Computed tomography image generation method and learning model generation method

Abstract

Provided is a computer tomographic image generation method and a learning model generation method capable of reducing blurring due to rotation of an object in tomographic image data. The computer tomographic image generation method includes: a step of acquiring a plurality of rotational projection image data (30) by performing X-ray imaging while rotating an object (90); a step of acquiring tomographic image data (31) by performing reconstruction processing based on the acquired plurality of rotational projection image data (30); and a step of inputting the tomographic image data (31) to a model (40a) as input image data (44), thereby acquiring corrected tomographic image data (32) in which blurring due to rotation of the object (90) in the tomographic image data (31) is corrected as output image data (45).

Description

Technical Field

[0001] This invention relates to a method for generating computed tomography images and a method for generating learning models. Background Technology

[0002] Previously, methods for generating CT (Computed Tomography) images were known (see, for example, Patent Document 1 and Patent Document 2).

[0003] Patent document 1 disclosed a method for generating CT images using an industrial CT scanner for non-destructive testing. The industrial CT scanner includes an X-ray tube, a detector, a rotary table for placing a subject and rotating it via a rotating mechanism, and a CPU (Central Processing Unit). The X-ray tube irradiates the subject placed on the rotary table with X-rays. The detector detects the X-rays irradiated from the X-ray tube. The detector acquires projected image data of the subject after it has rotated one revolution via the rotating mechanism. The CPU generates a reconstructed image (CT image) based on the projected image data acquired by the detector.

[0004] Furthermore, Patent Document 2 discloses the application of a filter to suppress rotational blur when generating volumetric data. Patent Document 2 discloses that the low-pass filter in the data region can be modeled as a convolution of two filters: a Gaussian filter and a top-hat filter. The Gaussian filter is used to model the X-ray source and voxel blur, while the top-hat filter is used to model the rotational blur. Furthermore, Patent Document 2 discloses that the rotational blur is caused by gantry movement during the integration of the detection signal based on the data acquisition circuit.

[0005] [Existing Technical Documents]

[0006] [Patent Literature]

[0007] [Patent Document 1] Japanese Patent Application Publication No. 62-284250

[0008] [Patent Document 2] Japanese Patent Application Publication No. 2016-198504 Summary of the Invention

[0009] [Technical problem to be solved]

[0010] Although not explicitly described in Patent Document 1, it is common practice to irradiate the subject with X-rays from an X-ray tube while rotating the subject placed on a rotating stage, thereby acquiring multiple projection image data (rotational projection image data). This shortens the X-ray imaging time for acquiring projection image data compared to intermittently irradiating the subject with X-rays from the X-ray tube while stopping the rotation of the rotating stage at each imaging angle. However, in X-ray imaging where X-rays are irradiated while the subject is rotated, blurring due to subject rotation occurs in the acquired projection image data. If reconstruction processing is performed based on projection image data containing this blur, blurring due to subject rotation also appears in the acquired volume data and tomographic image data. Furthermore, in the method of Patent Document 2, the effect of reducing blurring due to subject rotation in tomographic image data is insufficient. Therefore, it is desirable to reduce blurring due to subject rotation in tomographic image data.

[0011] The present invention was made to solve the above-mentioned problems. One object of the present invention is to provide a method for generating computed tomographic images and a method for generating learning models that can reduce blurring caused by subject rotation in tomographic image data.

[0012] [Technical solution to the problem]

[0013] A method for generating computed tomography images, comprising:

[0014] The step of rotating the subject while taking X-ray images to obtain multiple rotational projection image data;

[0015] The steps for obtaining tomographic image data include: reconstructing multiple acquired rotational projection image data; and...

[0016] The steps involve inputting tomographic image data into the model as input image data, thereby obtaining corrected tomographic image data that corrects the blur caused by the rotation of the subject in the tomographic image data, and using this as output image data.

[0017] In addition, a method for generating computed tomography images includes:

[0018] The step of rotating the subject while taking X-ray images to obtain multiple rotational projection image data;

[0019] The steps involve performing multiple reconstruction processes based on the acquired multiple rotational projection image data, and simultaneously obtaining intermediate reconstructed image data through intermediate stage reconstruction processes.

[0020] The steps of using a model to obtain corrected reconstructed image data that corrects for blur caused by subject rotation from the acquired intermediate reconstructed image data; and

[0021] The step of further reconstructing the obtained corrected and reconstructed image data to obtain the final reconstructed image data.

