Image analysis method, segmentation method, bone density measurement method, learning model generation method and image generation device

Abstract

The control unit (80) of the present invention comprises: an X-ray image generating unit (81); an X-ray image storage unit (82) storing X-ray images and training bone images for X-ray images; a DRR image generating unit (83) generating a DRR image of an area including bones; a DRR image storage unit (84) storing DRR images and training bone images for DRR images for machine learning; a learning unit (85) performing machine learning using X-ray images and training bone images for X-ray images, and performing machine learning using DRR images and training bone images for DRR images, thereby generating a learning model for identifying bones; and a bone image generating unit (86) converting an X-ray image of an area including bones of a subject using the learned model generated by the learning unit (85) to generate an image representing the bones.

Description

Technical Field

[0001] The present invention relates to an image analysis method, a segmentation method, a bone density measurement method, a learning model generation method and an image generation device. Background Art

[0002] In recent years, bone density measuring devices for measuring the bone density of subjects for the diagnosis of osteoporosis have been used. Patent document 1 discloses a bone salt quantitative analysis device, which is composed of the following mechanisms: a mechanism for generating radiation; a lattice irradiated with the radiation; a mechanism for aiming only radiation at two specified reflection angles among the radiation reflected by the lattice and irradiating the subject with two different energies of radiation at the same time; a radiation detection mechanism, the radiation of the two energies passing through the subject and entering the radiation detection mechanism; a wave height analysis mechanism, which separates the transmission data related to the radiation of each energy by performing wave height analysis output by the radiation detection mechanism; and a mechanism for calculating the bone density by performing arithmetic processing on the separated data.

[0003] Prior art literature

[0004] Patent Literature

[0005] Patent Document 1: Japanese Patent No. 2638875 Summary of the invention

[0006] Technical problem to be solved by the invention

[0007] Such bone density measurement targets the bone density of the lumbar spine and femur, which are clinically important. At this time, there are large individual differences in the shape of the femur, so it is important to determine the area of ​​the bone of the subject in order to implement stable tracking observation. In order to determine the area of ​​the bone, the operator must manually determine the area in advance, which not only makes the operation complicated, but also has the problem of deviation in the area determined by the operator.

[0008] In order to automatically perform segmentation for extracting the bone region from an image containing the bone of the subject, an algorithm using threshold processing based on a histogram may be considered, but it is difficult to correctly determine the bone region for segmentation of a femur, which has a particularly large individual difference in shape. Therefore, there is a problem that the accuracy of the final bone density measurement result is deteriorated.

[0009] The present invention is completed to solve the above-mentioned problems, and its purpose is to provide an image analysis method, a segmentation method, a bone density measurement method, a learning model generation method and an image generation device that can generate an image of an area containing an organ of a subject based on an X-ray image of an area containing an organ.

[0010] Solutions for solving the above technical problems

[0011] The invention described in Scheme 1 is an image analysis method, which performs segmentation for determining the region of an organ by analyzing an image of a region containing an organ of a subject, and is characterized in that machine learning is used as the segmentation method, and includes: a correction image generation step, which generates a correction image that changes the concentration of the region of the organ in the image containing the organ of the subject; and a learning model generation step, which generates a learning model for machine learning by using a learning process using an image containing the organ of the subject and the correction image generated by the correction image generation step.

[0012] The invention described in Option 2 is the invention described in Option 1, wherein an X-ray image of a region including an organ of the subject obtained by performing X-ray photography of the subject is converted using the learning model generated by the learning model generation process, thereby generating an image representing the organ.

[0013] The invention described in Scheme 3 is the invention described in Scheme 1, wherein the image of the region including the organ of the subject is a DRR image generated based on the CT image data of the subject, and in the correction image generation process, the region in which the CT value of the CT image data becomes a prescribed value is defined as the region of the organ and its concentration is changed.

[0014] The invention described in Scheme 4 is the invention described in Scheme 3, wherein when a DRR image is generated, at least one of the parameters of the projection coordinates and angles including the geometric conditions is changed, or at least one of image processing including rotation, deformation, and enlargement and reduction of the image is performed to generate multiple DRR images.

[0015] The invention described in claim 5 is the invention described in claim 3, wherein at least one of contrast change, noise addition, and edge enhancement is performed on the generated DRR image.

[0016] The invention described in Scheme 6 is the invention described in Scheme 1, wherein the image of the region including the organ of the subject is an X-ray image generated by performing X-ray photography of the subject, and in the correction image generation process, the image of the organ obtained using the X-ray image and dual-energy subtraction is used to change the concentration of the region of the organ.

