Learning device, method, and program, and radiation image processing device, method, and program

Abstract

To accurately estimate the composition of a subject in a learning device, a method, and a program, and a radiation image processing device, a method, and a program.SOLUTION: A neural network is mechanically learned using teacher data consisting of learning data and correct answer data, and when at least one object radiation image of a target subject is input, a learned neural network for outputting at least one of a composition and a body thickness of the target subject is constructed. The learning data includes a reference image, energy characteristics of a radiation, and thickness and quality of an object acquired by emitting the radiation to a reference object whose thickness and quality are known in a state that the object is interposed between the reference object and a radiation detector, and an imaging condition at the time when the reference image is acquired. The correct answer data includes the reference image, the energy characteristics of the radiation, the thickness and quality of the object, and at least one of the composition and the body thickness of the subject derived from at least one of the radiation images of the subject using the imaging condition.SELECTED DRAWING: Figure 3

Description

[Technical field]

[0001] The present disclosure relates to a learning device, method, and program, and a radiation image processing device, method, and program. [Background technology]

[0002] Conventionally, energy subtraction processing has been known, which uses two radiation images obtained by irradiating a subject with two types of radiation with different energy distributions, utilizing the fact that the attenuation of transmitted radiation differs depending on the material that constitutes the subject. The energy subtraction processing is a method of matching the pixels of the two radiation images obtained as described above, multiplying the pixels by an appropriate weighting coefficient, and then performing subtraction to obtain an image in which a specific structure contained in the radiation image is extracted. By performing such energy subtraction processing, for example, if a soft tissue image in which soft tissue is extracted from a radiation image obtained by photographing the chest is derived, shadows appearing in the soft tissue can be observed without being obstructed by bones. Conversely, if a bone image in which bone tissue is extracted is derived, shadows appearing in the bone can be observed without being obstructed by soft tissue.

[0003] Also, various methods have been proposed for deriving human body composition such as bone mineral content, fat, and muscle by energy subtraction processing. For example, Patent Document 1 proposes a method for generating soft-tissue images by extracting the soft tissue of a subject from a plurality of radiation images acquired by radiation having different energy distributions transmitted through the subject, estimating the body thickness distribution of the subject based on the imaging conditions when the soft-tissue images and the radiation images were acquired, calculating an approximate body thickness distribution by approximating the estimated body thickness distribution with a model corresponding to the human body, and calculating the distribution of body fat percentage within the subject based on the approximate body thickness distribution.

[0004] On the other hand, when capturing a radiographic image of a subject, particularly if the subject is thick, radiation is scattered within the subject, generating scattered rays, which can cause a problem of reducing the contrast of the acquired radiographic image. For this reason, a scattered radiation removal process is performed to remove scattered radiation components contained in the radiographic image (see, for example, Patent Document 2). According to the method described in Patent Document 2, the scattered radiation components of the radiographic image are derived based on the radiation attenuation coefficient of the subject, and the derived scattered radiation components are subtracted from the radiographic image, thereby performing the scattered radiation removal process. By applying such a scattered radiation removal process when performing the energy subtraction process, it is possible to obtain bone images and soft tissue images with more accurately separated compositions. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2018-153605 A [Patent Document 2] JP 2015-043959 A Summary of the Invention [Problem to be solved by the invention]

[0006] Incidentally, when actually performing radiography of a subject, a grid may be used between the subject and the radiation detector, or an object such as an imaging table or a top plate of the imaging table may be interposed. Since such objects have inherent radiological characteristics, the radiation attenuation coefficient changes due to beam hardening, which is a hardening of the radiation quality by passing through the object. Specifically, the radiation attenuation coefficient decreases due to beam hardening. Furthermore, the radiation attenuation coefficient decreases as the energy of the radiation irradiated to the subject increases. For this reason, the composition and even the body thickness of the subject cannot be accurately estimated unless the energy characteristics of the radiation and the radiological characteristics of the object interposed between the subject and the radiation detector are taken into consideration.

[0007] The present invention has been made in consideration of the above circumstances, and an object of the present invention is to accurately estimate at least one of the body thickness and composition of a subject. [Means for solving the problem]

[0008] A learning device according to the present disclosure includes at least one processor, the processor performs machine learning on the neural network using teacher data consisting of the learning data and the correct answer data, thereby constructing a trained neural network that outputs at least one of the composition of the target subject and the body thickness of the target subject when at least one target radiation image of the target subject is input; The learning data is A reference image obtained by irradiating a reference object of known thickness and material with radiation while an object is interposed between the reference object and the radiation detector; Energy characteristics of radiation, The thickness and material of the object, and Including the shooting conditions when the reference image was obtained, The correct data is The information includes at least one of a composition and a thickness of the subject derived from at least one radiographic image of the subject using the reference image, the energy characteristics of the radiation, the thickness and material of the object, and the imaging conditions.

[0009] In addition, in the learning device according to the present disclosure, the object may be at least one of the top plate of an imaging table on which a subject is placed in an imaging device, and an anti-scatter grid for removing scattered radiation components from radiation that has passed through the subject.

[0010] Furthermore, in the learning device according to the present disclosure, the target radiographic image may be a simple radiographic image based on radiation that has passed through a target subject including bones and soft tissue.

[0011] In addition, in the learning device according to the present disclosure, the target radiographic images may be a first radiographic image and a second radiographic image based on radiation having different energy distributions that have passed through a target subject including bones and soft tissue.

[0012] In addition, in the learning device according to the present disclosure, the composition of the subject and the composition of the target subject may be at least one of bone mineral mass and muscle mass.

[0013] A radiation image processing device according to the present disclosure includes at least one processor, The processor acquires at least one target radiographic image of the target object; Using a trained neural network constructed by the learning device according to the present disclosure and a target radiographic image, an estimation result of at least one of the composition of the target subject and the body thickness of the target subject is derived.

[0014] In the radiation image processing device according to the present disclosure, the target radiation image may be a simple radiation image based on radiation that has passed through a target subject including bones and soft tissue.

[0015] In the radiological image processing device according to the present disclosure, the target radiological images may be a first radiological image and a second radiological image based on radiation having different energy distributions transmitted through a target subject including bones and soft tissue.

[0016] In the radiation image processing device according to the present disclosure, the composition of the target subject may be at least one of bone salt mass and muscle mass.

[0017] A learning method according to the present disclosure is a learning method for constructing a trained neural network that outputs at least one of a composition of a target subject and a body thickness of a target subject when at least one target radiation image of a target subject is input, by machine learning a neural network using teacher data consisting of learning data and correct answer data, the method comprising: The learning data is A reference image obtained by irradiating a reference object of known thickness and material with radiation while an object is interposed between the reference object and the radiation detector; Energy characteristics of radiation, The thickness and material of the object, and Including the shooting conditions when the reference image was obtained, The correct data is The information includes at least one of a composition and a thickness of the subject derived from at least one radiographic image of the subject using the reference image, the energy characteristics of the radiation, the thickness and material of the object, and the imaging conditions.

[0018] The radiation image processing method according to the present disclosure includes acquiring at least one target radiation image of a target subject; Using a trained neural network constructed by the learning device according to the present disclosure and a target radiographic image, an estimation result of at least one of the composition of the target subject and the body thickness of the target subject is derived.

[0019] The learning method and radiation image processing method according to the present disclosure may be provided as a program executed by a computer. Effect of the Invention

[0020] According to the present disclosure, at least one of the body thickness and composition of a subject can be estimated with high accuracy. [Brief description of the drawings]

[0021] [Figure 1] FIG. 1 is a schematic block diagram showing a configuration of a radiation image capturing system to which a learning device and a radiation image processing device according to an embodiment of the present disclosure are applied; [Diagram 2] FIG. 1 is a diagram showing a schematic configuration of a learning device and a radiation image processing device according to an embodiment of the present invention; [Diagram 3] FIG. 1 is a diagram showing the functional configuration of a learning device and a radiation image processing device according to an embodiment of the present invention. [Figure 4] FIG. 1 is a diagram showing a schematic configuration of a neural network used in the present embodiment; [Diagram 5] Diagram showing training data [Figure 6] Diagram to explain neural network learning [Figure 7] Conceptual diagram of the processing performed by a trained neural network [Figure 8] Diagram showing the display screen [Figure 9] FIG. 2 is a diagram showing a functional configuration of a device for deriving correct answer data; [Figure 10] FIG. 1 is a diagram for explaining photographing of a reference object; [Figure 11] Diagram showing the energy spectrum of radiation [Figure 12] A diagram showing the radiation attenuation coefficients of soft tissue, bone tissue, and aluminum in the human body for radiation energy. [Figure 13] A diagram showing the relationship between the thickness of a reference object and the radiation attenuation coefficient. [Figure 14] Diagram showing the scattered radiation model [Figure 15] Figure showing bone images [Figure 16] Soft tissue images [Figure 17] A diagram showing the relationship between the contrast of bones and soft tissues and the body thickness of the subject [Figure 18] FIG. 1 is a diagram showing an example of a lookup table. [Figure 19] FIG. 1 shows an example of the energy spectrum of radiation after passing through muscle tissue and radiation after passing through fat tissue. [Figure 20] 1 is a flowchart of a learning process performed in this embodiment. [Figure 21] 1 is a flowchart of radiation image processing performed in this embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0022] Hereinafter, an embodiment of the present disclosure will be described with reference to the drawings. Fig. 1 is a schematic block diagram showing the configuration of a radiographic image capturing system to which a learning device and a radiographic image processing device according to an embodiment of the present disclosure are applied. As shown in Fig. 1, the radiographic image capturing system according to this embodiment includes a capturing device 1, and a learning device and a radiographic image processing device (hereinafter, sometimes referred to as the radiographic image processing device) 10 according to this embodiment.

