WO2019208037A1

WO2019208037A1 - Image analysis method, segmentation method, bone density measurement method, learning model creation method, and image creation device

Info

Publication number: WO2019208037A1
Application number: PCT/JP2019/011773
Authority: WO
Inventors: ▲高▼橋　渉; 翔太押川
Original assignee: 株式会社島津製作所
Priority date: 2018-04-24
Filing date: 2019-03-20
Publication date: 2019-10-31
Also published as: JP7092190B2; JPWO2019208037A1; KR20200142057A; KR102527440B1; CN112165900A

Abstract

A control unit (80) comprises: an X-ray image creation unit (81); an X-ray image memory unit (82) for storing an X-ray image and an X-ray image-use training bone part image; a DRR image creation unit (83) for creating a DRR image from a region containing a bone part; a DRR image memory unit (84) for storing the DRR image and a DRR image-use training bone part image for machine learning use; a learning unit (85) which uses the X-ray image and the X-ray image-use training bone part image to perform machine learning and uses the DRR image and the DRR image-use training bone part image to perform machine learning, thereby creating a learning model for recognizing the bone part; and a bone part image creation unit (86) which uses the learned model created in the learning unit (85) to convert an X-ray image of a region containing the bone part of a subject and create an image representing the bone part.

Description

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

The present invention relates to an image analysis method, a segmentation method, a bone density measurement method, a learning model creation method, and an image creation device.

In recent years, a bone density measuring device for measuring the bone density of a subject has been used for diagnosis of osteoporosis. Patent Document 1 discloses collimating only a means for generating radiation, a single crystal lattice irradiated with the radiation, and radiation having a predetermined two reflection angles out of the radiation reflected by the crystal lattice. By means of simultaneously irradiating the subject with radiation of different energies, radiation detection means through which the radiation of these two energies passes through the subject, and analyzing the pulse height of the output of the radiation detection means, respectively An apparatus for quantitatively analyzing bone mineral is disclosed which comprises a pulse height analyzing means for separating transmission data relating to radiation of different energy, and a means for calculating bone density by processing the separated data.

Japanese Patent No. 2638875

This measurement of bone density is targeted at the bone density of the lumbar vertebrae and femur, which requires clinical attention. At this time, the femur has a large individual difference in shape, and in order to perform stable follow-up observation, it is important to specify the bone region of the subject. In order to specify the region of the bone portion, the operator manually specifies the region, which is not only complicated, but also has a problem that the region specified by the operator varies.

In order to automatically perform segmentation for extracting the bone region from the image including the bone part of the subject, it may be possible to use an algorithm based on the threshold processing of the histogram. Regarding femoral segmentation, it was difficult to accurately identify the bone region. For this reason, the problem that the precision of the measurement result of a final bone density deteriorates has arisen.

The present invention has been made to solve the above problems, and an image analysis method capable of creating an image in which an organ region is accurately extracted from an X-ray image of a region including a subject's organ, An object of the present invention is to provide a segmentation method, a bone density measurement method, a learning model creation method, and an image creation device.

The invention according to claim 1 is an image analysis method for performing segmentation for identifying a region of the organ by analyzing an image of the region including the organ of the subject, and machine learning is used as the segmentation technique. And a modified image creating step of creating a modified image in which the density of the region of the organ in the image including the organ of the subject is changed, and an image including the organ of the subject and the modified image creating step And a learning model creation step of creating a machine learning learning model by a learning process using the corrected image created in the above.

The invention according to claim 2 is the learning model according to claim 1, wherein the learning model is used for an X-ray image of a region including the subject's organ obtained by X-ray imaging of the subject. An image representing the organ is created by performing conversion using the learning model created in the creation step.

The invention according to claim 3 is the invention according to claim 1, wherein the image of the region including the organ of the subject is a DRR image created from CT image data of the subject, and the correction In the image creating step, the density of the CT image data is changed using the region where the CT value of the CT image data is a predetermined value as the region of the organ.

According to a fourth aspect of the present invention, in the invention according to the third aspect, when a DRR image is created, a parameter including at least one of a projection coordinate and an angle of the geometric condition is changed, or an image is rotated. Image processing including at least one of deformation and enlargement / reduction is performed to create a plurality of DRR images.

According to a fifth aspect of the present invention, in the third aspect of the invention, at least one of contrast change, noise addition, and edge enhancement is performed on the created DRR image.