[0017] The invention described in Scheme 7 is the invention described in Scheme 2, wherein an X-ray image of a region including an organ of the subject obtained by performing X-ray photography of the subject and an image representing the organ obtained by converting the learning model generated by the learning model generating step are used for learning the learning model performed by the learning unit.

[0018] The invention described in Scheme 8 is the invention described in Scheme 1, wherein the organ has a shape that is left-right symmetrical with respect to the body axis of the subject, and in the learning model generation process, by flipping either the image of the organ on the right side or the image of the organ on the left side left-right, a machine learning learning model is generated for the images of the left and right organs together.

[0019] The invention described in claim 9 is a method for segmenting an organ, wherein the organ is a bone of the subject, and the image analysis method described in claim 1 is used to segment the bone region.

[0020] The invention described in claim 10 measures bone density of the bone region segmented by the segmentation method described in claim 9.

[0021] The invention described in Scheme 11 is a learning model generation method, which generates a learning model used when performing segmentation for determining the region of the organ by analyzing an image of a region containing an organ of a subject using machine learning, and is characterized in that the learning model is generated by performing machine learning using an image containing the organ of the subject and a correction image, wherein the correction image is generated by changing the concentration of the region of the organ in the image containing the organ of the subject.

[0022] The invention described in Scheme 12 is an image generating device that generates an image of a region of an organ extracted from an X-ray image of a region containing an organ of a subject, and is characterized in that it comprises: an X-ray image storage unit that stores a plurality of X-ray images obtained by X-raying the region containing the organ, and a plurality of X-ray image training images for machine learning; a DRR image generating unit that generates a DRR image of the region containing the organ; a DRR image storage unit that stores a plurality of DRR images generated by the DRR image generating unit, and a plurality of DRR image training images for machine learning generated based on the DRR images generated by the DRR image generating unit; an image generating unit that converts the X-ray image of the region containing the organ of the subject using a learning model for identifying the organ, thereby generating an image representing the organ, the learning model being pre-generated by performing machine learning using the plurality of X-ray images stored in the X-ray image storage unit and the plurality of X-ray image teacher images, and performing machine learning using the plurality of DRR images stored in the DRR image storage unit and the plurality of DRR image training images.

[0023] The invention described in claim 13 is the invention described in claim 12, wherein the DRR image generating unit generates a part of the plurality of DRR images as DRR images in which the density of the organ region in the region including the organ is changed.

[0024] The invention described in Scheme 14 is the invention described in Scheme 12, wherein a portion of the multiple X-ray images stored in the X-ray image storage unit are X-ray images obtained by using dual-energy subtraction to change the concentration of the organ region in the region containing the organ.

[0025] Effects of the Invention

[0026] According to the invention described in claims 1 to 8, since the corrected image obtained by changing the density of the organ region of the subject is used for machine learning, a learning model corresponding to subjects with low organ density can be generated. Therefore, the detection accuracy of the organ can be improved.

[0027] According to the invention described in Scheme 4, since the parameters of the projection coordinates and angles including the geometric perspective conditions are changed or image processing including rotation, deformation, and enlargement and reduction of the image is performed, the position of the bone can be correctly detected even when there is a slight deviation in the position and posture of the subject or the position of the imaging system during X-ray imaging. Then, since a large number of DRR images can be generated, it is possible to learn a customized identifier corresponding to each patient, and further, the position of the bone can be correctly detected even when using DRR images with a low frame rate.

[0028] According to the invention described in Scheme 5, since contrast change, noise addition, and edge enhancement are performed on the generated DRR image, the position of the bone can be accurately detected even when there is a difference in image quality between the DRR image and the X-ray image.

[0029] According to the invention described in Scheme 7, by reusing multiple X-ray images and images representing organs obtained by converting the learned model in the learning of the learning model, the learning images can be expanded to generate a more accurate learning model.

[0030] According to the invention described in claim 8, the detection accuracy of organs having a shape that is symmetrical with respect to the body axis of the subject can be made more uniform. In addition, since machine learning is performed on the images of the left and right organs at the same time, the learning images can be expanded to generate a more accurate learning model.

[0031] According to the invention described in claim 9, the bone region can be segmented with high accuracy.

[0032] According to the invention of claim 10, the bone density can be measured for the segmented bone region.

[0033] According to the invention described in claim 11, since the corrected image obtained by changing the density of the region of the subject's organ is used for machine learning, it is possible to generate a learning model that corresponds to subjects whose organ density is low.

[0034] According to the invention described in Schemes 12 to 14, the extraction accuracy of the organ region can be improved by extracting the region by machine learning. In this case, since machine learning is performed using both X-ray images and DRR images, the learning images can be expanded and the collection of clinical data for learning can be easily performed. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 This is a schematic front view of a bone image generating device according to an embodiment of the present invention that also functions as an X-ray imaging device.

