Radiation image analysis device, its operation method, and radiation image analysis program

Abstract

To provide a radiographic image analysis device, its operation method, and a radiographic image analysis program capable of easily measuring a bone quantity with a fixed reference by using a radiographic image.SOLUTION: A radiographic image analysis device 10 acquires an X-ray image 14 obtained by radiographing a subject containing a bone part, acquires a rotation angle of the bone part from a reference position based on the X-ray image 14, and corrects, on the basis of the rotation angle, pre-correction bone quantity information of the bone part obtained by converting a pixel value for each pixel of the X-ray image 14, thereby acquiring bone quantity information of the bone part.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to a radiographic image analyzing device for acquiring bone mass information from a radiographic image, a method for operating the same, and a radiographic image analyzing program. [Background technology]

[0002] In the medical field, bone density is measured by radiography. Bone density is used as important data for diagnosing osteoporosis or monitoring the progress of treatment. For example, DXA (Dual Energy X-ray Absorptiometry) examinations can obtain radiographic images that selectively capture bones. The image of the bones captured in these radiographic images correlates with the brightness (pixel value) of each pixel in the image, providing information on bone mass, which indicates the density of the minerals contained in the bones. Bone mass is also known as bone mineral content or bone density.

[0003] Methods for determining bone mass more accurately by radiography have been disclosed. For example, a bone densitometer is disclosed that includes a fixture for adjustably fixing the angle of the subject's legs relative to an X-ray imaging system and a registration means for registering the leg angles in order to improve the reproducibility of imaging examinations (Patent Document 1). Also, a bone densitometer is disclosed that can accurately set a region of interest for measuring the bone density of the forearm bone by correcting the direction of the region of interest (Patent Document 2). [Prior art documents] [Patent documents]

[0004] [Patent Document 1] International Publication No. 2017 / 026046 [Patent Document 2] Japanese Patent Application Laid-Open No. 2015-065999 Summary of the Invention [Problem to be solved by the invention]

[0005] The method of fixing the position of the subject, such as the angle, may take time to fix the subject in the correct position, and there is also a risk that it will be difficult to accurately compare bone mass between radiographic images taken in different positions. Furthermore, the method of correcting the direction of the region of interest may not be able to handle cases where the region of interest is shifted three-dimensionally.

[0006] An object of the present invention is to provide a radiographic image analysis device, an operating method thereof, and a radiographic image analysis program that can easily measure bone mass according to a certain standard using a radiographic image. [Means for solving the problem]

[0007] The present invention is a radiological image analysis device equipped with a processor, which acquires a radiological image obtained by performing radiography on a subject including a bone, acquires a rotation angle of the bone from a reference position based on the radiological image, and obtains bone mass information of the bone by correcting pre-correction bone mass information of the bone obtained by converting pixel values ​​for each pixel of the radiological image based on the rotation angle.

[0008] The radiographic image is preferably one obtained by photographing the subject in a frontal position.

[0009] Preferably, the bone is a proximal femur.

[0010] Preferably, the bone portion includes a predetermined region, and the processor acquires the rotation angle based on a region image of the region shown in the radiographic image.

[0011] Preferably, the bone portion is the proximal portion of the right or left femur, and the site is the lesser trochanter.

[0012] It is preferable that the processor calculates the lesser trochanter distance, which is the distance between the apex of the lesser trochanter and a tangent line that contacts the inner surface of the femur included in the proximal femur, based on the local image, and obtains the rotation angle based on the lesser trochanter distance.

[0013] Preferably, the processor acquires rotation angle correspondence information in which the lesser trochanter distance corresponds to the rotation angle, and acquires the rotation angle based on the lesser trochanter distance and the rotation angle correspondence information.

[0014] It is preferable that the processor obtains correction coefficient correspondence information in which the rotation angle corresponds to a correction coefficient for correcting the pre-correction bone mass information, and corrects the pre-correction bone mass information based on the rotation angle and the correction coefficient correspondence information.

[0015] It is preferable that the system is provided with a trained model that can estimate and output an estimation result related to the rotation angle by inputting input information having a radiological image, the trained model having parameters for outputting the estimation result based on the input information, and the processor obtains the rotation angle from the estimation result output by using the trained model.

[0016] When acquiring a radiographic image, the processor preferably estimates the scattered radiation component for each pixel based on the body thickness distribution of the subject, and removes the scattered radiation component from the radiographic image.

[0017] It is preferable that the bone portion includes cortical bone and cancellous bone, and the processor recognizes the cortical bone region and the cancellous bone region of the bone portion shown in the radiographic image, obtains pre-corrected cortical bone mass information based on the cortical bone region, obtains pre-corrected cancellous bone mass information based on the cancellous bone region, obtains the cortical bone mass information by correcting the pre-corrected cortical bone mass information based on the rotation angle, and obtains the cancellous bone mass information by correcting the pre-corrected cancellous bone mass information based on the rotation angle.

[0018] It is preferable that the processor obtains a reference value of bone mass information based on a radiographic image obtained by radiographing a phantom having a known bone mass, obtains bone mass correspondence information in which the known bone mass corresponds to the reference value of the bone mass information, and obtains the bone mass from the bone mass information based on the bone mass correspondence information.

[0019] The method of operating the radiographic image analysis device of the present invention includes the steps of acquiring a radiographic image obtained by radiographing a subject including a bone portion, acquiring a rotation angle of the bone portion from a reference position based on the radiographic image, and acquiring bone mass information of the bone portion by correcting pre-correction bone mass information of the bone portion obtained by converting pixel values ​​for each pixel of the radiographic image based on the rotation angle.

[0020] The radiographic image analysis program of the present invention causes a computer to perform the following functions: acquire a radiographic image obtained by radiographing a subject including a bone; acquire the rotation angle of the bone from a reference position based on the radiographic image; and acquire bone mass information of the bone by correcting pre-correction bone mass information of the bone obtained by converting pixel values ​​for each pixel of the radiographic image based on the rotation angle. [Effects of the Invention]

[0021] According to the present invention, bone mass can be easily measured based on a certain standard using a radiographic image. [Brief explanation of the drawings]

