Estimation device, method and program

Abstract

To provide an estimation device, method and program which can highly accurately predict the three-dimensional bone density.SOLUTION: A processor functions as a learned neural network which derives an estimation result related to the three-dimensional bone density from a simple radiation image acquired by simply imaging a subject including bone parts and a DXA scan image acquired by imaging the subject in a DXA method. The learned neural network is learned by using (i) the two-dimensional bone density of the bone parts included in the two radiation images or the like acquired by imaging the subject including the bone parts with the radiation with the different energy distributions and the two radiation images or the like, or (ii) the two-dimensional bone density of the bone parts included in the radiation image or the like of the subject, a bone part image expressing the bone parts of the subject, and the radiation image or the like or the bone part image and the three-dimensional bone density of the bone parts of the subject as teacher data.SELECTED DRAWING: Figure 3

Description

[Technical field]

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

[0002] DXA (Dual X-ray Absorptiometry) is one of the representative bone mineral quantification methods used to diagnose bone density in bone diseases such as osteoporosis. In DXA, radiation that enters and penetrates the human body has an attenuation coefficient μ (cm 2 / g) and its density ρ(g / cm 3 This is a method to calculate bone mineral density from the pixel values ​​of radiographic images taken with two types of radiation energy, taking advantage of the attenuation characterized by the radiation energy (cm) and thickness (t) of the bone.

[0003] Also, various methods for evaluating bone density using a radiographic image obtained by photographing a subject have been proposed. For example, Patent Document 1 proposes a method for estimating information on bone density, such as bone mineral density per unit volume, from an image showing a bone by using a trained neural network constructed by training a neural network. In the method described in Patent Document 1, a neural network is trained by using an image showing a bone obtained by plain radiography and a bone density obtained by DXA as training data.

[0004] Here, plain radiography is an imaging method in which a subject is irradiated with radiation once to obtain a single two-dimensional image that is a transmission image of the subject. In the following description, a radiation image obtained by plain radiography will be referred to as a plain radiation image. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] JP 2020-171785 A Summary of the Invention [Problem to be solved by the invention]

[0006] However, there is a demand for a method for estimating bone density with higher accuracy.

[0007] The present disclosure has been made in consideration of the above circumstances, and has an object to make it possible to estimate three-dimensional bone density with high accuracy. [Means for solving the problem]

[0008] The present disclosure provides an estimation apparatus comprising at least one processor, The processor functions as a trained neural network that derives an estimation result related to the three-dimensional bone density of a bone portion from a plain radiographic image obtained by plainly photographing a subject including a bone portion or a DXA scanned image obtained by photographing the subject using a DXA method; The trained neural network is trained using (i) two radiological images or DXA scanned images obtained by photographing a subject including a bone with radiation having different energy distributions, and two-dimensional bone density of the bone contained in the two radiological images or DXA scanned images, or (ii) a radiological image or DXA scanned image of the subject, a bone image representing the bone of the subject, and the two-dimensional bone density of the bone contained in the radiological image or DXA scanned image or bone image, and a three-dimensional bone density of the bone of the subject as training data.

[0009] In the estimation device according to the present disclosure, the three-dimensional bone density may be derived from a three-dimensional image of the subject.

[0010] In the estimation device according to the present disclosure, the three-dimensional image may be a CT image.

[0011] In addition, in the estimation device according to the present disclosure, the three-dimensional bone density may be derived by identifying a bone region in a CT image, deriving an attenuation coefficient of radiation in the bone region, and deriving the bone density at each position of the bone region based on the attenuation coefficient of radiation and the mass attenuation coefficient in the bone region.

[0012] Additionally, in the estimation apparatus according to the present disclosure, the two-dimensional bone mineral density may be derived from two radiological or DXA scan images.

[0013] In the estimation device according to the present disclosure, the two-dimensional bone mineral density is calculated based on a body thickness distribution of the subject estimated based on at least one of the two radiographic images or a DXA scanning image, The imaging conditions under which the two radiographic or DXA scans were obtained, and It may be derived based on pixel values ​​of bone regions in a bone image from which bones are extracted, the pixel values ​​being derived by energy subtraction processing that performs weighted subtraction on two radiographic images or DXA scan images.

[0014] In the estimation device according to the present disclosure, the two-dimensional bone density may be derived from a bone image.

[0015] In the estimation device according to the present disclosure, the two-dimensional bone density is calculated based on the body thickness distribution of the subject estimated based on a radiation image or a DXA scan image, The imaging conditions under which the radiographic image or DXA scan was obtained, and It may be derived based on pixel values ​​of a bone region in a bone image.

[0016] The estimation method according to the present disclosure is a method for deriving an estimation result related to a three-dimensional bone density of a bone portion from a simple radiographic image obtained by simply photographing a subject including a bone portion or a DXA scanned image obtained by photographing the subject using a DXA method, by using a trained neural network for deriving an estimation result related to a three-dimensional bone density of a bone portion, the method comprising: The trained neural network is trained using (i) two radiological images or DXA scanned images obtained by photographing a subject including a bone with radiation having different energy distributions, and two-dimensional bone density of the bone contained in the two radiological images or DXA scanned images, or (ii) a radiological image or DXA scanned image of the subject, a bone image representing the bone of the subject, and the two-dimensional bone density of the bone contained in the radiological image or DXA scanned image or bone image, and a three-dimensional bone density of the bone of the subject as training data.

[0017] The present disclosure provides an estimation program for causing a computer to execute a procedure for deriving an estimation result related to a three-dimensional bone density of a bone portion, using a trained neural network that derives an estimation result related to a three-dimensional bone density of a bone portion from a simple radiographic image obtained by simply photographing a subject including a bone portion, or a DXA scanned image obtained by photographing the subject using a DXA method, the estimation program comprising: The trained neural network is trained using (i) two radiological images or DXA scanned images obtained by photographing a subject including a bone with radiation having different energy distributions, and two-dimensional bone density of the bone contained in the two radiological images or DXA scanned images, or (ii) a radiological image or DXA scanned image of the subject, a bone image representing the bone of the subject, and the two-dimensional bone density of the bone contained in the radiological image or DXA scanned image or bone image, and a three-dimensional bone density of the bone of the subject as training data. Effect of the Invention

[0018] According to the present disclosure, three-dimensional bone density can be estimated with high accuracy. [Brief description of the drawings]

