X-ray device

Abstract

To provide an X-ray apparatus capable of executing fluoroscopy or radiography on an appropriate condition more reliably and quickly.SOLUTION: An X-ray apparatus includes: a condition storage part 57 for storing an image acquisition condition 63 corresponding to site information 61 of a subject M in association with the site information 61; a learning model storage part 55 for storing a learning model 51 for inferring a site of a human body shown in an image and outputting the site of the human body by machine learning; an image analysis part 47 for inferring the site of the subject M in an image for input and outputting the site of the subject by inputting an image of the subject to the learning model 51 as the image for input; a condition read-out part 49 for reading out the image acquisition condition 63 associated with the site information 61 by selecting the output site of the subject as the site information 61; and a main control part 39 for executing control of at least any one of control of an X-ray tube 5 and control of an image processing part 33 along the image acquisition condition 63 that has been read out.SELECTED DRAWING: Figure 9

Description

[Technical Field]

[0001] The present invention relates to an X-ray apparatus that irradiates an object with X-rays to perform X-ray fluoroscopy or X-ray photography. [Background technology]

[0002] In medical settings, when an X-ray device is used to obtain X-ray images of a subject by performing X-ray fluoroscopy or X-ray photography, it is important to set appropriate X-ray irradiation conditions or image processing conditions depending on the part of the subject to be imaged or the type of examination to be performed on the subject. In recent years, X-ray devices equipped with an anatomical program (APR) have been used as a configuration for setting appropriate X-ray irradiation conditions or image processing conditions.

[0003] An anatomical program is a data structure in which a series of X-ray irradiation conditions, such as tube voltage and tube current, and a series of image processing conditions, such as contrast processing conditions, are pre-associated with the subject's imaging region or the type of examination. Multiple anatomical programs are pre-set and stored according to the subject's imaging region or the type of examination. For example, in the case of general X-ray imaging in which the subject's imaging region is the chest, information on X-ray irradiation conditions and image processing conditions suitable for general X-ray imaging of the chest is stored in association with the general X-ray imaging in which the chest is the imaging region.

[0004] When performing X-ray fluoroscopy on a subject, multiple anatomical programs are displayed on a display unit, such as a liquid crystal panel, and the operator selects an appropriate one from the multiple anatomical programs (see, for example, Patent Documents 1 and 2).

[0005] As an example, when performing an endoscopic retrograde cholangiopancreatography (ERCP) examination on a subject's abdomen, the operator selects the "Abdomen / ERCP" program from multiple anatomical programs listed on the display, such as "Chest / General X-ray," "Abdomen / General X-ray," and "Abdomen / ERCP," which correspond to the examination site and type of examination. By performing this selection, the X-ray exposure conditions and image processing parameters appropriate for performing ERCP on the abdomen, which have been linked to "Abdomen / ERCP" in advance, are read and displayed on the display. The operator confirms the displayed X-ray exposure conditions, etc., and begins the examination on the subject. [Prior art documents] [Patent documents]

[0006] [Patent Document 1] Japanese Patent Application Laid-Open No. 2012-143443 [Patent Document 2] Japanese Patent Application Publication No. 2018-191983 Summary of the Invention [Problem to be solved by the invention]

[0007] However, the conventional example having such a configuration has the following problems.

[0008] In conventional configurations, when performing X-ray imaging or other procedures using an anatomical program, the operator must select an appropriate program from a list of programs. This manual operation prolongs the examination, which is problematic, particularly during emergency surgery. Furthermore, there is a concern that manual operation may result in inappropriate X-ray irradiation conditions and other parameters being set due to operator error. Furthermore, if the area of ​​the subject to be irradiated with X-rays changes during the procedure or examination, the operator must select a new appropriate APR depending on the new area, which increases the burden on the operator and increases the time required for the procedure.

[0009] The present invention has been made in view of the above circumstances, and has as its object to provide an X-ray apparatus that can perform X-ray fluoroscopy or X-ray photography more reliably, quickly and under appropriate conditions. [Means for solving the problem]

[0010] In order to achieve the above object, the present invention has the following configuration. That is, the X-ray apparatus according to the first aspect includes an X-ray tube that irradiates an object with X-rays, an X-ray detector that is disposed opposite the X-ray tube and detects the X-rays that have passed through the object, an image processing unit that generates an X-ray image by performing image processing using a detection signal output by the X-ray detector, a condition storage unit that stores image acquisition conditions including at least one of X-ray irradiation conditions and image processing conditions corresponding to each target part of the object, in association with the target part, a learning model storage unit that stores a learning model that infers and outputs the part of the human body shown in the image by performing machine learning using an image of the human body as a teacher image, and a learning model that stores the most recently obtained The system comprises a target part inference unit that inputs at least one of the X-ray image and the optical image of the subject as an input image into the learning model, thereby inferring and outputting the part shown in the input image; a condition reading unit that selects the part output by the target part inference unit as the target part, thereby reading out the image acquisition conditions stored in the condition memory unit in association with the target part; and a control unit that controls at least one of the X-ray tube and the image processing unit in accordance with the image acquisition conditions read out by the condition reading unit.

[0011] An X-ray apparatus according to a second aspect of the present invention includes an X-ray tube that irradiates an object with X-rays, an X-ray detector that is disposed opposite the X-ray tube and detects X-rays that have passed through the object, an image processing unit that generates an X-ray image by performing image processing using a detection signal output by the X-ray detector, a condition storage unit that stores image acquisition conditions including at least one of X-ray irradiation conditions and image processing conditions corresponding to each of examination items on the object, in association with the examination items, a learning model storage unit that stores a learning model that infers and outputs the type of examination in the examination image by performing machine learning using an examination image showing a human body as a teacher image, and inputs at least one of the X-ray image and optical image of the object that have been obtained most recently into the learning model as an input image, thereby Based on information including the presence or absence of a testing device shown in the input image or the presence or absence of a testing drug shown in the input image,The apparatus includes an examination type inference unit that infers and outputs the type of examination in the input image, a condition reading unit that selects the type of examination output by the examination type inference unit as the examination item, and thereby reads out the image acquisition conditions that are linked to the examination item and stored in the condition storage unit, and a control unit that controls at least one of the X-ray tube and the image processing unit in accordance with the image acquisition conditions read out by the condition reading unit. [Effects of the Invention]

[0012] According to a first aspect of the present invention, the X-ray device automatically sets image acquisition conditions, including at least one of X-ray irradiation conditions and image processing conditions, by using a learning model that infers a target region shown in an image and a mechanism that associates appropriate image acquisition conditions with each target region, such as an anatomical program. The learning model is configured to infer and output the region of the human body shown in the image through machine learning using a human body image as a training image. That is, the target region inference unit inputs an image of the subject as an input image to the learning model, and infers and outputs the region of the subject shown in the input image. The condition reading unit selects the output region information of the subject as the target region and automatically reads the image acquisition conditions associated with the target region and stored. Therefore, when an X-ray image of the subject is acquired, the target region inference unit and the condition reading unit automatically read image acquisition conditions appropriate for the irradiation field of the X-ray image. Therefore, even if the target region to be irradiated with X-rays on the subject is changed, the image acquisition conditions appropriate for the changed target region are automatically read. In other words, the process of the operator manually selecting the target region and setting the image acquisition conditions is no longer necessary, so X-ray fluoroscopy or X-ray photography can be performed more reliably and quickly under appropriate conditions.

[0013] According to a second aspect of the present invention, the X-ray device automatically sets image acquisition conditions, including at least one of X-ray irradiation conditions and image processing conditions, by using a learning model that infers the type of examination in an examination image of a human body as an examination item and a mechanism that associates appropriate image acquisition conditions with each examination item. The learning model is configured to infer and output the examination type of the image through machine learning using an examination image of a human body as a training image. That is, the examination type inference unit inputs an image of the subject as an input image to the learning model, and infers and outputs the type of examination for the input image. The condition reading unit selects the output examination type information as an examination item and automatically reads the image acquisition conditions associated with the examination item and stored. Therefore, when an X-ray image of the subject is acquired, the examination type inference unit and the condition reading unit automatically read image acquisition conditions appropriate for the examination item of the X-ray image. Therefore, even if the examination item for irradiating the subject with X-rays is changed, image acquisition conditions appropriate for the changed examination item are automatically read. In other words, the process of the operator manually selecting the examination items and setting the image acquisition conditions is no longer necessary, so X-ray fluoroscopy or X-ray photography can be performed more reliably and quickly under appropriate conditions. [Brief explanation of the drawings]

[0014] [Figure 1] 1 is a front view illustrating the overall configuration of an X-ray device according to a first embodiment. [Figure 2] 1 is a right side view illustrating the overall configuration of an X-ray device according to a first embodiment. [Figure 3] FIG. 1 is a perspective view illustrating the overall configuration of a foot switch according to a first embodiment. [Figure 4] FIG. 4 is a diagram illustrating an example of information displayed by a display unit according to the first embodiment. [Figure 5] 1 is a functional block diagram of an X-ray device according to a first embodiment. [Figure 6]1 is a diagram illustrating an APR in Example 1. (a) is a diagram showing the relationship between body part information and image acquisition conditions linked to the body part information, and (b) is a diagram showing body part information relating to each body part and examples of specific parameters of the image acquisition conditions linked to the body part information. [Figure 7] FIG. 10 is a schematic diagram showing a series of steps for reading out image acquisition conditions using a learning model and APR in the first embodiment. [Figure 8] 1 is a schematic diagram showing a series of steps for reading out image acquisition conditions before starting X-ray irradiation in Example 1. FIG. [Figure 9] FIG. 10 is a schematic diagram showing a series of steps for reading out image acquisition conditions when the hip joint is the irradiation field in Example 1. [Figure 10] 10 is a schematic diagram showing a series of steps for reading out image acquisition conditions when the abdomen is the irradiation field in Example 1. FIG. [Figure 11] 1 is a schematic diagram showing a series of steps for reading out image acquisition conditions when the chest is the irradiation field in Example 1. FIG. [Figure 12] FIG. 10 is a diagram showing a display screen of an APR according to a conventional example. [Figure 13] FIG. 10 is a diagram showing a state in which image acquisition conditions are read out using an APR according to a conventional example. [Figure 14] 1 is a diagram for explaining an APR in Example 1. (a) is a diagram showing an APR in which the examination item is general chest radiography, (b) is a diagram showing an APR in which the examination item is PCI of the chest, (c) is a diagram showing an APR in which the examination item is general chest radiography, (d) is a diagram showing an APR in which the examination item is ERCP of the abdomen, and (e) is a diagram showing an APR in which the examination item is UGI of the abdomen. [Figure 15] 10A and 10B are diagrams illustrating an APR in Example 2. FIG. 10A is a diagram showing the relationship between examination item information and image acquisition conditions linked to the examination item information, and FIG. 10B is a diagram showing examination item information related to each part and examples of specific parameters of the image acquisition conditions linked to the examination item information. [Figure 16]FIG. 10 is a schematic diagram showing a series of steps for reading out image acquisition conditions using a learning model and APR in the second embodiment. [Figure 17] FIG. 10 is a functional block diagram showing the main parts of an X-ray device according to a third embodiment. [Figure 18] 10A and 10B are diagrams illustrating erroneous analysis that occurs in a learning model in Example 3. (a) is a diagram illustrating a case where the learning model performs an appropriate image analysis, and (b) is a diagram illustrating a case where the learning model performs an erroneous image analysis. [Figure 19] 10 is a flowchart illustrating the main steps of the operation of the X-ray device according to the third embodiment. [Figure 20] This is a schematic diagram showing a series of steps for reading out image acquisition conditions using the learning model and APR in the case where the learning model properly performs image analysis in Example 3. [Figure 21] FIG. 11 is a schematic diagram showing a series of steps for reading out image acquisition conditions using the learning model and APR in the case where the learning model performs an incorrect image analysis in Example 3. [Figure 22] FIG. 10 is a functional block diagram showing the main parts of an X-ray device according to a fourth embodiment. [Figure 23] FIG. 13 is a diagram showing an example of a screen on which an approval key is displayed in the fourth embodiment. [Figure 24] 1 is a flowchart illustrating the operation of the X-ray device according to the embodiment, where (a) is a flowchart according to the first embodiment, and (b) is a flowchart according to the fourth embodiment. [Figure 25] FIG. 11 is a diagram showing the relationship between a plurality of timings and the position of the X-ray irradiation field at each timing in the fourth embodiment. [Figure 26] FIG. 11 is a block diagram showing the control relationship of the main parts of the X-ray device at timing M1 or timing M2 in the fourth embodiment. [Figure 27] FIG. 11 is a block diagram showing the control relationship of the main parts of the X-ray device at timing M3 in the fourth embodiment. [Figure 28] FIG. 11 is a block diagram showing the control relationship of the main parts of the X-ray device at timing M4 in the fourth embodiment. DETAILED DESCRIPTION OF THE INVENTION [Example]

[0015] A first embodiment of the present invention will be described below with reference to the drawings.