[0022] [Figure 1] FIG. 2 is an explanatory diagram illustrating functions of the radiation image analysis device. [Figure 2] 2A and 2B are explanatory diagrams illustrating the relationship between the shape of the lesser trochanter in an X-ray image and the rotation angle of the femur. FIG. 2A shows the case where the rotation angle is minus 20°, FIG. 2B shows the case where the rotation angle is plus or minus 0°, and FIG. 2C shows the case where the rotation angle is plus 20°. [Figure 3] FIG. 10 is an explanatory diagram illustrating the lesser trochanter distance a. [Figure 4] 4A and 4B are explanatory diagrams illustrating how the lesser trochanter distance a in X-ray images changes depending on the rotation angle of the femur. FIG. 4A shows the case where the rotation angle is minus 20°, FIG. 4B shows the case where the rotation angle is plus or minus 0°, and FIG. 4C shows the case where the rotation angle is plus 20°. [Figure 5] 1 is a graph showing the relationship between the lesser trochanter distance and the rotation angle in an X-ray image. [Figure 6] 1 is a graph showing the relationship between bone mineral density (BMD) and rotation angle. [Figure 7] 10 is a flowchart illustrating a processing flow of the radiation image analysis device. [Figure 8] FIG. 1 is a schematic diagram of a radiation image capturing system. [Figure 9] FIG. 2 is a block diagram showing the functions of the console. [Figure 10] FIG. 10 is an explanatory diagram for explaining generation of a bone tissue image. [Figure 11] FIG. 10 is a block diagram showing the functions of a console equipped with a body thickness distribution acquisition unit. [Figure 12] FIG. 2 is a block diagram showing the functions of a body thickness distribution acquisition unit. [Figure 13] FIG. 10 is a block diagram showing the functions of a rotation angle acquisition unit equipped with a trained model. [Figure 14] FIG. 1 is an explanatory diagram illustrating the functions of a trained model. [Figure 15] 10 is a lookup table showing conversion coefficients depending on body thickness and tube voltage, for tube voltages of 100 kV, 90 kV, and 80 kV. [Figure 16] FIG. 2 is a block diagram showing the functions of a bone mass acquisition unit. [Figure 17] FIG. 2 is an explanatory diagram illustrating cortical bone and cancellous bone. DETAILED DESCRIPTION OF THE INVENTION

[0023] An example of the basic configuration of the present invention will be described below. As shown in Fig. 1, a radiographic image analysis device (hereinafter referred to as "analysis device") 10 of the present invention includes an analysis image acquisition unit 11, a rotation angle acquisition unit 12, and a bone mass acquisition unit 13.

[0024] The analysis image acquisition unit 11 acquires a radiographic image obtained by performing radiography of a subject including bones. The radiography may be any radiographic image that uses radiation to obtain a radiographic image showing the bones. Therefore, radiography includes plain radiography and contrast radiography using radiation such as X-rays, as well as radiography using DXA and the like. The radiographic image obtained by performing radiography may be any radiographic image that shows the bones, and may include not only a radiographic image obtained directly by radiography, but also an image obtained after various image processing has been performed using the obtained radiographic image, and an image obtained after image synthesis or the like has been performed using multiple radiographic images. In this embodiment, plain X-ray radiography is performed as the radiography, and the analysis image acquisition unit 11 acquires a plain X-ray image (hereinafter referred to as an X-ray image) 14 as the radiographic image.

[0025] In this embodiment, the analysis image acquisition unit 11 acquires an X-ray image 14 of a proximal femur portion of a femur 15 as a bone portion. The proximal femur portion includes the lesser trochanter 16. The X-ray image 14 was taken of the proximal left femur in a frontal supine position (AP), which is a frontal position, without precise positioning adjustment. In FIG. 1, in the X-ray image 14, a region 15a that appears dark white is shown shaded, and a region 15b that appears light white is shown without shading.

[0026] The X-ray image 14 acquired by the analysis image acquisition unit 11 is used to acquire bone mass information by performing analysis based on the X-ray image 14. Therefore, it is preferable that the bones shown in the X-ray image 14 are free from various influences and only reflect the bones. In this embodiment, the analysis image acquisition unit 11 acquires the X-ray image 14 after image processing as a radiographic image so that only the bones are reflected as much as possible. Note that the X-ray image 14 acquired by the analysis image acquisition unit 11 may not be subjected to image processing. Hereinafter, the term "X-ray image 14" includes both the X-ray image 14 after image processing and the X-ray image 14 without image processing.

[0027] In addition, in this specification, "bone mass" means the same as bone mineral density (BMD), and its unit is g / cm. 2 is.

[0028] The rotation angle acquisition unit 12 acquires the rotation angle of the bone from a reference position based on the X-ray image 14. In this embodiment, the rotation angle of the femur 15 from the reference position is acquired based on the X-ray image 14. The reference position of the bone is a single position determined in advance. In this embodiment, when imaging one of the femurs 15 on either the left or right side in a frontal supine position, a positioning that is generally considered preferable is used as the reference position. For example, the reference position is a position in which the frontal plane of the pelvis is horizontal, the hip joint on the imaging side is aligned with the center of the image receiving surface, and the lower limbs are extended and slightly internally rotated.

[0029] The rotation angle refers to the angle of rotation of a rotatable bone when positioning the bone during imaging. Therefore, in the case where the bone is the femur 15, the rotation angle from the reference position refers to the angle of rotation from the reference position when the femur 15 is internally or externally rotated from the reference position due to rotation of the lower limbs, etc.

[0030] As shown in FIG. 2(B), the rotation angle when the femur 15 is in the reference position in the X-ray image 14 is 0°. In this case, the lesser trochanter 16 is visible in the X-ray image 14. In FIGS. 2(A), 2(B), and 2(C), the portion of the X-ray image 14 including the lesser trochanter 16 is shown enlarged at the bottom of each figure. As shown in FIG. 2(A), when the femur 15 is internally rotated and the rotation angle from the reference position is, for example, -20°, the shape of the lesser trochanter 16 visible in the X-ray image 14 is small. The rotation angle of internal rotation is represented by a negative angle, and the rotation angle of external rotation is represented by a positive angle. On the other hand, as shown in FIG. 2(C), when the femur 15 is externally rotated and the rotation angle from the reference position is, for example, 20°, the hidden lesser trochanter 16 becomes visible from the front, and the shape of the lesser trochanter 16 visible in the X-ray image 14 is noticeably large.

[0031] The bone mass is calculated based on the pixel values ​​of the X-ray image 14. Note that the pixel value is a value that represents the shade of a color, and since the X-ray image 14 is a monochrome image, it is also generally called the brightness value or contrast. The subject has bones and soft tissues that exist around the bones. According to the law of exponential function of attenuation of radiation, the pixel value of the bone region, which is the region in the radiographic image where the bones are captured, is calculated by multiplying the difference between the linear attenuation coefficient of the bone and the linear attenuation coefficient of the soft tissue by the linear attenuation coefficient of the bone. Department The linear attenuation coefficient of bone is a function of the density of the mineral components in the bone. Therefore, the pixel values ​​of the bone region in a radiographic image can be converted to bone mass, but the calculated bone mass may differ even for the same bone in X-ray images 14 taken at different rotation angles.

[0032] The rotation angle can be obtained using a method based on the acquired radiographic image. For example, a method can be used in which the rotation angle is obtained from a change in the shape of a portion of a bone portion shown in a radiographic image of a bone portion captured at a reference position and a radiographic image of the same bone portion captured in a measurement target. Alternatively, a trained model can be used in which the learning model is trained using radiographic images to which correct answer data for the rotation angle has been added, and the rotation angle can be obtained by inputting radiographic images of the same bone portion captured in a measurement target into the trained model.

[0033] In this embodiment, the rotation angle is obtained from a change in the shape of the lesser trochanter 16, which is a part of the femur 15 that appears in the radiographic X-ray image 14. As the change in shape, for example, the lesser trochanter distance can be used, which is the distance between the apex of the lesser trochanter 16 and a tangent line that contacts the inner side of the femur that is included in the femur 15, based on the local image of the lesser trochanter 16 that appears in the X-ray image 14.