[0019] [Figure 1] FIG. 1 is a schematic block diagram showing a configuration of a radiation image capturing system to which an estimation device according to an embodiment of the present disclosure is applied. [Diagram 2] FIG. 1 is a diagram showing a schematic configuration of an estimation device according to an embodiment of the present invention. [Diagram 3] FIG. 1 is a diagram showing a functional configuration of an estimation device according to an embodiment of the present invention. [Figure 4] FIG. 1 is a diagram showing a schematic configuration of a neural network used in the present embodiment; [Diagram 5] Diagram showing training data [Figure 6] FIG. 1 is a diagram showing a schematic configuration of an information derivation device according to an embodiment of the present invention; [Figure 7] FIG. 1 is a diagram showing a functional configuration of an information derivation device according to an embodiment of the present invention; [Figure 8] Figure showing bone images [Figure 9] A diagram showing the relationship between the contrast of bones and soft tissues and the body thickness of the subject [Figure 10] FIG. 1 is a diagram showing an example of a lookup table for acquiring a correction coefficient; [Figure 11] Diagram for explaining derivation of 3D bone density [Figure 12] Diagram for explaining derivation of 3D bone density [Figure 13] Diagram to explain CT value [Figure 14] A diagram showing the relationship between radiation energy and mass attenuation coefficient [Figure 15] Diagram to explain neural network learning [Figure 16] Conceptual diagram of the processing performed by a trained neural network [Figure 17] A diagram showing the display screen of the estimation results [Figure 18] 1 is a flowchart of a learning process performed in this embodiment. [Figure 19] 1 is a flowchart of an estimation process performed in this embodiment. [Figure 20] Another example of training data [Figure 21] Another example of training data [Figure 22] FIG. 13 is a diagram showing another example of a display screen for the estimation result. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Hereinafter, an embodiment of the present disclosure will be described with reference to the drawings. Fig. 1 is a schematic block diagram showing a configuration of a radiographic image capturing system to which an estimation device according to an embodiment of the present disclosure is applied. As shown in Fig. 1, the radiographic image capturing system according to this embodiment includes an imaging device 1, a CT device 8, an image storage system 9, an estimation device 10 according to this embodiment, and an information derivation device 50. The imaging device 1, the CT (Computed Tomography) device 8, the estimation device 10, and the information derivation device 50 are connected to the image storage system 9 via a network (not shown).

[0021] The imaging device 1 is an imaging device capable of performing energy subtraction by a so-called one-shot method in which radiation such as X-rays emitted from a radiation source 3 and transmitted through a subject H is irradiated to a first radiation detector 5 and a second radiation detector 6 with different energies. During imaging, as shown in Fig. 1, the first radiation detector 5, a radiation energy converting filter 7 made of a copper plate or the like, and the second radiation detector 6 are arranged in this order from the side closest to the radiation source 3, and the radiation source 3 is driven. The first and second radiation detectors 5 and 6 are in close contact with the radiation energy converting filter 7.

[0022] As a result, the first radiation detector 5 acquires a first radiation image G1 of the subject H using low-energy radiation including so-called soft rays. Meanwhile, the second radiation detector 6 acquires a second radiation image G2 of the subject H using high-energy radiation from which soft rays have been removed. Therefore, 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 and second radiation images G1, G2 are input to the estimation device 10. Both the first radiation image G1 and the second radiation image G2 are front images including the crotch area of ​​the subject H.

[0023] The first and second radiation detectors 5, 6 are capable of repeatedly recording and reading out radiation images, and may be so-called direct type radiation detectors that generate electric charges by directly receiving radiation, or so-called indirect type radiation detectors that convert radiation into visible light and then convert the visible light into an electric charge signal. As a method for reading out the radiation image signal, it is preferable to use a so-called TFT readout method in which the radiation image signal is read out by turning on and off a TFT (thin film transistor) switch, or a so-called optical readout method in which the radiation image signal is read out by irradiating the detector with readout light, but this is not limiting and other methods may also be used.

[0024] The imaging device 1 can also perform simple imaging of the subject H using only the first radiation detector 5 to obtain a simple radiation image G0, which is a simple two-dimensional image of the subject H. The imaging for obtaining the first and second radiation images G1, G2 is called energy subtraction imaging to distinguish it from simple imaging. In this embodiment, the first and second radiation images G1, G2 obtained by energy subtraction imaging are used as learning data, which will be described later. Furthermore, the simple radiation image G0 obtained by simple imaging is used to derive an estimation result related to bone density, as will be described later.

[0025] The CT device 8 obtains a plurality of tomographic images representing a plurality of tomographic planes in the subject H as a three-dimensional CT image V0. The CT value of each pixel (voxel) in the CT image is a numerical representation of the radiation absorption rate in the components that make up the human body. The CT value will be described later.

[0026] The image storage system 9 is a system that stores image data of radiographic images acquired by the imaging device 1 and image data of CT images acquired by the CT device 8. The image storage system 9 extracts images from the stored radiographic images and CT images in response to requests from the estimation device 10 and the information derivation device 50, and transmits the images to the requesting device. A specific example of the image storage system 9 is a PACS (Picture Archiving and Communication System). In this embodiment, the image storage system 9 stores a large amount of teacher data for training a neural network, which will be described later.

[0027] Next, the estimation device according to the present embodiment will be described. First, the hardware configuration of the estimation device according to the present embodiment will be described with reference to FIG. 2. As shown in FIG. 2, the estimation device 10 is a computer such as a workstation, a server computer, or a personal computer, and includes a CPU (Central Processing Unit) 11, a non-volatile storage 13, and a memory 16 as a temporary storage area. The estimation device 10 also includes a display 14 such as a liquid crystal display, an input device 15 such as a keyboard and a mouse, and a network I / F (InterFace) 17 connected to a network (not shown). The CPU 11, the storage 13, the display 14, the input device 15, the memory 16, and the network I / F 17 are connected to a bus 18. The CPU 11 is an example of a processor in the present disclosure.

[0028] The storage 13 is realized by a hard disk drive (HDD), a solid state drive (SSD), a flash memory, etc. The storage 13 as a storage medium stores the estimation program 12A and the learning program 12B installed in the estimation device 10. The CPU 11 reads out the estimation program 12A and the learning program 12B from the storage 13, expands them in the memory 16, and executes the expanded estimation program 12A and the learning program 12B.

[0029] The estimation program 12A and the learning program 12B are stored in a storage device of a server computer connected to a network or in a network storage in an externally accessible state, and are downloaded and installed in the computer constituting the estimation device 10 upon request. Alternatively, they are recorded on a recording medium such as a DVD (Digital Versatile Disc) or a CD-ROM (Compact Disc Read Only Memory) and distributed, and are installed in the computer constituting the estimation device 10 from the recording medium.

[0030] Next, the functional configuration of the estimation device according to this embodiment will be described. Fig. 3 is a diagram showing the functional configuration of the estimation device according to this embodiment. As shown in Fig. 3, 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. The CPU 11 executes an estimation program 12A to function as the image acquisition unit 21, the information acquisition unit 22, the estimation unit 23, and the display control unit 25. The CPU 11 executes a learning program 12B to function as the learning unit 24.

