Estimation device, method, and computer-readable storage medium

Abstract

The present application relates to a device, method and computer readable storage medium for estimating bone density. The device comprises at least one processor configured to function as a learned neural network configured to derive an estimation result related to bone density of a bone portion from a simple radiograph image acquired by simply photographing a subject including the bone portion. The learned neural network is learned using (i) two radiograph images acquired by photographing a subject including a bone portion using radiations having different energy distributions, (ii) a radiograph image of the subject and a bone portion image representing the bone portion of the subject, or (iii) a radiograph image of the subject and a bone density image representing bone density of the bone portion of the subject and information related to the bone density of the bone portion of the subject.

Description

Technical Field

[0001] This invention relates to an estimation device, method, and procedure. Background Technology

[0002] In osteoporosis and other skeletal diseases, DXA (Dual X-ray Absorptiometry) is a known method for quantifying bone mineral density, representing a representative approach for diagnosing bone mineral density. DXA works by utilizing the attenuation coefficient μ (cm²) of radiation incident on and transmitted through the human body, dependent on the constituent materials of the body (e.g., bone). 2 / g) and its density ρ (g / cm³) 3 The attenuation is characterized by thickness t (cm), and the amount of bone salt is calculated based on the pixel values ​​of the radiographic images obtained by taking radiographic images with two energies of radiation.

[0003] Furthermore, various methods for evaluating bone density using radiographic images acquired by photographing a subject have been proposed. For example, Patent Documents 1 and 2 propose methods for estimating information related to bone density based on radiographic images of bone using a fully trained neural network constructed through learning. In the method described in Patent Document 1, radiographic images of bone and bone density acquired simply by photographing are used as teacher data for learning the neural network. Moreover, in the method described in Patent Document 1, radiographic images of bone, bone density, and information related to bone density (e.g., age, sex, weight, drinking habits, smoking habits, fracture history, body fat percentage, and subcutaneous fat percentage) acquired simply by photographing are used as teacher data for learning the neural network.

[0004] Furthermore, simple imaging is a photographic method that obtains a two-dimensional image of the subject by irradiating it with radiation once. In the following description, the radiation image obtained by simple imaging will be referred to as a simple radiation image.

[0005] Patent Document 1: US Patent No. 6064716

[0006] Patent Document 2: International Publication No. 2020 / 054738

[0007] However, there is a desire for even higher precision in estimating bone mineral density. Summary of the Invention

[0008] The present invention was made in view of the above circumstances, and its object is to enable the estimation of bone mineral density with high accuracy.

[0009] The estimation device based on the present invention includes at least one processor.

[0010] The processor functions as a learned neural network that derives inferences related to bone density based on simple radiographic images of a subject containing bones.

[0011] The neural network was learned by learning from the following as teacher data: (i) two radiographic images of a subject containing bones obtained by taking pictures of the subject using radiation with different energy distributions, (ii) a radiographic image of the subject and a bone image representing the bones of the subject, or (iii) a radiographic image of the subject and a bone density image representing the bone density of the subject and information related to the bone density of the subject.

[0012] Furthermore, in the estimation device based on the present invention, the information related to bone mineral density can be derived as follows:

[0013] The body thickness distribution of the subject is inferred from at least one of two radiographic images obtained by photographing a subject containing bones and soft tissues using radiation with different energy distributions.

[0014] Photographic conditions for acquiring two radiographic images; and

[0015] The pixel values ​​of the bone region in the bone image are derived by energy subtraction processing, which is performed by weighted subtraction on two radiographic images.

[0016] Furthermore, in the estimation device based on the present invention, the bone image can be derived through the following steps:

[0017] Use at least one of the two radiographic images to identify the bony and soft tissue parts of the subject;

[0018] The attenuation coefficients for the bone and soft tissues are derived using the identification results of the bone and soft tissues and two radiographic images.

[0019] Energy subtraction is performed using an attenuation factor.

[0020] Furthermore, in the estimation device based on the present invention, the bone image can be derived through the following steps:

[0021] New weighting coefficients for weighted subtraction operations are derived based on the pixel values ​​of the bones contained in the bone image.

[0022] A new bone image is derived by performing a weighted subtraction operation on two radiographic images using new weighting coefficients.

[0023] Repeat the process of deriving a new weighting coefficient from the new bone image and deriving a new bone image from the new weighting coefficient.

[0024] Furthermore, in the estimation device based on the present invention, the bone image can be derived through the following steps:

[0025] While changing the attenuation coefficient, soft tissue thickness, and bone thickness of each energy distribution for different soft tissues from the initial values, the difference between the value of soft tissue attenuation coefficient × soft tissue thickness + bone tissue attenuation coefficient × bone thickness and the pixel value of the radiographic image is derived for each energy distribution.

[0026] The attenuation coefficients of the soft and bony parts for each energy distribution with differences that are minimized or less than a predefined threshold are derived.

[0027] Energy subtraction is performed using weighted coefficients derived from the attenuation coefficients of soft and bony parts.

[0028] Furthermore, in the estimation device based on the present invention, the bone image can be derived through the following steps:

[0029] Derive the compositional proportions of the multiple components contained in the soft parts of the subject;

[0030] Based on the composition ratio, the attenuation coefficient of each different energy distribution in the soft part is derived for each pixel of the two radiographic images;

[0031] Energy subtraction is performed using weighting coefficients derived from the attenuation coefficients of the soft tissue and the predefined attenuation coefficients of the bone tissue.

[0032] Furthermore, in the estimation device based on the present invention, the first body thickness and the second body thickness can be respectively derived from the body thickness of the subject for each pixel of the two radiation images.

[0033] Based on the first and second body thicknesses, the composition ratio is derived for each pixel of the radiographic image.

[0034] Furthermore, in the estimation device based on the present invention, the first body thickness and the second body thickness can be derived, in terms of compositional proportions, based on the attenuation coefficients of each energy distribution for each of the multiple components.

[0035] While changing the thickness of the components and the attenuation coefficient of each component, the first body thickness and the second body thickness are derived, and the component ratio is derived based on the difference between the first body thickness and the second body thickness, which is the thickness of the component below a predetermined threshold.

[0036] Furthermore, in the estimation apparatus based on the present invention, the following can be performed: for the bone image, a scattering removal process is performed to remove the scattering component of the radiation irradiating the subject from two radiation images, and the bone image is derived by energy subtraction processing of the two radiation images after the scattering component has been removed.

[0037] Furthermore, in the estimation apparatus based on the present invention, the scattered radiation removal process can be performed through the following steps:

[0038] Obtain the radiation characteristics of an object, corresponding to its thickness distribution, between the subject and the radiation detector that detects the radiation image;

[0039] The primary ray distribution and scattered ray distribution of the radiation contained in each of the two radiation images are derived using photographic conditions, thickness distribution, and the radiation characteristics of the object.

[0040] The error between the sum of the primary ray distribution and scattered ray distribution of each of the two radiation images and the pixel values ​​at each location in the two radiation images is derived. The volume thickness distribution is updated in such a way that the error is less than a predetermined threshold. The radiation characteristics are derived from the updated volume thickness distribution, and the primary ray distribution and scattered ray distribution contained in each of the two radiation images are derived again.

[0041] The distribution of scattered lines when the error is subtracted from each of the two radiographic images and becomes less than a predefined threshold.

[0042] Furthermore, in the estimation apparatus based on the present invention, the scattered radiation removal process can be performed through the following steps:

[0043] The distribution of the first primary rays and the scattered rays transmitted through the subject were derived using two radiographic images.

[0044] The second primary ray distribution and scattered ray distribution of the transmitted radiation are derived using the first primary ray distribution and scattered ray distribution, as well as the radiation characteristics of the object between the subject and the radiation detector that detects the radiation image.

[0045] The second primary ray distribution and scattered ray distribution are used to derive the radiation image after the transmission of the subject and object.

[0046] Furthermore, in the estimation apparatus based on the present invention, the scattered radiation removal process can be performed through the following steps:

[0047] Region detection images are derived by detecting the subject region where radiation penetrates the subject and reaches the radiation detector, and the direct radiation region where radiation does not penetrate the subject and directly reaches the radiation detector, in two radiation images.

[0048] Based on the region detection image and the scattering diffusion information related to the scattering ray diffusion, a scattering ray image related to the scattering ray component is derived.

[0049] Subtract the scattered radiation image from the two radiation images.

[0050] Furthermore, in the estimation device based on the present invention, the following can be performed: Regarding the bone image, derive the processing content of the first granular suppression processing for the first radiographic image with a high S / N ratio among the two radiographic images.

[0051] Based on the processing content of the first granular suppression processing, the processing content of the second granular suppression processing for the second radiographic image with low S / N is derived.

[0052] The first radiation image is subjected to grain suppression processing based on the processing content of the first grain suppression process.

[0053] The second radiation image is subjected to granular suppression processing based on the processing content of the second granular suppression process.

[0054] The bone image was derived using two radiographic images that had undergone granulation suppression.

[0055] In this case, the processing content of the first granular suppression process can be derived from the physical quantity mapping map of the subject derived from at least one of the first and second radiation images.

[0056] Furthermore, in the estimation device based on the present invention, the information related to bone density may include at least one of bone density, an assessment value of the subject's fracture risk, and information indicating the healing status of the bone after treatment.

[0057] The estimation method based on this invention uses a fully trained neural network to derive estimation results related to bone density from simple radiographic images. This fully trained neural network derives estimation results related to bone density from simple radiographic images obtained by simply photographing a subject containing bones.

[0058] The neural network was learned by learning from the following as teacher data: (i) two radiographic images of a subject containing bones obtained by taking pictures of the subject using radiation with different energy distributions, (ii) a radiographic image of the subject and a bone image representing the bones of the subject, or (iii) a radiographic image of the subject and a bone density image representing the bone density of the subject and information related to the bone density of the subject.

[0059] Alternatively, the presumed method based on the present invention may also be provided in the form of a program for causing a computer to execute.

[0060] Invention Effects

[0061] According to the present invention, bone mineral density can be estimated with high accuracy. Attached Figure Description

[0062] Figure 1 This is a schematic block diagram showing the structure of a radiographic imaging system for use with a presumed apparatus based on the first embodiment of the present invention.

[0063] Figure 2 This is a diagram showing the schematic structure of the estimated device based on the first embodiment.

[0064] Figure 3 This is a diagram showing the functional structure of the presumed device based on the first embodiment.

[0065] Figure 4 This is a diagram showing a schematic structure of the neural network used in this embodiment.

[0066] Figure 5 This is a graph representing teacher data.

[0067] Figure 6 This is a diagram showing a schematic structure of the information export device based on the first embodiment.

[0068] Figure 7 This is a diagram showing the functional structure of the information export device based on the first embodiment.

[0069] Figure 8 It is a diagram representing a bone structure.

[0070] Figure 9 It is a graph showing the relationship between the contrast of the bony and soft parts and the thickness of the subject.

[0071] Figure 10 This is a diagram illustrating an example of a lookup table used to obtain correction coefficients.

[0072] Figure 11 This is a diagram used to illustrate the learning process of a neural network.

[0073] Figure 12 This is a conceptual diagram of the processing performed after the neural network has been learned.

[0074] Figure 13 This is a diagram showing the estimated result.

[0075] Figure 14 This is a flowchart of the learning process performed in the first embodiment.

[0076] Figure 15 This is a flowchart of the presumption process performed in the first embodiment.

[0077] Figure 16 This is a diagram showing the functional structure of the information export device based on the second embodiment.

[0078] Figure 17 It is a graph used to illustrate the calculation of the index value representing the decay.

[0079] Figure 18 It is a graph used to illustrate the calculation of the index value representing the decay.

[0080] Figure 19 This is a diagram showing the functional structure of the information export device based on the third embodiment.

[0081] Figure 20 It is a graph of a table showing the relationship between the specified body thickness distribution and the initial weighting coefficients.

[0082] Figure 21 It is a graph showing the relationship between the pixel values ​​of a specified bone and the thickness of the bone.

[0083] Figure 22 It is a graph showing the relationship between the thickness of the soft tissue and the thickness of the bony tissue and the weighting coefficient.

[0084] Figure 23 This is a diagram showing the functional structure of the information export device based on the fourth embodiment.

[0085] Figure 24 This is a graph showing the relationship between the initial value of the specified thickness of the soft part and the initial value of the attenuation coefficient of the soft part.

[0086] Figure 25 It is a graph showing the relationship between the thickness of the soft tissue and the thickness of the bony tissue and the attenuation coefficient.

[0087] Figure 26 This is a diagram showing the functional structure of the information export device based on the fifth embodiment.

[0088] Figure 27 It is a graph used to illustrate the difference in volume thickness derived from low-energy and high-energy images.

[0089] Figure 28 This is a graph showing a table that establishes a correspondence between the body thickness difference derived from two radiographic images and the proportion of fat composition.

[0090] Figure 29 This is a diagram showing the functional structure of the scattering line removal unit in the information export device based on the sixth embodiment.

[0091] Figure 30 This is a photograph used to illustrate the sixth embodiment.

[0092] Figure 31 It is a graph used to illustrate the measurement of scattered light transmittance corresponding to the thickness of the subject.

[0093] Figure 32 It is a graph used to illustrate the measurement of scattered light transmittance corresponding to the thickness of the subject.

[0094] Figure 33 It is a table showing the relationship between the thickness distribution of the subject and the transmittance of scattered rays from an object between the subject and the radiation detector.

[0095] Figure 34 It is a graph used to illustrate the measurement of primary ray transmittance corresponding to the thickness of the subject.

[0096] Figure 35 It is a graph used to illustrate the measurement of primary ray transmittance corresponding to the thickness of the subject.

[0097] Figure 36 This is a table showing the relationship between the thickness distribution of the subject and the primary ray transmittance of an object between the subject and the radiation detector.

[0098] Figure 37 This diagram represents the state of the air layer between the top plate and the grid.

[0099] Figure 38 This is a diagram showing the functional structure of the scattering line removal unit in the information export device based on the seventh embodiment.

[0100] Figure 39 This is a diagram used to illustrate the function of the first output section in the seventh embodiment.

[0101] Figure 40 This is a diagram used to illustrate the method of estimating the thickness of a subject.

[0102] Figure 41 It is a graph representing the point spread function.

[0103] Figure 42 It is a diagram showing the path of a radiation beam.

[0104] Figure 43 This is a diagram showing the functional structure of the information export device based on the eighth embodiment.

[0105] Figure 44 This is a diagram representing the bilateral filter applied to the first radiation image.

[0106] Figure 45 It means and Figure 44The diagram shows a local region of the first radiation image corresponding to a local region of the second radiation image.

[0107] Figure 46 This is a diagram illustrating an example of a bilateral filter for the second radiation image.

[0108] Figure 47 This is a diagram showing the functional structure of the information export device based on the ninth embodiment.

[0109] Figure 48 This is a diagram showing an example of a bilateral filter for a mapping of physical quantities.

[0110] Figure 49 This is a diagram showing the functional structure of the information export device based on the 10th embodiment.

[0111] Figure 50 This is a diagram showing an example of the energy spectrum of radiation transmitted through muscle tissue and radiation transmitted through adipose tissue.

[0112] Figure 51 It is a chart showing the relationship between statistical values ​​and the incidence of fractures within 10 years.

[0113] Figure 52 This is another example of a graph representing teacher data.

[0114] Figure 53 This is a diagram showing the functional structure of the information export device based on the 11th embodiment.

[0115] Figure 54 This is an example of artificial bone implanted in the bone of the subject.

[0116] Figure 55 This is a chart illustrating the relationship between the distance from the femoral stem to the interior of the femur and the amount of bone salt at various postoperative stages.

[0117] Figure 56 This is a cross-sectional view showing an example of the cross-sectional structure of a human bone.

[0118] Figure 57 This is another example of a graph representing teacher data.

[0119] Figure 58 This is a diagram showing the functional structure of the scattering line removal unit in the information export device based on the 12th embodiment.

[0120] Figure 59 It is a diagram used to illustrate the detection of the subject area and the area of ​​direct radiation.

[0121] Figure 60 It is a graph representing a region detection image of a specific line segment.

[0122] Figure 61 It is a diagram representing the scattered lines of a specific line segment.

[0123] Figure 62 It is a graph that shows the thickness distribution of a specific line segment.

[0124] Figure 63 This is another example of a graph representing teacher data.

[0125] Figure 64 This is another example of a graph representing teacher data.

[0126] Figure 65 This is another example of a display screen showing the presumed result. Detailed Implementation

[0127] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Figure 1 This is a schematic block diagram illustrating the structure of a radiographic imaging system adapted to the presumption apparatus based on the first embodiment of the present invention. Figure 1 As shown, the radiographic imaging system based on the first embodiment includes an imaging device 1, an image storage system 9, an estimation device 10 based on the first embodiment, and an information export device 50. The imaging device 1, the estimation device 10, and the information export device 50 are connected to the image storage system 9 via a network (not shown).

[0128] The photographic apparatus 1 is a photographic apparatus capable of performing energy subtraction by altering the energy of radiation, such as X-rays emitted from a radiation source 3 and transmitted through the subject H, and then irradiating the first radiation detector 5 and the second radiation detector 6 using a so-called one-shot method. During the photographing process, such as... Figure 1 As shown, a first radiation detector 5, a radiation energy conversion filter 7 made of copper plates or the like, and a second radiation detector 6 are arranged sequentially from the side closest to the radiation source 3 to drive the radiation source 3. Furthermore, the first radiation detector 5 and the second radiation detector 6 are in close contact with the radiation energy conversion filter 7.

[0129] Therefore, in the first radiation detector 5, a first radiation image G1 of the subject H, formed by low-energy radiation including so-called soft filaments, is acquired. And in the second radiation detector 6, a second radiation image G2 of the subject H, formed by high-energy radiation from which the soft filaments have been removed, is acquired. Thus, the first radiation image G1 and the second radiation image G2 are acquired by photographing the subject H using radiation with different energy distributions. The first radiation image G1 and the second radiation image G2 are input into the estimation device 10. Both the first radiation image G1 and the second radiation image G2 are frontal images including the periphery of the crotch area of ​​the subject H.

[0130] The first radiation detector 5 and the second radiation detector 6 are radiation detectors capable of repeatedly recording and reading out radiation images. They can be either direct-type radiation detectors that generate charge by directly receiving radiation, or indirect-type radiation detectors that temporarily convert radiation into visible light and then convert that visible light into a charge signal. Furthermore, as the method for reading out the radiation image signal, a TFT readout method, which reads out the radiation image signal by turning a TFT (thin film transistor) switch on or off, or an optical readout method, which reads out the radiation image signal by irradiating a readout light, is preferred. However, they are not limited to these methods, and other methods may also be used.

[0131] Furthermore, the imaging device 1 can also acquire a simple radiographic image G0, which is a simple two-dimensional image of the subject H, by simply taking a picture of the subject H using only the first radiation detector 5. To distinguish it from simple imaging, the acquisition of the first radiographic image G1 and the second radiographic image G2 described above is referred to as energy subtraction imaging. In this embodiment, the first radiographic image G1 and the second radiographic image G2 acquired by energy subtraction imaging are used as learning data as described later. Furthermore, as described later, the simple radiographic image G0 acquired by simple imaging is used to derive an inference result related to bone density.

[0132] Image storage system 9 is a system for storing image data of radiographic images acquired by photographic device 1. Image storage system 9 extracts images from the stored radiographic images that meet the request from estimation device 10 and sends them to the requesting source device. A specific example of image storage system 9 is PACS (Picture Archiving and Communication Systems). Furthermore, in this embodiment, image storage system 9 stores a large amount of teacher data for learning the neural network described later.

[0133] Next, the estimation device according to the first embodiment will be described. First, refer to... Figure 2 The hardware structure of the estimation device according to the first embodiment will be described. For example... Figure 2As shown, the radiation image processing apparatus 10 is a computer such as a workstation, server computer, or personal computer, and includes a CPU (Central Processing Unit) 11, non-volatile memory 13, and internal memory 16 as a temporary storage area. Furthermore, the estimation device 10 includes a display 14 such as a liquid crystal display, input devices 15 such as a keyboard and mouse, and a network I / F (Interface) 17 connected to a network not shown. The CPU 11, memory 13, display 14, input devices 15, internal memory 16, and network I / F 17 are connected to a bus 18. The CPU 11 is an example of a processor in this invention.