[0034] As shown in FIG. 3, specifically, the lesser trochanter distance a is determined by drawing a tangent 18 to the inside of the femur near the part where the lesser trochanter 16 is formed in an X-ray image 14, and drawing a perpendicular line 19 from the apex of the lesser trochanter 16 to the tangent line 18. The distance from the apex of the lesser trochanter to the intersection of the perpendicular line 19 and the tangent line 18 is defined as the lesser trochanter distance a. The lesser trochanter distance a is the distance between the apex of the lesser trochanter 16 and the tangent line 18.

[0035] As shown in Figure 4(B), when the femur 15 is in the reference position, the lesser trochanter distance a is a2, whereas as shown in Figure 4(A), when the femur 15 is internally rotated 20° from the reference position, the lesser trochanter distance is a1, which is smaller than a2. On the other hand, as shown in Figure 4(C), when the femur 15 is externally rotated 20° from the reference position, the lesser trochanter distance is a3, which is larger than a2.

[0036] The lesser trochanter distance a measured based on the X-ray image 14 can be converted to a rotation angle by using pre-created rotation angle correspondence information that corresponds the lesser trochanter distance a to a rotation angle. The rotation angle correspondence information may be a relational expression obtained by performing polynomial approximation or the like on the lesser trochanter distance a and the rotation angle actually measured in the X-ray image 14 taken when the femur 15 is at a known rotation angle, or may be an LUT (Look Up Table) or the like. In this embodiment, the relational expression is used. The inventors have discovered that the rotation angle correspondence information makes it possible to convert the lesser trochanter distance a to a rotation angle while minimizing individual differences. That is, the subject measured to obtain the rotation angle correspondence information and the subject appearing in the radiographic image whose rotation angle is to be determined using the rotation angle correspondence information may be the same or different. Therefore, it is preferable to convert the lesser trochanter distance a to a rotation angle using the rotation angle correspondence information.

[0037] As shown in FIG. 5, the rotation angle correspondence information can be expressed as relational expression 22, which is a regression curve obtained by regression analysis of the measured lesser trochanter distance a and the rotation angle. Graph 21 plots data on multiple measured lesser trochanter distances a and rotation angles, with the vertical axis representing the lesser trochanter distance a and the horizontal axis representing the rotation angle. Specifically, the actual measurements were performed on multiple X-ray images 14 of the same femur 15, taken under the same conditions except for varying the rotation angle to a predetermined number of values. The vertical and horizontal axes represent the difference from a reference position. That is, the vertical axis represents the difference Δ (unit: mm) between the lesser trochanter distance a when the rotation angle is 0° as the reference (0 mm) and the measured lesser trochanter distance a, and the horizontal axis represents the rotation angle (unit: °) from the reference position, with the rotation angle of 0° as the reference position. A regression curve was obtained by regression analysis of these data, which was expressed as relational expression 22. Note that, for example, when the difference Δ in the lesser trochanter distance a is y (mm) and the rotation angle is x (°), Relational Expression 22 becomes the following Expression (1).

[0038] y=0.11x-0.27 (1)

[0039] Furthermore, relational expression 22 is the coefficient of determination R 2 was 0.96, which accurately showed the relationship between the lesser trochanter distance a and the rotation angle. Furthermore, from this relational expression 22, it was found that for both internal and external rotation, a difference Δ in the lesser trochanter distance a of approximately 1 mm occurs for every 10° of internal rotation angle.

[0040] The bone mass acquisition unit 13 converts the pixel value of each pixel of the radiographic image acquired by the analysis image acquisition unit 11 into bone mass information to obtain pre-correction bone mass information. Then, the bone mass information is obtained by correcting the pre-correction bone mass information based on the rotation angle acquired by the rotation angle acquisition unit 12. The bone mass calculated using the acquired bone mass information is regarded as the bone mass measured using the analysis device 10. In the correction, the pre-correction bone mass information is converted so that it becomes bone mass information equivalent to a rotation angle of 0°.

[0041] Bone mass information is information that indicates bone mass and is data that corresponds to the bone mass that is a measurement result. Therefore, by using the bone mass information, the bone mass that is a measurement result can be calculated. The bone mass that is a measurement result is expressed as BMD (unit: g / cm 2 ), bone mass can be said to be the absolute value of bone mass. Bone mass information can be any data that can be used to calculate bone mass, which is the measurement result, and may be a relative value representing bone mass or other value, or an absolute value representing bone mass. In this embodiment, bone mass information is information that represents the absolute value of bone mass obtained by measurement.

[0042] One method for converting pixel values ​​into bone mass information is to obtain a conversion coefficient for each pixel to convert the pixel values ​​of the bone area of ​​the bone shown in the X-ray image 14 into bone mass information by referring to an LUT in which pixel values ​​correspond to bone mass information, in order to perform correction based on imaging conditions such as tube voltage, and then multiply the pixel value by the conversion coefficient to obtain uncorrected bone mass information.

[0043] Next, the inventors discovered that, as a method for correcting pre-correction bone mass information based on the rotation angle, bone mass at the reference position can be accurately obtained by using correction coefficient correspondence information, in which the rotation angle corresponds to a correction coefficient for correcting the pre-correction bone mass information. The correction coefficient correspondence information may be a relational expression obtained by performing polynomial approximation or the like on the measured pre-correction bone mass information and the rotation angle, or may be an LUT or the like. When using a relational expression between the rotation angle and bone mass information obtained in advance, the relational expression is preferably obtained by polynomial approximation of the relationship between the rotation angle and the bone mass information obtained by actual measurement. This relational expression enables correction of the pre-correction bone mass information to bone mass information equivalent to a rotation angle of 0° while suppressing individual differences. The bone mass 17 based on the bone mass information obtained in this way was approximately the same as the bone mass measured when the image was captured at a rotation angle of 0°.

[0044] As shown in Figure 6, the correction coefficient correspondence information is based on the measured BMD (unit: g / cm 2) and the rotation angle, a polynomial approximation can be performed to obtain relational expression 32. Graph 31 plots multiple measured BMD and rotation angle data, with the vertical axis representing BMD and the horizontal axis representing rotation angle. Specifically, the actual measurements were performed on multiple X-ray images 14 of the same femur 15, acquired under the same conditions except for varying the rotation angle between a predetermined number of values. The bone mass was calculated by converting the pixel values ​​for each pixel of each image. Equation 32a represents the value of human bone a, and Equation 32b represents the value of human bone b, where human bone a and human bone b are bones of different individuals. When there is no need to distinguish between Equation 32a and Equation 32b, they are referred to as Equation 32. The vertical axis represents the measured value, and the horizontal axis represents the difference from the reference position. Negative values ​​represent the rotation angle due to internal rotation, and positive values ​​represent the rotation angle due to external rotation. That is, the horizontal axis represents the rotation angle (unit: °) from the reference position, with the rotation angle being 0°. A regression curve was obtained by regression analysis of this data, and this was set as Relational Formula 32. Relational Formula 32 is a polynomial, and it was possible to approximate it with the same polynomial even for different people.