[0016] <Explanation of overall configuration> 1 and 2, in the X-ray device 1 according to the first embodiment, an X-ray tube 5 and an X-ray detector 7 are arranged opposite each other across a top board 3 on which a subject M in a supine position is placed. The top board 3 is disposed on top of a top board support 4 that is configured to be movable up and down. The X-ray tube 5 irradiates X-rays onto the subject M. The X-ray detector 7 detects the X-rays that have been irradiated from the X-ray tube 5 onto and transmitted through the subject M, converts them into an electrical signal, and outputs it as an X-ray detection signal. An example of the X-ray detector 7 is an FPD (Flat Panel Detector).

[0017] The X-ray tube 5 and the X-ray detector 7 are respectively provided at one end and the other end of the C-arm 9. The C-arm 9 is held by an arm holding member 11 and is configured to rotate along an arc path of the C-arm 9 indicated by the symbol RA. That is, the C-arm 9 rotates around an axis in the y direction (the longitudinal direction of the tabletop 3) along the arc path RA.

[0018] Arm holding member 11 is disposed on the side of support column 13 and is configured to be rotatable around a horizontal axis P (arc-shaped path RB) that is parallel to the x-direction (the short-side direction of tabletop 3). C-arm 9 held by arm holding member 11 rotates around the axis in the x-direction following arm holding member 11. Because C-arm 9 is configured to be rotatable around two orthogonal axes along arc-shaped path RA and arc-shaped path RB, X-rays can be irradiated onto subject M from any direction.

[0019] The support column 13 is supported by a support base 15 disposed on the floor surface and is configured to be able to move horizontally in both the x and y directions along the upper surface of the support base 15. The arm support member 11 and C-arm 9 supported by the support column 13 move horizontally in the x or y direction in accordance with the horizontal movement of the support column 13. The collimator 17 is provided in the X-ray tube 5 and limits the X-rays emitted from the X-ray tube 5 to a predetermined shape. An example of a shape that limits the X-rays is a pyramidal cone.

[0020] An optical camera 19 is disposed on the collimator 17. The optical camera 19 is, for example, a digital camera, and captures an optical image of the subject M by photographing the subject M with visible light. In the X-ray device 1, the positions and orientations of the optical camera 19 and the X-ray tube 5 are adjusted so that the irradiation field of the optical camera 19 and the irradiation field of the X-ray tube 5 are the same range.

[0021] 2, a foot switch 21 is placed on the floor below the tabletop 3. The foot switch 21 is connected to a power source and a CPU constituting a main control unit 39 (described later) via a cable 23. In the first embodiment, the main control unit 39 is built into the tabletop support 4, and in FIG. 1, the cable 23 is connected to the tabletop support 4.

[0022] 3, the foot switch 21 includes a body 25, a switch 27 operated by the surgeon with his / her foot, and a bottom plate 29. The switch 27 is composed of three pedal switches: a fluoroscopy switch 27a, an imaging switch 27b, and a tabletop movement switch 27c. In the first embodiment, the number of switches 27 is three, but the number of switches 27 may be changed as appropriate.

[0023] The fluoroscopy switch 27a controls the on / off of X-ray fluoroscopy. That is, by stepping on the fluoroscopy switch 27a with a foot, X-ray fluoroscopy is started, in which a relatively low dose of X-rays is intermittently emitted from the X-ray tube 5 in accordance with the X-ray fluoroscopy conditions described below. The radiography switch 27b controls the on / off of X-ray imaging. That is, by stepping on the radiography switch 27b with a foot, X-ray imaging is started, in which a relatively high dose of X-rays is emitted from the X-ray tube 5 in accordance with the X-ray imaging conditions described below.

[0024] The top plate movement switch 27c controls the on / off of vertical movement of the top plate 3. That is, by stepping on the top plate movement switch 27 with a foot, the top plate support 4 expands and contracts in the z direction, changing the height of the top plate 3. By operating the top plate movement switch 27c, the top plate 3 moves up and down between a relatively low position (boarding and disembarking position) for the subject M to get on and off the top plate 3, and a relatively high position (treatment position) for the surgeon to perform medical procedures on the subject M in a supine position.

[0025] As shown in FIG. 4, the X-ray device 1 further includes an image processing unit 33, a display unit 35, an arm position detection unit 37, a main control unit 39, an operation console 41, and a storage unit 43.

[0026] The image processing unit 33 is provided after the X-ray detector 7 and generates an X-ray image based on the X-ray detection signal output from the X-ray detector 7. The display unit 35 displays the X-ray image generated by the image processing unit 33 and various information related to the X-ray device 1. Examples of the display unit 35 include a liquid crystal monitor or a high-precision display. Examples of the configuration in which the display unit 35 is disposed include a configuration in which it is suspended from the ceiling, a configuration in which it is mounted on a mobile cart, or a configuration in which it is disposed on the operation console 41.

[0027] Arm position detection unit 37 detects the amount of rotational movement of C-arm 9 on each of arcuate paths RA and RB, and also detects the amount of translation of C-arm 9 in the x and y directions, based on a position detector such as a potentiometer or encoder (not shown). By detecting the amount of rotational movement and translation of C-arm 9, arm position detection unit 37 detects the position of C-arm 9. By detecting the position of C-arm 9, arm position detection unit 37 can identify the position of the irradiation field of X-ray tube 5 relative to subject M.

[0028] The main control unit 39 includes an information processing unit such as a central processing unit (CPU) as an example. The main control unit 39 controls the various components of the X-ray device 1, such as the X-ray tube control unit 31, the image processing unit 33, and the display unit 35. The main control unit 39 corresponds to the control unit in the present invention.

[0029] The main control unit 39 includes a machine learning unit 45, an image analysis unit 47, and a condition reading unit 49. The machine learning unit 45 creates a learning model 51 by performing machine learning using X-ray images or optical images acquired in advance. The image analysis unit 47 uses the learning model 51 to analyze the X-ray images or optical images generated by the X-ray device 1 and determine the part of the human body shown in the image. The image analysis unit 47 according to the first embodiment corresponds to the target part inference unit of the present invention.

[0030] The condition reading unit 49 refers to the anatomical program 53 (APR53) to read, from the condition storage unit 43, the image acquisition conditions associated with the region determined by the image analysis unit 47. The condition reading unit 49 then transmits the read image acquisition conditions to the X-ray tube control unit 31 or the image processing unit 33, and causes the X-ray tube 5 to irradiate X-rays and the image processing unit 33 to generate an X-ray image in accordance with the transmitted conditions.

[0031] The operation console 41 is used to input instructions from an operator regarding the operation of the X-ray device 1, and the main control unit 39 performs overall control in accordance with the instructions input by the operator to the operation console 41. Examples of operation devices provided on the operation console 41 include a keyboard input panel, a touch input panel, a mouse, a dial, a changeover switch, and a push button switch. In the first embodiment, the operation console 41 may be attached to the side of the tabletop 3 as shown in Fig. 1, or may be provided on the top of the support column 13, or may be mounted on a mobile cart.

[0032] The storage unit 43 stores various X-ray images generated by the image processing unit 33, various information related to the operation of the X-ray device 1, and the like. An example of the storage unit 43 is a non-volatile memory. The storage unit 43 includes a learning model storage unit 55 and a condition storage unit 57. The learning model storage unit 55 stores the learning model 51 created by the machine learning unit 45. The condition storage unit 57 stores the APR 53.

[0033] Here, the APR 53 according to the first embodiment will be described. As shown in Fig. 6(a), the APR 53 is a program in which image acquisition conditions 63 are linked to each piece of body part information 61. The body part information 61 is information about a body part to be irradiated with X-rays, and examples thereof include the head and neck, chest, abdomen, hip joints, shoulders, knees, and feet. The image acquisition conditions 63 are a series of conditions related to the acquisition of X-ray images, and include X-ray irradiation conditions 65 and image processing conditions 67.

[0034] The X-ray irradiation conditions 65 are various parameters related to X-ray irradiation, and include X-ray fluoroscopy conditions 68 and X-ray imaging conditions 69. Examples of parameters related to X-ray irradiation include the tube voltage and tube current applied to the X-ray tube 5, the X-ray irradiation time, and the X-ray irradiation cycle. The image processing conditions 67 are parameters related to image processing performed on the electrical signal detected by the X-ray detector 7, and examples include a contrast processing value, a sharpening processing value, and an edge processing value. The image acquisition conditions 63 may further include setting conditions for the X-ray detector 7, such as a frame rate and a gain value.

[0035] The X-ray fluoroscopy conditions 68 are various parameters related to X-ray irradiation when performing X-ray fluoroscopy. In X-ray fluoroscopy, relatively weak X-rays are intermittently irradiated to obtain an X-ray fluoroscopic image (moving image). The X-ray imaging conditions 68 are various parameters related to X-ray irradiation when performing X-ray imaging. In X-ray imaging, relatively strong X-rays are irradiated for a short period of time to obtain an X-ray imaging image (still image). In the present invention, X-ray images include X-ray fluoroscopy images and X-ray imaging images.

[0036] 6(b) shows specific details of image acquisition conditions 63 linked to region information 61 in APR 53 according to this embodiment. As an example, among region information 61, region information 61a of the hip joint is linked in advance to image acquisition conditions 63a including X-ray irradiation conditions 65a and image processing conditions 67a. Image processing conditions 67a include parameters such as a contrast processing value of 10 and a sharpening processing value of 7. Among X-ray irradiation conditions 65a, X-ray fluoroscopy conditions 68a include parameters such as a tube voltage of 30 kV and a tube current of 2.0 mA. Among X-ray irradiation conditions 65a, X-ray imaging conditions 69a include parameters such as a tube voltage of 50 kV and a tube current of 3.0 mA.

[0037] Similarly, abdominal region information 61b in region information 61 is linked in advance to image acquisition conditions 63b including X-ray irradiation conditions 65b and image processing conditions 67b. X-ray irradiation conditions 65b include X-ray fluoroscopy conditions 68b and X-ray imaging conditions 69b. Chest region information 61c in region information 61 is linked in advance to image acquisition conditions 63c including X-ray irradiation conditions 65c and image processing conditions 67c. X-ray irradiation conditions 65c include X-ray fluoroscopy conditions 68c and X-ray imaging conditions 69c. Head and neck region information 61d in region information 61 is linked in advance to image acquisition conditions 63d including X-ray irradiation conditions 65d and image processing conditions 67d. X-ray irradiation conditions 65d include X-ray fluoroscopy conditions 68d and X-ray imaging conditions 69d. In this way, the APR 53 in which the appropriate image acquisition condition 63 is linked to each of the plurality of pieces of body part information 61 is set in advance and stored in the condition storage unit 57.

[0038] Here, a configuration for automatically setting image acquisition conditions 63 in the first embodiment will be described with reference to Fig. 7 and Fig. 24(a). Fig. 24(a) is a flowchart relating to the operation of the X-ray device 1 according to the first embodiment. First, a learning model 51 is created by performing machine learning in advance in the machine learning unit 45 (step M1). The machine learning unit 45 acquires images of various parts of the human body in advance as original images R1. The original images R1 are, for example, a group of many images including X-ray images, DRR images obtained by projecting three-dimensional CT images in various directions, and optical images acquired by an optical camera.

[0039] The machine learning unit 45 then performs image processing on the original image R1 to accommodate variations in X-ray conditions, such as increases / decreases in contrast, increases / decreases in brightness, increases / decreases in noise, etc., and obtains a large number of primary augmented images R2. The machine learning unit 45 then performs image processing on the primary augmented images R2 to accommodate variations in the position of the subject M or the position of the C-arm 9, such as rotation, enlargement, reduction, etc., and obtains a large number of secondary augmented images R3.

[0040] Finally, the machine learning unit 45 performs machine learning using the original image R1, the primary augmented image R2, and the secondary augmented image R3 as training images to create a learning model 51 that infers the parts of the human body that appear in the images. That is, by receiving an X-ray image F, an optical image D, or the like as input information, the learning model 51 infers which parts of the human body appear in the input image, and outputs the information on the parts of the human body obtained by inference. The trained learning model 51 is stored in the learning model storage unit 55.

[0041] Although the first embodiment illustrates a configuration in which the learning model 51 is created in the X-ray device 1, the learning model 51 may be created in advance in another device, and the program for the created learning model 51 may be stored in the learning model storage unit 55. In this case, the machine learning unit 45 can be omitted from the X-ray device 1.

[0042] Second, the image analysis unit 47 analyzes the image using the learning model 51, and identifies the part of the subject M that appears in the image obtained for the subject M. That is, the image analysis unit 47 reads out the learning model 51 stored in the learning model storage unit 55. Then, after the image processing unit 33 generates an X-ray image of the subject M by irradiating it with X-rays (step M2), the X-ray image or the like transmitted from the image processing unit 33 is input to the learning model 51 as an input image. The learning model 51 analyzes the input image using clues such as the components of the human body that appear in the input X-ray image, and infers the part of the subject M that appears in the input image. Information on the part obtained by inference is output from the learning model 51 as part information 61 (step M3).