[0031] The image acquisition unit 21 causes the imaging device 1 to perform energy subtraction imaging of the subject H, thereby acquiring a first radiation image G1 and a second radiation image G2, which are, for example, front images of the crotch area of ​​the subject H, from the first and second radiation detectors 5, 6. When acquiring the first radiation image G1 and the second radiation image G2, imaging conditions are set, such as the imaging dose, radiation quality, tube voltage, SID (Source Image Receptor Distance) which is the distance between the radiation source 3 and the surfaces of the first and second radiation detectors 5, 6, SOD (Source Object Distance) which is the distance between the radiation source 3 and the surface of the subject H, and the presence or absence of an anti-scatter grid.

[0032] The SOD and SID are used to calculate the body thickness distribution as described below. The SOD is preferably obtained by, for example, a TOF (Time Of Flight) camera. The SID is preferably obtained by, for example, a potentiometer, an ultrasonic distance meter, a laser distance meter, or the like.

[0033] The imaging conditions may be set by the operator through input device 15. The set imaging conditions are stored in storage 13. The first and second radiation images G1, G2 acquired by energy subtraction imaging and the imaging conditions are also transmitted to and stored in image storage system 9.

[0034] The image acquiring section 21 also acquires a simple radiographic image G0, which is a front image of the crotch area of ​​the subject H, by causing the imaging device 1 to perform simple imaging of the subject H using only the first radiation detector 5.

[0035] In this embodiment, the first and second radiographic images G1, G2 and the simple radiographic image G0 may be acquired by a program separate from the estimation program 12A and stored in the storage 13. In this case, the image acquisition unit 21 acquires the first and second radiographic images G1, G2 and the simple radiographic image G0 stored in the storage 13 by reading them out from the storage 13 for processing.

[0036] The information acquisition unit 22 acquires teacher data for training a neural network, which will be described later, from the image storage system 9 via the network I / F 17.

[0037] The estimation unit 23 derives an estimation result related to the three-dimensional bone density of the bone part included in the subject H from the simple radiographic image G0. In this embodiment, the estimation result related to the three-dimensional bone density is an estimation result of the three-dimensional bone density of the target bone in the bone part region included in the simple radiographic image G0. For this purpose, the estimation unit 23 derives the estimation result related to the three-dimensional bone density using a trained neural network 23A that outputs the three-dimensional bone density when the simple radiographic image G0 is input.

[0038] Here, the estimation result related to the three-dimensional bone density may be the bone density per unit volume of each pixel included in the bone part of the subject H in the plain radiographic image G0, or a representative value of the bone density per unit volume of each pixel included in the region of the target bone in the bone part, etc. In this embodiment, the target bone is the femur, and the representative value of the bone density per unit volume of each pixel included in the femur is derived as the estimation result related to the three-dimensional bone density.

[0039] The learning unit 24 constructs a trained neural network 23A by machine learning the neural network using the teacher data. Examples of the neural network include a simple perceptron, a multilayer perceptron, a deep neural network, a convolutional neural network, a deep belief network, a recurrent neural network, and a probabilistic neural network. In this embodiment, a convolutional neural network is used as the neural network.

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

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

[0042] FIG. 5 is a diagram showing an example of training data used for training the neural network. As shown in FIG. 5, the training data 40 is composed of training data 41 and correct answer data 42. In this embodiment, the data input to the trained neural network 23A to obtain an estimation result related to three-dimensional bone density is a simple radiographic image G0, but the training data 41 includes two radiographic images, a first radiographic image G1 and a second radiographic image G2, acquired by the above-mentioned energy subtraction imaging. In this embodiment, the training data 41 further includes a two-dimensional bone density Bs of the bone part of the subject from which the training data 41 was acquired.

[0043] The two-dimensional bone density may be the bone density per unit area of ​​each pixel in the bone part included in the first radiographic image G1 or the second radiographic image G2, or a representative value of the bone density per unit area of ​​each pixel included in the region of the target bone in the bone part, etc. In this embodiment, since the target bone is the femur, the representative value of the bone density per unit area of ​​each pixel included in the femur is used as the two-dimensional bone density Bs.

[0044] The correct answer data 42 is a three-dimensional bone density of the target bone (i.e., femur) of the subject from which the learning data 41 was obtained. In this embodiment, the three-dimensional bone density is based on bone density per unit volume, so the unit of bone density is (g / cm 3 On the other hand, the two-dimensional bone density Bs included in the learning data 41 is based on the bone density per unit area, so the unit is (g / cm 2 The two-dimensional bone density Bs, which is the learning data 41, and the three-dimensional bone density, which is the answer data 42, are derived by an information derivation device 50. The information derivation device 50 will be described below.

[0045] Fig. 6 is a schematic block diagram showing the configuration of the information derivation device according to this embodiment. As shown in Fig. 6, the information derivation device 50 according to this embodiment is a computer such as a workstation, a server computer, or a personal computer, and includes a CPU 51, a non-volatile storage 53, and a memory 56 as a temporary storage area. The information derivation device 50 also includes a display 54 such as a liquid crystal display, an input device 55 consisting of a pointing device such as a keyboard and a mouse, and a network I / F 57 connected to a network (not shown). The CPU 51, the storage 53, the display 54, the input device 55, the memory 56, and the network I / F 57 are connected to a bus 58.

[0046] The storage 53 is realized by an HDD, an SSD, a flash memory, or the like, similar to the storage 13. The storage 53 as a storage medium stores an information derivation program 52. The CPU 51 reads out the information derivation program 52 from the storage 53, expands it in the memory 56, and executes the expanded information derivation program 52.

[0047] Next, the functional configuration of the information derivation device according to this embodiment will be described. Fig. 7 is a diagram showing the functional configuration of the information derivation device according to this embodiment. As shown in Fig. 7, the information derivation device 50 according to this embodiment includes an image acquisition unit 61, a scattered radiation removal unit 62, a subtraction unit 63, a first bone density derivation unit 64, and a second bone density derivation unit 65. When the CPU 51 executes the information derivation program 52, the CPU 51 functions as the image acquisition unit 61, the scattered radiation removal unit 62, the subtraction unit 63, the first bone density derivation unit 64, and the second bone density derivation unit 65.

[0048] The image acquisition unit 61 acquires the first radiographic image G1 and the second radiographic image G2, which are to become the learning data 41 stored in the image storage system 9. The image acquisition unit 61 may acquire the first radiographic image G1 and the second radiographic image G2 by having the imaging device 1 capture an image of the subject H, similar to the image acquisition unit 21 of the estimation device 10. The image acquisition unit 61 acquires a CT image V0 for deriving the learning data 41 from the image storage system 9. The image acquisition unit 61 may acquire the CT image V0 by having the CT device 8 capture an image of the subject H, similar to the image acquisition unit 21 of the estimation device 10.

