 a and a second image 30 b that have been captured at different timings, thereby generating a movement vector 21 of the object 60 between the first image 30 a and the second image 30 b, and calculating the movement vector 21. the first radiographic image 31 and the second radiographic image 32 based on the acquired motion vector 21, and aligning the position of the object 91 in one of the first radiographic image 31 and the second radiographic image 32 with the position of the object 91 in the other image based on the acquired motion vector 21, and acquiring a difference image 34 by subtracting one of the first radiographic image 31 and the second radiographic image 32 from the other after alignment.

[0068] This makes it possible to provide an image processing method that can reduce the artifacts 80 caused by the positional displacement of the object 91 in the difference image 34, similar to the radiation imaging apparatus 100 described above.

[0069] Furthermore, in the first embodiment, the following additional effects can be obtained by configuring as follows.

[0070] That is, in the first embodiment, as described above, the trained model 20 is configured to generate both a first movement vector 21a, which is the movement vector 21 of the object 60 between the first image 30a and the second image 30b, and a second movement vector 21b, which is the movement vector 21 of the object 60 between the second image 30b and the first image 30a; the movement vector acquisition unit 2a is configured to acquire at least one of the first movement vector 21a and the second movement vector 21b generated by inputting the non-contrast agent image 31a (first radiographic image) and the contrast agent image 32a (second radiographic image) into the trained model 20; and the alignment processing unit 2b (control unit) is configured to align the non-contrast agent image 31a (first radiographic image) and the contrast agent image 32a (second radiographic image) based on either the first movement vector 21a or the second movement vector 21b. As a result, the trained model 20 generates both the first movement vector 21 a and the second movement vector 21 b, so that the object 91 can be accurately aligned in both a configuration in which the non-contrast image 31 a (first radiographic image) is aligned with the contrast image 32 a (second radiographic image) and a configuration in which the contrast image 32 a (second radiographic image) is aligned with the non-contrast image 31 a (first radiographic image). Therefore, in a configuration in which the object 91 needs to be aligned with an image acquired later, such as a moving image, the non-contrast image 31 a (first radiographic image) and the contrast image 32 a (second radiographic image) are aligned based on the first movement vector 21 a, thereby suppressing the occurrence of artifacts in the acquired difference image 34. Furthermore, in a configuration in which the object 91 may be aligned with either image, such as a still image, by aligning the non-contrast agent image 31a (first radiographic image) with the contrast agent image 32a (second radiographic image) based on either the first movement vector 21a or the second movement vector 21b, it is possible to prevent artifacts from occurring in the acquired difference image 34. Therefore, whether the radiographic image to be aligned is a moving image or a still image, it is possible to prevent artifacts from occurring in the acquired difference image 34.As a result, it is possible to improve usability for the user.

[0071] Furthermore, in the first embodiment, as described above, the object 91 is either the bone 92 or the diaphragm 93 appearing in the non-contrast image 31 a (first radiographic image) and the contrast image 32 a (second radiographic image), and the motion vector acquisition unit 2 a is configured to acquire a motion vector 21 of either the bone 92 or the diaphragm 93 between the non-contrast image 31 a (first radiographic image) and the contrast image 32 a (second radiographic image). This makes it possible to accurately align either the bone 92 or the diaphragm 93 appearing in the non-contrast image 31 a (first radiographic image) and the contrast image 32 a (second radiographic image) based on the motion vector 21. As a result, it is possible to prevent the occurrence of artifacts 80 (see FIG. 8 ) due to misalignment of either the bone 92 or the diaphragm 93, and therefore to prevent degradation of the image quality of the subtraction image 34 due to the artifacts 80.

[0072] Furthermore, in the first embodiment, as described above, the first radiographic image 31 is a non-contrast image 31a captured without using a contrast agent, the second radiographic image 32 is a contrast image 32a captured using a contrast agent, the object 91 is either a bone 92 or a diaphragm 93, the movement vector acquisition unit 2a is configured to acquire a first movement vector 21a of the object 91 that has moved from a position in the non-contrast image 31a in the contrast image 32a, which is generated by inputting the non-contrast image 31a and the contrast image 32a into the trained model 20, and the control unit 2 (subtraction processing unit 2c) is configured to align the non-contrast image 31a with the contrast image 32a based on the first movement vector 21a, and to subtract the aligned non-contrast image 33 from the contrast image 32a, thereby acquiring a subtraction image 34.

[0073] Here, blood vessels 95 are visible in the contrast agent image 32a but not in the non-contrast agent image 31a. Therefore, if there is no positional displacement of the object 91, such as the bone 92 and the diaphragm 93, only the blood vessels 95 are visible in the subtraction image 34 (first subtraction image 34a) obtained by subtracting the non-contrast agent image 31a from the contrast agent image 32a. On the other hand, if there is a positional displacement of the object 91, an artifact 80 (see FIG. 8 ) due to the positional displacement of the object 91 occurs. Therefore, the above-described configuration can prevent the artifact 80 from appearing in the subtraction image 34 (first subtraction image 34a) due to the positional displacement of either the bone 92 or the diaphragm 93. As a result, it is possible to prevent the artifact 80 from appearing at a position superimposed on the blood vessels 95 visible in the contrast agent image 32a, thereby preventing a decrease in the visibility of the blood vessels 95 in the subtraction image 34. The fact that it is possible to suppress the occurrence of artifacts 80 in the difference image 34 (first difference image 34a) has been confirmed by the inventors of the present application in an experiment described below.