[0045] As described above, analysis device 10 of the present invention can use X-ray image 14 taken without strict positioning adjustment to calculate bone mass equivalent to that obtained when the image is taken at a fixed position, taking into account the influence of rotation of femur 15. For example, when the bone part is femur 15, even if the rotation angle of femur 15 shown in X-ray image 14 varies, it is possible to determine the bone mass obtained when the image is taken at a standard position where the rotation angle of femur 15 is 0°. Therefore, analysis device 10 of the present invention can easily measure bone mass based on a fixed standard using X-ray image 14.

[0046] An example of processing by the analysis device 10 of the present invention will be described using the flowchart shown in FIG. 7. The analysis device 10 acquires an X-ray image 14 obtained by simple X-ray imaging of a subject including the proximal femur of the femur 15 (step ST110). Next, to obtain the rotation angle, the lesser trochanter distance a of the lesser trochanter 16, which is a bone portion, is obtained based on the X-ray image 14 (step ST120). Next, the rotation angle from the reference position of the femur 15 is obtained based on the X-ray image 14 (step ST130). In this embodiment, the rotation angle is obtained by using rotation angle correspondence information in which the lesser trochanter distance a corresponds to the rotation angle.

[0047] Next, pre-correction bone mass information of the femur 15 is obtained by converting the pixel value for each pixel of the X-ray image 14 (step ST140). In this embodiment, the pixel values ​​are converted using a previously acquired LUT. Next, a correction coefficient for correcting the pre-correction bone mass information is obtained based on the rotation angle (step ST150). In this embodiment, the correction coefficient is obtained using correction coefficient correspondence information in which the rotation angle corresponds to the correction coefficient for correcting the pre-correction bone mass information.

[0048] Then, the pre-correction bone mass information is corrected by multiplying the pre-correction bone mass information by the correction coefficient (step ST160). The corrected value of the pre-correction bone mass information is the bone mass information finally obtained by measuring the bone mass at the reference position based on the radiographic image (step ST170).

[0049] Bone mass measurements, such as those performed by DXA, are used to diagnose osteoporosis and measure the lumbar vertebrae and femur, which are prone to fragility fractures. The measured bone mass is a two-dimensional projection (unit: g / cm 2), there is a problem that measurement values ​​can fluctuate depending on the patient's posture. The outer cortical bone has more bone mass than the inner cancellous bone, and the amount of overlap of the cortical bone projected onto a two-dimensional surface changes as the bone rotates, resulting in greater fluctuations in measurement values. This has a particularly large impact on measurements of the femur, which is prone to rotational movement. To ensure accurate, repeatable measurements, radiologists use fixation devices and check pre-shot images to position the patient so that the bone is directly in front of the incident X-rays.

[0050] However, positioning adjustment can take time, and desirable positioning may not be possible for some patients. Furthermore, when diagnosing osteoporosis, etc., to obtain changes in bone mass from the same patient's past results, it is preferable to use radiographic images taken with the same positioning. Therefore, radiographic images taken with improper positioning may be rejected and require retaking. Furthermore, DXA testing, which has high measurement accuracy, requires specialized equipment, making it difficult to perform the test easily.

[0051] The analysis device 10 of the present invention can accurately obtain bone mass when radiographic images are taken without precise positioning, equivalent to a rotation angle of 0 degrees. This reduces the burden of positioning adjustment. It also prevents an increase in radiation exposure due to retaking images due to poor positioning.

[0052] Furthermore, because it can use plain radiographic images, there is no need for specialized equipment for DXA or other imaging, and existing facilities can be used. Furthermore, past data from radiographic images taken with existing equipment that had inappropriate positioning adjustments can be utilized. Furthermore, because there is no need for dedicated fixation devices for bone densitometry or internally rotated positioning for bone densitometry, it can be used in conjunction with radiographic imaging for other purposes, such as X-rays, contributing to the efficiency of examinations.

[0053] The analysis device 10 can also be preferably applied when dedicated equipment for performing imaging such as DXA is available. Even for radiographic images obtained by DXA, it is possible to calculate measurement values ​​corrected for the influence of bone rotation, making it possible to obtain bone mass measurements based on a fixed standard, i.e., at a fixed rotation angle.

[0054] Next, a radiation analyzer including the analysis device 10 of the present invention image An example of an embodiment of an imaging system etc. will be described. image The imaging system uses X-rays to capture the subject Obj. covered An X-ray image 14 of the object Obj is acquired, and bone mass is obtained based on the X-ray image 14.

[0055] 8, the radiation image capturing system 40 includes a radiation source 42 which is a radiation generating unit, a radiation capturing unit 43, and a console 41. The radiation capturing unit 43 detects radiation and generates radiation image data.

[0056] The console 41 is a main control device of the radiation image capturing system 40, and includes an analysis device (radiation analysis device) 10, an input unit 47, a display 48, and a PACS (Personal Assisted Communication System) for managing captured X-ray images 14 and information about the X-ray images 14. 50 Image servers such as the Picture Archiving and Communication System for medical applications, and HIS that registers and manages information such as patient information, medical information, examination information, accounting information, and imaging orders for each patient. 51 (Hospital Information System) and RIS 49 It connects to an information management server such as a Radiology Information System.

[0057] The radiographic imaging system 40 captures two X-ray images 14, consisting of a first radiographic image G1 and a second radiographic image G2, using radiation with different energy distributions. The two X-ray images 14 with different energy distributions can be captured by alternately irradiating the radiographic imaging unit 43 consisting of a single detector with X-rays having different energy distributions that have passed through the subject Obj, or by irradiating the radiographic imaging unit 43 consisting of two stacked detectors with X-rays that have passed through the subject Obj. This is known as the one-shot energy subtraction method. The one-shot energy subtraction method allows radiographic images of two different energy components to be recorded by irradiating the subject with radiation once (one shot). In this embodiment, the one-shot energy subtraction method is used. One or two of the captured two radiographic images are used as the X-ray images 14 for obtaining bone mass.

[0058] The radiation source 42 generates radiation used for radiography. In this embodiment, the radiation source 42 is an X-ray source that generates X-rays Ra. The object Obj is a human, and includes the proximal femur of the human femur 15.

[0059] The radiography unit 43 includes a first radiation detector 44 and a second radiation detector 45. Each of the first radiation detector 44 and the second radiation detector 45 is, for example, an FPD (Flat Panel Detector). The FPD detects X-rays Ra that have passed through the object Obj and converts them into electrical signals, thereby outputting an X-ray image 14 of the object Obj.

[0060] When the first radiation detector 44 and the second radiation detector 45 receive X-rays that have been emitted from the radiation source 42 and passed through the subject Obj, the first radiation detector 44 and the second radiation detector 45 each receive the X-rays with different energies. During imaging, the first radiation detector 44, an X-ray energy converting filter 46 made of a copper plate or the like, and the second radiation detector 45 are arranged in this order from the side closest to the radiation source 42, and the radiation source 42 is driven. It is preferable that the first radiation detector 44 and the second radiation detector 45 are placed as close as possible to the X-ray energy converting filter 46.

[0061] As a result, the first radiation detector 44 obtains a first radiographic image G1 of the subject Obj using low-energy X-rays including so-called soft rays. The second radiation detector 45 obtains a second radiographic image G2 of the subject Obj using high-energy X-rays from which the soft rays have been removed. The first radiographic image G1 and the second radiographic image G2 are received by the console 41.