[0049] The image acquiring unit 61 also acquires the imaging conditions when the first and second radiographic images G1, G2 were acquired and stored in the image storage system 9. The imaging conditions include the imaging dose, tube voltage, SID, SOD, and the presence or absence of an anti-scatter grid when the first radiographic image G1 and the second radiographic image G2 were acquired.

[0050] Here, each of the first radiographic image G1 and the second radiographic image G2 contains scattered radiation components based on radiation scattered within the subject H in addition to primary radiation components of the radiation that has passed through the subject H. For this reason, the scattered radiation removal section 62 removes the scattered radiation components from the first radiographic image G1 and the second radiographic image G2. For example, the scattered radiation removal section 62 may remove the scattered radiation components from the first radiographic image G1 and the second radiographic image G2 by applying the method described in JP 2015-043959 A. When using the method described in JP 2015-043959 A, the body thickness distribution of the subject H and the scattered radiation components for removing the scattered radiation components are derived simultaneously.

[0051] The removal of the scattered radiation components from the first radiographic image G1 will be described below, but the removal of the scattered radiation components from the second radiographic image G2 can be performed in the same manner. First, the scattered radiation 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 virtually representing the subject H in which the body thickness according to the initial body thickness distribution T0(x,y) is associated with the coordinate position of each pixel of the first radiographic image G1. The virtual model of the subject H having the initial body thickness distribution T0(x,y) may be stored in advance in the storage 53 of the information derivation device 50. In addition, the scattered radiation removal unit 62 may calculate the body thickness distribution T(x,y) of the subject H based on the SID and SOD included in the imaging conditions. In this case, the initial body thickness distribution T0(x,y) can be obtained by subtracting the SOD from the SID.

[0052] Next, the scattered radiation removal unit 62 generates an estimated image based on the virtual model by combining an estimated primary ray image obtained by estimating a primary ray image obtained by photographing the virtual model and an estimated scattered ray image obtained by estimating a scattered ray image obtained by photographing the virtual model, as an estimated image based on the first radiation image G1 obtained by photographing the subject H.

[0053] Next, the scattered radiation removal unit 62 corrects the initial body thickness distribution T0(x,y) of the virtual model so that the difference between the estimated image and the first radiographic image G1 becomes smaller. The scattered radiation removal unit 62 repeatedly generates the estimated image and corrects the body thickness distribution until the difference between the estimated image and the first radiographic image G1 satisfies a predetermined end condition. The scattered radiation removal unit 62 derives the body thickness distribution when the end condition is satisfied as the body thickness distribution T(x,y) of the subject H. In addition, the scattered radiation removal unit 62 removes the scattered radiation components contained in the first radiographic image G1 by subtracting the scattered radiation components when the end condition is satisfied from the first radiographic image G1.

[0054] The subtraction unit 63 performs energy subtraction processing to derive a bone image Gb from the first and second radiographic images G1, G2, from which the bones of the subject H are extracted. In the following processing, the first and second radiographic images G1, G2 are ones from which scattered radiation components have been removed. When deriving the bone image Gb, the subtraction unit 63 performs weighted subtraction between corresponding pixels of the first and second radiographic images G1, G2, as shown in the following formula (1), to generate a bone image Gb from which the bones of the subject H included in each of the radiographic images G1, G2 are extracted, as shown in FIG. 8. In formula (1), α is a weighting coefficient. The pixel value of each pixel in the bone region of the bone image Gb is the bone pixel value. Gb(x,y)=α·G2(x,y)-G1(x,y) (1)

[0055] The first bone density derivation unit 64 derives bone density for each pixel of the bone image Gb. In this embodiment, the first bone density derivation unit 64 derives bone density by converting each pixel value of the bone image Gb into a pixel value of a bone image acquired under standard imaging conditions. Specifically, the first bone density derivation unit 64 derives bone density by correcting each pixel value of the bone image Gb using a correction coefficient acquired from a lookup table described later.

[0056] Here, the higher the tube voltage in the radiation source 3 and the higher the energy of the radiation emitted from the radiation source 3, the smaller the contrast between soft tissue and bones in the radiological image. In addition, in the process of the radiation passing through the subject H, low-energy components of the radiation are absorbed by the subject H, causing beam hardening, in which the radiation becomes more energetic. The increase in the energy of the radiation due to beam hardening becomes greater as the body thickness of the subject H increases.

[0057] FIG. 9 is a diagram showing the relationship between the contrast between bones and soft tissues and the body thickness of the subject H. Note that FIG. 9 shows the relationship between the contrast between bones and soft tissues and the body thickness of the subject H at three tube voltages of 80 kV, 90 kV, and 100 kV. As shown in FIG. 9, the higher the tube voltage, the lower the contrast. Also, when the body thickness of the subject H exceeds a certain value, the larger the body thickness, the lower the contrast. Note that the larger the pixel value of the bone region in the bone image Gb, the higher the contrast between the bones and soft tissues. Therefore, the relationship shown in FIG. 9 shifts to the high contrast side as the pixel value of the bone region in the bone image Gb increases.

[0058] In this embodiment, a lookup table for acquiring correction coefficients for correcting the difference in contrast in the bone image Gb according to the tube voltage at the time of imaging and the decrease in contrast due to the effect of beam hardening is stored in the storage 53 of the information derivation device 50. The correction coefficients are coefficients for correcting each pixel value of the bone image Gb.

[0059] FIG. 10 is a diagram showing an example of a lookup table for acquiring a correction coefficient. In FIG. 10, a lookup table (hereinafter, simply referred to as a table) LUT1 in which a reference imaging condition is set to a tube voltage of 90 kV is illustrated. As shown in FIG. 10, in the table LUT1, a larger correction coefficient is set as the tube voltage increases and as the body thickness of the subject H increases. In the example shown in FIG. 10, since the reference imaging condition is a tube voltage of 90 kV, when the tube voltage is 90 kV and the body thickness is 0, the correction coefficient is 1. Note that, although the table LUT1 is shown two-dimensionally in FIG. 10, the correction coefficient differs depending on the pixel value of the bone region. Therefore, the table LUT1 is actually a three-dimensional table to which an axis representing the pixel value of the bone region is added.

