Image generation device, image generation method, and program

Abstract

An image generation device, an image generation method, and a program capable of improving the generation accuracy of a pseudo two dimensional image as compared with a case where an image generation model learned using a combination of a plurality of projection images or a plurality of tomographic images and a normal two dimensional image captured separately from tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images is used are obtained.SOLUTION: The image processing apparatus 16 as an image generation apparatus includes an image generation model that is learned in advance by a plurality of combinations of a plurality of projection images or a plurality of tomographic images and a normal two dimensional image captured by irradiating the breast compressed by the compression member with radiation in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images.SELECTED DRAWING: Figure 5

Description

[Technical Field]

[0001] The present disclosure relates to an image generation device, an image generation method, and a program. [Background technology]

[0002] Patent Document 1 discloses a learning device for an image generation model that aims to generate pseudo two-dimensional images that accurately reproduce the shape of breast lesions.

[0003] This learning device is a learning device for an image generation model that generates a pseudo 2D image from a series of multiple projection images obtained by tomosynthesis imaging of the breast or multiple tomographic images obtained from the series of multiple projection images, and is equipped with at least one processor. The processor acquires normal 2D images taken by irradiating the breast with radiation, and detects a first region of interest including calcifications in the breast and a second region of interest including lesions other than the calcifications based on either a synthetic 2D image obtained by synthesizing at least a portion of the series of multiple projection images or the multiple tomographic images, the tomographic image, and the normal 2D image. The processor then updates the weights of the network of the image generation model to reduce the loss based on the loss between the pseudo 2D image output by the image generation model, in which the weight of the first region of interest is set to be the heaviest and the weight of the second region of interest is set to be equal to or greater than the weights of other regions other than the first region of interest and the second region of interest, and the normal 2D image or the synthetic 2D image, or both. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2023-51400 Summary of the Invention [Problem to be solved by the invention]

[0005] Incidentally, the combination of images obtained by tomosynthesis imaging and standard 2D images used in training the image generation model is preferably a combination of images of the same breast obtained under the same compression conditions of the breast by the compression member. If the combination of images used in training is a combination of different breast images or images under different compression conditions, the breast conditions, such as size, thickness, shape, and position, will differ between the breast shown in the image obtained by tomosynthesis imaging and the breast shown in the standard 2D image. For this reason, training using such a combination of images can cause a decrease in the accuracy of generating pseudo 2D images by the image generation model.

[0006] However, in the technology disclosed in Patent Document 1, although the combinations of images obtained by tomosynthesis imaging and normal 2D images used for learning are combinations of images obtained by imaging the same breast, they include combinations of images that are not obtained under the same compression conditions. Therefore, the technology disclosed in Patent Document 1 has a problem in that the accuracy of generating pseudo 2D images using a trained image generation model is not necessarily high.

[0007] The present disclosure has been made in consideration of the above circumstances, and aims to provide an image generation device, an image generation method, and a program that can improve the accuracy of generating pseudo-two-dimensional images compared to using an image generation model trained using a combination of multiple projection images or multiple tomographic images and normal two-dimensional images taken separately from the tomosynthesis imaging that obtains the multiple projection images or multiple tomographic images. [Means for solving the problem]

[0008] In order to achieve the above object, an image generating device of a first aspect of the present disclosure is an image generating device that generates a pseudo two-dimensional image from a series of multiple projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, and is equipped with at least one image generation model, which has been trained in advance using multiple combinations of multiple projection images or multiple tomographic images and normal two-dimensional images taken by irradiating the breast while it is compressed by a compression member in tomosynthesis imaging to obtain the multiple projection images or multiple tomographic images.

[0009] An image generating device of a second aspect of the present disclosure is the image generating device of the first aspect, further comprising a processor, wherein the processor acquires multiple combinations of multiple projection images or multiple tomographic images and a normal two-dimensional image for each of predetermined divisions, and learns an image generation model individually for each division, thereby learning multiple image generation models.

[0010] An image generating device according to a third aspect of the present disclosure is the image generating device according to the second aspect, wherein the predetermined section is a section relating to a subject.

[0011] An image generating device of a fourth aspect of the present disclosure is an image generating device of the third aspect, wherein the categories relating to the subject include at least one category of the subject's height, weight, BMI (Body Mass Index), age, mammary gland information, and breast thickness.

[0012] An image generating device according to a fifth aspect of the present disclosure is the image generating device according to the second aspect, wherein the predetermined divisions are divisions relating to settings at the time of shooting.

[0013] An image generating device of a sixth aspect of the present disclosure is an image generating device of the fifth aspect, in which the categories related to the settings during imaging include at least one category of radiation irradiation angle in tomosynthesis imaging, resolution of the captured image, and radiation dose mode.

[0014] An image generating device of a seventh aspect of the present disclosure is an image generating device of the sixth aspect, in which the radiation dose mode classification is based on a combination of a first dose mode used during tomosynthesis imaging and a second dose mode used during normal two-dimensional imaging, the second dose mode having a dose equal to or greater than that of the first dose mode.

[0015] An image generating device of an eighth aspect of the present disclosure is an image generating device of the second aspect, wherein the predetermined division is a division related to the type of image processing technology for at least one of multiple projection images or multiple tomographic images and normal two-dimensional images.

[0016] An image generating device according to a ninth aspect of the present disclosure is the image generating device according to the eighth aspect, wherein the classification relating to the type of image processing technology is a classification for each manufacturer of an imaging device that performs tomosynthesis imaging.

[0017] An image generating device of a tenth aspect of the present disclosure is an image generating device of any one of the second to ninth aspects, in which a processor prohibits the use of an image of at least one of multiple combinations of multiple projection images or multiple tomographic images and a normal two-dimensional image for learning if the image quality of that combination is lower than a predetermined level.

[0018] An image generating device of an eleventh aspect of the present disclosure is an image generating device of the second aspect, in which, when a processor uses multiple tomographic images for learning, the multiple tomographic images are corrected so that they become tomographic images that would appear if a tube were virtually placed at the position of the tube that emitted radiation when capturing the corresponding normal two-dimensional image.

[0019] An image generating device of a twelfth aspect of the present disclosure is an image generating device in which, in the image generating device of the second aspect, a processor performs set imaging in which tomosynthesis imaging and normal 2D image imaging are performed under the same pressure by a compression member, and when a set image is obtained that is a combination of multiple projection images or multiple tomographic images and a normal 2D image, the set image is used as learning data for an image generation model.

[0020] An image generating device according to a thirteenth aspect of the present disclosure is the image generating device according to the twelfth aspect, wherein the processor accumulates a set of images to use in training the image generation model.

[0021] An image generating device according to a fourteenth aspect of the present disclosure is the image generating device according to the twelfth or thirteenth aspect, wherein the processor sequentially updates the image generation model using a set of images.

[0022] An image generating device of a fifteenth aspect of the present disclosure is an image generating device of the second aspect, in which a processor acquires a series of multiple projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, as images for generating a pseudo-two-dimensional image, and inputs the acquired images for generation into an image generation model to generate a pseudo-two-dimensional image corresponding to the images for generation.

[0023] An image generating device of a 16th aspect of the present disclosure is an image generating device of the 15th aspect, in which, when there are multiple image generation models trained individually for each predetermined division, the processor selectively uses an image generation model of a division from the multiple image generation models that corresponds to the pseudo 2D image to be generated to generate a pseudo 2D image.

[0024] An image generating device of a 17th aspect of the present disclosure is an image generating device of the first or second aspect, in which the image generation model is provided in an imaging device that performs only tomosynthesis imaging, or in an imaging device that performs tomosynthesis imaging and mammography imaging separately.

[0025] In addition, in order to achieve the above-mentioned object, the image generation method of the 18th aspect of the present disclosure is an image generation method in which a computer executes a process of generating a pseudo two-dimensional image using an image generation model from a series of multiple projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, and the image generation model has been trained in advance using multiple combinations of multiple projection images or multiple tomographic images and normal two-dimensional images taken by irradiating the breast while it is compressed by a compression member in tomosynthesis imaging to obtain the multiple projection images or multiple tomographic images.

[0026] In addition, in order to achieve the above-mentioned object, the program of the 19th aspect of the present disclosure is a program that causes a computer to execute a process of generating a pseudo two-dimensional image using an image generation model from a series of multiple projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, and the image generation model has been trained in advance using multiple combinations of multiple projection images or multiple tomographic images and normal two-dimensional images taken by irradiating the breast while it is compressed by a compression member in tomosynthesis imaging to obtain the multiple projection images or multiple tomographic images. [Effects of the Invention]

[0027] According to the present disclosure, the accuracy of generating pseudo-two-dimensional images can be improved compared to using an image generation model trained using a combination of multiple projection images or multiple tomographic images and normal two-dimensional images taken separately from the tomosynthesis imaging used to obtain the multiple projection images or multiple tomographic images. [Brief explanation of the drawings]