[0022] The step of obtaining corrected and reconstructed image data is to input intermediate reconstructed image data into the model as input image data, thereby obtaining corrected and reconstructed image data that corrects the blur caused by the rotation of the subject in the intermediate reconstructed image data as output image data.

[0023] In addition, a method for generating a learning model includes:

[0024] The step of obtaining a training image dataset, wherein the training image dataset consists of input training data and output training data, wherein the input training data includes tomographic image data of the subject obtained by reconstructing multiple rotational projection image data, and the output training data includes tomographic image data that has been corrected for blurring caused by subject rotation; and

[0025] Based on the input training data and the output training data, the step of generating a learning model from the tomographic image data through machine learning, wherein the learning model outputs corrected tomographic image data that corrects the blurring caused by the rotation of the subject in the tomographic image data.

[0026] [Invention Effects]

[0027] In the first computed tomography (CT) image generation method described above, by inputting tomographic image data based on rotational projection image data obtained by rotating the subject while performing X-ray imaging as input image data to the model, corrected tomographic image data that corrects for blurring caused by subject rotation in the tomographic image data can be obtained as output image data. Therefore, even if blurring due to subject rotation occurs in the tomographic image data based on rotational projection image data, by inputting tomographic image data to the model, the model outputs corrected tomographic image data that corrects for blurring caused by subject rotation as output result. Thus, corrected tomographic image data that corrects for blurring caused by subject rotation can be obtained. Therefore, blurring caused by subject rotation can be reduced in tomographic image data at any tomographic plane.

[0028] Furthermore, in the second computed tomography image generation method described above, during the intermediate stage of the reconstruction process involving multiple reconstruction steps, corrected reconstructed image data that corrects for blurring caused by subject rotation is obtained from the acquired intermediate reconstructed image data using a model. This allows for reconstruction processing of the corrected reconstructed image data, which corrects for blurring caused by subject rotation, during the intermediate stage of the reconstruction process. Therefore, it is possible to obtain reconstructed image data with high-precision correction for blurring caused by subject rotation as the final reconstructed image data. Consequently, blurring caused by subject rotation can be reduced in tomographic image data at any cross-sectional plane.

[0029] Furthermore, in the method for generating the learning model, based on input training data containing tomographic image data of the subject and output training data containing tomographic image data that has corrected for blurring caused by subject rotation, machine learning can be used to learn image processing techniques that can correct for blurring caused by subject rotation. Therefore, it is possible to easily generate a learning model for corrected tomographic image data that outputs corrections for blurring caused by subject rotation in tomographic image data, based on the tomographic image data itself. Attached Figure Description

[0030] Figure 1 This is a schematic diagram showing the overall structure of the X-ray imaging apparatus according to the first embodiment.

[0031] Figure 2 It is a diagram representing rotation information image data according to the first embodiment.

[0032] Figure 3 This is a diagram used to illustrate the generation of the learning model according to the first embodiment and the acquisition of corrected tomographic image data using the learning model.

[0033] Figure 4 This is a diagram illustrating the control for acquiring corrected tomographic image data using a learning model according to the first embodiment.

[0034] Figure 5 This is a flowchart illustrating the processing of the CT image generation method according to the first embodiment.

[0035] Figure 6 This is a flowchart illustrating the process of generating a learning model according to the first embodiment.

[0036] Figure 7 This is a diagram illustrating the comparison results between corrected tomographic image data obtained by the CT image generation method of the first embodiment and corrected tomographic image data obtained by the CT image generation method of a modified example of the first embodiment.

[0037] Figure 8 This is a schematic diagram showing the overall structure of the X-ray imaging apparatus according to the second embodiment.

[0038] Figure 9 This is a diagram used to illustrate the generation of the learning model according to the second embodiment and the acquisition of corrected tomographic image data using the learning model.

[0039] Figure 10 This is a flowchart illustrating the processing of the CT image generation method according to the second embodiment.

[0040] Figure 11 This is a flowchart illustrating the process of generating the learning model according to the second embodiment.

[0041] Figure 12 This is a diagram illustrating the comparison results between the corrected tomographic image data obtained by the CT image generation method of the second embodiment and the corrected tomographic image data obtained by the CT image generation method of the modified example of the second embodiment.