[0036] Figure 2 This is a side schematic diagram of a bone image generating device according to an embodiment of the present invention that also functions as an X-ray imaging device.

[0037] Figure 3 This is a block diagram showing a control system of the bone image generating device according to the embodiment of the present invention.

[0038] Figure 4 This is a schematic diagram for explaining a process of generating a bone image of a subject using machine learning by the bone image generating apparatus according to the embodiment of the present invention.

[0039] Figure 5 This is a flowchart showing the operation of the bone image generating apparatus according to the embodiment of the present invention to generate a bone image by extracting a bone region from an X-ray image of a region including the bone of a subject.

[0040] Figure 6 Schematic diagram of an X-ray image 101 generated by the X-ray image generating unit 81 .

[0041] Figure 7 This is a schematic diagram of an X-ray image training bone image 102 generated by the X-ray image generating unit 81 .

[0042] Figure 8 It is a schematic diagram showing the use of simulation Figure 1 1 is an explanatory diagram of a state in which a DRR image is generated by virtual projection of geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12 shown.

[0043] Fig. 9 Schematic diagram of the DRR image 103 generated by the DRR image generating unit 83 .

[0044] Fig.10This is a schematic diagram of a DRR image 104 generated by the DRR image generating unit 83 , in which the density of the bone region is changed to a smaller value.

[0045] Fig.11 This is a schematic diagram of a DRR image training bone image 105 generated by the DRR image generating unit 83 .

[0046] Fig.12 Schematic diagram of the X-ray image 106 generated by the X-ray image generating unit 81 .

[0047] Fig.13 Schematic diagram of the DRR image 107 generated by the DRR image generating unit 83 . DETAILED DESCRIPTION

[0048] Hereinafter, embodiments of the present invention will be described based on the drawings. Figure 1 is a front view schematic diagram of a bone image generating device according to an embodiment of the present invention that also functions as an X-ray imaging device. Figure 2 In this embodiment, a case where the present invention is applied to a bone image generating device that generates an image of a bone of a subject among bones and organs such as organs will be described.

[0049] The bone image generating device which also functions as an X-ray imaging device is also called an X-ray fluoroscopic imaging table, and comprises: a top plate 13; an X-ray tube holding component 15; an X-ray irradiation unit 11 arranged at the front end of the X-ray tube holding component 15; and an X-ray detection unit 12 having an X-ray detector such as a flat panel detector or an image intensifier (II) arranged on the opposite side of the X-ray irradiation unit 11 relative to the top plate 13.

[0050] The top plate 13, the X-ray tube holding member 15, the X-ray irradiation unit 11, and the X-ray detection unit 12 can be moved in a certain direction by a rotation mechanism 16 having a built-in motor (not shown). Figure 1 as well as Figure 2 The top plate 13 shown rotates between a lying position with the surface facing the horizontal direction and a standing position with the surface facing the vertical direction. In addition, the rotating mechanism 16 itself can be raised and lowered relative to a main support 17 erected on a base plate 18.

[0051] When the top board 13 is in the supine position, the subject in the supine state is subjected to X-ray photography. At this time, the subject is placed on the top board 13. In addition, when the top board 13 is in the standing position, the subject in the standing state is subjected to X-ray photography. At this time, the subject stands in front of the top board 13.

[0052] Next, the configuration of the bone image measuring device according to the embodiment of the present invention will be described. Figure 3 This is a block diagram showing a control system of the bone image generating device according to the embodiment of the present invention.

[0053] The bone image generating device is used to generate a bone image by extracting a bone area from an X-ray image of an area including the bone of a subject, and comprises: a CPU as a processor for performing logical operations; a ROM storing an action program required for controlling the device; a RAM for temporarily storing data during control, etc., and comprises a control unit 80 for controlling the entire device.

[0054] The control unit 80 includes the following units as functional components: an X-ray image generating unit 81 for generating an X-ray image; an X-ray image storing unit 82 for storing a plurality of X-ray images obtained by performing X-ray photography of a region including bones of a subject, etc., and a plurality of training bone images for X-ray images for machine learning; a DRR image generating unit 83 for generating a DRR (Digitally Reconstructed Image) of the region including bones by performing a virtual projection of the CT image data of the region including bones that simulates the geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12 when performing X-ray photography of the subject. Radiograph: Digital Reconfiguration Analog Image) image; a DRR image storage unit 84, storing the DRR images generated by the DRR image generation unit 83, and a plurality of DRR image training bone images for machine learning generated based on the DRR images generated by the DRR image generation unit 83; a learning unit 85, performing machine learning using a plurality of X-ray images and a plurality of X-ray image training bone images stored in the X-ray image storage unit 82, and performing machine learning using a plurality of DRR images and a plurality of DRR image training bone images stored in the DRR image storage unit 84, thereby generating a learning model for identifying bones; a bone image generation unit 86, converting an X-ray image of an area including the bones of the subject using the learned model learned by the learning unit 85, thereby generating an image representing the bones. The control unit 80 is composed of a computer with software installed. The functions of each unit included in the control unit 80 are realized by executing the software installed on the computer.