[0134] The memory 13 is implemented using HDD (Hard Disk Drive), SSD (Solid State Drive), or flash memory. The memory 13, serving as the storage medium, stores the estimation program 12A and the learning program 12B installed in the estimation device 10. The CPU 11 reads the estimation program 12A and the learning program 12B from the memory 13, expands them into the internal memory 16, and executes the expanded estimation program 12A and the learning program 12B.

[0135] Furthermore, the estimation program 12A and the learning program 12B are stored in a state accessible from the outside in the storage device or network storage of a server computer connected to a network, and are downloaded and installed into the computer constituting the estimation device 10 upon request. Alternatively, they are recorded on recording media such as DVD (Digital Versatile Disc) or CD-ROM (Compact Disc Read Only Memory) and distributed, and installed into the computer constituting the estimation device 10 from such recording media.

[0136] Next, the functional structure of the estimation device based on the first embodiment will be described. Figure 3 This is a diagram illustrating the functional structure of the presumed device based on the first embodiment. For example... Figure 3 As shown, the estimation device 10 includes an image acquisition unit 21, an information acquisition unit 22, an estimation unit 23, a learning unit 24, and a display control unit 25. Furthermore, the CPU 11 functions as the image acquisition unit 21, the information acquisition unit 22, the estimation unit 23, and the display control unit 25 by executing the estimation program 12A. Additionally, the CPU 11 functions as the learning unit 24 by executing the learning program 12B.

[0137] The image acquisition unit 21 acquires, for example, a frontal image of the subject H near its crotch, namely, a first radiation image G1 and a second radiation image G2, from the first radiation detector 5 and the second radiation detector 6 by causing the imaging device 1 to perform energy subtraction imaging of the subject H. When acquiring the first radiation image G1 and the second radiation image G2, imaging conditions such as imaging dose, line quality, tube voltage, distance between the radiation source 3 and the surfaces of the first radiation detector 5 and the second radiation detector 6 (SID, Source Image receptor Distance), distance between the radiation source 3 and the surface of the subject H (SOD, Source Object Distance), and whether or not there is a scattered radiation removal grid are set.

[0138] SOD and SID, as described later, are used to calculate volume thickness distribution. For SOD, it is preferable to obtain it using a TOF (Time of Flight) camera, for example. For SID, it is preferable to obtain it using a potentiometer, ultrasonic rangefinder, or laser rangefinder, for example.

[0139] The photographic conditions can be set by the operator through input device 15. The set photographic conditions are stored in memory 13. The first radiation image G1 and the second radiation image G2 acquired by energy subtraction imaging, as well as the photographic conditions, are also sent to and saved in image storage system 9. The first radiation image G1 and the second radiation image G2 are used to export the teacher data described later.

[0140] Furthermore, the image acquisition unit 21 acquires a frontal image of the subject H near its crotch, i.e., a simple radiation image G0, by having the photography device 1 simply take a picture of the subject H using only the first radiation detector 5.

[0141] Alternatively, in this embodiment, a program independent of the estimation program 12A can be used to acquire the first radiation image G1, the second radiation image G2, and the simple radiation image G0, and store them in the memory 13. In this case, the image acquisition unit 21 acquires these radiation images by reading the first radiation image G1, the second radiation image G2, and the simple radiation image G0 stored in the memory 13 for processing.

[0142] The information acquisition unit 22 acquires teacher data from the image storage system 9 via the network I / F17 for learning the neural network described later.

[0143] The estimation unit 23 derives an estimation result related to the bone density of the bone portion contained in the subject H based on the simple radiograph G0. In this embodiment, as the estimation result related to bone density, an estimation result of the bone density of the object bone in the bone region contained in the simple radiograph G0 is derived. Therefore, the estimation unit 23 uses the learned neural network 23A that outputs bone density when the simple radiograph G0 is input to derive the estimation result related to bone density.

[0144] The learning unit 24 constructs a learned neural network 23A by performing machine learning on the neural network using teacher data. Examples of neural networks include simple perceptrons, multilayer perceptrons, deep neural networks, convolutional neural networks, deep belief networks, recurrent neural networks, and probabilistic neural networks. In this embodiment, a convolutional neural network is used as the neural network.

[0145] Figure 4 This is a diagram representing the neural network used in this embodiment. For example... Figure 4 As shown, 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, multiple convolutional layers 35, multiple pooling layers 36, and a fully connected layer 37. In the neural network 30, the fully connected layer 37 is located before the output layer 33. Furthermore, in the neural network 30, the convolutional layers 35 and pooling layers 36 are alternately arranged between the input layer 31 and the fully connected layer 37.

[0146] Furthermore, the structure of neural network 30 is not limited to Figure 4 For example, neural network 30 could also have a convolutional layer 35 and a pooling layer 36 between the input layer 31 and the fully connected layer 37.

[0147] Figure 5 This is a diagram illustrating an example of teacher data used in the learning process of a neural network. For example... Figure 5 As shown, the teacher data 40 consists of learning data 41 and correct answer data 42. In this embodiment, the data input into the learned neural network 23A to obtain the estimated bone density is a simple radiographic image G0, but the learning data 41 includes two radiographic images: a first radiographic image G1 and a second radiographic image G2 obtained through the above-described energy subtraction imaging.

[0148] Correct data 42 refers to the bone mineral density (BMD) of the subject bone (i.e., femur) from which learning data 41 was acquired. Furthermore, in this embodiment, the BMD per unit area is estimated based on the two-dimensional simple radiographic image G0; therefore, the unit of BMD is (g / cm³). 2The bone density, which serves as the correct solution data 42, is exported by the information exporting device 50. In this embodiment, for example, the bone density that becomes the correct solution data 42 is exported using the method described in Japanese Patent Application Publication No. 2019-202035. Furthermore, the bone density that becomes the correct solution data 42 is an example of information related to the bone density of the subject's bones. The information exporting device 50 will be described below.

[0149] Figure 6 This is a schematic block diagram showing the structure of the information export device based on the first embodiment. For example... Figure 6 As shown, the information export device 50 based on the first embodiment is a computer such as a workstation, server computer, or personal computer, including a CPU 51, non-volatile memory 53, and internal memory 56 as a temporary storage area. Furthermore, the information export device 50 includes a display 54 such as a liquid crystal display, an input device 55 consisting of a keyboard and mouse, and a network I / F 57 connected to a network (not shown). The CPU 51, memory 53, display 54, input device 55, internal memory 56, and network I / F 57 are connected to a bus 58.

[0150] Similar to memory 13, memory 53 is implemented using HDD, SSD, flash memory, etc. Memory 53, as a storage medium, stores an information export program 52. CPU 51 reads the information export program 52 from memory 53, expands it into internal memory 56, and executes the expanded information export program 52.

[0151] Next, the functional structure of the information export device based on the first embodiment will be described. Figure 7 This is a diagram illustrating the functional structure of the information export device based on the first embodiment. For example... Figure 7 As shown, the information export device 50 based on the first embodiment includes an image acquisition unit 61, a scattering ray removal unit 62, a subtraction unit 63, and a bone density export unit 64. Furthermore, the CPU 51 executes the information export program 52, and the CPU 51 functions as the image acquisition unit 61, the scattering ray removal unit 62, the subtraction unit 63, and the bone density export unit 64.

[0152] The image acquisition unit 61 acquires the first radiation image G1 and the second radiation image G2, which are stored in the image storage system 9 as learning data 41. Furthermore, similar to the image acquisition unit 21 of the estimation device 10, the image acquisition unit 61 can acquire the first radiation image G1 and the second radiation image G2 by having the photographic device 1 capture an image of the subject H.

[0153] Furthermore, the image acquisition unit 61 also acquires the photographic conditions stored in the image storage system 9 for acquiring the first and second radiation images. The photographic conditions include the photographic dose, tube voltage, SID, SOD, and the presence or absence of a scattered radiation removal grid when acquiring the first radiation image G1 and the second radiation image G2.

[0154] Here, each of the first radiation image G1 and the second radiation image G2 contains, in addition to the primary ray component of the radiation transmitted through the subject H, a scattered ray component based on the radiation scattered within the subject H. Therefore, the scattered ray removal unit 62 removes the scattered ray component from the first radiation image G1 and the second radiation image G2. For example, the scattered ray removal unit 62 can use the method described in Japanese Patent Application Publication No. 2015-043959 to remove the scattered ray component from the first radiation image G1 and the second radiation image G2. When using the method described in Japanese Patent Application Publication No. 2015-043959, the volume thickness distribution of the subject H and the scattered ray component for removing the scattered ray component are derived simultaneously.

[0155] The following describes the removal of scattered ray components from the first radiation image G1, but the removal of scattered ray components from the second radiation image G2 can also be performed in the same way. First, the scattered ray removal unit 62 acquires a virtual model of the subject H having an initial body thickness distribution T0(x, y). The virtual model is data that virtually represents the subject H, establishing a correspondence between the body thickness of the initial body thickness distribution T0(x, y) and the coordinate positions of each pixel in the first radiation image G1. Furthermore, the virtual model of the subject H having the initial body thickness distribution T0(x, y) can be pre-stored in the memory 53 of the information export device 50. The scattered ray removal unit 62 can calculate the body thickness distribution T(x, y) of the subject H based on the SID and SOD included in the photographic conditions. In this case, the initial body thickness distribution T0(x, y) can be obtained by subtracting SOD from SID.

[0156] Next, the scattering removal unit 62 generates an image based on the virtual model as a presumed image of the first radiation image G1 obtained by photographing the subject H. This image is a composite of a presumed primary ray image obtained by photographing the primary ray image obtained by photographing the virtual model and a presumed scattering image obtained by photographing the scattering image obtained by photographing the virtual model.

[0157] Next, the scattering removal unit 62 corrects the initial thickness distribution TO(x, y) of the virtual model to minimize the difference between the estimated image and the first radiation image G1. The scattering removal unit 62 repeats the generation of the estimated image and the correction of the thickness distribution until the difference between the estimated image and the first radiation image G1 meets a predetermined termination condition. The scattering removal unit 62 derives the thickness distribution that meets the termination condition as the thickness distribution T(x, y) of the subject H. Furthermore, the scattering removal unit 62 removes the scattering components contained in the first radiation image G1 by subtracting the scattering components that meet the termination condition from the first radiation image G1.

[0158] The subtraction unit 63 extracts the bone image Gb of the subject H from the first radiation image G1 and the second radiation image G2 by performing energy subtraction processing. Furthermore, the first radiation image G1 and the second radiation image G2 in subsequent processing are radiation images with scattered radiation components removed. When extracting the bone image Gb, the subtraction unit 63 performs weighted subtraction operations between corresponding pixels on the first radiation image G1 and the second radiation image G2 as shown in equation (1) below. Figure 8 As shown, bone images Gb are generated from the bone regions of the subject H contained in each of the radiographic images G1 and G2. In equation (1), α is a weighting coefficient. In addition, the pixel value of each pixel in the bone region of the bone image Gb is called the bone pixel value.

[0159] Gb(x,y)=α·G2(x,y)-G1(x,y) (1)

[0160] The bone density extraction unit 64 extracts bone density for each pixel of the bone image Gb. In this embodiment, the bone density extraction unit 64 extracts bone density B by converting the pixel values ​​of the bone image Gb into the pixel values ​​of the bone image acquired using reference radiographic conditions. Specifically, the bone density extraction unit 64 extracts bone density by correcting the pixel values ​​of the bone image Gb using correction coefficients obtained from a lookup table described later.

[0161] Here, the higher the tube voltage in radiation source 3 and the higher the energy of the radiation emitted from radiation source 3, the smaller the contrast between soft and bony areas in the radiation image. Furthermore, during the transmission of radiation through the subject H, beam hardening occurs, where the low-energy components of the radiation are absorbed by the subject H, resulting in high-energy radiation. The greater the thickness of the subject H, the greater the high-energy radiation caused by beam hardening.

[0162] Figure 9 This is a graph showing the relationship between the contrast of the bony and soft parts and the thickness of the subject H. Additionally, in Figure 9The diagram shows the relationship between the contrast of the skeletal and soft parts relative to the body thickness H of the subject at three tube voltages: 80kV, 90kV, and 100kV. Figure 9 As shown, the higher the tube voltage, the lower the contrast. Furthermore, if the thickness of the subject H exceeds a certain value, the greater the thickness, the lower the contrast. Additionally, the larger the pixel value of the bone region in the bone image Gb, the greater the contrast between the bone and soft tissue. Therefore, the larger the pixel value of the bone region in the bone image Gb, the higher the contrast between the bone and soft tissue. Figure 9 The relationship shown shifts towards the higher contrast side.

[0163] In the first embodiment, the difference in contrast in the bone image Gb corresponding to the tube voltage at the time of imaging, and a lookup table for obtaining correction coefficients to correct the decrease in contrast caused by beam hardening, are stored in the memory 53 of the information export device 50. The correction coefficients are coefficients used to correct the pixel values ​​of the bone image Gb.

[0164] Figure 10 This is a diagram illustrating an example of a lookup table used to obtain correction coefficients. Figure 10 The example shown is a lookup table (hereinafter referred to as table) LUT1 that sets the reference imaging conditions to a tube voltage of 90kV. Figure 10 As shown in Table LUT1, the larger the tube voltage and the greater the thickness of the subject H, the larger the correction coefficient is set. Figure 10 In the example shown, the reference imaging condition is a tube voltage of 90kV. Therefore, when the tube voltage is 90kV and the volume thickness is 0, the correction factor becomes 1. Additionally, in Figure 10 In this table, LUT1 is shown in two dimensions, but the correction coefficients vary depending on the pixel values ​​of the bone region. Therefore, LUT1 is actually a three-dimensional table with axes representing the pixel values ​​of the bone region added.

[0165] The bone density extraction unit 64 extracts the body thickness distribution T(x, y) of the subject H from the table LUT1 and the correction coefficient CO(x, y) of each pixel corresponding to the imaging conditions including the tube voltage setting stored in the memory 13. Then, as shown in the following equation (2), the bone density extraction unit 64 extracts the bone density B(x, y) (g / cm³) of each pixel of the bone region in the bone image Gb by multiplying each pixel (x, y) of the bone region in the bone image Gb by the correction coefficient CO(x, y). 2 The bone mineral density B(x, y) thus derived represents the amount of bone salts in the bone region contained in a radiographic image obtained by photographing the subject H using a tube voltage of 90 kV as the reference imaging condition, and in which the effects of beam hardening are removed. Therefore, the bone mineral density derivation unit 64 derives a bone mineral density image with the derived bone mineral density as the pixel value of each pixel.

[0166] B(x,y)=CO(x,y)×Gb(x,y) (2)

[0167] Furthermore, in this embodiment, the bone density extraction unit 64 extracts representative values ​​of bone density B only for the target bone. For example, when the target bone is the femur, the bone density extraction unit 64 extracts representative values ​​of bone density B for the femur region by extracting representative values ​​of bone density B for each pixel in the femur region of the bone image Gb. As representative values, average, median, minimum, and maximum values ​​can be used. In this embodiment, the representative value of the bone density of the femur, which is the target bone, is used as the forward solving data 42.

[0168] Bone density data used as forward solving data 42 is exported at the same time as the acquisition of learning data 41 and sent to image storage system 9. In image storage system 9, the learning data 41 and forward solving data 42 are established as a correspondence and stored as teacher data 40. Furthermore, to improve the robustness of learning, additional teacher data 40 can be created and stored, which includes images of the same image that have undergone at least one of the following processing methods: scaling, contrast alteration, movement, in-plane rotation, inversion, and noise addition, as learning data 41.

[0169] Returning to the estimation device 10, the learning unit 24 uses a large amount of teacher data 40 to learn the neural network. Figure 11 This diagram illustrates the learning process of the neural network 30. During the learning process of the neural network 30, the learning unit 24 inputs learning data 41, namely the first radiographic image G1 and the second radiographic image G2, into the input layer 31 of the neural network 30. Furthermore, the learning unit 24 outputs the bone density of the target bone from the output layer 33 of the neural network 30 as output data 47. The learning unit 24 also derives the difference between the output data 47 and the correct solution data 42 as the loss L0.

[0170] The learning unit 24 learns the neural network 30 based on the loss L0. Specifically, the learning unit 24 adjusts the kernel coefficients in the convolutional layer 35, the connection weights between layers, and the connection weights in the fully connected layer 37 (hereinafter referred to as parameters 48) in a way that reduces the loss L0. For example, backpropagation can be used to adjust parameters 48. The learning unit 24 repeats the adjustment of parameters 48 until the loss L0 falls below a predetermined threshold. Thus, the learned neural network 23A is constructed by adjusting parameters 48 to output the bone density of the target bone when the input is a simple radiographic image G0. The constructed learned neural network 23A is stored in the memory 13.

[0171] Figure 12 This is a conceptual diagram of the processing performed after learning the neural network 23A. (Example) Figure 12 As shown, if a patient's simple radiographic image G0 is input into the learned neural network 23A constructed in the manner described above, the learned neural network 23A outputs the bone density of the object bone (i.e., femur) contained in the input simple radiographic image G0.

[0172] The display control unit 25 displays the estimated bone density result estimated by the estimation unit 23 on the display 14. Figure 13 This is a diagram representing the display screen of the estimated result. For example... Figure 13 As shown, the display screen 70 has an image display area 71 and a bone density display area 72. The image display area 71 displays a simple radiographic image G0 of the subject H. The bone density display area 72 displays a representative value of the bone density around the joint of the femur, as estimated by the estimation unit 23.

[0173] Next, the processing performed in the first embodiment will be described. Figure 14 This is a flowchart illustrating the learning process performed in the first embodiment. First, the information acquisition unit 22 acquires teacher data 40 from the image storage system 9 (step ST1). The learning unit 24 inputs the learning data 41 contained in the teacher data 40 into the neural network 30 and outputs bone density. It then learns the neural network 30 using a loss L0 based on the difference from the correct answer data 42 (step ST2) and returns to step ST1. Next, the learning unit 24 repeats steps ST1 and ST2 until the loss L0 reaches a predetermined threshold, and then ends the learning process. Alternatively, the learning unit 24 can end the learning process by repeating the learning process a predetermined number of times. Thus, the learning unit 24 constructs the learned neural network 23A.

[0174] Next, the estimation process in the first embodiment will be explained. Figure 15 This is a flowchart illustrating the estimation process in the first embodiment. Additionally, a simple radiographic image G0 is acquired by photography and stored in memory 13. If an instruction to start processing is input from input device 15, image acquisition unit 21 acquires the simple radiographic image G0 from memory 13 (step ST11). Next, estimation unit 23 derives an estimation result related to bone mineral density based on the simple radiographic image G0 (step ST12). Then, display control unit 25 displays the estimation result related to bone mineral density derived by estimation unit 23 along with the simple radiographic image G0 on display 14 (step ST13), and ends the processing.

[0175] Thus, in this embodiment, a learned neural network 23A, constructed by learning from the first radiographic image G1 and the second radiographic image G2 as teacher data, is used to derive an inference related to the bone density of the subject H contained in the simple radiographic image G0. Here, in this embodiment, both radiographic images, the first radiographic image G1 and the second radiographic image G2, are used in the learning of the neural network. Therefore, compared to using a single radiographic image and bone density-related information as teacher data, the learned neural network 23A can derive an inference related to bone density with greater accuracy from the simple radiographic image G0. Therefore, according to this embodiment, an inference related to bone density can be derived with greater accuracy.

[0176] Furthermore, in the first embodiment described above, the bone mineral density (BMD) as the positive solution data 42 can also be a value measured by DXA or ultrasound. In an apparatus that uses ultrasound to measure BMD, for example, ultrasound is applied to the lumbar region and the BMD of that region is measured, or ultrasound is applied to the heel and the BMD of that heel is measured.

