WO2022176280A1 - X-ray imaging apparatus, image processing device, and image processing method - Google Patents

Abstract

This X-ray imaging apparatus (100) detects a region of interest (11) including a target object (102) left behind in the body of a subject (101) from an X-ray image (10) using a detection model (51) generated by machine learning. On the basis of the detected region of interest (11) and an output result of image processing by a removal model (52) generated by machine learning, a portion corresponding to the region of interest (11) in the X-ray image (10) is cut out and a partially highlighted image (14) in which the target object (102) is highlighted is generated. A display unit (4) is made to display a display in which the partially highlighted image (14) can be identified.

Description

X-ray imaging device, image processing device, and image processing method

The present invention relates to an X-ray imaging apparatus, an image processing apparatus, and an image processing method.

Conventionally, biological image processing devices are known that extract objects contained in the body from images of subjects. Such a biological image processing device is disclosed, for example, in Japanese Patent Application Laid-Open No. 2018-89301.

The biometric image processing apparatus described in JP-A-2018-89301 performs image processing for extracting a target object from a biometric image of a subject including the target object. Specifically, this biometric image processing apparatus generates a biometric image that does not include the target from the biometric image of the subject that includes the target using a learned neural network. Then, this biometric image processing apparatus extracts the object by subtracting the biometric image not including the object, which is the output image of the image processing by the neural network, from the biometric image of the subject including the object. is configured to generate a subtracted image.

Japanese Unexamined Patent Application Publication No. 2018-89301

Here, although not described in JP-A-2018-89301, a worker such as a doctor is left behind in the body of a subject (subject) after surgery Hemostasis gauze (surgical gauze) In order to confirm the presence or absence of a foreign body (target object), an X-ray image is taken of the subject after surgery. Therefore, like the biological image processing apparatus described in JP-A-2018-89301, image processing using a neural network (learned model) is performed on the captured X-ray image (biological image). Extracting target objects such as surgical gauze included in the X-ray image of the subject after surgery by It is conceivable to generate an enhanced (enhanced) image.

However, when generating an image in which the target object is emphasized by performing image processing on the X-ray image of the subject using the trained model, human body structures such as bones are emphasized in the same way as the target object. Sometimes. In this case, there is a problem that confirmation of the target object included in the X-ray image becomes difficult due to the fact that structures other than the target object are emphasized in the same way as the target object.

The present invention has been made to solve the above problems, and one object of the present invention is to perform image processing using a trained model generated by machine learning to photograph a subject. To provide an X-ray imaging apparatus, an image processing apparatus, and an image processing method, which enable easy confirmation of a target object included in an X-ray image when the target object included in the X-ray image obtained from radiography is emphasized. is.

In order to achieve the above object, an X-ray imaging apparatus according to a first aspect of the present invention includes an X-ray irradiator that irradiates a subject with X-rays, and an X-ray that detects the X-rays emitted from the X-ray irradiator. a ray detector, an X-ray image generator for generating an X-ray image based on an X-ray detection signal detected by the X-ray detector, and a controller, wherein the controller controls the X-ray image. a region detection unit for detecting a region of interest from an X-ray image by a first trained model generated by machine learning so as to detect a region of interest including a target object in the body of a subject; and detection by the region detection unit corresponding to the region of interest in the X-ray image based on the obtained region of interest and the output result of image processing by the second trained model generated by machine learning to remove or enhance the object of interest in the X-ray image. a partial image generation unit for generating an enhanced partial image in which a target object is emphasized while a portion is cut out; an image output unit for causing a display unit to display a display in which the enhanced partial image generated by the partial image generation unit is identifiable; including. It should be noted that the "removal of the target object" referred to here is described as a broad concept including weakening (suppressing) the image of the target object.

An image processing apparatus according to a second aspect of the present invention detects a region of interest including a target object inside the body of a subject in an X-ray image generated based on a detection signal of X-rays irradiated to the subject. By the first learned model generated by machine learning so as to remove or An enhanced partial image in which a portion corresponding to the region of interest in the X-ray image is cut out and the target object is enhanced based on the output result of image processing by the second trained model generated by machine learning so as to enhance. and an image output unit configured to cause a display unit to display an identifiable display of the enhanced partial image generated by the partial image generation unit.

In the image processing method according to the third aspect of the present invention, a region of interest including a target object inside the body of the subject is detected in an X-ray image generated based on a detection signal of X-rays irradiated to the subject. detecting a region of interest from the X-ray image by a first trained model generated by machine learning to remove or enhance the detected region of interest and the object of interest in the X-ray image; generating an enhanced partial image in which a portion corresponding to the region of interest in the X-ray image is cut out and the target object is emphasized, based on the output result of image processing by the second trained model generated by and causing the display unit to display a display capable of identifying the highlighted partial image.

In the X-ray imaging apparatus according to the first aspect, the image processing apparatus according to the second aspect, and the image processing method according to the third aspect, the portion corresponding to the region of interest in the X-ray image is cut out and the target object is extracted. generates an enhanced partial image in which is enhanced. Then, a display that allows identification of the generated emphasized partial image is displayed on the display unit. As a result, only the portion corresponding to the region of interest containing the target object is cut out and emphasized, so that structures such as bones of the human body in regions other than the portion containing the target object are emphasized in the same way as the target object. can be suppressed. Therefore, it is possible to prevent difficulty in identifying the target object due to enhancement of the human body structure other than the portion including the target object. As a result, when the target object included in the X-ray image of the subject is emphasized by performing image processing using a trained model generated by machine learning, the target object included in the X-ray image can be easily identified. can be verified.

1 is a diagram for explaining the configuration of an X-ray imaging apparatus according to a first embodiment; FIG. 1 is a block diagram for explaining the configuration of an X-ray imaging apparatus according to a first embodiment; FIG. It is a figure which shows an example of the X-ray image of the subject in which a target object is contained in the body. FIG. 4 is a diagram for explaining generation of an identification image according to the first embodiment; FIG. 4 is a diagram for explaining detection of a region of interest according to the first embodiment; FIG. 4 is a diagram for explaining generation of a removed image according to the first embodiment; FIG. 4 is a diagram for explaining synthesis of an enhanced partial image and a contour partial image according to the first embodiment; It is the figure which showed the display of the display part by 1st Embodiment. FIG. 3 is a flowchart for explaining an image processing method according to the first embodiment; FIG. 3 is a block diagram for explaining the configuration of an X-ray imaging apparatus according to a second embodiment; FIG. FIG. 10 is a diagram for explaining generation of an identification image according to the second embodiment; FIG. 11 is a block diagram for explaining the configuration of an X-ray imaging apparatus according to a third embodiment; FIG. FIG. 11 is a diagram for explaining generation of an identification image according to the third embodiment; FIG. 11 is a diagram for explaining generation of an enhanced image according to the third embodiment; FIG.

An embodiment embodying the present invention will be described below based on the drawings.

[First embodiment]
(Overall configuration of X-ray imaging apparatus)
An X-ray imaging apparatus 100 according to a first embodiment of the present invention will be described with reference to FIGS. 1 to 8. FIG.

As shown in FIG. 1, the X-ray imaging apparatus 100 performs X-ray imaging to identify a target object 102 inside the body of a subject 101 . For example, the X-ray imaging apparatus 100 applies X-rays to a subject 101 after undergoing laparotomy in an operating room to confirm whether or not a target object 102 (foreign matter) remains in the body. take a picture. The X-ray imaging apparatus 100 is, for example, a mobile X-ray imaging apparatus in which the entire apparatus is movable. Target objects 102 include, for example, surgical gauze, suture needles, and forceps (such as hemostatic forceps).