[0028] [Figure 1] 1 is a configuration diagram schematically illustrating an example of the overall configuration of a radiation image capturing system according to an embodiment; [Figure 2] FIG. 1 is a diagram illustrating an example of tomosynthesis imaging. [Figure 3] FIG. 1 is a diagram illustrating an example of an image generation model. [Figure 4] FIG. 4 is a diagram illustrating an example of an intermediate layer of the image generation model illustrated in FIG. 3. [Figure 5] FIG. 1 is a block diagram illustrating an example of a configuration of an image processing apparatus according to an embodiment. [Figure 6] FIG. 1 is a schematic diagram for explaining an outline of the learning flow of each image generation model of an image generation model group in an image processing apparatus according to an embodiment. [Figure 7] FIG. 2 is a functional block diagram of an example of a configuration related to a function for learning an image generation model in the image processing apparatus according to the embodiment. [Figure 8] FIG. 2 is a schematic diagram illustrating an example of a configuration of a model selection information database according to the embodiment. [Figure 9] 10 is a flowchart illustrating an example of the flow of a learning process performed by the image processing apparatus according to the embodiment. [Figure 10] 1 is a schematic diagram for explaining an outline of a flow of generating a pseudo two-dimensional image using an image generation model in an image processing apparatus according to an embodiment. [Figure 11] 2 is a functional block diagram of an example of a configuration related to a function for generating a pseudo two-dimensional image in the image processing apparatus according to the embodiment. FIG. [Figure 12] 10 is a flowchart illustrating an example of the flow of an image generation process performed by the image processing apparatus according to the embodiment. [Figure 13] FIG. 10 is a schematic diagram illustrating an example of another configuration of a model selection information database according to the embodiment. [Figure 14] FIG. 2 is a schematic diagram illustrating an example of a configuration of a model selection information database according to the embodiment. [Figure 15] FIG. 10 is a schematic diagram illustrating an example of another configuration of a model selection information database according to the embodiment. [Figure 16] FIG. 2 is a schematic diagram illustrating an example of a configuration of a model selection information database according to the embodiment. [Figure 17]FIG. 10 is a schematic diagram illustrating an example of another configuration of a model selection information database according to the embodiment. [Figure 18A] 4A and 4B are diagrams showing an example of a radiation irradiation position and a normal two-dimensional image when normal imaging is performed according to the embodiment. [Figure 18B] 10A and 10B are diagrams illustrating an example of a radiation irradiation position and a projected image when performing tomosynthesis imaging according to an embodiment, in which the reference position is shifted from that during normal imaging. DETAILED DESCRIPTION OF THE INVENTION

[0029] Hereinafter, embodiments of the present disclosure will be described in detail with reference to the drawings. Note that the technology of the present disclosure is not limited to the embodiments.

[0030] [First embodiment] First, an example of the overall configuration of a radiographic imaging system according to this embodiment to which the technology of the present disclosure is applied will be described. Fig. 1 shows a configuration diagram illustrating an example of the overall configuration of a radiographic imaging system 1 according to this embodiment. As shown in Fig. 1, the radiographic imaging system 1 according to this embodiment includes a mammography apparatus 10, a console 12, a PACS (Picture Archiving and Communication Systems) 14, and an image processing device 16. The console 12, the PACS 14, and the image processing device 16 are connected via a network 17 by wired or wireless communication.

[0031] First, the mammography device 10 of this embodiment will be described. Figure 1 shows a side view of an example of the appearance of the mammography device 10 of this embodiment. Note that Figure 1 shows an example of the appearance of the mammography device 10 when viewed from the left side of the subject.

[0032] The mammography device 10 of this embodiment operates under the control of the console 12 and is an apparatus that takes a radiographic image of a subject's breast by irradiating the breast with radiation R (e.g., X-rays) from a radiation source 29. The mammography device 10 of this embodiment also has the function of performing normal imaging, in which imaging is performed with the radiation source 29 positioned at an irradiation position normal to the detection surface 20A (see also FIG. 2) of the radiation detector 20, and so-called tomosynthesis imaging (described in detail below), in which imaging is performed by moving the radiation source 29 to each of multiple irradiation positions.

[0033] As shown in FIG. 1, the mammography apparatus 10 includes an imaging table 24, a base 26, an arm 28, and a compression unit 32.

[0034] A radiation detector 20 is disposed inside the imaging table 24. As shown in Fig. 2, in the mammography apparatus 10 of this embodiment, when imaging is performed, the breast U of the subject is positioned on the imaging surface 24A of the imaging table 24 by the user.

[0035] The radiation detector 20 detects radiation R that has passed through the subject's breast U. In detail, the radiation detector 20 detects radiation R that has entered the subject's breast U and the imaging table 24 and reached the detection surface 20A of the radiation detector 20, generates a radiographic image based on the detected radiation R, and outputs image data representing the generated radiographic image. Hereinafter, the series of operations of irradiating radiation R from the radiation source 29 and generating a radiographic image by the radiation detector 20 may be referred to as "imaging." The type of radiation detector 20 in this embodiment is not particularly limited, and may be, for example, an indirect conversion type radiation detector that converts radiation R into light and then converts the converted light into electric charges, or a direct conversion type radiation detector that directly converts radiation R into electric charges.

[0036] The compression plate 30, which serves as a compression member of the present disclosure and is used to compress the breast when imaging, is attached to a compression unit 32 provided on the imaging table 24, and is moved toward or away from the imaging table 24 (hereinafter referred to as the "up and down direction") by a compression plate drive section (not shown) provided on the compression unit 32. By moving in the up and down direction, the compression plate 30 compresses the breast of the subject between itself and the imaging table 24.

[0037] The arm 28 can rotate relative to the base 26 via the shaft 27. The shaft 27 is fixed to the base 26, and the shaft 27 and arm 28 rotate together. The shaft 27 and the compression unit 32 of the imaging table 24 are each provided with a gear, and by switching between an engaged state and a disengaged state of these gears, the compression unit 32 of the imaging table 24 and the shaft 27 can be switched between a state in which they are connected and rotate together, and a state in which the shaft 27 is separated from the imaging table 24 and rotates freely. Note that the switching between transmitting and disengaging power to the shaft 27 is not limited to the gear, and various mechanical elements can be used. The arm 28 and the imaging table 24 can rotate independently relative to the base 26, with the shaft 27 as the rotation axis.

[0038] When performing tomosynthesis imaging in mammography apparatus 10, radiation source 29 is moved sequentially to each of a plurality of irradiation positions with different irradiation angles by rotating arm unit 28. Radiation source 29 has a radiation tube (not shown) as a tube of the present disclosure that generates radiation R, and the radiation tube is moved to each of the plurality of irradiation positions in accordance with the movement of radiation source 29.

[0039] 2 is a diagram for explaining an example of tomosynthesis imaging. Note that the compression paddle 30 is not shown in FIG. 2. In this embodiment, as shown in FIG. 2, the radiation source 29 is positioned at irradiation positions 19 whose irradiation angles are different in increments of a predetermined angle β. t (t=1, 2, . . . , the maximum value is 7 in FIG. 2), in other words, the radiation detector 20 is moved to a position where the radiation R irradiates the detection surface 20A of the radiation detector 20 at different irradiation angles. tIn the radiation imaging system 1, radiation R is irradiated from the radiation source 29 toward the breast U in response to an instruction from the console 12, and a radiation image is captured by the radiation detector 20. t and move to each of the irradiation positions 19 t In the example of Fig. 2, when tomosynthesis imaging is performed to capture radiation images at each irradiation position 19, seven radiation images are obtained. In the following, in tomosynthesis imaging, a radiation image captured at each irradiation position 19 is referred to as a "projection image" when distinguishing it from other radiation images, and a plurality of projection images captured in one tomosynthesis imaging is referred to as a "series of a plurality of projection images." In addition to projection images, radiation images, regardless of the type such as tomographic images and normal two-dimensional images, which will be described later, are collectively referred to simply as "radiation images." In the following, each irradiation position 19 is referred to as a "projection image." t Projection images taken at irradiation position 19 t For the images corresponding to the above, the symbols representing each image are followed by the irradiation position 19. t The symbol "t" is added to indicate this.

[0040] 2, the irradiation angle of radiation R refers to the angle α formed between a normal CL to the detection surface 20A of the radiation detector 20 and a radiation axis RC. The radiation axis RC is an axis connecting the focal point of the radiation source 29 at each irradiation position 19 and a predetermined position such as the center of the detection surface 20A. Here, the detection surface 20A of the radiation detector 20 is assumed to be a surface that is approximately parallel to the imaging surface 24A.

[0041] On the other hand, when performing normal imaging in the mammography device 10, the radiation source 29 is positioned at the irradiation position 19 where the irradiation angle α is 0 degrees. t (Irradiation position along the normal direction 19 t 2 ), the radiation source 29 irradiates the patient with radiation R, and the radiation detector 20 captures a radiographic image. In this embodiment, a radiographic image captured in normal radiography is referred to as a "normal two-dimensional image" to distinguish it from other radiographic images.

[0042] The mammography device 10 and the console 12 are connected by wired or wireless communication. A radiographic image captured by the radiation detector 20 in the mammography device 10 is output to the console 12 by wired or wireless communication via a communication I / F (Interface) unit (not shown).

[0043] As shown in FIG. 1, the console 12 of this embodiment includes a control unit 40, a storage unit 42, a user I / F unit 44, and a communication I / F unit 46.

[0044] As described above, the control unit 40 of the console 12 has the function of controlling the capture of radiographic images of the breast by the mammography device 10. The control unit 40 may be, for example, a computer system equipped with a CPU (Central Processing Unit), ROM (Read Only Memory), and RAM (Random Access Memory).

[0045] The storage unit 42 has a function of storing information related to radiographic image capture, radiographic images acquired from the mammography apparatus 10, etc. The storage unit 42 is a non-volatile storage unit, such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive).

[0046] The user I / F unit 44 includes input devices such as various buttons and switches operated by users such as technicians in relation to radiographic imaging, and display devices such as lamps and displays that display information about imaging, radiographic images, etc.

[0047] The communication I / F unit 46 communicates various data, such as information related to radiographic image capture and radiographic images obtained by capture, with the mammography apparatus 10 via wired or wireless communication. The communication I / F unit 46 also communicates various data, such as radiographic images, with the PACS 14 and image processing device 16 via the network 17 via wired or wireless communication.