[0177] Furthermore, in the first embodiment described above, when deriving the bone density of the correct solution data 42, the bone image Gb can also be derived through the following steps: using at least one of the first radiographic image G1 and the second radiographic image G2 to identify the bone and soft tissue of the subject H, using the identification results of the bone and soft tissue and the first radiographic image G1 and the second radiographic image G2 to derive the attenuation coefficients of the radiographs of the bone and soft tissue, and using the derived attenuation coefficients to perform energy subtraction processing. Hereinafter, this will be described as the second embodiment. In addition, the energy subtraction processing in the second embodiment is described, for example, in International Publication No. 2020 / 175319.

[0178] Figure 16 This is a diagram illustrating the functional structure of the information export device based on the second embodiment. Additionally, in Figure 16 In the middle, to and Figure 7 The same structure is assigned the same reference symbols, and detailed descriptions are omitted. For example... Figure 16 As shown, the information export device 50A based on the second embodiment further includes a structure recognition unit 65 and a weight coefficient export unit 66 compared to the information export device 50 based on the first embodiment.

[0179] The structure recognition unit 65 uses at least one of the first radiation image G1 and the second radiation image G2 to identify structures contained in the subject H. In the second embodiment, the structures to be identified are bone and soft tissue. In the second embodiment, the structure recognition unit 65 uses both the first radiation image G1 and the second radiation image G2 acquired by the image acquisition unit 61 for recognition processing, but it can also use either the first radiation image G1 or the second radiation image G2 for recognition processing. Furthermore, the structure recognition unit 65 can use the first radiation image G1 and the second radiation image G2 after scattering removal processing to identify structures, or it can use the first radiation image G1 and the second radiation image G2 before scattering removal processing to identify structures.

[0180] The structure recognition unit 65 identifies the position, size, and / or shape of structures contained in the subject H photographed in the first radiographic image G1 and the second radiographic image G2. That is, the recognition process performed by the structure recognition unit 65 is the process of determining the position, size, and / or shape of structures having boundaries with other tissues, etc., within the subject H photographed in the radiographic images. In the second embodiment, the bone and soft tissue of the subject H are identified as structures.

[0181] The weighting coefficient derivation unit 66 uses the recognition results from the structure recognition unit 65 and the first radiation image G1 and the second radiation image G2 to derive an index value representing the attenuation of radiation for the structures identified by the structure recognition unit 65. This index value is used as the weighting coefficient α during subtraction processing. The attenuation coefficient is a so-called radiation attenuation coefficient, which represents the degree (proportion) of radiation attenuation due to absorption or scattering. The attenuation coefficient varies depending on the specific composition (density, etc.) and thickness (mass) of the structure through which the radiation is transmitted.

[0182] The weighting coefficient derivation unit 66 uses the ratio or difference of the pixel values ​​at corresponding pixels in the first radiation image G1 and the second radiation image G2 to calculate an index value representing attenuation. Figure 17 and Figure 18 This is a graph used to illustrate the calculation of the index values ​​representing decay. Additionally, in Figure 17 and Figure 18 The calculation of the attenuation index values ​​for the three structures is explained. For example... Figure 17 As shown, in the first radiation image G1, there are structures composed of three elements: "Ca", "Cb", and "Cc". Assume that their pixel values ​​in the first radiation image G1 are all "V1". On the other hand, the corresponding pixels in the second radiation image G2 have pixel values ​​of "Va", "Vb", and "Vc", respectively. The degree of reduction in these pixel values ​​corresponds to the degree of radiation attenuation caused by each structure (component). Therefore, as... Figure 18As shown, the weighting coefficient derivation unit 66 can calculate the index value μa representing the attenuation of the structure constituting "Ca", the index value μb representing the attenuation of the structure constituting "Cb", and the index value μc representing the attenuation of the structure constituting "Cc" using the ratio or difference between the pixel values ​​of the first radiographic image G1 and the corresponding pixel values ​​of the second radiographic image G2. In the second embodiment, the weighting coefficient derivation unit 66 calculates the index values ​​representing the attenuation of the bone and the attenuation of the soft tissue by setting the components as bone and soft tissue.

[0183] Furthermore, if the ratio or difference between the pixel values ​​in the first radiation image G1 and the corresponding pixel values ​​in the second radiation image G2 is known, then the index value μ representing attenuation can be calculated. Therefore, for the sake of explanation, it is assumed that the pixel values ​​of each component "Ca", "Cb", and "Cc" are the same "V1" in the first radiation image G1, but the pixel values ​​of each component "Ca", "Cb", and "Cc" need not be the same in the first radiation image G1.

[0184] In this embodiment, the bone image Gb is derived in the subtraction unit 63. Therefore, the weighting coefficient derivation unit 66 derives the ratio Gs1(x,y) / Gs2(x,y) of the pixel value Gs1(x,y) of the first radiation image G1 (a low-energy image) and the pixel value Gs2(x,y) of the second radiation image G2 (a high-energy image) for the soft region in the first radiation image G1 and the second radiation image G2, as an index value representing attenuation, i.e., the weighting coefficient α in the above formula (1). In addition, the ratio Gs1(x,y) / Gs2(x,y) represents the ratio μls / μhs of the attenuation coefficient μls for low-energy radiation relative to the attenuation coefficient μhs for high-energy radiation regarding the soft region.

[0185] In the second embodiment, the subtraction unit 63 performs energy subtraction processing using the weighting coefficient α derived by the weighting coefficient derivation unit 66 to derive the bone image Gb. The weighting coefficient α in equation (1) is derived based on the attenuation coefficients of the bone and the soft tissue derived by the weighting coefficient derivation unit 66.

[0186] Then, in the second embodiment, the bone density exporting unit 64 uses the bone image Gb exported by the subtraction unit 63 to export the bone density B and a representative value of the bone density B in the target bone.

[0187] Next, the third embodiment of the present invention will be described. Figure 19 This is a diagram illustrating the functional structure of the information export device based on the third embodiment. Additionally, in Figure 19 In the middle, to and Figure 7The same structure is assigned the same reference symbols, and detailed descriptions are omitted. In the third embodiment, when deriving the bone density as the positive solution data 42, the bone image Gb is derived by the following steps: deriving new weighting coefficients for weighted subtraction operations based on the pixel values ​​of the bone contained in the bone image Gb; deriving a new bone image by performing weighted subtraction operations on the first radiographic image G1 and the second radiographic image G2 using the new weighting coefficients; and repeating the process of deriving another new weighting coefficient based on the new bone image and deriving another new bone image based on the new weighting coefficient.

[0188] Therefore, as Figure 19 As shown, the information export device 50B based on the third embodiment further includes an initial weight coefficient setting unit 67 and a weight coefficient export unit 68 compared to the information export device 50 based on the first embodiment.

[0189] The initial weight coefficient setting unit 67 sets the initial value of the weight coefficient when the subtraction unit 63 performs subtraction processing based on the body thickness distribution when the scattering removal unit 62 meets the termination condition. Here, in the third embodiment, the subtraction unit 63 uses the initial weight coefficient set by the initial weight coefficient setting unit 67 and the weight coefficient α derived by the weight coefficient derivation unit 68 to derive the bone image Gb using the above formula (1).

[0190] The initial weighting coefficient setting unit 67 sets the initial value of the weighting coefficient, i.e., the initial weighting coefficient α0, based on the volume thickness distribution when the scattering removal unit 62 meets the termination condition. In the third embodiment, as... Figure 20 As shown, a table LUT2 specifying the relationship between the body thickness distribution and the initial weighting coefficient α0 is stored in memory 53. The initial weighting coefficient setting unit 67 refers to table LUT2 and sets the initial weighting coefficient α0 according to the body thickness distribution.

[0191] Here, the degree of beam hardening depends on the thickness ts of the soft part and the thickness tb of the bone part within the subject H. Therefore, the attenuation coefficient μs of the soft part and the attenuation coefficient μb of the bone part can be defined as μs(ts, tb) and μb(ts, tb) as functions of ts and tb, respectively.

[0192] In energy subtraction processing, there are two images with different energy distributions. Therefore, the attenuation coefficient of the soft portion of the low-energy image (the first radiographic image G1 in this embodiment) can be expressed as μls(ts, tb), and the attenuation coefficient of the bone portion can be expressed as μlb(ts, tb). Similarly, the attenuation coefficient of the soft portion of the high-energy image (the second radiographic image G2 in this embodiment) can be expressed as μhs(ts, tb), and the attenuation coefficient of the bone portion can be expressed as μhb(ts, tb).

[0193] To derive the bone image Gb, it is necessary to eliminate the contrast of the soft tissue contained in the radiographic image. Therefore, the weighting coefficient α can be obtained by using the ratio of the attenuation coefficients of the soft tissue, using d = μls(ts, tb) / μhs(ts, tb). That is, the weighting coefficient d can be expressed as a function of the thickness ts of the soft tissue and the thickness tb of the bone.

[0194] In the third embodiment, the subtraction unit 63 first performs a weighted subtraction operation on the first radiation image G1 and the second radiation image G2 between corresponding pixels using the initial weight coefficient α0 set by the initial weight coefficient setting unit 67. Then, as described later, the subtraction operation is performed using the weight coefficient αnew derived by the weight coefficient derivation unit 68.

[0195] The weighting coefficient derivation unit 68 derives a new weighting coefficient αnew based on the pixel values ​​Gb(x, y) of the bone contained in the bone image Gb. Here, the pixel values ​​Gb(x, y) of the bone correspond to the thickness of the bone in the subject H. Therefore, in the third embodiment, a reference radiometric image of a reference object is obtained by pre-capturing a reference object simulating bone with various thicknesses. Then, using the relationship between the pixel values ​​of the reference object's region in the reference radiometric image and the thickness of the reference object, a table specifying the relationship between the pixel values ​​and thickness of the bone is pre-derived and stored in the memory 53. Figure 21 It is a graph showing the relationship between the pixel values ​​of a specified bone and the thickness of the bone. Figure 21 The LUT3 table shown indicates that the lower the pixel value Gb(x,y) of the bone (i.e., the higher the brightness), the thicker the bone.

[0196] The weighting coefficient derivation unit 68 refers to table LUT3 and derives the bone thickness tb at each pixel of the bone image Gb based on the pixel values ​​Gb(x, y) of the bone image Gb. Furthermore, regions in the bone image Gb that do not contain bone are composed only of soft tissue, therefore the bone thickness tb is 0. On the other hand, the weighting coefficient derivation unit 68 derives the soft tissue thickness ts in the bone image Gb by subtracting the bone thickness tb from the thickness distribution when the scattering line removal unit 62 satisfies the termination condition at pixels where the bone thickness tb is not 0.

[0197] As described above, the weighting coefficient α can be represented as a function of the thickness ts of the soft part and the thickness tb of the bone part. In the third embodiment, a table specifying the relationship between the thickness ts of the soft part and the thickness tb of the bone part and the weighting coefficient α is stored in the memory 53. Figure 22 This is a graph showing the relationship between the thickness ts of the soft tissue and the thickness tb of the bony tissue, and the weighting coefficient α. (See table below.) Figure 22As shown, Table LUT4 three-dimensionally represents the relationship between the thickness ts of the soft tissue and the thickness tb of the bone tissue and the weighting coefficient α. In Table LUT4, the larger the thickness ts of the soft tissue and the thickness tb of the bone tissue, the smaller the weighting coefficient α becomes.

[0198] Furthermore, in the third embodiment, multiple tables LUT4 are prepared based on the energy distribution of the radiation used during imaging, and these tables are stored in the memory of the information export device 50B. The weighting coefficient export unit 68 acquires information on the energy distribution of the radiation used during imaging based on the imaging conditions, and reads the table LUT4 corresponding to the acquired energy distribution information from the memory, using it to export weighting coefficients. Then, the weighting coefficient export unit 68 exports new weighting coefficients αnew based on the exported bone thickness tb and soft tissue thickness ts, referring to the table LUT4.

[0199] The subtraction section 63 uses the new weight coefficient αnew derived by the weight coefficient derivation section 68 to derive a new bone image Gbnew using the above formula (1).

[0200] Alternatively, the new bone image Gbnew can be used as the final bone image Gb, but in the third embodiment, the export of the weighting coefficient α and the subtraction process are repeated.

[0201] That is, the weight coefficient derivation unit 68 derives a new bone thickness tbnew based on the pixel values ​​of the bone in the new bone image Gbnew, referring to table LUT3. Then, the weight coefficient derivation unit 68 derives the difference Δtb between the new bone thickness tbnew and the bone thickness tb obtained in the previous processing, and determines whether the difference Δtb becomes less than a predetermined threshold. When the difference Δtb is above the threshold, the weight coefficient derivation unit 68 derives a new soft tissue thickness tsnew based on the new bone thickness tbnew, and derives another new weight coefficient αnew based on the new bone thickness tbnew and the new soft tissue thickness tsnew, referring to table LUT4.

[0202] Subtraction section 63 uses a new weighting coefficient αnew for subtraction processing, deriving a new bone image Gbnew.

[0203] Then, the weight coefficient derivation unit 68 derives another new bone thickness tbnew based on the new bone image Gbnew, and derives the difference Δtb between the new bone thickness tbnew and the bone thickness tb obtained in the previous processing.

[0204] In the third embodiment, the subtraction process and the derivation of the weight coefficient αnew are repeatedly performed by the subtraction unit 63 and the weight coefficient derivation unit 68 until the difference Δtb derived by the weight coefficient derivation unit 68 is less than a predetermined threshold.

[0205] Then, in the third embodiment, the bone density extraction unit 64 uses the bone image Gb when the difference Δtb becomes less than a threshold to extract the bone density B and a representative value of the bone density B in the target bone.

[0206] Next, the fourth embodiment of the present invention will be described. Figure 23 This is a diagram illustrating the functional structure of the information export device based on the fourth embodiment. Additionally, in Figure 23 In the middle, to and Figure 7 The same structure is assigned the same reference symbols, and detailed descriptions are omitted. In the fourth embodiment of the present invention, when deriving the bone density as the positive solution data 42, the bone image Gb is derived by the following steps: while changing the attenuation coefficient, soft tissue thickness, and bone thickness of each energy distribution for different soft tissues from initial values, the difference between the value of soft tissue attenuation coefficient × soft tissue thickness + bone tissue attenuation coefficient × bone thickness and the pixel values ​​of the radiographic image is derived for each energy distribution; the attenuation coefficients of the soft tissue and bone for each energy distribution with differences that are minimized or less than a predetermined threshold are derived; and energy subtraction is performed using weighting coefficients derived from the attenuation coefficients of the soft tissue and bone.

[0207] Therefore, as Figure 23 As shown, the information export device 50C based on the fourth embodiment further includes an initial value export unit 81, an attenuation coefficient export unit 82, and a weighting coefficient export unit 83 compared to the information export device 50 based on the first embodiment.

[0208] The initial value derivation unit 81 derives initial values ​​for the weighting coefficients used in energy subtraction processing, the attenuation coefficients for each energy distribution of the soft part, the thickness of the soft part, the attenuation coefficients for each energy distribution of the bone part, and the thickness of the bone part. Specifically, it derives initial values ​​for the attenuation coefficient μls of the soft part for low-energy radiation, the attenuation coefficient μhs of the soft part for high-energy radiation, the thickness ts of the soft part, the attenuation coefficient μlb of the bone part for low-energy radiation, the attenuation coefficient μhb of the bone part for high-energy radiation, and the thickness tb of the bone part: μls0, μhs0, ts0, μ1b0, μhb0, tb0.

[0209] In the fourth embodiment, the subtraction unit 63 performs a weighted subtraction operation between corresponding pixels of the first radiation image G1 and the second radiation image G2 using the weight coefficient derived by ...

[0210] Here, as described above, the attenuation coefficient μs of the soft tissue and the attenuation coefficient μb of the bone tissue can be defined as μs(ts, tb) and μb(ts, tb) as functions of ts and tb, respectively. Therefore, in the fourth embodiment, similarly to the third embodiment, the attenuation coefficient of the soft tissue in the low-energy image (the first radiographic image G1 in this embodiment) can also be expressed as μls(ts, tb), and the attenuation coefficient of the bone tissue can also be expressed as μlb(ts, tb). Furthermore, the attenuation coefficient of the soft tissue in the high-energy image (the second radiographic image G2 in this embodiment) can be expressed as μhs(ts, tb), and the attenuation coefficient of the bone tissue can be expressed as μhb(ts, tb).

[0211] To derive the bone image Gb, it is necessary to eliminate the contrast of soft tissues contained in the radiographic image. Therefore, the weighting coefficient α can be obtained by using the ratio of the attenuation coefficients of the soft tissues, using α = μls(ts, tb) / μhs(ts, tb). Furthermore, in the following explanation, the attenuation coefficients μls(ts, tb), μhs(ts, tb), μlb(ts, tb), and μhb(ts, tb) will be omitted (ts, tb) and simply represented as attenuation coefficients μls, μhs, μlb, and hb.

[0212] The initial value derivation unit 81 uses the thickness distribution when the scattering removal unit 62 meets the termination condition as the initial value ts0 of the soft part thickness ts. Furthermore, in the fourth embodiment, the thickness distribution used during the scattering removal process is set to the thickness distribution assuming the subject H is composed only of soft parts. Therefore, the initial value tb0 of the bone thickness tb is 0. And, as the initial values ​​μls0, μhs0, μlb0, and μhb0 of the attenuation coefficients, values ​​corresponding to the initial values ​​ts0 and tb0 of the soft and bone thicknesses ts and tb are derived. In the fourth embodiment, since the initial value tb0 of the bone thickness tb is 0, the attenuation coefficients μlb0 and μhb0 of the bone are 0. The attenuation coefficients μls0 and μhs0 of the soft part are derived as values ​​corresponding to the initial value ts0 of the soft part thickness ts. Therefore, in the fourth embodiment, a table specifying the relationship between the initial value tsO of the thickness ts of the soft part and the initial values ​​μlsO and μhsO of the attenuation coefficient of the soft part is stored in the memory 53.

[0213] Figure 24 This is a diagram showing the relationship between the initial value tsO of the specified soft part thickness ts and the initial values ​​μls0 and μhs0 of the soft part attenuation coefficients. The initial value derivation unit 81 refers to table LUT5 stored in memory 53 and derives the initial values ​​μls0 and μhs0 of the soft part attenuation coefficients corresponding to the initial value tsO of the soft part thickness ts.

[0214] The attenuation coefficient derivation unit 82 derives the attenuation coefficients μls and μhs for the soft parts and μlb and μhb for the bone parts, each with different energy distributions. Here, in order to perform energy subtraction processing, the subject H is photographed using radiation with different energy distributions to obtain low-energy and high-energy images. In this embodiment, the first radiation image G1 is a low-energy image and the second radiation image G2 is a high-energy image. The pixel values ​​G1(x,y) of each pixel in the first radiation image G1 (low-energy image) and G2(x,y) of each pixel in the second radiation image G2 (high-energy image) are represented by the thickness ts(x,y) of the soft parts, the thickness tb(x,y) of the bone parts, and the attenuation coefficients μls(x,y), μhs(x,y), μlb(x,y), and μhb(x,y) at the corresponding pixel positions, as shown in equations (3) and (4) below. In addition, the description of (x,y) is omitted in equations (3) and (4).

[0215] G1=μls×ts+μlb×tb (3)

[0216] G2=μhs×ts+μhb×tb (4)

[0217] To derive the weighting coefficient α used for energy subtraction, it is necessary to derive the attenuation coefficients μls(x,y), μhs(x,y), μlb(x,y), and μhb(x,y). As mentioned above, the attenuation coefficients μls(x,y), μhs(x,y), μlb(x,y), and μhb(x,y) are expressed as functions of the soft tissue thickness ts and the bone thickness tb. Therefore, to derive the attenuation coefficients μls(x,y), μhs(x,y), μlb(x,y), and μhb(x,y), it is necessary to derive the soft tissue thickness ts and the bone thickness tb. If equations (3) and (4) are solved with respect to ts and tb, they become equations (5) and (6) as follows.