[0023] The imaging device 1 is capable of performing energy subtraction by a so-called one-shot method in which radiation such as X-rays emitted from a radiation source 2 and transmitted through a subject H lying on a supine position on an imaging table 3 is irradiated to a first radiation detector 5 and a second radiation detector 6 with different energies. During imaging, as shown in FIG. 1, a scatter reduction grid (hereinafter simply referred to as a grid) 4, a first radiation detector 5, a radiation energy conversion filter 7 made of a copper plate or the like, and a second radiation detector 6 are arranged in this order from the side closest to the radiation source 2, and the radiation source 2 is driven. The first and second radiation detectors 5, 6 and the radiation energy conversion filter 7 are in close contact with each other. The grid 4, the first radiation detector 5, the radiation energy conversion filter 7, and the second radiation detector 6 are removably attached to the bottom of a top plate 3A of the imaging table 3 by attachment parts 3B.

[0024] As a result, the first radiation detector 5 acquires a first radiation image G1 of the subject H using low-energy radiation including so-called soft rays. Meanwhile, the second radiation detector 6 acquires a second radiation image G2 of the subject H using high-energy radiation from which the soft rays have been removed. Therefore, the first and second radiation images G1, G2 are radiation images with different energy distributions. The first and second radiation images G1, G2 are input to the radiation image processing device 10.

[0025] The first and second radiation detectors 5, 6 are capable of repeatedly recording and reading out radiation images, and may be so-called direct type radiation detectors that generate electric charges by directly receiving radiation, or so-called indirect type radiation detectors that convert radiation into visible light and then convert the visible light into an electric charge signal. As a method for reading out the radiation image signal, it is preferable to use a so-called TFT readout method in which the radiation image signal is read out by turning on and off a TFT (thin film transistor) switch, or a so-called optical readout method in which the radiation image signal is read out by irradiating the detector with readout light, but this is not limiting and other methods may also be used.

[0026] In the present embodiment, in the imaging device 1, there are cases where only one radiation detector is attached to the attachment portion 3B to image the subject H. The imaging in this case is referred to as simple imaging, and the radiographic image acquired by the simple imaging is referred to as simple radiographic image G0.

[0027] The grid 4 is made up of alternately arranged radiation-opaque materials such as lead and interspace materials such as aluminum and fiber that are easily transparent to radiation, at a fine grid density of, for example, about 4.0 lines / mm. By using the grid 4, it is possible to remove the scattered ray components of the radiation that has passed through the subject H, but it is not possible to remove them completely. For this reason, the first and second radiation images G1, G2 or the simple radiation image G0 contain not only the primary ray components of the radiation that has passed through the subject H, but also the scattered ray components.

[0028] The primary ray component is a signal component of a pixel value represented by radiation that has passed through the subject H and reached the radiation detector without being scattered by the subject H. On the other hand, the scattered ray component is a signal component of a pixel value represented by radiation that has passed through the subject H and reached the radiation detector after being scattered by the subject H.

[0029] The radiation image processing device 10 is connected to an image storage system 9 via a network (not shown). The image storage system 9 is a system that stores image data of radiation images captured by the imaging device 1. The image storage system 9 extracts an image from the stored radiation images in response to a request from the radiation image processing device 10, and transmits the image to the device that issued the request. A specific example of the image storage system 9 is a PACS (Picture Archiving and Communication System).

[0030] Next, the learning device and the radiation image processing device according to the present embodiment will be described. First, referring to FIG. 2, the hardware configuration of the learning device and the radiation image processing device according to the present embodiment will be described. As shown in FIG. 2, the radiation image processing device 10 is a computer such as a workstation, a server computer, or a personal computer, and includes a CPU (Central Processing Unit) 11, a non-volatile storage 13, and a memory 16 as a temporary storage area. The learning device and the radiation image processing device 10 also include a display 14 such as a liquid crystal display, an input device 15 such as a keyboard and a mouse, and a network I / F (InterFace) 17 connected to a network (not shown). The CPU 11, the storage 13, the display 14, the input device 15, the memory 16, and the network I / F 17 are connected to a bus 18. The CPU 11 is an example of a processor in the present disclosure.

[0031] The storage 13 is realized by a hard disk drive (HDD), a solid state drive (SSD), a flash memory, etc. The storage 13 as a storage medium stores a learning program 12A and a radiation image processing program 12B installed in the learning device and the radiation image processing device 10. The CPU 11 reads out the learning program 12A and the radiation image processing program 12B from the storage 13, expands them in the memory 16, and executes the expanded learning program 12A and the radiation image processing program 12B.

[0032] The learning program 12A and the radiation image processing program 12B are stored in a storage device of a server computer connected to a network or in a network storage in an externally accessible state, and are downloaded and installed in the computer constituting the learning device and the radiation image processing device 10 upon request. Alternatively, they are recorded on a recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) and distributed, and are installed from the recording medium in the computer constituting the learning device and the radiation image processing device 10.

[0033] Next, the functional configuration of the learning device and the radiation image processing device according to this embodiment will be described. Fig. 3 is a diagram showing the functional configuration of the learning device and the radiation image processing device according to this embodiment. As shown in Fig. 3, the learning device and the radiation image processing device 10 includes an image acquisition section 21, an information acquisition section 22, a learning section 23, an estimation section 24, and a display control section 25. The CPU 11 functions as the information acquisition section 22 and the learning section 23 by executing the learning program 12A. The CPU 11 also functions as the image acquisition section 21, the estimation section 24, and the display control section 25 by executing the radiation image processing program 12B.

[0034] When deriving the estimated results of the body thickness, bone mineral mass, and muscle mass of the subject H, which is the target of the radiation image processing device 10, the image acquisition unit 21 causes the imaging device 1 to perform simple imaging of the subject H, thereby acquiring a simple radiation image G0 of the subject H from the first radiation detector 5. Note that by causing the imaging device 1 to perform imaging of the subject H, a first radiation image G1 and a second radiation image G2, which have different energy distributions for the subject H, may be acquired from the first and second radiation detectors 5, 6. When acquiring the simple radiation image G0, the first radiation image G1, and the second radiation image G2, imaging conditions such as the imaging dose, tube voltage, SID (Source Image Receptor Distance) which is the distance between the radiation source 2 and the surfaces of the first and second radiation detectors 5, 6, SOD (Source Object Distance) which is the distance between the radiation source 2 and the surface of the subject H, and the presence or absence of an anti-scatter grid are set.

[0035] The SOD and SID are used to calculate the body thickness distribution as described below. The SOD is preferably obtained by, for example, a TOF (Time Of Flight) camera. The SID is preferably obtained by, for example, a potentiometer, an ultrasonic distance meter, a laser distance meter, or the like.

[0036] The imaging conditions may be set by the operator through input device 15. The set imaging conditions are stored in storage 13. In this embodiment, the simple radiographic image G0 or the first and second radiographic images G1, G2 may be acquired and stored in storage 13 by a program separate from the radiographic image processing program 12B. In this case, the image acquisition unit 21 acquires the simple radiographic image G0 or the first and second radiographic images G1, G2 stored in storage 13 by reading them out from the storage 13 for processing.

[0037] The information acquiring unit 22 acquires training data for training a neural network, which will be described later, from the image storage system 9 via the network I / F 17. The training data will be described later.

[0038] The learning unit 23 performs machine learning on the neural network using the training data acquired by the information acquisition unit 22, thereby constructing a trained neural network 24A that outputs the body thickness, bone mineral content, and muscle mass of the subject H when a simple radiographic image G0 of the target subject H is input.

[0039] Examples of neural networks include a simple perceptron, a multi-layer perceptron, a deep neural network, a convolutional neural network, a deep belief network, a recurrent neural network, and a probabilistic neural network, etc. In this embodiment, a convolutional neural network is used as the neural network.

[0040] Fig. 4 is a diagram showing a neural network used in this embodiment. As shown in Fig. 4, the neural network 30 includes an input layer 31, an intermediate layer 32, and an output layer 33. The intermediate layer 32 includes, for example, a plurality of convolution layers 35, a plurality of pooling layers 36, and a fully connected layer 37. In the neural network 30, the fully connected layer 37 exists in front of the output layer 33. In the neural network 30, the convolution layers 35 and the pooling layers 36 are alternately arranged between the input layer 31 and the fully connected layer 37.

[0041] The configuration of the neural network 30 is not limited to the example shown in Fig. 4. For example, the neural network 30 may include one convolution layer 35 and one pooling layer 36 between the input layer 31 and the fully connected layer 37.

[0042] Fig. 5 is a diagram showing an example of training data used for training a neural network. As shown in Fig. 5, training data 40 consists of training data 41 and correct answer data 42. Training data 41 includes a reference image 41A, radiation energy characteristics 41B, object characteristics 41C, and imaging conditions 41D. Correct answer data 42 includes a subject's body thickness 42A, bone mineral content 42B, and muscle mass 42C derived based on a radiation image of the subject. Note that body thickness 42A, bone mineral content 42B, and muscle mass 42C are values ​​obtained for each pixel of the radiation image.

[0043] The radiation energy characteristics 41B include at least one of the radiation energy spectrum, tube voltage, aluminum half-value layer, and total filtration amount. The radiation energy characteristics 41B will be described later. The object characteristics 41C include the material and thickness of each of the top plate 3A and grid 4, which are objects interposed between the subject H (or a reference object described later) and the radiation detector 5. An example of the material of the top plate 3A is acrylic. The material of the grid 4 is the material of the interspace material that constitutes the grid 4, and an example of this is aluminum. The shooting conditions 41D include at least one of the dose, radiation quality, SID, and SOD during shooting.

[0044] The teacher data 40 includes a plurality of learning data 41, which are obtained by varying combinations of the thickness of a reference object, radiation energy characteristics, object characteristics, and imaging conditions, and a plurality of correct answer data 42, which are obtained by varying combinations of body thickness, bone mineral content, and muscle mass. The derivation of the correct answer data 42 included in the teacher data 40 will be described later.