[0048] 1, the PACS 14 of this embodiment includes a storage unit 50 and a communication I / F unit (not shown) that store a radiographic image group 52. The radiographic image group 52 includes radiographic images captured by the mammography device 10 and acquired from the console 12 via the communication I / F unit (not shown).

[0049] The image processing device 16 is used when a doctor or the like (hereinafter simply referred to as "doctor") interprets a radiological image. The image processing device 16 of this embodiment has a function of generating a pseudo two-dimensional image equivalent to a normal two-dimensional image from a plurality of tomographic images using an image generation model. Note that the image processing device 16 of this embodiment is an example of an image generation device of the present disclosure, and the image generation method by the image processing device 16 is an example of an image generation method of the present disclosure.

[0050] First, an example of an image generation model used to generate a pseudo two-dimensional image in the image processing device 16 of this embodiment will be described. FIG. 3 shows an example of a plurality of image generation models (hereinafter simply referred to as "image generation models") 67 included in the image generation model group 66 of this embodiment (see also FIG. 5). The image generation model 67 of this embodiment uses a convolutional neural network (CNN) that has undergone machine learning through deep learning. The image generation model 67 uses a plurality of tomographic images 100 (k images in FIG. 3: tomographic images 1001 to 1002) obtained from a series of a plurality of projection images. k ) is input and a pseudo two-dimensional image 102 is output.

[0051] 3 includes an input layer 200, an intermediate layer 202, and an output layer 204, each of which is provided for each of the k tomographic images 100. In this embodiment, the input layers 200 provided for each of the tomographic images 100 have the same configuration.

[0052] A tomographic image 100 is input to the input layer 200. The input layer 200 includes a plurality of nodes 300, each corresponding to a pixel of the tomographic image 100. The input layer 200 performs convolution processing when propagating information for each pixel (each pixel) of the input tomographic image 100 to the intermediate layer 202. For example, if the size of the image to be processed is 28 pixels x 28 pixels and is grayscale data, the size of the data propagated from the input layer 200 to the intermediate layer 202 is 28 x 28 x 1 = 784.

[0053] Note that the present invention is not limited to this embodiment, and a form in which information is input to the input layer 200 in units of voxels extracted from a plurality of tomographic images 100 may also be used.

[0054] As shown in Fig. 4, the intermediate layer 202 includes an encoder 203e and a decoder 203d. The encoder 203e includes a plurality of convolutional layers conv that perform convolutional processing and pooling layers pool that perform pooling processing, the number of which corresponds to the number of layers of the encoder 203e (the number of layers illustrated in the encoder 203e in Fig. 4 is "2"). Note that the plurality of nodes 3021 illustrated in Fig. 3 correspond to the plurality of nodes included in the first layer convolutional layer conv included in the encoder 203e.

[0055] In the convolution process, a three-dimensional convolution operation is performed on each pixel to output a pixel value Icp(x, y, z) corresponding to each pixel of interest Ip. In this way, three-dimensional output data Dc is output, which includes multiple pieces of output data DIc having pixel values ​​Icp(x, y, z) arranged two-dimensionally. One piece of output data Dc is output for each 3x3x3 filter F. When multiple filters F of different types are used, output data Dc is output for each filter F. The filter F represents a neuron (node) in the convolution layer, and since the features that can be extracted are determined for each filter F, the number of features that can be extracted from one piece of input data D in the convolution layer is the number of filters F.

[0056] In addition, the pooling layer pool performs a pooling process that reduces the original image while preserving its features. In other words, the pooling layer pool performs a pooling process that selects a local representative value, reduces the resolution of the input image, and reduces the image size. For example, if the pooling process, which selects a representative value from a 2x2 pixel block, is performed with a stride of "1," that is, shifted by one pixel, a reduced image that is half the size of the input image is output.

[0057] On the other hand, the decoder 203d includes a plurality of convolution layers conv that perform convolution processing and upsampling layers upsmp that perform upsampling processing, in accordance with the number of layers of the decoder 203d (the number of layers illustrated in the decoder 203d in FIG. 4 is "2"). u correspond to a plurality of nodes included in the first layer convolutional layer conv included in the decoder 203d.

[0058] The convolution layer conv included in the decoder 203d performs the same processing as the convolution layer conv included in the encoder 203e. On the other hand, the upsampling layer upsmp receives the output of the encoder 203e as input and performs processing so that the image size becomes equal to that of the pseudo 2D image 102.

[0059] On the other hand, the output layer 204 is composed of all nodes 302 included in the convolution layer conv arranged at the end of the intermediate layer 202. u The image size in the output layer 204 is the same as the size of the pseudo 2D image 102 output from the image generation model 67, and multiple nodes 304 included in the output layer 204 correspond to each pixel of the pseudo 2D image 102.

[0060] In this way, the image generation model 67 of this embodiment outputs a pseudo two-dimensional image 102 when a plurality of tomographic images 100 are input.

[0061] Fig. 5 is a block diagram showing an example of the configuration of the image processing device 16 of this embodiment. As shown in Fig. 5, the image processing device 16 of this embodiment includes a control unit 60 as a computer, a storage unit 62, a display unit 70, an operation unit 72, and a communication I / F unit 74. The control unit 60, the storage unit 62, the display unit 70, the operation unit 72, and the communication I / F unit 74 are connected via a bus 79 such as a system bus or a control bus so that various information can be exchanged between them.

[0062] The control unit 60 controls the overall operation of the image processing device 16. The control unit 60 includes a CPU 60A as a processor, a ROM 60B, and a RAM 60C. The ROM 60B stores various programs and the like for the CPU 60A to perform control. The RAM 60C temporarily stores various data.

[0063] The storage unit 62 is a non-volatile storage unit, specific examples of which include an HDD, an SSD, etc. The storage unit 62 stores various information such as a learning program 63A, an image generation program 63B, learning data 64 for learning an image generation model 67, an image generation model group 66 composed of multiple image generation models 67, and a model selection information database 68, the details of which will be described later.

[0064] The display unit 70 displays radiographic images and various types of information. The display unit 70 is not particularly limited and may be any of various displays. The operation unit 72 is used by the user to input instructions and various types of information for a doctor to diagnose breast lesions using radiographic images. The operation unit 72 is not particularly limited and may be, for example, any of various switches, a touch panel, a touch pen, a mouse, etc. The display unit 70 and the operation unit 72 may be integrated into a touch panel display.

[0065] The communication I / F unit 74 communicates various types of information with the console 12 and the PACS 14 via the network 17 by wireless or wired communication.

[0066] Next, the functions of the image processing device 16 of this embodiment will be described. There is a learning phase in which each image generation model 67 of the image generation model group 66 is trained, and an operation phase in which a pseudo two-dimensional image is generated from multiple tomographic images by selectively using an image generation model 67 of the image generation model group 66.

[0067] (Learning phase) First, an example of the learning phase of the image processing device 16 of this embodiment will be described. Fig. 6 shows a schematic diagram for explaining an outline of the learning flow of each image generation model 67 of the image generation model group 66 in the image processing device 16 of this embodiment.

[0068] As shown in Figure 6, the training data 64 is composed of a set of multiple tomographic images 101 obtained from a series of multiple projection images obtained by tomosynthesis imaging using the mammography device 10, and a normal two-dimensional image 111 obtained by normal imaging after irradiating radiation R onto the breast U in a state compressed by the compression paddle 30 in the tomosynthesis imaging. In this case, the projection images and the normal two-dimensional image 111 may be captured in either order before or after the normal two-dimensional image 111. Hereinafter, the combination of tomosynthesis imaging and capturing the normal two-dimensional image 111 while the same compression is being performed by the compression paddle 30 is referred to as a "set imaging," and the images obtained by the set imaging are referred to as "set images."

[0069] In the learning phase, a plurality of tomographic images 101 of the learning data 64 are input to each image generation model 67 in the image generation model group 66. In response to this, the image generation model 67 outputs a pseudo two-dimensional image 103 as described above.

[0070] Loss function calculation unit 85 calculates a loss function, which is a function that represents the degree of difference between pseudo 2D image 103 output from image generation model 67 and normal 2D image 111. The closer the value of the loss function is to 0, the more similar pseudo 2D image 103 is to normal 2D image 111, and the more accurately the shape of the lesion is reproduced in pseudo 2D image 103.

[0071] The weight update unit 86 updates the weights of the network in the image generation model 67 in accordance with the loss function calculated by the loss function calculation unit 85. Specifically, in accordance with the loss function calculated by the loss function calculation unit 85, the weights indicating the strength of connections between nodes in the previous and next layers, which are coefficients of each filter F in the image generation model 67 described above, and the weight w of the difference connecting the node 304 in the output layer 204 in each tomographic image 101 to a node in the connection layer, etc. are changed by backpropagation, stochastic gradient descent, etc.

[0072] In the learning phase, a series of processes, including inputting multiple tomographic images 101 of learning data 64 to an image generation model 67, outputting a pseudo two-dimensional image 103 from the image generation model 67, calculating a loss function, and updating the weights, is repeated to reduce the loss function.

[0073] 7 shows a functional block diagram of an example of a configuration related to a function for learning an image generation model 67 in the image processing device 16 of this embodiment. As shown in Fig. 7, the image processing device 16 includes a learning data acquisition unit 80 and a learning unit 84. As an example, in the image processing device 16 of this embodiment, the CPU 60A of the control unit 60 executes a learning program 63A stored in the storage unit 62, so that the CPU 60A functions as the learning data acquisition unit 80 and the learning unit 84.