[0074] Furthermore, in the first embodiment, as described above, the radiation imaging unit 1 is configured to capture moving images of the subject 90, and the motion vector acquisition unit 2a is configured to acquire, each time a plurality of contrast agent images 32a are captured, a first motion vector 21a of the object 91 that has moved from its position in the non-contrast agent image 31a in each of the plurality of contrast agent images 32a. The control unit 2 (the alignment processing unit 2b and the subtraction processing unit 2c) is configured to align the non-contrast agent image 31a with each of the plurality of contrast agent images 32a based on each of the first motion vectors 21a, and to subtract the aligned non-contrast agent image 33 from each of the plurality of contrast agent images 32a, thereby obtaining a plurality of difference images 34. This makes it possible to suppress the occurrence of artifacts in each of the difference images 34 when the plurality of difference images 34 are displayed as a moving image. As a result, when the difference images 34 are displayed as a moving image, it is possible to suppress a decrease in the visibility of an object that the user wants to check, such as a blood vessel 95, in the moving image.

[0075] First Example (Effect Confirmation Experiment) In order to confirm the effect of the first embodiment described above, the following experiment was conducted: That is, the difference in artifacts 80 caused by the object 91 between the difference image 34 of the first comparative example and the first difference image 34a of the first example was compared.

[0076] The difference image 35 of the first comparative example shown in Figure 8 is an image obtained by aligning the object 91 in the non-contrast image 31a with the object 91 in the contrast image 32a based on a movement vector obtained by a method based on pixel values, and subtracting the aligned non-contrast image 31a from the contrast image 32a.

[0077] Note that a motion vector obtained by a method based on pixel values ​​is a motion vector obtained based on the results of comparing the pixel value of one pixel in one of two images taken at different times with the pixel values ​​of each of multiple pixels belonging to an area containing the corresponding pixel in the other image, without using a trained model.

[0078] In the subtraction image 35 of the first comparative example, there are positional deviations of the bone 92, the diaphragm 93, and the catheter 94. Therefore, the subtraction image 35 shows, as artifacts 80 caused by the object 91, an artifact 81 caused by the positional deviation of the bone 92, an artifact 82 caused by the positional deviation of the diaphragm 93, and an artifact 83 caused by the positional deviation of the catheter 94. Note that the artifact 81 caused by the positional deviation of the bone 92 includes an artifact 81a caused by the positional deviation of the spine 92a and an artifact 81b caused by the positional deviation of the rib 92b.

[0079] In the region 40 shown by the dashed line in the difference image 35 of the first comparative example, an artifact 81 a caused by the misalignment of the spine 92 a and an artifact 83 caused by the catheter 94 are captured in addition to the blood vessel 95. Therefore, it was confirmed that the visibility of the blood vessel 95 was reduced in the region 40.

[0080] On the other hand, the first difference image 34a shown in Figure 9 is an image obtained by aligning the object 91 in the non-contrast image 31a with the object 91 in the contrast image 32a based on the first movement vector 21a generated by the trained model 20, and subtracting the aligned non-contrast image 33 from the contrast image 32a.

[0081] In the first subtraction image 34a according to the first embodiment, the positional displacements of the bone 92, the diaphragm 93, and the catheter 94 are slight, and therefore, as shown in Fig. 9, an artifact 81 caused by the positional displacement of the bone 92, an artifact 82 caused by the positional displacement of the diaphragm 93, and an artifact 83 caused by the positional displacement of the catheter 94 are suppressed. Therefore, in the first subtraction image 34a, an artifact 80 caused by the object 91 does not occur near the blood vessel 95, and it was confirmed that the visibility of the blood vessel 95 is not reduced.

[0082] In the example shown in Figure 9, the artifact 81 caused by the misalignment of the bone 92, the artifact 82 caused by the misalignment of the diaphragm 93, and the artifact 83 caused by the misalignment of the catheter 94 are illustrated with dashed lines, thereby indicating that the artifact 81 caused by the misalignment of the bone 92, the artifact 82 caused by the misalignment of the diaphragm 93, and the artifact 83 caused by the misalignment of the catheter 94 are suppressed.

[0083] As shown in Figures 8 and 9, it was confirmed that the first subtraction image 34a of the first embodiment has fewer artifacts 81 caused by the displacement of the bone 92, artifacts 82 caused by the displacement of the diaphragm 93, and artifacts 83 caused by the displacement of the catheter 94 than the subtraction image 35 of the first comparative example.

[0084] Second Embodiment The configuration of a radiation imaging apparatus 200 according to a second embodiment will be described with reference to Figures 10 to 14. Note that components similar to those included in the radiation imaging apparatus 100 according to the first embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted.

[0085] 10 , the radiation imaging apparatus 200 includes a radiation imaging unit 1 and a control unit 201. The radiation imaging apparatus 100 also includes a storage unit 3, a display unit 4, and an input receiving unit 5.

[0086] The control unit 201 differs from the control unit 2 of the first embodiment in that it is equipped with a motion vector acquisition unit 201a, a position alignment processing unit 201b, and a difference processing unit 201c instead of the motion vector acquisition unit 2a, the position alignment processing unit 2b, and the difference processing unit 2c, and in that it does not have a moving image generation unit 2d.

[0087] (First Radiation Image and Second Radiation Image) Next, the first radiation image 31 and the second radiation image 32 in the second embodiment will be described with reference to FIGS. 11 and 12. FIG.

[0088] In the second embodiment, the first radiographic image 31 is a first tube voltage radiographic image 31b captured at a predetermined tube voltage. The first tube voltage radiographic image 31b shows a bone 92 and intestinal gas 96. The bone 92 is the spine.

[0089] In the second embodiment, the second radiographic image 32 is a second tube voltage radiographic image 32b captured at a lower tube voltage than the first tube voltage radiographic image 31b. The second tube voltage radiographic image 32b shows a bone 92 and intestinal gas 96. In the second tube voltage radiographic image 32b, the bone 92 is depicted with a dashed line, and the hatching applied to the intestinal gas 96 is different from the hatching applied to the intestinal gas 96 in the first tube voltage radiographic image 31b, thereby indicating the difference in the bone 92 and the intestinal gas 96 due to the difference in tube voltage compared to the first tube voltage radiographic image 31b.