[0218] ts={μhb×G1-μlb×G2} / {μls×μhb-μlb×μhs} (5)

[0219] tb={μls×G2-μhs×G1} / {μls×μhb-μlb×μhs} (6)

[0220] Here, the attenuation coefficients μls(x,y), μhs(x,y), μlb(x,y), and μhb(x,y) on the right side of equations (5) and (6) are expressed as functions of the thickness ts of the soft tissue and the thickness tb of the bone tissue, so equations (5) and (6) cannot be solved algebraically.

[0221] Therefore, in the fourth embodiment, error functions EL and EH are set as shown in equations (7) and (8) below. The error functions EL and EH correspond to the differences between the values ​​of the soft part attenuation coefficient × soft part thickness + bone part attenuation coefficient × bone part thickness in different energy distributions and the pixel values ​​of the radiographic image. Then, in order to minimize both error functions EL and EH, in the fourth embodiment, error function E0 is set as shown in equation (9). Then, while changing the soft part thickness ts, bone part thickness tb, and attenuation coefficients μls, μhs, μlb, and μhb from their initial values, a combination of soft part thickness ts and bone part thickness tb that minimizes error function E0 or is less than a predetermined threshold is derived. At this time, it is preferable to use optimization algorithms such as the steepest descent method and the conjugate gradient method to derive the soft part thickness ts and bone part thickness tb. The initial values ​​of the soft part thickness ts, the bone part thickness tb, and the attenuation coefficients μls, μhs, μ1b, and μhb used at this time are ts0, tb0, μls0, μhs0, μlb0, and hb0 derived by the initial value derivation unit 81.

[0222] EL=G1-{μlS×ts+μlb×tb} (7)

[0223] EH=G2-{μhs×ts+μhb×tb} (8)

[0224] E0 = EL 2 +EH 2 (9)

[0225] Furthermore, the attenuation coefficient used in deriving the thicknesses ts of the soft tissue and tb of the bone tissue is derived by referring to a pre-defined table relating these two thicknesses. This table is stored in the memory 53 of the information deriving device.

[0226] Figure 25 This is a graph showing the relationship between the thickness ts of the soft tissue and the thickness tb of the bony tissue and the attenuation coefficient. (See table for example.) Figure 25 As shown, Table LUT6 three-dimensionally represents the relationship between the thickness ts of the soft tissue and the thickness tb of the bony tissue and the attenuation coefficient μ. Furthermore, in Figure 25 Only one LUT6 is shown, but tables for the attenuation coefficients μls, μhs, μlb, and μhb are prepared and stored in memory. Here, in table LUT6, the larger the thickness ts of the soft part and the thickness tb of the bone part, the smaller the attenuation coefficient μ becomes.

[0227] If the thickness ts of the soft part and the thickness tb of the bone part are derived such that the error function E0 becomes the minimum or becomes less than a predetermined threshold, then the attenuation coefficient derivation section 82 derives the attenuation coefficients μls, μhs, μlb, and μhb from reference table LUT6.

[0228] The weighting coefficient derivation unit 83 derives the weighting coefficient α used by the subtraction unit 63 during subtraction processing. That is, the weighting coefficient derivation unit 83 derives the weighting coefficient α by performing the calculation α = μls / μhs using the attenuation coefficients μls, μhs, μlb, and μhb derived by the attenuation coefficient derivation unit 82.

[0229] In the fourth embodiment, the subtraction unit 63 uses the weighting coefficient α derived by the weighting coefficient derivation unit 83 to derive the bone image Gb using the above formula (1). Then, the bone density derivation unit 64 derives the bone density B and the representative value of the bone density B in the target bone.

[0230] Next, the fifth embodiment of the present invention will be described. Figure 26 This is a diagram illustrating the functional structure of the information export device based on the fifth embodiment. Additionally, in Figure 26 In the middle, to and Figure 7 The same structures are assigned the same reference symbols, and detailed descriptions are omitted. In the fifth embodiment of the present invention, when deriving the bone density as positive resolution data 42, the bone image Gb is derived by the following steps: deriving the compositional proportions of the multiple components contained in the soft part of the subject H; deriving the attenuation coefficients of each energy distribution of the soft part for each pixel of the first radiographic image G1 and the second radiographic image G2 according to the compositional proportions; and performing subtraction processing using weighting coefficients derived from the attenuation coefficients of the soft part and the pre-defined attenuation coefficients of the bone part.

[0231] Therefore, as Figure 26 As shown, the information export device 50D based on the fifth embodiment further includes a component ratio export unit 84 and an attenuation coefficient setting unit 85 compared to the information export device 50 based on the first embodiment.

[0232] The composition ratio derivation unit 84 obtains the composition ratio of the subject H. In the fifth embodiment, the composition ratio derivation unit 84 obtains the composition ratio by deriving the composition ratio of the subject H from the first radiographic image G1 and the second radiographic image G2. Furthermore, in the fifth embodiment, the composition ratio of fat is derived as the composition ratio. Therefore, although the subject H includes bone, for the sake of explanation, it is assumed that only soft tissue is included in the first radiographic image G1 and the second radiographic image G2, excluding bone.

[0233] To derive the composition ratio, the composition ratio derivation unit 84 first derives the body thickness of the subject H for each pixel of the first radiation image G1 and the second radiation image G2, respectively, as the first body thickness and the second body thickness. Specifically, the composition ratio derivation unit 84 assumes that the brightness distribution of the first radiation image G1 is consistent with the body thickness distribution of the subject H, and derives the first body thickness t1 of the subject H by converting the pixel values ​​of the first radiation image G1 into thickness using the attenuation coefficient of the radiation in the muscles of the subject H. Similarly, the composition ratio derivation unit 84 assumes that the brightness distribution of the second radiation image G2 is consistent with the body thickness distribution of the subject H, and derives the second body thickness t2 of the subject H by converting the pixel values ​​of the second radiation image G2 into thickness using the attenuation coefficient in the muscles of the subject H.

[0234] Here, the degree of beam hardening depends on the thickness tf of fat and the thickness tm of muscle in the subject H. Therefore, the attenuation coefficients μf of fat and μm of muscle can be defined as μf(tf, tm) and μm(tf, tm) as nonlinear functions of the thickness tf of fat and the thickness tm of muscle.

[0235] In the fifth embodiment, the attenuation coefficient of fat in the first radiographic image G1, which is a low-energy image, can be expressed as μlf(tf, tm), and the attenuation coefficient of muscle can be expressed as μlm(tf, tm). Furthermore, the attenuation coefficient of fat in the second radiographic image G2, which is a high-energy image, can be expressed as μhf(tf, tm), and the attenuation coefficient of muscle can be expressed as μhm(tf, tm).

[0236] Furthermore, the pixel values ​​G1(x,y) of each pixel in the soft region of the first radiometric image G1 (a low-energy image) and G2(x,y) of each pixel in the soft region of the second radiometric image G2 (a high-energy image) are represented by the fat thickness tf(x,y), muscle thickness tm(x,y), and attenuation coefficients μlf(x,y), μhf(x,y), μlm(x,y), and μhm(x,y) at the corresponding pixel locations, respectively, by the following equations (10) and (11). In addition, the expression for (x,y) is omitted in equations (10) and (11).

[0237] G1=μlf×tf+μlm×tm (10)

[0238] G2=μhf×tf+μhm×tm (11)

[0239] As described above, in the fifth embodiment, when deriving the first body thickness t1 and the second body thickness t2, the pixel values ​​of the first radiographic image G1 and the second radiographic image G2 are converted into thickness using the attenuation coefficient of the muscles in the subject H. Therefore, in the fifth embodiment, the component ratio deriving unit 84 derives the first body thickness t1 and the second body thickness t2 using the following formulas (12) and (13). In addition, the description of (x, y) is omitted in formulas (12) and (13).

[0240] t1=G1 / μlm (12)

[0241] t2=G2 / μhm (13)

[0242] At the pixel positions where the first body thickness t1 and the second body thickness t2 are derived, the first body thickness t1 and the second body thickness t2 are the same when the subject H contains only muscle. However, in the actual subject H, at the same pixel positions in the first radiometric image G1 and the second radiometric image G2, both muscle and fat are present. Therefore, the first body thickness t1 and the second body thickness t2 derived using equations (12) and (13) are not consistent with the actual body thickness of the subject H. Furthermore, in the first body thickness t1 derived from the first radiometric image G1 as a low-energy image and the second body thickness t2 derived from the second radiometric image G2 as a high-energy image, the first body thickness t1 becomes a value greater than the second body thickness t2.

[0243] For example, such as Figure 27 As shown, assuming an actual body thickness of 100 mm, with fat and muscle thicknesses of 30 mm and 70 mm respectively, the first body thickness t1 derived using a first radiation image G1 obtained from low-energy radiation is, for example, 80 mm, and the second body thickness t2 derived using a second radiation image G2 obtained from high-energy radiation is, for example, 70 mm. Furthermore, the greater the proportion of fat, the greater the difference between the first body thickness t1 and the second body thickness t2.

[0244] Here, the difference between the first body thickness t1 and the second body thickness t2 varies depending on the composition ratio of fat and muscle in the subject H. Therefore, in the fifth embodiment, a subject model with varying fat composition ratios is photographed using radiation with different energy distributions. The body thickness is derived from the two radiation images obtained, and a table showing the correspondence between the body thickness difference derived from the two radiation images and the fat composition ratio is pre-created and stored in the memory 53.

[0245] Figure 28 This is a graph showing the correspondence between the body thickness difference derived from two radiographic images and the proportion of fat composition. For example... Figure 28As shown in Table LUT7, the horizontal axis represents the body thickness difference derived from the two radiographic images, and the vertical axis represents the proportion of fat composition. Figure 28 As shown, the greater the difference in body thickness derived from the two radiographic images, the greater the proportion of fat composition. Furthermore, a table establishing the correspondence between the difference in body thickness derived from the two radiographic images and the proportion of fat composition is prepared according to each energy distribution of the radiation used during imaging, and stored in memory 53.

[0246] The composition ratio extraction unit 84 extracts the difference between the first body thickness t1 and the second body thickness t2, and extracts the fat composition ratio R(x, y) by referring to the LUT7 stored in the memory of the information extraction device 50D. In addition, the muscle composition ratio can be extracted by subtracting the extracted fat composition ratio R(x, y) from 100%.

[0247] The attenuation coefficient setting unit 85 sets the attenuation coefficient of the radiation used when obtaining the first radiation image G1 and the second radiation image G2 for each pixel of the first radiation image G1 and the second radiation image G2 according to the fat composition ratio R(x, y). Specifically, the attenuation coefficient of the soft part of the subject H is set. Here, in the fifth embodiment, the first radiation image G1 is equivalent to a low-energy image, and the second radiation image G2 is equivalent to a high-energy image. Therefore, the attenuation coefficient setting unit 85 derives the attenuation coefficient μls(x, y) of the soft part of the low-energy image and the attenuation coefficient μhs(x, y) of the soft part of the high-energy image using the following formulas (14) and (15). In addition, the description of (x, y) is omitted in formulas (14) and (15). Furthermore, μlm is the attenuation coefficient of muscle in the low-energy image, μlf is the attenuation coefficient of fat in the low-energy image, μhm is the attenuation coefficient of muscle in the high-energy image, and μhf is the attenuation coefficient of fat in the high-energy image.

[0248] μls=(1-R)×μlm+R×μlf (14)

[0249] μhs=(1-R)×μhm+R×μhf (15)

[0250] In the fifth embodiment, the subtraction unit 63 performs subtraction processing using the attenuation coefficients μls and μhs set by the attenuation coefficient setting unit 85. At this time, the subtraction unit 63 derives the weighting coefficient α in the above formula (1). Using the attenuation coefficients set by the attenuation coefficient setting unit 85, the weighting coefficient α used in the fifth embodiment is derived using α = μls / μhs. Then, the subtraction unit 63 uses the derived weighting coefficient α to derive the bone image Gb using the above formula (1). Then, the bone density deriving unit 64 derives the bone density B and a representative value of the bone density B in the target bone.

[0251] Next, the information export apparatus based on the sixth embodiment of the present invention will be described. Furthermore, the structure of the information export apparatus based on the sixth embodiment is the same as that of the information export apparatus 50 based on the first embodiment, except that the processing performed by the scattering line removal unit 62 is different; therefore, a detailed description is omitted here. Figure 29 This is a diagram illustrating the functional structure of the scattering line removal unit in the information export apparatus based on the sixth embodiment. (See diagram for example.) Figure 29 As shown, the scattering line removal unit 62A in the information export device based on the sixth embodiment includes a photography condition acquisition unit 91, a volume thickness export unit 92, a characteristic acquisition unit 93, a line distribution export unit 94, and a calculation unit 95.

[0252] Furthermore, in the sixth embodiment, by using Figure 30 The photographic apparatus 1A shown also uses a diffused radiation removal grid 8 to photograph a subject H lying supine on a photographic table 4 to acquire a first radiation image G1 and a second radiation image G2, and uses the first radiation image G1 and the second radiation image G2 as teacher data 40. Additionally, in Figure 30 In the imaging stage 4, there is a top plate 4A. Below the top plate 4A, from the side of the radiation source 3, there are a scattered radiation removal grid (hereinafter referred to as grid) 8, a first radiation detector 5, a radiation energy conversion filter 7 and a second radiation detector 6, which are installed below the top plate 4A via a mounting part 4B.

[0253] When generating teacher data 40, the photography condition acquisition unit 91 acquires the photography conditions used for energy subtraction imaging of the subject H from the image storage system 9 in order to acquire the first radiation image G1 and the second radiation image G2, which are used as learning data 41. Furthermore, in the sixth embodiment, the photography conditions also include the line quality of the radiation. The line quality is defined using one or more of the following: tube voltage [kV] of the radiation generator in the radiation source 3, total filtration amount [mmAl equivalent], and half-value layer [mmAl]. Tube voltage refers to the maximum value of the generated radiation energy distribution. Total filtration amount is a value obtained by converting the filtration amount of each component of the imaging device 1, such as the radiation generator and collimator in the radiation source 3, into the thickness of aluminum. The larger the total filtration amount, the greater the beam hardening effect in the imaging device 1, and the more high-energy components 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 relative to the generated radiation energy distribution. The thicker the aluminum in the half-value layer, the more high-energy components in the wavelength distribution of the radiation.

[0254] The thickness derivation unit 92 derives the thickness distribution of the subject H based on at least one of the first radiographic image G1 and the second radiographic image G2, and the photographic conditions. Hereinafter, the thickness distribution derived by the thickness derivation unit 92 is referred to as the initial thickness distribution T0. The derivation of the initial thickness distribution T0 will be explained below. Furthermore, in the following explanation, the derivation of the thickness and the removal of scattered rays using the first radiographic image G1 will be described, but the thickness derivation and scattered ray removal can also be performed using the second radiographic image G2.

[0255] First, in the absence of a subject H, when the radiation source 3 is driven to irradiate the radiation detector 5, the arrival dose I0(x, y) of the radiation emitted from the radiation source 3 to the radiation detector 5 is represented by the following equation (16). In equation (16), mAs included in the imaging conditions is the tube current-time product, and kV is the tube voltage. In addition, when the half-value layer is also considered, the arrival dose I0(x, y) is represented by the following equation (16-1). Here, F is a nonlinear function representing the amount of radiation arriving at the radiation detector 5 when the dose (e.g., 1 mAs) that will become the reference at the reference SID (e.g., 100 cm) irradiates the radiation detector 5 in the absence of a subject H. F varies according to each tube voltage or depends on the tube voltage and the half-value layer. Furthermore, the arrival dose I0 is derived for each pixel of the radiation image acquired by the radiation detector 5, so (x, y) represents the pixel position of each pixel. Furthermore, in the following explanation, both cases are included, with and without considering the half-value layer. Therefore, as shown in equation (16-2) below, mmAl is included in the parentheses in each equation.

[0256] I0(x, y) = mAs × F(kV) / SID 2 (16)

[0257] I0(x, y)=mAs×F(kV, mmAl) / SID 2 (16-1)

[0258] I0(x,y)=mAs×F(kV(,mmAl)) / SID 2 (16-2)

[0259] Furthermore, if the initial body thickness distribution is set to T0, the attenuation coefficient of the subject H with the initial body thickness distribution T0 is set to μ(T0), and the ratio of the amount of scattered rays to the amount of primary rays in the radiation transmitted through the subject H with the initial body thickness distribution T0 without considering the diffusion of scattered rays is set to STPR(T0), then the dose I1 after transmission through the subject H is expressed by the following equation (17). In addition, in equation (17), the initial body thickness distribution T0, the dose I0 reached, and the dose I1 are derived for each pixel of the simple radiation image G0, but (x, y) are omitted. Furthermore, STPR is a nonlinear function that depends not only on the body thickness but also on the tube voltage [kV] and the half-value layer [mmAl], but the notation of kV and mmAl is omitted in equation (17).

[0260] I1=I0×exp{-μ(T0)×T0}×{1+STPR(TO)} (17)

[0261] In equation (17), dose I1 is the pixel value at each pixel of the simple radiation image G0, and the dose I0 is derived using equations (16) and (16-1) above. On the one hand, since F and STPR are nonlinear functions, equation (17) cannot be solved algebraically with respect to T0. Therefore, the thickness derivation unit 92 defines the error function E1 shown in equation (18) or equation (18-1) below. Then, T0 that minimizes the error function E1 or is less than a predetermined threshold is derived as the initial thickness distribution. At this time, the thickness derivation unit 92 uses optimization algorithms such as the steepest descent method and the conjugate gradient method to derive the initial thickness distribution T0.

[0262] E1=[I1-I0×exp{-μ(T0)×T0}×{1+STPR(T0)}] 2 (18)

[0263] E1=|I1-I0×exp{-μ(T0)×T0}×{1+STPR(T0)}| (18-1)

[0264] The characteristic acquisition unit 93 acquires the radiation characteristics of an object located between the subject H and the first radiation detector 5 and the second radiation detector 6 during the imaging process. Here, when radiation transmitted through the subject H is transmitted through the object between the subject H and the radiation detector 5, the transmittance of the radiation varies depending on the quality of the radiation transmitted through the subject H. Furthermore, the transmittance of the primary rays and scattered rays included in the radiation transmitted through the subject H differs depending on the direction of radiation travel and the quality of the rays. Therefore, in the sixth embodiment, the transmittance of the primary rays and the transmittance of the scattered rays of the object are used as the radiation characteristics of the object.

[0265] Furthermore, as described above, when radiation transmitted through the subject H is transmitted through an object between the subject H and the radiation detector 5, the transmittance of the radiation varies depending on the quality of the radiation transmitted through the subject H. Moreover, the quality of the radiation transmitted through the subject H depends on the thickness distribution T of the subject H. Therefore, the primary ray transmittance and the scattered ray transmittance can be expressed as functions of the thickness distribution T of the subject H as Tp(T) and Ts(T), respectively.

[0266] Furthermore, the quality of the radiation transmitted through the subject H also depends on the quality of the radiation source 3 included in the photographic conditions. The quality depends on the tube voltage and the half-value layer. Therefore, strictly speaking, the primary ray transmittance and the scattered ray transmittance are expressed as Tp(kV(, mmAl), T) and Ts(kV(, mmAl), T), respectively. However, in the following explanation, the primary ray transmittance and the scattered ray transmittance are sometimes simply expressed as Tp and Ts.

[0267] Here, as described above, the primary ray transmittance Tp and scattered ray transmittance Ts of the object between the subject H and the radiation detector 5 depend on the body thickness distribution T of the subject H. Therefore, in the sixth embodiment, the primary ray transmittance Tp and scattered ray transmittance Ts of the object corresponding to the body thickness distribution T of the subject H are measured using phantoms of various thicknesses that simulate the body thickness distribution T of the subject H. Based on the measurement results, a table specifying the relationship between the body thickness distribution T of the subject H and the primary ray transmittance Tp and scattered ray transmittance Ts of the object is generated and stored in the memory of the information export device based on the sixth embodiment. Hereinafter, the measurement of the primary ray transmittance Tp and scattered ray transmittance Ts of the object corresponding to the body thickness distribution T of the subject H will be described.