[0042] Figure 13 This is a schematic diagram showing the overall structure of the X-ray imaging apparatus according to the third embodiment.

[0043] Figure 14 This is a diagram used to illustrate the reconstruction process according to the third embodiment.

[0044] Figure 15 This is a flowchart illustrating the processing of the CT image generation method according to the third embodiment.

[0045] Figure 16 This is a diagram used to illustrate the effect of the final reconstructed image data obtained by the CT image generation method of the third embodiment.

[0046] Explanation of icon numbers

[0047] 30 Rotational Projection Image Data

[0048] 31, 41a tomographic image data

[0049] 31c Intermediate Reconstructed Image Data

[0050] 31d final reconstructed image data

[0051] 32 Corrected tomographic image data

[0052] 33, 41b Rotation information image data

[0053] 35 Correcting and reconstructing image data

[0054] Learning models 40a and 40b (models)

[0055] 41 Input training data

[0056] 42 Output training data

[0057] 42a Tomographic image data corrected for blurring caused by subject rotation (correct image data)

[0058] 43 Training Image Dataset

[0059] 44 Input image data

[0060] 45 Output image data

[0061] 90 subjects Detailed Implementation

[0062] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings.

[0063] [First Implementation Method]

[0064] Reference Figure 1 This describes the overall structure of the X-ray imaging apparatus 100 according to the first embodiment.

[0065] like Figure 1 As shown, the X-ray imaging apparatus 100 is an apparatus for capturing X-ray images of a subject 90 and generating CT images. The X-ray imaging apparatus 100 of the first embodiment is used, for example, for non-destructive testing purposes. The subject 90 to be inspected is not particularly limited. The X-ray imaging apparatus 100 acquires rotational projection image data 30 (X-ray image data) of the subject 90 from the entire circumference of the subject placement section 3 where the subject 90 is placed, and constructs a tomographic image (CT image) based on the acquired rotational projection image data 30.

[0066] The X-ray imaging apparatus 100 includes an X-ray tube 1, a detector 2, a subject placement section 3, a rotation mechanism 4, and a control device 20. The X-ray tube 1 and the detector 2 constitute the imaging section 5 for capturing X-ray images.

[0067] X-ray tube 1 is configured to irradiate X-rays 99 onto a subject 90 disposed in subject placement section 3. Specifically, X-ray tube 1 continuously irradiates X-rays 99 onto the subject 90, which rotates due to the rotation of subject placement section 3 on which the subject 90 is placed. X-ray tube 1 is configured to generate X-rays 99 by applying a high voltage. X-ray tube 1 is opposite to detector 2 via subject placement section 3. X-ray tube 1, subject placement section 3, and detector 2 are arranged side by side in the horizontal direction.

[0068] Detector 2 is configured to detect X-rays 99 emitted from X-ray tube 1. The X-rays 99 emitted from X-ray tube 1 pass through the subject 90 and incident on the detection surface of detector 2. Detector 2 is configured to convert the detected X-rays 99 into an electrical signal. Thus, an X-ray image reflecting the transmission of X-rays 99 through the subject 90 is obtained. Detector 2 is, for example, an FPD (Flat Panel Detector). Detector 2 consists of multiple conversion elements (not shown) and pixel electrodes (not shown) disposed on the multiple conversion elements. The multiple conversion elements and pixel electrodes are arranged in a matrix within the detection surface at a predetermined period (pixel spacing). The detection signal (image signal) of detector 2 is sent to image processing unit 23.

[0069] The subject placement unit 3 is disposed between the X-ray tube 1 and the detector 2 and is configured to place the subject 90. The subject placement unit 3 consists of a subject stage for placing the subject 90.

[0070] The rotation mechanism 4 rotates one of the imaging section 5, which includes the X-ray tube 1 and the detector 2, and the subject placement section 3. The rotation mechanism 4 rotates either the imaging section 5 or the subject placement section 3 about a rotation axis 4a. In the first embodiment, the rotation mechanism 4 rotates the subject placement section 3 in a horizontal plane about the rotation axis 4a. The rotation mechanism 4 does not rotate the imaging section 5. The rotation axis 4a passes through the subject placement section 3 and is vertical. The rotation axis 4a is orthogonal to a straight line (representative line of the X-ray beam) from the X-ray tube 1 through the subject 90 on the subject placement section 3 toward the detector 2. The rotation mechanism 4 includes a motor (not shown) and a reducer (not shown) for rotating the subject placement section 3.