The invention according to claim 6 is the invention according to claim 1, wherein the image of the region including the organ of the subject is an X-ray image created by X-ray imaging of the subject. In the modified image creating step, the density of the organ region is changed using the X-ray image and the image of the organ obtained by using dual energy subtraction.

According to a seventh aspect of the present invention, in the second aspect of the present invention, an X-ray image of a region including the subject's organ obtained by X-raying the subject and the learning model creating step The image representing the organ obtained by performing the conversion using the learning model created in the above is used for learning of the learning model by the learning unit.

The invention according to claim 8 is the invention according to claim 1, wherein the organ has a symmetrical shape with respect to the body axis of the subject, and in the pre-learning model creation step, the right organ A machine learning learning model is created collectively for the left and right organ images by horizontally flipping either the left image or the left organ image.

According to the ninth aspect of the present invention, the region of the bone part is segmented using the image analysis method according to the first aspect, wherein the organ is the bone part of the subject.

In the invention described in claim 10, the bone density is measured for the bone region segmented by the segmentation method described in claim 9.

The invention according to claim 11 is a learning model used when performing segmentation for specifying the region of the organ by analyzing an image of the region including the organ of the subject using machine learning. An image including the subject's organ and a modified image generated by changing the density of the region of the organ in the image including the subject's organ. Used to create a learning model by executing machine learning.

The invention according to claim 12 is an image creation device for creating an image obtained by extracting an area of the organ from an X-ray image of the area including the organ of the subject, and X-rays the area including the organ. An X-ray image storage unit that stores a plurality of X-ray images obtained in this manner, a plurality of X-ray image teacher images for machine learning, and a DRR image generation unit that generates a DRR image of an area including the bone part A DRR image that stores a plurality of DRR images created by the DRR image creation unit and a plurality of machine learning DRR image teacher images created based on the DRR image created by the DRR image creation unit Machine learning is performed using the storage unit, the plurality of X-ray images stored in the X-ray image storage unit, and the plurality of X-ray image teacher images, and is stored in the DRR image storage unit. The plurality of DRR images X-rays of a region including the organ of the subject using a learning model for recognizing the organ, which is created in advance by performing machine learning using the plurality of teacher images for DRR images And an image creation unit that creates an image representing the organ by performing conversion on the image.

According to a thirteenth aspect of the present invention, in the invention according to the eleventh aspect, the DRR image creation unit may include a part of the plurality of DRR images as a part of the region including the bone part. Created as a DRR image in which the density of the region is changed.

The invention described in claim 14 is the invention described in claim 11, wherein a part of the plurality of X-ray images stored in the X-ray image storage unit uses dual energy subtraction. This is an X-ray image in which the density of the organ region of the region including the organ is changed.

According to the first to eighth aspects of the invention, since the corrected image in which the density of the organ region of the subject is changed is used for machine learning, it can be applied to a subject having a low organ density. It is possible to create a learning model. For this reason, it becomes possible to improve the detection accuracy of an organ.

According to the fourth aspect of the present invention, the parameters including the projection coordinates and angles of the geometric perspective conditions are changed, or image processing including rotation, deformation, and enlargement / reduction of the image is performed. Even when the position and orientation of the X-ray or the position of the imaging system during X-ray imaging are slightly deflected, the position of the bone part can be accurately detected. Since a large amount of DRR images can be created, it is possible to learn a custom-made discriminator corresponding to each patient. Furthermore, even when using a low frame rate DRR image, The position can be accurately detected.

According to the fifth aspect of the present invention, since contrast change, noise addition, and edge enhancement are performed on the created DRR image, even when there is a difference in image quality between the DRR image and the X-ray image. In addition, the position of the bone part can be accurately detected.

According to the seventh aspect of the present invention, by reusing a plurality of X-ray images and an image representing an organ obtained by performing conversion using a learned learned model for learning of a learning model, It is possible to create a learning model with higher accuracy by expanding the learning image.

According to the invention described in claim 8, it is possible to make the detection accuracy uniform for an organ having a symmetrical shape with respect to the body axis of the subject. Since machine learning is performed on the images of the left and right organs in a lump, it is possible to create a learning model with higher accuracy by expanding the learning image.

According to the invention described in claim 9, it is possible to accurately perform segmentation of the bone region.

According to the invention of the tenth aspect, it is possible to execute the bone density measurement on the segmented bone region.