[0268] First, the calculation of the scattered radiation transmittance Ts is explained. Figure 31 and Figure 32 This is a graph used to illustrate the measurement of scattered light transmittance Ts corresponding to the body thickness of the subject H. First, as... Figure 31 As shown, a human-like phantom 101 is placed on the surface of the radiation detector 5, and a lead plate 102 is further placed on the phantom 101. Here, the phantom 101 has various thicknesses such as 5 cm, 10 cm, and 20 cm, and is made of materials such as acrylic, which has the same radiation transmittance as water. In this state, by driving the radiation source 3 to irradiate the radiation detector 5, the characteristic acquisition unit 93 acquires a measurement radiation image K0. The signal value of the radiation image K0 is relatively large in the area of ​​direct radiation irradiation in the radiation detector 5, and the signal value decreases sequentially from the area of ​​the phantom 101 to the area of ​​the lead plate 102.

[0269] Furthermore, the lead plate 102 does not transmit radiation, therefore the signal value in the region of the lead plate 102 in the radiation image K0 should be 0. However, the radiation scattered by the phantom 101 reaches the region of the radiation detector 5 corresponding to the lead plate 102. Therefore, the region of the lead plate 102 in the radiation image K0 has a signal value S0 corresponding to the component of scattered radiation generated by the phantom 101.

[0270] Next, as Figure 32 As shown, a phantom 101 is placed on the top plate 4A of the photographic apparatus 1, and a lead plate 102 is further placed on the phantom 101. Then, similar to the case of photographing the subject H, with the radiation detector 5 and grid 8 arranged below the top plate 4A, the radiation source 3 is driven to irradiate radiation into the radiation detector 5, thereby the characteristic acquisition unit 93 acquires a measurement radiation image K1. Similar to the radiation image K0, the signal value of the radiation image K1 is larger in the area of ​​direct irradiation of radiation in the radiation detector 5, and the signal value decreases in the order of the area of ​​the phantom 101 and the area of ​​the lead plate 102. Here, as Figure 32 As shown, when the image is taken with the top plate 4A and the grid 8 positioned between the phantom 101 and the radiation detector 5, not only the radiation scattered by the phantom 101, but also the radiation scattered by the top plate 4A and the grid 8 reaches the region of the radiation detector 5 corresponding to the lead plate 102. Therefore, the region of the lead plate 102 in the radiation image K1 has a signal value S1 corresponding to the scattered radiation component generated by the phantom 101, the top plate 4A, and the grid 8.

[0271] In addition, the signal value S1 includes the scattered line component generated by the top plate 4A and the grid 8, and therefore becomes greater than Figure 31 The signal value S0 is shown. Therefore, when photographing a phantom 101 of thickness t, the scattered radiation transmittance Ts of the object between the subject H and the radiation detector 5, namely the top plate 4A and the grid 8, can be calculated using S1 / S0.

[0272] In the sixth embodiment, the characteristic acquisition unit 93 uses at least two phantoms with different thicknesses to... Figure 31 and Figure 32 The scattered ray transmittance Ts corresponding to each thickness is calculated in the manner shown. Furthermore, for the scattered ray transmittance Ts of thicknesses not present in the phantom 101, the characteristic acquisition unit 93 derives it by interpolating the measured scattered ray transmittance Ts for multiple thicknesses. Then, the characteristic acquisition unit 93 interpolates the scattered ray transmittance for thicknesses between each thickness, thereby obtaining the scattered ray transmittance as shown. Figure 33 As shown, Table LUT8 is generated, which represents the relationship between the thickness of the subject H and the scattered light transmittance Ts of the object between the subject H and the radiation detector 5.

[0273] Next, the calculation of primary ray transmittance will be explained. Figure 34 and Figure 35 This is a graph used to illustrate the measurement of primary ray transmittance Tp corresponding to the body thickness of the subject H. First, as... Figure 34 As shown, a phantom 101 simulating a human body is placed on the surface of the radiation detector 5. Here, the phantom 101 used is the same phantom used in the case of deriving the scattered ray transmittance Ts. Then, in this state, by driving the radiation source 3 to irradiate the radiation detector 5, the characteristic acquisition unit 93 acquires a measurement radiation image K2. The signal value S2 of the region corresponding to the phantom 101 in the radiation image K2 includes both the primary ray component and the scattered ray component of the radiation transmitted through the phantom 101. Here, the scattered ray component of the radiation transmitted through the phantom 101 is obtained through the above-described... Figure 31 The signal value S0 in the radiation image KO is obtained by the method shown. Therefore, the primary ray component of the radiation transmitted through phantom 101 is derived from S2-S0.

[0274] Next, as Figure 35 As shown, a phantom 101 is mounted on the top plate 4A of the imaging device 1. Similar to the case of photographing the subject H, with the radiation detector 5 and grid 8 positioned below the top plate 4A, the radiation source 3 is driven to irradiate the radiation detector 5, thereby the characteristic acquisition unit 93 acquires a measurement radiation image K3. The signal value S3 of the region corresponding to the phantom 101 in the radiation image K3 includes both the primary ray component and the scattered ray component of the radiation transmitted through the phantom 101, the top plate 4A, and the grid 8. Here, the scattered ray component of the radiation transmitted through the phantom 101, the top plate 4A, and the grid 8 is obtained through the aforementioned... Figure 32 The signal value S1 in the radiation image K1 obtained by the method shown is thus derived from S3-S1. Therefore, the primary ray component of the radiation transmitted through the phantom 101, the top plate 4A, and the grid 8 is derived from S3-S1.

[0275] Therefore, the primary ray transmittance Tp of the top plate 4A and the grid 8 between the subject H and the radiation detector 5 when photographing the phantom 101 can be calculated using (S3-S1) / (S2-S0). Then, in the sixth embodiment, the characteristic acquisition unit 93 uses at least two phantoms with different thicknesses to... Figure 34 and Figure 35 The primary ray transmittance Tp corresponding to each thickness is calculated in the manner shown. Furthermore, for the primary ray transmittance Tp of thicknesses not present in the phantom 101, the characteristic acquisition unit 93 derives it by interpolating the measured primary ray transmittance Tp for multiple thicknesses. Then, the characteristic acquisition unit 93... Figure 36The table LUT9 shows the relationship between the thickness of the subject H and the primary ray transmittance Tp of the object between the subject H and the radiation detector 5.

[0276] The tables LUT8 and LUT9 generated as described above are stored in the memory 53 of the information export device based on the sixth embodiment. In addition, tables are generated according to various imaging conditions (i.e., line quality, dose, and distance from the radiation source) and the type of grid 8 to be used, and stored in the memory 13.

[0277] The characteristic acquisition unit 93 acquires the primary ray transmittance Tp(T0) and scattered ray transmittance Ts(T0) of the object between the subject H and the radiation detector 5, corresponding to the initial body thickness distribution T0, by referring to tables LUT8 and LUT9 stored in memory 53 based on the photographic conditions acquired by the photographic conditions acquisition unit 91. Furthermore, the primary ray transmittance Tp and scattered ray transmittance Ts also depend on the ray quality; therefore, the primary ray transmittance Tp and scattered ray transmittance Ts can be expressed as Tp(kV(, mmAl), T0) and Ts(kV(, mmAl), T0), respectively.

[0278] The line distribution derivation unit 94 uses photographic conditions, volume thickness distribution, and the radiation characteristics of the object between the subject H and the radiation detector 5 to derive the primary ray distribution and scattered ray distribution of the radiation detected by the radiation detector 5. Here, when the volume thickness distribution is set to T, the primary ray distribution Ip0 and the scattered ray distribution Is0 of the radiation transmitted through the subject H can be represented by the following equations (19) and (20). In equation (20), PSF is the point spread function representing the distribution of scattered rays diffusing from one pixel, defined according to the line quality and volume thickness. And, * indicates convolution operation. The primary ray distribution Ip0 and the scattered ray distribution Is0 are derived for each pixel of the simple radiation image G0, but (x, y) are omitted in equations (19) and (20). Furthermore, in the sixth embodiment, as described later, the derivation of the volume thickness distribution, the primary ray distribution Ip0, and the scattered ray distribution Is0 is repeated. When the primary ray distribution Ip0 and the scattered ray distribution Is0 are derived for the first time, the volume thickness distribution T uses the initial volume thickness distribution T0.

[0279] Ip0=I0×exp{-μ(T)×T} (19)

[0280] Is0=Ip0×STPR(kV(,mmAl),T)*PSF(kV(,mmAl),T) (20)

[0281] Furthermore, the line distribution derivation unit 94 uses the primary ray transmittance Tp and scattered ray transmittance Ts of the object between the subject H and the radiation detector 5 to derive the primary ray distribution Ip1 and scattered ray distribution Is1 reaching the radiation detector 5 using the following equations (21) and (22). Furthermore, the sum Iw1 of the primary ray distribution Ip1 and the scattered ray distribution Is1 is derived using the following equation (23). In equations (21) and (22), the initial thickness distribution T is also used when deriving the primary ray distribution Ip1 and the scattered ray distribution Is1 for the first time.

[0282] Ip1=Ip0×Tp(kV(,mmAl),T) (21)

[0283] Is1=Is0×Ts(kV(,mmAl),T) (22)

[0284] Iw1=Ip1+Is1 (23)

[0285] The computation unit 95 derives the error E2 between the sum Iw1 of the primary ray distribution Ip1 and the scattered ray distribution Is1 and the dose, i.e., the pixel value I1, at each pixel position of the first radiation image G1. The error E2 is derived using the following equation (24) or equation (24-1). In equations (24) and (24-1), N represents the number of pixels in the first radiation image G1, and ∑ represents the sum of all pixels in the first radiation image G1. Furthermore, equation (24-1) performs the calculation of I1 / Iw1 within log, thus enabling the derivation of the error E2 independent of the dose I0 reached when the irradiated subject H is irradiated.

[0286] E2=(1 / N)×∑{I1-Iw1} 2 (twenty four)

[0287] E2=(1 / N)×∑|log{I1 / Iw1} (24-1)

[0288] Then, the calculation unit 95 updates the volume thickness distribution T in a manner that minimizes the error E2 or makes the error E2 less than a predetermined threshold. The calculation unit 95 then repeats the process of obtaining the primary ray transmittance Tp and the scattered ray transmittance Ts based on the updated volume thickness distribution, and deriving the primary ray distribution Ip1 and the scattered ray distribution Is1. Here, the calculation performed by the calculation unit 95 is referred to as repeated calculation. Furthermore, in the sixth embodiment, the calculation unit 95 performs repeated calculations in a manner that makes the error E2 less than a predetermined threshold. Then, the calculation unit 95 outputs a first radiation image Gc1, which is processed with the primary ray distribution Ipc derived from the volume thickness distribution Tc of the subject H with an error E2 less than the predetermined threshold as the pixel value.

[0289] In addition, the repeated acquisition of primary ray transmittance Tp and scattered ray transmittance Ts, as well as the repeated derivation of primary ray distribution Ip1 and scattered ray distribution Is1, are performed by the characteristic acquisition unit 93 and the line distribution derivation unit 94, respectively.

[0290] On the other hand, for the second radiation image G2, the primary ray distribution Ipc is also derived in the same way as for the first radiation image G1. Furthermore, the primary ray distribution for the first radiation image G1 is set to Ipc-1, and the primary ray distribution for the second radiation image G2 is set to Ipc-2. Then, the processing unit 95 outputs the second radiation image Gc2, which has been processed with the primary ray distribution Ipc-2 as pixel values.

[0291] In the sixth embodiment, the subtraction unit 63 uses the processed first radiographic image Gc1 and the second radiographic image Gc2 to derive a bone image Gb. Then, the bone density deriving unit 64 uses the bone image Gb derived by the subtraction unit 63 to derive the bone density B and a representative value of the bone density B in the target bone.

[0292] Furthermore, the learned neural network 23A, constructed by using the bone density B derived in the sixth embodiment as forward solution data for learning, if the input is... Figure 30 The photographic device 1A shown simply captures a simple radiometric image G0 of the subject H, and outputs the bone density of the object between the subject H and the radiometric detector.

[0293] Furthermore, in the sixth embodiment described above, the top plate 4A of the imaging stage 4 and the grid 8 are used as the object between the subject H and the radiation detector 5, but as Figure 37 As shown, sometimes the air layer 103 is located between the top plate 4A and the grid 8. In this case, in the online distribution derivation unit 94, it is preferable to include the air layer 103 in the object between the subject H and the radiation detector 5 to derive the primary ray distribution Ip1 and the scattered ray distribution Is1. In this case, as shown in equations (21-1) and (22-1) below, the primary ray distribution Ip1 and the scattered ray distribution Is1 can be derived by performing a convolution operation on the point spread function PSFair(kV(, mmAl), tair) corresponding to the thickness tair of the air layer 103 with respect to the above equations (21) and (22). In addition, the thickness tair of the air layer 103 is the distance between the lower surface of the top plate 4A and the surface of the grid 8 on the subject H side.

[0294] Ip1=Ip0×Tp(kV(,mmAl),T)*PSF air (kV(,mmAl),t air (21-1)

[0295] Is1=Is0×Ts(kV(,mmAl),T)*PSF air (kV(,mmAl),t air (22-1)

[0296] Next, the information export apparatus based on the seventh embodiment of the present invention will be described. Furthermore, the structure of the information export apparatus based on the seventh embodiment is the same as that based on the first embodiment, except that the processing performed by the scattering line removal unit 62 is different; therefore, a detailed description is omitted here. Figure 38 This is a diagram illustrating the functional structure of the scattering line removal unit in the information export apparatus based on the seventh embodiment. (See diagram for example.) Figure 38 As shown, the scattering line removal unit 62B in the information export device according to the seventh embodiment includes a first export unit 97, a second export unit 98, and an image generation unit 99.

[0297] The information export apparatus based on the seventh embodiment performs scattering removal processing through the following steps: The first export unit 97 uses the first radiation image G1 and the second radiation image G2 to export the first primary ray distribution and the scattering distribution of radiation transmitted through the subject H; the second export unit 98 uses the first primary ray distribution, the scattering distribution, and the radiation characteristics of the object between the subject H and the radiation detector that detects the radiation image to export the second primary ray distribution and the scattering distribution of radiation transmitted through the object; the image generation unit 99 uses the second primary ray distribution and the scattering distribution to export the radiation image after transmission through the subject and the object. Furthermore, the scattering removal processing performed by the information export apparatus based on the seventh embodiment is described in International Publication No. 2020 / 241664. Hereinafter, the scattering removal processing from the first radiation image G1 will be described, but the scattering removal processing can also be performed on the second radiation image G2.

[0298] The first export unit 97 uses the first radiation image G1 to estimate the components of the radiation transmitted through the subject H (primary ray component and scattered ray component). In the export process performed by the first export unit 97 (hereinafter referred to as the first export process), "radiation transmitted through the subject H" refers to the radiation transmitted through the subject H and in front of objects such as the top plate 4A and the grid 8.

[0299] Furthermore, specifically, the component of radiation transmitted through the subject H refers to the component of radiation transmitted through the subject H and / or the component of radiation scattered by the subject H. That is, the component of radiation transmitted through the subject H refers to the primary ray component after transmission through the subject H. The component of radiation scattered by the subject H refers to the scattered ray component after transmission through the subject H. Regarding radiation incident on the subject H at any position X, if the subject H is considered as the operator g1 generating the primary ray component and the operator h1 generating the scattered ray component, then as follows... Figure 39 As shown, the primary ray component after passing through the subject H is g1(X), and the scattered ray component after passing through the subject H is h1(X).

[0300] The first extraction unit 97 uses the first radiation image G1 to estimate the primary ray component g1(X) after passing through the subject H, the scattered ray component h1(X) after passing through the subject H, or both. In the seventh embodiment, the first extraction unit 97 uses the first radiation image G1 to estimate the primary ray component g1(X) after passing through the subject H and the scattered ray component h1(X) after passing through the subject H, respectively.

[0301] Furthermore, when estimating the primary ray component g1(X) after transmission through the subject H based on the first radiation image G1, the first derivation unit 97 estimates the scattered ray component h1(X) after transmission through the subject H by subtracting the estimated primary ray component g1(X) from the first radiation image G1. Also, when estimating the scattered ray component h1(X) after transmission through the subject H based on the first radiation image G1, the first derivation unit 97 estimates the primary ray component g1(X) after transmission through the subject H by subtracting the estimated scattered ray component h1(X) from the first radiation image G1.

[0302] The first export process performed by the first export unit 97 can, for example, be performed by estimating the thickness of the subject H using the first radiation image G1 and estimating the components of radiation transmitted through the subject H using the estimated thickness of the subject H. In this case, the first export unit 97 estimates, based on the estimated thickness of the subject H, the primary ray component g1(X) of radiation transmitted through the subject H and the scattered ray component h1(X) of radiation scattered by the subject H for each pixel of the first radiation image G1 (or for each defined segment composed of multiple pixels).

[0303] For example, such as Figure 40As shown, the pixel value V2 when there is a subject H ("with subject") is less than the pixel value V1 when there is no subject H ("without subject"). This is caused by absorption by the subject H, etc. Therefore, their difference Δ (=V1-V2) is related to the thickness of the subject H. On the other hand, the pixel value V1 when there is no subject H can be obtained from the pixel value of the area (direct area) where the radiation reaches the radiation detector 5 without passing through the subject H, or from a prior experiment (imaging without the subject H). Therefore, the first derivation unit 97 can estimate the thickness of the subject H based on the pixel value V2 of the first radiation image G1 captured with the subject H in between.

[0304] Furthermore, both the primary ray component g1(X) and the scattered ray component h1(X) after transmission through the subject H are related to the thickness of the subject H. For example, the thicker the subject H, the less the primary ray component g1(X) becomes due to absorption by the subject H, and the more the scattered ray component h1(X) becomes relative to the incident radiation. Such properties of the subject H, namely the amount of transmission and scattering of the subject H relative to radiation with a specific energy, can be determined in advance through experiments before radiography.

[0305] Therefore, the first derivation unit 97 possesses characteristics related to its transmission and scattering amounts (hereinafter referred to as subject scattering characteristics) in the form of functions or tables, for example, for each subject H or for each photographic part of the subject H. Then, by using the energy of the radiation used in the photograph and the estimated actual thickness of the subject H, the transmission and scattering amounts of the radiation are calculated to estimate the primary ray component g1(X) and the scattered ray component h1(X) after transmission through the subject H.

[0306] The export result (hereinafter referred to as the first export result) output by the first export unit 97 is the primary ray component g1(X) at position P1 after passing through the subject H, the scattered ray component h1(X) at position P1 after passing through the subject H, or the intensity distribution f1(X) of the radiation at position P1 after passing through the subject H. The intensity distribution f1(X) of the radiation at position P1 is, for example, the sum or weighted sum of the primary ray component g1(X) and the scattered ray component h1(X). In the seventh embodiment, the first export unit 97 outputs the intensity distribution f1(X) of the radiation at position P1 after passing through the subject H as the first export result, for example, in the form of an image or in the form of a collection of data capable of constructing an image. Alternatively, the first export unit 97 can also output either the primary ray component g1(X) or the scattered ray component h1(X) after passing through the subject H as the export result.

[0307] The second export unit 98 uses the export results from the first export unit 97 and the scattering characteristics of objects such as the top plate 4A and grid 8, which transmit radiation further after it has passed through the subject H, to estimate the component of the radiation that has been transmitted through the object. In the export process of the second export unit 98 (hereinafter referred to as the second export process), "transmitted through the object" means that the subject H transmits radiation through a certain position and then through the object. Therefore, depending on the specific shape of the subject H, it includes the case where the radiation does not transmit through the subject H but directly transmits through the object.

[0308] Specifically, the second derivation section 98 presumes the component of radiation that is transmitted through the subject H and the object, or the component of radiation that is scattered by at least one of the subject H or the object. The component of radiation that is transmitted through the subject H and the object refers to the primary ray component after transmission through the object. The component of radiation that is scattered by at least one of the subject H or the object refers to the scattered ray component after transmission through the object.