[0045] The learning unit 23 learns the neural network using a large amount of teacher data 40. Fig. 6 is a diagram for explaining the learning of the neural network 30. When learning the neural network 30, the learning unit 23 inputs, to the input layer 31 of the neural network 30, a reference image 41A, radiation energy characteristics 41B, object characteristics 41C, and imaging conditions 41D, which are learning data 41.

[0046] For the reference image 41A, the pixel values ​​of each pixel of the reference image 41A are input. When the radiation energy characteristics 41B are an energy spectrum, the relationship itself of the relative number of radiation photons to the radiation energy (i.e., the value of the relative number of radiation photons to the radiation energy) is input. When the radiation energy characteristics 41B are an aluminum half-value layer, a combination of the tube voltage and the aluminum half-value layer is input. When the radiation energy characteristics 41B are a total filtration amount, a combination of the tube voltage and the total filtration amount is input. For the object characteristics 41C, the thickness and material of the object, i.e., the top plate 3A and the grid 4, are input. For the imaging conditions 41D, the dose, radiation quality, SID, and SOD values ​​at the time of imaging are input.

[0047] Then, the learning unit 23 causes the output layer 33 of the neural network 30 to output the body thickness, bone salt mass, and muscle mass as output data 47. The learning unit 23 then derives the difference between the output data 47 and the correct answer data 42 as a loss L0. Note that the loss L0 is the sum of the differences between the body thickness, bone salt mass, and muscle mass included in the output data 47 and the body thickness 42A, bone salt mass 42B, and muscle mass 42C included in the correct answer data 42.

[0048] The learning unit 23 learns the neural network 30 based on the loss L0. Specifically, the learning unit 23 adjusts the kernel coefficients in the convolution layer 35, the connection weights between the layers, the connection weights in the fully connected layer 37, and the like (hereinafter referred to as parameters 48) so as to reduce the loss L0. As a method for adjusting the parameters 48, for example, the backpropagation method can be used. The learning unit 23 repeats the adjustment of the parameters 48 until a termination condition is satisfied. The termination condition may be that the loss L0 is equal to or less than a predetermined threshold value, or that learning is performed a predetermined number of times.

[0049] As a result, when a simple radiographic image G0 of a target subject H is input, the parameters 48 are adjusted so as to output the body thickness, bone mineral content, and muscle content of the subject H at each pixel of the input simple radiographic image G0, and the trained neural network 24A is constructed. The trained neural network 24A thus constructed is stored in the storage 13.

[0050] Fig. 7 is a conceptual diagram of the process performed by the trained neural network 24A. As shown in Fig. 7, when a simple radiographic image G0 of a patient is input to the trained neural network 24A constructed as described above, the trained neural network 24A outputs the body thickness, bone mineral content, and muscle mass (i.e., body thickness distribution, bone mineral content distribution, and muscle mass distribution) for each pixel of the input simple radiographic image G0.

[0051] The estimation unit 24 derives estimation results of the body thickness, bone salt mass, and muscle mass of the subject H from a simple radiographic image G0 acquired by photographing the target subject H. In this embodiment, the estimation unit 24 inputs the simple radiographic image G0 to the trained neural network 24A and causes the trained neural network 24A to output the body thickness, bone salt mass, and muscle mass, thereby deriving estimation results of the body thickness, bone salt mass, and muscle mass of the subject H.

[0052] The display control unit 25 displays the estimation results of the body thickness, bone salt mass, and muscle mass estimated by the estimation unit 24 on the display 14. FIG. 8 is a diagram showing a display screen of the estimation results. As shown in FIG. 8, a simple radiological image G0 of the subject H, a body thickness image Gt which is an estimation result of the body thickness, a bone salt image Ge which is an estimation result of the bone salt mass, and a muscle image Gm which is an estimation result of the muscle mass are displayed on the display screen 50. A pattern corresponding to the body thickness is added to the body thickness image Gt. A pattern corresponding to the bone salt mass distribution is added to the bone salt image Ge. A pattern corresponding to the muscle mass distribution is added to the muscle image Gm.

[0053] In Fig. 8, for the sake of simplicity, a pattern representing the bone salt amount is applied only to the femur in the bone salt image Ge. A reference 51 indicating the magnitude of the bone salt amount for the applied pattern is displayed below the bone salt image Ge. The operator can easily recognize the bone salt amount by reading the bone salt image Ge while referring to the reference 51. Instead of a pattern, different colors according to the bone salt amount may be applied to the bone salt image Ge.

[0054] For the sake of explanation, the outline of the bone region of the muscle image Gm is shown by a broken line in the muscle image Gm. For the sake of simplicity, a pattern representing muscle mass is added only to the vicinity of the femur in FIG. 8. A reference 52 indicating the magnitude of muscle mass for the added pattern is displayed below the muscle image Gm. The operator can easily recognize the muscle mass by reading the muscle image Gm while referring to the reference 52. Instead of a pattern, different colors according to muscle mass may be added to the muscle image Gm.

[0055] Hereinafter, the derivation of the body thickness, bone mineral mass, and muscle mass contained in the correct answer data 42 of the training data 40 will be described.

[0056] Fig. 9 is a diagram showing the functional configuration of a derivation device that derives body thickness, bone salt mass, and muscle mass contained in the supervised data 42. The derivation device 60 is a computer such as a workstation, a server computer, or a personal computer, and has a hardware configuration similar to that of the learning device and radiation image processing device according to this embodiment shown in Fig. 2. In the derivation device 60, the CPU executes a derivation program to derive a scattered radiation model from a reference image, and the body thickness, bone salt mass, and muscle mass are derived using the scattered radiation model.

[0057] 9, the derivation device 60 includes an image acquisition unit 61, an information acquisition unit 62, an attenuation coefficient derivation unit 63, a primary ray component derivation unit 64, a scattered ray component derivation unit 65, a model derivation unit 66, a scattered ray removal unit 67, a subtraction unit 68, and a composition derivation unit 69. When the CPU of the derivation device 60 executes a derivation program, the CPU of the derivation device 60 functions as the image acquisition unit 61, the information acquisition unit 62, the attenuation coefficient derivation unit 63, the primary ray component derivation unit 64, the scattered ray component derivation unit 65, the model derivation unit 66, the scattered ray removal unit 67, the subtraction unit 68, and the composition derivation unit 69.

[0058] When deriving a scattered radiation model, the image acquisition unit 61 acquires a reference image K0 by having the imaging device 1 capture an image of a reference object simulating a human body. At this time, only one radiation detector may be used. Note that, if the reference image K0 is stored in the image storage system 9, the image acquisition unit 61 acquires the reference image K0 from the image storage system 9.

[0059] FIG. 10 is a diagram for explaining the photographing of a reference object. The photographing of the reference object is performed by the photographing device 1 shown in FIG. 1. As shown in FIG. 10, the reference object 70 has a stepwise structure with different thicknesses, such as 5 cm, 10 cm, and 20 cm, and is made of a material having a radiation transmittance similar to that of the soft tissues (fat and muscle) of the human body. Therefore, the reference object 70 simulates the radiation characteristics of the human body. Here, the soft tissue is made of a mixture of muscle and fat at a certain ratio. The mixture ratio of muscle and fat differs depending on gender, body type, etc., but can be defined by the average body fat percentage (65%). Therefore, a material such as acrylic, which corresponds to a composition where 0.75% of muscle and 0.65% of fat are mixed, is used as the reference object.

[0060] 10, to acquire the reference image K0, a reference object 70 is placed on the top plate 3A of the imaging table 3, and the radiation source 2 is driven to irradiate a radiation detector (here, the first radiation detector 5) with radiation that has passed through the reference object 70 via the grid 4, thereby causing the image acquisition section 61 to acquire the reference image K0. The pixel value of each pixel in the reference image K0 includes a primary ray component based on the radiation that has traveled straight through the reference object 70 and a scattered ray component based on the radiation scattered by the reference object 70.

[0061] It should be noted that the reference object 70 is not limited to one object having different thicknesses as shown in Fig. 10. A plurality of reference objects each having a different thickness may be used. In this case, the reference image K0 may be obtained by photographing the plurality of reference objects at once, or the reference images corresponding to the respective reference objects may be obtained by photographing the plurality of reference objects separately.

[0062] On the other hand, when deriving the composition of the subject H (i.e., bone mineral content and muscle mass), the image acquisition unit 61 causes the imaging device 1 to capture an image of the subject H, thereby acquiring a first radiation image G1 and a second radiation image G2 of the subject H from the first and second radiation detectors 5, 6. Note that if the first radiation image G1 and the second radiation image G2 are stored in the image storage system 9, the image acquisition unit 61 acquires the first radiation image G1 and the second radiation image G2 from the image storage system 9.

[0063] When acquiring the reference image K0, the first radiographic image G1, and the second radiographic image G2, imaging conditions such as the radiation dose, radiation quality, SID and SOD, and the presence or absence of a grid 4 during imaging are set.

[0064] The information acquiring unit 62 acquires the imaging conditions set at the time of imaging. The information acquiring unit 62 also acquires the radiation energy characteristics at the time of imaging the reference object 70. The radiation energy characteristics may be acquired from the imaging device 1, or may be stored in the image storage system 9 and acquired from the image storage system 9. Note that, although the nominal value of the imaging device 1 may be used for the energy characteristics, since there are individual differences in characteristics for each device, it is preferable to measure them in advance using a semiconductor dosimeter.