[0055] In the above configuration, when the DRR image training bone image and the X-ray image are label images, both may be the same image.

[0056] In the above configuration, the learning unit 85 may perform machine learning before the device is delivered and store the results in advance, or may perform machine learning in an additional manner after the device is delivered to a medical institution, etc. In this case, the learning unit 85 generates a recognizer based on various learning methods using any method such as FCN (Fully Convolutional Networks), Neural Machine Network, Support Vector Machine (SVM), and Boosting.

[0057] The control unit 80 is connected to the X-ray irradiation unit 11 and the X-ray detection unit 12. In addition, the control unit 80 is connected to a display unit 21 that is composed of a liquid crystal display panel and displays various images including X-ray images, and an operation unit 22 that has various input devices such as a keyboard and a mouse. Furthermore, the control unit 80 is connected online or offline to a CT image storage unit 70 that stores images obtained by CT imaging of the subject. The CT image storage unit 70 may be included in a CT imaging device or in a treatment planning device that generates a treatment plan for the subject.

[0058] Next, the operation of generating a bone image by extracting a bone region from an X-ray image of a region including the bone of a subject using the bone image generating device having the above configuration will be described.

[0059] First, the basic concept for generating bone images is described. Figure 4 This is a schematic diagram for explaining a process of generating a bone image of a subject using machine learning by the bone image generating apparatus according to an embodiment of the present invention.

[0060] In order to determine the position of the bone by machine learning, a learning model must first be generated. In this learning model generation process, the X-ray image and the DRR image of the region containing the bone are used as the input layer, the X-ray image showing the bone is used to train the bone image and the DRR image is used to train the bone image as the output layer, and the convolutional layer used as the learning model is learned by machine learning. Then, a bone image is generated. In this bone image generation process, the X-ray perspective image taken is used as the input layer, and the previously learned learning model is used to convert, thereby generating an image of the bone image obtained by extracting the region of the bone as the output layer.

[0061] Next, the operation of generating a bone image through such a process will be described in detail. Figure 5This is a flowchart showing the operation of generating a bone image by extracting a bone region from an X-ray image of a region including the bone of a subject using the bone image generating apparatus according to the embodiment of the present invention.

[0062] When the bone image generation device according to the embodiment of the present invention is used to generate a bone image, first, an X-ray image generation process (step S1) is performed. Figure 3 The X-ray image generation unit 81 shown uses Figure 1 The X-ray irradiation unit 11 and the X-ray detection unit 12 shown in the figure perform X-ray photography of the subject on the top plate 13, thereby generating a plurality of X-ray images. In addition, the X-ray images may be images acquired by other X-ray photography devices, or may be generated by performing X-ray photography of a phantom instead of the subject. The generated X-ray images are stored in Figure 3 The X-ray image is stored in the X-ray image storage unit 82 as shown (step S2).

[0063] Next, a training bone image for X-ray images used in machine learning is generated (step S3). The training bone image for X-ray images is generated by the X-ray image generation unit 81 by cropping the bone area of ​​the subject from the previously generated X-ray image. In addition, when generating the training bone image for X-ray images, an image is also generated by slightly translating, rotating, deforming, or scaling the cropped X-ray image. The image obtained by translating, rotating, deforming, or scaling the cropped X-ray image is also used for learning in order to cope with the following situation where the subject moves during X-ray shooting or the X-ray irradiation unit 11 and the X-ray detection unit 12 move. The generated training bone image for X-ray images is stored in Figure 3 The X-ray image is stored in the X-ray image storage unit 82 as shown (step S4).

[0064] In addition, at this time, translation, rotation, deformation, and enlargement and reduction are performed on both the X-ray image and the training bone image for X-ray image under the same conditions.

[0065] Figure 6 is a schematic diagram of an X-ray image 101 generated by the X-ray image generating unit 81. Figure 7 This is a schematic diagram of an X-ray image training bone image 102 generated by the X-ray image generating unit 81 .