[0071] During continuous irradiation of the X-ray tube 1 with X-rays 99, the rotating mechanism 4 does not stop the rotation of the subject placement section 3 on which the subject 90 is placed. That is, the rotating mechanism 4 rotates the subject placement section 3 on which the subject 90 is placed and continuously irradiates the X-ray tube 1 with X-rays 99, thereby taking an X-ray image of the subject 90 while the subject 90 is rotating.

[0072] The control device 20 includes a control unit 21, a storage unit 24, and an input / output unit 25. The control device 20 may be configured as, for example, a PC (personal computer). The control device 20 is connected to a display device 26 and an input device 27.

[0073] The control unit 21 is a computer that includes a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), ROM (Read Only Memory), and RAM (Random Access Memory). The control unit 21 executes a predetermined program 70 via the CPU to perform predetermined control. The control unit 21 includes a capture control unit 22 and an image processing unit 23 as its functional structure. That is, the control unit 21 executes the predetermined program 70 via the CPU, thereby functioning as the capture control unit 22 and the image processing unit 23.

[0074] The imaging control unit 22 executes the program 70 stored in the storage unit 24 to set the imaging conditions in the X-ray imaging apparatus 100 and control the start and stop of X-ray imaging. That is, the imaging control unit 22 controls the operation of the X-ray tube 1. Furthermore, the imaging control unit 22 controls the operation of the rotating mechanism 4.

[0075] The image processing unit 23 acquires multiple rotational projection image data 30 from the detector 2. The image processing unit 23 generates multiple rotational projection image data 30 based on the detection signal (image signal) from the detector 2. As described above, the subject 90 is subjected to X-ray imaging while being rotated by the imaging unit 5. The image processing unit 23 acquires multiple rotational projection image data 30 based on the detection signal (image signal) acquired by the detector 2 while the subject 90 is rotated during X-ray imaging.

[0076] The image processing unit 23 acquires multiple rotational projection image data 30 from the detector 2 at various angles of multiple shooting angles, wherein the multiple shooting angles are set based on the rotation speed and frame rate (frames per second) of the subject placement unit 3 of the rotating mechanism 4. The rotational projection image data 30 is data of the X-ray image acquired at each shooting angle. The acquisition of the rotational projection image data 30 at each shooting angle is performed within a predetermined angle range. The predetermined angle range is 360 degrees (1 revolution). Furthermore, the number of rotational projection image data 30 acquired is based on a predetermined frame rate. In addition, the predetermined angle range is not limited to 360 degrees (1 revolution), and is not particularly limited as long as it is 180 degrees (half a revolution) or more.

[0077] The image processing unit 23 performs reconstruction processing based on the acquired multiple rotational projection image data 30 to obtain tomographic image data 31. The tomographic image data 31 can be tomographic image data 31 obtained by cutting the rotational projection image data 30 at any position. In this embodiment, preferably, the tomographic image data 31 is a tomographic image cut in a direction perpendicular to the rotation axis of the subject 90. Because the blurring caused by the rotation of the subject 90 in such a tomographic image is particularly large, the blur correction effect of the present invention is significant. Alternatively, multiple tomographic image data 31 can be stacked to form 3D volume data. The image processing unit 23 performs reconstruction processing on a set of rotational projection image data 30 for each shooting angle in 360 degrees (called a projection dataset) to generate the tomographic image data 31.

[0078] Image processing unit 23, as an example, performs reconstruction processing using an iterative approximation method. However, the reconstruction processing is not limited to using an iterative approximation method; known reconstruction processes can be implemented. For example, the reconstruction processing can be an analytical method using the FDK (Feldkamp-Davis-Kress) method, or it can be a reconstruction processing based on other analytical methods besides the FDK method.

[0079] Image processing unit 23 acquires rotation information image data 33 (see reference) Figure 2 Rotation information image data 33 is image data that reflects the rotation information of the subject 90 when acquiring multiple rotation projection image data 30. Further details of rotation information image data 33 will be described later.