[0065] Here, the energy characteristics are defined by any one of (i) the energy spectrum of the radiation emitted from the radiation source 2, (ii) the tube voltage [kV] and the total filtration amount [mmAl equivalent], and (iii) the tube voltage [kV] and the aluminum half-value layer [mmAl]. The radiation energy spectrum is a plot of the relationship between the radiation energy [keV] and the relative number of radiation photons. The tube voltage means the maximum value of the generated radiation energy distribution. The total filtration amount is the filtration amount of each component of the imaging device 1, such as the radiation generator and collimator in the radiation source 2, converted into the aluminum thickness. The larger the total filtration amount, the greater the effect of beam hardening in the imaging device 1, and the more high-energy components there are in the wavelength distribution of the radiation. The half-value layer is defined by the thickness of aluminum required to attenuate the dose by half for the generated radiation energy distribution. The thicker the aluminum in the half-value layer, the more high-energy components there are in the wavelength distribution of the radiation.

[0066] Figure 11 is a diagram showing the energy spectrum of radiation. In Figure 11, the energy spectrum corresponds to a tube voltage of 90 kV and a total filtration amount of 2.5 mmAl. The total filtration amount of 2.5 mmAl corresponds to a half-value layer of 2.96 mmAl.

[0067] The attenuation coefficient derivation unit 63 uses the energy characteristics of the radiation to derive the relationship between the thickness of the reference object 70 and the radiation attenuation coefficient of the reference object 70, reflecting the effects of beam hardening of objects present between the reference object 70 and the radiation detector 5.

[0068] The attenuation coefficient derivation unit 63 first derives the energy spectrum of the radiation from the energy characteristics of the radiation acquired by the information acquisition unit 62, using the well-known Tucker approximation formula or the like. Note that if the energy characteristics acquired by the information acquisition unit 62 are the energy spectrum of the radiation, the acquired energy spectrum may be used as is.

[0069] Furthermore, the attenuation coefficient deriving unit 63 derives a radiation attenuation coefficient that depends on the thickness of the reference object 70 by simulating a radiation spectrum using the radiation attenuation characteristics of the soft tissue of the human body.

[0070] Here, when the energy spectrum of the radiation emitted from the radiation source 2 is Sin(E) and the thickness of the reference object 70 is t, the radiation dose Xbody(t) after passing through the reference object 70 can be calculated by the following formula (1) using the radiation attenuation characteristic μSoft(E) of the soft tissue of the human body. The radiation attenuation coefficients of the soft tissue, bone tissue, and aluminum of the human body with respect to the radiation energy are known, for example, as shown in FIG. 12. Aluminum is the interspace material of the grid 4. Here, FIG. 12 also shows the radiation attenuation coefficient of acrylic (Polymethyl methacrylate, PMMA), which is the material of the reference object 70. As shown in FIG. 12, the radiation attenuation coefficient of acrylic is almost the same as the radiation attenuation coefficient of the soft tissue of the human body.

number

[0071] On the other hand, as shown in Fig. 1, when a reference object 70 is imaged in the imaging device 1, a top plate 3A and a grid 4 are present between the reference object 70 and the radiation detectors 5 and 6. The material of the top plate 3A is acrylic, and the interspace material of the grid 4 is aluminum. If the radiation attenuation coefficient of acrylic is μPMMA(E), the thickness of the top plate 3A (i.e., the thickness of the acrylic) is tPMMA, the radiation attenuation characteristic of aluminum is μAl(E), and the thickness of the grid 4 (i.e., the aluminum) is tAl, the amount of X-rays Xout(t) after passing through the top plate 3A and the grid 4 is expressed by the following formula (2).

number

[0072] In addition, when the material of the tabletop 3A and the interspace material of the grid 4 are unknown, the X-ray dose Xout(t) after passing through the tabletop 3A and the grid 4 cannot be derived from the above formula (2). In this case, the energy characteristics (kV, TF0) of the radiation emitted from the radiation source 2 and the energy characteristics (kV, TF1) of the radiation after passing through the tabletop 3A and the grid 4 are measured using a dosimeter, and the X-ray dose Xout(t) after passing through the tabletop 3A and the grid 4 can be derived from the following formula (2-1) using the energy characteristics (kV, TF0) and the energy characteristics (kV, TF1). In addition, the energy characteristics in formula (2-1) represent the total filtration amount (Total Filtering) [mmAl equivalent] of the radiation emitted by a certain tube voltage [kV].

number

[0073] The radiation attenuation coefficient of the reference object 70 in the imaging system including the top plate 3A and the grid 4 is expressed as an exponential decay of the rate attenuation of radiation after passing through the reference object 70, using the radiation dose when the reference object 70 is not present (i.e., when the thickness of the reference object 70 is 0) as a reference, as shown in the following equation (3).

number

[0074] By solving equation (3) for the radiation attenuation coefficient μSoft(t) of soft tissue as shown in the following equation (4), the relationship between the thickness t of the reference object 70 and the radiation attenuation coefficient can be derived.

number

[0075] The reference object 70 has a plurality of different thicknesses in stages. Therefore, the attenuation coefficient deriving unit 63 derives the radiation attenuation coefficient for each of the plurality of thicknesses of the reference object 70 by the formula (4). Then, the attenuation coefficient deriving unit 63 derives the relationship between the thickness t of the reference object 70 and the radiation attenuation coefficient by performing an interpolation calculation using the radiation attenuation coefficient of the thickness of the reference object 70 for the radiation attenuation coefficient of the thickness not in the reference object 70. FIG. 13 is a diagram showing the relationship between the thickness t of the reference object 70 and the radiation attenuation coefficient. FIG. 13 shows the relationship between the thickness of the reference object 70 and the radiation attenuation coefficient in the case of a tube voltage of 90 kV and a total filtration amount of 2.5 mmAl. The attenuation coefficient deriving unit 63 derives the relationship between the thickness of the reference object 70 and the radiation attenuation coefficient for each energy characteristic of radiation, and stores the derived relationship in the storage of the derivation device 60.

[0076] The primary ray component derivation unit 64 derives a radiation attenuation coefficient corresponding to the thickness of the reference object 70 based on the relationship between the thickness of the reference object 70 and the radiation attenuation coefficient derived by the attenuation coefficient derivation unit 63. Furthermore, based on the radiation attenuation coefficient corresponding to the thickness of the reference object 70, the primary ray component included in the reference image K0 is derived.

[0077] Here, assuming that the pixel value of each pixel of the reference image K0 is I0o(x,y), the thickness of the reference object 70 corresponding to each pixel of the reference image K0 is T0(x,y), and the radiation attenuation coefficient derived by the above formula (4) for the thickness T0(x,y) of each pixel of the reference image K0 is μSoft0(x,y), the primary ray component derivation unit 64 derives the primary ray component I0p(x,y) contained in the pixel value of each pixel of the reference image K0 by the following formula (5). Since the reference object 70 has a plurality of thicknesses in stages, the primary ray component derivation unit 64 derives the primary ray component I0p(x,y) for each thickness in the reference object 70. For the primary ray components corresponding to thicknesses not in the reference object 70, the primary ray component derivation unit 64 may derive the relationship between the thickness of the reference object 70 and the primary ray components by performing an interpolation calculation using the primary ray components of the thicknesses in the reference object 70. I0p(x,y) = I0o(x,y)×exp(-μSoft0(x,y)×T0(x,y)) (5)

[0078] The scattered ray component derivation unit 65 derives the scattered ray components contained in the reference object 70 based on the difference between the pixel value of the reference image K0 and the primary ray components. That is, the scattered ray component derivation unit 65 derives the scattered ray components I0s(x,y) by the following formula (6). Since the reference object 70 has a plurality of thicknesses in stages, the scattered ray components I0s(x,y) corresponding to the stepped thicknesses of the reference object 70 are derived. For the scattered ray components corresponding to thicknesses not present in the reference object 70, the scattered ray component derivation unit 65 may derive the relationship between the thickness of the reference object 70 and the scattered ray components by performing an interpolation calculation using the scattered ray components of the thicknesses present in the reference object 70. I0s(x,y) = I0o(x,y)-I0p(x,y) (6)

[0079] The model derivation unit 66 derives a scattered radiation model that represents the relationship between the thickness of the reference object 70 and the ratio of the scattered radiation component I0s(x,y) to the primary radiation component I0p(x,y). That is, the model derivation unit 66 calculates the ratio of the scattered radiation component I0s(x,y) to the primary radiation component I0p(x,y) (i.e., I0s(x,y) / I0p(x,y)) as the STPR (Scatter-to-Primary Ratio) for each thickness of the reference object 70, and derives the scattered radiation model by plotting the relationship between the thickness of the reference object 70 and the STPR. Note that since the thickness of the reference object 70 differs stepwise, the STPR at a thickness not present in the reference object 70 may be derived by an interpolation calculation using the STPR at a thickness present in the reference object 70.

[0080] FIG. 14 is a diagram showing a scattered radiation model. FIG. 14 shows the relationship between the thickness of the reference object 70 and STPR when the tube voltage is 90 kV and the total filtration amount is 2.5 mmAl. The model derivation unit 66 stores the derived scattered radiation model in the storage of the derivation device 60. The reference object 70 simulates the radiation characteristics of the human body. Therefore, the scattered radiation model shown in FIG. 14 represents the relationship between the thickness of the subject H and STPR.

[0081] The scattered radiation model may be derived for each energy characteristic of radiation that can be emitted by the radiation source 2 of the imaging device 1 and stored in the storage of the deriving device 60.

[0082] The scattered radiation removal section 67 removes scattered radiation components from each of the first radiographic image G1 and the second radiographic image G2 acquired by the image acquisition section 61, using the scattered radiation model derived by the model derivation section 66. The removal of scattered radiation components will be described below. As a method for removing scattered radiation components, any method, such as the method described in JP 2015-043969 A, can be used. The scattered radiation removal process using the method described in JP 2015-043969 A will be described below. In the following description, the first and second radiographic images from which scattered radiation components have been removed will also be referred to as G1 and G2, respectively, as reference characters.

[0083] First, the scattered radiation removal unit 67 acquires a virtual model of the subject H having an initial body thickness distribution Ts(x,y). The virtual model is data virtually expressing the subject H in which the body thickness according to the initial body thickness distribution Ts(x,y) is associated with the coordinate position of each pixel of the first radiographic image G1. Note that the virtual model of the subject H having the initial body thickness distribution Ts(x,y) is stored in advance in the storage of the derivation device 60, but may be acquired from an external server in which the virtual model is saved.