[0066] The X-ray image 101 displays the femur 51, the pelvis 52, and the soft tissue region 53. In addition, the X-ray image training bone image 102 displays the femur 51 and the pelvis 52.

[0067] Then, by Figure 3The DRR image generating unit 83 shown in FIG. 1 simulates the CT image data acquired from the CT image storage unit 70. Figure 1 The virtual projection of the geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12 shown in the figure generates a plurality of DRR images showing the area including the bone (step S5), and the DRR images are stored in the DRR image storage unit 84 (step S6), and a plurality of training bone images for DRR images showing the area including the bone are generated (step S7), and the training bone images for DRR images are stored in the DRR image storage unit 84 (step S8). Here, when generating training bone images for DRR images showing the bone, the area with a CT value of a certain value or more is regarded as the area of ​​the bone, and the training bone image for DRR images is generated. For example, the area with a CT value of 200HU (Hounsfield Unit) or more is identified as the area of ​​the bone, and the training bone image for DRR images is generated.

[0068] Figure 8 It is schematically shown by simulating Figure 1 1 and 12. This is an explanatory diagram of a state in which a DRR image is generated by virtual projection obtained based on geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12 shown in FIG.

[0069] In the figure, reference numeral 300 indicates CT image data. The CT image data 300 is three-dimensional voxel data as a collection of a plurality of two-dimensional CT image data. The CT image data 300 has, for example, a direction transverse to the subject (along the Figure 8 The structure is obtained by convolving about 200 two-dimensional images of 512×512 pixels on the line segment L1 or L2 shown in the figure.

[0070] When the DRR image is generated by the DRR image generating unit 83, the CT image data 300 is virtually projected. At this time, the three-dimensional CT image data 300 is configured on the computer. Then, the geometric structure as the geometric configuration of the X-ray imaging system is reproduced on the computer. In this embodiment, the X-ray irradiation unit 11 and the X-ray detection unit 12 are configured on both sides of the CT image data 300. The configuration of these CT image data 300, the X-ray irradiation unit 11, and the X-ray detection unit 12 presents the same geometric structure as the configuration of the subject, the X-ray irradiation unit 11, and the X-ray detection unit 12 when performing X-ray imaging. Here, the geometric structure represents the geometric configuration relationship between the imaging object, the X-ray irradiation unit 11, and the X-ray detection unit 12.

[0071] In this state, a plurality of line segments L are set to connect the pixels of the X-ray irradiation unit 11 and the pixels of the X-ray detection unit 12 via the pixels of the CT image data 300. Figure 8In FIG. 1 , two line segments L1 and L2 are shown for convenience of explanation. Then, a plurality of calculation points are set on the line segment L, and the CT value of each calculation point is calculated. When calculating the CT value, interpolation of the CT value in the CT data voxel around the calculation point is performed. Then, the CT value of each calculation point on the line segment L is accumulated. The accumulated value is converted into a line integral of a linear attenuation coefficient, and the attenuation of X-rays is calculated, thereby generating a DRR image.

[0072] When generating the DRR image, the parameters for generating the DRR image including at least one of the projection coordinates and the angle of the CT image data 300 are changed to generate the DRR image. Alternatively, image processing including at least one of slight translation, rotation, deformation, and scaling is performed. The translation, rotation, deformation, scaling is performed to cope with the following situation where the subject moves or the X-ray irradiation unit 11 and the X-ray detection unit 12 move during X-ray shooting.

[0073] In addition, at least one of contrast change, noise addition, and edge enhancement is performed on the generated DRR image to absorb the difference in image quality between the DRR image and the X-ray image, thereby more reliably identifying the bone region.

[0074] The projection coordinates, angles, and other parameters used to generate the DRR images, or the contrast changes, noise addition, and edge enhancement described above are implemented by randomly or at equal intervals within a fixed range to give various changes. Therefore, a large number of DRR images can be generated based on the CT image data 300 of one subject. Therefore, a large number of DRR images generated in this way can be used to generate a customized learning model corresponding to each patient. In addition, the DRR images of a large number of patients can also be used to generate a learning model.

[0075] In addition, when generating a DRR image and when generating a training bone image for the DRR image, the parameters of the projection coordinates and angles including the geometric perspective conditions are changed under the same conditions, or image processing including image rotation, deformation, and enlargement and reduction is performed under the same conditions.

[0076] In addition, when generating a DRR image and a DRR image in a training bone image for a DRR image, the DRR image generating unit 83 generates a portion of the DRR images in a plurality of DRR images as a DRR image obtained by changing the concentration of the bone region in the region containing the bone. More specifically, the CT value of the region of the bone having a CT value above a certain value is set to a value smaller than the actual CT value. Therefore, a DRR image of the bone simulating a decrease in bone density can be obtained. Therefore, machine learning can be performed using the DRR image of the bone simulating a decrease in bone density, and the accuracy of bone extraction can be improved for patients including patients with decreased bone density and even osteoporosis.