[0043] Third, the condition readout unit 49 sets appropriate image acquisition conditions 63 using the region information 61 obtained by the image analysis unit 47 and the APR 53. The condition readout unit 49 inputs the region information 61 obtained by the image analysis unit 47 to the APR 53, and reads and outputs the image acquisition conditions 63 that are associated with the region information 61 and stored in the APR 53 (step M4). The condition readout unit 49 sets the output image acquisition conditions 63 as conditions to be used for acquiring an X-ray image that will be performed immediately afterwards (step M5).

[0044] The image acquisition conditions 63 set as conditions for acquiring an X-ray image are transmitted to the X-ray tube 5 and the image processing unit 33, and a new X-ray image of the subject M is generated according to various parameters of the set image acquisition conditions 63 (step M6). In this way, the X-ray device 1 automatically sets appropriate image acquisition conditions 63 by using the learning model 51 and the APR 53. After a predetermined time has elapsed (step M7), the process returns to step S2, an input image is generated, and the region information 61 is inferred using the learning model 51. An example of the predetermined time is the time it takes to generate 10 frames of X-ray images. In this case, inference using the learning model 51 is performed every time 10 frames of X-ray images are generated.

[0045] <Explanation of operation> Here, a specific description will be given of an operation of examining a subject M using the X-ray device 1 with the learning model 51 and the APR 53 stored in the storage unit 43. In Example 1, a case of performing a percutaneous coronary intervention (PCI) will be described as an example of an examination. In the coronary intervention according to Example 1, a catheter Ch is inserted from the groin, and while the catheter Ch is sequentially confirmed by X-ray fluoroscopy, the catheter Ch is guided to the coronary artery and a stent is placed. That is, as shown in FIGS. 8 to 11, the target region of the subject M to be irradiated with X-rays shifts in the order of the hip joint La, abdomen Lb, and chest Lc.

[0046] When starting an examination of subject M by coronary intervention, the operator first sets image acquisition conditions 63 for acquiring the first X-ray image. As shown in FIG. 8, the operator places subject M on the tabletop 3 and moves the C-arm 9 so that the position of the irradiation field is the hip joint La. Then, the optical camera 19 is used to capture an image of the hip joint La of subject M. Through this capture, the optical camera 19 generates an optical image D1 of the hip joint La. Data of the generated optical image D1 is sent to the image analysis unit 47 (see symbol T1).

[0047] The image analysis unit 47 inputs the optical image D1 as an input image to the learning model 51. The learning model 51 analyzes the optical image D1 and estimates and outputs the region shown in the optical image D1, which is the input information. In this case, using the outline of the subject M and the presence of the feet shown in the optical image D1 as clues, the learning model 51 estimates that the region shown in the optical image D1 is most likely to be the hip joint. As a result, the learning model 51 outputs region information 61a related to the hip joint. The region information 61a output from the learning model 51 is sent to the condition reading unit 49 (see symbol T2).

[0048] The condition reading unit 49 searches for appropriate image acquisition conditions 63 for the hip joint La associated with the region information 61a. That is, by inputting the region information 61a to the APR 53, the image acquisition conditions 63 associated with the region information 61a and stored in the APR 53, i.e., the image acquisition conditions 63a, are read and output. As a result, the X-ray irradiation conditions 63a included in the image acquisition conditions 63a are set as control parameters for the X-ray tube 5, and the image processing conditions 67a are set as control parameters for the image processing unit 33.

[0049] In this way, by performing the operation of acquiring the optical image D1 with the optical camera 19, the image acquisition conditions 63a are automatically set using the learning model 51 and the APR 53. The image acquisition conditions 63a set by inputting the optical image D1 correspond to the image acquisition conditions 63 for acquiring the first X-ray image.

[0050] After confirming that the image acquisition conditions 63a have been set, the operator starts X-ray fluoroscopy by stepping on the fluoroscopy switch 27a of the foot switch 21. The operator inserts the catheter Ch into the groin of the subject M while checking the X-ray fluoroscopic image of the hip joint La obtained by X-ray fluoroscopy.

[0051] At this time, the main controller 39 controls the X-ray tube 5 in accordance with X-ray fluoroscopy conditions 68a included in the image acquisition conditions 63a. That is, under conditions such as a tube voltage of 30 kV and a tube current of 2.0 mA, the X-ray tube 5 irradiates the hip joint La with X-rays as shown in FIG. 9. The main controller 39 then controls the image processor 33 in accordance with image acquisition conditions 67a included in the image acquisition conditions 63a. That is, the image processor 33 performs various image processing operations on the detection signal from the X-ray detector 7 under conditions such as a contrast value of 10 and sharpness of 7, and generates an X-ray fluoroscopic image (X-ray image F1) of the hip joint La as the target site. The operator begins manipulating the catheter Ch while checking the position of the catheter Ch shown in the X-ray image F1.

[0052] The X-ray device 1 repeats the operation of automatically setting the image acquisition conditions 63 at a predetermined timing. FIG. 9 shows the process of automatically setting the image acquisition conditions 63 by irradiating the hip joint La with X-rays. By irradiating the hip joint La with X-rays from the X-ray tube 5, the image processing unit 33 generates an X-ray image F1 of the hip joint La as the target site. Data of the generated X-ray image F1 is sent to the image analysis unit 47.

[0053] The image analysis unit 47 inputs the X-ray image F1 as an input image to the learning model 51. The learning model 51 estimates and outputs the region shown in the X-ray image F1, which is input information. In this case, using the pelvis Ba and femur Bc of the subject M shown in the X-ray image F1 as clues, the learning model 51 estimates that the region shown in the X-ray image F1 is most likely to be a hip joint. As a result, the learning model 51 outputs region information 61a, which is information that the region is a hip joint. The region information 61a output from the learning model 51 is sent to the condition reading unit 49.

[0054] The condition reading unit 49 searches for appropriate image acquisition conditions 63 for the hip joint La associated with the region information 61a. By inputting the region information 61a to the APR 53, the image acquisition conditions 63 stored in association with the region information 61a in the APR 53, i.e., the image acquisition conditions 63a preset as appropriate conditions (parameters) for generating an X-ray image of the hip joint La, are read and output. As a result, the X-ray tube 5 irradiates X-rays based on the X-ray irradiation conditions 63a, and the image processing unit 33 performs image processing based on the image processing conditions 67a. In this way, when the position of the C-arm 9 is determined so that the X-ray irradiation field is positioned at the hip joint La, the image acquisition conditions 63a are automatically set as conditions for acquiring X-ray images, and X-ray images F1 are continuously generated. The operator manipulates the catheter Ch to advance it toward the heart while checking the position of the catheter Ch reflected in the X-ray image F1.

[0055] The generated X-ray image F1 is transmitted to the image analysis unit 47 and displayed on the display unit 35. FIG. 4 shows information displayed on the display unit 35. The display unit 35 has an image display area K1, an analysis result display area K2, a selected region display area K3, and an application condition display area K4. The image display area K1 displays the most recently acquired X-ray image F or optical image D. When X-rays are being irradiated to the hip joint La, the most recently acquired image is the X-ray image F1, and therefore the X-ray image F1 is displayed in the image display area K1.

[0056] The analysis result display area K2 displays information output by the learning model 51 as the analysis result. The learning model 51 outputs information about the body part shown in the input image (here, X-ray image F1) that is the input information, along with the accuracy. As an example, the learning model 51 outputs information that the accuracy that the body part shown in the X-ray image F1 is the hip joint is 91.4%, the accuracy that the body part is the head and neck is 1.4%, the accuracy that the body part is the chest is 4.2%, and the accuracy that the body part is the abdomen is 0.1%. The image analysis unit 47 selects the body part with the highest accuracy as body part information 61 and transmits the body part information 61 to the condition reading unit 49. The image analysis unit 47 also displays a predetermined number of body parts in the analysis result display area K2 in descending order of accuracy. In the first embodiment, the analysis result display area K2 displays three body parts in descending order of accuracy. That is, the information about the "hip joint", "chest", and "head and neck" is displayed in a list along with the numerical accuracy values. The number of regions displayed in the analysis result display area K2 is not limited to three and may be changed as appropriate.

[0057] The selected region display area K3 displays the region selected as the region information 61. Here, the region information 61a of the hip joint La is selected, so "hip joint" is displayed in the selected condition display area K3. The applied condition display area K4 displays the image acquisition conditions 63 currently applied. Here, the image acquisition conditions 63a linked to the region information 61a of the hip joint La are applied, so the image acquisition conditions 63a are displayed in the applied condition display area K4. For convenience of explanation, in FIG. 4, information on the tube voltage and tube current among the image acquisition conditions 63a is displayed in the applied condition display area K4. An adjustment key NB is displayed in the applied condition display area K4. By operating a mouse or the like to appropriately input data into the adjustment key NB, the operator can, for example, perform an adjustment to increase or decrease the tube voltage value from the initial value set in the image acquisition conditions 63a.

[0058] The operator directs his / her gaze at the display unit 35 and checks the contents of the image acquisition conditions 63 displayed in the application condition display area K4, the contents of the region displayed in the selection condition display area K3, and the accuracy information displayed in the analysis result display area K2, thereby determining whether the image acquisition conditions 63 automatically set by the learning model 51 and APR 53 are appropriate. The operator also advances the catheter Ch toward the coronary artery while checking the X-ray image F1 and other information displayed on the display unit 35.

[0059] The operator advances the catheter Ch and moves the position of the C-arm 9 in the y direction to match the position of the catheter Ch, thereby changing the position of the X-ray irradiation field. Therefore, as the catheter Ch advances from the area of ​​the hip joint La to the area of ​​the abdomen Lb, the target area to be irradiated with X-rays shifts from the hip joint La to the abdomen Lb. In the X-ray device 1, new image acquisition conditions 63 are set as the target area to be irradiated with X-rays shifts. Figure 10 shows the process of automatically setting the image acquisition conditions 63 by irradiating the abdomen Lb with X-rays.

[0060] The image processing unit 33 generates an X-ray image F2 of the target region of the abdomen Lb by irradiating the abdomen Lb with X-rays from the X-ray tube 5. Data of the generated X-ray image F2 is sent to the image analysis unit 47.

[0061] The image analysis unit 47 inputs the X-ray image F2 as an input image into the learning model 51 and analyzes the X-ray image F2. That is, in the image analysis unit 47, the learning model 51 estimates and outputs the region shown in the X-ray image F2, which is the input information. In this case, using the stomach Ga and lumbar vertebrae Bh of the subject M shown in the X-ray image F2 as clues, the learning model 51 estimates that the region shown in the X-ray image F2 is most likely the abdomen. As a result, the learning model 51 outputs region information 61b of the abdomen. The region information 61b output from the learning model 51 is sent to the condition reading unit 49.

[0062] The condition reading unit 49 searches for appropriate image acquisition conditions 63 for the abdomen Lb associated with the region information 61b. That is, by inputting the region information 61b to the APR 53, the image acquisition conditions 63 associated with the region information 61b and stored in the APR 53, i.e., the image acquisition conditions 63b, are read and output. As a result, the X-ray tube 5 irradiates X-rays based on the X-ray irradiation conditions 63b, and the image processing unit 33 performs image processing based on the image processing conditions 67b.

[0063] Because the operator operates the fluoroscopy switch 27a with his / her foot, the X-ray tube 5 irradiates X-rays based on the X-ray fluoroscopy conditions 68b, and X-ray fluoroscopy is performed. When an X-ray image is required, the operator changes the operated switch to the imaging switch 27b, and the X-ray tube 5 irradiates X-rays to perform X-ray imaging in accordance with the X-ray imaging conditions 69b. By irradiating X-rays, an X-ray image of the abdomen Lb generated under the appropriate conditions can be obtained.

[0064] In this way, when the position of the C-arm 9 is determined so that the X-ray irradiation field is located on the abdomen Lb, the image acquisition conditions 63b are automatically set and the X-ray image F2 continues to be generated. When the X-ray image F generated by the image processing unit 33 changes from the X-ray image F1 that shows the hip joint La to the X-ray image F2 that shows the abdomen Lb, the image analysis unit 47 and the condition reading unit 49 quickly change the image acquisition conditions 63. That is, the image acquisition conditions 63a that are suitable for the hip joint Ka are automatically and quickly changed to the image acquisition conditions 63b that are suitable for the abdomen Lb. Therefore, the operator can check the X-ray image F2 generated under the appropriate conditions on the display unit 35 without interrupting the operation of the catheter Ch.