According to the invention of claim 11, since the corrected image in which the density of the organ region of the subject is changed is used for machine learning, a learning model corresponding to the subject having a low organ density is also provided. It becomes possible to create.

According to the invention described in claims 12 to 14, it is possible to improve the extraction accuracy by extracting an organ region by machine learning. At this time, since machine learning is performed using both the X-ray image and the DRR image, the learning image can be expanded, and the collection of learning clinical data can be easily performed.

1 is a schematic front view of a bone image creating apparatus according to an embodiment of the present invention that also functions as an X-ray imaging apparatus. 1 is a schematic side view of a bone image creating apparatus according to an embodiment of the present invention that also functions as an X-ray imaging apparatus. It is a block diagram which shows the control system of the bone part image creation apparatus which concerns on embodiment of this invention. It is a schematic diagram for demonstrating the process of producing the bone part image of a subject using machine learning with the bone part image creation apparatus which concerns on embodiment of this invention. It is a flowchart which shows the operation | movement which produces the bone part image which extracted the area | region of the bone part from the X-ray image of the area | region containing a subject's bone part by the bone part image creation apparatus which concerns on embodiment of this invention. 3 is a schematic diagram of an X-ray image 101 created by an X-ray image creation unit 81. FIG. FIG. 6 is a schematic diagram of a teacher bone image for X-ray image 102 created by an X-ray image creation unit 81. It is explanatory drawing which shows typically the state which produces a DRR image by the virtual projection which simulated the geometric condition of the X-ray irradiation part 11 and the X-ray detection part 12 which are shown in FIG. 6 is a schematic diagram of a DRR image 103 created by a DRR image creation unit 83. FIG. FIG. 10 is a schematic diagram of a DRR image 104 in which the density of the bone region created by the DRR image creating unit 83 is changed to a small value. 6 is a schematic diagram of a DRR image teacher bone image 105 created by a DRR image creation unit 83. FIG. 3 is a schematic diagram of an X-ray image 106 created by an X-ray image creation unit 81. FIG. 6 is a schematic diagram of a DRR image 107 created by a DRR image creation unit 83. FIG.

Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 is a schematic front view of a bone image creating apparatus according to an embodiment of the present invention that also functions as an X-ray imaging apparatus, and FIG. 2 is a schematic side view thereof. In this embodiment, a case will be described in which the present invention is applied to a bone image creating apparatus that creates an image of a bone part of a subject among organs such as a bone part and an organ.

This bone image creating apparatus that also functions as an X-ray imaging apparatus is also referred to as an X-ray fluoroscopic imaging table, and includes a top plate 13, an X-ray tube holding member 15, and an X-ray tube holding member 15. X-ray irradiation unit 11 disposed at the tip, and X-rays such as a flat panel detector and an image intensifier (II) disposed on the opposite side of the X-ray irradiation unit 11 with respect to the top plate 13 And an X-ray detection unit 12 having a detector.

The top plate 13, the X-ray tube holding member 15, the X-ray irradiation unit 11, and the X-ray detection unit 12 are shown in FIG. 1 and FIG. 2 by the action of a rotation mechanism 16 incorporating a motor (not shown). It is possible to rotate between a recumbent position where the surface of 13 faces the horizontal direction and a standing position where the surface of the top plate 13 faces the vertical direction. Further, the rotation mechanism 16 itself can be moved up and down with respect to the main column 17 erected on the base plate 18.

When the top 13 is in the prone position, X-ray imaging is performed on the subject in the prone position. At this time, the subject is placed on the top board 13. When the top 13 is in the standing position, X-ray imaging is performed on the subject in the standing position. At this time, the subject stands up in front of the top board 13.

Next, the configuration of the bone image measuring apparatus according to the embodiment of the present invention will be described. FIG. 3 is a block diagram showing a control system of the bone image creating apparatus according to the embodiment of the present invention.

This bone part image creating apparatus is for creating a bone part image obtained by extracting a bone region from an X-ray image of a region including a bone part of a subject, and a CPU as a processor for executing a logical operation And a ROM that stores an operation program necessary for controlling the apparatus, a RAM that temporarily stores data and the like during control, and a control unit 80 that controls the entire apparatus.