In general, when a surgical operation such as laparotomy is performed, a worker such as a doctor leaves a target object 102 such as surgical gauze, a suture needle, and forceps in the body of the subject 101 after the wound is closed ( X-ray imaging for confirmation is performed on the subject 101 so that there is no residual. A worker such as a doctor confirms that the target object 102 is not left inside the body of the subject 101 by viewing the captured X-ray image 10 (see FIG. 3).

<Parts of X-ray equipment>
As shown in FIG. 2, the X-ray imaging apparatus 100 includes an X-ray irradiation unit 1, an X-ray detection unit 2, an X-ray image generation unit 3, a display unit 4, a storage unit 5, and a control unit 6. The control unit 6 is an example of the "control unit" and the "image processing device" in the claims.

The X-ray irradiation unit 1 irradiates the subject 101 after surgery with X-rays. The X-ray irradiation unit 1 includes an X-ray tube that emits X-rays when a voltage is applied.

The X-ray detection unit 2 detects X-rays emitted by the X-ray irradiation unit 1 and transmitted through the subject 101 . Then, the X-ray detector 2 outputs a detection signal based on the detected X-rays. The X-ray detection unit 2 includes, for example, an FPD (Flat Panel Detector). The X-ray detector 2 is configured as a wireless type X-ray detector and outputs a detection signal as a radio signal. Specifically, the X-ray detection unit 2 is configured to be able to communicate with an X-ray image generation unit 3, which will be described later, through a wireless connection such as a wireless LAN. Outputs a detection signal.

As shown in FIG. 3 , the X-ray image generator 3 generates an X-ray image 10 based on the X-ray detection signal detected by the X-ray detector 2 . The X-ray image generation unit 3 controls X-ray imaging by controlling the X-ray irradiation unit 1 and the X-ray detection unit 2 . The X-ray image generator 3 is configured to be able to communicate with the X-ray detector 2 through a wireless connection such as a wireless LAN. The X-ray image generator 3 includes a processor such as an FPGA (field-programmable gate array). Then, the X-ray image generator 3 outputs the generated X-ray image 10 to the controller 6, which will be described later.

The X-ray image 10 is an image obtained by X-raying the abdomen of the subject 101 after surgery. For example, X-ray image 10 includes surgical gauze as target object 102 . The surgical gauze is woven with a contrast thread that does not easily transmit X-rays so that it can be visually recognized in the X-ray image 10 obtained by radiography after surgery.

The display unit 4 includes, for example, a touch panel type liquid crystal display. Then, the display unit 4 displays the captured X-ray image 10 . The display unit 4 also displays an identification image 20 (see FIGS. 4 and 8) output by the image output unit 65, which will be described later. Further, the display unit 4 is configured to receive an input operation for operating the X-ray imaging apparatus 100 by an operator such as a doctor, based on the operation on the touch panel.

The storage unit 5 is composed of a storage device such as a hard disk drive, for example. The storage unit 5 stores image data such as an X-ray image 10 generated by the X-ray image generation unit 3 and an identification image 20 (see FIG. 4) generated by the control unit 6, which will be described later. The storage unit 5 is also configured to store various setting values for operating the X-ray imaging apparatus 100 . The storage unit 5 also stores a program used for control processing of the X-ray imaging apparatus 100 by the control unit 6 . The storage unit 5 also stores in advance a detection model 51 and a removal model 52, which will be described later. Note that the detection model 51 is an example of a "first trained model" in the scope of claims. Also, the elimination model 52 is an example of a "second trained model" in the scope of claims.

The control unit 6 is a computer including, for example, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), a ROM (Read Only Memory) and a RAM (Random Access Memory). The control unit 6 includes an area detection unit 61, a partial image generation unit 62, a contour image generation unit 63, an identification image generation unit 64, and an image output unit 65 as functional configurations. That is, the area detection unit 61, the partial image generation unit 62, the contour image generation unit 63, the identification image generation unit 64, and the image output unit 65 are function blocks as software in the control unit 6, and are hardware. is configured to function by executing a predetermined control program.

(Generation of identification image)
As shown in FIG. 4, the control unit 6 is configured to execute image processing for generating an identification image 20 for identifying a target object 102 included in the X-ray image 10 from the X-ray image 10. . Specifically, the controller 6 acquires the X-ray image 10 generated by the X-ray image generator 3 . Then, the control unit 6 performs processes of detecting a region of interest 11 , generating a removed image 12 , and generating a contour image 15 on the X-ray image 10 . The region of interest 11 is a region of the X-ray image 10 that includes the target object 102 inside the body of the subject 101 . The removed image 12 is an image obtained by removing the target object 102 from the X-ray image 10 . The contour image 15 is an image showing the contour (extracted contour) of the subject 101 (and the bones of the subject 101, etc.) in the X-ray image 10 . The control unit 6 also generates an enhanced whole image 13 in which the target object 102 is enhanced (extracted) based on the removed image 12 and the X-ray image 10 . Then, the control unit 6 generates an enhanced partial image 14 by cutting out a portion corresponding to the region of interest 11 in the enhanced whole image 13, and synthesizes the enhanced partial image 14 with the contour image 15 to form the identification image 20. Generate.

<Detection of region of interest>
As shown in FIGS. 4 and 5, in the first embodiment, the region detection unit 61 (control unit 6) detects the region of interest 11 from the X-ray image 10 using the detection model 51. FIG. Specifically, the region detection unit 61 acquires the position (coordinates) and size (size) of the region of interest 11 based on the output from the detection model 51 to which the X-ray image 10 is input. The detection model 51 is a trained model (algorithm) generated by machine learning using deep learning so as to detect the region of interest 11 including the target object 102 in the X-ray image 10 .

The detection model 51 is generated in advance by a learning device 103 separate from the X-ray imaging apparatus 100 . The learning device 103 is, for example, a computer including a CPU, GPU, ROM, and RAM. The learning device 103 generates a detection model 51 by machine learning using a plurality of teacher input X-ray images 10a and a plurality of teacher output regions of interest 11a as teacher data (training set). The teacher input X-ray image 10a is generated so as to simulate the X-ray image 10 obtained by imaging the subject 101 with the target object 102 left inside the body. Also, the teacher input X-ray image 10 a is generated so as to have the same conditions (size, etc.) as the X-ray image 10 used for input in the inference using the detection model 51 . The teacher output region of interest 11a is information (annotation) indicating the position and size of the region corresponding to the target object 102 in the teacher input X-ray image 10a.

As a region detection (object detection) algorithm in the detection model 51, for example, FasterR-CNN (Faster Region-based Convolutional Neural Network) is used. In object detection using FasterR-CNN, one or more target objects are detected by being surrounded by a rectangular area (bounding box) in an input image. The likelihood (decision value) for each target object detection (inference) result is also estimated at the same time. The likelihood is a numerical value indicating the possibility (likelihood) that the target object is included in the detected rectangular area. In the first embodiment, the detection model 51 detects one region with the highest likelihood among multiple regions estimated to contain the target object 102 . That is, the region detection unit 61 is configured to detect one region with the highest likelihood from the X-ray image 10 by the detection model 51 as the region of interest 11 .

<Generation of enhanced partial image>
As shown in FIGS. 4 and 6, in the first embodiment, the partial image generation unit 62 of the control unit 6 combines the region of interest 11 detected by the region detection unit 61 with the output result of image processing by the removal model 52. Based on this, an enhanced partial image 14 is generated. The partial image generation unit 62 (control unit 6 ) generates the removal image 12 from the X-ray image 10 as an output result of image processing by the removal model 52 . Specifically, the partial image generation unit 62 generates the removal image 12 corresponding to the entire X-ray image 10 from the entire X-ray image 10 using the removal model 52 . The removal model 52 is machine learning using deep learning so as to remove (suppress) the target object 102 in the X-ray image 10 to generate the removed image 12 in which the target object 102 is removed from the X-ray image 10. It is a trained model (algorithm) generated by

The removal model 52 is generated in advance by a learning device 103 separate from the X-ray imaging apparatus 100, similar to the detection model 51. The learning device 103 generates a removal model 52 by machine learning using a plurality of teacher input X-ray images 10b and a plurality of teacher output removal images 12b as teacher data (training set). The teacher input X-ray image 10b is generated so as to simulate the X-ray image 10 obtained by imaging the subject 101 with the target object 102 left inside the body. The teacher output removed image 12b is an image obtained by removing the target object 102 from the teacher input X-ray image 10b. The teacher input X-ray image 10b and the teacher output removal image 12b are generated so as to have the same conditions (size, etc.) as the X-ray image 10 used as an input in inference using the removal model 52 .