[0065] The operator continues to operate the catheter Ch, advancing it closer to the heart. The position of the C-arm 9 is then moved in the y direction in accordance with the advancement of the catheter Ch. Therefore, as the catheter Ch advances from the abdominal Lb region to the chest Lc region where the heart is located, the target region to be irradiated with X-rays shifts from the abdominal Lb to the chest Lc. As the target region shifts to the chest Lc, new image acquisition conditions 63 are set. FIG. 11 shows the process of automatically setting the image acquisition conditions 63 by irradiating the chest Lc with X-rays.

[0066] The image processing unit 33 generates an X-ray image F3 of the target region of the chest Lc by irradiating the chest Lc with X-rays from the X-ray tube 5. Data of the generated X-ray image F3 is sent to the image analysis unit 47.

[0067] The image analysis unit 47 inputs the X-ray image F2 as an input image to the learning model 51. The learning model 51 estimates and outputs the region shown in the X-ray image F2, which is input information. In this case, using the heart H, lungs Lu, and thoracic vertebrae (not shown) of the subject M shown in the X-ray image F2 as clues, the learning model 51 estimates that the region shown in the X-ray image F2 is most likely the chest. As a result, the learning model 51 outputs region information 61c related to the chest. The region information 61c output from the learning model 51 is sent to the condition reading unit 49.

[0068] The condition reading unit 49 searches for appropriate image acquisition conditions 63 for the chest Lc related to the region information 61c. That is, by inputting the region information 61c to the APR 53, the image acquisition conditions 63 associated with the region information 61c and stored in the APR 53, i.e., the image acquisition conditions 63c, are read and output. The read image acquisition conditions 63c are then set as conditions for acquiring an X-ray image, and among the image acquisition conditions 63c, X-ray irradiation conditions 65c are sent to the X-ray tube 5 and image processing conditions 67c are sent to the image processing unit 33.

[0069] As a result, the X-ray tube 5 irradiates X-rays based on the X-ray irradiation conditions 65c, and the image processor 33 performs image processing of the X-ray image F based on the image processing conditions 67c. Therefore, the image acquisition conditions 63c are automatically set quickly as the target area shifts to the chest Lc, and the visibility of the X-ray image F3 is quickly improved using the appropriate image acquisition conditions 63c. The operator places a stent in the coronary artery while checking the X-ray image F3, and then completes the PCI procedure.

[0070] In this way, in the X-ray device 1 of Example 1, by combining the learning model 51 and APR 53, the X-ray device 1 has a configuration in which the image acquisition conditions 63 are automatically changed in accordance with changes in the target area to be irradiated with X-rays.

[0071] In conventional devices using APR, an operator manually selects a target region to be irradiated with X-rays, and image acquisition conditions linked to information about the target region are read out. As an example, as shown in Figure 12, the conventional device is equipped with an operation touch panel TP that displays a large number of icons Ac for specifying the target region. The operator selects and presses an icon from the icon group Ac to specify the region to be irradiated with X-rays (for example, icon Ab specifying the chest).

[0072] By pressing the icon Ab specifying the chest, the image acquisition conditions Na associated with the target region information "chest" are read and set, and the information on the image acquisition conditions Na is displayed on the touch panel TP, as shown in Fig. 13. In the conventional example shown in Fig. 13, a tube voltage of 40 kV and a tube current of 2.5 mA are displayed as the image acquisition conditions Na. As such, in order to set image acquisition conditions using APR in conventional devices, the operator must manually specify the target region.

[0073] For example, when PCI is performed with a conventional device, the operator must manually specify the target area by operating the touch panel TP each time the target area to be irradiated with X-rays changes from the hip joint to the abdomen to the chest. Because the catheterization procedure must be interrupted each time such a manual operation is performed, the PCI procedure takes longer and it becomes difficult for the operator to concentrate on the procedure, which is a problem.

[0074] One method for reducing the number of manual operations is to select the chest APR from the beginning and read out image acquisition conditions suitable for the chest. However, even if the target area for X-ray irradiation is the hip joint or abdomen, X-ray images are acquired using image acquisition conditions suitable for the chest, making it difficult to obtain highly visible X-ray images in the early or middle stages of PCI.

[0075] In contrast to such a conventional configuration, the X-ray device 1 according to the first embodiment includes an image analysis unit 47 that analyzes a most recently acquired image using a learning model 51 and estimates the part of the subject M that appears in the image, and a condition reading unit 49 that reads, using an APR 53, image acquisition conditions 63 appropriate for the part estimated by the image analysis unit 47. The learning model 51 is trained in advance to infer the part of the human body that appears in the image by using an image of the human body as training information. Therefore, by using an X-ray image F or an optical image D that shows the subject M as an input image, the learning model 51 infers the part of the subject M that appears in the input image and outputs it as part information 61.

[0076] In the APR 53, image acquisition conditions 63 that can appropriately acquire an X-ray image of the target region are associated with each piece of region information 61 that identifies the target region. Therefore, by sending the image of the subject M most recently acquired using the X-ray tube 5 or the optical camera 19 as input information to the image analysis unit 47 and the condition reading unit 49, the image acquisition conditions 63 appropriate for the region targeted by the image of the subject M are automatically set.

[0077] As described above, the X-ray device 1 according to the first embodiment is configured to automatically change the image acquisition conditions 63 in response to changes in the target region of X-ray irradiation by using the learning model 51 that estimates the region of the human body shown in the image and the APR 53 in which the image acquisition conditions 63 are linked to the region information 61. Therefore, even when performing an examination in which the region to be irradiated with X-rays changes in a short period of time, such as PCI, the image acquisition conditions 63 are automatically changed to appropriate parameters in response to changes in the target region. Therefore, the operator can check highly visible X-ray images acquired under appropriate conditions while concentrating on the examination.

[0078] Furthermore, by combining the learning model 51 and the APR 53, it is possible to realize a configuration in which an image of the subject M is used as input information and appropriate image acquisition conditions 63 are acquired as output information. With current machine learning, it is difficult to accurately infer image acquisition conditions 63 that can appropriately generate an image of the training information using an image of a human body as training information. In other words, it is difficult to use a learning model to directly infer image acquisition conditions from an image of the subject.

[0079] As a result of intensive research, the inventors have found that it is possible to perform machine learning that uses an image of the subject M as input information and accurately infers the body part shown in the image. By combining the APR 53, in which the body part and the image acquisition conditions 63 are linked, with the learning model 51, they have realized a configuration that accurately infers appropriate image acquisition conditions 63 starting from the image of the subject M. [Example]

[0080] Next, a second embodiment of the present invention will be described. The overall configuration of an X-ray device 1A according to the second embodiment is basically the same as the overall configuration of the X-ray device 1 according to the first embodiment as shown in Fig. 1. Therefore, in the second embodiment, the same components as those in the first embodiment are denoted by the same reference numerals, and detailed description thereof will be omitted.

[0081] The X-ray device 1A according to the second embodiment differs from the first embodiment in that the X-ray device 1A infers the inspection items in the image using the learning model 51A, and then reads out the image acquisition conditions 63 corresponding to the inspection items. Hereinafter, a configuration for automatically reading out the image acquisition conditions 63 in the second embodiment will be described.

[0082] Generally, when the target region shown in the X-ray image differs, the various parameters of the image acquisition conditions 63 suitable for acquiring the X-ray image differ. However, even if the target region of the X-ray image is the same, when the type of examination (also called examination item or procedure) performed on the target region differs, the various parameters of the image acquisition conditions 63 suitable for acquiring the X-ray image also differ. As an example, the image acquisition conditions 63 suitable for acquiring the X-ray image differ between when performing general X-ray photography without introducing anything into the abdomen and when performing ERCP with an endoscope inserted into the abdomen.

[0083] Therefore, the learning model 51A according to the second embodiment is configured to output inspection items using an image as input information. The learning model 51A according to the second embodiment is common to the learning model 51 according to the first embodiment in that the machine learning unit 45 performs machine learning using the original image R1, the primary augmented image R2, and the secondary augmented image R3 as training images. However, the learning model 51 according to the first embodiment recognizes the bones, organs, contours, etc. of a human body shown in an image, and infers the parts of the human body shown in the image using the structure of the human body as a clue.

[0084] On the other hand, the learning model 51A according to the second embodiment is configured to detect not only the structure of the human body, such as bones and organs, but also examination equipment, such as a catheter or an endoscope, and an examination agent, such as a contrast agent, that appear in the image. The learning model 51A infers what examination item the image relates to, using information about the structure of the human body as well as information about the presence or absence of examination equipment or examination agent as clues. That is, the learning model 51A is configured to receive an image of the subject M as input information and output information specifying the examination item related to the image as examination item information 71.

[0085] Each diagram in FIG. 14 illustrates the relationship between an input image and an inference result by the learning model 51A. When an X-ray image W1 such as that shown in FIG. 14(a) is input to the learning model 51A as an input image, the learning model 51A analyzes the X-ray image W1 to detect the heart H, lungs Lu, thoracic vertebrae (not shown), and the contours of the human body. Based on information such as the presence of the heart H, lungs Lu, and the absence of examination equipment such as a catheter, the learning model 51A infers that the X-ray image W1 is an image of a general X-ray radiography with the chest as the target region. As a result, when the X-ray image W1 is used as input information, the learning model 51A outputs examination item information 71 (examination item information 71a) indicating that the examination item of the input image is "general chest radiography."

[0086] 14(b) is input to the learning model 51A as an input image, the learning model 51A analyzes the X-ray image W2 to detect the heart H, lungs Lu, catheter Ch, etc., and uses these as clues to infer that the X-ray image W2 is an image of a catheterization procedure targeting the chest, i.e., PCI, as an examination item. As a result, when the learning model 51A receives the X-ray image W2 as input information, it outputs examination item information 71 (examination item information 71b) indicating that the examination item for the input image is "thoracic PCI."

[0087] 14(c) is input to the learning model 51A as an input image, the learning model 51A analyzes the X-ray image W2 and detects the lumbar vertebrae Bh, stomach Ga, etc. Then, based on the information that the lumbar vertebrae Bh, stomach Ga, etc. are present and the information that examination equipment such as a catheter is not present, the learning model 51A infers that the X-ray image W3 is an image for which the examination item is general abdominal X-ray photography. As a result, when the learning model 51A receives the X-ray image W3 as input information, it outputs examination item information 71 (examination item information 71c) that reads "general abdominal X-ray photography."

[0088] When an X-ray image W4 as shown in FIG. 14(d) is input to the learning model 51A as an input image, the learning model 51A detects the lumbar vertebrae Bh, stomach Ga, endoscope Es, etc. Using these as clues, it infers that the X-ray image W4 is an image from an endoscopic examination of the abdomen, i.e., an ERCP examination. As a result, when the learning model 51A receives the X-ray image W4 as input information, it outputs examination item information 71 (examination item information 71d) that reads "abdominal ERCP."

[0089] When an X-ray image W5 such as that shown in FIG. 14(e) is input to the learning model 51A as an input image, the learning model 51A detects the lumbar vertebrae Bh, the stomach Ga, and the contrast agent Ct that has accumulated in accordance with the shape of the stomach. Using these as clues, the learning model 51A infers that the X-ray image W5 is an examination in which the stomach is contrasted, i.e., an upper gastrointestinal series (UGI) examination is the examination item. As a result, when the learning model 51A receives the X-ray image W5 as input information, it outputs examination item information 71 (examination item information 71e) indicating "abdominal UGI."

[0090] In this way, by inputting an X-ray image F or an optical image D as input information, machine learning of the learning model 51A is performed so that the image as input information is inferred as to what examination item the image as input information relates to, and the inferred information on the examination item 71 is output. The trained learning model 51A is stored in the learning model storage unit 55.

[0091] Here, the APR 53A according to Example 2 will be described. The learning model 51A takes an image as input information and outputs examination item information 71, and therefore, as shown in Fig. 15(a), the APR 53A is configured so that image acquisition conditions 63 are associated with each piece of examination item information 71. That is, by inputting the examination item information 71 to the APR 53A, the image acquisition conditions 63 according to the examination item related to the examination item information 71 are output from the APR 53A.

[0092] 15B shows specific contents of the image acquisition conditions 63 linked to the examination item information 71 in the APR 53A according to the second embodiment. As an example, among the examination item information 71, examination item information 71a having general chest X-ray imaging as the examination item is linked in advance to image acquisition conditions 63e including image processing conditions 67e, X-ray fluoroscopy conditions 68e, and X-ray imaging conditions 69e. Examination item information 71b having chest PCI as the examination item is linked in advance to image acquisition conditions 63f. Examination item information 71c having abdominal general X-ray imaging as the examination item is linked in advance to image acquisition conditions 63g. Examination item information 71d having abdominal ERCP as the examination item is linked in advance to image acquisition conditions 63h. In this way, in the second embodiment, the APR 53A in which appropriate image acquisition conditions 63 are linked to each of a plurality of examination item information 71 is set in advance and stored in the condition storage unit 57.