[0084] Next, the scattered radiation removal unit 67 derives an estimated primary radiation image Ip(x,y) that is an estimate of a primary radiation image obtained by imaging the virtual model, and an estimated scattered radiation image Is(x,y) that is an estimate of a scattered radiation image obtained by imaging the virtual model, based on the virtual model, as shown in the following formulas (7) and (8). Furthermore, the scattered radiation removal unit 67 derives an image obtained by combining the estimated primary radiation image Ip(x,y) and the estimated scattered radiation image Is(x,y) as an estimated image Im(x,y) that is an estimate of the first radiation image G1 obtained by imaging the subject H, as shown in the following formula (9). Ip(x,y) = Io(x,y)×exp(-μSoft(x,y)×T(x,y)) (7) Is(x,y) = Io(x,y)×STPR(T(x,y))*PSF(T(x,y)) (8) Im(x,y) = Is(x,y) + Ip(x,y) (9)

[0085] Here, (x, y) are the coordinates of the pixel position of the first radiographic image G1, Io(x, y) is the pixel value of the first radiographic image G1 at the pixel position (x, y), Ip(x, y) is the primary ray component at the pixel position (x, y), and Is(x, y) is the scattered ray component at the pixel position (x, y). When deriving the first estimated image Im(x, y), the initial body thickness distribution Ts(x, y) is used as the body thickness distribution T(x, y) in equations (7) and (8). Furthermore, μSoft(T(x, y)) in equation (7) is derived by referring to the relationship between the thickness t of the reference object 70 and the radiation attenuation coefficient shown in FIG. 13, which is derived by the attenuation coefficient derivation unit 63. In addition, STPR(T(x, y)) in equation (8) is derived by referring to the scattered radiation model shown in FIG. 14, which is derived by the model derivation unit 66 based on the energy characteristics of the radiation used when photographing the subject H.

[0086] Moreover, PSF(T(x,y)) in formula (8) is a point spread function that represents the distribution of scattered rays spreading from one pixel according to the body thickness distribution T(x,y), and is defined according to the energy characteristics of the radiation. Moreover, * is an operator indicating a convolution operation. The PSF also varies depending on the distribution of the irradiation field in the imaging device 1, the distribution of the composition of the subject H, the irradiation dose during imaging, the tube voltage, the imaging distance, and the characteristics of the radiation detectors 5 and 6. For this reason, the PSF may be experimentally obtained in advance for each energy characteristic of the radiation used by the imaging device 1 according to the irradiation field information, the subject information, the imaging conditions, and the like, and may be stored in the storage of the derivation device 60.

[0087] Next, the scattered radiation removal unit 67 corrects the initial body thickness distribution Ts(x,y) of the virtual model so that the difference between the estimated image Im and the first radiographic image G1 becomes small. The scattered radiation removal unit 67 updates the body thickness distribution T(x,y), the scattered radiation component Is(x,y) and the primary radiation component Ip(x,y) by repeating the derivation of the body thickness distribution T(x,y), the scattered radiation component Is(x,y) and the primary radiation component Ip(x,y) until the difference between the estimated image Im and the first radiographic image G1 satisfies a predetermined termination condition. The scattered radiation removal unit 67 subtracts the scattered radiation component Is(x,y) derived by the formula (8) from the first radiographic image G1 when the termination condition is satisfied. This removes the scattered radiation component contained in the first radiographic image G1. The body thickness distribution T(x,y) when the termination condition is satisfied is used as the body thickness 42A of the correct answer data 42 contained in the teacher data 40.

[0088] On the other hand, the scattered radiation removal section 67 also performs the scattered radiation removal process on the second radiographic image G2 in the same manner as the first radiographic image G1.

[0089] The subtraction unit 68 performs energy subtraction processing to derive a bone image Gb from which bones of the subject H are extracted and a soft tissue image Gs from which soft tissue is extracted, from the first and second radiological images G1, G2 that have been subjected to scattered radiation removal processing. Note that the first and second radiological images G1, G2 in the subsequent processing are processed radiological images from which scattered radiation components have been removed.

[0090] When deriving the bone image Gb, the subtraction unit 68 performs weighted subtraction between corresponding pixels of the first and second radiographic images G1, G2 as shown in the following formula (10), thereby generating a bone image Gb in which the bones of the subject H included in each of the radiographic images G1, G2 are extracted, as shown in Fig. 15. In formula (10), β1 is a weighting coefficient, and is set to a value that allows the bones of the subject H included in each of the radiographic images G1, G2 to be extracted by formula (10) based on the radiation attenuation coefficients of bone tissue and soft tissue. The pixel value of each pixel in the bone region of the bone image Gb becomes the bone pixel value. Gb(x, y)=G1(x, y)-β1×G2(x, y) (10)

[0091] On the other hand, when deriving a soft tissue image Gs, the subtraction unit 68 performs weighted subtraction between corresponding pixels of the first and second radiographic images G1, G2 as shown in the following equation (11) to generate a soft tissue image Gs from which the soft tissue of the subject H contained in each of the radiographic images G1, G2 has been extracted, as shown in Fig. 16. In equation (11), β2 is a weighting coefficient, which is set to a value that enables extraction of the soft tissue of the subject H contained in each of the radiographic images G1, G2 by equation (11), based on the radiation attenuation coefficients of bone tissue and soft tissue. Gs(x, y)=G1(x, y)-β2×G2(x, y) (11)

[0092] The soft tissue image Gs represents a soft tissue region of the soft tissue of the subject H. In this embodiment, the "soft tissue" of the subject H refers to tissue other than bone tissue, and specifically includes muscle tissue, fat tissue, blood, and water.

[0093] The composition derivation unit 69 derives the composition of the subject H. Specifically, the composition derivation unit 69 derives the bone mineral content as the composition for each pixel of the bone region of the bone image Gb based on the pixel values ​​of the bone image Gb. The composition derivation unit 69 also derives the muscle content as the composition for each pixel of the soft tissue region in the soft tissue image Gs. First, the derivation of the bone mineral content will be described.

[0094] The composition deriving unit 69 derives bone mineral mass for each pixel of the bone image Gb. In this embodiment, the composition deriving unit 69 derives bone mineral mass by converting each pixel value of the bone image Gb into a pixel value of a bone image acquired under standard imaging conditions. Specifically, the composition deriving unit 69 derives bone mineral mass by correcting each pixel value of the bone image Gb using a correction coefficient acquired from a lookup table described later.

[0095] Here, the higher the tube voltage in the radiation source 2 and the higher the energy of the radiation emitted from the radiation source 2, the smaller the contrast between soft tissue and bone in the radiological image. In addition, in the process of the radiation passing through the subject H, low-energy components of the radiation are absorbed by the subject H, causing beam hardening, in which the radiation becomes more energetic. The increase in the energy of the radiation due to beam hardening becomes greater as the body thickness of the subject H increases.

[0096] FIG. 17 is a diagram showing the relationship between the contrast of bones and soft tissues and the body thickness of subject H. FIG. 17 shows the relationship between the contrast of bones and soft tissues and the body thickness of subject H at three tube voltages of 80 kV, 90 kV, and 100 kV. As shown in FIG. 17, the higher the tube voltage, the lower the contrast. Furthermore, when the body thickness of subject H exceeds a certain value, the larger the body thickness, the lower the contrast. Furthermore, the larger the pixel value of the bone region in bone image Gb, the higher the contrast between bones and soft tissues. Therefore, the relationship shown in FIG. 17 shifts to the high contrast side as the pixel value of the bone region in bone image Gb increases.

[0097] In this embodiment, a lookup table for acquiring correction coefficients for correcting the difference in contrast in the bone image Gb according to the tube voltage at the time of imaging and the decrease in contrast due to the influence of beam hardening is stored in the storage of the derivation device 60. The correction coefficients are coefficients for correcting each pixel value of the bone image Gb.

[0098] FIG. 18 is a diagram showing an example of a lookup table stored in the storage of the derivation device 60. In FIG. 18, a lookup table LUT1 in which a reference imaging condition is set to a tube voltage of 90 kV is illustrated. As shown in FIG. 18, in the lookup table LUT1, a larger correction coefficient is set as the tube voltage increases and as the body thickness of the subject H increases. In the example shown in FIG. 18, since the reference imaging condition is a tube voltage of 90 kV, when the tube voltage is 90 kV and the body thickness is 0, the correction coefficient is 1. Note that, although the lookup table LUT1 is shown in two dimensions in FIG. 18, the correction coefficient differs depending on the pixel value of the bone region. Therefore, the lookup table LUT1 is actually a three-dimensional table to which an axis representing the pixel value of the bone region is added.