[0074] The learning data acquisition unit 80 has a function of acquiring learning data 64 from the storage unit 62. Although one set of learning data 64 is illustrated in FIG. 6, in reality, the storage unit 62 stores a sufficient amount of learning data 64 for training the image generation model 67. The learning data acquisition unit 80 outputs the acquired learning data 64 to the learning unit 84.

[0075] The learning unit 84 includes a loss function calculation unit 85 and a weight update unit 86. The learning unit 84 has a function of generating an image generation model 67 that receives a plurality of tomographic images 101 as input and outputs a pseudo two-dimensional image 103 by performing machine learning on a machine learning model using the learning data 64 as described above.

[0076] As described above, the loss function calculation unit 85 calculates a loss function that represents the degree of difference between the pseudo 2D image 103 output from the image generation model 67 and the normal 2D image 111 of the training data 64.

[0077] As described above, the weight update unit 86 updates the weights of the network in the image generation model 67 in accordance with the loss function calculated by the loss function calculation unit 85.

[0078] The learning unit 84 stores the generated image generation model 67 in the storage unit 62.

[0079] In the learning phase according to this embodiment, an image generation model 67 is learned individually for each of the predetermined categories, thereby learning a plurality of image generation models 67, and the plurality of image generation models 67 are stored in the memory unit 62 as an image generation model group 66.

[0080] Therefore, the learning data 64 according to this embodiment stores a plurality of pairs of tomographic images 101 and normal two-dimensional images 111 for each of the above-mentioned predetermined divisions (hereinafter referred to as "setting divisions"). The learning data acquisition unit 80 according to this embodiment acquires the learning data 64 from the storage unit 62 for each setting division, and the learning unit 84 according to this embodiment uses the acquired learning data 64 to learn an image generation model 67 for each setting division. Through this learning for each setting division, an image generation model 67 for that setting division is generated.

[0081] In this embodiment, subject-related classifications are applied as the set classifications. Note that in this embodiment, the subject's height, weight, BMI, and age are applied as the subject-related classifications, but this is not limited to this. For example, any one of these classifications, or a combination of multiple classifications excluding all of the classifications, may be applied as the set classification, or a subject-related classification other than these classifications that affects the pseudo two-dimensional image may be applied as the set classification.

[0082] Furthermore, when the above-described set imaging is performed, in which tomosynthesis imaging for obtaining multiple tomographic images 101 and capturing of normal two-dimensional images 111 corresponding to the multiple tomographic images 101 are performed under the same compression state by the compression paddle 30, and the above-described set image, which is a combination of the multiple tomographic images 101 and the normal two-dimensional image 111, is obtained, the learning unit 84 according to this embodiment includes the set image in the learning data 64. As a result, when set imaging is performed by the mammography apparatus 10, the set image obtained by the set imaging can be used for training the image generation model 67.

[0083] Here, the learning unit 84 according to this embodiment accumulates a predetermined amount of set images to be used in learning the image generation model 67, and uses the set images to successively update the image generation model 67. This makes it possible to prevent excessive learning, which would otherwise occur if the image generation model 67 were to be learned every time a set image is obtained.

[0084] Furthermore, the learning unit 84 according to this embodiment prohibits the use of images of a combination of a plurality of tomographic images 101 and a normal two-dimensional image 111 for learning when the image quality of at least one image is lower than a predetermined level. In this embodiment, to determine whether the image quality of the image is lower than a predetermined level, both a determination of whether the amount of noise in the image is greater than a predetermined amount (hereinafter referred to as "first determination") and a determination of whether the amount of body movement of the subject in the image is greater than a predetermined amount (hereinafter referred to as "second determination") are applied.

[0085] Note that, as a method for deriving the amount of noise in an image used in the first determination, a conventionally known method described in JP 2016-190012 A, JP 2009-82498 A, etc. can be applied. Furthermore, as a method for deriving the amount of body movement of a subject in an image used in the second determination, a conventionally known method described in JP 2022-125356 A, JP 2020-48991 A, etc. can be applied. Therefore, further explanation of the method for deriving these physical quantities will be omitted.

[0086] Next, the model selection information database 68 according to this embodiment will be described with reference to Fig. 8. Fig. 8 is a schematic diagram showing an example of the configuration of the model selection information database 68 according to this embodiment.

[0087] The model selection information database 68 according to this embodiment is a database in which information for selectively applying the image generation model 67 generated for each of the above-mentioned setting categories is registered. As an example, as shown in Fig. 8, the model selection information database 68 according to this embodiment stores information on subject categories and model IDs (Identifications) in association with each other.

[0088] The setting category is information indicating the setting category, and includes the subject's height, weight, BMI, and age, as described above. The model ID is information previously assigned to each of the image generation models 67 included in the image generation model group 66 in order to individually identify the image generation models 67.

[0089] In this way, in this embodiment, the model ID is used as an identifier for the image generation model 67 included in the image generation model group 66. Therefore, each image generation model 67 included in the image generation model group 66 is stored in association with the corresponding model ID.

[0090] Next, the operation of the image processing device 16 of this embodiment in the learning phase will be described with reference to Fig. 9. The learning process shown in Fig. 9 is performed by the CPU 60A executing the learning program 63A stored in the storage unit 62. To avoid confusion, the case where the number of set images that have not been used in learning for each setting category has been accumulated to the predetermined amount or more at the start of execution of this learning process will be described here.

[0091] 9, the learning data acquisition unit 80 acquires the learning data 64 from the storage unit 62, as described above. At this time, the learning data acquisition unit 80 acquires the learning data 64 by reading it from the storage unit 62, the learning data 64 corresponding to any one of the set classifications (hereinafter referred to as the "application classification").

[0092] In the next step S102, the learning unit 84 performs the first and second judgments as described above to determine whether or not there are any images with image quality lower than a predetermined level among the multiple tomographic images 101 and normal two-dimensional images 111 in the learning data 64 acquired in step S100. If there are any images with image quality lower than the predetermined level, the learning unit 84 deletes the combination of multiple tomographic images 101 and normal two-dimensional images 111 that includes the images, thereby executing a use prohibition process that prohibits the use of the images of the combination in learning.

[0093] In the next step S104, the learning unit 84 inputs the multiple tomographic images 101 contained in the learning data 64 that have undergone the prohibition processing of step S102, as described above, into an image generation model 67 corresponding to the application category (hereinafter referred to as the "application model").

[0094] In the next step S106, the loss function calculation unit 85 acquires the pseudo 2D image 103 output from the applied model, as described above, and calculates a loss function that represents the degree of difference between the pseudo 2D image 103 output from the applied model and the corresponding normal 2D image 111.

[0095] In the next step S108, the learning unit 84 determines whether or not to end learning. In this embodiment, the applied model that has the smallest output of the loss function after learning has been performed a predetermined number of times is ultimately adopted. Therefore, the learning unit 84 determines whether or not learning has been performed a predetermined number of times. Specifically, the learning unit 84 determines whether or not the processes of steps S104 and S106 have been performed a predetermined number of times. If learning has not been performed a predetermined number of times, in other words, if the number of times learning has been performed is less than the predetermined number of times, learning has not yet ended, so the determination in step S108 is negative, and the process proceeds to step S110.

[0096] In step S110, the weight update unit 86 updates the network weights in the applied model in accordance with the loss function calculated in step S106, as described above.

[0097] After the weights of the network of the applied model are updated by the process of step S110, the process returns to step S104, and learning is performed again by repeating the processes of steps S104 to S108. As a result, multiple applied models are obtained according to the number of times learning has been performed.

[0098] On the other hand, if the number of times learning has been performed is equal to or greater than the predetermined number of times, learning is terminated, so the determination in step S108 becomes positive and the process proceeds to step S112.

[0099] In step S112, the learning unit 84 selects, from among a plurality of application models corresponding to the number of times of learning, the one with the smallest loss function calculated in step S106 above, as the application model finally obtained by learning.

[0100] In step S114, the learning unit 84 determines whether the above processing has been completed for all the sections in the set classification, and if the determination is negative, the process returns to step S100, whereas if the determination is positive, the learning processing ends. Note that when repeatedly executing the processing of steps S100 to S114, the learning unit 84 sets any section in the set classification that has not been set as an applicable section as an applicable section.

[0101] Through the above learning process, the image generation model 67 corresponding to all application categories is learned.

[0102] It should be noted that this is not limited to this learning process. For example, a threshold may be set depending on whether the pseudo 2D image 103 output from the image generation model 67 is sufficiently close to the normal 2D image 111, and when the output of the loss function falls below the threshold, the learning may be terminated, and the model at that time may be used as the final application model obtained by learning.

[0103] In recent years, the number of facilities that perform only tomosynthesis imaging or facilities that perform tomosynthesis imaging and mammography imaging separately has been increasing. Furthermore, these facilities often have only imaging devices that perform only tomosynthesis imaging or imaging devices that perform tomosynthesis imaging and mammography imaging separately.

[0104] Although not shown, in the radiographic imaging system 1 according to this embodiment, these imaging devices are connected to a network 17, and the image generation model group 66 obtained by the above learning process is transferred to the imaging devices. Then, in these imaging devices, an image generation model 67 included in the image generation model group 66 is used to generate a pseudo two-dimensional image.

[0105] (Operational phase) Next, the operation phase in which a pseudo two-dimensional image is generated using image generation model 67 of image generation model group 66 trained as described above will be described.

[0106] 10 is a schematic diagram illustrating an outline of the flow of generating a pseudo two-dimensional image 102 using the image generation model 67 in the image processing device 16 of this embodiment. As shown in FIG. 10, the image processing device 16 generates the pseudo two-dimensional image 102 by inputting a plurality of tomographic images 100 into the image generation model 67 and outputting the pseudo two-dimensional image 102. Note that when the image generation model 67 is used, the pseudo two-dimensional image 102 may be generated by inputting a plurality of tomographic images 100 in voxel units, creating patches of the pseudo two-dimensional image 102, and combining the patches.