The removal model 52 is generated, for example, based on U-Net, which is one type of full convolution network (FCN). The removal model 52 transforms the pixels estimated to be the target object 102 from among the pixels of the input X-ray image 10, thereby removing the portion estimated to be the target object 102 from the X-ray image 10. is generated by learning to perform an image transformation (image reconstruction) that removes

Then, in the first embodiment, the partial image generation unit 62 (control unit 6) generates the enhanced partial image 14 based on the X-ray image 10, the removed image 12, and the region of interest 11. The enhanced partial image 14 is an image obtained by cutting out a portion corresponding to the region of interest 11 in the X-ray image 10 and enhancing the target object 102 . Specifically, the partial image generator 62 generates the enhanced whole image 13 in which the target object 102 in the entire X-ray image 10 is emphasized based on the difference between the X-ray image 10 and the removed image 12 . That is, the partial image generator 62 subtracts the removed image 12 from the X-ray image 10 to generate the enhanced whole image 13 having the same size (region) as the X-ray image 10 . The enhanced whole image 13 is an image in which the target object 102 included in the X-ray image 10 is enhanced (extracted). In addition, the enhanced whole image 13 includes structures other than the target object 102, such as bones of the subject 101. FIG. The partial image generation unit 62 is configured to generate the emphasized partial image 14 by cutting out the portion corresponding to the region of interest 11 from the generated emphasized whole image 13 . That is, in the first embodiment, the enhanced partial image 14 has the same position and size as the position and size in the X-ray image 10 of the region of interest 11 detected by the detection model 51 .

<Generation of contour image>
As shown in FIG. 4, in the first embodiment, the contour image generator 63 (controller 6) generates the contour image 15. FIG. The contour image generator 63 generates a contour image 15 by executing image processing for contour extraction (edge detection) on the X-ray image 10 . For example, in the first embodiment, the contour image generation unit 63 blurs the X-ray image 10 . Then, the contour image generator 63 generates the contour image 15 based on the difference between the X-ray image 10 after blurring and the X-ray image 10 before blurring. The generated contour image 15 is an image showing high frequency components of the X-ray image 10 . The contour image generator 63 is configured to be able to extract a contour with a lower intensity (lower contrast) than the enhanced whole image 13 from the X-ray image 10 by executing the blurring process.

<Synthesis of emphasized partial image and contour image>
As shown in FIGS. 4 and 7, in the first embodiment, the identification image generator 64 (controller 6) generates the identification image 20 based on the emphasized partial image 14 and the contour image 15. FIG. Specifically, the identification image generation unit 64 cuts out a region having the same position and size as the region of interest 11 (a region having the same position and size as the emphasized partial image 14) from the contour image 15, thereby extracting the contour portion. Generate image 16 . Then, the identification image generator 64 combines the outline partial image 16 and the emphasized partial image 14 to generate a combined partial image 17 . Then, the identification image generator 64 generates the identification image 20 by synthesizing the combined partial image 17 with the clipped portion of the contour image 15 .

Further, in the first embodiment, the identification image generation unit 64 (control unit 6) generates the pixel values of the emphasized partial image 14 and the pixel values of the portion (outline partial image 16) corresponding to the emphasized partial image 14 in the contour image 15. and are each weighted, and the contour image 15 and the emphasized partial image 14 are synthesized to generate the identification image 20 . For example, the identification image generator 64 weights the transparency of each pixel of the outline partial image 16 and the emphasized partial image 14 based on the luminance value of the pixels of the weighted image 18 . In other words, the identification image generation unit 64 displays the emphasized partial image 14 in the portion (white portion) of the weighted image 18 where the luminance value is large, and displays the outline in the portion (black portion) of the weighted image 18 where the luminance value is small. A composite partial image 17 is generated so that the image 15 (contour partial image 16) is displayed. The weighted image 18 is generated by the identification image generator 64 so as to have the same size as the region of interest 11 . Also, the weighted image 18 is configured such that the brightness value is large near the center and gradually decreases circularly (radially) toward the periphery.

(Display on display)
As shown in FIG. 8, in the first embodiment, the image output unit 65 (control unit 6) causes the display unit 4 to display a display that allows the generated emphasized partial image 14 to be identified. Specifically, the image output unit 65 is configured to display the identification image 20 generated by the identification image generation unit 64 and the X-ray image 10 on the display unit 4 . For example, the image output unit 65 is configured to display the X-ray image 10 and the identification image 20 side by side on the display unit 4 . Note that the image output unit 65 can display the identification image 20 and the X-ray image 10 side by side based on an input operation on the touch panel of the display unit 4, or can display only the X-ray image 10 without displaying the identification image 20. It may be configured to switch between causing and causing.

(Image processing method according to the first embodiment)
Next, with reference to FIG. 9, a control processing flow regarding the image processing method according to the first embodiment will be described. Further, steps 401 to 403 indicate control processing by the X-ray image generation unit 3, and steps 404 to 410 indicate control processing by the control unit 6. FIG.

First, in step 401, the subject 101 is irradiated with X-rays in order to detect the target object 102 inside the body of the subject 101 after surgery. Next, in step 402, emitted x-rays are detected. Next, in step 403, an X-ray image 10 is generated based on the detected X-ray detection signal.

Next, in step 404 , the detection model 51 detects the region of interest 11 from the X-ray image 10 . Then, the size (magnitude) of the detected region of interest 11 and the position (coordinates) in the X-ray image 10 are acquired.

Next, in step 405 , a removal image 12 is generated from the X-ray image 10 as an output result of image processing by the removal model 52 . An enhanced global image 13 is then generated in step 406 based on the difference between the X-ray image 10 and the removed image 12 . Then, in step 407 , based on the enhanced whole image 13 and the region of interest 11 , the enhanced partial image 14 is generated by cutting out the region corresponding to the region of interest 11 from the enhanced whole image 13 . That is, through the processing in steps 405 to 407, the enhanced partial image 14 is generated based on the detected region of interest 11 and the output result of the image processing by the removal model 52. FIG.

Next, in step 408, a contour image 15 is generated from the X-ray image 10. Then, in step 409 , an identification image 20 is generated based on the enhanced partial image 14 and the contour image 15 . Specifically, based on the weighted image 18 , the contour partial image 16 cut out from the contour image 15 and the emphasized partial image 14 are combined to generate a combined partial image 17 . An identification image 20 is generated by synthesizing the synthetic partial image 17 with the outline image 15 .

Next, in step 410, the display unit 4 displays an identifiable display of the emphasized partial image 14. That is, the identification image 20 is displayed on the display section 4 . Specifically, the X-ray image 10 and the identification image 20 are displayed side by side on the display unit 4 .

Any of the detection of the region of interest 11 in step 404, the generation of the removed image 12 in step 405, and the generation of the contour image 15 in step 408 may be executed first.

(Effect of the first embodiment)
The following effects can be obtained in the first embodiment.