As a functional configuration, the control unit 80 includes an X-ray image creation unit 81 for creating an X-ray image, and a plurality of X-ray images obtained by X-ray imaging of a region including a bone part such as a subject. And X-ray image storage unit 82 that stores a plurality of X-ray image teacher bone images for machine learning, and X-ray imaging of a subject with respect to CT image data of a region including the bone portion By creating a virtual projection that simulates the geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12, a DRR (Digitally Reconstructed Radiograph) image of the region including the bone part is created. DRR image creation unit 83, a plurality of DRR images created by DRR image creation unit 83, and a plurality of machines created based on the DRR image created by DRR image creation unit 83 A DRR image storage unit 84 for storing a training DRR image teacher bone image, a plurality of X-ray images and a plurality of X-ray image teacher bone images stored in the X-ray image storage unit 82 are used. Machine learning and recognizing a bone part by executing machine learning using a plurality of DRR images and a plurality of DRR image teacher bone images stored in the DRR image storage unit 84. An image representing a bone part by performing conversion using a learned model learned by the learning unit 85 on an X-ray image of a region including the bone part of the subject and a learning unit 85 that creates a learning model A bone image creating unit 86 for creating The control unit 80 is composed of a computer in which software is installed. The functions of each unit included in the control unit 80 are realized by executing software installed in the computer.

In the above-described configuration, when the DRR image teacher bone image and the X-ray image are label images, both may be the same image.

In the above-described configuration, the learning unit 85 may perform machine learning at the stage before delivery of the device and store the result in advance, and additionally machine learning after delivery of the device to a medical institution or the like. May be executed. At this time, the learning unit 85 creates a discriminator by various learnings using arbitrary methods such as FCN (Fully Convolutional Networks), a neural machine network, a support vector machine (SVM), and boosting.

The control unit 80 is connected to the X-ray irradiation unit 11 and the X-ray detection unit 12 described above. The control unit 80 is connected to a display unit 21 configured by a liquid crystal display panel or the like and displaying various images including an X-ray image, and an operation unit 22 having various input means such as a keyboard and a mouse. Yes. Further, the control unit 80 is connected online or offline to a CT image storage unit 70 that stores a CT image obtained by CT imaging of the subject. The CT image storage unit 70 may be included in a CT imaging apparatus, or may be included in a treatment planning apparatus that creates a treatment plan for a subject.

Next, an operation for creating a bone part image obtained by extracting a bone region from an X-ray image of a region including the bone part of a subject by using the bone part image creation device having the above-described configuration. explain.

First, the basic concept for creating a bone image will be described. FIG. 4 is a schematic diagram for explaining a process of creating a bone image of a subject using machine learning by the bone image creating apparatus according to the embodiment of the present invention.

In order to specify the position of the bone using machine learning, first, a learning model is created. In this learning model creation step, an X-ray image and a DRR image of a region including a bone part are used as an input layer, an X-ray image teacher bone image indicating a bone part and a DRR image teacher bone part image are used as an output layer, A convolutional layer used as a learning model is learned by machine learning. Next, a bone part image is created. In this bone part image creation step, a bone obtained by extracting a bone part region as an output layer by using a captured X-ray fluoroscopic image as an input layer and performing conversion using a learning model learned earlier. An image showing a partial image is created.

Next, the operation of creating a bone image by such a process will be described in detail. FIG. 5 is a flowchart showing an operation of creating a bone portion image obtained by extracting a bone region from an X-ray image of a region including a bone portion of a subject by the bone portion image creating apparatus according to the embodiment of the present invention. .

When creating a bone part image by the bone part image creating apparatus according to the embodiment of the present invention, first, an X-ray image creating step is executed (step S1). In this image creation process, an X-ray image of the subject on the top 13 is obtained by using the X-ray irradiation unit 11 and the X-ray detection unit 12 shown in FIG. 1 by the X-ray image creation unit 81 shown in FIG. By doing so, a plurality of X-ray images are created. Note that an X-ray image may be obtained by taking an image taken by another X-ray imaging apparatus, or may be created by taking an X-ray image of a phantom instead of the subject. The created X-ray image is stored in the X-ray image storage unit 82 shown in FIG. 3 (step S2).

Next, a teacher bone image for X-ray image used for machine learning is created (step S3). This X-ray image teacher bone part image is created by the X-ray image creation part 81 by trimming the region of the bone part of the subject with respect to the previously created X-ray image. Further, when the X-ray image teacher bone image is created, an image obtained by slightly translating, rotating, deforming, and enlarging / reducing the trimmed X-ray image is also created. An image obtained by translating, rotating, transforming, and enlarging / reducing the trimmed X-ray image is also used for learning because the subject moves during X-ray imaging described later, or the X-ray irradiation unit 11 and the X-ray detection unit This is to cope with a case where 12 moves. The created X-ray image teacher bone image is stored in the X-ray image storage unit 82 shown in FIG. 3 (step S4).