[0090] (Object) In the second embodiment, the object 91 is intestinal gas 96 that appears in the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b.

[0091] The control unit 201 according to the second embodiment generates a subtraction image 34 (see FIG. 13 ) by subtracting the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b, which are obtained using different tube voltages. The control unit 201 then calculates bone density based on the generated subtraction image 34. When capturing the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b, a member for suppressing the subject's movement is attached to the subject 90. Therefore, displacement of the bones 92 and other components of the subject 90 due to their movement is unlikely to occur in the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b. However, intestinal gas 96 moves regardless of the subject's movement. Although the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b are captured at short time intervals, the intestinal gas 96 may move during this time. Therefore, if one of the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b is subtracted directly from the other, an artifact 84 (see FIG. 13 ) caused by misalignment of the intestinal gas 96 occurs. Therefore, in this embodiment, the position of the intestinal gas 96 in the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b is aligned to suppress the artifact 84 caused by misalignment of the intestinal gas 96. The control unit 201 then subtracts one of the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b from the other after alignment, thereby obtaining a subtraction image 34.

[0092] (Acquisition of Difference Image) Next, with reference to FIG. 13, a configuration in which the control unit 201 (see FIG. 10) according to the second embodiment acquires the difference image 34 will be described.

[0093] The motion vector acquiring unit 201a is configured to acquire a motion vector 21 (see FIG. 10 ) of intestinal gas 96 between the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b by inputting the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b into the trained model 20. Specifically, the motion vector acquiring unit 201a is configured to acquire at least one of a first motion vector 21a of the object 91 (see FIG. 12 ) that has moved from its position in the second tube voltage radiographic image 32b to the first tube voltage radiographic image 31b, and a second motion vector 21b (see FIG. 5 ) of the object 91 that has moved from its position in the first tube voltage radiographic image 31b to the second tube voltage radiographic image 32b, which are generated by inputting the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b into the trained model 20. In the second embodiment, the motion vector acquiring unit 201a acquires the first motion vector 21a. Then, the movement vector acquisition unit 201a outputs the acquired first movement vector 21a to the registration processing unit 201b.

[0094] The registration processing unit 201b registers the first tube voltage radiographic image 31b with the second tube voltage radiographic image 32b based on the first movement vector 21a. Specifically, the registration processing unit 201b moves each pixel in the first tube voltage radiographic image 31b by the movement amount and movement direction of the corresponding position of the first movement vector 21a. Note that when registering the object 91 in the first tube voltage radiographic image 31b, if a state occurs in which no pixel corresponds to the moved position, the registration processing unit 201b interpolates the pixel value of the pixel that does not correspond to the moved position based on the pixel values ​​of surrounding pixels.

[0095] Thereafter, the alignment processing unit 201b outputs the aligned first tube voltage radiation image 36 to the subtraction processing unit 201c.

[0096] The subtraction processing unit 201c is configured to obtain a subtraction image 34 by subtracting the aligned first tube voltage radiographic image 31b from the second tube voltage radiographic image 32b, or to align the second tube voltage radiographic image 32b with the first tube voltage radiographic image 31b based on the second movement vector 21b, or to obtain a subtraction image 34 (second subtraction image 34b) by subtracting the aligned second tube voltage radiographic image 32b from the first tube voltage radiographic image 31b. In the second embodiment, the subtraction processing unit 201c is configured to obtain a second subtraction image 34b by subtracting the aligned first tube voltage radiographic image 36 from the second tube voltage radiographic image 32b.

[0097] As shown in FIG. 13, the second subtraction image 34b contains a slight artifact 84 caused by the displacement of the intestinal gas 96.

[0098] In the second embodiment, the control unit 201 is configured to calculate bone density based on pixel values ​​of the area of ​​bone 92 shown in the difference image 34 and pixel values ​​of the areas of soft tissue 41 located on both sides of the area of ​​bone 92.

[0099] (Difference Image Acquisition Process) Next, with reference to FIG. 14, a process in which the control unit 201 (see FIG. 10) according to the second embodiment acquires the difference image 34 (see FIG. 13) will be described.

[0100] In step 210, the control unit 201 acquires a first radiographic image 31 (first tube voltage radiographic image 31b) of the subject 90 captured by the radiation imaging unit 1, and a second radiographic image 32 (second tube voltage radiographic image 32b) captured at a different timing from the first radiographic image 31.

[0101] In step 211, the motion vector acquisition unit 201a (see FIG. 13 ) inputs the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b to the trained model 20, and acquires the motion vector 21 (see FIG. 10 ) of the object 91 (see FIG. 12 ) between the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b. In the second embodiment, the motion vector acquisition unit 201a acquires the first motion vector 21a (see FIG. 13 ).

[0102] In step 212, the registration processing unit 201b (see FIG. 13 ) aligns the position of the object 91 in one of the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b with the position of the object 91 in the other image, based on the movement vector 21 acquired by the movement vector acquisition unit 201a. In the second embodiment, the registration processing unit 201b aligns the position of the object 91 in the first tube voltage radiographic image 31b with the position of the object 91 in the second tube voltage radiographic image 32b, based on the first movement vector 21a.

[0103] In step 212, the subtraction processing unit 201c (see FIG. 13 ) obtains a subtraction image 34 by subtracting one of the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b after alignment from the other. In the second embodiment, the subtraction processing unit 201c obtains a second subtraction image 34b (see FIG. 13 ) by subtracting the first tube voltage radiographic image 36 after alignment (see FIG. 13 ) from the second tube voltage radiographic image 32b. Then, the processing ends.