As described above, the X-ray imaging apparatus 100 of the first embodiment generates the enhanced partial image 14 in which the portion corresponding to the region of interest 11 in the X-ray image 10 is clipped and the target object 102 is enhanced. Then, the display section 4 is caused to display a display that allows the generated emphasized partial image 14 to be identified. As a result, only the portion corresponding to the region of interest 11 including the target object 102 is cut out and emphasized, so that structures such as human bones in regions other than the portion including the target object 102 are similar to the target object 102. can be suppressed. Therefore, it is possible to suppress difficulty in identifying the target object 102 due to enhancement of the human body structure other than the portion including the target object 102 . As a result, when the target object 102 included in the X-ray image 10 obtained by imaging the subject 101 is emphasized by performing image processing using a trained model generated by machine learning, the target object 102 included in the X-ray image 10 is The target object 102 can be easily confirmed.

In addition, in the above-described first embodiment, the following further effects can be obtained by configuring as follows.

That is, in the first embodiment, as described above, the control unit 6 controls the target object 102 included in the X-ray image 10 based on the enhanced partial image 14 generated by the partial image generation unit 62 (control unit 6). The image output unit 65 is configured to display the identification image 20 generated by the identification image generation unit 64 on the display unit 4. . With this configuration, a worker such as a doctor can easily visually recognize the target object 102 included in the X-ray image 10 from the identification image 20 displayed on the display unit 4 . Therefore, an operator such as a doctor can easily confirm the target object 102 included in the X-ray image 10 by viewing the display unit 4 .

Further, in the first embodiment, as described above, the control unit 6 includes the contour image generation unit 63 that generates the contour image 15 showing the contour of the subject 101 in the X-ray image 10, and the identification image generation unit 64 (Control section 6) generates identification image 20 based on contour image 15 generated by contour image generation section 63 and enhanced partial image 14 generated by partial image generation section 62 (control section 6). is configured as With this configuration, since the identification image 20 is generated based on the contour image 15 from which the contour of the object 101 is extracted and the enhanced partial image 14, a portion of the identification image 20 other than the target object 102 (enhanced partial image) is generated. 14) can be used as a contour image 15 showing the contour of the subject 101. FIG. Therefore, the identification image 20 can be generated so that the portion of the target object 102 is emphasized while the portion other than the outline of the subject 101 included in the X-ray image 10 is removed. Parts of the object 102 can be made more emphasized. As a result, since the identification image 20 can be generated so that the target object 102 can be more effectively identified, the target object 102 included in the X-ray image 10 can be more easily confirmed.

Further, in the first embodiment, as described above, the identification image generation unit 64 (the control unit 6) generates the pixel values of the emphasized partial image 14 and the pixel values of the portion of the contour image 15 corresponding to the emphasized partial image 14. are weighted, and the contour image 15 and the emphasized partial image 14 are synthesized to generate the identification image 20 . With this configuration, weighting is applied so that the boundary between the enhanced partial image 14 and the contour image 15 is smoothly switched, thereby enhancing the contour image 15 while smoothing the boundary line (outer peripheral portion) of the enhanced partial image 14. Partial images 14 can be synthesized. Therefore, it is possible to prevent the boundary between a portion that is emphasized as including the target object 102 in the identification image 20 and a portion that is not emphasized from becoming unnecessarily conspicuous. Therefore, it is possible to prevent the generated identification image 20 from being significantly different from the original X-ray image 10 due to the conspicuous boundary between the emphasized portion and the non-emphasized portion in the identification image 20. can be done. As a result, the difference between the identification image 20 and the X-ray image 10 can be reduced, so that the target object 102 included in the X-ray image 10 can be more easily confirmed.

Further, in the first embodiment, as described above, the contour image generation unit 63 (the control unit 6) performs the blurring process on the X-ray image 10, and the X-ray image 10 after the blurring process is performed. , a contour image 15 is generated based on the difference from the X-ray image 10 before blurring. With this configuration, by executing the blurring process to generate the contour image 15, the contour image 15 can be generated in a state where the intensity (contrast) of the contour of the subject 101 in the X-ray image 10 is relatively small. can be done. Therefore, the identification image 20 can be generated based on the contour image 15 in which the contrast of the structure included in the image is relatively low and the enhanced partial image 14 in which the target object 102 is emphasized and the contrast is relatively high. 20, the visibility of the part including the target object 102 can be relatively improved. As a result, the target object 102 included in the X-ray image 10 can be confirmed more easily.

Further, in the first embodiment, the image output unit 65 (control unit 6) is configured to output the X-ray image 10 and the identification image 20 (220) to the display unit 4 as described above. With this configuration, an operator such as a doctor can easily compare the identification image 20 and the X-ray image 10 in which the target object 102 is emphasized so that the object 102 can be easily identified. Therefore, an operator such as a doctor can more easily confirm the target object 102 included in the X-ray image 10 by comparing the identification image 20 and the X-ray image 10 .

Further, in the first embodiment, as described above, the partial image generation unit 62 (control unit 6) determines the position and It is configured to generate an enhanced partial image 14 having a position and size equal to the size. With this configuration, it is possible to generate the enhanced partial image 14 having the same position and size as the portion estimated to be the target object 102 by the detection model 51, so that the position and the size of the detected target object 102 can be generated. Areas equal in size can be emphasized. Therefore, regardless of whether the size of the target object 102 included in the X-ray image 10 is relatively small or relatively large, it is possible to suppress the deterioration of the visibility of the target object 102. The target object 102 included in the line image 10 can be easily confirmed.

Further, in the first embodiment, as described above, the partial image generation unit 62 (control unit 6) removes the target object 102 from the entire X-ray image 10 using the removal model 52 (second learned model). Based on the difference between the X-ray image 10 and the removed image 12, an enhanced whole image 13 in which the target object 102 in the entire X-ray image 10 is enhanced is generated. At the same time, by cutting out a portion corresponding to the region of interest 11 from the generated enhanced whole image 13, an enhanced partial image 14 is generated. With this configuration, the enhanced overall image 13 can be easily generated from the entire X-ray image 10 without specifying in advance the portion containing the target object 102 . As a result, the enhanced partial image 14 cut out from the enhanced whole image 13 can be easily generated, so that the target object 102 included in the X-ray image 10 can be easily confirmed.

Further, in the first embodiment, as described above, the region detection unit 61 (control unit 6) determines that the target object 102 is included in the X-ray image 10 by the detection model 51 (first learned model). It is configured to detect one region with the highest likelihood as the region of interest 11 from among the plurality of estimated regions. Here, when multiple regions of interest 11 are detected, multiple regions corresponding to the multiple detected regions of interest 11 are considered to be emphasized. In this case, it becomes difficult to identify a portion including the target object 102 in the X-ray image 10 due to detection of a region other than the target object 102 . On the other hand, in the first embodiment, the region detection unit 61 detects one region of interest 11 from the X-ray image 10 that has the highest likelihood that the detection model 51 estimates that the target object 102 is included. Configure to detect. With this configuration, unlike the case where a plurality of regions of interest 11 are detected, only the portion corresponding to one region of interest 11 is emphasized, making it difficult to identify the portion containing the target object 102. can be suppressed. As a result, the part including the target object 102 can be easily identified by viewing the display on the display unit 4 , so the target object 102 included in the X-ray image 10 can be more easily confirmed.

(Effect of the image processing method of the first embodiment)
With the image processing method of the first embodiment, the following effects can be obtained.

In the image processing method of the first embodiment, by configuring as described above, the portion corresponding to the region of interest 11 in the X-ray image 10 is cut out and the enhanced partial image 14 in which the target object 102 is enhanced is generated. Then, the display section 4 is caused to display a display that allows the generated emphasized partial image 14 to be identified. As a result, only the portion corresponding to the region of interest 11 including the target object 102 is cut out and emphasized, so that structures such as human bones in regions other than the portion including the target object 102 are similar to the target object 102. can be suppressed. Therefore, it is possible to suppress difficulty in identifying the target object 102 due to enhancement of the human body structure other than the portion including the target object 102 . As a result, when the target object 102 included in the X-ray image 10 obtained by imaging the subject 101 is emphasized by performing image processing using a trained model generated by machine learning, the target object 102 included in the X-ray image 10 is It is possible to provide an image processing method that allows the target object 102 to be easily confirmed.