At this time, parallel movement, rotation, deformation, and enlargement / reduction are executed under the same conditions for both the X-ray image and the X-ray image teacher bone part image.

FIG. 6 is a schematic diagram of the X-ray image 101 created by the X-ray image creation unit 81, and FIG. 7 is a schematic diagram of the teacher bone image 102 for the X-ray image created by the X-ray image creation unit 81. It is.

In the X-ray image 101, a femur 51, a pelvis 52, and a soft part region 53 are displayed. Further, the femur 51 and the pelvis 52 are displayed in the teacher bone image 102 for X-ray images.

Next, geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12 shown in FIG. 1 are applied to the CT image data acquired from the CT image storage unit 70 by the DRR image creation unit 83 shown in FIG. By performing virtual projection that simulates the above, a plurality of DRR images showing the region including the bone part are created (step S5), and the DRR image is stored in the DRR image storage unit 84 (step S6). A plurality of teacher bone part images for DRR images showing a region including the image are created (step S7), and the teacher bone part images for DRR images are stored in the DRR image storage unit 84 (step S8). Here, when creating a DRR image teacher bone image indicating a bone part, a DRR image teacher bone image is created with a region having a CT value equal to or greater than a certain value as a bone region. For example, a DRR image teacher bone image is created by identifying a region having a CT value of 200 HU (Hounsfield Unit) or more as a bone region.

FIG. 8 is an explanatory diagram schematically showing a state in which a DRR image is created by virtual projection simulating the geometric conditions of the X-ray irradiation unit 11 and the X-ray detection unit 12 shown in FIG.

In this figure, reference numeral 300 indicates CT image data. The CT image data 300 is three-dimensional voxel data that is a set of a plurality of two-dimensional CT image data. The CT image data 300 has a structure in which, for example, about 200 two-dimensional images of 512 × 512 pixels are stacked in a direction crossing the subject (direction along the line segment L1 or L2 shown in FIG. 8). .

When the DRR image creating unit 83 creates a DRR image, it virtually projects the CT image data 300. At this time, the three-dimensional CT image data 300 is arranged on the computer. Then, the geometry which is the geometric arrangement 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 disposed on both sides of the CT image data 300. The arrangement of the CT image data 300, the X-ray irradiation unit 11, and the X-ray detection unit 12 is such that the subject when performing X-ray imaging, the X-ray irradiation unit 11, and the X-ray detection unit 12 are arranged. It has the same geometry as the arrangement. Here, the term “geometry” means a geometric arrangement relationship between the imaging target, the X-ray irradiation unit 11 and the X-ray detection unit 12.

In this state, a large number of line segments L connecting the X-ray irradiation unit 11 and each pixel of the X-ray detection unit 12 via each pixel of the CT image data 300 are set. In FIG. 8, two line segments L1 and L2 are shown for convenience of explanation. 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 is performed using the CT value in CT data voxels around the calculation point. Thereafter, the CT values of the calculation points on the line segment L are accumulated. This accumulated value is converted into a line integral of a line attenuation coefficient, and a DRR image is created by calculating attenuation of X-rays.

At the time of creating this DRR image, the DRR image is created by changing the parameters for creating the DRR image including at least one of the projection coordinates and the angle with respect to the CT image data 300. Alternatively, image processing including at least one of slight translation, rotation, deformation, and enlargement / reduction is executed. The parallel movement, rotation, deformation, and enlargement / reduction are executed in order to correspond to the case where the subject moves during the X-ray imaging described later, or the X-ray irradiation unit 11 and the X-ray detection unit 12 move. It is.

Also, at least one of contrast change, noise addition, and edge enhancement is executed on the created DRR image. This contrast change, noise addition, and edge enhancement are performed in order to absorb the difference in image quality between the DRR image and the X-ray image and to more reliably recognize the bone region.

The above-described changes in parameters for creating a DRR image such as projection coordinates and angles, or changes in contrast, addition of noise, and edge enhancement are varied in a predetermined range at random or at equal intervals. Will be implemented. Thereby, a large amount of DRR images can be created from the CT image data 300 of one subject. Therefore, it is possible to create a custom-made learning model corresponding to each patient using a large amount of DRR images created in this way. It is possible to create a learning model using DRR images of a large number of patients.