[0107] 11 shows a functional block diagram of an example of a configuration related to a function of generating a pseudo two-dimensional image 102 in the image processing device 16. As shown in FIG. 11, the image processing device 16 includes a tomographic image generating unit 90, a pseudo two-dimensional image generating unit 92, and a display control unit 94. As an example, in the image processing device 16 of this embodiment, the CPU 60A of the control unit 60 executes an image generating program 63B stored in the storage unit 62, so that the CPU 60A functions as the tomographic image generating unit 90, the pseudo two-dimensional image generating unit 92, and the display control unit 94.

[0108] The tomographic image generating unit 90 has the function of generating multiple tomographic images from a series of multiple projection images. Based on an instruction to generate a pseudo 2D image, the tomographic image generating unit 90 acquires a desired series of multiple projection images from the console 12 of the mammography apparatus 10 or the PACS 14. The tomographic image generating unit 90 then generates multiple tomographic images 100, each at a different height from the imaging plane 24A, from the acquired series of multiple projection images. Note that the method by which the tomographic image generating unit 90 generates the multiple tomographic images 100 is not particularly limited. For example, the tomographic image generating unit 90 can generate the multiple tomographic images 100 by reconstructing the series of multiple projection images using a back projection method such as the FBP (Filtered Back Projection) method or an iterative reconstruction method. The tomographic image generating unit 90 outputs the generated multiple tomographic images 100 to the pseudo 2D image generating unit 92.

[0109] 10, the pseudo 2D image generation unit 92 has a function of generating a pseudo 2D image 102 using the image generation model 67. The pseudo 2D image generation unit 92 inputs a plurality of tomographic images 100 to the image generation model 67. As a result, the pseudo 2D image 102 is output from the image generation model 67, as described above. The pseudo 2D image generation unit 92 acquires the pseudo 2D image 102 output from the image generation model 67, and outputs it to the display control unit 94.

[0110] Here, the pseudo 2D image generator 92 according to this embodiment includes a model selector 92A. The model selector 92A selects an image generation model 67 corresponding to the subject's category in the set category from the image generation model group 66. Then, the pseudo 2D image generator 92 generates a pseudo 2D image 102 using the image generation model 67 selected by the model selector 92A.

[0111] The display control unit 94 has a function of controlling the display of the pseudo two-dimensional image 102 generated by the pseudo two-dimensional image generating unit 92 on the display unit 70 .

[0112] Next, the operation of generating the pseudo two-dimensional image 102 in the image processing device 16 of this embodiment will be described with reference to Fig. 12. The image generation process shown in Fig. 12 is performed by the CPU 60A executing the image generation program 63B stored in the storage unit 62.

[0113] In step S200 of FIG. 12, the tomographic image generating unit 90 acquires a series of multiple projection images from the console 12 of the mammography apparatus 10 or the PACS 14, as described above.

[0114] In the next step S202, the tomographic image generating unit 90 generates a plurality of tomographic images 100 from the series of a plurality of projection images acquired in the above step S200, as described above.

[0115] In the next step S204, the model selection unit 92A identifies the corresponding subject's category. In step S206, the model selection unit 92A selectively acquires a model ID corresponding to the identified category from the model selection information database 68, thereby selecting an image generation model 67 corresponding to the identified category from the image generation model group 66. In the radiation imaging system 1 according to this embodiment, information indicating the subject's category is registered in the header area of ​​the tomographic image 100, and the subject's category is identified by referring to the information. However, this is not a limitation. For example, the subject's category may be identified via a private tag of the image defined by DICOM (Digital Imaging and Communications in Medicine), a common standard for medical images. Alternatively, the subject's category may be identified by acquiring information indicating the subject's category from a patient information management system (not shown) connected to the network 17.

[0116] In the next step S208, the pseudo 2D image generation unit 92 generates the pseudo 2D image 102 using the selected image generation model 67, as described above. Specifically, the pseudo 2D image 102 output from the image generation model 67 is acquired by inputting the multiple tomographic images 100 generated in step S202 to the selected image generation model 67.

[0117] In the next step S210, the display control unit 94 controls the display unit 70 to display the pseudo two-dimensional image 102 obtained in step S208. The display format for displaying the pseudo two-dimensional image 102 on the display unit 70 is not particularly limited. For example, only the pseudo two-dimensional image 102 may be displayed on the display unit 70, or a plurality of tomographic images 100 and the pseudo two-dimensional image 102 may be displayed on the display unit 70. When the processing of step S210 is completed, the image generation processing shown in FIG. 12 is completed.

[0118] As described above, according to this embodiment, an image generation model is applied that has been trained in advance using a plurality of combinations of a plurality of tomographic images and normal two-dimensional images captured by irradiating a breast compressed by a compression member in tomosynthesis imaging to obtain the plurality of tomographic images. Therefore, the accuracy of generating a pseudo two-dimensional image can be improved compared to using an image generation model trained using a combination of a plurality of tomographic images and normal two-dimensional images captured separately from the tomosynthesis imaging to obtain the plurality of tomographic images.

[0119] Furthermore, according to this embodiment, multiple combinations of multiple tomographic images and normal 2D images are acquired for each of the predetermined sections, and an image generation model is individually trained for each of the sections, thereby training multiple image generation models. Therefore, by using the multiple image generation models for each of the sections, it is possible to generate pseudo 2D images with higher accuracy than when the multiple image generation models are not used.

[0120] Furthermore, according to this embodiment, the predetermined divisions are divisions related to the subject, so that pseudo two-dimensional images corresponding to each division of the subject can be generated with high accuracy.

[0121] In particular, according to this embodiment, classifications including at least one of the subject's height, weight, BMI, and age are applied as classifications related to the subject. Since the subject's physique indicated by these classifications affects the breast composition, it is possible to generate a pseudo two-dimensional image corresponding to the classification indicating the subject's physique with high accuracy.

[0122] Furthermore, according to this embodiment, if the image quality of at least one of the combinations of multiple tomographic images and normal 2D images is lower than a predetermined level, the image of that combination is prohibited from being used in learning, thereby improving the accuracy of generating pseudo 2D images compared to when this prohibition is not implemented.

[0123] Furthermore, according to this embodiment, when a set of images is obtained by performing a set of tomosynthesis imaging and normal two-dimensional image imaging under the same pressure from a compression member, the set of images is used as learning data for the image generation model. Therefore, learning data for the image generation model can be obtained every time a set of images is performed.

[0124] In particular, according to this embodiment, a predetermined amount of set images are accumulated and used for training the image generation model, and the image generation model is successively updated using the set images. Therefore, excessive training, which would otherwise be required to train the image generation model every time a set image is obtained, can be prevented.

[0125] Furthermore, according to this embodiment, multiple tomographic images obtained from a series of multiple projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member are acquired as images for generating a pseudo 2D image. Then, according to this embodiment, the acquired images for generation are input into an image generation model to generate a pseudo 2D image corresponding to the images for generation. Therefore, the accuracy of generating a pseudo 2D image can be improved compared to when an image generation model is used to generate a pseudo 2D image, which is trained using a combination of multiple tomographic images and normal 2D images captured separately from the tomosynthesis imaging used to obtain the multiple tomographic images.

[0126] In particular, according to this embodiment, a pseudo 2D image is generated by selectively using an image generation model for a division corresponding to the pseudo 2D image to be generated from among a plurality of image generation models individually trained for each predetermined division. Therefore, a pseudo 2D image can be generated with higher accuracy than when a plurality of image generation models are not selectively used for each of the divisions.

[0127] Furthermore, according to this embodiment, the above image generation model is provided in an imaging device that performs only tomosynthesis imaging and an imaging device that performs tomosynthesis imaging and mammography imaging separately. Therefore, even in these imaging devices, the accuracy of generating pseudo 2D images can be improved compared to when generating images using an image generation model trained using a combination of multiple tomographic images and normal 2D images captured separately from the tomosynthesis imaging that obtains the multiple tomographic images.

[0128] In the present embodiment, at least one of height, weight, BMI, and age, which indicate the physique of the subject, is used as the predetermined category of the present disclosure, but the present disclosure is not limited to this. For example, at least one of mammary gland information, which indicates the type of mammary gland in the subject's breast, and the thickness of the subject's breast may be used as the predetermined category of the present disclosure.

[0129] That is, as disclosed in JP 2019-58606 A, breast gland types are classified into four types: high-density type, fatty type, scattered breast gland type, and heterogeneously high-density type. Therefore, by generating an image generation model for each of these four breast types and applying it to the generation of pseudo-2D images, pseudo-2D images specialized for each breast type can be generated with high accuracy. In this embodiment, the subject's breast type can be identified from a corresponding series of multiple projection images or multiple tomographic images by applying a conventionally known method, such as the method disclosed in JP 2019-58606 A, for example.

[0130] Furthermore, if the thickness of the breast differs, the tube voltage of the tube in radiation source 29 during imaging will differ, and the amount of scattered radiation generated by radiation will also differ. Therefore, by generating an image generation model for each breast thickness and applying it to the generation of a pseudo 2D image, it is possible to generate a pseudo 2D image corresponding to the breast thickness with high accuracy.

[0131] [Second embodiment] In the first embodiment, the predetermined classification of the present disclosure is applied to classifications related to subjects. In contrast, in the present embodiment, the predetermined classification of the present disclosure is applied to classifications related to settings during imaging. In particular, in the present embodiment, the radiation irradiation angle in tomosynthesis imaging and the resolution of the captured image are applied as classifications related to settings during imaging.