[0062] The first radiation detector 44 and the second radiation detector 45 can repeatedly record and read out radiation images and may be so-called direct type radiation detectors that generate electric charges upon direct exposure to radiation. Alternatively, the first radiation detector 44 and the second radiation detector 45 may be indirect type radiation detectors that convert radiation into visible light and then convert the visible light into electric charge signals. The radiation image signal is preferably read out by turning a TFT (Thin Film Transistor) switch on and off, a so-called optical readout method.

[0063] The console 41 is a main control device (so-called computer) of the radiographic imaging system 40, and is, for example, a personal computer or a computer such as a workstation on which an application program for executing predetermined functions is installed. The analysis device 10 is also, for example, a personal computer or a computer such as a workstation on which an application program for executing predetermined functions is installed. The computer is equipped with a CPU (Central Processing Unit) which is a processor, memory, storage, etc., and realizes various functions by programs, etc. stored in the storage.

[0064] In this embodiment, the computer of the console 41 executes the functions of the main control device of the radiographic imaging system 40 and the functions of the analysis device 10. When the computer of the console 41 executes the functions of the analysis device 10, a radiographic image analysis unit (hereinafter referred to as the analysis unit) 10a included in the console 41 executes the functions of the analysis device 10. Therefore, the console 41 includes the analysis unit 10a. Note that the analysis device 10 may be a computer common to the computer of the console 41, as in this embodiment, or may be a computer other than the console 41. In other words, the analysis device 10 may be included in another device connected to the console 41, or may be a stand-alone device.

[0065] Specifically, the functions of the analysis unit 10a are realized by the operation of the computer of the console 41 by a radiographic image analysis program. For example, the computer of the console 41 stores the radiographic image analysis program in a storage device (not shown) or the like and executes it. The radiographic image analysis program causes the computer to execute the following functions: acquire a radiographic image obtained by simply capturing an image of a subject including a bone; acquire a rotation angle of the bone from a reference position based on the radiographic image; and calculate the bone mass of the bone by correcting, based on the rotation angle, pre-correction bone mass information of the bone obtained by converting pixel values ​​for each pixel of the radiographic image.

[0066] The input unit 47 connected to the console 41 is a user interface that receives operations related to processing of the imaging menu and analysis by the radiography and analysis unit 10a, and receives input from the user. The user interface can be a keyboard, a mouse, a touch panel, or the like (none of which are shown). The touch panel can be a display 48 connected to the console 41, and may accept various operations in cooperation with a GUI (Graphical User Interface) displayed on the display 48.

[0067] The display 48 connected to the console 41 displays the imaging menu sent from the RIS, a GUI screen for instructing or controlling imaging, the captured X-ray images 14 and measured bone mass, and a GUI for operating or controlling the analysis unit 10a, etc. When the display 48 is connected to various information management servers such as a RIS, it can communicate with the servers to display patient information and past X-ray images 14, etc. The console 41 also transmits the captured X-ray images 14 and analysis results by the analysis unit 10a to an image server such as a PACS or various information management servers such as a RIS. The console 41 not only receives captured X-ray images 14 from the radiography unit 43, but may also receive previously captured X-ray images 14 stored in a server such as a RIS. In this case, the analysis unit 10a can analyze the previously captured X-ray images 14 to obtain bone mass based on the previous X-ray images 14 and display the bone mass on the display 48, etc.

[0068] 9, the console 41 includes an image acquisition unit 52, an image processing unit 53, and an analysis unit 10a that functions as the analysis device 10. The image acquisition unit 52 acquires a radiographic image detected by a radiation detector of the radiography unit 43 by imaging. The image processing unit 53 performs image processing on the radiographic image acquired by the image acquisition unit 52.

[0069] In this embodiment, the image processing unit 53 acquires the captured first radiographic image G1 and second radiographic image G2, processes them using appropriate coefficients, and then extracts bones or soft tissues by subtraction to generate a bone tissue image or soft tissue image. Alternatively, the bone tissue image and the soft tissue image may be combined to generate a composite image in which the bone tissue image and the soft tissue image are separated and easily visible. Therefore, the X-ray image 14 after image processing by the image processing unit 53 includes at least one of a bone tissue image, a soft tissue image, and a composite image. In this embodiment, the bone tissue image generated by the image processing unit 53 is used as the X-ray image 14 after image processing to be input to the analysis unit 10a.

[0070] As shown in FIG. 10, the image processing unit 53 performs image processing on the acquired first radiographic image G1 and second radiographic image G2 using an energy subtraction method. The radiographic image G1 and second radiographic image G2 are plain X-ray images. A bone tissue image 14b is generated by image processing that extracts bones by utilizing the characteristic difference between the two X-ray energies used when the first radiographic image G1 and the second radiographic image G2 were acquired. Similarly, a soft tissue image 14c is generated by image processing that extracts soft tissues. The generated bone tissue image 14b is input to the analysis image acquisition unit 11 (see FIG. 1) of the analysis unit 10a. Note that the bone tissue image 14b and the soft tissue image 14c may be combined to generate a composite image 14a in which the soft tissues and bones are separated and easily visible.

[0071] In some cases, in addition to the bone tissue image 14b, the composite image 14a may be input to the analysis image acquisition unit 11. Therefore, the X-ray image 14 acquired by the analysis image acquisition unit 11 includes the bone tissue image 14b and the composite image 14a in addition to the X-ray image 14 that has not been subjected to image processing for analysis.

[0072] If an anti-scatter grid that removes scattered components of X-rays that have passed through the subject Obj is used when capturing an image of the subject Obj, the first radiographic image G1 and the second radiographic image G2 will contain the primary ray components of the X-rays that have passed through the subject Obj. On the other hand, if an anti-scatter grid is not used when capturing an image of the subject Obj, the first radiographic image G1 and the second radiographic image G2 will contain the primary ray components and scattered ray components of the X-rays.

[0073] Furthermore, if the X-ray image 14 contains scattered radiation components, it is preferable to estimate and remove the scattered radiation generated for each thickness based on the body thickness distribution of the subject Obj. The X-ray image 14 from which the scattered radiation is estimated and removed is preferably the first radiographic image G1 or the second radiographic image G2 obtained by imaging and not yet subjected to image processing such as bone extraction. This is because more preferable image processing results can be obtained by performing various image processing such as bone extraction after the scattered radiation has been removed.

[0074] As shown in FIG. 11 , in this case, the console 41 includes a body thickness distribution acquisition unit 54. The thickness distribution of the object Obj may be a measured value or an estimated value. Therefore, the body thickness distribution acquisition unit 54 acquires a measured value of the thickness distribution of the object Obj, or an estimated value of the thickness distribution of the object Obj. The body thickness distribution acquisition unit 54 estimates scattered rays for each pixel based on the acquired measured or estimated body thickness distribution, and removes the scattered rays from the first radiographic image G1 and the second radiographic image G2, respectively. The first radiographic image G1 and the second radiographic image G2 from which scattered rays have been removed are input to the image processing unit 53 as the first radiographic image G1 and the second radiographic image G2 after image processing, such as bone extraction.