[0077] Fig. 9 is a schematic diagram of a DRR image 103 generated by the DRR image generating unit 83. Fig.10 1 is a schematic diagram of a DRR image 104 in which the density of the bone region generated by the DRR image generating unit 83 is changed to a smaller value. Fig.11 This is a schematic diagram of a DRR image training bone image 105 generated by the DRR image generating unit 83 .

[0078] The DRR images 103 and 104 display the femur 51, the pelvis 52, and the soft tissue region 53. In addition, the DRR image training bone image 105 displays the femur 51 and the pelvis 52.

[0079] If the above process is completed, the learning unit 85 is used to Figure 6 The X-ray image 101 shown is used as the input layer. Figure 7 The X-ray image shown performs machine learning with the training bone image 102 as the output layer and Fig. 9 The DRR image 103 shown and Fig.10 The DRR image 104 obtained by changing the concentration of the bone region shown in FIG. 1 to a smaller value is used as the input layer and Fig.11 The DRR image shown in the figure uses the training bone image 105 as the output layer to perform machine learning, thereby generating a learning model (step S9) for identifying bones (femur 51 and pelvis 52). In this machine learning, for example, FCN can be used. The convolutional neural network used in FCN is the above-mentioned Figure 4 That is, when generating a learning model, the input layer is the X-ray image 101 and the DRR images 103 and 104, and the output layer is the X-ray image training bone image 102 and the DRR image training bone image 105.

[0080] If the learning model is generated by the above process, then the subject is subjected to X-ray photography (step S10). Then, the bone image generation unit 86 uses the previously generated learning model (convolutional layer) to transform the photographed X-ray image, performs segmentation, and generates an image of the bone (femur 51 and pelvis 52) (step S11). That is, for the X-ray image obtained by X-ray photography, the previously generated learning model is used to generate an image representing the bone as an output layer. Then, using the bone area determined by the segmentation, the bone density is measured by various methods.

[0081] In this specification, etc., "segmentation" is a concept including a process of determining the contour or outer shape of a bone or the like in addition to the process of determining the region of the bone or the like in this embodiment.

[0082] If the generation of the bone image is completed through the above process, the operator will correct the generated bone image as needed. Then, the corrected bone image and its original X-ray image are used for the generation or re-learning of the learning model performed by the learning unit 85. In this way, the learning image containing the failure case can be expanded to generate a learning model with higher accuracy.

[0083] As described above, according to the bone image generation device of the embodiment of the present invention, the extraction accuracy can be improved by extracting the bone area by machine learning. At this time, since both the X-ray image and the DRR image are used for machine learning, the learning image can be expanded, and the collection of clinical data for learning can be easily performed. In addition, by using the DRR image obtained by changing the concentration of the bone area, machine learning can be performed by simulating the DRR image of the bone with reduced bone density, and the extraction accuracy of the bone can be improved for patients including those with reduced bone density and osteoporosis.

[0084] In addition, in the above-mentioned embodiment, the X-ray image may be input into the learning model after being blurred by a Gaussian filter or the like. Generally, since the DRR image is generated from a low-resolution CT image, it has a low resolution compared to the X-ray image. Therefore, by blurring the X-ray image, reducing the noise of the X-ray image, or setting the resolution to be the same as that of the DRR image during learning, the bone part can be more reliably identified. In addition, in the above-mentioned embodiment, the DRR image and the X-ray image input into the learning model may be input after contrast normalization has been performed in advance. In addition, a local contrast normalization layer or a local response normalization layer may be added to the intermediate layer.

[0085] Next, another embodiment of the present invention will be described.

[0086] In the above-mentioned embodiment, a portion of the DRR images among the plurality of DRR images is generated as a DRR image obtained by changing the concentration of the bone region in the region including the bone, thereby generating a DRR image of the bone that simulates a decrease in bone density and using it for machine learning. In contrast, in this embodiment, a portion of the X-ray images among the plurality of X-ray images is set as an X-ray image obtained by changing the concentration of the bone region in the region including the bone by using dual energy subtraction, wherein the dual energy subtraction is performed by subtracting an X-ray image (high voltage image) taken when a high voltage is applied to the X-ray tube and an X-ray image (low voltage image) taken when a low voltage is applied to the X-ray tube.