[Second embodiment]
Next, an X-ray imaging apparatus 200 according to a second embodiment will be described with reference to FIGS. 10 and 11. FIG. Unlike the first embodiment configured to generate the ablation image 12 from the entire X-ray image 10 by the ablation model 52, in the second embodiment, the portion of the X-ray image 10 corresponding to the region of interest 11 is (X-ray partial image 210 ) is cut out to generate a removed image 212 . In addition, in 2nd Embodiment, the same code|symbol is attached|subjected regarding the structure similar to the said 1st Embodiment, and description is abbreviate|omitted.

As shown in FIG. 10, the X-ray imaging apparatus 200 of the second embodiment includes a control section 206. As shown in FIG. As with the control unit 6 of the first embodiment, the control unit 206 performs image processing for generating an identification image 220 for identifying the target object 102 included in the X-ray image 10 from the X-ray image 10. is configured to Further, the control unit 206 includes an area detection unit 61, a partial image generation unit 262, a contour image generation unit 63, an identification image generation unit 264, and an image output unit 65 as functional configurations. Note that the control unit 206 is an example of a “control unit” and an “image processing device” in the claims.

As shown in FIG. 11, the region detection unit 61 (control unit 206) detects the region of interest 11 from the X-ray image 10 using the detection model 51, as in the first embodiment. Also, the contour image generator 63 (controller 206 ) generates the contour image 15 from the X-ray image 10 .

In the X-ray imaging apparatus 200 according to the second embodiment, the partial image generation unit 262 (control unit 206) cuts out a portion corresponding to the region of interest 11 of the X-ray image 10 from the X-ray image 10 based on the region of interest 11. to generate an X-ray partial image 210 . The X-ray partial image 210 has a predetermined (constant) size centered on the position of the region of interest 11 in the X-ray image 10 regardless of the size of the region of interest 11 . For example, if the target object 102 is surgical gauze, the predetermined size is set to be slightly larger than the size of the surgical gauze in the X-ray image 10 .

In the second embodiment, the partial image generation unit 262 (control unit 206) cuts out a portion corresponding to the region of interest 11 from the X-ray partial image 210 as an output result of the image processing by the removal model 252, and It is configured to generate an ablation image 212 with the object 102 removed. That is, the partial image generator 262 uses the removal model 252 to generate a removal image 212 of a predetermined size equal to the X-ray partial image 210 from the X-ray partial image 210 cut out to a predetermined size. The removal model 252 is a trained model generated by machine learning using deep learning so as to generate the removal image 212 in which the target object 102 is removed from the X-ray partial image 210 . The removal model 252 is generated in advance by the learning device 103 separate from the X-ray imaging apparatus 200 and stored in the storage unit 5, similarly to the removal model 52 according to the first embodiment. The removed model 252 is an example of a "second trained model" in the claims.

Then, the partial image generation unit 262 (control unit 206) creates a portion (X-ray partial image 210) corresponding to the region of interest 11 of the X-ray image 10 and a removed image 212 obtained by clipping the portion corresponding to the region of interest 11. is configured to generate an enhanced partial image 214 based on the difference between . Similar to the X-ray partial image 210 , the enhanced partial image 214 has a predetermined size centered on the position of the detected region of interest 11 in the X-ray image 10 regardless of the size of the region of interest 11 . That is, the partial image generator 262 generates an enhanced partial image 214 having the same size as the X-ray partial image 210 and the removed image 212 .

The identification image generation unit 264 (control unit 206) generates the identification image 220 based on the emphasized partial image 214 and the contour image 15, as in the first embodiment. Specifically, the identification image generator 264 generates a combined partial image 217 by combining the contour partial image 216 cut out from the contour image 15 and the emphasized partial image 214 . A contour partial image 216 is generated by cutting out a region having a predetermined size equal to that of the emphasized partial image 214 from a position corresponding to the region of interest 11 in the contour image 15 . Further, the identification image generator 264 synthesizes the contour image 15 (contour partial image 216) and the emphasized partial image 214 in a weighted state based on the weighted image 218, as in the first embodiment. The weighted image 218 likewise has the same size as the contour partial image 216 and the enhanced partial image 214 . Weighting based on the weight image 218 in synthesizing the contour image 15 (contour partial image 216) and the enhanced partial image 214 (generating the identification image 220) is the same as in the first embodiment.

Further, as in the first embodiment, the image output unit 65 (control unit 206) is configured to display the X-ray image 10 and the identification image 220 side by side on the display unit 4. Other configurations of the second embodiment are the same as those of the first embodiment.

(Effect of Second Embodiment)
The following effects can be obtained in the second embodiment.

In the second embodiment, as described above, the partial image generation unit 262 (control unit 206) generates a predetermined is configured to generate an enhanced partial image 214 having a size of . With this configuration, since the emphasized partial image 214 has a predetermined size, the size of the region emphasized in the display on the display unit 4 can be made constant. Therefore, it is possible to suppress the size of the region to be emphasized from being too small and from being too large, so it is possible to suppress deterioration in the visibility of the target object 102 in the display on the display unit 4 .

Further, in the second embodiment, as described above, the partial image generation unit 262 (control unit 206) uses the removal model 252 (second trained model) to generate a portion ( A removed image 212 is generated by cutting out a portion corresponding to the region of interest 11 from the X-ray partial image 210) and removing the target object 102, which corresponds to the region of interest 11 of the X-ray image 10). Based on the difference between the part (X-ray partial image 210) and the removed image 212 in which the part corresponding to the region of interest 11 is cut out, an enhanced partial image 214 is generated. With this configuration, since the partial X-ray image 210, which is the clipped portion of the X-ray image 10, is used to generate the removed image 212, the entire X-ray image 10 is The processing load of image processing using the removal model 252 by the partial image generation unit 262 can be reduced compared to the case where the image processing is performed using the partial image generation unit 262 . Other effects of the second embodiment are the same as those of the first embodiment.

[Third embodiment]
Next, an X-ray imaging apparatus 300 according to a third embodiment will be described with reference to FIGS. 12 to 14. FIG. Unlike the first embodiment, which was configured to use an ablation model 52 trained to generate the ablation image 12 from the X-ray image 10, the third embodiment generates an enhanced image 313 from the X-ray image 10. It is configured to use an enhancement model 352 that has been trained to do so. Note that the emphasis model 352 is an example of a "second trained model" in the scope of claims. In addition, in the third embodiment, the same reference numerals are given to the same configurations as in the first embodiment, and the description thereof is omitted.

As shown in FIG. 12, the X-ray imaging apparatus 300 of the third embodiment includes a control section 306. As shown in FIG. Similar to the control unit 6 of the first embodiment, the control unit 306 generates an identification image 320 (see FIG. 13) for identifying the target object 102 included in the X-ray image 10 from the X-ray image 10. configured to do the work. Further, the control unit 306 includes an area detection unit 61, a partial image generation unit 362, a contour image generation unit 63, an identification image generation unit 364, and an image output unit 65 as functional configurations. Note that the control unit 306 is an example of a “control unit” and an “image processing device” in the claims.

As shown in FIG. 13, the region detection unit 61 (control unit 306) detects the region of interest 11 from the X-ray image 10 using the detection model 51, as in the first embodiment. Further, the contour image generator 63 (controller 306 ) generates the contour image 15 from the X-ray image 10 .