[0104] The other configurations of the second embodiment are similar to those of the first embodiment.

[0105] Effect of Second Embodiment In the second embodiment, similarly to the first embodiment, the radiation imaging apparatus 200 includes a radiation irradiating unit 1 a that irradiates radiation onto the subject 90 and a radiation detecting unit 1 b that detects radiation that has passed through the subject 90, a radiation imaging unit 1 that images the subject 90, a storage unit 3 that stores a trained model 20 that receives a first image 30 a and a second image 30 b that have been captured at different times, generates a movement vector 21 of the object 60 between the first image 30 a and the second image 30 b, and generates an image 30 c at an intermediate point between the time when the first image 30 a was captured and the time when the second image 30 b was captured based on the movement vector 21, and a first radiation image 31 of the subject 90 captured by the radiation imaging unit 1 and a second image 30 b that has been captured at different times. The radiographic image 31 and a second radiographic image 32 captured at a different timing are input to a trained model 20 stored in the storage unit 3, and the control unit 201 includes a motion vector acquisition unit 201a that acquires a motion vector 21 of the object 91 between the first radiographic image 31 and the second radiographic image 32. The control unit 201 is configured to align the position of the object 91 in one of the first radiographic image 31 and the second radiographic image 32 with the position of the object 91 in the other image based on the motion vector 21 acquired by the motion vector acquisition unit 201a, and to acquire a difference image 34 obtained by subtracting one of the first radiographic image 31 and the second radiographic image 32 from the other after alignment. This reduces artifacts 80 in the difference image 34 that are caused by misalignment of the object 91. The inventors of the present application have confirmed in experiments described below that the artifacts 80 in the difference image 34 that are caused by misalignment of the object 91 can be reduced.

[0106] Furthermore, in the second embodiment, as described above, the object 91 is intestinal gas 96 appearing in the first tube voltage radiographic image 31b (first radiographic image) and the second tube voltage radiographic image 32b (second radiographic image), and the motion vector acquisition unit 201a is configured to acquire a motion vector 21 of the intestinal gas 96 between the first tube voltage radiographic image 31b (first radiographic image) and the second tube voltage radiographic image 32b (second radiographic image). This makes it possible to accurately align the intestinal gas 96 appearing in the first tube voltage radiographic image 31b (first radiographic image) and the second tube voltage radiographic image 32b (second radiographic image) based on the motion vector 21. As a result, it is possible to prevent the occurrence of artifacts 84 due to misalignment of the intestinal gas 96, and therefore to prevent degradation of the image quality of the subtraction image 34 due to the artifacts 84.

[0107] In the second embodiment, as described above, the first radiographic image 31 is a first tube voltage radiographic image 31b captured at a predetermined tube voltage, the second radiographic image 32 is a second tube voltage radiographic image 32b captured at a tube voltage lower than that of the first tube voltage radiographic image 31b, the object 91 is intestinal gas 96, and the movement vector acquisition unit 201a acquires a first movement vector 21a of the object 91 that has moved from its position in the second tube voltage radiographic image 32b in the first tube voltage radiographic image 31b, which are generated by inputting the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b into the trained model 20, and a second movement vector 21b of the object 91 that has moved from its position in the first tube voltage radiographic image 31b in the second tube voltage radiographic image 32b in the trained model 20. The control unit 201 is configured to acquire at least one of a first movement vector 21a and a second movement vector 21b of the object 91 that has moved from the first tube voltage radiation image 31b, and is configured to either align the first tube voltage radiation image 31b with the second tube voltage radiation image 32b based on the first movement vector 21a, and subtract the aligned first tube voltage radiation image 31b from the second tube voltage radiation image 32b to acquire a difference image 34, or to align the second tube voltage radiation image 32b with the first tube voltage radiation image 31b based on the second movement vector 21b, and subtract the aligned second tube voltage radiation image 32b from the first tube voltage radiation image 31b to acquire a difference image 34.

[0108] Here, the subtraction image 34 (second subtraction image 34b) obtained by subtracting one of the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b from the other is used, for example, to calculate bone density. Bone density is calculated based on pixel values ​​of a region of the subtraction image 34 that captures the bone 92 and pixel values ​​of a region 41 that captures the soft tissue surrounding the bone 92. If artifacts 84 resulting from intestinal gas 96 are captured in either the region capturing the bone 92 or the region capturing the soft tissue 41, the accuracy of the bone density calculation decreases. Therefore, with the above configuration, the positions of the intestinal gas 96 captured in the first tube voltage radiographic image 31b and the intestinal gas 96 captured in the second tube voltage radiographic image 32b can be accurately aligned using either the first movement vector 21a or the second movement vector 21b. Therefore, the occurrence of artifacts 84 resulting from misalignment of the intestinal gas 96 in the subtraction image 34 (second subtraction image 34b) can be suppressed. As a result, it is possible to prevent artifacts 84 from appearing in the subtraction image 34 (second subtraction image 34b), thereby preventing a decrease in the accuracy of bone density calculation. Note that the fact that it is possible to prevent artifacts 84 caused by intestinal gas 96 from appearing in the subtraction image 34 (second subtraction image 34b) has been confirmed in an experiment described below by the present inventors.

[0109] Other effects of the second embodiment are similar to those of the first embodiment.

[0110] Second Example (Effect Confirmation Experiment) In order to confirm the effect of the second embodiment described above, the following experiment was conducted: That is, the difference in artifacts 80 caused by the object 91 was compared between the difference image 37 of the second comparative example and the second difference image 34b of the second example.