[0087] That is, in order to diagnose osteoporosis, when measuring the bone density of the subject, a structure for measuring the bone density using dual-energy subtraction is adopted, wherein the dual-energy subtraction is to perform subtraction processing on an X-ray image taken when a high voltage is applied to the X-ray tube and an X-ray image taken when a low voltage is applied to the X-ray tube. When determining the bone image, the dual-energy subtraction is also used to weight the X-ray image taken when a high voltage is applied to the X-ray tube and the X-ray image taken when a low voltage is applied to the X-ray tube, and then a dual-energy subtraction image representing the bone is generated by obtaining their difference. Then, by subtracting the dual-energy subtraction image from the X-ray image (high-voltage image or low-voltage image), an image (equivalent to an image of an X-ray image of a bone in a state where the bone density is reduced) in which the concentration of the bone region is reduced in the region including the bone is obtained. By using the X-ray image with a low concentration of the bone region for machine learning, machine learning can be performed by simulating the X-ray image of the bone with reduced bone density, and the accuracy of bone extraction can be improved for patients including patients with osteoporosis.

[0088] At this time, as the X-ray image used in machine learning, any one of a high-pressure image, a low-pressure image, or a dual-energy subtraction image can be used, or an image obtained by connecting these images along the channel direction can be used. In addition, an X-ray image of a bone portion simulating reduced bone density can be obtained by adjusting the parameters of the dual-energy subtraction image instead of subtracting the dual-energy subtraction image from the X-ray image (high-pressure image or low-pressure image).

[0089] Next, another embodiment of the present invention will be described. Fig.12 is a schematic diagram of an X-ray image 106 generated by the X-ray image generating unit 81. Fig.13 Schematic diagram of the DRR image 107 generated by the DRR image generating unit 83 .

[0090] This embodiment is used when generating a bone image of a bone such as a femur that has a shape that is bilaterally symmetrical with respect to the body axis of the subject. Figure 6 is a schematic diagram of an X-ray image 101 near the right leg of the subject. Fig. 9 is a schematic diagram of a DRR image 103 near the right leg of the subject. Fig.12 is a schematic diagram of an X-ray image 106 of the vicinity of the left leg of the subject. Fig.13 1 is a schematic diagram of a DRR image 107 near the left leg of the subject.

[0091] Thus, when the object is a bone part (femur 51 and pelvis 52) having a shape that is symmetrical with respect to the body axis of the subject, the learning unit 85 performs machine learning on the images of the right and left bones together by flipping either the image of the right bone part or the image of the left bone part. Fig.12 The X-ray image 106 of the vicinity of the left leg of the subject shown is flipped left and right. Figure 6 The X-ray image 101 of the vicinity of the right leg of the subject shown in FIG. 1 is also used for machine learning. Fig.13 The DRR image 107 near the left leg of the subject shown is flipped left and right. Fig. 9 The DRR image 103 of the vicinity of the right leg of the subject shown is also used for machine learning.

[0092] By adopting such a configuration, the detection accuracy of bones having a bilaterally symmetrical shape with respect to the body axis of the subject can be uniform. In addition, by performing machine learning on the images of the left and right bones at the same time, the learning images can be expanded to generate a more accurate learning model.

[0093] In the above-mentioned embodiment, machine learning is performed using both X-ray images and DRR images. However, machine learning may be performed using either X-ray images or DRR images.

[0094] In addition, in the above-mentioned embodiment, the bone part is used as the object as the organ, but for example, organs such as internal organs can also be used as the object. For example, when the subject has a lot of visceral fat, the concentration of the internal organ region becomes low when X-rays are taken. According to the present invention, even in such a case, a learning model corresponding to the subject whose internal organ concentration is low can be generated. Therefore, the detection accuracy of the internal organs can be improved.

[0095] Description of Reference Numerals

[0096] 11 X-ray irradiation department

[0097] 12 X-ray inspection unit

[0098] 13 Top Plate

[0099] 14 Pillars

[0100] 15 X-ray tube holding parts

[0101] 16 Rotation mechanism

[0102] 17 Main Pillar

[0103] 18 Base Plate

[0104] 21 Display

[0105] 22 Operation Department

[0106] 70 CT image storage unit

[0107] 80 Control Department

[0108] 81 X-ray image generation unit

[0109] 82 X-ray image storage unit

[0110] 83 DRR Image Generation Unit

[0111] 84 DRR image storage unit

[0112] 85 Learning Department

[0113] 86 Bone Image Generation Unit

[0114] 300 CT image data.