[0060] The first bone density derivation unit 64 extracts, from the table LUT1, a correction coefficient C0(x, y) for each pixel according to the imaging conditions including the body thickness distribution T(x, y) of the subject H and the set value of the tube voltage stored in the storage 13. Then, as shown in the following formula (2), the first bone density derivation unit 64 multiplies each pixel (x, y) of the bone region in the bone image Gb by the correction coefficient C0(x, y) to obtain a two-dimensional bone density Bs(x, y) (g / cm 2 ) is derived. The two-dimensional bone density Bs(x, y) derived in this manner represents pixel values ​​of the bone region included in a radiographic image acquired by imaging the subject H under a tube voltage of 90 kV, which is the standard imaging condition, and from which the effects of beam hardening have been removed. Therefore, a bone density image in which the derived bone density is the pixel value of each pixel is derived by the first bone density derivation unit 64. Bs(x,y)=C0(x,y)×Gb(x,y) (2)

[0061] Furthermore, in this embodiment, the first bone density derivation unit 64 derives a representative value of the two-dimensional bone density Bs only for the target bone. For example, when the target bone is the femur, the first bone density derivation unit 64 derives a representative value of the two-dimensional bone density Bs of the femur region by deriving a representative value of the two-dimensional bone density Bs of each pixel in the femur region in the bone image Gb. The representative value may be an average value, a median value, a minimum value, a maximum value, or the like. In this embodiment, the representative value of the two-dimensional bone density Bs of the femur, which is the target bone, is used as the two-dimensional bone density Bs of the learning data 41.

[0062] The second bone density derivation unit 65 derives three-dimensional bone density from the CT image V0. Fig. 11 is a diagram for explaining the derivation of three-dimensional bone density. Note that in Fig. 11, the three-dimensional CT image V0 is shown in two dimensions for the sake of explanation. As shown in Fig. 11, a subject H is included in the three-dimensional space represented by the CT image V0. The subject H is composed of multiple components, including bones, fat, muscles, and internal organs.

[0063] Here, the CT value V0(x, y, z) of each pixel of the CT image V0 can be expressed by the following formula (3) using the attenuation coefficient μi of the composition at that pixel and the attenuation coefficient μw of water. (x, y, z) are coordinates that represent the pixel position of the CT image V0. In the following explanation, the attenuation coefficient means the linear attenuation coefficient unless otherwise specified. The attenuation coefficient represents the degree (proportion) of attenuation of radiation due to absorption or scattering, etc. The attenuation coefficient differs depending on the specific composition (density, etc.) and thickness (mass) of the structure through which the radiation passes. V0(x,y,z)=(μi-μw) / μw×1000 (3)

[0064] The attenuation coefficient μw of water is known. Therefore, by solving equation (3) for μi, the attenuation coefficient μi of each composition can be calculated as shown in equation (4) below. μi=V0(x,y,z)×μw / 1000+μw (4)

[0065] When deriving the three-dimensional bone density, the second bone density deriving unit 65 derives the three-dimensional bone density by virtually irradiating the subject H with radiation at an exposure dose I0 from the front of the subject H as shown in FIG. 11, and virtually detecting the radiation that has passed through the subject H with a radiation detector (not shown) installed on a virtual plane 67 to derive a composite two-dimensional image CG. The exposure dose I0 and radiation energy of the virtual radiation are set according to predetermined imaging conditions. Specifically, the exposure dose I0 may be set by preparing a table corresponding to imaging conditions such as tube voltage, mAs value, and SID, and referring to this table. Also, the radiation energy may be set by preparing a table corresponding to the tube voltage, and referring to this table. At this time, the reaching dose I1(x,y) for each pixel of the composite two-dimensional image CG passes through one or more compositions in the subject H. For this reason, the reaching dose I1(x,y) can be derived by the following formula (5) using the attenuation coefficient μi of one or more compositions through which the radiation at the exposure dose I0 passes. The arriving dose I1(x, y) becomes the pixel value of each pixel of the composite two-dimensional image CG. I1(x,y)=I0×exp(-∫μi·dt) (5)

[0066] If the irradiating radiation source is assumed to be a surface light source, the attenuation coefficient μi used in equation (5) may be derived from the CT values ​​of each pixel arranged in the vertical direction as shown in Fig. 11 using equation (4). If the irradiating radiation source is assumed to be a surface light source as a point light source, as shown in Fig. 12, pixels on the path of the radiation that reaches each pixel may be identified based on the geometric positional relationship between the point light source and each position on a virtual plane 67, and the attenuation coefficient μi derived from the CT value of the identified pixel using equation (4) may be used.

[0067] The composite two-dimensional image CG is derived based on radiation virtually irradiated from the front of the subject H. Therefore, the composite two-dimensional image CG can be positioned to correspond to the first and second radiation images G1, G2, and further to the bone image Gb. Therefore, the second bone density derivation unit 65 aligns the composite two-dimensional image CG with the bone image Gb, and associates corresponding pixel positions with each other.

[0068] The second bone density derivation unit 65 uses the CT image V0 to derive the three-dimensional bone density Bv of the subject H for each pixel of the composite two-dimensional image CG. Here, the CT value will be explained. FIG. 13 is a diagram for explaining the CT value. The CT value is a numerical representation of the X-ray absorption rate in the human body. Specifically, as shown in FIG. 13, the CT value is determined according to the composition constituting the human body, such that water has a CT value of 0 and air has a CT value of -1000 (unit: HU).

[0069] The second bone density derivation unit 65 first identifies a bone region in the CT image V0 based on the CT value of the CT image V0. Specifically, a region consisting of pixels with a CT value of 100 to 1000 is identified as the bone region by threshold processing. Note that instead of threshold processing, the bone region may be identified by using a trained neural network that has been trained to detect the bone region from the CT image V0. Alternatively, the CT image V0 may be displayed on the display 54, and the bone region may be identified by manually specifying the bone region in the displayed CT image V0.

[0070] Here, the density per unit volume of the composition at each pixel of the CT image V0 is ρ (g / cm 3 ) is the attenuation coefficient μi(1 / cm) of a composition and the mass attenuation coefficient μe(cm 2 / g) according to the following formula (6). ρ=μi / μe (6)

[0071] FIG. 14 is a diagram showing the relationship between radiation energy and mass attenuation coefficient for various compositions of the human body. FIG. 14 shows the relationship between radiation energy and mass attenuation coefficient for bones, muscles, etc., and fat. Note that muscles, etc. refer to muscles, blood, and water. In this embodiment, the relationship between radiation energy and mass attenuation coefficient shown in FIG. 14 is stored in the storage 53 as a table. In this embodiment, since the mass attenuation coefficient of bones is required, the mass attenuation coefficient of bones is obtained by referring to the relationship for bones in the table shown in FIG. 14 based on virtual radiation energy. In addition, the attenuation coefficient μb in each pixel of the bone region is derived by the above formula (4). Then, the bone density ρ per unit volume in each pixel of the bone region included in the CT image V0 is derived by the above formula (6).