[0080] like Figure 3 As shown, the image processing unit 23 inputs tomographic image data 31 as input image data 44 to the learning model 40a stored in the storage unit 24, thereby obtaining corrected tomographic image data 32 as output image data 45. The corrected tomographic image data 32 is image data that corrects the blurring in the tomographic image data 31 caused by the rotation of the subject 90. In the first embodiment, the image processing unit 23 inputs tomographic image data 31 and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining corrected tomographic image data 32, which corrects the blurring in the tomographic image data 31 caused by the rotation of the subject 90, as output image data 45. That is, the image processing unit 23 inputs tomographic image data 31 and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining corrected tomographic image data 32 as the output result (inference result). Details of the learning model 40a will be described later. Furthermore, the learning model 40a is an example of the "model" in the claims.

[0081] like Figure 1As shown, the storage unit 24 is configured to include both volatile and non-volatile storage devices. The storage unit 24 stores a program 70, various setting information (not shown) related to X-ray image acquisition by the X-ray imaging apparatus 100, rotational information image data 33, and a learning model 40a, etc. Furthermore, the storage unit 24 stores multiple acquired rotational projection image data 30, tomographic image data 31 generated based on the rotational projection image data 30, and corrected tomographic image data 32 generated using the learning model 40a.

[0082] The input / output unit 25 is composed of various interfaces for inputting and outputting signals to the control device 20. The input / output unit 25 is connected to the display device 26 and the input device 27. The display device 26 is, for example, a liquid crystal display. The input device 27 includes a keyboard and a mouse. The image processing unit 23 acquires the detection signal (image signal) from the detector 2 via the input / output unit 25.

[0083] (Rotational information image data)

[0084] Reference Figure 2 The image processing unit 23 acquires rotation information image data 33 that reflects the rotation information of the subject 90. That is, the image processing unit 23 generates rotation information image data 33 that reflects the rotation information of the subject 90.

[0085] Specifically, the rotation information image data 33 is image data that reflects the rotation information of the subject 90 when acquiring multiple rotational projection image data 30 for each pixel of the reconstructed tomographic image data 31. The rotation information of the subject 90 includes the rotation speed and rotation angle of the subject 90. That is, the rotation information image data 33 is motion data representing the amount of movement of each pixel in the tomographic image data 31, based on the rotation speed of the subject 90 as rotation information of the subject 90 and the rotation angle of the subject 90 among the multiple rotational projection image data 30.

[0086] The rotational speed of the subject 90 is the rotational speed of the subject 90 when acquiring multiple rotational projection image data 30. That is, the rotational speed of the subject 90 is the amount of rotation per unit time of the subject placement section 3 caused by the rotation mechanism 4 when acquiring multiple rotational projection image data 30. The rotational speed of the subject placement section 3 caused by the rotation mechanism 4 can be set before acquiring multiple rotational projection image data 30, or it can be set as a fixed value of the X-ray imaging device 100.

[0087] The rotation angle of the subject 90 is based on the frame rate set before acquiring multiple rotational projection image data 30, and is the rotation angle of the subject 90 between multiple rotational projection image data 30. That is, the rotation angle of the subject 90 is the rotation angle of the subject placement part 3 about the rotation axis 4a between the acquired rotational projection image data 30 and the next rotational projection image data 30 to be acquired.

[0088] Here, when acquiring rotational projection image data 30 while continuously irradiating a rotating subject 90 with X-rays 99, blurring (jitter of the subject 90) occurs in the acquired rotational projection image data 30 due to the rotation of the subject 90. That is, the movement of the subject 90 during the acquisition of one rotational projection image data 30 is manifested as blurring (jitter of the subject 90) in the acquired rotational projection image data 30. Then, if reconstruction processing is performed based on the rotational projection image data 30 containing the blurring (jitter of the subject 90) caused by the rotation of the subject 90, blurring (jitter of the subject 90) will also occur in the tomographic image data 31 on any cross-section obtained by the reconstruction processing. In other words, in the tomographic image data 31 obtained by the reconstruction processing, part of the subject will also be blurred due to the rotation of the subject 90.

[0089] Rotational information image data 33 represents the amount of movement of the subject 90 in each pixel of the tomographic image data 31 obtained through reconstruction processing. That is, rotational information image data 33 represents the amount of movement of each pixel per unit time. In rotational information image data 33, darker colors (darker areas, near-black areas) indicate smaller pixel movement, while lighter colors (brighter areas, near-white areas) indicate larger pixel movement.