[0111] 15 is a subtraction image 37 according to a second comparative example, which was obtained by subtracting the first tube voltage radiographic image 31b from the second tube voltage radiographic image 32b without aligning the position of the intestinal gas 96, which is the object 91, between the first tube voltage radiographic image 31b and the second tube voltage radiographic image 32b. In the subtraction image 37 according to the second comparative example, an artifact 84 caused by the intestinal gas 96 has occurred at the position of the region 42.

[0112] 16 is a second subtraction image 34b according to the second embodiment, which is obtained by aligning the object 91 in the first tube voltage radiographic image 31b with the object 91 in the second tube voltage radiographic image 32b based on the movement vector 21 generated by the trained model 20, and then subtracting the aligned first tube voltage radiographic image 36 from the second tube voltage radiographic image 32b. In the second subtraction image 34b according to the second embodiment, an artifact 84 caused by intestinal gas 96 has occurred at the position of the region 43.

[0113] Fig. 17 is an enlarged image 38 obtained by enlarging a region 42 of the difference image 37 according to the second comparative example. Fig. 18 is an enlarged image 34c obtained by enlarging a region 43 of the second difference image 34b according to the second embodiment.

[0114] Comparing the number and area of ​​artifacts 84 caused by intestinal gas 96 between enlarged image 34c and enlarged image 38, it was confirmed that enlarged image 34c had fewer artifacts 84 caused by intestinal gas 96 and also had a smaller area than enlarged image 38. In other words, it was confirmed that it was possible to suppress the occurrence of artifacts 84 caused by intestinal gas 96 in subtraction image 34 (second subtraction image 34b).

[0115] [Modifications] The embodiments and examples disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the description of the embodiments and examples above, and further includes all modifications (modifications) within the meaning and scope equivalent to the claims.

[0116] For example, in the first and second embodiments described above, the control unit 2 (201) included in the radiographic apparatus 100 (200) aligns the position of the object 91 in one of the first radiographic image 31 and the second radiographic image 32 with the position of the object 91 in the other image based on the movement vector 21 generated by the trained model 20, and acquires a difference image 34 obtained by subtracting one of the first radiographic image 31 and the second radiographic image 32 from the other after the alignment. However, the present invention is not limited to this. For example, as in an image processing device 300 according to a modified example shown in FIG. 19 , a device other than the radiographic apparatus 100 (200) may be configured to acquire the difference image 34 based on the movement vector 21 generated by the trained model 20.

[0117] The image processing device 300 according to the modified example includes a control unit 301, a storage unit 3, a display unit 4, an input reception unit 5, and an image acquisition unit 11. The storage unit 3, the display unit 4, and the input reception unit 5 have the same configurations as the storage unit 3, the display unit 4, and the input reception unit 5 included in the radiation imaging device 100 according to the first embodiment, and therefore detailed description thereof will be omitted.

[0118] The image acquisition unit 11 is configured to acquire a first radiographic image 31 of the subject 90 captured by a radiation imaging unit 1 including a radiation irradiation unit 1a that irradiates radiation onto the subject 90 and a radiation detection unit 1b that detects radiation that has passed through the subject 90, and a second radiographic image 32 that is captured at a different timing from the first radiographic image 31. The image acquisition unit 11 is, for example, an input / output interface.

[0119] The control unit 301 differs from the control unit 2 of the first embodiment in that it has a motion vector acquisition unit 301a, a position alignment processing unit 301b, a difference processing unit 301c, and a motion image generation unit 301d instead of the motion vector acquisition unit 2a, the position alignment processing unit 2b, the difference processing unit 2c, and the motion image generation unit 2d.

[0120] The movement vector acquisition unit 301a is configured to input the first radiographic image 31 and the second radiographic image 32 acquired by the image acquisition unit 11 into the learned model 20 stored in the memory unit 3, and to acquire the movement vector 21 of the object 91 between the first radiographic image 31 and the second radiographic image 32.

[0121] The movement vector acquisition unit 301a according to the modified example differs from the movement vector acquisition unit 2a according to the first embodiment in that the movement vector acquisition unit 301a acquires the movement vector 21 using the first radiographic image 31 and the second radiographic image 32 acquired by the image acquisition unit 11. In other words, the movement vector acquisition unit 301a and the movement vector acquisition unit 2a differ only in that the latter uses images captured by the radiation imaging unit 1 or images acquired by the image acquisition unit 11, and the configuration for acquiring the movement vector 21 is the same.

[0122] The alignment processing unit 301b, the difference processing unit 301c, and the moving image generating unit 301d differ only in whether they use images captured by the radiation imaging unit 1 or images acquired by the image acquisition unit 11, and the configuration for performing actual processing is the same as the configuration of the alignment processing unit 2b, the difference processing unit 2c, and the moving image generating unit 301d in the first embodiment described above.

[0123] That is, in the image processing device 300 according to the modified example, the control unit 301 is configured to align the position of the object 91 in either the first radiographic image 31 or the second radiographic image 32 with the position of the object 91 in the other image based on the movement vector 21 acquired by the movement vector acquisition unit 301a, and to acquire a difference image 34 obtained by subtracting one of the first radiographic image 31 and the second radiographic image 32 from the other after the alignment.

[0124] If a moving image of the difference image 34 is not generated, the control unit 301 does not need to include the moving image generating unit 301d.

[0125] By configuring in this manner, an image processing device 300 can be provided that, similar to the radiation imaging device 100 (200) according to the first and second embodiments described above, is capable of reducing artifacts 80 (81) in the differential image 34 that are caused by misalignment of the object 91.

[0126] Furthermore, in the first and second embodiments, an example was described in which the trained model 20 is configured to generate both the first movement vector 21a and the second movement vector 21b. However, the present invention is not limited to this. The trained model 20 may be configured to generate only one of the first movement vector 21a and the second movement vector 21b. However, in a configuration in which a moving image of the difference image 34 is generated, as in the first embodiment, it is necessary to align the first radiographic image 31 with the second radiographic image 32 using the second movement vector 21b. Therefore, if the trained model 20 is configured to generate only the second movement vector 21b, it will be impossible to generate a moving image of the difference image 34. Therefore, it is preferable that the trained model 20 be configured to generate both the first movement vector 21a and the second movement vector 21b.