In the third embodiment, the partial image generation unit 362 (control unit 306) produces an image in which the target object 102 is emphasized (extracted) from the X-ray image 10 as the output result of image processing by the emphasis model 352. An enhanced image 313 is generated. Specifically, the partial image generator 362 generates an enhanced image 313 corresponding to the entire X-ray image 10 from the entire X-ray image 10 by the enhancement model 352 . The enhancement model 352 is a trained model (algorithm) generated by machine learning using deep learning so as to enhance the target object 102 in the X-ray image 10 . Note that the enhanced image 313 includes not only an image in which only the target object 102 is extracted, but also an image including structures such as bones of the subject 101 other than the target object 102 .

As shown in FIG. 14, the enhancement model 352 is generated in advance by the learning device 103 separate from the X-ray imaging device 100, like the removal model 52 according to the first embodiment. The learning device 103 generates an enhancement model 352 by machine learning using a plurality of teacher input X-ray images 310b and a plurality of teacher output enhanced images 313b as teacher data (training set). The teacher input X-ray image 310b is generated so as to simulate the X-ray image 10 obtained by imaging the subject 101 including the target object 102 inside the body. The teacher output enhanced image 313b is an image obtained by extracting only the target object 102 from the teacher input X-ray image 310b. The teacher input X-ray image 310b and the teacher output enhanced image 313b are generated so as to have the same conditions (size, etc.) as the X-ray image 10 used for input in inference using the enhancement model 352 .

The enhancement model 352 is generated, for example, based on U-Net, which is one type of full convolution network (FCN). The enhancement model 352 transforms the pixels of the input X-ray image 10 excluding the portion estimated to be the target object 102 , thereby extracting the background other than the target object 102 from the X-ray image 10 . It is generated by learning to perform an image transformation (image reconstruction) that removes (suppresses) parts that are presumed to be. Also, the generated enhancement model 352 is pre-stored in the storage unit 5 .

Then, as shown in FIG. 13, in the third embodiment, the partial image generation unit 362 (control unit 306) generates an enhanced partial image 314 based on the enhanced image 313 generated by the enhancement model 352 and the region of interest 11. is configured to generate The partial image generator 362 generates an enhanced partial image 314 by cutting out a portion corresponding to the region of interest 11 from the enhanced image 313, as in the first embodiment.

The identification image generation unit 364 (control unit 306) generates the identification image 320 based on the emphasized partial image 314 and the contour image 15, as in the first embodiment. That is, the identification image generation unit 264 generates the contour partial image 16 and the weighted image 18 by the same processing as in the first embodiment, weights the weighted image 18, and generates the contour image 15 (the contour partial image 16 ) and the enhanced partial image 314 to generate a synthesized partial image 317 . Then, the identification image generator 364 generates the identification image 320 based on the combined partial image 317 and the contour image 15 .

Further, the image output unit 65 (control unit 306) is configured to display the X-ray image 10 and the identification image 320 side by side on the display unit 4, as in the first embodiment. Other configurations of the third embodiment are the same as those of the first embodiment.

(Effect of the third embodiment)
The following effects can be obtained in the third embodiment.

In the third embodiment, as described above, the partial image generation unit 362 (control unit 306) generates an enhanced image 313 in which the target object 102 in the X-ray image 10 is enhanced by the enhancement model 352 (second trained model). and is configured to generate an enhanced partial image 314 based on the enhanced image 313 and the region of interest 11 . Here, an image from which the target object 102 has been removed is generated using a learned model that has been learned to remove the target object 102 included in the X-ray image 10, and the image from which the target object 102 has been removed and the X-ray By generating the enhanced image 313 directly from the X-ray image 10 using the enhancement model 352, as compared to enhancing the target object 102 contained in the X-ray image 10 by taking the difference with the image 10. , the target object 102 contained in the X-ray image 10 can be enhanced with higher accuracy. That is, by using the enhancement model 352, the portion (background portion) other than the target object 102 in the X-ray image 10 can be suppressed (removed) with higher accuracy. Therefore, in the third embodiment, the partial image generation unit 362 is configured to generate the enhanced image 313 using the enhancement model 352 and to generate the enhanced partial image 314 based on the enhanced image 313 and the region of interest 11. The enhanced partial image 314 can be generated such that the target object 102 is more easily visually recognizable. As a result, an operator such as a doctor can more easily confirm the target object 102 included in the X-ray image 10 . Other effects of the third embodiment are similar to those of the first and second embodiments.

[Modification]
It should be noted that the embodiments disclosed this time should be considered as examples and not restrictive in all respects. The scope of the present invention is indicated by the scope of the claims rather than the above description of the embodiments, and includes all modifications (modifications) within the scope and meaning equivalent to the scope of the claims.

For example, in the first to third embodiments, the identification image generators 64, 264, 364 ( controllers 6, 206, 306) are identified based on the emphasized partial image 14 (214, 314) and the contour image 15. Although an example configured to generate the image 20 (220, 320) has been shown, the present invention is not limited to this. In the present invention, the display unit 4 may be configured to display an identifiable display of the enhanced partial image 14 (214, 314) based on the X-ray image 10 and the enhanced partial image 14 (214, 314). . For example, by synthesizing the emphasized partial image 14 (214, 314) with the X-ray image 10 with reduced brightness (brightness), the display unit 4 can display the portion including the target object 102 so as to be identifiable. may be configured to be displayed on

In the above-described first to third embodiments, the identification image generators 64, 264, 364 ( controllers 6, 206, 306) generate the pixel values of the emphasized partial image 14 (214, 314) and the emphasized portion in the contour image 15. Synthesizing the contour image 15 and the enhanced partial image 14 (214, 314) in a state where the pixel values of the portions (contour partial images 16, 216) corresponding to the image 14 (214, 314) are respectively weighted. Although an example is shown in which the identification image 20 (220, 320) is generated by, the present invention is not limited to this. For example, the identification image 20 (220, 320) may be generated by synthesizing the outline image 15 and the enhanced partial image 14 (214, 314) without weighting.

In the first to third embodiments described above, the outline image generation unit 63 (the control units 6, 206, and 306) is configured to generate the X-ray image 10 after blurring and the X-ray before blurring. Although an example is shown in which the contour image 15 is generated based on the difference from the image 10, the present invention is not limited to this. For example, the contour image 15 may be generated by performing edge detection processing such as a Sobel filter or a Laplacian filter.

In the first and third embodiments described above, the partial image generators 62 and 362 (controllers 6 and 306) are configured to detect the position of the region of interest 11 in the X-ray image 10 detected by the detection model 51 (first trained model). , and an example configured to generate an enhanced partial image 14 (314) having a position and size equal to the detected An example has been shown in which the position of the region of interest 11 in the X-ray image 10 is used as the center to generate the enhanced partial image 214 having a predetermined size regardless of the size of the region of interest 11. However, the present invention is not limited to this. is not limited to For example, an enhanced partial image that is larger than the size of the detected region of interest 11 by a predetermined ratio may be generated.

In the first to third embodiments, the image output unit 65 ( control units 6, 206, 306) is configured to display the X-ray image 10 and the identification image 20 (220, 320) side by side on the display unit 4. Although an example is shown, the present invention is not limited to this. For example, only the identification image 20 (220, 320) may be displayed without displaying the X-ray image 10. FIG. Further, the display of the identification image 20 (220, 320) and the display of the X-ray image 10 may be switched based on the operation on the operation unit (touch panel).

In the above-described first to third embodiments, the region detection unit 61 ( control units 6, 206, 306) is configured to detect one region from the X-ray image 10 as the region of interest 11. Although shown, the invention is not so limited. For example, it may be configured to detect multiple regions of interest.

Further, in the first to third embodiments, an example is shown in which the weighted image 18 (218) is configured such that the luminance value is large near the center and gradually (circularly) decreases toward the periphery. However, the present invention is not limited to this. For example, the weighted image 18 (218) may be configured to have a rectangular shape instead of a circular shape so that the luminance value is gradually changed.