[0072] Since the CT image V0 is a three-dimensional image, the unit of the bone mineral density per unit volume ρ derived by the above formula (6) is (g / cm 3 ). In this embodiment, the second bone density derivation section 65 derives the bone density per unit volume for each pixel of the composite two-dimensional image CG as a three-dimensional bone density Bv. For this reason, the second bone density derivation section 65 derives the three-dimensional bone density Bv for each pixel of the composite two-dimensional image CG by projecting the bone density per unit volume ρ derived by the above formula (6) onto a virtual plane 67 in the same manner as when the composite two-dimensional image CG is derived. The unit of the three-dimensional bone density Bv is (g / cm 3 ).

[0073] In addition, when projecting, a representative value of bone density per unit volume of each pixel of the CT image V0 on a path from a virtual radiation source to each pixel of the composite two-dimensional image CG may be derived. The representative value may be an integrated value, an average value, a maximum value, a median value, a minimum value, or the like. Furthermore, in this embodiment, the second bone density derivation unit 65 derives a representative value of the three-dimensional bone density Bv only for the target bone. For example, since the target bone is the femur in this embodiment, the second bone density derivation unit 65 derives a representative value of the three-dimensional bone density Bv of the femur region by deriving a representative value of the three-dimensional bone density Bv of each pixel in the femur region in the composite two-dimensional image CG. The representative value may be an average value, a median value, a minimum value, a maximum value, or the like. In this embodiment, the representative value of the three-dimensional bone density Bv of the femur, which is the target bone, is used as the correct answer data 42. As a result, the two-dimensional bone density Bs derived by the first bone density derivation unit 64 and the three-dimensional bone density Bv derived by the second bone density derivation unit 65 are the two-dimensional bone density and the three-dimensional bone density, respectively, of the same target bone.

[0074] The two-dimensional bone density used as the learning data 41 and the three-dimensional bone density used as the answer data 42 are derived at the same time as the learning data 41 is acquired, and transmitted to the image storage system 9. In the image storage system 9, the learning data 41 and the answer data 42 are associated with each other and stored as the teacher data 40. Note that, in order to improve the robustness of learning, teacher data 40 including images obtained by performing at least one of the following on the same image, such as enlarging / reducing, changing the contrast, moving, rotating in a plane, inverting, and adding noise, may be additionally created and stored as the learning data 41.

[0075] Returning to the estimation device 10, the learning unit 24 learns the neural network using a large amount of teacher data 40. FIG. 15 is a diagram for explaining the learning of the neural network 30. When learning the neural network 30, the learning unit 24 inputs learning data 41, i.e., the first and second radiographic images G1 and G2 and the two-dimensional bone density Bs, to the input layer 31 of the neural network 30. Then, the learning unit 24 causes the output layer 33 of the neural network 30 to output the three-dimensional bone density of the target bone as output data 47. Then, the learning unit 24 derives the difference between the output data 47 and the correct answer data 42 as a loss L0.

[0076] 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 convolution layer 35, the connection weights between the layers, and the connection weights in the fully connected layer 37 (hereinafter referred to as parameters 48) so as to reduce the loss L0. For example, the backpropagation method can be used as a method for adjusting the parameters 48. The learning unit 24 repeats the adjustment of the parameters 48 until the loss L0 becomes equal to or less than a predetermined threshold value. As a result, when a plain radiographic image G0 is input, the parameters 48 are adjusted so as to output the bone density of the target bone, and the trained neural network 23A is constructed. The constructed trained neural network 23A is stored in the storage 13.

[0077] Fig. 16 is a conceptual diagram of the process performed by the trained neural network 23A. As shown in Fig. 16, when a plain radiographic image G0 of a patient is input to the trained neural network 23A constructed as described above, the trained neural network 23A outputs a three-dimensional bone density of the target bone (i.e., the femur) included in the input plain radiographic image G0.

[0078] The display control unit 25 displays the bone density estimation result estimated by the estimation unit 23 on the display 14. Fig. 17 is a diagram showing a display screen of the estimation result. As shown in Fig. 17, a display screen 70 has an image display area 71 and a bone density display area 72. A plain radiographic image G0 of the subject H is displayed in the image display area 71. Furthermore, a representative value of the three-dimensional bone density around the femoral joint for the bone density estimated by the estimation unit 23 is displayed in the bone density display area 72.

[0079] Next, the process performed in this embodiment will be described. FIG. 18 is a flowchart showing the learning process performed in this embodiment. First, the information acquisition unit 22 acquires the teacher data 40 from the image storage system 9 (step ST1), and the learning unit 24 inputs the learning data 41 included in the teacher data 40 to the neural network 30 to output the three-dimensional bone density, and learns the neural network 30 using the loss L0 based on the difference from the correct answer data 42 (step ST2), and returns to step ST1. Then, the learning unit 24 repeats the processes of steps ST1 and ST2 until the loss L0 becomes a predetermined threshold value, and ends the learning process. Note that the learning unit 24 may end the learning process by repeating the learning a predetermined number of times. In this way, the learning unit 24 constructs the trained neural network 23A.

[0080] Next, the estimation process in this embodiment will be described. Fig. 19 is a flowchart showing the estimation process in this embodiment. It is assumed that the simple radiographic image G0 is acquired by imaging and stored in the storage 13. When an instruction to start the process is input from the input device 15, the image acquisition unit 21 acquires the simple radiographic image G0 from the storage 13 (step ST11). Next, the estimation unit 23 derives an estimation result related to the three-dimensional bone density from the simple radiographic image G0 (step ST12). Then, the display control unit 25 displays the estimation result related to the three-dimensional bone density derived by the estimation unit 23 on the display 14 together with the simple radiographic image G0 (step ST13), and the process ends.

[0081] In this manner, in the present embodiment, the trained neural network 23A constructed by learning using the first and second radiographic images G1, G2 and the two-dimensional bone density Bs as teacher data is adapted to derive an estimation result related to the three-dimensional bone density of the subject H included in the simple radiographic image G0. Here, in the present embodiment, two radiographic images, the first and second radiographic images G1, G2, are used for training the neural network. Furthermore, in the present embodiment, the two-dimensional bone density Bs is also used for training the neural network. Therefore, compared to a case where one radiographic image and information related to bone density are used as teacher data, the trained neural network 23A is capable of deriving an estimation result related to the three-dimensional bone density from the simple radiographic image G0 with greater accuracy. Therefore, according to the present embodiment, it is possible to derive an estimation result related to the three-dimensional bone density with greater accuracy.