[0093] In the second embodiment, the image analysis unit 47 uses the learning model 51A to identify the examination item related to the image obtained of the subject M. That is, by inputting an X-ray image or the like transmitted from the image processing unit 33 as an input image to the learning model 51A, the learning model 51A infers the examination item being performed in the X-ray image using clues such as the components of the human body shown in the input X-ray image. Information obtained by the inference is output from the learning model 51A as examination item information 71. The condition reading unit 49 inputs the examination item information 71 obtained by the image analysis unit 47 to the APR 53A, thereby setting image acquisition conditions 63 appropriate for the examination item related to the examination item information 71. The image analysis unit 47 according to the second embodiment corresponds to the examination type inference unit in the present invention.

[0094] A series of steps for automatically setting the image acquisition conditions 63 in Example 2 will now be described with reference to Fig. 16. As a specific example of an examination item, a case will be described in which the image acquisition conditions 63 are set in a state in which a catheter Ch is advanced into the chest Lc and PCI is being performed.

[0095] While the catheter Ch is advanced into the chest Lc, the chest Lc is irradiated with X-rays from the X-ray tube 5, and the image processing unit 33 generates an X-ray image W2 of the target region of the chest Lc. Data of the generated X-ray image W2 is sent to the image analysis unit 47.

[0096] The image analysis unit 47 inputs the X-ray image W2 to the learning model 51A as an input image. The learning model 51 estimates and outputs the parts of the body that appear in the X-ray image W2, which is the input information. In this case, using the heart H, lungs Lu, and catheter Ch of the subject M that appear in the X-ray image W2 as clues, the learning model 51A estimates that the X-ray image W2 is most likely an image in which the examination item is thoracic PCI. As a result, the learning model 51A outputs information that the examination item of the input image is thoracic PCI, i.e., examination item information 71b. The examination item information 71b output from the learning model 51A is sent to the condition reading unit 49.

[0097] The condition reading unit 49 searches for image acquisition conditions 63 appropriate for acquiring X-ray images by chest PCI related to the examination item information 71b. That is, by inputting the examination item information 71b selected as an examination item to the APR 53A, the image acquisition conditions 63 stored in association with the examination item information 71b in the APR 53A, i.e., the image acquisition conditions 63f, are read and output.

[0098] As a result, the X-ray tube 5 irradiates X-rays based on the X-ray irradiation conditions 65f, and the image processor 33 performs image processing of the X-ray image F based on the image processing conditions 67f. Therefore, the image acquisition conditions 63f are set automatically without the need for manual selection of examination items, and the visibility of the X-ray image W2 is improved promptly using the appropriate image acquisition conditions 63f. The operator places a stent in the coronary artery while checking the X-ray image W2, and then completes the PCI procedure.

[0099] In the second embodiment, by using the learning model 51A and the APR 53A, the examination item being performed on the most recently acquired image of the subject M is inferred as input information, and the image acquisition conditions 63 appropriate for the examination item can be read and set. That is, the image acquisition conditions 63 are automatically and appropriately changed not only when the target site for X-ray irradiation is changed, but also when an operation such as inserting the endoscope Es or injecting the contrast agent Ct is performed.

[0100] Therefore, when performing an examination using examination equipment or an examination agent, the burden on the operator and the subject M can be reduced. As an example, when performing a contrast examination using a contrast agent Ct, the image acquisition conditions 63 appropriate for general X-ray imaging are automatically selected before the injection of the contrast agent Ct, and are automatically changed to the image acquisition conditions 63 appropriate for contrast imaging after the injection of the contrast agent Ct. In other words, since there is no need to manually select the examination items before and after the injection of the contrast agent Ct, the operator can concentrate on the procedure of the contrast examination. Furthermore, when the region to be X-rayed is continuously changed to track the contrast agent Ct injected into the digestive tract or blood vessels, the examination items also change with the change in the target region, and the image acquisition conditions 63 are automatically changed in accordance with the change in the examination items. Therefore, the risk of losing track of the contrast agent due to manually setting the image acquisition conditions 63 can be more reliably avoided. [Example]

[0101] Next, a third embodiment of the present invention will be described. As shown in Fig. 17, an X-ray apparatus 1B according to the third embodiment differs from the first embodiment in that a main controller 39 further includes an erroneous analysis detector 73.

[0102] The misanalysis detection unit 73 is disposed after the image analysis unit 47 and before the condition reading unit 49. The misanalysis detection unit 73 is configured to receive the position information of the C-arm 9 detected at any time by the arm position detection unit 37, and to receive the body part information 61 output from the image analysis unit 47. The misanalysis detection unit 73 detects whether or not an analysis error has occurred by the learning model 51, based on whether or not the C-arm position information and the body part information 61 have changed.

[0103] 18 is a diagram illustrating a case where an erroneous analysis by the learning model 51 occurs in the image analysis unit 47. Fig. 18(a) shows a state in which the learning model 51 correctly analyzes an X-ray image F2 showing the abdomen. In this case, the learning model 51 correctly analyzes the X-ray image F2 obtained at a predetermined timing Ta, determines that the X-ray image F2 is an image showing the abdomen, and outputs abdominal region information 61b.

[0104] On the other hand, due to noise or pixel value fluctuations occurring in the X-ray image F, which is the input information, the learning model 51 may erroneously recognize the region appearing in the input image. As an example, as shown in FIG. 18(b), assume that an abdominal X-ray image F4, which shows a ring-shaped artifact At, is acquired at time Tb, which is after time Ta. When the abdominal X-ray image F4, which shows the ring-shaped artifact At, is input to the learning model 51 as input information, the learning model 51 may erroneously recognize the image of the stomach Ga and the artifact At as images of the lungs, and may erroneously analyze the X-ray image F4, which actually shows the abdomen, as an image of the chest. When such an erroneous analysis occurs, the learning model 51 will output chest region information 61c, even though the abdominal X-ray image F4 was input.

[0105] When the region information 61 output due to such an erroneous analysis is sent to the condition reading unit 49, the condition reading unit 49 outputs and sets image acquisition conditions 63c suitable for acquiring chest X-ray images. As a result, the image acquisition conditions 63c that are inappropriate for acquiring abdominal X-ray images are applied, and the visibility of subsequently generated abdominal X-ray images F4 is reduced.

[0106] Therefore, the X-ray device 1B according to the third embodiment is provided with an erroneous analysis detection unit 73, thereby preventing the generation of an X-ray image F by applying an inappropriate image acquisition condition 63 due to an erroneous analysis by the learning model 51. Fig. 19 is a flowchart of the process in which the erroneous analysis detection unit 73 detects whether or not an erroneous analysis has occurred by the learning model 51.

[0107] The misanalysis detection unit 73 compares the position information of the X-ray irradiation field with the region information 61 output by the learning model 51 for each frame of the X-ray image F. That is, for an X-ray image acquired at a predetermined timing T1, the position information of the X-ray irradiation field when generating the X-ray image and the region information 61 output by the learning model 51 are sent to the misanalysis detection unit 73 (step S1).

[0108] In the third embodiment, the position of the X-ray irradiation field is specified by the position of the C-arm 9. That is, the position of the X-ray irradiation field is detected by the arm position detection unit 37 that detects the position of the C-arm 9, and the position information of the X-ray irradiation field is transmitted from the arm position detection unit 37 to the erroneous analysis detection unit 73. In addition, the part information 61 output by the learning model 51 is transmitted from the image analysis unit 47 to the erroneous analysis detection unit 73. The arm position detection unit 37 corresponds to the irradiation position detection unit in this invention.

[0109] Then, for the X-ray image acquired at time T2, one frame's worth of time after time T1, the position information of the X-ray irradiation field and the part information 61 output by the learning model 51 are sent to the misanalysis detection unit 73 (step S2).

[0110] The misanalysis detection unit 73 first compares the region information 61 of the X-ray image acquired at timing T1 with the region information 61 of the X-ray image acquired at timing T2, and determines whether the region information 61 has changed between timing T1 and timing T2 (step S3). If there is no change in the region information 61, the misanalysis detection unit 73 determines that no misanalysis of the learning model 51 has occurred, and transmits the region information 61 acquired at timing T2 to the condition reading unit 49, causing it to read the image acquisition conditions 63 (step SR1). Therefore, the X-ray tube 5 and image processing unit 33 are controlled using the image acquisition conditions 63 acquired at timing T2, and the acquisition of X-ray images from timing T2 onwards is performed.

[0111] On the other hand, if the region information 61 has changed between timing T1 and timing T2, the process proceeds to step S4. In step S4, the erroneous analysis detection unit 73 compares the position information of the X-ray irradiation field acquired at timing T1 with the position information of the X-ray irradiation field acquired at timing T2, and determines whether the position of the X-ray irradiation field has changed between timing T1 and timing T2.

[0112] If it is determined that the X-ray irradiation field has shifted between timing T1 and timing T2, misanalysis detection unit 73 determines that no misanalysis of learning model 51 has occurred. Then, misanalysis detection unit 73 transmits region information 61 obtained at timing T2 to condition reading unit 49, causing image acquisition conditions 63 to be read (step SR1). That is, since it is considered that the region information 61 output by learning model 51 has changed as a result of the X-ray irradiation field shifting and the region targeted for X-ray irradiation being changed, misanalysis detection unit 73 determines that no misanalysis of learning model 51 has occurred. Therefore, X-ray tube 5 and image processing unit 33 are controlled using image acquisition conditions 63 obtained at timing T2, and X-ray images are acquired from timing T2 onward.

[0113] On the other hand, if it is determined in step S4 that the X-ray irradiation field has not shifted between timing T1 and timing T2, misanalysis detection unit 73 determines that a misanalysis of learning model 51 has occurred (step S5). Then, misanalysis detection unit 73 does not use region information 61 acquired at timing T2, but sends region information 61 acquired at timing T1 to condition reading unit 49 to read image acquisition conditions 63 (step SR2). Therefore, X-ray tube 5 and image processing unit 33 are controlled using image acquisition conditions 63 acquired at timing T1, and X-ray images are acquired from timing T2 onwards.

[0114] If the X-ray irradiation field has not shifted, the target area of ​​X-ray irradiation has not changed, and therefore the area information 61 output by the learning model 51 should not change either. Therefore, if the area information 61 output by the learning model 51 changes even though the X-ray irradiation field has not shifted between timing T1 and timing T2, it is considered that the learning model 51 has performed an erroneous analysis. Therefore, if it is determined in step S4 that there has been no change in the position information of the X-ray irradiation field, it can be determined that the area information 61 acquired at timing T2 is erroneous information. In this way, by determining whether the X-ray irradiation field has shifted and whether the area information 61 has changed for each frame, the erroneous analysis detection unit 73 can detect whether an erroneous analysis has occurred in the learning model 51.

[0115] The process of making a determination by the erroneous analysis detection unit 73 in Example 3 will be specifically described with reference to Figs. 20 and 21. Fig. 20 shows the process when no erroneous analysis has occurred. At timing T1, the C-arm 9 is positioned at abdomen Lb, and an X-ray image F2 showing the abdomen is generated. The image analysis unit 47 inputs the X-ray image F2 as input information into the learning model 51, thereby outputting abdominal region information 61b. The abdominal region information 61b is transmitted to the erroneous analysis detection unit 73 as region information 61 relating to timing T1. At this time, the arm position detection unit 37 also transmits position information to the erroneous analysis detection unit 73 indicating that the position of the X-ray irradiation field at timing T1 is abdomen Lb.

[0116] At timing T2, which corresponds to the frame following timing T1, C-arm 9 moves to chest Lc, and an X-ray image F3 of the chest is generated. Image analysis unit 47 inputs X-ray image F3 as input information to learning model 51, thereby outputting chest region information 61c. Chest region information 61c is transmitted to erroneous analysis detection unit 73 as region information 61 relating to timing T2. Arm position detection unit 37 then transmits position information to erroneous analysis detection unit 73 indicating that the position of the X-ray irradiation field at timing T2 is chest Lc.

[0117] The misanalysis detection unit 73 compares the region information 61 relating to each of timings T1 and T2 with the position information of the X-ray irradiation field. First, by comparing the region information 61b relating to timing T1 with the region information 61c relating to timing T2, the misanalysis detection unit 73 determines that there has been a change in the region information 61 between timings T1 and T2 (step S3). Next, by comparing the position information Lb of the X-ray irradiation field relating to timing T1 with the position information Lc of the X-ray irradiation field relating to timing T2, the misanalysis detection unit 73 determines that there has been a change in the position of the X-ray irradiation field between timings T1 and T2 (step S4).

[0118] Because there are changes in both the region information 61 and the position information of the X-ray irradiation field, the misanalysis detection unit 73 determines that there was no misanalysis in the learning model 51 and accepts the analysis result of the learning model 51 at timing T2 (region information 61c). The misanalysis detection unit 73 then transmits the region information 61c to the condition reading unit 49 (step SR1). In other words, when proceeding to step SR1, the misanalysis detection unit 73 transmits the image acquisition conditions 63 output by the learning model 51 after the content of the region information 61 changed to the condition reading unit 49. As a result, the image acquisition conditions 63c are read in the condition reading unit 49, and X-ray images at timing T2 and after are generated using the image acquisition conditions 63c.