It should be noted that when the DRR image is created and when the DRR image teacher bone image is created, the parameters including the projection coordinates and angle of the geometric perspective condition are changed under the same conditions, or the image is rotated, deformed, or enlarged. Image processing including reduction is performed under the same conditions.

Further, when creating a DRR image out of the DRR image and the DRR image teacher bone part image, the DRR image creation part 83 selects a part of the DRR images from the plurality of DRR images as a bone in the region including the bone part. It is created as a DRR image in which the density of the partial area is changed. More specifically, the CT value of the bone region where the CT value is a certain value or more is set to a value smaller than the actual CT value. Thereby, it is possible to obtain a DRR image simulating a bone part having a reduced bone density. Therefore, it is possible to perform machine learning using a DRR image simulating a bone part having a decreased bone density, and to improve the extraction accuracy of the bone part for patients including a patient having a decreased bone density and resulting in osteoporosis. It becomes possible.

FIG. 9 is a schematic diagram of the DRR image 103 created by the DRR image creation unit 83. FIG. 10 shows a DRR image 104 in which the density of the bone region created by the DRR image creation unit 83 is changed to a small value. FIG. 11 is a schematic diagram of the DRR image teacher bone part image 105 created by the DRR image creation unit 83.

In the

DRR images

103 and 104, the femur 51, the pelvis 52, and the soft part region 53 are displayed. Further, the femur 51 and the pelvis 52 are displayed in the DRR image teacher bone part image 105.

When the above steps are completed, the learning unit 85 executes machine learning using the X-ray image 101 shown in FIG. 6 as an input layer and the X-ray image teacher bone image 102 shown in FIG. 7 as an output layer. Machine learning using the DRR image 103 shown in FIG. 9 and the DRR image 104 in which the density of the bone region shown in FIG. 10 is changed to a small value as an input layer, and the teacher bone image 105 for DRR image shown in FIG. 11 as an output layer. By executing the above, a learning model for recognizing the bone part (the femur 51 and the pelvis 52) is created (step S9). For this machine learning, for example, FCN is used. The convolutional neural network used in the FCN is configured as shown in FIG. That is, when creating a learning model, the input layer is an X-ray image 101 and

DRR images

103 and 104, and the output layer is an X-ray image teacher bone image 102 and a DRR image teacher bone image 105.

If a learning model is created by the above steps, next, X-ray imaging is performed on the subject (step S10). Then, the bone part image creation unit 86 converts the captured X-ray image by using the learning model (convolution layer) created earlier, thereby executing segmentation, and the bone part (femur) 51 and the pelvis 52) are created (step S11). That is, the learning model created previously is used for an X-ray image obtained by X-ray imaging, and an image representing a bone part is created as an output layer. Then, the bone density is measured by various methods using the bone region specified by the segmentation.

In this specification and the like, “segmentation” is a concept including a process of specifying an outline of a bone or the like or a process of specifying an outline of a bone or the like in addition to the process of specifying a region such as a bone in this embodiment. is there.

If creation of the bone part image is completed through the above steps, the operator corrects the created bone part image as necessary. Then, the corrected bone image and the original X-ray image are used for creating a learning model by the learning unit 85 or for re-learning. As a result, it is possible to create a learning model with higher accuracy by expanding learning images including failure examples.

As described above, according to the bone image creating apparatus according to the embodiment of the present invention, the extraction accuracy can be improved by extracting the bone region by machine learning. At this time, since machine learning is performed using both the X-ray image and the DRR image, the learning image can be expanded, and the collection of learning clinical data can be easily performed. In addition, by using a DRR image in which the density of the bone region is changed, it is possible to perform machine learning with a DRR image simulating a bone portion having a reduced bone density, resulting in a decrease in bone density and osteoporosis. Bone extraction accuracy can be improved for patients including patients.

In the above-described embodiment, the X-ray image may be input to the learning model after being blurred by a Gaussian filter or the like. In general, since a DRR image is created from a low-resolution CT image, it has a lower resolution than an X-ray image. For this reason, it is possible to more reliably identify bone parts by blurring the X-ray image, reducing noise in the X-ray image, and setting the resolution to be equivalent to that of the DRR image at the time of learning. In the above-described embodiment, the DRR image and the X-ray image input to the learning model may be input after performing contrast normalization in advance. Further, a local contrast normalization layer or a local response normalization layer may be added to the intermediate layer.