[0082] In the above embodiment, the second bone density derivation unit 65 derives the three-dimensional bone density by projecting the bone density per unit volume for each pixel of the CT image V0 onto each pixel of the composite two-dimensional image CG, but the present invention is not limited to this. A representative value of the bone density ρ of each pixel in the three-dimensional region of the target bone in the CT image V0 derived by the above formula (6) may be derived as the three-dimensional bone density of the region of the target bone. As the representative value, the average value, median value, minimum value, maximum value, etc. of the bone density ρ of each pixel may be used. In this case, the bone density for each pixel of the composite two-dimensional image CG derived by the second bone density derivation unit 65 in the above embodiment may be used as the two-dimensional bone density Bs in the learning data 41.

[0083] In the above embodiment, the first and second radiographic images G1 and G2 are used as the learning data 41 of the teacher data 40, but the present invention is not limited to this. As in teacher data 40A shown in Fig. 20, a bone image Gb may be used as the learning data 41A instead of the second radiographic image G2. In this case, the bone image Gb may be one derived according to the above embodiment.

[0084] In the above embodiment, instead of the representative value of the two-dimensional bone density Bs, a two-dimensional bone density image Gds in which each pixel has a two-dimensional bone density value may be used as the learning data 41B, as in the teacher data 40B shown in FIG. 21. In this case, the correct answer data 42 may be a three-dimensional bone density image Gdv in which each pixel has a three-dimensional bone density value. When such teacher data 40B is used, the estimation unit 23 of the estimation device 10 derives a three-dimensional bone density image from the plain radiographic image G0 as an estimation result related to the three-dimensional bone density. When the three-dimensional bone density image is derived in this way, the bone density image may be displayed on the display screen.

[0085] FIG. 22 is a diagram showing another example of the display screen of the estimation result. As shown in FIG. 22, the display screen 70A has an image display area 71 similar to the display screen 70 shown in FIG. 17. The image display area 71 displays a three-dimensional bone density image Gdv, which is an estimation result of the three-dimensional bone density in the simple radiation image G0 of the subject H. In the three-dimensional bone density image Gdv, a pattern is added to the bone area according to the three-dimensional bone density. In FIG. 22, for the sake of simplicity, a pattern representing the bone salt amount is added only to the femur. A reference 73 indicating the magnitude of the bone salt amount for the added pattern is displayed below the image display area 71. The operator can easily recognize the three-dimensional bone density of the patient by reading the three-dimensional bone density image Gdv while referring to the reference 73. In place of the pattern, a different color may be added to the three-dimensional bone density image Gdv according to the three-dimensional bone density.

[0086] In the above embodiment, information related to the three-dimensional bone density of the femur near the hip joint is estimated, but the target bone is not limited to the femur. The technology disclosed herein can also be applied to estimating information related to the three-dimensional bone density of any bone part, such as the femur and tibia near the knee joint, vertebrae such as the lumbar vertebrae, the calcaneus, and the metacarpals.

[0087] In the above embodiment, the estimation result related to bone density is derived from the simple radiation image G0, but the present invention is not limited to this. For example, the technology of the present disclosure can be applied to the case where the estimation result related to bone density is derived from the DXA scan image acquired by photographing the subject with the DXA imaging device described in Japanese Patent Application Laid-Open No. 9-108206 and Japanese Patent Application Laid-Open No. 2006-271437. The DXA scan image refers to a radiation image captured by a radiation detector, in which a thin collimated high-energy radiation beam and a low-energy radiation beam are irradiated to the subject while switching and scanning. The thin collimated radiation beam is, for example, a radiation beam formed into a pencil beam, a narrow fan beam, a wide fan beam, or the like, using a collimator between the radiation source and the subject. The low-energy radiation refers to radiation with energy relatively lower than that of the high-energy radiation.

[0088] The trained neural network 23A may be constructed using the DXA scan image as a teacher image. In this case, for example, instead of the first and second radiation images G1 and G2 shown in FIG. 5, a DXA scan image acquired by a low-energy radiation beam and a DXA scan image acquired by a high-energy radiation beam may be used. The two-dimensional bone density Bs may be a two-dimensional bone density calculated from the DXA scan image. Instead of the first radiation image G1 and bone image Gb shown in FIG. 21, a DXA scan image acquired by a low-energy radiation beam and a bone image calculated from the DXA scan image may be used.

[0089] In the above embodiment, the two-dimensional bone density of the learning data is derived by energy subtraction processing, but the present invention is not limited to this. The two-dimensional bone density may be derived by the DXA method using a DXA scan image acquired at the same time as the first and second radiographic images G1 and G2 included in the learning data were acquired.

[0090] In the above embodiment, the trained neural network 23A is constructed by training the neural network in the estimation device 10, but the present invention is not limited to this. The trained neural network 23A constructed in a device other than the estimation device 10 may be used in the estimation unit 23 of the estimation device 10 in this embodiment.

[0091] In the above embodiment, the first and second radiographic images G1, G2 are acquired by a one-shot method when performing the energy subtraction process for deriving bone density, but the present invention is not limited to this. The first and second radiographic images G1, G2 may be acquired by a so-called two-shot method in which imaging is performed twice using only one radiation detector. In the case of the two-shot method, the position of the subject H included in the first radiographic image G1 and the second radiographic image G2 may be shifted due to the body movement of the subject H. For this reason, it is preferable to align the position of the subject in the first radiographic image G1 and the second radiographic image G2 before performing the process of this embodiment.

[0092] The alignment process may be performed using, for example, the method described in JP 2011-255060 A. The method described in JP 2011-255060 A generates a plurality of first band images and a plurality of second band images representing structures having different frequency bands for each of the first and second radiographic images G1 and G2, obtains the amount of positional deviation between corresponding positions in the first band image and the second band image of the corresponding frequency band, and aligns the first radiographic image G1 and the second radiographic image G2 based on the amount of positional deviation.

[0093] In the above embodiment, the bone density is derived as the correct answer data of the teacher data and information related to the bone density is estimated using the radiation images acquired in the system that captures the image of the subject H using the first and second radiation detectors 5 and 6. However, the technology of the present disclosure can be applied to a case where the first and second radiation images G1 and G2 are acquired using stimulable phosphor sheets instead of radiation detectors. In this case, two stimulable phosphor sheets are stacked and irradiated with radiation that has passed through the subject H, and the radiation image information of the subject H is accumulatively recorded on each stimulable phosphor sheet, and the radiation image information is photoelectrically read from each stimulable phosphor sheet to acquire the first and second radiation images G1 and G2. The two-shot method may also be used when the first and second radiation images G1 and G2 are acquired using stimulable phosphor sheets.

[0094] Moreover, the radiation in the above embodiment is not particularly limited, and in addition to X-rays, α-rays, γ-rays, etc. can be used.