[0099] The composition derivation unit 69 extracts, from the lookup table LUT1, a correction coefficient C0(x, y) for each pixel that corresponds to the body thickness distribution T(x, y) of the subject H and the setting value of the tube voltage of the radiation source 2 included in the imaging conditions stored in the storage of the derivation device 60. For the body thickness distribution T(x, y) of the subject H, the body thickness distribution T(x, y) obtained when the end condition is satisfied in the scattered radiation removal unit 67 is used. Then, as shown in the following formula (12), the composition derivation unit 69 multiplies the pixel value Gb(x, y) of each pixel in the bone region in the bone image Gb by the correction coefficient C0(x, y) to obtain the bone mineral content B(x, y) (g / cm) for each pixel of the bone image Gb. 2 ) is derived. The bone mineral content B(x, y) derived in this manner represents the pixel values ​​of the bone region contained in the radiographic image obtained by imaging the subject H under the standard imaging condition of a tube voltage of 90 kV and from which the effects of beam hardening have been removed. The bone mineral content B(x, y) derived by equation (12) is used as the bone mineral content 42B of the correct answer data 42 contained in the training data 40. B(x,y)=C0(x,y)×Gb(x,y) (12)

[0100] Here, the correction coefficient C0(x,y) can be derived as follows. First, the radiation attenuation coefficient of the bone tissue according to the thickness of the soft tissue of the subject H is derived. Assuming that the soft tissue and the bone tissue are overlapped on the radiation transmission path, and the thickness of the soft tissue is tsoft, the radiation attenuation coefficient can be derived as a function depending on the thickness of the soft tissue. When the energy spectrum of the radiation emitted from the radiation source 2 is Sin(E) and the thickness of the soft tissue of the subject H is tsoft, the radiation dose Xout1(tsoft) after passing through the subject H in the case where there is no bone tissue can be calculated by the following formula (13) for each thickness t of the subject H using the radiation attenuation characteristic μSoft(E) of the soft tissue of the human body. Note that in formula (13), the radiation attenuation coefficient of the object (i.e., the tabletop 3A and the grid 4) existing between the subject H and the radiation detectors 5 and 6 is taken into account, as in the above formula (2).

number

[0101] The radiation dose Xout2(t) in the presence of bone tissue is further derived from the following equation (14) using the radiation attenuation coefficient μBone(E) of the bone tissue.

number

[0102] The radiation attenuation coefficient of bone tissue is expressed as an exponential decay of the radiation dose due to bone tissue, based on the radiation dose when there is no bone tissue, as shown in the following equation (15).

number

[0103] By solving equation (15) for μBone(t) as shown in the following equation (16), the relationship between the thickness t of the subject H and the radiation attenuation coefficient of bone tissue can be derived, where tBone is the thickness of the bone tissue.

number

[0104] The correction coefficient C0(t) is calculated by multiplying the inverse of the attenuation coefficient of the bone region for each thickness of the subject H by the following equation (17), so that the bone contrast of the attenuation coefficient μBone(tbase) of the bone region at the average thickness tbase of the subject H defined according to the imaging area is used as the reference.

number

[0105] Next, derivation of muscle mass will be described. As described above, soft tissue includes muscle tissue, fat tissue, blood, and water. In this embodiment, tissue other than fat tissue in soft tissue is considered to be muscle tissue. That is, in this embodiment, non-fatty tissue including blood and water is included in muscle tissue and treated as muscle tissue.

[0106] The composition deriving unit 69 separates muscle and fat in the soft tissue image Gs by utilizing the difference in energy characteristics between muscle tissue and fat tissue. Here, as shown in FIG. 19, the dose of radiation after passing through the subject H, which is a human body, is lower than that of radiation before entering the subject H. In addition, since muscle tissue and fat tissue absorb different energies and have different radiation attenuation coefficients, the energy spectrum of the radiation after passing through the subject H is different from that of the radiation after passing through the muscle tissue. As shown in FIG. 19, the energy spectrum of the radiation that passes through the subject H and is irradiated to each of the first radiation detector 5 and the second radiation detector 6 depends on the body composition of the subject H, specifically, the ratio of muscle tissue to fat tissue. Since fat tissue is more easily permeable to radiation than muscle tissue, the dose of radiation after passing through the human body is lower when the ratio of muscle tissue is higher than that of fat tissue.

[0107] Therefore, the composition deriving unit 69 separates muscle and fat in the soft tissue image Gs by utilizing the difference in the energy characteristics of the muscle tissue and fat tissue described above. That is, the composition deriving unit 69 generates a muscle image and a fat image from the soft tissue image Gs. The composition deriving unit 69 also derives the muscle mass of each pixel based on the pixel value of the muscle image.

[0108] The specific method by which the composition deriving unit 69 separates muscle and fat from the soft tissue image Gs is not limited. As an example, the composition deriving unit 69 generates a muscle image from the soft tissue image Gs using the following formulas (18) and (19). Specifically, the composition deriving unit 69 first derives the muscle ratio rm(x, y) at each pixel position (x, y) in the soft tissue image Gs using formula (18). In formula (18), μm is a weighting coefficient according to the radiation attenuation coefficient of muscle tissue, and μA is a weighting coefficient according to the radiation attenuation coefficient of fat tissue. Also, Δ(x, y) represents the concentration difference distribution. The concentration difference distribution is the distribution on the image of the concentration change seen from the concentration obtained by the radiation reaching the first radiation detector 5 and the second radiation detector 6 without penetrating the subject H. The distribution of density changes in the image is calculated by subtracting the density of each pixel in the area of ​​the subject H from the density in the non-radiated area obtained by directly irradiating the first radiation detector 5 and the second radiation detector 6 with radiation in the soft tissue image Gs. rm(x,y)={μA-Δ(x,y) / T(x,y)} / (μA-μm) (18)

[0109] Furthermore, the composition deriving unit 69 generates a muscle image Gm from the soft tissue image Gs by the following formula (19). Gm(x,y)=rm(x,y)×Gs(x,y) (19)

[0110] Then, the composition derivation unit 69 multiplies each pixel (x, y) of the muscle image Gm by a correction coefficient C1(x, y) that represents a relationship between a predetermined pixel value and muscle mass, as shown in the following formula (20), to calculate the muscle mass M(x, y) (g / cm) for each pixel of the muscle image Gm. 2The muscle mass M(x, y) derived by the formula (20) is used as the muscle mass 42C of the correct answer data 42 included in the training data 40. M(x,y)=C1(x,y)×Gm(x,y) (20)

[0111] Here, the correction coefficient C1(x, y) can be derived as follows: First, derive the attenuation coefficient of bone according to the thickness of the soft tissue of the subject H. Assuming that fat and muscle overlap on the radiation transmission path, and assuming that the thickness of fat is tA, the radiation attenuation coefficient can be derived as a function that depends on the thickness of fat.

[0112] When the energy spectrum of radiation emitted from the radiation source 2 is Sin(E) and the thickness of fat of the subject H is tA, the radiation dose Xout3(t) after passing through the subject H in the absence of muscle can be calculated by the following formula (21) for each thickness t of the subject H, using the radiation attenuation characteristics μA(E) of human fat. Note that formula (21) takes into account the radiation attenuation coefficient of objects (i.e., the tabletop 3A and grid 4) that exist between the subject H and the radiation detectors 5 and 6, as in formula (2).

number

[0113] The radiation dose Xout4(t) when muscle is present is further derived using the radiation attenuation coefficient μM(E) of muscle according to the following equation (22).

number

[0114] The radiation attenuation coefficient of muscle is expressed as an exponential decay of the radiation dose due to muscle, based on the radiation dose when there is no muscle, as shown in the following equation (23).

number

[0115] The relationship between fat thickness tA and muscle attenuation coefficient can be derived by solving equation (23) for μM(t) as shown in equation (24) below, where tM is muscle thickness.

number

[0116] The correction coefficient C1(t) is calculated by multiplying the inverse of the muscle attenuation coefficient for each thickness of the subject H using the muscle contrast of the muscle radiation attenuation coefficient μM(tbase) at the average thickness tbase of the subject H defined according to the imaging area as the standard, as shown in the following formula (25).

number

[0117] Next, the process performed in this embodiment will be described. FIG. 20 is a flow chart showing the learning process performed in this embodiment. First, the information acquisition unit 22 acquires the teacher data 40 from the image storage system 9 (step ST1), and the learning unit 23 inputs the learning data 41 included in the teacher data 40 to the neural network 30 to output the body thickness, bone mineral content, and muscle mass, and learns the neural network 30 using the loss L0 based on the difference from the correct answer data 42 (step ST2), and returns to step ST1. Then, the learning unit 23 repeats the processes of steps ST1 and ST2 until the end condition is satisfied, and ends the learning process. As a result, the learning unit 23 constructs a trained neural network 24A.

[0118] Next, the radiation image processing for deriving the estimation results of the body thickness, bone salt mass, and muscle mass in this embodiment will be described. FIG. 21 is a flowchart showing the radiation image processing in this embodiment. It is assumed that the simple radiation image G0 to be processed is acquired by photography and stored in the storage 13. When an instruction to start the processing is input from the input device 15, the image acquisition unit 21 acquires the simple radiation image G0 from the storage 13 (step ST11). Next, the estimation unit 24 derives the estimation results of the body thickness, bone salt mass, and muscle mass from the simple radiation image G0 (step ST12). Then, the display control unit 25 displays the estimation results together with the simple radiation image G0 on the display 14 (step ST13), and the processing ends.

[0119] In this embodiment, the trained neural network 24A is constructed using training data 41 including a reference image 41A, radiation energy characteristics 41B, object characteristics 41C, and imaging conditions 41D, and teacher data 40 including correct answer data 42 including a body thickness 42A, bone mineral content 42B, and muscle mass 42C derived from a radiation image of a subject using the reference image 41A, radiation energy characteristics 41B, object characteristics 41C, and imaging conditions 41D. The trained neural network 24A is then used to derive estimated results of the body thickness, bone mineral content, and muscle mass for a simple radiation image G0 acquired by imaging a target subject H.

[0120] In this embodiment, when deriving the body thickness 42A, bone salt mass 42B and muscle mass 42C as the correct answer data 42, the relationship between the thickness of the reference object 70 reflecting the influence of beam hardening of objects such as the tabletop 3A and the grid 4 interposed between the subject H and the first and second radiation detectors 5 and 6 and the radiation attenuation coefficient is derived using the energy characteristics of the radiation in the imaging device 1, and the primary ray components and scattered ray components corresponding to the thickness of the reference object 70 contained in the reference image K0 are derived based on the derived relationship, and a scattered ray model expressing the relationship between the thickness of the reference object 70 and the ratio of the scattered ray components to the primary ray components is derived. The scattered ray components are then removed using the derived scattered ray model, and the body thickness, bone salt mass and muscle mass are derived using the first and second radiation images G1 and G2 from which the scattered ray components have been removed, taking into consideration the energy characteristics of the radiation irradiated to the subject H and the objects interposed between the subject and the radiation detectors.