[0075] The X-ray image 14 from which scattered rays are removed is not limited to either the first radiographic image G1 or the second radiographic image G2, but can also be any of the bone tissue image 14b, the soft tissue image 14c, the composite image 14a, etc. Since the first radiographic image G1 and the second radiographic image G2 are X-ray images 14 obtained by the one-shot energy subtraction method, the body thickness distribution of the subject Obj can be estimated based on either the first radiographic image G1 or the second radiographic image G2, and the estimated value can be used as the body thickness distribution of the second radiographic image G2.

[0076] After performing scattered radiation removal on the first radiographic image G1 or the second radiographic image G2, which have not been subjected to image processing, image processing for bone extraction or soft tissue extraction is performed to generate a bone tissue image 14b or a soft tissue image 14c, etc. Therefore, image processing such as bone extraction is performed while suppressing the effects of scattered radiation, so that a more accurate bone tissue image 14b or soft tissue image 14c can be generated.

[0077] 12, the body thickness distribution acquisition unit 54 includes a body thickness distribution measurement value acquisition unit 61, a body thickness distribution estimated value acquisition unit 62, and a scattered radiation removal unit 63. The body thickness distribution acquisition unit 54 acquires the measured body thickness distribution. The body thickness distribution measurement value acquisition unit 61 acquires measurement values ​​obtained by actually measuring the body thickness distribution, and sets these as the body thickness distribution of the subject Obj.

[0078] The body thickness distribution estimated value acquisition unit 62 acquires an estimate of the thickness distribution of the subject using a known method. As a method for acquiring an estimate of the thickness distribution or a method for performing image processing for scattering removal using the acquired estimate of the thickness distribution, it is preferable to employ, for example, a method using a virtual model as described in Japanese Patent Application Laid-Open No. 2015-043959.

[0079] The method using a virtual model first acquires a virtual model having a predetermined body thickness distribution, and generates an estimated image by combining an estimated primary ray image and an estimated scattered ray image obtained by radiography of the virtual model. Next, the body thickness distribution of the virtual model is corrected so as to reduce an error value representing the difference in pixel values ​​of pixels at corresponding positions between the subject image obtained by radiography of the subject Obj and the estimated image. The corrected body thickness distribution of the virtual model is determined as the body thickness distribution of the subject. In this way, even for a subject imaged without using a grid, the influence of scattered rays can be suppressed and a more accurate estimate of the body thickness distribution can be obtained.

[0080] The scattered radiation removal unit 63 acquires the measured or estimated values ​​of the body thickness distribution of the subject Obj as the body thickness distribution of the subject Obj, and removes scattered radiation from the X-ray image 14 based on the acquired body thickness distribution. A known method can be used to remove scattered radiation, and for example, an estimated scattered radiation image estimated from the body thickness distribution may be subtracted from the captured X-ray image 14 to generate an X-ray image 14 from which scattered radiation components have been removed.

[0081] The X-ray image 14 obtained as described above is sent to the analysis image acquisition unit 11 of the analysis unit 10a. In this embodiment, the X-ray image 14 acquired by the analysis image acquisition unit 11 is a bone tissue image 14b obtained by performing image processing on the first radiographic image G1 and the second radiographic image G2, which have been subjected to image processing for removing scattered rays. The analysis image acquisition unit 11 sends the acquired bone tissue image 14b to the rotation angle acquisition unit 12, and the rotation angle acquisition unit 12 acquires the rotation angle of the bone from a reference position based on the bone tissue image 14b.

[0082] In addition, the rotation angle acquisition unit 12 may obtain the rotation angle from changes in the shape of the lesser trochanter 16, which is a part of the femur 15 in the bone tissue image 14b, or may use machine learning technology to obtain the rotation angle from a trained model trained on radiographic images to which correct data for the rotation angle has been added.

[0083] As shown in Fig. 13, in this case, the rotation angle acquisition unit 12 (see Fig. 1) includes a trained model 71. The trained model 71 is a mathematical formula that determines parameters in a training model consisting of a preset algorithm, and the parameters are adjusted so that the trained model 71 can estimate and output an estimation result related to the rotation angle by inputting input information having a bone tissue image 14b, which is an X-ray image 14. Therefore, the trained model 71 has parameters for outputting an estimation result related to the rotation angle based on input information having a bone tissue image 14b.

[0084] The pre-set learning model is preferably one that can output favorable results for input information consisting of an image. Therefore, it is preferable to use a multilayer neural network, particularly a convolutional neural network (CNN), which is considered suitable for image recognition, as the learning model, and it is preferable to adopt deep learning. These learning models are input with bone tissue images 14b to which correct rotation angle data is added, and learning is performed to adjust parameters so that the correct rotation angle is output. The parameters are adjusted by repeating learning multiple times, resulting in a trained model 71 in which the estimated and output rotation angle when the bone tissue image 14b is input is approximately the correct rotation angle. When generating the trained model 71, it is preferable to generate and use a trained model 71 for each type of input X-ray image 14. The type of input X-ray image 14 can be distinguished by the type of bone shown in the X-ray image 14, the type of image processing applied to the X-ray image 14, or the like.

[0085] Furthermore, when learning a learning model or outputting estimation results related to the rotation angle using the trained model 71, techniques used in machine learning can be appropriately adopted to obtain desirable estimation results, such as using multiple learning models, devising processing for the bone tissue image 14b to be trained, or suppressing overlearning.

[0086] As shown in FIG. 14, trained model 71 is obtained by adjusting parameters by using CNN to train on bone tissue image 14b showing femur 15 and ground truth data on rotation angle. By inputting bone tissue image 14b showing femur 15 with an unknown rotation angle into trained model 71, for example, a rotation angle of minus 10° is output as estimation result 72. That is, femur 15 with an unknown rotation angle shown in bone tissue image 14b input to trained model 71 is estimated to have a rotation angle of minus 10°.

[0087] The rotation angle acquisition unit 12 acquires, as the rotation angle, the estimated result output by using the trained model 71. The acquired rotation angle is used to correct the pre-correction bone mass information using the correction coefficient correspondence information, just as in the case where the rotation angle is acquired from the shape of the lesser trochanter 16.

[0088] Next, the bone mass acquisition unit 13 converts the pixel value of each pixel in the bone tissue image 14b to obtain pre-correction bone mass information. The bone mass acquisition unit 13 (see FIG. 1) acquires bone mass information for each pixel in the bone region included in the bone tissue image 14b. The bone mass information acquired here is used as pre-correction bone mass information.

[0089] To correct for the effect of changes in contrast of X-ray image 14 due to imaging conditions, bone mass acquisition unit 13 acquires bone mass information by converting pixel values ​​of the bone region included in bone tissue image 14b into pixel values ​​of the bone region acquired under standard imaging conditions. This correction allows bone mass to be calculated as if the X-ray image 14 was acquired under standard imaging conditions, even for X-ray image 14 acquired under different imaging conditions, allowing for more accurate comparison or diagnosis.