[0309] The scattering characteristics of an object define the distribution of the amount of radiation transmitted through the object and / or the amount of radiation scattered by the object. In the seventh embodiment, this is a scattering characteristic f2(X) that includes a first characteristic g2(X) defining the distribution of the amount of radiation transmitted through the object and a second characteristic h2(X) defining the distribution of the amount of radiation scattered by the object. Specifically, the scattering characteristic f2(X) is the sum or weighted sum of the first characteristic g2(X) and the second characteristic h2(X), for example, f2(X) = g2(X) + h2(X).

[0310] The first characteristic g2(X) is a function or table that determines the transmitted dose of radiation incident directly onto an object from any position X without passing through the subject H. Similarly, the second characteristic h2(X) is a function or table that determines the transmitted dose of radiation incident directly onto an object from any position X without passing through the subject H. For example, when the object is simply the top plate 4A of the radiography stage 4, the first characteristic g2(X) determines the distribution of transmitted radiation from the top plate 4A, and the second characteristic h2(X) determines the distribution of scattered radiation from the top plate 4A. The specific structure of the object (whether the radiography stage 4 is used or not, etc.) is known before radiography. Therefore, the first characteristic g2(X) and the second characteristic h2(X) can be determined in advance, for example, by experimentation, for each specific structure of the object or each combination of objects. Furthermore, if it is assumed that an object generates primary ray components and scattered ray components from incident radiation, then the first characteristic g2(X) is the operator that generates the primary ray components corresponding to the incident radiation, and the second characteristic h2(X) is the operator that generates the scattered ray components corresponding to the incident radiation.

[0311] In the seventh embodiment, the second derivation unit 98 pre-possesses, for example, the first characteristic g2(X) and the second characteristic h2(X) according to each specific structure of the object. As a result, the second derivation unit 98 pre-possesses the scattering characteristic f2(X) of the object. However, the second derivation unit 98 can acquire the first characteristic g2(X), the second characteristic h2(X), and / or the scattering characteristic f2(X) as needed.

[0312] like Figure 41 As shown, the intensity distribution of radiation (X) incident towards an arbitrary point X0 after passing through an object can be approximated using a point spread function (PSF) 120. The PSF 120 is, for example, a Gaussian function. Furthermore, the component of the radiation (X) incident towards the object that reaches the arbitrary point X0 and its vicinity is the distribution 121 of the primary ray component, and the portion of the PSF 120 after removing this primary ray component distribution 121 is the distribution 122 of the scattered ray component.

[0313] Furthermore, since the energy of the radiation used in the imaging, as well as the material (density, etc.) and thickness (mass) of the top plate 4A, etc., which is the object, are known, the peak height and half-width, and the specific shape of the PSF120 are predetermined. Therefore, for example, by performing the aforementioned deconvolution operation of the scattered ray component distribution 122 on the radiation image obtained without the subject H in the imaging, the second characteristic h2(X) can be predetermined. Moreover, by subtracting the second characteristic h2(X) from the radiation image obtained without the subject H in the imaging, or by performing the deconvolution operation of the primary ray component distribution 121, the first characteristic g2(X) can be predetermined.

[0314] The second derivation unit 98 estimates the component of radiation transmitted through the object by applying the scattering characteristics of the object to the first derivation result of the first derivation unit 97, i.e., the first derivation result. Specifically, the object receives incident radiation distributed as represented by the first derivation result. Therefore, the second derivation unit 98 estimates the component of radiation transmitted through the object by using the independent variable of the object's scattering characteristics f2(X) as the first derivation result (f1(X)). That is, the second derivation unit 98 estimates the component of radiation transmitted through the object by performing a calculation based on the following equation (25). In the seventh embodiment, the first derivation result is f1(X) = g1(X) + h1(X), so the above equation (25) can be expressed as in the following equation (26) and can be expanded as in equation (27).

[0315] f2(f1(X))=g2(f1(X))+h2(f1(X)) (25)

[0316] f2(f1(X))=g2(g1(X)+h1(X))+h2(g1(X)+h1(X)) (26)

[0317] f2(f1(X))=g2g1(X)+g2h1(X)+h2g1(X)+h2h1(X) (27)

[0318] like Figure 42 As shown, the first term on the right side of the above equation (27), "g2g1(X)", represents the radiation used in the photograph that transmits to the subject H and the object (in) Figure 42 The diagram shows the ray Ra1 that reaches pixel P(X0) at any point X0 from the top plate 4A. The second term "g2h1(X)" on the right side of equation (27) represents the ray Ra2 that is scattered by the scatterer D1 contained in the subject H and then passes through the object to reach pixel P(X0) at any point X0. The third term "h2g1(X)" on the right side of equation (27) represents the ray Ra3 that is transmitted through the subject H and then scattered by the scatterer D3 contained in the object to reach pixel P(X0) at any point X0. Furthermore, the fourth term "h2h1(X)" on the right side of equation (27) represents the ray Ra4 that is scattered by the scatterer D2 contained in the subject H and then further scattered by the scatterer D4 contained in the object to reach pixel P(X0) at any point X0.

[0319] Based on the above, the second derivation unit 98 calculates the first term "g2g1(X)" and / or the sum of the second to fourth terms "g2h1(X)+h2gl(X)+h2h1(X)" of equation (27). This is because the first term "g2g1(X)" of equation (27) represents the distribution of the primary ray component after passing through the object, and the sum of the second to fourth terms "g2h1(X)+h2g1(X)+h2h1(X)" represents the distribution of the scattered ray component after passing through the object. In the seventh embodiment, the second derivation unit 98 calculates the distribution g2g1(X) of the primary ray component after passing through the object and outputs it as the derivation result.

[0320] The image generation unit 99 uses the export result of the second export unit 98 to generate a processed radiation image of the subject H formed by radiation transmitted through the subject H and the object. When the second export unit 98 estimates that the radiation transmitted through the primary ray component of the subject H and the object, the image generation unit 99 generates the processed radiation image by imagerizing the export result of the second export unit 98, i.e., the second export result. Furthermore, when the second export unit 98 estimates that the subject H or the object scattered the scattered ray component of the radiation, the image generation unit 99 generates the processed radiation image by subtracting the export result of the second export unit 98, i.e., the second export result, from the first radiation image G1.

[0321] In the seventh embodiment, the second output unit 98 outputs the distribution of the primary radiation component after passing through the object, and the image generation unit 99 generates a processed radiation image by imagerizing it. Therefore, the distribution g2g1(X) of the primary radiation component output by the second output unit 98 is essentially a processed radiation image with the scattered radiation component removed. Furthermore, the image generation unit 99 can perform various image processing operations (e.g., contrast adjustment processing or structure emphasis processing) on ​​the generated processed radiation image as needed.

[0322] Alternatively, the scattered radiation removal process based on the sixth and seventh embodiments can be performed in the first to fifth embodiments described above. In this case, the information export devices 50A to 50D replace the scattered radiation removal unit 62 and have a scattered radiation removal unit 62A or a scattered radiation removal unit 62B.

[0323] Next, the eighth embodiment of the present invention will be described. Figure 43 This is a diagram illustrating the functional structure of the information export device based on the eighth embodiment. Additionally, in Figure 43 In the middle, to and Figure 7 The same structures are assigned the same reference numerals, and detailed descriptions are omitted. In the seventh embodiment of the present invention, grain suppression processing is performed on the first radiation image G1 and the second radiation image G2, and energy subtraction processing is performed on the first radiation image G1 and the second radiation image G2 after grain suppression processing. Therefore, as Figure 43 As shown, the information export device 50E based on the eighth embodiment further includes a content export unit 86 and a particle suppression processing unit 87 compared to the information export device 50 based on the first embodiment.

[0324] The processing content export unit 86 exports the processing content of the first grain suppression processing for the first radiation image G1, and exports the processing content of the second grain suppression processing for the second radiation image G2 based on the processing content of the first grain suppression processing.

[0325] In this embodiment, a first radiation image G1 and a second radiation image G2 are obtained by photographing the subject H using a single irradiation method. In the single irradiation method, radiation transmitted through the subject H irradiates two radiation detectors 5 and 6 that overlap, separated by a radiation energy conversion filter 7. Therefore, the second radiation detector 6, located further away from the radiation source 3, receives a smaller radiation dose compared to the first radiation detector 5, which is located closer to the radiation source 3. As a result, the second radiation image G2 has more quantum noise and a lower signal-to-noise ratio (S / N) compared to the first radiation image G1. Therefore, especially for the second radiation image G2, a particle suppression process is needed to suppress the granularity caused by quantum noise.

[0326] The processing content export unit 86 exports processing content for a first grain suppression process for the first radiation image G1, which has a high S / N ratio in both the first radiation image G1 and the second radiation image G2. Then, based on the processing content of the first grain suppression process, the processing content export unit 86 exports processing content for a second grain suppression process for the second radiation image G2. The export of the processing content will be explained below.

[0327] As a graininess suppression process, filtering based on smoothing filters such as Gaussian filters can be cited. These smoothing filters have a predetermined size of 3×3 or 5×5 centered on the pixel of interest. However, if a Gaussian filter is used, the edges of structures contained in the first radiation image G1 and the second radiation image G2 may be blurred. Therefore, in the eighth embodiment, a graininess suppression process is performed using an edge-preserving smoothing filter that prevents edge blurring and suppresses graininess. As an edge-preserving smoothing filter, a bilateral filter is used, which, for pixels adjacent to the pixel of interest, uses a normal distribution as a weight, where the weight decreases as the pixel moves further away from the pixel of interest, and the weight decreases as the difference between the pixel value and the pixel of interest increases.

[0328] Figure 44 This is a diagram illustrating an example of a bilateral filter for the first radiation image G1. Additionally, Figure 44 The image shows a 5×5 pixel local region A1 near the edge of one of the two first-radio images G1, side by side. Figure 44 The two local regions A1 shown are identical, but the positions of the pixels of interest are different. Figure 44 In the local region A1 on the left, the low-density pixels that are in contact with the edge boundary are the pixels of interest P11. As mentioned above, for pixels adjacent to the pixels of interest, the bilateral filter uses a normal distribution as the weighting, where the weight decreases as the pixel is further away from the pixels of interest and the weight decreases as the difference between the pixel value and the pixel of interest increases.

[0329] Therefore, the content export unit 86 determines the filter size of the bilateral filter based on the difference between the pixel value of the pixel of interest and the pixel values ​​of the pixels surrounding the pixel of interest. For example, the smaller the difference between the pixel value of the pixel of interest and the pixel values ​​of the pixels surrounding the pixel of interest, the larger the filter size is. Furthermore, the content export unit 86 determines the weights of the bilateral filter based on the difference between the pixel value of the pixel of interest and the pixel values ​​of the pixels surrounding the pixel of interest. For example, the smaller the difference between the pixel value of the pixel of interest and the pixel values ​​of the pixels surrounding the pixel of interest, the larger the weight is set for pixels closer to the pixel of interest compared to pixels farther away from the pixel of interest.

[0330] Therefore, the content export section processes 86 pairs of data. Figure 44 The left-side local region A1, the focus pixel P11, is exported with the following characteristics: Figure 44 The weighted 3×3 bilateral filter F11 shown on the left is used as the processing content for the first granular suppression process.

[0331] exist Figure 44 In the local region A1 on the right, the high-density pixels that are in contact with the edge boundary are the pixels of interest P12. Therefore, the content exporting unit 86 processes the content. Figure 44 The local region A1 on the right side of the focus pixel P12 is exported with the following characteristics: Figure 44 The weighted 3×3 bilateral filter F12 shown on the right is used as the processing content for the first granular suppression process.

[0332] Figure 45 It means and Figure 44 The diagram shows a local region A1 in the first radiation image, corresponding to a local region A2 in the second radiation image. Additionally, Figure 45 The diagram shows a 5×5 pixel local region A2 near an edge contained in the second radiographic image G2. Local region A2 is... Figure 44 The region at the same location as local region A1 in the first radiographic image G1 shown. Figure 45 The two local regions A2 shown are identical, but the positions of the pixels of interest are different. Figure 45 In the local region A2 on the left, with Figure 44 The pixel corresponding to the pixel of interest P11 shown in the local area A1 on the left is the pixel of interest P21. Figure 45 In the local region A2 on the right, with Figure 44 The pixel corresponding to the pixel of interest P12 shown in the local area A1 on the right is the pixel of interest P22.

[0333] Here, the radiation dose of the second radiation detector 6 acquiring the second radiation image G2 is lower than that of the first radiation detector 5 acquiring the first radiation image G1. Therefore, compared to the first radiation image G1, the second radiation image G2 has more quantum noise and poor graininess, resulting in less distinct edges. Furthermore, due to the influence of quantum noise, high-concentration regions near edge boundaries contain low-concentration pixels, or vice versa. Therefore, based on the second radiation image G2, it is impossible to properly determine a bilateral filter that preserves edges and suppresses graininess as effectively as in the first radiation image G1.

[0334] In this case, a smoothing filter such as a Gaussian filter is considered. However, if such a smoothing filter is used, it is impossible to simultaneously suppress noise and preserve edges. As a result, the edges are buried by noise, making it impossible to recover the edges of the structures contained in the second radiographic image G2.

[0335] Therefore, in the eighth embodiment, the processing content exporting unit 86 exports the processing content for the second grain suppression processing of the second radiation image based on the processing content of the first grain suppression processing of the first radiation image G1. That is, the processing content exporting unit 86 exports the processing content for the second grain suppression processing in the same manner as the processing content of the second grain suppression processing performed on each pixel of the second radiation image G2 and the processing content of the first grain suppression processing performed on the pixels of the first radiation image G1 corresponding to each pixel of the second radiation image G2. Specifically, the processing content exporting unit 86 exports a bilateral filter having the same size and weight as the bilateral filter determined for each pixel of the first radiation image G1 as the processing content for the second grain suppression processing of the second radiation image G2.

[0336] Figure 46 This is a diagram illustrating an example of a bilateral filter for the second radiation image G2. Additionally, in Figure 46 In, with Figure 44 Similarly, a 5×5 pixel local region A2 near the edge is shown in the second radiographic image G2. Figure 46 As shown, for the pixel of interest P21 in the local region A2 of the second radiation image G2, the processing content exporting unit 86 exports a bilateral filter F21 with the same size and weight as the bilateral filter F11 exported for the pixel of interest P11 in the local region A1 of the first radiation image G1, as the processing content for the second grain suppression processing.

[0337] Furthermore, for the pixel of interest P22 in the local region A2 of the second radiation image G2, the processing content exporting unit 86 exports a bilateral filter F22 with the same size and weight as the bilateral filter F12 exported for the pixel of interest P12 in the local region A1 of the first radiation image G1, as the processing content for the second grain suppression processing.

[0338] Furthermore, when exporting the processing content of the second granular suppression process, it is necessary to establish a correspondence between the pixel positions of the first radiation image G1 and the second radiation image G2. Therefore, it is preferable to align the first radiation image G1 with the second radiation image G2.

[0339] The graininess suppression processing unit 87 performs graininess suppression processing on the first radiation image G1 and the second radiation image G2. That is, it performs graininess suppression processing on the first radiation image G1 and the second radiation image G2 according to the processing content derived by the processing content derivation unit 86. Specifically, it performs filtering processing on the first radiation image G1 based on a bilateral filter derived from the first radiation image G1. Furthermore, it performs filtering processing on the second radiation image G2 based on a bilateral filter derived from the first radiation image G1.

[0340] In the eighth embodiment, the subtraction unit 63 uses the first radiographic image G1 and the second radiographic image G2, which have undergone grain suppression processing, to derive the bone image Gb using the above formula (1). Then, the bone density deriving unit 64 derives the bone density B and a representative value of the bone density B in the target bone.

[0341] Next, the ninth embodiment of the present invention will be described. Figure 47 This is a diagram illustrating the functional structure of the information export device based on the ninth embodiment. Additionally, in Figure 47 In the middle, to and Figure 43 The same structures are given the same reference numerals, and detailed descriptions are omitted. The information export apparatus 50F based on the ninth embodiment of the present invention, compared to the information export apparatus 50E based on the eighth embodiment, further includes a mapping map export unit 88 that exports a physical quantity mapping map of the subject H based on at least one of the first radiation image G1 and the second radiation image G2. The difference from the eighth embodiment is that, in the processing content export unit 86, processing content for the second grain suppression processing of the second radiation image G2 is exported based on the physical quantity mapping map.

[0342] The mapping diagram export unit 88 exports a physical quantity mapping diagram of the subject H. Examples of physical quantities include the body thickness and bone density of the subject H. The body thickness can be obtained using the body thickness distribution when the scattering removal unit 62 satisfies the termination conditions. The bone density can be obtained using the bone density export unit 64.

[0343] Here, the contrast of the structures contained in the first radiographic image G1 and the second radiographic image G2, which are the targets of granular suppression processing, varies depending on the photographic conditions. Therefore, when using a bilateral filter for edge-preserving smoothing, it is necessary to control the intensity of the edges to be preserved according to the photographic conditions. On the other hand, in the thickness map representing the thickness of the subject H, the contrast of the structures contained within the map is represented by a thickness (mm) independent of the photographic conditions.

[0344] Therefore, in the ninth embodiment, the processing content exporting unit 86 exports the processing content for the first grain suppression processing of the first radiation image G1 based on the physical quantity mapping diagram. Figure 48 This is a diagram illustrating an example of a bilateral filter for a mapping of physical quantities. Additionally, Figure 48 The image shows a 5×5 pixel local region A3 near the edge of the two physical quantity mapping maps side by side. Figure 48 The two local regions A3 shown are identical, but the positions of the pixels of interest are different. Figure 48 The local area A3 on the left side, with high-density pixels on the edge, becomes the pixel of interest P31. Therefore, for Figure 48 The focus pixel P31 of the local area A3 on the left side is exported to have the following characteristics: Figure 48 The weighted 3×3 bilateral filter F31 shown on the left is used as the processing content for the first granular suppression process.

[0345] exist Figure 48 The local area A3 on the right side, with low-density pixels in contact with the edge boundary, is the pixel of interest P32. Therefore, for Figure 48 The local area A3 on the right side, focusing on pixel P32, is exported with the following characteristics: Figure 48 The weighted 3×3 bilateral filter F32 shown on the right is used as the processing content for the first granular suppression process.

[0346] In the ninth embodiment, the processing content export unit 86 also exports a bilateral filter identical to the bilateral filter determined for each pixel of the first radiation image G1 as the processing content for the second grain suppression processing of the second radiation image G2. That is, in the second radiation image G2, in the local region corresponding to the local region A3 of the first radiation image G1, a bilateral filter with the same size and weight as the bilateral filter F31 is exported for the pixel corresponding to the pixel of interest P31 in the local region A3 as the processing content for the second grain suppression processing. Furthermore, in the second radiation image G2, a bilateral filter with the same size and weight as the bilateral filter F32 is exported for the pixel corresponding to the pixel of interest P32 in the local region A3 of the first radiation image G1 as the processing content for the second grain suppression processing.

[0347] Then, in the ninth embodiment, the grain suppression processing unit 87 performs grain suppression processing on the first radiographic image G1 and the second radiographic image G2 according to the processing content exported by the processing content export unit 86. The subtraction unit 63 uses the first radiographic image G1 and the second radiographic image G2 that have undergone grain suppression processing to export the bone image Gb using the above formula (1). Then, the bone density export unit 64 exports the bone density B and a representative value of the bone density B in the target bone.

[0348] Alternatively, the particle suppression treatment performed in the above-described 8th and 9th embodiments can also be performed in the above-described 1st to 7th embodiments.

[0349] Furthermore, in the above embodiments, the bone mineral density of the simple radiographic image G0 was estimated as information related to bone mineral density, but this is not a limitation. For example, a fracture risk assessment value can also be derived as an estimation result related to bone mineral density. Hereinafter, this will be described as the tenth embodiment. The fracture risk assessment value can be derived, for example, using the method described in International Publication No. 2020 / 166561.

[0350] Figure 49 This is a diagram illustrating the functional structure of the information export device based on the tenth embodiment. Additionally, in Figure 49 In the middle, to and Figure 7 The same structures are assigned the same reference symbols, and detailed descriptions are omitted. In the 10th embodiment of the invention, the fracture risk assessment value is derived instead of the bone mineral density derived as the positive solution data 42. Therefore, as Figure 49 As shown, the information export device 50G based on the 10th embodiment, compared with the information export device 50 based on the 1st embodiment, further includes a muscle mass export unit 110, a statistical value export unit 111, and an evaluation value export unit 112.