[0119] Figure 21 shows the steps when an erroneous analysis occurs. At time T1, the C-arm 9 is positioned at abdomen Lb, and an X-ray image F2 of the abdomen is generated. The learning model 51 uses the X-ray image F2 as input information and outputs abdominal region information 61b. The abdominal region information 61b is sent to the erroneous analysis detection unit 73. The arm position detection unit 37 sends position information to the erroneous analysis detection unit 73 indicating that the position of the C-arm 9 at time T1 is abdomen Lb.

[0120] At timing T2, the C-arm 9 has not moved from the abdomen Lb, and an X-ray image F4 is generated with the abdomen Lb as the X-ray irradiation field. Here, the learning model 51 uses the X-ray image F4 as input information, and as a result, erroneously analyzes the image in the X-ray image F4 due to artifacts At, etc., and outputs region information 61c of the chest. The chest region information 61c output due to the erroneous analysis is transmitted to the erroneous analysis detection unit 73 as region information 61 relating to timing T2. The arm position detection unit 37 then transmits position information to the erroneous analysis detection unit 73 indicating that the position of the X-ray irradiation field at timing T2 is the abdomen Lb.

[0121] The misanalysis detection unit 73 compares the region information 61 relating to each of timings T1 and T2 with the position information of the X-ray irradiation field. First, by comparing the region information 61b relating to timing T1 with the region information 61c relating to timing T2, the misanalysis detection unit 73 determines that the region information 61 has changed between timings T1 and T2 (step S3). Next, by comparing the position information Lb of the X-ray irradiation field relating to timing T1 with the position information Lb of the X-ray irradiation field relating to timing T2, the misanalysis detection unit 73 determines that the position of the X-ray irradiation field has not changed between timings T1 and T2 (step S4).

[0122] Because there is a change in region information 61 while there is no change in the position information of the X-ray irradiation field, misanalysis detection unit 73 determines that a misanalysis has occurred in learning model 51 and ignores or discards the analysis result of learning model 51 for timing T2 (region information 61c) as erroneous information (step S5). Then, misanalysis detection unit 73 transmits region information 61b obtained at timing T1, which is immediately before timing T2, to condition reading unit 49 (step SR2). In other words, in step SR2, misanalysis detection unit 73 transmits image acquisition conditions 63 output by learning model 51 before the content of region information 61 changed to condition reading unit 49. As a result, image acquisition conditions 63b are read in condition reading unit 49, and X-ray images for timing T2 and later are generated using image acquisition conditions 63b.

[0123] In the third embodiment, an erroneous analysis detection unit 73 is further provided, thereby enabling detection of an erroneous analysis in the learning model 51. The erroneous analysis detection unit 73 determines whether or not the results output by the learning model 51 have changed and whether or not the position of the X-ray irradiation field has changed. If the erroneous analysis detection unit 73 detects that the results output by the learning model 51 have changed but the position of the X-ray irradiation field has not changed, it determines that an erroneous analysis has occurred in the learning model 51 and discards the most recently obtained output result of the learning model 51 (here, the region information 61 obtained at timing T2). The most recently obtained output result of the learning model 51 (here, the region information 61 obtained at timing T1) is then sent to the condition reading unit 49, and the image acquisition conditions 63 are read.

[0124] The misanalysis detection unit 73 constantly determines whether the position of the X-ray irradiation field has changed and whether the output result of the learning model 51 has changed, so that even if the learning model 51 performs an erroneous analysis due to noise in the input information or the like and outputs erroneous region information 61, the erroneous region information 61 is ignored. Therefore, even if the learning model 51 performs an erroneous analysis due to noise in the input information or the like, it is possible to avoid a situation in which inappropriate image acquisition conditions 63 are read out based on the erroneous region information 61 output by the erroneous analysis. Therefore, the influence of erroneous analysis by the learning model 51 due to fluctuations or the like can be reduced, and appropriate image acquisition conditions 63 can be automatically read out. [Example]

[0125] Next, a fourth embodiment of the present invention will be described. In the first and second embodiments, the image acquisition conditions 63 read by the condition readout unit 49 are automatically set to generate the X-ray image F. That is, the latest image acquisition conditions 63 read by the condition readout unit 49 are automatically transmitted to the X-ray tube 5 or the image processing unit 33, and the X-ray tube 5 or the image processing unit 33 is always controlled according to the latest image acquisition conditions 63.

[0126] On the other hand, in the fourth embodiment, when predetermined conditions are satisfied, an approval step is generated as a preliminary step to a step of controlling the X-ray tube 5 or the image processing unit 33 using the image acquisition conditions 63 read by the condition reading unit 49. The approval step is a step in which the operator decides whether or not to set the latest image acquisition conditions 63 read by the condition reading unit 49.

[0127] As shown in FIG. 22, the X-ray device 1C of Example 4 differs from the configuration of Example 1 etc. in that the main control unit 39 further includes an approval condition determination unit 75, an approval key display control unit 77, and a condition setting control unit 79.

[0128] The approval condition determination unit 75 determines whether a predetermined condition (approval condition) that requires an approval step has occurred. In the fourth embodiment, the predetermined condition is a change in the content of the information output by the learning model 51. When the approval condition determination unit 75 determines that the predetermined condition is met, it transmits information to the approval key display control unit 77 that an approval step is required.

[0129] The approval key display control unit 77 is provided after the approval condition determination unit 75 and controls the display unit 35 to display the approval key GK. When information indicating that an approval step is required is transmitted from the approval condition determination unit 75, the approval key display control unit 77 controls the display unit 35 to display the approval key GK. Figure 23 shows an example of the display form of the approval key GK on the display unit 35. Details of the approval key GK will be described later.

[0130] The condition setting control unit 79 is provided before the condition reading unit 49 and after the approval condition determination unit 75. When a predetermined condition that requires an approval step occurs, the condition setting control unit 79 controls whether or not the latest image acquisition conditions 63 read by the condition reading unit 49 can be set. The condition setting control unit 79 is configured to control the condition reading unit 49 so that the latest image acquisition conditions 63 are set when the approval key GK is operated by the operator as a trigger.

[0131] In this embodiment, the condition setting control unit 79 is configured to transmit a signal (inhibition signal ST) that inhibits the latest image acquisition conditions 63 from being set as conditions for X-ray image acquisition to the condition reading unit 49 under the control of the approval condition determination unit 75. The condition setting control unit 79 is also controlled so that transmission of the inhibition signal ST is stopped when the approval key GK is operated by the operator as a trigger.

[0132] Fig. 24(b) is a flowchart showing a series of operations of the X-ray device 1C according to Example 4. In Example 1 and the like, as shown in Fig. 24(a), when the condition reading unit 49 reads out the image acquisition conditions 63 in step M4, the process automatically and unconditionally proceeds to step M5. That is, the most recently read image acquisition conditions 63 are automatically set as conditions for X-ray image acquisition and transmitted to the X-ray tube 5 or the image processing unit 33. The process then proceeds to step M6, where the next and subsequent frames of X-ray images F are generated according to the latest image acquisition conditions 63.

[0133] On the other hand, in the fourth embodiment, an approval condition determination step Q1 is present between step M4 and step M5. The approval condition determination step Q1 is a step for determining whether or not an approval condition has occurred. If it is determined that an approval condition has not occurred, that is, in this embodiment, if the content of the information output by the learning model 51 has not changed, the process proceeds to step M5. In other words, if an approval condition has not occurred, the latest image acquisition conditions 63 read by the condition reading unit 49 are set as the conditions for acquiring an X-ray image.

[0134] If the approval condition determination unit 75 determines in approval condition determination step Q1 that an approval condition has occurred, the process proceeds from step Q1 to step M4a, where the approval key GK is displayed. As an example, the approval key GK is displayed in the approval condition occurrence area K5 on the display unit 35, as shown in FIG. 23. In addition to the approval key GK and the rejection key DK, the approval condition occurrence area K5 displays information J1 indicating that an approval condition has occurred and information J2 prompting the user to select whether or not to approve the latest image acquisition conditions 63. The presence and content of the information J1 and information J2 may be selected as appropriate.

[0135] After the approval key GK is displayed, the process proceeds to step Q2. In step Q2, the operator selects whether or not to approve the latest image acquisition conditions 63. If the latest image acquisition conditions 63 are approved, the operator operates the operation console 41 or the like to select the approval key GK. Examples of the operation of selecting the approval key GK include operating the mouse on the operation console 41 to click the approval key GK, or touching the approval key GK displayed on the display unit 35, which is a touch panel. The operation console 41 corresponds to the approval instruction input unit in this invention.

[0136] When the operation of selecting the approval key GK is performed, the process proceeds from step Q2 to step M5, where the latest image acquisition conditions 63 are set as the conditions for acquiring an X-ray image. Then, the process proceeds to step M6, where an X-ray image is generated using the set image acquisition conditions 63. That is, the X-ray image F from the next frame onward is generated using the latest image acquisition conditions 63.

[0137] On the other hand, the operator may determine that using the latest image acquisition conditions 63 may actually decrease the visibility of the X-ray image F. In addition, the operator may determine that there is no need to approve the latest image acquisition conditions 63 because sufficient visibility is obtained in the X-ray image F obtained using the image acquisition conditions 63 currently set as conditions for X-ray image acquisition (the most recently set image acquisition conditions 63).

[0138] In this way, if the operator decides not to approve the latest image acquisition conditions 63, the operator either leaves the approval key GK unselected or selects the rejection key DK. If the rejection key DK is selected, the information such as the approval key GK that was displayed in the approval condition generation area K5 is hidden.

[0139] If the approval key GK is not selected, the process proceeds to step M5a. When the process proceeds to step M5a, the latest image acquisition conditions 63 are not set as the conditions for X-ray image acquisition, and the most recently set image acquisition conditions 63 continue to be set as the conditions for X-ray image acquisition. In other words, the image acquisition conditions 63 that have already been set as the most recently set conditions for X-ray image acquisition are continuously transmitted to the X-ray tube 5 or the image processing unit 33. The process then proceeds to step M6, where an X-ray image is generated using the set image acquisition conditions 63. In other words, the most recently set image acquisition conditions 63 are continuously used to generate the X-ray images F for the next frame and thereafter.

[0140] Here, a series of steps for setting image acquisition conditions 63 in Example 4 will be described with reference to Figs. 25 to 28, taking a specific example. Fig. 25 shows the position of the X-ray irradiation field at each timing. In Example 4, as a specific example, it is assumed that X-ray image F is acquired from timings P1 to P4. Then, it is assumed that the C-arm 9 is displaced just before timing P3, and the X-ray irradiation field is displaced from the abdomen Lb to the chest Lc. That is, the position of the X-ray irradiation field is the abdomen Lb at times P1 and P2, and the position of the X-ray irradiation field is the chest Lc at times P3 and P4. It is also assumed that the approval key GK is operated at timing P4.

[0141] First, the steps of setting image acquisition conditions 63 at timings P1 and P2 will be described. The steps up to step M4 are the same as those in Example 1 shown in FIG. 10. That is, X-rays are irradiated from X-ray tube 5 to acquire X-ray image F2 of abdomen Lb (step M2), and learning model 51 outputs region information 61b using X-ray image F2 as an input image (step M3). Then, APR 53 reads image acquisition conditions 63b linked to region information 61b as the latest image acquisition conditions 63 (step M4).

[0142] In the fourth embodiment, the process proceeds from step S4 to step Q1 as shown in FIG. 24(b). At timings P1 and P2, the position of the X-ray irradiation field is not changed. Therefore, the approval condition determination unit 75 determines that the approval condition does not occur at timings P1 and P2. If the approval condition does not occur, as shown in FIG. 26, the approval condition determination unit 75 does not perform any special control over the approval key display control unit 77 and the condition setting control unit 79. Therefore, the approval key display control unit 77 does not perform any special control over the display unit 35, and the approval key GK is not displayed.

[0143] Furthermore, if the approval condition has not occurred, the condition setting control unit 79 is not under the control of the approval condition determination unit 75, and therefore the condition setting control unit 79 does not transmit an inhibition signal ST to the condition readout unit 49. Therefore, the latest image acquisition conditions 63 (here, image acquisition conditions 63b) read out by the condition readout unit 49 are set as the conditions for X-ray image acquisition (step M5). That is, at times P1 and P2, the latest image acquisition conditions 63b are transmitted to the X-ray tube 5 or the image processing unit 33, and an X-ray image F is acquired in accordance with the image acquisition conditions 63b (step M6).

[0144] Next, the process of setting image acquisition conditions 63 at timing P3 will be described. The process up to step M4 is the same as the process in embodiment 1 shown in Fig. 11. That is, X-rays are irradiated from X-ray tube 5 to acquire an X-ray image F3 of chest Lc (step M2), and learning model 51 outputs region information 61c using X-ray image F3 as an input image (step M3). Then, APR 53 reads image acquisition conditions 63c linked to region information 61c as the latest image acquisition conditions 63 (step M4).