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

In the embodiment described above, the bone density is reduced by creating a part of the plurality of DRR images as a DRR image in which the density of the bone region of the region including the bone portion is changed. A DRR image simulating the bone part is created and used for machine learning. On the other hand, in this embodiment, an X-ray image (high-voltage image) obtained by imaging a part of the plurality of X-ray images with a high voltage applied to the X-ray tube; By using dual energy subtraction that performs subtraction processing on an X-ray image (low-pressure image) taken with a low voltage applied to the X-ray tube, the bone region of the region including the bone portion An X-ray image having a changed density is used.

That is, when measuring the bone density of a subject for the diagnosis of osteoporosis, an X-ray image taken with a high voltage applied to the X-ray tube and a low voltage applied to the X-ray tube A configuration is adopted in which bone density is measured by dual energy subtraction for performing subtraction processing on the X-ray image. Even at the time of specifying the bone part image, this dual energy subtraction was used to capture an X-ray image taken with a high voltage applied to the X-ray tube and a low voltage applied to the X-ray tube. After weighting the X-ray image, a differential energy subtraction image representing the bone part is created by taking the difference between them. Then, by subtracting the dual energy subtraction image from the X-ray image (high pressure image or low pressure image), an image in which the density of the bone region is reduced among the regions including the bone portion (the bone portion in a state where the bone density is reduced). An image corresponding to an X-ray image) is obtained. By using this X-ray image with a low bone region density for machine learning, machine learning can be performed with an X-ray image simulating a bone part with reduced bone density, which can be applied to patients including osteoporosis patients. On the other hand, it is possible to improve the extraction accuracy of the bone part.

At this time, as an X-ray image used for machine learning, either a high pressure image, a low pressure image, or a dual energy subtraction image may be used, or an image obtained by connecting these images in the channel direction may be used. Good. In addition, instead of subtracting the dual energy subtraction image from the X-ray image (high-pressure image or low-pressure image), parameter adjustment is performed on the dual energy subtraction image, thereby simulating the bone portion having a reduced bone density. May be obtained.

Next, still another embodiment of the present invention will be described. FIG. 12 is a schematic diagram of the X-ray image 106 created by the X-ray image creation unit 81, and FIG. 13 is a schematic diagram of the DRR image 107 created by the DRR image creation unit 83.

This embodiment is used when a bone part image is created for a bone part having a symmetrical shape with respect to the body axis of a subject such as a femur. FIG. 6 described above is a schematic diagram of the X-ray image 101 near the subject's right foot, and FIG. 9 is a schematic diagram of the DRR image 103 near the subject's right foot. In contrast, FIG. 12 is a schematic diagram of an X-ray image 106 near the left foot of the subject, and FIG. 13 is a schematic diagram of a DRR image 107 near the left foot of the subject.

Thus, in the case where a bone part (femur 51 and pelvis 52) having a symmetrical shape with respect to the body axis of the subject is a target, the learning unit 85 performs an image of the right bone part on the left side. The machine learning is executed on the left and right bone images collectively by flipping one of the two bone images horizontally. For example, the X-ray image 106 near the subject's left foot shown in FIG. 12 is reversed left and right, and used together with the X-ray image 101 near the subject's right foot shown in FIG. 6 for machine learning. Similarly, the DRR image 107 near the subject's left foot shown in FIG. 13 is reversed left and right to be used together with the DRR image 103 near the subject's right foot shown in FIG. 9 for machine learning.

By adopting such a configuration, it is possible to make the detection accuracy uniform with respect to a bone portion having a symmetrical shape with respect to the body axis of the subject. And since machine learning is performed collectively on the images of the left and right bone parts, it is possible to expand the learning image and create a more accurate learning model.

In the above-described embodiment, machine learning is performed using both X-ray images and DRR images. However, machine learning may be performed using either one of the X-ray image and the DRR image.

In the above-described embodiment, a bone part is targeted as an organ, but an organ such as an organ may be targeted. For example, when the subject has a large amount of visceral fat, the concentration of the organ region is low during X-ray imaging. According to the present invention, even in such a case, it is possible to create a learning model corresponding to a subject whose organ concentration is low. For this reason, it becomes possible to improve the detection accuracy of an organ.