[0090] The influence of imaging conditions includes tube voltage and body thickness. Regarding tube voltage, the higher the tube voltage applied to the radiation source 42 and the higher the energy of the X-rays, the smaller the contrast between soft tissue and bone in the X-ray image 14. Furthermore, in the process of X-rays passing through the subject Obj, low-energy components of the X-rays are absorbed by the subject Obj, causing beam hardening, in which the X-rays become more energetic. The increase in the energy of the X-rays due to beam hardening becomes greater as the body thickness of the subject Obj increases.

[0091] The bone mass acquisition unit 13 calculates bone mass based on the body thickness distribution, pixel values ​​of the bone region, and the imaging conditions when the body thickness distribution and the radiographic image were acquired. As shown in FIG. 15 , the bone mass acquisition unit 13 refers to the LUT 80 and acquires a conversion coefficient Cb(x, y) for each pixel to convert the pixel values ​​of the bone region into bone mass information. It is assumed that the body thickness of the subject is associated with each position on the xy plane, and the body thickness distribution of the subject is T(x, y). The conversion coefficient Cb(x, y) is determined based on the imaging conditions and the body thickness distribution T(x, y). Then, as shown in the following equation (2), the pixel value Gb(x, y) of each pixel of the bone region is multiplied by the conversion coefficient Cb(x, y) to acquire bone mass information B(x, y) for each pixel of the bone region.

[0092] B(x, y)=Cb(x, y)×Gb(x, y) (2)

[0093] LUT80 defines the relationship between the body thickness included in the body thickness distribution and the conversion coefficient Cb(x, y). As shown in this relationship, the higher the tube voltage included in the imaging conditions and the greater the body thickness, the larger the value of the conversion coefficient. Even if the imaging conditions vary, calculating bone mass information when imaging under standard imaging conditions allows for more accurate comparison or diagnosis using bone mass. Therefore, when the imaging condition tube voltage is 90 kV, which is the standard imaging condition, the conversion coefficient Cb(x, y) is set to "1" when the thickness is "0."

[0094] Similarly, when the imaging condition is a tube voltage of 100 kV, which is higher than the reference imaging condition, and the body thickness is "0," the conversion coefficient Cb(x, y) is larger than "1." This is because the higher the tube voltage, the lower the contrast between bones and soft tissue. In this embodiment, the pixel values ​​of bones are corrected using the conversion coefficient to address the decrease in contrast. Furthermore, when the imaging condition is a tube voltage of 80 kV, which is lower than the reference imaging condition, and the body thickness is "0," the conversion coefficient Cb(x, y) is smaller than "1." This is because the lower the tube voltage, the higher the contrast between bones and soft tissue. In this embodiment, the pixel values ​​of bones are corrected using the conversion coefficient to address the increase in contrast.

[0095] Furthermore, it is preferable that the pre-correction bone mass information be calculated by calculating cortical bone mass information and trabecular bone mass information for each of the cortical bone and the cancellous bone of the bone portion. If the bone portion does not include either cortical bone or cancellous bone, either the cortical bone mass information or the cancellous bone mass information may be calculated. As shown in FIG. 16 , in this case, the bone mass acquisition unit 13 includes a bone composition recognition unit 81. Based on the bone tissue image 14b, the bone composition recognition unit 81 distinguishes between a cortical bone region where cortical bone is captured and a cancellous bone region where cancellous bone is captured in the bone tissue image 14b. Known methods can be used to distinguish and recognize the cortical bone region and the cancellous bone region, and specifically, a method of performing a filtering process for region identification can be used.

[0096] As shown in FIG. 17, in the bone tissue image 14b, a cortical bone region 82 covering the outside of the femur 15 is distinguished from a remaining trabecular bone region 83. The bone mass acquisition unit 13 calculates cortical bone mass information and trabecular bone mass information based on the cortical bone region 82 and the trabecular bone region 83. The calculated cortical bone mass information and trabecular bone mass information are set as pre-correction cortical bone mass information and pre-correction trabecular bone mass information, respectively, and are corrected based on the determined rotation angle to obtain the final measured values ​​of the cortical bone mass information and trabecular bone mass information. The method of correction based on the rotation angle is the same as described above.

[0097] Measurement of cortical bone mass and trabecular bone mass in bones is important for diagnosis and treatment. For example, methods for assessing osteoporosis or fracture risk, such as hip structural analysis (HSA) or trabecular bone structure analysis (TBS), are used. HSA uses cortical bone mass, and TBS uses trabecular bone structure. Therefore, X-ray images 14 are preferred because they can easily measure cortical bone mass and trabecular bone mass information based on a set standard, which leads to simple and accurate evaluations based on each method.

[0098] Furthermore, when converting pixel values ​​to bone mass information or bone mass information to bone mass, calibration may be performed using a phantom with a known bone composition before conversion. It is preferable to perform calibration before capturing the X-ray image 14, and obtain a correction coefficient for correcting the relationship between pixel values ​​and bone mass or the relationship between bone mass information and bone mass.

[0099] It is preferable to use various types of phantoms depending on the application. It is also preferable to use a phantom whose composition is distinguished between cortical bone and cancellous bone, because this allows for more detailed calibration by separating the cortical bone and cancellous bone.

[0100] A case where a phantom is used when converting bone mass information into bone mass will be described. The analysis unit 10a acquires bone mass information based on an X-ray image 14 obtained by performing radiography on a phantom having a known bone mass. Here, the acquired bone mass information is bone mass information of a subject whose bone mass is known, and therefore is used as a reference value for bone mass information.

[0101] Next, bone mass correspondence information is obtained by matching the known bone mass with a reference value of the bone mass information. For example, the absolute value of bone mass can be calculated by using the bone mass information obtained from the X-ray image 14 and the bone mass correspondence information. A more accurate measurement value of bone mass can be obtained by using a phantom selected to enable more accurate calculation of bone mass depending on the type of subject, etc. Therefore, the method using the bone mass correspondence information is effective whether the bone mass information is an absolute value, a relative value, or some other value.

[0102] In the above embodiment, the femur 15 was used as the bone portion. However, even when the target bone portion is the forearm or lumbar vertebrae, etc., which is used for the examination or diagnosis of osteoporosis, the analysis device 10 can similarly measure bone mass based on a fixed standard by correcting the pixel values ​​of the X-ray image 14 based on various positioning to the pixel values ​​in the X-ray image 14 obtained by positioning at a reference position.

[0103] Furthermore, in the above embodiment, a trained model is used to obtain the rotation angle of the X-ray image 14, but a trained model may also be used to obtain bone mass from pixel values.

[0104] In the above embodiment, a single radiographic image may be subjected to a filtering process for region identification to identify bone regions and soft tissue regions, or to identify cortical bone and cancellous bone. For example, if a soft tissue region has a specific spatial frequency, it is preferable to identify the soft tissue region by a frequency filtering process that extracts a range of the specific spatial frequency. Alternatively, if soft tissue regions contain many low-frequency regions and it is difficult to identify the soft tissue region by a frequency filtering process for extracting the soft tissue region, the radiographic image may be subjected to a frequency filtering process to extract bone regions with higher frequencies than the soft tissue region, and the frequency-filtered radiographic image may then be subtracted from the original radiographic image to identify the soft tissue region. The same applies to cortical bone and cancellous bone.