[0127] In the first and second embodiments, the motion vector acquisition unit 2a (201a) acquires either the first motion vector 21a or the second motion vector 21b, but the present invention is not limited to this. The motion vector acquisition unit may be configured to acquire both the first motion vector and the second motion vector.

[0128] In the first embodiment, the first radiographic image 31 is a non-contrast agent image 31a and the second radiographic image 32 is a contrast agent image 32a, and in the second embodiment, the first radiographic image 31 is a first tube voltage radiographic image 31b and the second radiographic image 32 is a second tube voltage radiographic image 32b. However, the present invention is not limited to this. The first radiographic image and the second radiographic image may be images using radiation other than X-rays. For example, the first radiographic image and the second radiographic image may be images using electron beams or particle beams other than X-rays as radiation.

[0129] In the first embodiment, the moving image generator 2d generates moving images at 15 fps, but the present invention is not limited to this. For example, the frame rate of the moving images generated by the moving image generator may be 7.5 fps. The frame rate of the moving images generated by the moving image generator may be any frame rate as long as it is appropriate for the moving images to be displayed when an operator, such as a doctor, treats the subject 90 using the catheter 94.

[0130] In the first embodiment, the motion vector acquisition unit 2 a acquires the first motion vector 21 a, and the registration processing unit 2 b registers the non-contrast agent image 31 a with the contrast agent image 32 a based on the first motion vector 21 a. However, the present invention is not limited to this. For example, if no moving images are acquired, the motion vector acquisition unit may acquire the second motion vector, and the registration processing unit may register the contrast agent image with the non-contrast agent image based on the second motion vector.

[0131] In the second embodiment, the motion vector acquisition unit 201a acquires the first motion vector 21a, and the alignment processing unit 201b aligns the first tube voltage radiographic image 31b with the second tube voltage radiographic image 32b based on the first motion vector 21a. However, the present invention is not limited to this. For example, the motion vector acquisition unit may acquire the second motion vector, and the alignment processing unit may align the second tube voltage radiographic image with the first tube voltage radiographic image based on the second motion vector. In this case, the subtraction processing unit may be configured to subtract the aligned second tube voltage radiographic image from the first tube voltage radiographic image to obtain a subtraction image.

[0132] Furthermore, in the first and second embodiments, an example has been described in which the alignment processing unit 2b (201b) aligns the first radiographic image 31 and the second radiographic image 32 using only the first movement vector 21a. However, the present invention is not limited to this. For example, the alignment processing unit may be configured to align the first radiographic image and the second radiographic image based on both the first movement vector and the second movement vector. In this case, the alignment processing unit may be configured to align the first radiographic image and the second radiographic image based on a movement vector obtained by weighting the first movement vector and the second movement vector, as shown in Equation (1) below. Here, V_total is a motion vector obtained by weighting the first motion vector and the second motion vector. α is a weighting coefficient for the first motion vector. β is a weighting coefficient for the second motion vector. V_1 is the first motion vector. V_2 is the second motion vector. α and β are, for example, "0.5."

[0133] Here, for example, when aligning the first radiographic image and the second radiographic image based on a movement vector obtained by weighting the first movement vector and the second movement vector, changing the combination of positive and negative values ​​of the weighting coefficients makes it possible to easily obtain the movement vector of the object between the first radiographic image and the second radiographic image, and the movement vector of the object between the second radiographic image and the first radiographic image. Therefore, with the above configuration, in a configuration where it is necessary to align the object with a later acquired image, such as a moving image, the first radiographic image and the second radiographic image can be aligned using the movement vector of the object between the first radiographic image and the second radiographic image. Furthermore, in a configuration where it is acceptable to align the object with either image, such as a still image, the first radiographic image and the second radiographic image can be aligned using either the movement vector of the object between the first radiographic image and the second radiographic image, or the movement vector of the object between the second radiographic image and the first radiographic image. Therefore, whether the radiographic images to be aligned are moving images or still images, it is possible to prevent artifacts from occurring in the acquired subtraction image, thereby improving user convenience (usability).

[0134] In the first and second embodiments, for convenience of explanation, the processing performed by the control unit 2 (201) is described using a flow-driven flowchart in which the processing is performed in order according to a processing flow, but the present invention is not limited to this. In the present invention, the processing performed by the control unit may be an event-driven processing in which processing is performed on an event-by-event basis. In this case, the processing may be completely event-driven, or may be a combination of event-driven and flow-driven.

[0135] Aspects It will be appreciated by those skilled in the art that the exemplary embodiments described above are examples of the following aspects.

[0136] (Item 1) A radiological imaging system includes a radiation irradiator that irradiates radiation onto a subject and a radiation detector that detects radiation that has passed through the subject, and images the subject; a storage that stores a trained model that receives a first image and a second image captured at different times to generate a motion vector of an object between the first image and the second image, and generates an image at an intermediate point between the time of capturing the first image and the time of capturing the second image based on the motion vector; and a control unit that inputs the first radiological image of the subject captured by the radiological imaging unit and a second radiological image captured at a different time from the first radiological image into the trained model stored in the storage unit, and acquires the motion vector of the object between the first radiological image and the second radiological image, the control unit is configured to align the position of the object in one of the first radiographic image and the second radiographic image with the position of the object in the other image based on the movement vector acquired by the movement vector acquisition unit, and to acquire a difference image by subtracting one of the first radiographic image and the second radiographic image from the other after the alignment.