DESCRIPTION OF SYMBOLS 11 X-ray irradiation part 12 X-ray detection part 13 Top plate 14 Support | pillar 15 X-ray tube holding member 16 Rotation mechanism 17 Main support | pillar 18 Base plate 21 Display part 22 Operation part 70 CT image memory | storage part 80 Control part 81 X-ray image creation part 82 X-ray image storage unit 83 DRR image generation unit 84 DRR image storage unit 85 learning unit 86 bone image generation unit 300 CT image data

Claims

An image analysis method for performing segmentation for identifying an area of an organ by analyzing an image of an area including an organ of a subject,
While using machine learning as the segmentation technique,
A modified image creating step for creating a modified image in which the density of the region of the organ in the image including the organ of the subject is changed;
A learning model creating step of creating a learning model for machine learning by a learning process using an image including the organ of the subject and the modified image created in the modified image creating step;
An image analysis method comprising:
The image analysis method according to claim 1,
By performing conversion using the learning model created in the learning model creation step, the X-ray image of the region including the organ of the subject obtained by X-ray imaging of the subject, An image analysis method for creating an image representing an organ.
The image analysis method according to claim 1,
The image of the region including the organ of the subject is a DRR image created from the CT image data of the subject,
An image analysis method in which, in the modified image creation step, an area where the CT value of the CT image data is a predetermined value is defined as an area of the organ and the density thereof is changed.
The image analysis method according to claim 3,
At the time of creating the DRR image, a parameter including at least one of the projection coordinates and angle of the geometric condition is changed, or image processing including at least one of rotation, deformation, and enlargement / reduction of the image is performed, An image analysis method for creating a DRR image.
The image analysis method according to claim 3,
An image analysis method for executing at least one of contrast change, noise addition, and edge enhancement on a created DRR image.
The image analysis method according to claim 1,
The image of the region including the subject's organ is an X-ray image created by X-raying the subject,
An image analysis method in which the density of the region of the organ is changed using the X-ray image and the image of the organ obtained by using dual energy subtraction in the modified image creation step.
The image analysis method according to claim 2,
The organ obtained by performing conversion using an X-ray image of a region including the organ of the subject obtained by X-ray imaging of the subject and the learning model created in the learning model creation step An image analysis method that uses an image representing the image for learning of a learning model by the learning unit.
The image analysis method according to claim 1,
The organ has a bilaterally symmetric shape with respect to the body axis of the subject, and in the pre-learning model creation step, the right-side organ image or the left-side organ image is horizontally reversed. An image analysis method that creates a machine learning learning model for images of left and right organs at once.
The segmentation method of segmenting the region of the bone using the image analysis method according to claim 1, wherein the organ is a bone of the subject.
A bone density measuring method for measuring bone density with respect to a region of a bone portion segmented by the segmentation method according to claim 9.
This is a learning model creation method for creating a learning model used when performing segmentation for specifying a region of the organ by analyzing an image of the region including the organ of the subject using machine learning. And
Learning by performing machine learning learning using an image including the organ of the subject and a modified image created by changing the density of the region of the organ in the image including the organ of the subject A learning model creation method characterized by creating a model.
An image creation device for creating an image obtained by extracting an area of an organ from an X-ray image of an area including an organ of a subject,
An X-ray image storage unit for storing a plurality of X-ray images obtained by X-ray imaging of the region including the organ and a plurality of X-ray image teacher images for machine learning;
A DRR image creation unit for creating a DRR image of a region including the bone part;
DRR image storage for storing a plurality of DRR images created by the DRR image creation unit and a plurality of machine learning DRR image teacher images created based on the DRR image created by the DRR image creation unit And
Machine learning is performed using the plurality of X-ray images and the plurality of X-ray image teacher images stored in the X-ray image storage unit, and the plurality of the plurality of X-ray images stored in the DRR image storage unit A region including the organ of the subject using a learning model for recognizing the organ created in advance by performing machine learning using a DRR image and the plurality of DRR image teacher images An image creating unit that creates an image representing the organ by converting the X-ray image of
An image creating apparatus comprising:
The image creating apparatus according to claim 11.
The DRR image creation unit creates an DRR image of 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 bone part is changed.
The image creating apparatus according to claim 11.
Some X-ray images of the plurality of X-ray images stored in the X-ray image storage unit change the concentration of the organ region of the region including the organ by using dual energy subtraction. An image creation device that is an X-ray image.