Claims

1. An image analysis method for analyzing an image of a region containing an organ of a subject to perform segmentation for determining the region of the organ, characterized in that: Machine learning is used as the segmentation method, and includes: a corrected image generating step of generating a corrected image by changing the density of a region of the organ in an image including the organ of the subject; a learning model generating step of generating a learning model for machine learning by performing a learning process as an input layer including an image of the organ of the subject and a corrected image generated by the corrected image generating step, The image of the region including the organ of the subject is a DRR image generated based on CT image data of the subject. In the corrected image generating step, a region in which the CT value of the CT image data becomes a predetermined value is set as the region of the organ, and the density thereof is changed.

2. The image analysis method according to claim 1, characterized in that: An X-ray image of a region including an organ of the subject obtained by performing X-ray imaging of the subject is converted using the learning model generated in the learning model generating step, thereby generating an image representing the organ.

3. The image analysis method according to claim 1, wherein: When generating a DRR image, at least one of the parameters of the projection coordinates and angles including geometric conditions is changed, or image processing including at least one of rotation, deformation, and enlargement and reduction of the image is performed to generate a plurality of DRR images.

4. The image analysis method according to claim 1, wherein: At least one of contrast change, noise addition, and edge enhancement is performed on the generated DRR image.

5. The image analysis method according to claim 2, wherein: An X-ray image of a region including an organ of the subject obtained by performing X-ray photography of the subject and an image representing the organ converted using the learning model generated by the learning model generating step are used for learning the learning model by the learning unit.

6. The image analysis method according to claim 1, wherein: The organ has a shape that is symmetrical left-right relative to the body axis of the subject. In the learning model generation process, by flipping either the image of the organ on the right side or the image of the organ on the left side left-right, a learning model for machine learning is generated for the images of the left and right organs together.

7. An image analysis method for analyzing an image of a region containing an organ of a subject to perform segmentation for determining the region of the organ, characterized in that: Machine learning is used as the segmentation method, and includes: a corrected image generating step of generating a corrected image by changing the density of a region of the organ in an image including the organ of the subject; a learning model generating step of generating a learning model for machine learning by performing a learning process as an input layer including an image of the organ of the subject and a corrected image generated by the corrected image generating step, The image of the region including the organ of the subject is an X-ray image generated by photographing the subject with X-rays. In the corrected image generating step, the density of the region of the organ is changed using the X-ray image and the image of the organ obtained by dual-energy subtraction.

8. A segmentation method, characterized in that: The organ is a bone of the subject, and the region of the bone is segmented using the image analysis method according to claim 1 or 7.

9. A method for measuring bone density, characterized in that: Bone density is measured for the bone region segmented by the segmentation method according to claim 8.

10. A learning model generation method, which uses machine learning to analyze an image of a region containing an organ of a subject, thereby generating a learning model used when performing segmentation for determining the region of the organ, characterized in that: The image including the organ of the subject and the corrected image generated by changing the density of the region of the organ in the image including the organ of the subject are used as input layers, and machine learning is performed to generate a learning model. The image of the region including the organ of the subject is a DRR image generated based on CT image data of the subject. a region in which the CT value of the CT image data becomes a predetermined value as the region of the organ and changes the concentration thereof; or, The image of the region including the organ of the subject is an X-ray image generated by photographing the subject with X-rays. The density of a region of the organ is changed using the X-ray image and the image of the organ obtained by dual-energy subtraction.

11. An image generating device for generating an image by extracting a region of an organ of a subject from an X-ray image of a region including the organ of the subject, characterized in that: have: An X-ray image storage unit storing a plurality of X-ray images obtained by performing X-ray imaging of a region including the organ and a plurality of X-ray image training images for machine learning; A DRR image generating unit, generating a DRR image of a region including the organ; A DRR image storage unit storing a plurality of DRR images generated by the DRR image generation unit and a plurality of DRR image training images for machine learning generated based on the DRR images generated by the DRR image generation unit; An image generating unit converts an X-ray image of a region containing the organ of the subject using a learning model for identifying the organ, thereby generating an image representing the organ, wherein the learning model is pre-generated by performing machine learning using the multiple X-ray images stored in the X-ray image storage unit and the multiple X-ray image training images, and performing machine learning using the multiple DRR images stored in the DRR image storage unit and the multiple DRR image training images.

12. The image generating device according to claim 11, wherein: The DRR image generating unit generates a part of the plurality of DRR images as a DRR image in which the density of an organ region in a region including the organ is changed, The DRR image is a DRR image generated based on the CT image data of the subject, A region in which the CT value of the CT image data reaches a predetermined value is defined as the region of the organ, and the density thereof is changed.

13. The image generating device according to claim 11, wherein: Some of the plurality of X-ray images stored in the X-ray image storage unit are X-ray images obtained by changing the density of an organ region in a region including the organ using dual-energy subtraction.

Images