[0132] That is, as an example, as disclosed in Japanese Patent Application Laid-Open No. 2014-166357, the radiation irradiation angle and the resolution of the captured image in tomosynthesis imaging may differ for each of multiple predetermined imaging modes.

[0133] For example, in the example described in JP 2014-166357 A, a diagnostic mode and an imaging mode are provided as imaging modes, which can be selected by a user such as a doctor. The diagnostic mode is a mode in which a user roughly images a subject so that the user can perform a medical examination, diagnosis, etc., and, as an example, imaging is performed in 1-degree increments within a range of +10 degrees to -10 degrees. The imaging mode is a mode for performing imaging with higher resolution than imaging in the diagnostic mode, and, as an example, imaging is performed in 1-degree increments within a range of +20 degrees to -20 degrees. Thus, in this example, when imaging to obtain a high-resolution image, imaging is performed at a larger angle to increase the amount of information (amount of image information).

[0134] Therefore, in the image processing device 16 according to this embodiment, the radiation irradiation angle in tomosynthesis imaging and the resolution of the captured image are applied as setting categories that are categories of the image generation model 67. Note that the configuration of the radiographic image capturing system 1 according to this embodiment is substantially the same as that of the first embodiment except for the configuration of the model selection information database 68, so the model selection information database 68 according to this embodiment will be described below with reference to Fig. 13. Fig. 13 is a schematic diagram showing an example of the configuration of the model selection information database 68 according to this embodiment.

[0135] The model selection information database 68 according to this embodiment is a database in which information for selectively applying the image generation model 67 generated for each setting category is registered. As an example, as shown in FIG. 13, each piece of information on the shooting setting category and the model ID is stored in association with each other.

[0136] The imaging setting category is information indicating the type of setting category, and includes information on the resolution of the captured image and the irradiation angle of the radiation, as described above. The model ID is the same information as the model ID in the model selection information database 68 according to the first embodiment. The example shown in Fig. 13 illustrates a case where two levels of resolution, a medium resolution and a high resolution, are applied as the resolution, and angles in 1-degree increments are applied as the irradiation angle, but it goes without saying that the present invention is not limited to this.

[0137] Therefore, in the learning phase of the image processing device 16 according to this embodiment, a plurality of image generation models 67 are generated that have been learned for each combination of resolution and illumination angle indicated in the shooting setting section. Also, in the operation phase of the image processing device 16 according to this embodiment, the pseudo 2D image 102 is generated using the image generation model 67 for each combination of the shooting setting section.

[0138] In the radiographic imaging system 1 according to this embodiment, information indicating the resolution and irradiation angle is registered in the header area of ​​the tomographic image 100, and the setting category of the tomographic image 100 is identified by referring to this information, but this is not limited to this. For example, the setting category may be identified via a private tag of the image defined by DICOM. Alternatively, the setting category may be identified by acquiring information indicating the resolution and irradiation angle from a patient information management system (not shown) connected to the network 17.

[0139] As described above, according to this embodiment, the predetermined classification is based on the settings at the time of shooting, and therefore, a pseudo 2D image corresponding to the classification of the settings at the time of shooting can be generated with high accuracy.

[0140] In particular, according to this embodiment, the radiation irradiation angle and the resolution of the captured image in tomosynthesis imaging are applied as classifications related to the settings during imaging, and therefore, a pseudo two-dimensional image corresponding to the radiation irradiation angle and resolution can be generated with high accuracy.

[0141] In the present embodiment, the case where both the radiation irradiation angle in tomosynthesis imaging and the resolution of the captured image are applied in combination as classifications of the settings during imaging has been described, but this is not limiting. For example, only one of the radiation irradiation angle in tomosynthesis imaging and the resolution of the captured image may be applied as classifications of the settings during imaging.

[0142] [Third embodiment] In the second embodiment, the predetermined classification of the present disclosure applies to classifications related to settings during imaging, and at least one of the radiation irradiation angle and the resolution of the captured image in tomosynthesis imaging is applied as the classification related to settings during imaging. In contrast, in the present embodiment, the radiation dose mode is applied as the classification related to settings during imaging.

[0143] That is, in tomosynthesis imaging and normal two-dimensional imaging, it is often possible to apply a dose mode in which the dose of radiation irradiated during imaging is divided into a plurality of stages.

[0144] Therefore, in the image processing device 16 according to this embodiment, the classification of the dose mode in tomosynthesis imaging and normal two-dimensional image imaging is applied as the setting classification, which is the classification of the image generation model 67. Note that the configuration of the radiographic image capturing system 1 according to this embodiment is substantially the same as that of the first embodiment except for the configuration of the model selection information database 68, and therefore, the model selection information database 68 according to this embodiment will be described below with reference to Fig. 14. Fig. 14 is a schematic diagram showing an example of the configuration of the model selection information database 68 according to this embodiment.

[0145] The model selection information database 68 according to this embodiment is a database in which information for selectively applying the image generation model 67 generated for each setting category is registered. As an example, as shown in FIG. 14, the information on the dose category and the model ID is stored in association with each other.

[0146] The dose category is information indicating the type of setting category, and includes, as described above, information indicating the type of dose mode for tomosynthesis imaging and information indicating the type of dose mode for capturing normal two-dimensional images. Note that the model ID is the same information as the model ID in the model selection information database 68 according to the first embodiment. The example shown in Fig. 14 illustrates a case where three levels of modes, namely, low dose mode, medium dose mode, and high dose mode, are applied as the dose mode, but it goes without saying that the present invention is not limited to this.

[0147] Therefore, in the learning phase of the image processing device 16 according to this embodiment, a plurality of image generation models 67 are generated that are trained for each combination of the dose mode in tomosynthesis imaging indicated in the dose category and the dose mode in imaging of a normal two-dimensional image. Also, in the operation phase of the image processing device 16 according to this embodiment, the image generation model 67 is used for each combination of dose categories to generate a pseudo two-dimensional image 102.

[0148] In the radiographic imaging system 1 according to this embodiment, information indicating the dose mode applied during imaging is registered in the header area of ​​the tomographic image 100 and the normal two-dimensional image 111, and the dose mode of the tomographic image 100 and the normal two-dimensional image 111 is identified by referring to this information, but this is not limited to this. For example, the dose mode may be identified via a private tag of the image defined by DICOM. Alternatively, the dose mode may be identified by acquiring information indicating the dose mode from a patient information management system (not shown) connected to the network 17.

[0149] As described above, according to this embodiment, the radiation dose mode is applied as a classification related to the settings during imaging, and therefore, a pseudo two-dimensional image corresponding to the radiation dose mode during imaging can be generated with high accuracy.

[0150] In this embodiment, the classification of settings during imaging has been described as applying all combinations of dose modes for tomosynthesis imaging and dose modes for imaging of normal 2D images, but this is not limited to this.

[0151] For example, when generating a pseudo two-dimensional image from a series of multiple tomographic images, it may be desirable to ensure that the image quality of the pseudo two-dimensional image is equal to or higher than that of the tomographic images. In this case, as shown in FIG. 15 , for multiple dose modes in tomosynthesis imaging, an image generation model 67 may be prepared that corresponds only to dose modes for capturing normal two-dimensional images that have a dose equal to or higher than the corresponding dose for tomosynthesis imaging. In this embodiment, the dose mode for tomosynthesis imaging shown in FIG. 15 corresponds to the first dose mode of the present disclosure, and the dose mode for capturing normal two-dimensional images shown in FIG. 15 corresponds to the second dose mode with a dose equal to or higher than the first dose mode of the present disclosure. Note that FIG. 15 is a schematic diagram illustrating another example of the configuration of the model selection information database 68 according to this embodiment.

[0152] According to this embodiment, the number of image generation models can be reduced compared to when generating image generation models corresponding to all combinations of dose modes in tomosynthesis imaging and dose modes in normal 2D image imaging.

[0153] [Fourth embodiment] In the above embodiments, the predetermined classification of the present disclosure is based on classifications related to subjects or classifications related to settings during imaging. In contrast, in the present embodiment, the predetermined classification of the present disclosure is based on classifications related to the types of image processing techniques for at least one of a plurality of tomographic images and a normal two-dimensional image.

[0154] That is, the series of projection images described above, the multiple tomographic images obtained from the projection images, and the normal two-dimensional image are generally subjected to image processing unique to each manufacturer of the mammography apparatus 10. Note that the image processing referred to here includes, for example, image processing to reduce noise in the image and image processing to sharpen the image.

[0155] Therefore, in the image processing device 16 according to this embodiment, classifications related to the types of image processing techniques for the tomographic image 100 and the normal two-dimensional image 111 are applied as setting classifications that are classifications of the image generation model 67. Note that the configuration of the radiographic image capturing system 1 according to this embodiment is substantially the same as that of the first embodiment except for the configuration of the model selection information database 68, and therefore, the model selection information database 68 according to this embodiment will be described below with reference to Fig. 16. Fig. 16 is a schematic diagram showing an example of the configuration of the model selection information database 68 according to this embodiment.

[0156] The model selection information database 68 according to this embodiment is a database in which information for selectively applying the image generation model 67 generated for each setting category is registered. As an example, as shown in FIG. 16, each piece of information on the image processing category and the model ID is stored in association with each other.

[0157] The image processing category is information indicating the type of setting category, and as described above, includes information indicating the type of image processing technique to be applied to the tomographic image 100 and information indicating the type of image processing technique to be applied to the normal two-dimensional image 111. The model ID is the same information as the model ID in the model selection information database 68 according to the first embodiment.