[0145] However, at timing P3, the process from step M4 onwards differs from those at timings P1 and P2. The position of the X-ray irradiation field at timing P2 is abdomen Lb, whereas the position of the X-ray irradiation field at timing P3 is chest Lc. That is, as shown by symbol L1 in Figure 25, the information output by learning model 51 at timing P2 is region information 61b, whereas the information output by learning model 51 at timing P3 has changed to region information 61c. Therefore, at timing P3, in step Q1, approval condition determination unit 75 determines that the approval condition has occurred.

[0146] If it is determined that an approval condition has occurred, the approval condition determination unit 75 transmits a control signal to the approval key display control unit 77 and the condition setting control unit 79. The approval key display control unit 77 causes the display unit 35 to display the approval key GK in accordance with the control signal from the approval condition determination unit 75 (step M4a). Specifically, as shown in Fig. 23 , the approval key GK, rejection key DK, etc. are displayed in the approval condition occurrence area K5 of the display unit 35.

[0147] Furthermore, while the approval condition has occurred, the approval key GK has not been operated at timing P3 (step Q2). Therefore, the condition setting control unit 79 transmits an inhibition signal ST to the condition reading unit 49 in accordance with the control signal from the approval condition determination unit 75. Therefore, the latest image acquisition conditions 63c read out by the condition reading unit 49 are not set as the conditions for X-ray image acquisition, and the image acquisition conditions 63 set at the most recent timing continue to be set as the conditions for X-ray image acquisition (step M5a). In other words, the image acquisition conditions 63c read out at timing P3 are not transmitted to the X-ray tube 5 or the image processing unit 33 (see symbol RP in FIG. 27).

[0148] The timing immediately preceding timing P3 is timing P2. Therefore, image acquisition conditions 63b that were set as conditions for acquiring an X-ray image at timing P2 continue to be set as conditions for acquiring an X-ray image at timing P3. That is, as shown by symbol L2 in Fig. 25, even though image acquisition conditions 63c are read out by condition reading unit 49 at timing P3, image acquisition conditions 63b are set as conditions for acquiring an X-ray image, and an X-ray image F is acquired using image acquisition conditions 63b (step M6).

[0149] Finally, the process of setting image acquisition conditions 63 at timing P4 will be described. The process up to step M4 is the same as that at timing P3. That is, learning model 51 outputs region information 61c (step M3). Then, APR 53 reads image acquisition conditions 63c linked to region information 61c as the latest image acquisition conditions 63 (step M4).

[0150] However, while the approval key GK is not operated at timing P3, at timing P4 the operator checks the approval key GK displayed on the display unit 35 and operates the approval key GK using the operation console 41 or the like (step M4a, step Q2). By operating the approval key GK, an instruction to approve the most recently read image acquisition conditions 63 is input.

[0151] When the approval key GK is operated in step Q2, a signal AN indicating that the transmission of the inhibition signal ST is to be stopped is sent to the condition setting control unit 79 via the console 41, as shown in Fig. 28. The condition setting control unit 79 stops the transmission of the inhibition signal ST upon receiving the signal AN. Therefore, when the approval key GK is operated at timing P4, the image acquisition conditions 63c most recently read by the condition reading unit 49 are set as the conditions for X-ray image acquisition (step M5). Therefore, the most recent image acquisition conditions 63c are sent from the condition reading unit 49 to the X-ray tube 5 or the image processing unit 33, and an X-ray image F is acquired in accordance with the image acquisition conditions 63c (step M6).

[0152] As described above, in the fourth embodiment, an approval step including steps Q1, M4a, and Q2 is executed as a preliminary step for setting the image acquisition conditions 63 as conditions for acquiring an X-ray image. The image acquisition conditions 63 linked to the body part information 61 in the APR 53 are identified as appropriate parameters when the subject's physique is within a general range. Therefore, depending on various conditions such as the subject's physique, the image acquisition conditions 63 read out using the APR 53 may not be optimal as parameters for actually generating an X-ray image of the subject. In this case, if the image acquisition conditions 63 read out from the condition readout unit 49 are automatically set as conditions for acquiring an X-ray image, the visibility of the X-ray image for the operator would be reduced.

[0153] Therefore, in this embodiment, an approval step is provided, and an operation by the operator to approve the settings of the most recently read image acquisition conditions 63 is used as a trigger to generate an X-ray image F using parameters related to the most recent image acquisition conditions 63. By providing the approval step, it is possible to avoid a situation in which image acquisition conditions 63 that are not actually optimal are automatically set as conditions for acquiring an X-ray image against the operator's will.

[0154] <Effects of the configuration of the embodiment> (Item 1) The X-ray device (1) according to this embodiment includes an X-ray tube (5) that irradiates a subject M with X-rays, an X-ray detector (7) that detects X-rays that have passed through the subject M, an image processing unit (33) that generates an X-ray image by performing image processing using a detection signal output by the X-ray detector (7), a condition storage unit (57) that stores image acquisition conditions (63) including at least one of X-ray irradiation conditions (65) and image processing conditions (67) corresponding to each target region of the subject M in association with the target region, a learning model storage unit (55) that stores a learning model (51) that infers and outputs a region of the human body shown in an image by performing machine learning using an image of the human body as a teacher image, and a direct The apparatus includes a target region inference unit (47) that inputs at least one of an X-ray image (F) and an optical image (D) of a subject M obtained recently into a learning model (51) as an input image, thereby inferring and outputting the region shown in the input image; a condition reading unit (49) that selects the region output by the target region inference unit (47) as region information, thereby reading out image acquisition conditions (63) that are linked to the region information and stored in a condition storage unit (57); and a control unit (39) that controls at least one of the X-ray tube (5) and the image processing unit (33) in accordance with the image acquisition conditions (63) read out by the condition reading unit (49).

[0155] According to the X-ray device 1 described in paragraph 1, by using a learning model 51 that infers a target region shown in an image and an APR 53 to which appropriate image acquisition conditions are linked according to the target region, image acquisition conditions 63 including at least one of X-ray irradiation conditions 65 and image processing conditions 67 are automatically set. The learning model 51 is configured to infer and output the target region of the human body shown in the image through machine learning using a human body image as a training image. That is, the image analysis unit 47 inputs an image of the subject M as an input image to the learning model 51, and the target region of the subject shown in the input image is inferred and output. The condition reading unit 49 selects the output target region of the subject M as region information 61, thereby automatically reading out the image acquisition conditions 63 linked to the region information 61 and stored. Therefore, when an X-ray image F of the subject is acquired, the image analysis unit 47 and the condition reading unit 49 automatically read out the image acquisition conditions 63 appropriate for the irradiation field of the X-ray image F. Therefore, even if the target region of the subject M to be irradiated with X-rays is changed, the image acquisition conditions 63 appropriate for the changed target region are automatically read out. In other words, the process of the operator manually selecting the target region and setting the image acquisition conditions 63 is no longer necessary, so X-ray fluoroscopy or X-ray imaging can be performed more reliably and quickly under appropriate conditions.

[0156] (Item 2) The X-ray device described in item 1 further includes an irradiation position detection unit (37) that constantly detects the position of the X-ray irradiation field relative to the subject (M), and an erroneous analysis detection unit (73) that, when the content of the region output by the learning model (51) has changed and the irradiation position detection unit (37) has detected a displacement of the X-ray irradiation field, selects the region output from the learning model (51) after the change as the target region, thereby reading out the image acquisition conditions (63) from the condition reading unit (49), and when the content of the region (61) output by the learning model (51) has changed and the irradiation position detection unit (37) has not detected a displacement of the X-ray irradiation field, selects the region output from the learning model (51) before the change, thereby reading out the image acquisition conditions (63) from the condition reading unit (49).

[0157] According to the X-ray apparatus 1B described in paragraph 2, the misanalysis detection unit 73 compares the position of the X-ray irradiation field relative to the subject M with the content of the region information 61 output by the learning model 51, and detects whether or not a misanalysis has occurred in the learning model 51. If the content of the region information 61 output by the learning model 51 has changed and the X-ray irradiation field has been displaced, it can be determined that the change in the region information 61 output by the learning model 51 is not due to a misanalysis by the learning model 51. Therefore, the misanalysis detection unit 73 selects the region information 61 after the change and causes the condition reading unit 49 to read out the image acquisition conditions 63.

[0158] On the other hand, if the content of region information 61 output by learning model 51 changes and the X-ray irradiation field has not shifted, it can be determined that the change in region information 61 output by learning model 51 is due to an erroneous analysis by learning model 51. Therefore, erroneous analysis detection unit 73 selects region information 61 output before the change, rather than selecting region information 61 after the change, and causes condition reading unit 49 to read image acquisition conditions 63. By providing such an erroneous analysis detection unit 73, even if an erroneous analysis by learning model 51 occurs due to fluctuations such as noise, and the content of region information 61 output by learning model 51 changes, the content of region information 61 output due to the erroneous analysis can be reliably ignored. Therefore, it is possible to avoid a situation in which inappropriate image acquisition conditions 63 are read out due to an erroneous analysis by learning model 51.

[0159] (Item 3) The X-ray device described in item 1 or 2 further includes an approval instruction input unit (41) that allows an operator to input an instruction to approve the image acquisition conditions (63) read by the condition reading unit (49), and the control unit (39) is configured to perform at least one of the following controls when an instruction to approve the image acquisition conditions (63) is input by the approval instruction input unit (41): controlling the X-ray tube (5) so that X-rays are irradiated in accordance with the X-ray irradiation conditions (65) read by the condition reading unit (49), and controlling the image processing unit (33) so that an X-ray image is generated in accordance with the image processing conditions (67).

[0160] According to the X-ray device 1C described in paragraph 3, when the operator approves the image acquisition conditions 63 read out by the condition reading unit 49, the X-ray tube 5 or the image processing unit 33 is controlled in accordance with the contents of the image acquisition conditions 63. There are cases where the image acquisition conditions 63 linked to the body part information 61 output by the learning model 51 are not actually appropriate as conditions for acquiring an X-ray image of the subject, for example, because the subject's physique is not within a typical range. In this case, the operator can avoid a situation where image acquisition conditions 63 that are not actually optimal are automatically set as conditions for acquiring an X-ray image against the operator's will by not approving the image acquisition conditions 63 read out by the condition reading unit 49.

[0161] (Item 4) The X-ray device (1A) according to this embodiment includes an X-ray tube (5) that irradiates X-rays onto a subject M, an X-ray detector (7) that detects X-rays that have passed through the subject M, an image processing unit (33) that generates an X-ray image by performing image processing using a detection signal output by the X-ray detector M, a condition storage unit (57) that stores image acquisition conditions (63) including at least one of X-ray irradiation conditions (65) and image processing conditions (67) corresponding to each of examination items (71) on the subject M in association with the examination items (71), and a learning model storage unit (51A) that stores a learning model (51A) that infers and outputs the type of examination in an examination image by performing machine learning using an examination image showing a human body as a teacher image. 55), an examination type inference unit (47) that inputs at least one of an X-ray image (F) and an optical image (D) of the subject into a learning model (51A) as an input image, thereby inferring and outputting the type of examination for the input image, a condition reading unit (49) that selects the type of examination output by the examination type inference unit (47) as an examination item (71), thereby reading out image acquisition conditions (63) that are linked to the examination item (71) and stored in the condition storage unit (57), and a control unit (39) that controls at least one of the X-ray tube (5) and the image processing unit (33) in accordance with the image acquisition conditions (63) read out by the condition reading unit (49).

[0162] According to the X-ray device 1A described in paragraph 4, the learning model 51A infers the type of examination in an examination image, and the APR 53A, to which appropriate image acquisition conditions 63 are linked according to examination item information 71 related to the type of examination, are used to automatically set image acquisition conditions 63 including at least one of X-ray irradiation conditions 65 and image processing conditions 67. The learning model 51 is configured to infer and output the type of examination in the examination image through machine learning using an examination image depicting a human body as a training image. That is, the image analysis unit 47 inputs an image of the subject M as an input image to the learning model 51, and infers and outputs the type of examination in the input image. The condition reading unit 49 selects the output type of examination of the subject M as examination item information 71, thereby automatically reading out the image acquisition conditions 63 linked to the examination item information 71 and stored. Therefore, when an X-ray image F of the subject is acquired, the image analysis unit 47 and the condition reading unit 49 automatically read out the image acquisition conditions 63 appropriate for the examination item of the X-ray image F. Therefore, even if the examination item information 71 for the subject M is changed, the image acquisition conditions 63 appropriate for the changed examination item information 71 are automatically read out. In other words, the process of the operator manually selecting the target region and setting the image acquisition conditions 63 is no longer necessary, so that X-ray fluoroscopy or X-ray imaging can be performed more reliably and quickly under appropriate conditions.