[0137] (Item 2) The radiographic apparatus of item 1, wherein the trained model is configured to generate both a first movement vector, which is the movement vector of the object between the first image and the second image, and a second movement vector, which is the movement vector of the object between the second image and the first image; the movement vector acquisition unit is configured to acquire at least one of the first movement vector and the second movement vector generated by inputting the first radiographic image and the second radiographic image to the trained model; and the control unit is configured to align the first radiographic image and the second radiographic image based on either the first movement vector or the second movement vector or both the first movement vector and the second movement vector.

[0138] (Item 3) The radiographic imaging device according to Item 2, wherein the object is any one of a bone, a diaphragm, and intestinal gas that appears in the first radiographic image and the second radiographic image, and the movement vector acquisition unit is configured to acquire the movement vector of any one of the bone, the diaphragm, and the intestinal gas between the first radiographic image and the second radiographic image.

[0139] (Item 4) The radiographic apparatus of item 3, wherein the first radiographic image is a non-contrast image captured without using a contrast agent, the second radiographic image is a contrast image captured using a contrast agent, the object is either the bone or the diaphragm, the movement vector acquisition unit is configured to acquire the first movement vector of the object in the contrast image generated by inputting the non-contrast image and the contrast image into the trained model, the movement vector being of the object having moved from a position in the non-contrast image, and the control unit is configured to align the non-contrast image with the contrast image based on the first movement vector, and to acquire the difference image by subtracting the aligned non-contrast image from the contrast image.

[0140] (Item 5) The radiation imaging device according to Item 4, wherein the radiation imaging unit is configured to capture moving images of the subject, the motion vector acquisition unit is configured to acquire the first motion vector of the object in each of the plurality of contrast agent images that has moved from a position in the non-contrast agent image each time the plurality of contrast agent images are captured, and the control unit is configured to align the non-contrast agent image with each of the plurality of contrast agent images based on each of the first motion vectors, and to acquire the plurality of difference images by subtracting the aligned non-contrast agent image from each of the plurality of contrast agent images.

[0141] (Item 6) The first radiographic image is a first tube voltage radiographic image captured at a predetermined tube voltage, the second radiographic image is a second tube voltage radiographic image captured at a tube voltage lower than that of the first tube voltage radiographic image, and the object is the intestinal gas, and the movement vector acquisition unit is configured to acquire at least one of the first movement vector of the object moved in the first tube voltage radiographic image from a position in the second tube voltage radiographic image and the second movement vector of the object moved in the second tube voltage radiographic image from a position in the first tube voltage radiographic image, which are generated by inputting the first tube voltage radiographic image and the second radiographic image to the trained model, Item 4. The radiographic apparatus according to item 3, wherein the control unit is configured to: align the first tube voltage radiographic image with the second tube voltage radiographic image based on the first movement vector, and obtain the difference image by subtracting the aligned first tube voltage radiographic image from the second tube voltage radiographic image; or align the second tube voltage radiographic image with the first tube voltage radiographic image based on the second movement vector, and obtain the difference image by subtracting the aligned second tube voltage radiographic image from the first tube voltage radiographic image.

[0142] (Item 7) A radiological imaging system comprising: an image acquisition unit that acquires a first radiological image of a subject captured by a radiation imaging unit including a radiation irradiator that irradiates radiation onto the subject and a radiation detector that detects radiation that has passed through the subject, and a second radiological image that is captured at a timing different from that of the first radiological image; a storage unit that stores a trained model that generates a movement vector of an object between the first image and the second image by inputting the first image and the second image captured at different timings, and generates an image at an intermediate point in time between the capture of the first image and the capture of the second image based on the movement vector; and a control unit that inputs the first radiological image and the second radiological image captured by the image acquisition unit into the trained model stored in the storage unit, and acquires the movement vector of the object between the first radiological image and the second radiological image, the control unit is configured to align a position of the object in one of the first radiographic image and the second radiographic image with a position of the object in the other image based on the movement vector acquired by the movement vector acquisition unit, and to acquire a difference image by subtracting one of the first radiographic image and the second radiographic image from the other after the alignment.

[0143] (Item 8) An image processing method comprising: acquiring a first radiographic image of a subject captured by a radiographic imaging unit including a radiation irradiator that irradiates the subject with radiation and a radiation detector that detects radiation that has passed through the subject, and a second radiographic image captured at a timing different from that of the first radiographic image; generating a movement vector of an object between the first image and the second image by inputting the first image and the second image captured at different timings, and inputting the first radiographic image and the second radiographic image into a trained model that generates an image at an intermediate point in time between the capture of the first image and the capture of the second image based on the movement vector, thereby acquiring the movement vector of the object between the first radiographic image and the second radiographic image; and aligning a position of the object in one of the first radiographic image and the second radiographic image with a position of the object in the other image based on the acquired movement vector, and acquiring a difference image by subtracting one of the first radiographic image and the second radiographic image after alignment from the other.