In addition, in the above-described first to third embodiments, when synthesizing the outline image 15 (outline partial images 16 and 216) and the emphasized partial image 14 (214 and 314), an example of synthesizing by changing the transparency is given. Although shown, the invention is not so limited. For example, contour image 15 (contour partial images 16, 216) and enhanced partial image 14 (214, 314) may be combined by adding or multiplying them.

Also, in the second embodiment, an example of generating the emphasized partial image 214 having a predetermined size regardless of the size of the detected region of interest 11 has been described, but the present invention is not limited to this. For example, the detection model 51 (first trained model) may be configured to output a region of interest having a predetermined size (constant size).

Further, in the above-described first to third embodiments, the target object 102 left behind in the body of the subject 101 includes surgical gauze, suture needle, and forceps, but the present invention is limited to this. do not have. For example, the detected target object 102 may include bolts, fixing clips, and the like.

Further, in the first to third embodiments, the X-ray image generation unit 3 and the control unit 6 (206, 306), which are configured as separate hardware, control processing for generation of the X-ray image 10 and , control processing for generation of the enhanced partial image 14 (214, 314) are respectively performed, but the present invention is not limited to this. For example, one common control unit (hardware) may be configured to generate the X-ray image 10 and the enhanced partial image 14 (214, 314).

Further, in the first to third embodiments, the region detection unit 61, the partial image generation unit 62 (262, 362), the contour image generation unit 63, the identification image generation unit 64 (264, 364), and the image output unit 65 is configured as a functional block (software) in one piece of hardware (control unit 6), but the present invention is not limited to this. For example, the region detection unit 61, the partial image generation unit 62 (262, 362), the contour image generation unit 63, the identification image generation unit 64 (264, 364), and the image output unit 65 are provided with separate hardware (computation circuit).

Further, in the first to third embodiments described above, the learning device 103 separate from the X-ray imaging apparatus 100 (200, 300) uses the detection model 51 (first trained model), the removal model 52 (252), and the Although an example in which the enhancement model 352 (second trained model) is generated has been shown, the present invention is not limited to this. For example, the detection model 51, the removal model 52 (252) and the enhancement model 352 may be generated by the X-ray imaging apparatus 100 (200, 300). Also, the detection model 51, the removal model 52 (252), and the enhancement model 352 may be generated by different learning devices (PCs).

Further, in the first to third embodiments, an example in which FasterR-CNN is used as the detection model 51, which is a trained model (algorithm) for area detection (object detection), is shown, but the present invention is limited to this. can't For example, YOLO (You Only Look Once) may be used as an algorithm for area detection. Also, FastR-CNN, R-CNN, SSD (Single Shot MultiBox Detector), etc. may be used as algorithms for area detection.

In addition, in the above-described first to third embodiments, the removal model 52 (252) and the enhancement model 352 (second trained model) are U-Nets, which are one type of full-layer convolution network (FCN). Although an example generated based on is shown, the present invention is not limited to this. For example, the removal model 52 (252) and the enhancement model 352 may be generated based on a CNN (Convolutional Neural Network) including fully connected layers. Also, the removal model 52 (252) and enhancement model 352 may be generated based on Encoder-Decoder models other than U-Net such as SegNet or PSPNet.

In the first and second embodiments, the partial image generation units 62 and 262 (control units 6 and 206) obtain the difference between the X-ray image 10 and the removed image 12 (212) (subtraction processing is performed). ) to generate images in which the target object 102 contained in the X-ray image 10 is enhanced (the enhanced whole image 13 and the enhanced partial image 214), but the present invention is directed to this. Not limited. For example, an image in which the target object 102 is emphasized may be generated by performing division processing of the removed image 12 (212) from the X-ray image 10 instead of subtraction processing.

Further, in the first to third embodiments, an example of displaying the generated identification image 20 (220, 320) on the display unit 4 provided in the X-ray imaging apparatus 100 (200, 300) is shown. The present invention is not limited to this. For example, the identification image 20 (220, 320) may be displayed on a display device such as an external monitor provided separately from the X-ray imaging apparatus 100 (200, 300).

Further, in the first to third embodiments, the X-ray imaging apparatus 100 (200, 300) is an X-ray imaging apparatus for rounds, but the present invention is not limited to this. For example, it may be a general X-ray imaging apparatus installed in an X-ray imaging room.

Further, in the first to third embodiments, image processing for generating the identification image 20 (220, 320) is performed by the control unit 6 (206, 306) provided in the X-ray imaging apparatus 100 (200, 300). However, the present invention is not limited to this. For example, the identification image may be generated by an image processing device (for example, a personal computer) that performs image processing separate from the X-ray imaging device.

Further, in the third embodiment, the partial image generation unit 362 (control unit 306) is changed from the entire X-ray image 10 by the enhancement model 352 (second learned model) to the enhancement corresponding to the entire X-ray image 10. Although an example configured to generate the image 313 has been shown, the present invention is not limited to this. For example, the partial image generation unit 362 may be configured to use an enhancement model (second trained model) trained to generate an enhanced partial image from a portion corresponding to the region of interest 11 of the X-ray image 10. good. That is, the partial image generation unit 362 may be configured to generate an enhanced partial image from a portion of the X-ray image 10 corresponding to the region of interest 11 as an output result of image processing by the second trained model.

[Aspect]
It will be appreciated by those skilled in the art that the exemplary embodiments described above are specific examples of the following aspects.

(Item 1)
an X-ray irradiation unit that irradiates an object with X-rays;
an X-ray detection unit that detects X-rays emitted from the X-ray irradiation unit;
an X-ray image generation unit that generates an X-ray image based on an X-ray detection signal detected by the X-ray detection unit;
a control unit;
The control unit
A region for detecting the region of interest from the X-ray image by a first trained model generated by machine learning so as to detect the region of interest including the target object inside the body of the subject in the X-ray image. a detection unit;
Based on the region of interest detected by the region detection unit and the output result of image processing by a second trained model generated by machine learning to remove or emphasize the target object in the X-ray image, a partial image generation unit that generates an enhanced partial image in which a portion corresponding to the region of interest in the X-ray image is cut out and the target object is emphasized;
an image output unit that causes a display unit to display an identifiable display of the enhanced partial image generated by the partial image generation unit.

(Item 2)
the control unit further includes an identification image generation unit that generates an identification image for identifying the target object included in the X-ray image based on the enhanced partial image generated by the partial image generation unit;
The X-ray imaging apparatus according to Item 1, wherein the image output section is configured to cause the display section to display the identification image generated by the identification image generation section.

(Item 3)
The control unit further includes a contour image generation unit that generates a contour image representing the contour of the subject in the X-ray image,
The identification image generation section is configured to generate the identification image based on the contour image generated by the contour image generation section and the enhanced partial image generated by the partial image generation section. The X-ray imaging apparatus according to item 2, wherein

(Item 4)
The identification image generation unit weights the pixel values of the emphasized partial image and the pixel values of the portion of the contour image corresponding to the emphasized partial image, and generates the contour image and the emphasized partial image. 4. The X-ray imaging apparatus according to item 3, configured to generate the identification image by combining .

(Item 5)
The contour image generation unit executes blurring processing on the X-ray image, and based on the difference between the X-ray image after the blurring processing is executed and the X-ray image before the blurring processing is executed 5. An X-ray imaging device according to item 3 or 4, adapted to generate the contour image by means of a

(Item 6)
6. The X-ray imaging apparatus according to any one of items 2 to 5, wherein the image output unit is configured to display the X-ray image and the identification image on the display unit.

(Item 7)
The partial image generation unit generates the enhanced partial image having a position and size equal to the position and size in the X-ray image of the region of interest detected by the first trained model, and the detected region of interest. Any one of items 1 to 6, configured to generate one of the enhanced partial image having a predetermined size centered on the position in the X-ray image regardless of the size of the region of interest. 1. The X-ray imaging apparatus according to claim 1.