[0158] Therefore, in the learning phase of the image processing device 16 according to this embodiment, a plurality of image generation models 67 are generated that have been trained for each combination of the image processing technique for the tomographic image 100 indicated in the image processing section and the image processing technique for the normal two-dimensional image 111. Also, in the operation phase of the image processing device 16 according to this embodiment, the image generation model 67 is used for each combination of the image processing sections to generate the pseudo two-dimensional image 102.

[0159] In the radiographic imaging system 1 according to this embodiment, information indicating the applied image processing technology is registered in the header area of ​​the tomographic image 100 and the normal two-dimensional image 111, and the image processing technology of the tomographic image 100 and the normal two-dimensional image 111 is identified by referring to this information, but this is not limited to this. For example, the image processing technology may be identified using a private tag of the image defined by DICOM. Alternatively, the image processing technology may be identified by acquiring information indicating each image processing technology from a patient information management system (not shown) connected to the network 17.

[0160] As described above, according to this embodiment, the predetermined classification is based on the type of image processing technology used for multiple tomographic images and normal two-dimensional images, and therefore, a pseudo two-dimensional image corresponding to the type of image processing technology used can be generated with high accuracy.

[0161] In this embodiment, the case where the predetermined classification is based on the type of image processing technology for multiple tomographic images and ordinary two-dimensional images has been described, but the present invention is not limited to this. For example, as shown in Fig. 17, the classification based on the type of image processing technology may be based on the manufacturer of the mammography device 10 that performs tomosynthesis imaging. Fig. 17 is a schematic diagram showing another example of the configuration of the model selection information database 68 according to this embodiment.

[0162] According to this aspect, the number of image generation models can be reduced compared to when a classification related to the type of image processing technology is applied to a plurality of tomographic images and a normal two-dimensional image as predetermined classifications.

[0163] [Fifth embodiment] Incidentally, the position of the tube of radiation source 29 (the position on normal line CL in the example shown in FIG. 2, hereinafter referred to as the "reference position"), which serves as the reference during imaging, may differ between tomosynthesis imaging and imaging of a normal two-dimensional image, i.e., normal imaging. In this case, even when a set of images of the breast of the same subject is taken, there will be a discrepancy in the position of the same tissue, such as calcification or mammary gland, between the projection image or tomographic image obtained by tomosynthesis imaging and the normal two-dimensional image.

[0164] Fig. 18A is a diagram showing an example of a radiation irradiation position P1 when performing normal imaging and a normal two-dimensional image 111. Fig. 18B is a diagram showing an example of a radiation irradiation position P2 (i.e., the position of irradiation position 194 in Fig. 2) and a projection image 120 when performing tomosynthesis imaging and the reference position is shifted from that during normal imaging.

[0165] 18B, projection image 120 obtained by tomosynthesis imaging of breast U with the reference position shifted from irradiation position P1 during normal imaging to irradiation position P2 shows a change in the positional relationship of lesion L compared to normal two-dimensional image 111 shown in Fig. 18A. Therefore, when a series of projection images obtained in this state is used to generate tomographic image 101, the positional relationship of lesion L shown in tomographic image 101 will also be changed from that shown in normal two-dimensional image 111.

[0166] On the other hand, the higher the correlation between the multiple tomographic images 101 used to train the image generation model 67 and the corresponding normal two-dimensional image 111, the higher the accuracy of generating the pseudo two-dimensional image.

[0167] Therefore, the learning unit 84 of the image processing device 16 according to this embodiment corrects the multiple tomographic images 101 used in learning the image generation model 67 so that they become tomographic images obtained when a tube is virtually placed at the position of the tube that emitted radiation when capturing the corresponding normal two-dimensional image 111.

[0168] As shown in FIG. 18B as an example, the learning unit 84 according to this embodiment derives a corresponding virtual projection position P3 from an irradiation position P1 during normal imaging in order to align the positional relationship of a lesion L during tomosynthesis imaging with a normal two-dimensional image 111. Specifically, for example, the difference between irradiation position P1 and irradiation position P2 may be calculated, and the irradiation position during tomosynthesis imaging may be corrected according to the calculated difference. The virtual projection position P3 is calculated for each irradiation position during tomosynthesis imaging based on the calculated difference. The virtual projection position P3 is the position onto which the breast U is virtually projected during tomosynthesis imaging.

[0169] The learning unit 84 generates the corrected tomographic image 101 based on the derived virtual projection position P3. At this time, the learning unit 84 generates a plurality of tomographic images 101 with the magnification ratio corrected, centered on the virtual projection position P3.

[0170] In the radiographic imaging system 1 according to this embodiment, information indicating the irradiation position applied during imaging is registered in the header area of ​​the tomographic image 101 and the normal two-dimensional image 111, and the irradiation position is identified by referring to this information, but the present invention is not limited to this. For example, the irradiation position may be identified via a private tag of the image defined by DICOM. Alternatively, the irradiation position may be identified by obtaining the information from a patient information management system (not shown) connected to the network 17.

[0171] The operation of the learning phase and operation phase of the image processing device 16 according to this embodiment is the same as that of the first embodiment except that the above corrections are made to the tomographic images 101 used in learning the image generation model 67, and therefore will not be described here.

[0172] As described above, according to this embodiment, when multiple tomographic images are used for learning, the multiple tomographic images are corrected so that they become tomographic images obtained when a tube is virtually placed at the position of the tube that emitted radiation when capturing the corresponding normal two-dimensional image. Therefore, a pseudo two-dimensional image can be generated with higher accuracy than when this correction is not performed.

[0173] In the above embodiments, the CPU 60A provided in the image processing device 16 is used as the processor of the technology of the present disclosure, but the present disclosure is not limited to this. For example, the CPU of the control unit provided in the mammography device 10 or the CPU of the control unit 40 provided in the console 12 may be used as the processor of the technology of the present disclosure.

[0174] Furthermore, in the above-described embodiments, a case has been described in which a plurality of tomographic images are applied as information to be input to the image generation model 67 in the learning phase and the operation phase, but the present invention is not limited to this. For example, a series of a plurality of projection images obtained by tomosynthesis imaging may be applied as information to be input to the image generation model 67. In this case, it is possible to train the image generation model 67 or generate pseudo two-dimensional images without generating tomographic images.

[0175] In addition, in the above embodiments, a CNN model is used as the image generation model 67, but the present invention is not limited to this. For example, a recurrent neural network (RNN) model or other artificial intelligence (AI) model may be used as the image generation model 67.

[0176] In addition, in the above embodiments, the image generation model 67 is provided in the image processing device 16, but this is not limiting. For example, the image generation model 67 may be provided in the mammography device 10.

[0177] In each of the above embodiments, the following various processors may be used as the hardware structure of processing units that perform various processes, such as the learning data acquisition unit 80, the learning unit 84, the loss function calculation unit 85, the weight update unit 86, the tomographic image generation unit 90, the pseudo two-dimensional image generation unit 92, the model selection unit 92A, and the display control unit 94. As described above, the various processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as dedicated electrical circuits, such as programmable logic devices (PLDs) whose circuit configuration can be changed after manufacture, such as field programmable gate arrays (FPGAs), and application specific integrated circuits (ASICs), which are processors having a circuit configuration designed specifically for performing specific processes.

[0178] A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, a combination of multiple FPGAs, or a combination of a CPU and an FPGA). Also, multiple processing units may be configured with a single processor.

[0179] Examples of configuring multiple processing units with a single processor include, first, a form in which one processor is configured with a combination of one or more CPUs and software, and this processor functions as multiple processing units, as typified by computers such as client and server. Second, a form in which a processor is used to realize the functions of the entire system including multiple processing units with a single IC (Integrated Circuit) chip, as typified by systems on chips (SoCs). In this way, various processing units are configured using one or more of the above-mentioned various processors as a hardware structure.

[0180] Furthermore, more specifically, the hardware structure of these various processors can be an electric circuit that combines circuit elements such as semiconductor elements.

[0181] In the above embodiments, the learning program 63A and the image generation program 63B are pre-stored (installed) in the storage unit 62 of the image processing device 16, but this is not limiting. The programs may be provided in a form recorded on a recording medium such as a CD-ROM (Compact Disc Read Only Memory), a DVD-ROM (Digital Versatile Disc Read Only Memory), or a USB (Universal Serial Bus) memory. The programs may also be downloaded from an external device via a network.

[0182] The present invention is also applicable to programs and program products.

[0183] From the above description, the invention described in the following appendix can be understood.