[0351] In addition, the subtraction unit 63 of the information export device 50G based on the tenth embodiment exports not only the bone image Gb, but also a soft tissue image Gs from which the soft tissue of the subject H was extracted from the first radiographic image G1 and the second radiographic image G2. The soft tissue image Gs is derived by the following equation (28). In equation (25), β is a weighting coefficient used to extract the soft tissue.

[0352] Gs(x,y)=G1(x,y)-β×G2(x,y) (28)

[0353] The muscle mass extraction unit 110 extracts muscle mass based on the pixel value of each pixel in the soft tissue region of the soft tissue image Gs. Soft tissue includes muscle tissue, adipose tissue, blood, and water. In the muscle mass extraction unit 110 of the tenth embodiment, tissue other than adipose tissue in the soft tissue is considered as muscle tissue. That is, in the muscle mass extraction unit 110 of the tenth embodiment, non-adipose tissue, including blood and water, is treated as muscle tissue.

[0354] The muscle volume extraction unit 110 separates muscle and fat from the soft tissue image Gs by utilizing the differences in energy properties between muscle and adipose tissue. For example... Figure 50 As shown, the amount of radiation transmitted through subject H is reduced compared to the radiation incident before it reaches the human body. Furthermore, muscle tissue and adipose tissue absorb different amounts of energy and have different attenuation coefficients; therefore, the energy spectra of radiation transmitted through muscle tissue and adipose tissue differ. For example... Figure 50 As shown, the energy spectrum of the radiation emitted by the subject H and irradiated by the first radiation detector 5 and the second radiation detector 6 depends on the body composition of the subject H, specifically, on the ratio of muscle tissue to adipose tissue. Adipose tissue transmits radiation more readily than muscle tissue; therefore, when the proportion of muscle tissue is higher than that of adipose tissue, the dose of radiation transmitted through the body is lower.

[0355] Therefore, the muscle mass extraction unit 110 separates muscle and fat from the soft tissue image Gs by utilizing the difference in energy characteristics between the muscle tissue and adipose tissue. That is, the muscle mass extraction unit 110 generates a muscle image from the soft tissue image Gs. Furthermore, the muscle mass extraction unit 110 extracts the muscle mass of each pixel based on the pixel values ​​of the muscle image.

[0356] Furthermore, the specific method by which the muscle mass extraction unit 110 separates muscle and fat from the soft tissue image Gs is not limited. As an example, the muscle mass extraction unit 110 of the tenth embodiment generates a muscle image from the soft tissue image Gs using the following formulas (29) and (30). Specifically, firstly, the muscle mass extraction unit 110 extracts the muscle rate rm(x,y) at each pixel position (x,y) within the soft tissue image Gs using formula (29). In addition, μm in formula (29) is a weighting coefficient corresponding to the attenuation coefficient of muscle tissue, and μf is a weighting coefficient corresponding to the attenuation coefficient of adipose tissue. Furthermore, Δ(x,y) represents the concentration difference distribution. The concentration difference distribution is the distribution of concentration changes observed from radiation that has not penetrated the subject H and has reached the first radiation detector 5 and the second radiation detector 6 on the image. The distribution of concentration changes in the image is calculated by subtracting the concentration of each pixel in the region of the subject H from the concentration in the unexposed region obtained by directly irradiating the first radiation detector 5 and the second radiation detector 6 with radiation from the soft image Gs.

[0357] rm(x,y)={μf-Δ(x,y) / T(x,y)} / (μf-μm) (29)

[0358] Furthermore, the muscle volume extraction unit 110 generates a muscle image Gm from the soft tissue image Gs using the following formula (30). In addition, x and y in formula (30) are the coordinates of each pixel in the muscle image Gm.

[0359] Gm(x,y)=rm(x,y)×Gs(x,y) (30)

[0360] Furthermore, as shown in equation (31) below, the muscle mass derivation unit 110 derives the muscle mass M(x,y) (g / cm²) of each pixel in the muscle image Gm by multiplying each pixel (x,y) of the muscle image Gm by a pre-defined coefficient C1(x,y) representing the relationship between pixel value and muscle mass. 2 ).

[0361] M(x,y)=C1(x,y)×Gm(x,y) (31)

[0362] The statistical value derivation unit 111 calculates a statistical value related to the subject based on the bone density and muscle mass derived by the bone density derivation unit 64. As described later, the statistical value is used in the calculation of the fracture risk assessment value used to evaluate fracture risk. Specifically, as shown in the following formula (32), the statistical value derivation unit 111 derives a statistical value Q based on the bone density distribution index value Bd related to the spatial distribution of bone density and the muscle mass distribution index value Md related to the spatial distribution of muscle mass.

[0363] Q = W1 × Bd + W2 × Md (32)

[0364] In equation (32), W1 and W2 are weighting coefficients, which are obtained by collecting a large number of bone density distribution index values ​​and muscle mass distribution index values ​​and following the rules of regression analysis.

[0365] Bone mineral density (BMD) distribution metrics indicate how bone density values ​​are distributed. Examples of BMD distribution metrics include the value per unit area, average, median, maximum, and minimum values. Muscle mass distribution metrics indicate how muscle mass values ​​are distributed. Examples of muscle mass distribution metrics include the average, median, maximum, and minimum values.

[0366] Furthermore, in addition to bone density and muscle mass, the statistical value derivation unit 111 can also calculate the statistical value Q based on at least one of the subject's height, weight, age, and fracture history. For example, when calculating the statistical value based on bone density, muscle mass, and age, the statistical value Q is calculated using the following formula (33) based on the bone density distribution index value Bd, the muscle mass distribution index value Md, and the age Y.

[0367] Q = W1 × Bd + W2 × Md + W3 × Y (33)

[0368] In Equation (33), W1, W2, and W3 are weighting coefficients. A large amount of data on bone density distribution index, muscle mass distribution index, and age-related data corresponding to these indexes are collected, and the weighting coefficients W1, W2, and W3 are determined according to regression analysis based on these data. In addition, when the subject's height, weight, and fracture history are included in addition to age to obtain statistical values, it is preferable to multiply the weighting coefficients and then add them together.

[0369] When using equation (33), if the weighting coefficients obtained through regression analysis based on a large amount of data are "W1 = 2.0, W2 = 0.1, W3 = -0.01", and the bone mineral density distribution index value Rd of a certain subject H is "1.0 g / cm³", then... 2 If the muscle mass distribution index is “20kg” and the age Y is “40 years old”, then using the following formula (34), the statistical value Q becomes “3.6”.

[0370] Q=2.0×1.0+0.1×20+(-0.01)×40=3.6 (34)

[0371] The evaluation value derivation unit 112 calculates a fracture risk assessment value for evaluating the fracture risk of the subject H based on the statistical value Q. The relationship between the statistical value Q and the fracture risk assessment value is obtained based on a large amount of diagnostic data, therefore, the evaluation value derivation unit 112 uses this relationship to calculate the fracture risk assessment value. The relationship between the statistical value Q and the fracture risk assessment value can be pre-derived and stored in the memory 53 in the form of a table.

[0372] For example, as a fracture risk assessment value, there is the probability of fracture incidence within 10 years from the time of diagnosis of subject H (when the first radiographic image G1 and the second radiographic image G2 were acquired). Then, as mentioned above, when equation (33) is used in the calculation of the statistical value Q and the weighting coefficients obtained by regression analysis based on a large amount of data are "W1 = 2.0, W2 = 0.1, W3 = -0.01", such as Figure 51 As shown, the relationship between "the probability of fracture within 10 years" and "statistical value Q" indicates that the larger the statistical value Q becomes, the lower the probability of fracture.

[0373] In the 10th embodiment, the fracture risk assessment value exported by the information export device 50G is used as the correct answer data for the teacher data. Figure 52 This is a graph representing the teacher data exported in the 10th embodiment. For example... Figure 52 As shown, the teacher data 40A consists of learning data 41 containing the first radiographic image G1 and the second radiographic image G2, and the positive solution data 42A of the fracture risk assessment value.

[0374] By using Figure 52 The teacher data 40A shown is used to learn a neural network that is able to construct a learned neural network 23A that outputs a fracture risk assessment value as an inference result related to bone density when inputting a simple radiographic image G0.

[0375] Next, the 11th embodiment of the present invention will be described. Figure 53 This is a diagram illustrating the functional structure of the information export device based on the 11th embodiment. Additionally, in Figure 53 In the middle, to and Figure 7 The same structure is assigned the same reference numerals, and detailed descriptions are omitted. In the 11th embodiment of the invention, information representing the healing state of the bone after treatment can also be derived instead of deriving bone density as the positive solution data 42. Therefore, as Figure 53 As shown, the information export device 50H based on the 11th embodiment, compared to the information export device 50 based on the 1st embodiment, further includes a healing information export unit 114. Furthermore, in the 11th embodiment, as a treatment for the bone, it is assumed that a surgery is performed to implant an artificial bone or other artificial material into the bone.

[0376] The healing information export unit 114 exports information indicating the state of the subject's bone after the artificial bone or other artificial material is implanted into the bone of the subject H, based on the bone density around the implanted artificial bone or other artificial material. The artificial bone or other artificial material replaces bone lost due to comminuted fractures or tumors, and is therefore surgically implanted into the living body.

[0377] Figure 54 This is an example of artificial bone implanted in the bone portion of a subject. Figure 54 The illustration shows the bone of a subject H who has undergone artificial hip arthroplasty, in which a femoral stem 131 of an artificial joint is implanted in the femur 130 of the subject H.

[0378] Methods for fixing the femoral stem 131 include direct fixation (cementless fixation) and indirect fixation (cement fixation). In the direct fixation method, the femoral stem 131 is inserted into the cavity inside the femur 130 without the use of cement. The shape of the cavity inside the femur 130 is pre-shaped to match the femoral stem 131. The surface of the femoral stem 131 is roughened to allow bone tissue to gradually grow towards the interior of the femoral stem 131 through infiltration. That is, immediately after the femoral stem 131 is implanted into the femur 130, a cavity exists between the femoral stem 131 and the femur 130, but as the femur 130 recovers, the cavity shrinks and disappears with the growth of bone tissue. Therefore, by obtaining the bone density around the femoral stem 131, the degree of recovery of the femur 130 after surgery can be assessed.

[0379] Figure 55 This is a chart illustrating the relationship between the distance from the femoral stem 131 inside the femur 130 and bone density at various postoperative stages. Figure 55 The horizontal axis of the chart shown is along Figure 54 The position of line L in the diagram. Figure 55 In the diagram, solid lines correspond to the initial stage immediately after the femoral stem 131 is implanted into the femoral head 130, dotted lines correspond to the mid-recovery stage, and dashed lines correspond to the full recovery stage. For example... Figure 55 As shown, in the initial postoperative stage, the femur 130 and femoral stem 131 are not tightly adhered, and the bone density near the femoral stem 131 is extremely low. With recovery, bone tissue gradually grows inward towards the femoral stem 131 via infiltration, thereby increasing the bone density near the femoral stem 131. On the other hand, the bone density distal to the femoral stem 131 remains approximately constant throughout the postoperative stages. During the healing phase, the bone density near the femoral stem 131 becomes approximately equal to that at the distal location.

[0380] The following describes the method by which the Healing Information Export Department 114 exports healing information. Figure 54The case of total hip arthroplasty shown is used as an example for illustration. The healing information export unit 114 exports information from a position L relatively close to the femoral stem 131. A Bone mineral density at the site B A X, a position relatively far from the femoral stem 131 B Bone mineral density at the site B B The difference in bone density is used as the corresponding numerical value ΔB as healing information. For example, the healing information derivation unit 114 can derive the difference in bone density (ΔB = B). B -B A This is used as healing information. In this case, the value derived as healing information decreases as recovery progresses, gradually approaching 0. Furthermore, the healing information deriving unit 114 can also derive the ratio of bone mineral density (ΔB = B). A / B B This is used as healing information. In this case, the value ΔB derived as healing information increases as the bone recovers, gradually approaching 1. That is, it is related to bone mineral density B. A With B B The numerical value ΔB corresponding to the difference can be considered a value representing the degree of bone recovery after surgery. Therefore, by deriving the numerical value ΔB as healing information, the degree of recovery of the femur 130 after surgery can be quantitatively determined.

[0381] Furthermore, the healing information export unit 114 uses the bone density B of each pixel in the bone image Gb exported by the bone density export unit 64 to export healing information. In any of the first radiographic image G1, the second radiographic image G2, and the bone image Gb, the pixel value of the femoral stem 131 is significantly different from the pixel value in the bone region, thus the region where the femoral stem 131 is located can be determined in each of the aforementioned images. Therefore, the healing information export unit 114 can determine the distance from the femoral stem 131 based on any of the first radiographic image G1, the second radiographic image G2, and the bone image Gb.

[0382] Figure 56 This is a sectional view showing an example of the cross-sectional structure of a human bone. For example... Figure 56 As shown, human bone consists of cancellous bone 140 and cortical bone 141 covering the outer side of the cancellous bone 140. Cortical bone 141 is harder and denser than cancellous bone 140. Cancellous bone 140 is an aggregate of small bone columns called trabeculae that diffuse into the medullary cavity. Trabeculae have a plate-like and rod-like structure and are interconnected. The bone density of cancellous bone 140 is significantly different from that of cortical bone 141; therefore, cortical bone 141 and cancellous bone 140 can be distinguished based on the bone density B of each pixel in the bone image Gb.

[0383] When the artificial implant is placed in the cancellous bone 140, the healing information export unit 114 can determine the region of the cancellous bone 140 based on the bone density B of each pixel in the bone image Gb, and export healing information based on the bone density of the cancellous bone 140 surrounding the artificial implant. Specifically, the healing information export unit 114 can export the position X within the cancellous bone 140 that is relatively close to the artificial implant. A Bone mineral density at the site B A Location X within 140 degrees of cancellous bone at a relatively distant distance from the artificial object B Bone mineral density at the site B B The difference in value ΔB is used as the healing information.

[0384] On the other hand, when the artificial object is implanted into the cortical bone 141, the healing information exporting unit 114 preferably determines the region of the cortical bone 141 based on the bone density B of each pixel in the bone image Gb, and exports healing information based on the bone density of the cortical bone 141 surrounding the artificial object. Specifically, the healing information exporting unit 114 can export the position X within the cortical bone 141 that is relatively close to the artificial object. A Bone mineral density at the site B A Location X within the cortical bone 141, which is relatively far from the artificial object B Bone mineral density at the site B B The difference in value ΔB is used as the healing information.

[0385] Furthermore, when the artificial implanted in the bone of the subject H involves both cancellous bone 140 and cortical bone 141, the regions of cancellous bone 140 and cortical bone 141 can be determined based on the bone density B of each pixel in the bone image Gb, and healing information is derived based on the bone density of both cancellous bone 140 and cortical bone 141 surrounding the artificial. Specifically, the healing information deriving unit 114 can derive the position L within the cancellous bone 140 that is relatively close to the artificial. A1 Bone mineral density at the site B A1 Position L within 140 degrees of cancellous bone, relatively far from the artificial object B1 Bone mineral density at the site B B1 The difference in value ΔB1 is used as healing information, and the position L within the cortical bone 141, which is relatively close to the artificial object, is derived. A2 Bone mineral density at the site B A2 Position L within the cortical bone 141, which is relatively far from the artificial object B2 Bone mineral density at the site B B2The difference in bone density is used to determine the healing information, with the corresponding value ΔB2. Furthermore, when the implanted bone in the subject H involves both cancellous bone 140 and cortical bone 141, healing information can also be derived based on the bone density of either the cancellous bone 140 or the cortical bone 141 surrounding the implant. That is, one of the values ​​ΔB1 and ΔB2 can be derived as the healing information.

[0386] In the 11th embodiment, the healing information exported by the information exporting device 50H based on the 11th embodiment is used as the correct solution data for the teacher data. Figure 57 This is a graph representing the teacher data exported in embodiment 11. For example... Figure 57 As shown, the teacher data 40B consists of learning data 41 containing the first radiation image G1 and the second radiation image G2, and positive solution data 42B as the numerical value of the cure information.

[0387] By using Figure 57 The teacher data 40B shown is used to learn the neural network, which can construct a neural network 23A that outputs information representing the cure status when the input is a simple radiographic image G0.

[0388] Next, the information export apparatus based on the 12th embodiment of the present invention will be described. Furthermore, the structure of the information export apparatus based on the 12th embodiment is the same as that based on the 1st embodiment, except that the processing performed by the scattering line removal unit 62 is different; therefore, a detailed description is omitted here. Additionally, in the 12th embodiment, similar to the 6th embodiment, the method using… Figure 30 The first radiation image G1 and the second radiation image G2 obtained by the photographic device 1A shown in the figure are used as teacher data 40.

[0389] Figure 58 This is a diagram illustrating the functional structure of the scattering line removal unit in the information export apparatus based on the 12th embodiment. (See diagram for example.) Figure 58 As shown, the scattering line removal unit 62C in the information export device based on the 12th embodiment includes a region detection unit 150, a scattering line image export unit 151, a pixel value calculation unit 152, a pixel value replacement unit 153, a boundary position adjustment unit 154, and a calculation unit 155.

[0390] The area detection unit 150 obtains area detection images by detecting in the first radiation image G1 and the second radiation image G2 the area of ​​the subject where radiation passes through the subject H and reaches the first radiation detector 5 and the second radiation detector 6, and the area of ​​the direct radiation where radiation does not pass through the subject H but only passes through the top plate 4A and directly reaches the first radiation detector 5 and the second radiation detector 6.

[0391] Specifically, such as Figure 59 As shown, the region detection unit 150 detects regions in the first radiation image G1 whose pixel values ​​are less than a region threshold and designates them as subject regions 160, and detects regions in the first radiation image G1 whose pixel values ​​are greater than the region threshold and designates them as direct radiation regions 161. For example, when the region detection image is made into a binarized image with subject region 160 set to "0" and direct radiation region 161 set to "1", the following is obtained on a specific line 162 of the first radiation image G1: Figure 60 The distribution is shown. Alternatively, the region detection unit 150 can use a learned neural network that has been trained to detect both subject regions and direct radiation regions, instead of threshold-based region detection.

[0392] Furthermore, a region threshold is specified for each photographic condition, and it is preferable to use the region threshold corresponding to the photographic conditions when photographing the subject H. Moreover, the region detection processing can be performed on both the first radiation image G1 and the second radiation image G2, or on either one.

[0393] The scattered ray image extraction unit 151 obtains a scattered ray image related to the scattered ray components based on the region detection image and scattered ray diffusion information related to the diffusion of scattered rays. The calculation unit 155 obtains a radiation image with scattered ray components removed by subtracting the scattered ray image from the first radiation image G1 and the second radiation image G2. Furthermore, the scattered ray diffusion information used in the scattered ray image extraction unit 151 can, for example, utilize... Figure 41 The scattered line diffusion information is shown.

[0394] Then, in the scattering ray image exporting unit 151, a scattering ray image containing scattering ray components is obtained by performing a "region detection image * PSF (* is the convolution operator)" operation, which convolves the region detection image and the PSF. Furthermore, the scattering ray image exporting unit 151 exports scattering ray images from the first radiation image G1 and the second radiation image G2, respectively.

[0395] In the scattered ray image, the pixel values ​​of the scattered ray component in the direct ray region can be directly used from the pixel values ​​of the direct ray region 161 in the first ray image G1 and the second ray image G2. However, the pixel values ​​of the direct ray region 161 are often saturated (exceeding the pixel values ​​that the camera sensors of the first ray detector 5 and the second ray detector 6 can receive light). Therefore, as the pixel values ​​of the direct ray region in the scattered ray image, it is preferable to use unsaturated pixel values ​​that are theoretically calculated to have no saturation and take into account the influence of scattered rays, i.e., unsaturated scattered ray pixel values, and replace these theoretically calculated unsaturated scattered ray pixel values ​​with the pixel values ​​of the direct ray region in the scattered ray image.