(Item 8)
The partial image generation unit is configured to generate a removed image in which the target object is removed from the entire X-ray image by the second trained model, and the X-ray image and the removed image are Based on the difference between the above and the X-ray imaging apparatus according to any one of items 1 to 7, configured to generate an enhanced partial image.

(Item 9)
The partial image generation unit generates a removed image in which a portion corresponding to the region of interest is cut out from a portion corresponding to the region of interest of the X-ray image and the target object is removed by the second trained model. and generating the enhanced partial image based on a difference between a portion of the X-ray image corresponding to the region of interest and the removed image obtained by clipping the portion corresponding to the region of interest. The X-ray imaging apparatus according to any one of items 1 to 7, which is configured to

(Item 10)
The partial image generation unit is configured to generate an enhanced image in which the target object in the X-ray image is enhanced by the second trained model, and based on the enhanced image and the region of interest 8. The X-ray imaging apparatus according to any one of items 1 to 7, configured to generate the enhanced partial image.

(Item 11)
The region detection unit detects, from the X-ray image, one region having the highest likelihood among a plurality of regions estimated to include the target object by the first trained model as the region of interest. The X-ray imaging apparatus according to any one of items 1 to 10, which is configured to

(Item 12)
A first learning generated by machine learning so as to detect a region of interest including a target object inside the body of the subject in an X-ray image generated based on a detection signal of X-rays irradiated to the subject. a region detection unit that detects the region of interest from the X-ray image using the finished model;
Based on the region of interest detected by the region detection unit and the output result of image processing by a second trained model generated by machine learning to remove or emphasize the target object in the X-ray image, a partial image generation unit that generates an enhanced partial image in which a portion corresponding to the region of interest in the X-ray image is cut out and the target object is emphasized;
an image output unit that causes a display unit to display an identifiable display of the enhanced partial image generated by the partial image generation unit.

(Item 13)
A first learning generated by machine learning so as to detect a region of interest including a target object inside the body of the subject in an X-ray image generated based on a detection signal of X-rays irradiated to the subject. detecting the region of interest from the X-ray image with a pre-defined model;
Based on the detected region of interest and the output result of image processing by a second trained model generated by machine learning to remove or enhance the target object in the X-ray image, in the X-ray image generating an enhanced partial image in which a portion corresponding to the region of interest is clipped and the target object is enhanced;
and causing a display unit to display a display that allows identification of the generated emphasized partial image.

1 X-ray irradiation unit 2 X-ray detection unit 3 X-ray image generation unit 4 Display unit 6, 206, 306 Control unit 10 X-ray image 11 Region of interest 12, 212 Removal image 13 Enhanced whole image 14, 214, 314 Enhanced partial image 15 contour image 20, 220, 320 identification image 51 detection model (first trained model)
52, 252 elimination model (second trained model)
61 region detection unit 62, 262, 362 partial image generation unit 63 contour image generation unit 64, 264, 364 identification image generation unit 65 image output unit 100, 200, 300 X-ray imaging apparatus 101 subject 102 target object 313 enhanced image 352 Emphasis model (2nd trained model)

Claims (13)

1. an X-ray irradiation unit that irradiates an object with X-rays;
   an X-ray detection unit that detects X-rays emitted from the X-ray irradiation unit;
   an X-ray image generation unit that generates an X-ray image based on an X-ray detection signal detected by the X-ray detection unit;
   a control unit;
   The control unit
   A region for detecting the region of interest from the X-ray image by a first trained model generated by machine learning so as to detect the region of interest including the target object inside the body of the subject in the X-ray image. a detection unit;
   Based on the region of interest detected by the region detection unit and the output result of image processing by a second trained model generated by machine learning to remove or emphasize the target object in the X-ray image, a partial image generation unit that generates an enhanced partial image in which a portion corresponding to the region of interest in the X-ray image is cut out and the target object is emphasized;
   an image output unit that causes a display unit to display an identifiable display of the enhanced partial image generated by the partial image generation unit.
2. the control unit further includes an identification image generation unit that generates an identification image for identifying the target object included in the X-ray image based on the enhanced partial image generated by the partial image generation unit;
   The X-ray imaging apparatus according to claim 1, wherein the image output section is configured to display the identification image generated by the identification image generation section on the display section.
3. The control unit further includes a contour image generation unit that generates a contour image representing the contour of the subject in the X-ray image,
   The identification image generation section is configured to generate the identification image based on the contour image generated by the contour image generation section and the enhanced partial image generated by the partial image generation section. 3. The radiographic apparatus of claim 2, wherein
4. The identification image generation unit weights the pixel values of the emphasized partial image and the pixel values of the portion of the contour image corresponding to the emphasized partial image, and generates the contour image and the emphasized partial image. 4. The X-ray imaging apparatus according to claim 3, configured to generate said identification image by synthesizing .
5. The contour image generation unit executes blurring processing on the X-ray image, and based on the difference between the X-ray image after the blurring processing is executed and the X-ray image before the blurring processing is executed 4. An X-ray imaging apparatus according to claim 3, configured to generate said contour image by means of a
6. The X-ray imaging apparatus according to claim 2, wherein the image output unit is configured to display the X-ray image and the identification image on the display unit.
7. The partial image generation unit generates the enhanced partial image having a position and size equal to the position and size in the X-ray image of the region of interest detected by the first trained model, and the detected region of interest. 2. The enhanced partial image according to claim 1, configured to generate either one of said enhanced partial image having a predetermined size centered on a position in said X-ray image regardless of the size of said region of interest. X-ray equipment.
8. The partial image generation unit is configured to generate a removed image in which the target object is removed from the entire X-ray image by the second trained model, and the X-ray image and the removed image are Based on the difference between the above and the 2. An X-ray imaging device as claimed in claim 1, arranged to generate an enhanced partial image.
9. The partial image generation unit generates a removed image in which a portion corresponding to the region of interest is cut out from a portion corresponding to the region of interest of the X-ray image and the target object is removed by the second trained model. and generating the enhanced partial image based on a difference between a portion of the X-ray image corresponding to the region of interest and the removed image obtained by clipping the portion corresponding to the region of interest. 2. The radiographic apparatus of claim 1, configured to.
10. The partial image generation unit is configured to generate an enhanced image in which the target object in the X-ray image is enhanced by the second trained model, and based on the enhanced image and the region of interest , the radiographic apparatus of claim 1, adapted to generate the enhanced partial image.
11. The region detection unit detects, from the X-ray image, one region having the highest likelihood among a plurality of regions estimated to include the target object by the first trained model as the region of interest. 2. The radiographic apparatus of claim 1, configured to.
12. A first learning generated by machine learning so as to detect a region of interest including a target object inside the body of the subject in an X-ray image generated based on a detection signal of X-rays irradiated to the subject. a region detection unit that detects the region of interest from the X-ray image using the finished model;
    Based on the region of interest detected by the region detection unit and the output result of image processing by a second trained model generated by machine learning to remove or emphasize the target object in the X-ray image, a partial image generation unit that generates an enhanced partial image in which a portion corresponding to the region of interest in the X-ray image is cut out and the target object is emphasized;
    an image output unit that causes a display unit to display an identifiable display of the enhanced partial image generated by the partial image generation unit.
13. A first learning generated by machine learning so as to detect a region of interest including a target object inside the body of the subject in an X-ray image generated based on a detection signal of X-rays irradiated to the subject. detecting the region of interest from the X-ray image with a pre-defined model;
    Based on the detected region of interest and the output result of image processing by a second trained model generated by machine learning to remove or enhance the target object in the X-ray image, in the X-ray image generating an enhanced partial image in which a portion corresponding to the region of interest is clipped and the target object is enhanced;
    and causing a display unit to display a display that allows identification of the generated emphasized partial image.

Images