[0105] In addition, in the above embodiment, the pre-correction bone mass information is corrected based on the rotation angle, but the pixel value of each pixel in the acquired X-ray image 14 may be corrected based on the rotation angle, and the corrected pixel value may then be converted into bone mass information of the bone portion.

[0106] Although the above embodiment uses the proximal femur as the bone, other bones may be used. For example, using a bone used in diagnosing osteoporosis, or a bone that is prone to fracture, such as the spine or the distal radius, is preferable because it allows for more accurate testing of osteoporosis over time. Specifically, to measure the bone density of the distal radius, the rotation angle of the distal radius may be calculated using the positional relationship between the radius and ulna.

[0107] In the above embodiment, the hardware structure of processing units such as the analytical image acquisition unit 11, rotation angle acquisition unit 12, and bone mass acquisition unit 13 included in the analysis device 10, or the image acquisition unit 52, image processing unit 53, radiographic image analysis unit 10a, and body thickness distribution acquisition unit 54 included in the console 41, is various processors as shown below. The various processors include a CPU (Central Processing Unit), which is a general-purpose processor that executes software (programs) to function as various processing units, a programmable logic device (PLD), such as an FPGA (Field Programmable Gate Array), whose circuit configuration can be changed after manufacture, and a dedicated electrical circuit, which is a processor with a circuit configuration designed specifically for executing various processes.

[0108] A single processing unit may be configured with one of these various processors, or may be configured with a combination of two or more processors of the same or different types (for example, multiple FPGAs, or a combination of a CPU and an FPGA). Also, multiple processing units may be configured with a single processor. Examples of multiple processing units configured with a single processor include, first, a configuration in which one processor is configured with a combination of one or more CPUs and software, as typified by client or server computers, and this processor functions as multiple processing units. Second, a configuration in which a processor is used to realize the functions of an entire system including multiple processing units on a single IC (Integrated Circuit) chip, as typified by a System on Chip (SoC). In this way, the various processing units are configured with one or more of the above-mentioned various processors as a hardware structure.

[0109] Furthermore, the hardware structure of these various processors is, more specifically, an electric circuit in the form of a combination of circuit elements such as semiconductor elements. [Explanation of symbols]

[0110] 10. Radiation image analysis device 10a Radiation Image Analysis Unit 11 Image acquisition unit for analysis 12 Rotation angle acquisition section 13 Bone mass acquisition department 14 X-ray image 15 Femur 15a Area where white appears dark 15b Area where white appears pale 16 Lesser trochanter 17 Bone mass 18 tangent 19 Perpendicular 21 graphs 22 Relational Expressions 31 graphs 32, 32a, 32b Relations 40 Radiation imaging system 41 Console 42 Radiation source 43 Radiography Department 44 First Radiation Detector 45 Second Radiation Detector 46 X-ray energy conversion filter 47 Input section 48 Display 49RIS 50 PACS 51 HIS 52 Image acquisition unit 53 Image processing section 54 Body thickness distribution acquisition part 61 Body thickness distribution measurement acquisition unit 62 Body thickness distribution estimation unit 63 Scattered radiation removal section 71 trained models 72 Estimation results 80 LUT 81 Bone composition recognition section 82 Cortical bone area 83 Cancellous bone area a, a1, a2, a3 distance G1 First radiographic image G2 Second radiographic image Obj Subject Ra X-ray ST110~ST170 Step

Claims

1. A radiological image analysis device including a processor, The processor: obtaining a radiographic image obtained by radiographing a subject including a bone portion; obtaining a rotation angle of the bone portion from a reference position based on the radiographic image; A radiological image analysis device that acquires bone mass information of the bone portion by correcting pre-correction bone mass information of the bone portion, which is obtained by converting pixel values ​​for each pixel of the radiological image, based on the rotation angle.

2. The radiation image analyzing device according to claim 1 , wherein the radiation image is an image of the subject taken in a frontal position.

3. 3. The radiographic image analyzing device according to claim 1, wherein the bone portion is a proximal part of a femur.

4. the bone portion includes a predetermined portion, The radiographic image analyzing device according to claim 1 , wherein the processor acquires the rotation angle based on a local image of the local area shown in the radiographic image.

5. the bone portion is a right or left proximal femur; The radiographic image analyzing device according to claim 4 , wherein the region is the lesser trochanter.

6. The processor: a lesser trochanter distance, which is the distance between the apex of the lesser trochanter and a tangent line that contacts the inner surface of the femur included in the proximal femur, based on the local image; The radiographic image analyzing device according to claim 5 , wherein the rotation angle is acquired based on the lesser trochanter distance.

7. The processor: acquiring rotation angle correspondence information in which the lesser trochanter distance corresponds to the rotation angle; The radiographic image analyzing device according to claim 6 , wherein the rotation angle is acquired based on the lesser trochanter distance and the rotation angle correspondence information.

8. The processor: acquiring correction coefficient correspondence information in which the rotation angle corresponds to a correction coefficient for correcting the pre-correction bone mass information; 8. The radiographic image analyzing device according to claim 1, wherein the uncorrected bone mass information is corrected based on the rotation angle and the correction coefficient correspondence information.

9. a trained model capable of estimating and outputting an estimation result related to the rotation angle by inputting input information having the radiation image; the trained model has parameters for outputting the estimation result based on the input information, The radiological image analysis device according to claim 1 , wherein the processor acquires the rotation angle from the estimation result output by using the trained model.

10. 10. The radiographic image analysis device according to claim 1, wherein the processor, when acquiring the radiographic image, estimates scattered radiation components for each pixel based on the body thickness distribution of the subject, and removes the scattered radiation components from the radiographic image.

11. the bone portion includes cortical bone and cancellous bone, The processor: Recognizing the cortical bone region and the cancellous bone region of the bone portion shown in the radiographic image; Obtaining uncorrected cortical bone mass information based on the region of the cortical bone, and obtaining uncorrected trabecular bone mass information based on the region of the trabecular bone; 11. The radiographic image analysis device according to claim 1, wherein the cortical bone mass information is obtained by correcting the pre-correction cortical bone mass information based on the rotation angle, and the trabecular bone mass information is obtained by correcting the pre-correction trabecular bone mass information based on the rotation angle.

12. The processor: obtaining a reference value of the bone mass information based on the radiographic image obtained by performing radiography on a phantom having a known bone mass; obtaining bone mass correspondence information in which the known bone mass corresponds to a reference value of the bone mass information; The radiological image analyzing device according to claim 1 , wherein the bone mass is acquired from the bone mass information based on the bone mass correspondence information.

13. acquiring a radiographic image obtained by performing radiography on a subject including a bone portion; acquiring a rotation angle of the bone portion from a reference position based on the radiographic image; and correcting pre-correction bone mass information of the bone portion, obtained by converting pixel values ​​for each pixel of the radiographic image, based on the rotation angle, to obtain bone mass information of the bone portion.

14. a function of acquiring a radiographic image obtained by performing radiography of a subject including a bone; a function of acquiring a rotation angle of the bone portion from a reference position based on the radiographic image; and a function of acquiring bone mass information of the bone portion by correcting, based on the rotation angle, pre-correction bone mass information of the bone portion obtained by converting pixel values ​​for each pixel of the radiographic image.

Images