[0184] [Appendix 1] An image generating device that generates a pseudo two-dimensional image from a series of multiple projection images obtained by tomosynthesis imaging while a breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, at least one image generation model; The image generation model is The system has been trained in advance using a plurality of combinations of the plurality of projection images or the plurality of tomographic images and normal two-dimensional images captured by irradiating radiation onto the breast in a state compressed by the compression member in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images. Image generating device. [Appendix 2] further comprising a processor; The processor: acquiring a plurality of combinations of the plurality of projection images or the plurality of tomographic images and the ordinary two-dimensional image for each of predetermined sections; training the image generation model individually for each of the sections, thereby training a plurality of the image generation models; 2. The image generating device of claim 1. [Appendix 3] The predetermined division is a division related to a subject. 3. The image generating apparatus of claim 2. [Appendix 4] The category regarding the subject includes at least one category of the subject's height, weight, BMI, age, breast information, and breast thickness; 4. The image generation device of claim 3. [Appendix 5] The predetermined classification is a classification related to settings at the time of shooting. 3. The image generating apparatus of claim 2. [Appendix 6] The classification regarding the setting contents during imaging includes at least one classification of an irradiation angle of radiation in the tomosynthesis imaging, a resolution of an imaging image, and a radiation dose mode. 6. The image generating apparatus of claim 5. [Appendix 7] The classification of the radiation dose mode is a classification for each combination of a first dose mode during the tomosynthesis imaging and a second dose mode during the imaging of the normal two-dimensional image, the second dose mode having a dose equal to or greater than the first dose mode. 7. The image generating apparatus of claim 6. [Appendix 8] the predetermined classification is a classification related to the type of image processing technique for at least one of the plurality of projection images or the plurality of tomographic images and the ordinary two-dimensional image; 3. The image generating apparatus of claim 2. [Appendix 9] The classification of the type of image processing technology is a classification by manufacturer of the imaging device that performs the tomosynthesis imaging. 9. The image generation apparatus of claim 8. [Appendix 10] The processor: If the image quality of at least one image among the plurality of combinations of the plurality of projection images or the plurality of tomographic images and the normal two-dimensional image is lower than a predetermined level, the image of the combination is prohibited from being used in the learning. 10. The image generating device according to any one of Supplementary Note 2 to Supplementary Note 9. [Appendix 11] The processor: When the plurality of tomographic images are used for the learning, the plurality of tomographic images are corrected to become tomographic images obtained when a tube is virtually placed at the position of a tube that emitted radiation when capturing the corresponding normal two-dimensional image. 11. The image generating device according to any one of claims 2 to 10. [Appendix 12] The processor: When a set of images is obtained in which the tomosynthesis imaging and the normal two-dimensional image are captured under the same pressure by the compression member, and a set of images is obtained that is a combination of the plurality of projection images or the plurality of tomographic images and the normal two-dimensional image, the set of images is used as learning data for the image generation model. 12. The image generating device according to any one of claims 2 to 11. [Appendix 13] The processor: The set of images is stored and used in training the image generation model. 13. The image generation apparatus of claim 12. [Appendix 14] The processor: Sequentially updating the image generation model using the set of images; 14. The image generating device according to claim 12 or 13. [Appendix 15] The processor: a series of projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member, or a series of tomographic images obtained from the series of projection images, are acquired as images for generating the pseudo two-dimensional image; The acquired image for generation is input into the image generation model to generate the pseudo two-dimensional image corresponding to the image for generation. 15. The image generating device according to any one of claims 2 to 14. [Appendix 16] The processor: When there are a plurality of image generation models trained individually for each predetermined section as the image generation model, the pseudo 2D image is generated by selectively using an image generation model of a section corresponding to the pseudo 2D image to be generated from among the plurality of image generation models. 16. The image generation apparatus of claim 15. [Appendix 17] The image generation model is provided in an imaging device that performs only tomosynthesis imaging or an imaging device that performs tomosynthesis imaging and mammography imaging separately. 17. The image generating device according to any one of claims 1 to 16. [Appendix 18] An image generation method in which a computer executes a process of generating a pseudo two-dimensional image using an image generation model from a series of multiple projection images obtained by tomosynthesis imaging in a state in which a breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, The image generation model is The system has been trained in advance using a plurality of combinations of the plurality of projection images or the plurality of tomographic images and normal two-dimensional images captured by irradiating radiation onto the breast in a state compressed by the compression member in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images. Image generation method. [Appendix 19] A program that causes a computer to execute a process of generating a pseudo two-dimensional image using an image generation model from a series of multiple projection images obtained by tomosynthesis imaging while a breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, The image generation model is The system has been trained in advance using a plurality of combinations of the plurality of projection images or the plurality of tomographic images and normal two-dimensional images captured by irradiating radiation onto the breast in a state compressed by the compression member in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images. program. [Explanation of symbols]

[0185] 1. Radiography system 10 Mammography equipment 12 Console 14 PACS 16 Image processing device 17 Network 191-197, 19 t Irradiation position 20 Radiation detector, 20A detection surface 24 imaging table, 24A imaging surface 26 Foundation 27 Shaft 28 Arm section 29 Radiation source 30 Compression Plate 32 Compression Unit 40, 60 Control unit 42, 50, 62 storage section 44 User I / F section 46, 74 Communication I / F section 52 Radiography images 60A CPU, 60B ROM, 60C RAM 63A Learning Program, 63B Image Generation Program 64 training data 66 Image generation models 67 Image Generation Model 70 Display section 72 Control section 79 Bus 80 Learning data acquisition unit 84 Learning Department 85 Loss function calculation unit 86 Weight update unit 90 Tomographic image generation unit 92 Pseudo 2D image generation unit 92A Model Selection Section 94 Display control unit 100, 1001, 100 k , 101 Tomographic images 102, 103 Pseudo 2D images 111 Normal 2D image 120 Projected Images 200 input layers 202 Middle Class 203d decoder 203e Encoder 204 Output Layer 300, 3021, 302 u , 304 nodes F Filter Ip Pixel of interest Icp pixel value D input data, Dc, DIc output data U Breast R radiation, RC radiation axis conv convolutional layer pool pooling layer upsmp upsampling layer α, β angles P1, P2 irradiation position P3 Virtual projection position CL normal L lesion

Claims

1. An image generating device that generates a pseudo two-dimensional image from a series of multiple projection images obtained by tomosynthesis imaging while a breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, at least one image generation model; The image generation model is The system has been trained in advance using a plurality of combinations of the plurality of projection images or the plurality of tomographic images and normal two-dimensional images captured by irradiating radiation onto the breast in a state compressed by the compression member in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images. Image generating device.

2. further comprising a processor; The processor: acquiring a plurality of combinations of the plurality of projection images or the plurality of tomographic images and the ordinary two-dimensional image for each of predetermined sections; training the image generation model individually for each of the sections, thereby training a plurality of the image generation models; The image generating device of claim 1 .

3. The predetermined division is a division related to a subject. The image generating device of claim 2 .

4. The category regarding the subject includes at least one category of the subject's height, weight, BMI, age, breast information, and breast thickness; The image generating device according to claim 3 .

5. The predetermined classification is a classification related to settings at the time of shooting. The image generating device of claim 2 .

6. the classification relating to the setting contents during imaging includes at least one classification of an irradiation angle of radiation in the tomosynthesis imaging, a resolution of an image to be captured, and a dose mode of radiation; The image generating device according to claim 5 .

7. the classification of the radiation dose mode is a classification for each combination of a first dose mode during the tomosynthesis imaging and a second dose mode during the imaging of the normal two-dimensional image, the second dose mode having a dose equal to or greater than that of the first dose mode; 7. The image generating device of claim 6.

8. the predetermined classification is a classification related to the type of image processing technique for at least one of the plurality of projection images or the plurality of tomographic images and the ordinary two-dimensional image; The image generating device of claim 2 .

9. The classification of the type of image processing technology is a classification by manufacturer of the imaging device that performs the tomosynthesis imaging.

9. The image generating device of claim 8.

10. The processor: If the image quality of at least one image among a plurality of combinations of the plurality of projection images or the plurality of tomographic images and the ordinary two-dimensional image is lower than a predetermined level, the image of the combination is prohibited from being used in the learning.

10. The image generating device according to claim 2.

11. The processor: When the plurality of tomographic images are used for the learning, the plurality of tomographic images are corrected to become tomographic images obtained when a tube is virtually placed at the position of a tube that emitted radiation when capturing the corresponding normal two-dimensional image. The image generating device of claim 2 .

12. The processor: When a set of images is obtained in which the tomosynthesis imaging and the normal two-dimensional image are captured under the same pressure by the compression member, and a set of images is obtained that is a combination of the plurality of projection images or the plurality of tomographic images and the normal two-dimensional image, the set of images is used as learning data for the image generation model. The image generating device of claim 2 .

13. The processor: The set of images is stored and used in training the image generation model.

13. The image generating device of claim 12.

14. The processor: Sequentially updating the image generation model using the set of images; 14. The image generating device according to claim 12 or 13.

15. The processor: a series of projection images obtained by tomosynthesis imaging while the breast is compressed by a compression member, or a series of tomographic images obtained from the series of projection images, are acquired as images for generating the pseudo two-dimensional image; The acquired image for generation is input into the image generation model to generate the pseudo two-dimensional image corresponding to the image for generation. The image generating device of claim 2 .

16. The processor: When there are a plurality of image generation models trained individually for each predetermined section as the image generation model, the pseudo two-dimensional image is generated by selectively using an image generation model of a section corresponding to the pseudo two-dimensional image to be generated from among the plurality of image generation models.

16. The image generating device of claim 15.

17. The image generation model is provided in an imaging device that performs only tomosynthesis imaging or an imaging device that performs tomosynthesis imaging and mammography imaging separately.

3. The image generating device according to claim 1 or 2.

18. An image generation method in which a computer executes a process of generating a pseudo two-dimensional image using an image generation model from a series of multiple projection images obtained by tomosynthesis imaging in a state in which a breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, comprising: The image generation model is The system has been trained in advance using a plurality of combinations of the plurality of projection images or the plurality of tomographic images and normal two-dimensional images captured by irradiating radiation onto the breast in a state compressed by the compression member in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images. Image generation method.

19. A program that causes a computer to execute a process of generating a pseudo two-dimensional image using an image generation model from a series of multiple projection images obtained by tomosynthesis imaging in a state in which a breast is compressed by a compression member, or multiple tomographic images obtained from the series of multiple projection images, The image generation model is The system has been trained in advance using a plurality of combinations of the plurality of projection images or the plurality of tomographic images and normal two-dimensional images captured by irradiating radiation onto the breast in a state compressed by the compression member in tomosynthesis imaging for obtaining the plurality of projection images or the plurality of tomographic images. program.

Images