[0121] Therefore, according to this embodiment, by using the constructed trained neural network 24A, it is possible to accurately estimate the body thickness, bone mineral content, and muscle mass of the target subject H, taking into account the energy characteristics of the radiation irradiated to the subject H and objects interposed between the subject H and the radiation detector.

[0122] In the above embodiment, the trained neural network 24A is constructed so as to output the body thickness, bone mineral content, and muscle mass in response to the input of the simple radiographic image G0 of the subject H, but the present invention is not limited to this. The trained neural network 24A may be constructed so as to output the body thickness, bone mineral content, and muscle mass in response to the input of the first and second radiographic images G1 and G2 of the subject H, which have different energy distributions.

[0123] In the above embodiment, the trained neural network 24A is constructed to output body thickness, bone salt mass, and muscle mass, but is not limited thereto. The trained neural network 24A may be constructed to output any one of body thickness, bone salt mass, and muscle mass, or a combination of two of them. In this case, teacher data 40 including any one of body thickness, bone salt mass, and muscle mass, or a combination of two of them as correct answer data 42 may be prepared. The trained neural network 24A constructed in this manner outputs any one of body thickness, bone salt mass, and muscle mass, or a combination of two of them, when the plain radiographic image G0 or the first and second radiographic images G1 and G2 are input.

[0124] In the above embodiment, the objects interposed between the subject H and the radiation detectors 5, 6 are the tabletop 3A and the grid 4, but are not limited thereto. When the grid 4 is not used, the object interposed between the subject H and the radiation detectors 5, 6 is only the tabletop 3A. When the subject H is photographed in an upright position, not in a supine position, the tabletop 3A may not be used and only the grid 4 may be used. In this case, the object interposed between the subject H and the radiation detectors 5, 6 is only the grid 4. In this way, when the object interposed between the subject H and the radiation detectors 5, 6 is only the tabletop 3A or only the grid 4, the relationship between the thickness and the radiation attenuation coefficient of the reference object 70, which reflects the influence of the radiation attenuation coefficient of only the tabletop 3A or the beam hardening of only the grid 4, may be derived to derive the correct answer data 42.

[0125] In the above embodiment, when deriving the muscle mass 42C as the correct answer data 42, an energy subtraction process may be performed to derive a muscle image Gm in which muscles in the soft tissues of the subject H are extracted from the first and second radiographic images G1, G2 that have been subjected to the scattered radiation removal process. When deriving the muscle image Gm, the subtraction unit 68 may perform weighted subtraction between corresponding pixels on the first and second radiographic images G1, G2 as shown in the following formula (26) to generate a muscle image Gm in which the muscles of the subject H included in each of the radiographic images G1, G2 are extracted. In formula (26), β3 is a weighting coefficient, and is set to a value that allows the muscles of the subject H included in each of the radiographic images G1, G2 to be extracted by formula (26) based on the radiation attenuation coefficients of fat and muscle. Gm(x, y)=G1(x, y)-β3×G2(x, y) (26)

[0126] In the above embodiment, the first and second radiographic images G1, G2 are acquired by a one-shot method, but the present invention is not limited thereto. The first and second radiographic images G1, G2 may be acquired by a so-called two-shot method in which imaging is performed twice. In the case of the two-shot method, the position of the subject H included in the first radiographic image G1 and the second radiographic image G2 may be shifted due to the body movement of the subject H. For this reason, it is preferable to perform the processing of this embodiment after aligning the position of the subject in the first radiographic image G1 and the second radiographic image G2. For example, the method described in JP-A-2011-655060 can be used as the alignment processing. The method described in JP 2011-655060 A involves generating a plurality of first band images and a plurality of second band images representing structures in different frequency bands for each of the first and second radiographic images G1, G2, obtaining the amount of positional shift between corresponding positions in the first band image and the second band image of the corresponding frequency band, and aligning the first radiographic image G1 and the second radiographic image G2 based on the amount of positional shift.

[0127] In the above embodiment, the learning process and the radiation image process are performed using the radiation image of the subject acquired in the system that captures the image of the subject H using the first and second radiation detectors 5 and 6. However, the technology of the present disclosure can be applied to a system that acquires the radiation image of the subject using a stimulable phosphor sheet instead of a radiation detector. In this case, the simple radiation image G0 may be acquired by irradiating one stimulable phosphor sheet with radiation that has passed through the subject H, accumulating and recording the radiation image information of the subject H on the stimulable phosphor sheet, and photoelectrically reading the radiation image information from the stimulable phosphor sheet. The first and second radiation images G1 and G2 may be acquired by irradiating two stimulable phosphor sheets that are stacked together with radiation that has passed through the subject H, accumulating and recording the radiation image information of the subject H on each stimulable phosphor sheet, and photoelectrically reading the radiation image information from each stimulable phosphor sheet. The two-shot method may also be used when acquiring the first and second radiation images G1 and G2 using stimulable phosphor sheets.

[0128] Moreover, the radiation in the above embodiment is not particularly limited, and in addition to X-rays, α-rays, γ-rays, etc. can be used.

[0129] In the above embodiment, the following various processors can be used as the hardware structure of the processing units that execute various processes, such as the image acquisition unit 21, the information acquisition unit 22, the learning unit 23, the estimation unit 24, and the display control unit 25. As described above, the above various processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as a programmable logic device (PLD), which is a processor whose circuit configuration can be changed after manufacture, such as an FPGA (Field Programmable Gate Array), a dedicated electric circuit, which is a processor having a circuit configuration designed specifically for executing specific processes, such as an ASIC (Application Specific Integrated Circuit), etc.

[0130] 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 (e.g., 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.

[0131] As an example of configuring multiple processing units with one processor, first, there is a form in which one processor is configured with a combination of one or more CPUs and software, as typified by computers such as client and server, and this processor functions as multiple processing units. Second, there is a form in which a processor is used that realizes the functions of the entire system including multiple processing units with one IC (Integrated Circuit) chip, as typified by System On Chip (SoC), etc. In this way, the various processing units are configured using one or more of the above various processors as a hardware structure.

[0132] Furthermore, more specifically, the hardware structure of these various processors can be an electric circuit that combines circuit elements such as semiconductor elements.