[0144] REFERENCE SIGNS LIST 1 Radiation imaging unit 1a Radiation irradiation unit 1b Radiation detection unit 2, 201, 301 Control unit 2a, 201a, 301a Movement vector acquisition unit 11 Image acquisition unit 20 Trained model 21 Movement vector 21a First movement vector 21b Second movement vector 30a First image 30b Second image 30c Image at an intermediate point between the time of capturing the first image and the time of capturing the second image 31 First radiographic image 31a Non-contrast agent image 31b First tube voltage radiographic image 32 Second radiographic image 32a Contrast agent image 32b Second tube voltage radiographic image 33 Non-contrast agent image after alignment (first radiographic image after alignment) 34 Difference image 36 First tube voltage radiographic image after alignment (first radiographic image after alignment) 60 Object 90 Subject 91 Object 92 Bone 93 Diaphragm 96 Intestinal gas 100, 200 Radiography device 300 Image processing device

Claims

1. A radiological imaging unit that images the subject, including a radiation irradiation unit that irradiates radiation onto the subject and a radiation detection unit that detects radiation that has passed through the subject; a memory unit that stores a trained model that receives a first image and a second image captured at different times to generate a movement vector of an object between the first image and the second image, and generates an image at an intermediate point between the time of capturing the first image and the time of capturing the second image based on the movement vector; and a control unit that inputs the first radiological image of the subject captured by the radiological imaging unit and a second radiological image captured at a different time from the first radiological image into the trained model stored in the memory unit, and acquires the movement vector of the object between the first radiological image and the second radiological image, the control unit is configured to align the position of the object in one of the first radiographic image and the second radiographic image with the position of the object in the other image based on the movement vector acquired by the movement vector acquisition unit, and to acquire a difference image by subtracting one of the first radiographic image and the second radiographic image from the other after the alignment.

2. The radiographic imaging device of claim 1, wherein the trained model is configured to generate both a first movement vector, which is the movement vector of the object between the first image and the second image, and a second movement vector, which is the movement vector of the object between the second image and the first image; the movement vector acquisition unit is configured to acquire at least one of the first movement vector and the second movement vector generated by inputting the first radiographic image and the second radiographic image into the trained model; and the control unit is configured to align the first radiographic image and the second radiographic image based on either the first movement vector or the second movement vector, or both the first movement vector and the second movement vector.

3. The radiographic imaging device of claim 2, wherein the object is any one of a bone, a diaphragm, and intestinal gas that appears in the first radiographic image and the second radiographic image, and the movement vector acquisition unit is configured to acquire the movement vector of any one of the bone, the diaphragm, and intestinal gas between the first radiographic image and the second radiographic image.

4. The radiographic apparatus of claim 3, wherein the first radiographic image is a non-contrast image captured without using a contrast agent, the second radiographic image is a contrast image captured using a contrast agent, the object is either the bone or the diaphragm, the movement vector acquisition unit is configured to acquire the first movement vector of the object in the contrast image generated by inputting the non-contrast image and the contrast image into the trained model, the object having moved from its position in the non-contrast image, and the control unit is configured to align the non-contrast image with the contrast image based on the first movement vector, and to acquire the difference image by subtracting the aligned non-contrast image from the contrast image.

5. The radiation imaging device of claim 4, wherein the radiation imaging unit is configured to capture moving images of the subject, the movement vector acquisition unit is configured to acquire the first movement vector of the object in each of the plurality of contrast agent images that has moved from its position in the non-contrast agent image each time the plurality of contrast agent images are captured, and the control unit is configured to align the non-contrast agent image with each of the plurality of contrast agent images based on each of the first movement vectors, and to acquire the plurality of difference images by subtracting the aligned non-contrast agent image from each of the plurality of contrast agent images.

6. The first radiographic image is a first tube voltage radiographic image captured at a predetermined tube voltage, the second radiographic image is a second tube voltage radiographic image captured at a tube voltage lower than that of the first tube voltage radiographic image, and the object is the intestinal gas, and the movement vector acquisition unit is configured to acquire at least one of the first movement vector of the object in the first tube voltage radiographic image, which has moved from a position in the second tube voltage radiographic image, and the second movement vector of the object in the second tube voltage radiographic image, which are generated by inputting the first tube voltage radiographic image and the second radiographic image into the trained model, 4. The radiographic apparatus according to claim 3, wherein the control unit is configured to: align the first tube voltage radiographic image with the second tube voltage radiographic image based on the first movement vector, and subtract the aligned first tube voltage radiographic image from the second tube voltage radiographic image to obtain the difference image; or align the second tube voltage radiographic image with the first tube voltage radiographic image based on the second movement vector, and subtract the aligned second tube voltage radiographic image from the first tube voltage radiographic image to obtain the difference image.

7. An imaging system including: an image acquisition unit that acquires a first radiographic image of the subject captured by a radiographic imaging unit including a radiation irradiation unit that irradiates radiation onto the subject and a radiation detection unit that detects radiation that has passed through the subject; and a second radiographic image that is captured at a timing different from that of the first radiographic image; a memory that stores a trained model that generates a movement vector of an object between the first image and the second image by inputting the first image and the second image captured at different timings, and generates an image at an intermediate point in time between the capture of the first image and the capture of the second image based on the movement vector; and a control unit that inputs the first radiographic image and the second radiographic image captured by the image acquisition unit into the trained model stored in the memory unit, and acquires the movement vector of the object between the first radiographic image and the second radiographic image, the control unit is configured to align a position of the object in one of the first radiographic image and the second radiographic image with a position of the object in the other image based on the movement vector acquired by the movement vector acquisition unit, and to acquire a difference image by subtracting one of the first radiographic image and the second radiographic image from the other after the alignment.

8. An image processing method comprising the steps of: acquiring a first radiographic image of a subject captured by a radiographic imaging unit including a radiation irradiation unit that irradiates the subject with radiation and a radiation detection unit that detects radiation that has passed through the subject, and a second radiographic image that is captured at a different time from the first radiographic image; generating a movement vector of an object between the first image and the second image by inputting the first image and the second image that are captured at different times, and inputting the first radiographic image and the second radiographic image into a trained model that generates an image at an intermediate point in time between the capture of the first image and the capture of the second image based on the movement vector, thereby acquiring a movement vector of the object between the first radiographic image and the second radiographic image; and aligning a position of the object in one of the first radiographic image and the second radiographic image with a position of the object in the other image based on the acquired movement vector, and acquiring a difference image by subtracting one of the first radiographic image and the second radiographic image after alignment from the other.

Images