[0095] In the above embodiment, the following various processors can be used as the hardware structure of the processing units that perform various processes, such as the image acquisition unit 21, the information acquisition unit 22, the estimation unit 23, the learning unit 24, and the display control unit 25 of the estimation device 10, and the image acquisition unit 61, the scattered radiation removal unit 62, the subtraction unit 63, the first bone density derivation unit 64, and the second bone density derivation unit 65 of the information derivation device 50. As described above, the above various processors include a CPU, which is a general-purpose processor that executes software (programs) and functions as various processing units, as well as a programmable logic device (PLD), which is a processor whose circuit configuration can be changed after manufacture, such as an FPGA (Field Programmable Gate Array), and a dedicated electric circuit, which is a processor having a circuit configuration designed specifically for performing specific processes, such as an ASIC (Application Specific Integrated Circuit).

[0096] A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (e.g., a combination of multiple FPGAs or a combination of a CPU and an FPGA). Also, multiple processing units may be configured with a single processor.

[0097] As an example of configuring multiple processing units with one processor, first, there is a form in which one processor is configured with a combination of one or more CPUs and software, as typified by computers such as client and server, and this processor functions as multiple processing units. Second, there is a form in which a processor is used that realizes the functions of the entire system including multiple processing units with one IC (Integrated Circuit) chip, as typified by System On Chip (SoC), etc. In this way, the various processing units are configured using one or more of the above various processors as a hardware structure.

[0098] Furthermore, more specifically, the hardware structure of these various processors can be an electric circuit that combines circuit elements such as semiconductor elements. [Explanation of symbols]

[0099] 1. Imaging device 3 Radiation source 5,6 Radiation detectors 7 Radiation Energy Conversion Filter 8 CT device 9. Image Storage System 10 Estimation device 11, 51 CPU 12A Estimation Program 12B Study Program 13, 53 Storage 14,54 Display 15, 55 Input Devices 16,56 memory 17, 57 Network I / F 18, 58 Bus 21 Image acquisition unit 22 Information Acquisition Department 23 Estimation part 23A Trained Neural Networks 24 Learning Department 25 Display control section 30 Neural Networks 31 Input Layer 32 Middle Class 33 Output layer 35 Convolutional Layer 36 Pooling Layer 37 Fully connected layer 40, 40A, 40B Teacher data 41, 41A, 41B Learning data 42 Correct data 47 Output Data 48 Parameters 50 Information derivation device 52 Information derivation program 61 Image acquisition unit 62 Scattered radiation removal section 63 Subtraction Section 64 1st bone density derivation section 65 Second bone density derivation section 67 plane 70, 70A display screen 71 Image display area 72 Bone density display area 73 References Bs 2D bone density Bv 3D bone density CG composite 2D image G0 Plain radiographic image G1, G2 radiological images Gb bone image Gds 2D bone density image Gdv 3D bone density image H Subject I0 irradiation dose I1 Dose achieved L0 loss V0 CT image

Claims

1. At least one processor; The processor, a trained neural network that derives an estimation result related to a three-dimensional bone density of a bone portion from a simple radiographic image obtained by simply photographing a subject including a bone portion or a DXA scanned image obtained by photographing the subject using a DXA method; the trained neural network is trained using (i) two radiological images or DXA scanned images obtained by photographing a subject including a bone with radiation having different energy distributions, and the two-dimensional bone density of the bone contained in the two radiological images or DXA scanned images, and the three-dimensional bone density of the bone of the subject, as training data, or (ii) a radiological image or DXA scanned image of the subject, a bone image representing the bone of the subject, and the two-dimensional bone density of the bone contained in the radiological image or DXA scanned image or the bone image, and the three-dimensional bone density of the bone of the subject, as training data.

2. The estimation apparatus according to claim 1 , wherein the three-dimensional bone density is derived from a three-dimensional image of the subject.

3. The estimation device according to claim 2 , wherein the three-dimensional image is a CT image.

4. 4. The estimation device of claim 3, wherein the three-dimensional bone density is derived by identifying a bone region in the CT image, deriving an attenuation coefficient of radiation in the bone region, and deriving the three-dimensional bone density at each position of the bone region based on the attenuation coefficient of the radiation and the mass attenuation coefficient in the bone region.

5. The apparatus according to claim 1 , wherein the two-dimensional bone mineral density is derived from the two radiological images or the DXA scan image.

6. The two-dimensional bone density is a body thickness distribution of the subject estimated based on at least one of the two radiation images or the DXA scanning image; The imaging conditions under which the two radiographic images or the DXA scan image were obtained; and The estimation device according to claim 5 , wherein the bone is derived based on pixel values ​​of a bone region in a bone image from which the bone is extracted, the bone being derived by energy subtraction processing that performs weighted subtraction on the two radiation images or the DXA scan image.

7. The estimation device according to claim 1 , wherein the two-dimensional bone density is derived from the bone image.

8. The two-dimensional bone density is a body thickness distribution of the subject estimated based on the radiation image or the DXA scan image; The imaging conditions under which the radiographic image or the DXA scanned image was obtained; and The estimation device according to claim 7 , wherein the estimation value is derived based on pixel values ​​of a bone region in the bone image.

9. 1. An estimation method for deriving an estimation result related to a three-dimensional bone density of a bone portion from a simple radiographic image obtained by simple imaging of a subject including a bone portion or a DXA scanned image obtained by imaging the subject using a DXA method, using a trained neural network for deriving an estimation result related to a three-dimensional bone density of the bone portion, comprising: the trained neural network is trained using (i) two radiological images or DXA scanned images obtained by photographing a subject including a bone with radiation having different energy distributions, and the two-dimensional bone density of the bone contained in the two radiological images or DXA scanned images, and the three-dimensional bone density of the bone of the subject, as training data, or (ii) a radiological image or DXA scanned image of the subject, a bone image representing the bone of the subject, and the two-dimensional bone density of the bone contained in the radiological image or DXA scanned image or the bone image, and the three-dimensional bone density of the bone of the subject, as training data.

10. An estimation program for causing a computer to execute a procedure for deriving an estimation result related to a three-dimensional bone density of a bone portion, using a trained neural network that derives an estimation result related to a three-dimensional bone density of the bone portion from a simple radiographic image obtained by simple imaging of a subject including a bone portion, or a DXA scanned image obtained by imaging the subject using a DXA method, the estimation program comprising: the trained neural network is trained using (i) two radiological images or DXA scanned images obtained by photographing a subject including a bone with radiation having different energy distributions, and the two-dimensional bone density of the bone contained in the two radiological images or DXA scanned images, and the three-dimensional bone density of the bone of the subject, as training data, or (ii) the radiological image or DXA scanned image of the subject, a bone image representing the bone of the subject, and the two-dimensional bone density of the bone contained in the radiological image or DXA scanned image or the bone image, and the three-dimensional bone density of the bone of the subject, as training data.

Images