[0163] (Item 5) In the X-ray device according to item 4, an irradiation position detection unit (37) for detecting the position of the X-ray irradiation field relative to the subject M at any time; and an erroneous analysis detection unit (73) that, when the content of the inspection item (71) output by the learning model (51) has changed and the irradiation position detection unit (37) has detected a displacement of the X-ray irradiation field, causes the condition reading unit (49) to read out the image acquisition condition (63) by selecting the inspection item (71) output after the change, and when the content of the inspection item (71) output by the learning model (51) has changed and the irradiation position detection unit (37) has not detected a displacement of the X-ray irradiation field, causes the condition reading unit (49) to read out the image acquisition condition (63) by selecting the inspection item (71) output before the change.

[0164] According to the X-ray apparatus 1B described in paragraph 5, the erroneous analysis detection unit 73 compares the position of the X-ray irradiation field relative to the subject M with the contents of the examination item information 71 output by the learning model 51, and detects whether or not an erroneous analysis has occurred in the learning model 51. If the contents of the examination item information 71 output by the learning model 51 have changed and the X-ray irradiation field has been displaced, it can be determined that the change in the examination item information 71 output by the learning model 51 is not due to an erroneous analysis by the learning model 51. Therefore, the erroneous analysis detection unit 73 selects the examination item information 71 after the change and causes the condition reading unit 49 to read out the image acquisition conditions 63.

[0165] On the other hand, if the content of the examination item information 71 output by the learning model 51 changes and the X-ray irradiation field does not shift, it can be determined that the change in the examination item information 71 output by the learning model 51 is due to an erroneous analysis by the learning model 51. Therefore, the erroneous analysis detection unit 73 selects the examination item information 71 output before the change, rather than selecting the examination item information 71 after the change, and causes the condition reading unit 49 to read out the image acquisition conditions 63. By providing such an erroneous analysis detection unit 73, even if an erroneous analysis by the learning model 51 occurs due to fluctuations such as noise, and the content of the examination item information 71 output by the learning model 51 changes, the content of the examination item information 71 output due to the erroneous analysis can be reliably ignored. Therefore, it is possible to avoid a situation in which inappropriate image acquisition conditions 63 are read out due to an erroneous analysis by the learning model 51.

[0166] (Item 6) The X-ray device described in either item 4 or 5 further includes an approval instruction input unit (41) that allows an operator to input an instruction to approve the image acquisition conditions (63) read out by the condition reading unit (49), and the control unit (39) is configured to perform at least one of the following controls when an instruction to approve the image acquisition conditions 63 is input by the approval instruction input unit (41): controlling the X-ray tube 5 so that X-rays are irradiated in accordance with the X-ray irradiation conditions (65) read out by the condition reading unit 49; and controlling the image processing unit (33) so that an X-ray image F is generated in accordance with the image processing conditions (67).

[0167] According to the X-ray device 1C described in paragraph 6, when the operator approves the image acquisition conditions 63 read out by the condition reading unit 49, the X-ray tube 5 or the image processing unit 33 is controlled in accordance with the contents of the image acquisition conditions 63. There are cases where the image acquisition conditions 63 linked to the body part information 61 output by the learning model 51 are not actually appropriate as conditions for acquiring an X-ray image of the subject, for example, because the subject's physique is not within a typical range. In this case, the operator can avoid a situation where image acquisition conditions 63 that are not actually optimal are automatically set as conditions for acquiring an X-ray image against the operator's will by not approving the image acquisition conditions 63 read out by the condition reading unit 49.

[0168] <Other embodiments> It should be noted that the embodiments disclosed herein are illustrative in all respects and are not limiting. The scope of the present invention includes the claims and all modifications within the meaning and scope of the claims. For example, the present invention can be modified as follows:

[0169] (1) In each of the above-described embodiments, the image acquisition conditions 63 are not limited to a configuration including both the X-ray irradiation conditions 65 and the image processing conditions 67, but may be a configuration including either one of them. Furthermore, the X-ray irradiation conditions 65 are not limited to a configuration including both the X-ray fluoroscopy conditions 68 and the X-ray imaging conditions 69, but may be a configuration including either one of them.

[0170] (2) In each of the above-described embodiments, the X-ray image F used as input information to the learning model 51 may be an X-ray fluoroscopic image or an X-ray radiographic image. Furthermore, X-ray radiography may be performed using the X-ray fluoroscopic image as input information and the image acquisition conditions 63 read out by the learning model 51 and the APR 53. Furthermore, X-ray fluoroscopy may be performed using the X-ray radiographic image as input information and the image acquisition conditions 63 read out by the learning model 51 and the APR 53.

[0171] (3) In each of the above-described embodiments, an X-ray fluoroscopic imaging device equipped with a C-arm 9 is used as an example of the X-ray device 1, but this is not limited thereto, and the configuration of the present invention can be applied to any radiation imaging device, such as an X-ray imaging device for general X-ray imaging or a tomography imaging device.

[0172] (4) In the above-described third embodiment, a configuration is exemplified in which the displacement of the X-ray irradiation field is detected based on the position of the C-arm 9 detected by the arm position detection unit 37. However, the configuration is not limited to one in which the displacement of the X-ray irradiation field is detected by the arm position detection unit 37. For example, if the X-ray irradiation field is displaced due to horizontal movement of the tabletop 3, a configuration for identifying the position of the tabletop 3 is required to detect the displacement of the X-ray irradiation field.

[0173] (5) In the above-described fourth embodiment, the approval condition is not limited to a change in the information output by the learning model 51. Other examples of the approval condition include a change in the position of the irradiation field relative to the subject, or the S / N ratio of the most recently obtained X-ray image being equal to or less than a predetermined threshold. A condition for a change in the position of the irradiation field includes a change in the position of the C-arm 9 or the tabletop 3.

[0174] (6) In the above-described fourth embodiment, the approval key GK is displayed on the display unit 35 together with the X-ray image, but this is not limiting. If the display unit 35 has multiple monitors, the X-ray image and analysis results may be displayed on one monitor, and the approval key GK may be displayed on another monitor. Alternatively, a monitor or touch panel that displays the approval key GK may be disposed on the operation console 41, and the approval key GK may be displayed on the other monitor.

[0175] (7) In the above-described fourth embodiment, the configuration for inputting an instruction to approve the image acquisition conditions 63 is not limited to the configuration for operating the approval key GK displayed on the display unit 35 or the like, and other configurations, such as a switch or a button, may be used as appropriate. Furthermore, the method by which the condition setting control unit 79 controls the condition reading unit 49 when an approval condition occurs is not limited to the configuration for inhibiting the condition reading unit 49 from transmitting the image acquisition conditions 63 to the X-ray tube 5 or the like using the inhibition signal ST. The configuration controlled by the condition setting control unit 79 may be changed as appropriate as long as the control is configured such that, when an approval condition occurs, input of an instruction to approve the latest image acquisition conditions 63 is used as a trigger to set the latest image acquisition conditions 63 as the conditions for X-ray image acquisition. One example is a configuration in which, when an instruction to approve the latest image acquisition conditions 63 is input by the operator, the condition setting control unit 79 transmits a signal to the condition reading unit 49 instructing the latest image acquisition conditions 63 to be transmitted to the X-ray tube 5 or the like. [Explanation of symbols]

[0176] 1...X-ray device 3. Top plate 5...X-ray tube 7...X-ray detector 9...C-arm 17...Collimator 19...Optical camera 21...Foot switch 33...Image processing section 35...Display section 37...Arm position detection unit 39...Main control unit 41...Operation console 43...Storage section 45...Machine Learning Department 47...Image analysis section 49...Condition reading section 51...Learning Model 53 …APR 55...Learning model memory section 57…Condition storage section 61 …part information 63...Image acquisition conditions 65...X-ray irradiation conditions 67...Image processing conditions 68…X-ray fluoroscopy conditions 69... X-ray conditions 71...Inspection item information 73 ... Misanalysis detection unit 75 ... Approval condition determination section 77 ... Approval key display control section 79...Condition setting control section

Claims

1. an X-ray tube that irradiates an object with X-rays; an X-ray detector that detects X-rays that have passed through the subject; an image processing unit that generates an X-ray image by performing image processing using the detection signal output by the X-ray detector; a condition storage unit that stores image acquisition conditions, including at least one of X-ray irradiation conditions and image processing conditions, corresponding to each of target regions in the subject, in association with the target regions; a learning model storage unit that stores a learning model that infers and outputs a part of the human body shown in an image by performing machine learning using an image of the human body as a teacher image; a target region inference unit that inputs at least one of the X-ray image and the optical image of the subject as an input image into the learning model, thereby inferring and outputting a region shown in the input image; a condition reading unit that reads out the image acquisition conditions stored in the condition storage unit in association with the target portion by selecting the portion output by the target portion inference unit as the target portion; a control unit that controls at least one of the X-ray tube and the image processing unit in accordance with the image acquisition conditions read by the condition reading unit; an irradiation position detection unit that constantly detects the position of the X-ray irradiation field with respect to the subject; an erroneous analysis detection unit that, when the content of the region output by the learning model has changed and the irradiation position detection unit has detected a displacement of the X-ray irradiation field, selects the region output after the change as the target region, thereby causing the condition reading unit to read out the image acquisition conditions; and, when the content of the region output by the learning model has changed and the irradiation position detection unit has not detected a displacement of the X-ray irradiation field, selects the region output before the change as the target region, thereby causing the condition reading unit to read out the image acquisition conditions; An X-ray device comprising:

2. In the X-ray device according to claim 1, an approval instruction input unit that inputs an instruction from an operator to approve the image acquisition conditions read by the condition reading unit; The control unit When an instruction to approve the image acquisition conditions is input by the approval instruction input unit, the control unit is configured to perform at least one of the following controls: control the X-ray tube so that X-rays are irradiated in accordance with the X-ray irradiation conditions read by the condition reading unit; and control the image processing unit so that the X-ray image is generated in accordance with the image processing conditions. X-ray equipment.

3. an X-ray tube that irradiates an object with X-rays; an X-ray detector that detects X-rays that have passed through the subject; an image processing unit that generates an X-ray image by performing image processing using the detection signal output by the X-ray detector; a condition storage unit that stores image acquisition conditions, including at least one of X-ray irradiation conditions and image processing conditions, corresponding to each of the examination items of the subject, in association with the examination items; a learning model storage unit that stores a learning model that infers and outputs the type of examination in an examination image by performing machine learning using an examination image showing a human body as a teacher image; an examination type inference unit that inputs at least one of the X-ray image and the optical image of the subject into the learning model as an input image, and infers and outputs the type of examination in the input image based on information including the presence or absence of an examination device or a testing agent shown in the input image; a condition reading unit that reads out the image acquisition conditions stored in the condition storage unit in association with the examination item by selecting the examination type output by the examination type inference unit as the examination item; a control unit that controls at least one of the X-ray tube and the image processing unit in accordance with the image acquisition conditions read by the condition reading unit; An X-ray device comprising:

4. An X-ray tube that irradiates an object with X-rays; an X-ray detector that detects X-rays that have passed through the subject; an image processing unit that generates an X-ray image by performing image processing using the detection signal output by the X-ray detector; a condition storage unit that stores image acquisition conditions, including at least one of X-ray irradiation conditions and image processing conditions, corresponding to each of the examination items of the subject, in association with the examination items; a learning model storage unit that stores a learning model that infers and outputs the type of examination in an examination image by performing machine learning using an examination image showing a human body as a teacher image; an examination type inference unit that inputs at least one of the X-ray image and the optical image of the subject as an input image into the learning model, thereby inferring and outputting the type of examination for the input image; a condition reading unit that reads out the image acquisition conditions stored in the condition storage unit in association with the examination item by selecting the examination type output by the examination type inference unit as the examination item; a control unit that controls at least one of the X-ray tube and the image processing unit in accordance with the image acquisition conditions read by the condition reading unit; an irradiation position detection unit that constantly detects the position of the X-ray irradiation field with respect to the subject; an erroneous analysis detection unit that, when the content of the type of examination output by the learning model has changed and the irradiation position detection unit has detected a displacement of the X-ray irradiation field, selects the type of examination output after the change as the examination item, thereby reading out the image acquisition conditions from the condition reading unit; and, when the content of the type of examination output by the learning model has changed and the irradiation position detection unit has not detected a displacement of the X-ray irradiation field, selects the type of examination output before the change as the examination item, thereby reading out the image acquisition conditions from the condition reading unit; An X-ray device comprising:

5. In the X-ray apparatus according to claim 3 or claim 4, an approval instruction input unit that inputs an instruction from an operator to approve the image acquisition conditions read by the condition reading unit; The control unit When an instruction to approve the image acquisition conditions is input by the approval instruction input unit, the control unit is configured to perform at least one of the following controls: control the X-ray tube so that X-rays are irradiated in accordance with the X-ray irradiation conditions read by the condition reading unit; and control the image processing unit so that the X-ray image is generated in accordance with the image processing conditions. X-ray equipment.

Images