[0133] The following are appended claims to this disclosure. (Additional note 1) At least one processor; The processor, obtaining at least one reference image representing a reference object having a different thickness, the at least one reference image being obtained by imaging the reference object with radiation while an object is interposed between the reference object and a radiation detector; deriving a relationship between a thickness of the reference object and a radiation attenuation coefficient of the reference object, the relationship reflecting the influence of the reference object and beam hardening caused by the object, corresponding to the energy characteristics of the radiation; deriving a primary ray component corresponding to a thickness of the reference object included in the reference image based on a relationship between a thickness of the reference object and a radiation attenuation coefficient of the reference object; deriving a scattered ray component corresponding to a thickness of the reference object included in the reference image based on a difference between the reference image and the primary ray component; A scattered ray model derivation device that derives a scattered ray model that indicates a relationship between a thickness of the reference object and a ratio of the scattered ray component to the primary ray component. (Additional note 2) The scattered ray model derivation device described in Appendix 1, wherein the processor derives the relationship between the thickness of the reference object and the radiation attenuation coefficient of the reference object based on the energy characteristics of the radiation, the radiation attenuation coefficient of the reference object corresponding to the energy characteristics, the thickness of the reference object, the radiation attenuation coefficient of the object corresponding to the energy characteristics, and the thickness of the object. (Additional note 3) The processor of the scattered ray model derivation device of appendix 1 derives a relationship between the thickness of the reference object and the radiation attenuation coefficient of the reference object based on the energy characteristics of the radiation, the radiation attenuation coefficient of the reference object corresponding to the energy characteristics, the thickness of the reference object, the radiation attenuation coefficient of the object corresponding to the energy characteristics, and the energy characteristics of the radiation before and after passing through the object. (Additional note 4) The processor derives a relationship between the thickness of the reference object and the radiation attenuation coefficient of the reference object for a thickness not present in the reference object by interpolating a relationship between the thickness of the reference object and the radiation attenuation coefficient of the reference object for a thickness present in the reference object; deriving primary line components for thicknesses not present in the reference object by interpolating primary line components for thicknesses present in the reference object; deriving a scattered radiation component for a thickness not present in the reference object by interpolating a scattered radiation component for a thickness present in the reference object; 4. The scattered radiation model derivation device according to any one of claims 1 to 3, which derives a scattered radiation model for a thickness not present in the reference object by interpolating a scattered radiation model for a thickness present in the reference object. (Additional note 5) The scattered ray model derivation device according to any one of appendix 1 to 4, wherein the object is at least one of a top plate of an imaging table on which a subject is placed in an imaging device, and a scattered ray anti-grid for removing scattered ray components from radiation that has passed through the subject. (Additional note 6) at least one processor; The processor, acquiring at least one radiological image of the subject; Obtaining a body thickness distribution of the subject; 6. A radiation image processing device that derives at least one processed radiation image by removing scattered radiation components contained in the radiation image using a scattered radiation model derived by the scattered radiation model derivation device according to any one of appended items 1 to 5 and a body thickness distribution of the subject. (Additional note 7) the processor derives a primary ray component included in the at least one radiation image using the scattered ray model and a body thickness distribution of the subject; The radiation image processing device according to claim 6, wherein the at least one processed radiation image is derived by updating the body thickness distribution, the scattered radiation component, and the primary radiation component until a difference between the primary radiation component and the processed radiation image satisfies a predetermined condition. (Additional note 8) the radiation images of the subject are a first radiation image and a second radiation image based on radiation having different energy distributions that have passed through the subject including a bone and a soft tissue, the processor derives a first processed radiographic image and a second processed radiographic image for the first radiographic image and the second radiographic image, respectively, by removing scattered radiation components from the first radiographic image and the second radiographic image, respectively, using the scattered radiation model; 8. The radiation image processing device according to claim 6 or 7, wherein a composition of the subject is derived from the first processed radiation image and the second processed radiation image. (Additional note 9) the processor derives a bone image by extracting a bone of the subject from the first processed radiographic image and the second processed radiographic image; 9. The radiological image processing device according to claim 8, further comprising: a bone mineral content for each pixel of a bone region of the bone image being derived as the composition based on pixel values ​​of the bone image. (Additional note 10) The radiological image processing device according to claim 9, wherein the processor derives bone mineral content for each pixel of the bone region by correcting pixel values ​​of the bone image using a correction coefficient derived based on the radiation attenuation coefficient of the bone. (Additional note 11) 11. The radiological image processing device according to claim 10, wherein the correction coefficient is derived also based on a radiation attenuation coefficient of an object interposed between the subject and a radiation detector that acquires the first and second radiological images. (Additional note 12) The processor derives muscle images by extracting muscles of the subject from the first processed radiation image and the second processed radiation image; The radiological image processing device according to claim 8, further comprising: a muscle mass deriving unit for deriving the composition as the composition based on pixel values ​​of the muscle image. (Additional note 13) the processor derives a soft part image by extracting a soft part of the subject from the first processed radiographic image and the second processed radiographic image; deriving the muscle image from the soft tissue image; The radiological image processing device according to claim 12, wherein a muscle mass is derived for each pixel of the muscle image by correcting pixel values ​​of the muscle image with a correction coefficient derived based on a radiation attenuation coefficient of the muscle. (Additional note 14) 14. The radiological image processing device according to claim 13, wherein the correction coefficient is derived also based on a radiation attenuation coefficient of an object interposed between the subject and a radiation detector that acquires the first and second radiological images. (Additional note 15) 15. The radiation image processing device according to claim 8, wherein the processor displays the composition on a display. (Additional note 16) obtaining at least one reference image representing a reference object having a different thickness, the at least one reference image being obtained by imaging the reference object with radiation while an object is interposed between the reference object and a radiation detector; deriving a relationship between a thickness of the reference object and a radiation attenuation coefficient of the reference object, the relationship reflecting the influence of the reference object and beam hardening caused by the object, corresponding to the energy characteristics of the radiation; deriving a primary ray component corresponding to a thickness of the reference object included in the reference image based on a relationship between a thickness of the reference object and a radiation attenuation coefficient of the reference object; deriving a scattered ray component corresponding to a thickness of the reference object included in the reference image based on a difference between the reference image and the primary ray component; A method for deriving a scattered ray model that derives a scattered ray model that represents a relationship between a thickness of the reference object and a ratio of the scattered ray component to the primary ray component. (Additional note 17) acquiring at least one radiological image of the subject; Obtaining a body thickness distribution of the subject; 6. A radiological image processing method for deriving at least one processed radiological image by removing scattered radiation components contained in the radiological image using a scattered radiation model derived by the scattered radiation model derivation device according to any one of appended items 1 to 5 and a body thickness distribution of the subject. (Additional note 18) obtaining at least one reference image representing a reference object of a different thickness, the at least one reference image being obtained by imaging the reference object with radiation while an object is interposed between the reference object and a radiation detector; deriving a relationship between a thickness of the reference object and a radiation attenuation coefficient of the reference object, the relationship reflecting the influence of the reference object and beam hardening caused by the object, corresponding to the energy characteristics of the radiation; deriving a primary ray component corresponding to a thickness of the reference object included in the reference image based on a relationship between a thickness of the reference object and a radiation attenuation coefficient of the reference object; deriving a scattered ray component corresponding to a thickness of the reference object included in the reference image based on a difference between the reference image and the primary ray component; and a procedure for deriving a scattered ray model that represents a relationship between a thickness of the reference object and a ratio of the scattered ray component to the primary ray component. (Additional note 19) acquiring at least one radiological image of a subject; acquiring a body thickness distribution of the subject; and a step of deriving at least one processed radiographic image by removing scattered radiation components contained in the radiographic image using the scattered radiation model derived by the scattered radiation model derivation device according to any one of appended items 1 to 5 and the body thickness distribution of the subject. [Explanation of symbols]

[0134] 1. Imaging device 2 Radiation source 3. Shooting stand 3A Top Board 3B Mounting part 4. Grid 5,6 Radiation detectors 7 Radiation Energy Conversion Filter 10 Radiation image processing device 11 CPU 12. Memory 13. Storage 14 Display 15 Input Devices 16 Memory 17 Network Interface 18 Bus 21 Image acquisition unit 22 Information Acquisition Department 23 Learning Department 24 Estimation part 24A Trained Neural Network 25 Display control section 30 Neural Networks 31 Input Layer 32 Middle Class 33 Output layer 35 Convolutional Layer 36 Pooling Layer 37 Fully connected layer 40 Teacher Data 41 Training Data 41A Reference Image 41B Radiation Energy Characteristics 41C Object special 41D Shooting conditions 42 Correct data 42A body thickness 42B Bone mineral content 42C Muscle Mass 47 Output Data 48 Parameters 50 display screen 51,52 References 60 Derivation device 61 Image acquisition unit 62 Information Acquisition Department 63 Attenuation coefficient derivation part 64 Primary line component derivation part 65 Scattered radiation component derivation part 66 Model Derivation Section 67 Scattered radiation removal section 68 Subtraction Section 69 Composition derivation part 70 Reference object G1 First radiological image G2 Second radiographic image Ge bone mineral image Gb bone image Gm muscle images Gs Soft tissue image Gt body thickness image K0 Reference Image LUT1 Lookup table

Claims

1. At least one processor; the processor performs machine learning on a neural network using teacher data consisting of learning data and correct answer data, thereby constructing a trained neural network that outputs at least one of a composition of the target subject and a body thickness of the target subject when at least one target radiation image of the target subject is input; The learning data is a reference image obtained by irradiating a reference object, which has a thickness simulating a human body and is made of a material having a radiation transmittance similar to that of the soft tissue of the human body, with radiation being placed between the reference object and a radiation detector; an energy characteristic of the radiation; The thickness and material of said object, and The photographing conditions when the reference image was acquired, The correct answer data is the reference image, the energy characteristics of the radiation, the thickness and material of the object, and at least one of a composition and a body thickness of the subject derived from at least one radiation image of the subject using the imaging conditions; Learning device.

2. 2. The learning device according to claim 1, wherein the object is at least one of a top plate of an imaging table on which a subject is placed in an imaging device and an anti-scatter grid for removing scattered radiation components from radiation that has passed through the subject.

3. 3. The learning device according to claim 1, wherein the target radiation image is a simple radiation image based on radiation that has passed through the target subject including bones and soft tissue.

4. 3. The learning device according to claim 1, wherein the target radiographic images are a first radiographic image and a second radiographic image based on radiation having different energy distributions transmitted through the target subject including a bone and a soft tissue.

5. The learning device according to claim 1 , wherein the composition of the subject and the composition of the target subject are at least one of bone mineral mass and muscle mass.

6. At least one processor; The processor, obtaining at least one target radiographic image of a target-subject; 6. A radiation image processing device that derives an estimation result of at least one of a composition of the target subject and a body thickness of the target subject, using a trained neural network constructed by the learning device according to claim 1 and the target radiation image.

7. The radiation image processing apparatus according to claim 6 , wherein the target radiation image is a simple radiation image based on radiation transmitted through the target subject including bones and soft tissue.

8. The radiation image processing apparatus according to claim 6 , wherein the target radiation images are a first radiation image and a second radiation image based on radiation having different energy distributions transmitted through the target subject including a bone and a soft tissue.

9. The radiation image processing apparatus according to claim 6 , wherein the composition of the target subject is at least one of a bone mineral content and a muscle mass.

10. A learning method for constructing a trained neural network by machine learning using teacher data consisting of learning data and correct answer data, the trained neural network being configured to output at least one of a composition of a target subject and a body thickness of the target subject when at least one target radiation image of the target subject is input, the method comprising: The learning data is a reference image obtained by irradiating a reference object, which has a thickness simulating a human body and is made of a material having a radiation transmittance similar to that of the soft tissue of the human body, with radiation being placed between the reference object and a radiation detector; an energy characteristic of the radiation; The thickness and material of said object, and The photographing conditions when the reference image was acquired, The correct answer data is the reference image, the energy characteristics of the radiation, the thickness and material of the object, and at least one of a composition and a body thickness of the subject derived from at least one radiation image of the subject using the imaging conditions; How to learn.

11. obtaining at least one target radiographic image of a target-subject; 6. A radiological image processing method for deriving an estimation result of at least one of a composition of the target subject and a body thickness of the target subject, using a trained neural network constructed by the learning device according to claim 1 and the target radiological image.

12. A learning program for causing a computer to execute a procedure for constructing a trained neural network by machine learning using teacher data consisting of learning data and correct answer data, the trained neural network outputting at least one of a composition of a target subject and a body thickness of the target subject when at least one target radiation image of the target subject is input, the program comprising: The learning data is a reference image obtained by irradiating a reference object, which has a thickness simulating a human body and is made of a material having a radiation transmittance similar to that of the soft tissue of the human body, with radiation being placed between the reference object and a radiation detector; an energy characteristic of the radiation; The thickness and material of said object, and The photographing conditions when the reference image was acquired, The correct answer data is the reference image, the energy characteristics of the radiation, the thickness and material of the object, and at least one of a composition and a body thickness of the subject derived from at least one radiation image of the subject using the imaging conditions; Study program.

13. acquiring at least one target radiographic image of a target-subject; and a step of deriving an estimation result of at least one of a composition of the target subject and a body thickness of the target subject, using the trained neural network constructed by the learning device according to any one of claims 1 to 5 and the target radiographic image.

Images