[0396] The pixel value calculation unit 152 uses a relationship between the radiation dose and the unsaturated pixel value obtained when there is no pixel value saturation and the effect of scattered rays is taken into account, i.e., the unsaturated scattered ray pixel value, to calculate the unsaturated scattered ray pixel value corresponding to the imaging dose. Then, the pixel value replacement unit 153 replaces the pixel value of the direct radiation area with the unsaturated scattered ray pixel value. Specifically, as follows... Figure 61 As shown, in the scattering ray image exported by the scattering ray image exporting unit 151, the pixel values ​​of the direct radiation region 161 are replaced with the unsaturated scattering ray pixel values ​​calculated by the pixel value calculation unit 152.

[0397] In a direct radiation area table (not shown), the relationship between unsaturated scattered radiation pixel values ​​is predefined for each photographic condition and stored in memory 53. The pixel value calculation unit 152 refers to the direct radiation area table and uses any unsaturated scattered radiation pixel value relationship corresponding to the photographic conditions when photographing the subject H, or combines multiple unsaturated scattered radiation pixel value relationships that satisfy the photographic conditions when photographing the subject H. Furthermore, the unsaturated scattered radiation pixel value is preferably calculated, for example, using the formula: "Unsaturated scattered radiation pixel value = Unsaturated pixel value obtained when there are no pixel values ​​× Scattered radiation content of the direct radiation area".

[0398] As described above, in the region detection unit 150, a region threshold is used to distinguish between the subject region 160 and the direct radiation region 161. However, the direct radiation region 161 detected by the region threshold sometimes also includes a region that transmits light through a soft area close to the skin in the subject H. When it is assumed that scattered rays are also emitted from such soft areas, if a subtraction operation is performed on the scattered ray image from the first radiation image G1 and the second radiation image G2, the scattered ray component will be excessively removed. Therefore, it is possible to adjust the boundary position between the subject region 160 and the direct radiation region 161 by only a specific width, and obtain a scattered ray image based on the region detection image after adjusting the boundary position and the scattered ray diffusion information.

[0399] The boundary position adjustment unit 154 adjusts the boundary position of the subject area 160 and the direct radiation area 161 by only a specific width. The boundary position adjustment is preferably performed when the pixel value of the direct radiation area exceeds a pixel value threshold or when the radiation dose exceeds a dose threshold. Furthermore, the specific width is preferably defined based on at least one of the location of the subject H, the imaging method of the subject H, and the imaging conditions. Moreover, the specific width is preferably defined based on the pixel values ​​near the boundary position in the radiation image.

[0400] Specifically, when a specific width is defined based on the shape of the subject H, i.e., the thickness distribution of the subject H, in a given location, such as... Figure 62As shown in the thickness distribution of the specific line 162 portion, the portion Px with a thickness exceeding 0 in the direct radiation region 161 detected by the region using a threshold is set as the specific width. By adjusting the boundary position BP only by the amount corresponding to the specific width of this portion Px, the subject region 160 only expands by the amount of the specific width, while the direct radiation region 161 only shrinks by the amount of the specific width. Furthermore, when the specific width is defined based on the pixel values ​​near the boundary position, it is preferable to define the specific width based on the pixel distance between the pixel at the boundary position and the nearby pixels located within a specific range from the boundary position and whose pixel values ​​have a specific range.

[0401] In the 12th embodiment, the subtraction unit 63 uses the processed first radiographic image G1 and second radiographic image G2 to derive a bone image Gb. Then, the bone density deriving unit 64 uses the bone image Gb derived by the subtraction unit 63 to derive the bone density B and a representative value of the bone density B in the target bone.

[0402] Alternatively, the scattering line removal process based on the 12th embodiment can be performed in the 1st to 5th and 8th to 11th embodiments described above.

[0403] Furthermore, in the above embodiments, the learning data 41 used as teacher data 40 utilizes the first radiographic image G1 and the second radiographic image G2, but is not limited to this. For example... Figure 63 As shown in teacher data 40C, the bone image Gb can also be used as learning data 41C instead of the second radiographic image G2. In this case, the bone image Gb can be derived through any of the embodiments described in embodiments 1 to 11 above. Furthermore, as... Figure 64 As shown in teacher data 40D, a bone density image Gd, in which each pixel is a value of bone density, can also be used as learning data 41D instead of the second radiographic image G2. In this case, the bone density image Gd can use the bone density values ​​derived in any of the embodiments described in the first to twelfth embodiments as pixel values.

[0404] Furthermore, in the above embodiments, the correct answer data 42 for teacher data 40 can also be a bone density image with bone density B exported by bone density exporting unit 64 as pixel values. In this case, the estimation unit 23 of estimation device 10 exports a bone density image from simple radiographic image G0 as an estimation result related to bone density. Thus, when a bone density image is exported, the bone density image can be displayed on the display screen.

[0405] Figure 65 This is another example of a display showing the estimated result. For example... Figure 65 As shown, display screen 70A has the same... Figure 13The same image display area 71 as the display screen 70 shown. The image display area 71 displays the estimated bone density result, i.e., the bone density image Gd, from the simple radiometric image G0 of the subject H. In the bone density image Gd, patterns are applied to the bone regions based on the bone density. Additionally, in... Figure 65 For simplicity, only the femur is given a pattern indicating the amount of bone mineral. Below the image display area 71, a reference 73 indicating the size of the bone mineral content based on the given pattern is displayed. The operator can easily identify the patient's bone density by referring to the reference 73 while reading the bone mineral density image Gd. Alternatively, instead of the pattern, different colors can be assigned to the bone mineral density image Gd according to the bone density.

[0406] Furthermore, in the above embodiments, information relating to bone mineral density of the femur near the hip joint is estimated, but the bone to be considered is not limited to the femur. The technology of the present invention can also be applied when estimating information relating to bone mineral density of any bone near the knee joint, such as the femur, tibia, vertebrae such as the lumbar vertebrae, calcaneus, and metacarpals.

[0407] Furthermore, in the above embodiments, bone density, fracture risk, and healing information are used as the positive solution data included in the teacher data used to learn the neural network. Therefore, the bone density-related information estimated by the estimation unit 23 based on the simple radiograph G0 includes bone density, fracture risk, and healing information in the simple radiograph G0, but is not limited to these. The learned neural network 23A can also be constructed using YAM, T score, or Z score as positive solution data, and YAM, T score, and Z score can be estimated from the simple radiograph G0 as bone density-related information. In the estimation unit 23, the bone density-related information to be estimated can be the detection results of the presence or absence of fractures, tumors, or implants, or the determination results of osteoporosis. Furthermore, bone diseases related to bone density, such as multiple myeloma, rheumatism, arthritis, and chondrodysplasia, can also be estimated as bone density-related information. In this case, the learned neural network 23A can be constructed using teacher data containing this bone density-related information as the positive solution data.

[0408] Furthermore, in the above embodiments, when deriving bone density as the correct solution data, the first radiographic image G1 and the second radiographic image G2 themselves are used, but this is not a limitation. Alternatively, a moving average with surrounding pixels can be calculated for each pixel of the first radiographic image G1 and the second radiographic image G2, and bone density can be derived using the first radiographic image G1 and the second radiographic image G2 with the moving average as the pixel value for each pixel. Here, cortical bone becomes important information when determining bone density; therefore, to maintain a resolution that allows cortical bone identification, for example, a resolution of 2 mm or less based on the actual size of the subject, a moving average with surrounding pixels can be calculated for each pixel. In this case, the pixels used in the moving average can be appropriately determined based on information about the distances between the radiation source 3, the subject H, and the radiation detectors 5 and 6, as well as information about the pixel sizes of the radiation detectors 5 and 6.

[0409] Furthermore, in the above embodiments, a learned neural network 23A is constructed by learning a neural network in the estimation device 10, but this is not a limitation. The learned neural network 23A constructed in a device other than the estimation device 10 may also be used in the estimation unit 23 of the estimation device 10 in this embodiment.

[0410] Furthermore, in the above embodiments, when performing energy subtraction processing for deriving bone density, the first radiographic image G1 and the second radiographic image G2 are obtained using a single irradiation method, but this is not a limitation. The first radiographic image G1 and the second radiographic image G2 can also be obtained using a so-called two-irradiation method, which involves taking two images using only one radiation detector. In the case of the two-irradiation method, the position of the subject H included in the first radiographic image G1 and the second radiographic image G2 may shift due to body movement of the subject H. Therefore, in the first radiographic image G1 and the second radiographic image G2, it is preferable to perform the processing of this embodiment after subject alignment.

[0411] As a alignment process, for example, the method described in Japanese Patent Application Publication No. 2011-255060 can be used. In the method described in Japanese Patent Application Publication No. 2011-255060, multiple first-band images and multiple second-band images of structures representing different frequency bands are generated, respectively relating to the first radiation image G1 and the second radiation image G2. The position offsets of corresponding positions in the first-band images and second-band images of the corresponding frequency bands are obtained, and the alignment of the first radiation image G1 and the second radiation image G2 is performed based on the position offsets.

[0412] Furthermore, in the above embodiments, bone density was derived as the correct answer data for teacher data and information related to bone density was estimated using the radiation images acquired in the system that uses the first radiation detector 5 and the second radiation detector 6 to capture the subject H. However, the technology of the present invention can also be applied when a fluorophore is used instead of a radiation detector to acquire the first radiation image G1 and the second radiation image G2. In this case, two fluorophores are overlapped and the radiation transmitted through the subject H is irradiated, and the radiation image information of the subject H is accumulated and recorded in each fluorophore. The first radiation image G1 and the second radiation image G2 are acquired by photoelectrically reading the radiation image information from each fluorophore. In addition, when using a fluorophore to acquire the first radiation image G1 and the second radiation image G2, a two-stage irradiation method can also be used.

[0413] Furthermore, the radiation used in the above embodiments is not particularly limited; in addition to X-rays, alpha rays or gamma rays can also be used.

[0414] Furthermore, in the above embodiments, the hardware structure of the processing units (processing units) that perform various processes, such as the image acquisition unit 61, scattering removal unit 62, subtraction unit 63, and bone density extraction unit 64 of the estimation device 10 (image acquisition unit 21, information acquisition unit 22, estimation unit 23, learning unit 24, and display control unit 25) and the information export device 50, can use various processors as shown below. These various processors include, in addition to general-purpose processors (CPUs) that execute software (programs) to function as various processing units, as described above, processors such as FPGAs (Field Programmable Gate Arrays) whose circuit structure can be changed after manufacturing (Programmable Logic Devices: PLDs), and processors such as ASICs (Application Specific Integrated Circuits) with circuit structures specifically designed for performing specific processes (dedicated circuits).

[0415] A processing unit can consist of one of these various processors, or it can consist of 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). Furthermore, a single processor can also constitute multiple processing units.

[0416] As examples of a single processor comprising multiple processing units, the first approach is as follows: As exemplified by client computers and servers, a processor is constructed from a combination of one or more CPUs and software, and this processor functions as multiple processing units. The second approach is as follows: As exemplified by System-on-Chip (SoC), a processor that implements the overall system functionality including multiple processing units using a single integrated circuit (IC) chip. In this way, various processing units are constructed as hardware using one or more of the aforementioned processors.

[0417] Furthermore, as the hardware architecture of these various processors, more specifically, it is possible to use circuits composed of circuit elements such as semiconductor elements.

[0418] Symbol Explanation

[0419] 1. 1A - Imaging device; 3 - Radiation source; 4 - Imaging stage; 4A - Top plate; 4B - Mounting unit; 5, 6 - Radiation detector; 7 - Radiation energy conversion filter; 8 - Scattered radiation removal grid; 9 - Image storage system; 10 - Estimation device; 11, 51 - CPU; 12A - Estimation program; 12B - Learning program; 13, 53 - Memory; 14, 54 - Display; 15, 55 - Input device; 16, 56 - Internal memory; 17, 57 - Network I / F; 18, 58 - Bus; 21 - Image acquisition unit; 22 - Information acquisition unit; 23 - Estimation unit; 23A - Completed learning neural network; 24 - Learning unit; 25 - Display control unit; 30 - Neural network; 31 - Input layer 32-Intermediate layer, 33-Output layer, 35-Convolutional layer, 36-Pooling layer, 37-Fully connected layer, 40, 40A, 40B, 40C, 40D-Teacher data, 41, 41C, 41D-Learning data, 42, 42A, 42B, 42C-Forward solving data, 47-Output data, 48-Parameters, 50, 50A, 50B, 50C, 50D, 50E, 50G, 50H-Information export device, 52-Information export program, 61-Image acquisition unit, 62, 62A, 62B-Scattering ray removal unit, 63-Subtraction unit, 64-Bone density export unit, 65-Structure recognition unit, 66-Weight coefficient export unit, 67-Initial weight coefficient setting unit, 68-Weight coefficient export unit 70, 70A - Display screen; 71 - Image display area; 72 - Bone density display area; 73 - Reference data; 81 - Initial value derivation unit; 82 - Attenuation coefficient derivation unit; 83 - Weighting coefficient derivation unit; 84 - Composition ratio derivation unit; 85 - Attenuation coefficient setting unit; 86 - Processing content derivation unit; 87 - Granularity suppression processing unit; 88 - Mapping map derivation unit; 91 - Photographic condition acquisition unit; 92 - Body thickness derivation unit; 93 - Characteristic acquisition unit; 94 - Linear distribution derivation unit; 95 - Calculation unit; 97 - First derivation unit; 98 - Second derivation unit; 99 - Image generation unit; 101 - Phantom; 102 - Lead plate; 103 - Air layer; 110 - Muscle mass derivation unit; 111 - Statistical value derivation unit; 112 - Evaluation value export section, 114 - Healing information export section, 120 - PSF, 121 - Distribution of primary radiation components, 122 - Distribution of scattered radiation components, 130 - Femur, 131 - Femoral stem, 140 - Cancellous bone, 141 - Cortical bone, 150 - Region detection section, 151 - Scattered radiation image export section, 152 - Pixel value calculation section, 153 - Pixel value replacement section, 154 - Boundary position adjustment section, 155 - Calculation section, 160 - Subject area, 161 - Direct radiation area, 162 - Specific line, A1~A3 - Local area, BP - Boundary position, C0 - Correction coefficient, D1~D4 - Scatterer, F11, F12, F21, F22, F31, F32 - Bilateral filtersG0 - Simple radiographic image, G1, G2 - Radiographic images, Gb - Bone image, Gd - Bone density image, K1-K3 - Radiographic images, L - Distance, LUT1-LUT9 - Tables, P11, P12, P21, P22, P31, P32 - Pixels of interest, Q - Statistical value, Ra - Radiometric radiation.

Claims

1. A estimation device comprising at least one processor, The processor functions as a learned neural network that derives an inference related to the bone density of a bone by simply capturing radiographic images of a subject containing bones. The learned neural network is trained using the following data as teacher data: (i) two radiographic images of a subject containing a bone portion obtained by photographing a subject with different energy distributions; (ii) a radiographic image of the subject and a bone image representing the bone portion of the subject; or (iii) a radiographic image of the subject and a bone density image representing the bone density of the bone portion of the subject; and Information related to bone density of the skeletal portion of the subject. The information related to the bone mineral density is derived from the following: The body thickness distribution of the subject is inferred from at least one of two radiographic images obtained by photographing a subject containing bones and soft tissues using radiation with different energy distributions. The photographic conditions under which the two radiation images were acquired; and The pixel values ​​of the bone region in the extracted bone image are derived by energy subtraction processing, which is performed by weighted subtraction on the two radiographic images. The bone image is exported through the following steps: The initial values ​​of the weighting coefficients used in the weighted subtraction operation are set according to the thickness distribution of the subject. The initial weighting coefficients are used to perform the energy subtraction process to export the bone image; The thickness of the bone and the thickness of the soft tissue at each pixel in the bone image are derived based on the pixel values ​​of the bone contained in the bone image. Based on the derived thickness of the bone and the thickness of the soft tissue, new weighting coefficients are derived for the weighted subtraction operation. A new bone image is derived by performing the weighted subtraction operation on the two radiographic images using the new weighting coefficients. and Repeat the process of deriving a new weighting coefficient from the new bone image and deriving a new bone image from the new weighting coefficient until the difference between the thickness of the bone in the new bone image and the thickness of the bone in the bone image obtained in the previous processing is less than a predetermined threshold.

2. The estimation device according to claim 1, wherein, Information related to the bone density includes at least one of bone density, an assessment of the subject's fracture risk, and information indicating the healing status of the bone after treatment.

3. An estimation method using a learned neural network to derive an estimation result related to bone density from a simple radiographic image, wherein the learned neural network derives the estimation result related to the bone density of the bone by simply photographing a subject containing bones, wherein... The learned neural network is trained using the following data as teacher data: (i) two radiographic images of a subject containing a bone portion obtained by photographing a subject using radiation with different energy distributions; (ii) a radiographic image of the subject and a bone image representing the bone portion of the subject; or (iii) a radiographic image of the subject and a bone density image representing the bone density of the bone portion of the subject; and Information related to bone density of the skeletal portion of the subject. The information related to the bone mineral density is derived from the following: The body thickness distribution of the subject is inferred from at least one of two radiographic images obtained by photographing a subject containing bones and soft tissues using radiation with different energy distributions. The photographic conditions under which the two radiation images were acquired; and The pixel values ​​of the bone region in the extracted bone image are derived by energy subtraction processing, which is performed by weighted subtraction on the two radiographic images. The bone image is exported through the following steps: The initial values ​​of the weighting coefficients used in the weighted subtraction operation are set according to the thickness distribution of the subject. The initial weighting coefficients are used to perform the energy subtraction process to export the bone image; The thickness of the bone and the thickness of the soft tissue at each pixel in the bone image are derived based on the pixel values ​​of the bone contained in the bone image. Based on the derived thickness of the bone and the thickness of the soft tissue, new weighting coefficients are derived for the weighted subtraction operation. A new bone image is derived by performing the weighted subtraction operation on the two radiographic images using the new weighting coefficients. and Repeat the process of deriving a new weighting coefficient from the new bone image and deriving a new bone image from the new weighting coefficient until the difference between the thickness of the bone in the new bone image and the thickness of the bone in the bone image obtained in the previous processing is less than a predetermined threshold.

4. A computer-readable storage medium storing an estimation program that causes a computer to perform the following steps: using a learned neural network to derive an estimation result related to bone density from a simple radiographic image, wherein the learned neural network derives the estimation result related to the bone density of the bone from the simple radiographic image obtained by simply photographing a subject containing bones, wherein... The learned neural network is trained using the following data as teacher data: (i) two radiographic images of a subject containing a bone portion obtained by photographing a subject using radiation with different energy distributions; (ii) a radiographic image of the subject and a bone image representing the bone portion of the subject; or (iii) a radiographic image of the subject and a bone density image representing the bone density of the bone portion of the subject; and Information related to bone density of the skeletal portion of the subject. The information related to the bone mineral density is derived from the following: The body thickness distribution of the subject is inferred from at least one of two radiographic images obtained by photographing a subject containing bones and soft tissues using radiation with different energy distributions. The photographic conditions under which the two radiation images were acquired; and The pixel values ​​of the bone region in the extracted bone image are derived by energy subtraction processing, which is performed by weighted subtraction on the two radiographic images. The bone image is exported through the following steps: The initial values ​​of the weighting coefficients used in the weighted subtraction operation are set according to the thickness distribution of the subject. The initial weighting coefficients are used to perform the energy subtraction process to export the bone image; The thickness of the bone and the thickness of the soft tissue at each pixel in the bone image are derived based on the pixel values ​​of the bone contained in the bone image. Based on the derived thickness of the bone and the thickness of the soft tissue, new weighting coefficients are derived for the weighted subtraction operation. A new bone image is derived by performing the weighted subtraction operation on the two radiographic images using the new weighting coefficients. and Repeat the process of deriving a new weighting coefficient from the new bone image and deriving a new bone image from the new weighting coefficient until the difference between the thickness of the bone in the new bone image and the thickness of the bone in the bone image obtained in the previous processing is less than a predetermined threshold.

Images