[0090] The central portion of the rotation information image data 33 is the rotation center of the subject 90, so the amount of pixel movement is almost negligible. That is, the central portion of the rotation information image data 33 is darker in color. Conversely, in the rotation information image data 33, the amount of pixel movement increases as the portion moving away from the central portion approaches the outer periphery of the subject 90. That is, in the rotation information image data 33, the portion moving away from the central portion and approaching the outer periphery of the subject 90 is lighter in color. Furthermore, the portion of the rotation information image data 33 corresponding to the outer part of the subject 90 has no pixel movement due to no rotation, and is therefore black.

[0091] The rotation information image data 33 represents the amount of movement of the subject 90 in each pixel of the tomographic image data 31 obtained through reconstruction processing, but does not include the direction of movement of the subject 90 in each pixel of the tomographic image data 31 obtained through reconstruction processing. However, since the central portion of the rotation information image data 33 is the rotation center of the subject 90, the direction of movement of the subject 90 in each pixel is automatically determined based on the rotation direction of the subject 90.

[0092] The rotational speed of the subject 90 and the rotational angle of the subject 90 relative to the plurality of rotational projection image data 30 are set before the start of X-ray imaging by the imaging unit 5 for acquiring the plurality of rotational projection image data 30. The image processing unit 23 acquires rotational information image data 33 based on the set rotational speed of the subject 90 and the rotational angle of the subject 90 relative to the plurality of rotational projection image data 30. Each time X-ray imaging is performed on the subject 90, the image processing unit 23 acquires rotational information image data 33 based on the set rotational speed of the subject 90 and the rotational angle of the subject 90 relative to the plurality of rotational projection image data 30.

[0093] Furthermore, the acquisition of rotational information image data 33 by image processing unit 23 can be performed at any time, as long as it occurs before the acquisition of corrected tomographic image data 32 using learning model 40a. For example, the acquisition of rotational information image data 33 by image processing unit 23 can occur before the start of X-ray imaging by imaging unit 5, after the acquisition of multiple rotational projection image data 30 by image processing unit 23, or after the acquisition of tomographic image data 31 by performing reconstruction processing based on image processing unit 23.

[0094] (Learning models and methods for generating learning models)

[0095] like Figure 3 As shown, the learning model 40a takes the tomographic image data 31 and the rotation information image data 33 as input image data 44, and outputs the corrected tomographic image data 32, which corrects the blur caused by the rotation of the subject 90 in the tomographic image data 31, as output image data 45.

[0096] The methods for generating learning model 40a include:

[0097] The step of obtaining a training image dataset 43 consisting of input training data 41 and output training data 42, wherein the input training data 41 includes tomographic image data 41a of the subject 90 obtained by reconstruction processing based on multiple rotation projection image data 30, and the output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blurring caused by rotation of the subject 90; and

[0098] Based on the input training data 41 and the output training data 42, the step of generating a learning model 40a through machine learning based on the tomographic image data 31, wherein the learning model 40a outputs corrected tomographic image data 32 that corrects the blur caused by the rotation of the subject 90 in the tomographic image data 31.

[0099] That is, in the first embodiment, the method for generating the learning model 40a includes:

[0100] The step of obtaining a training image dataset 43 consisting of input training data 41 and output training data 42, wherein the input training data 41 includes tomographic image data 41a and rotation information image data 41b of the subject 90 obtained by reconstruction processing based on multiple rotation projection image data 30; and the output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blurring caused by the rotation of the subject 90; and

[0101] Based on the input training data 41 and the output training data 42, the step of generating a learning model 40a through machine learning based on the tomographic image data 31, wherein the learning model 40a outputs corrected tomographic image data 32 that corrects the blur caused by the rotation of the subject 90 in the tomographic image data 31.

[0102] The tomographic image data 41a in the input training data 41 contains multiple tomographic image data 41a. These multiple tomographic image data 41a contain blur caused by the rotation of the subject 90°. Furthermore, the rotation information image data 41b in the input training data 41 contains multiple rotation information image data 41b corresponding to each of the multiple tomographic image data 41a in the input training data 41. Additionally, the tomographic image data 42a (correct image data) in the output training data 42, which corrects for blur caused by the rotation of the subject 90°, contains multiple tomographic image data 42a (correct image data) corresponding to each of the multiple tomographic image data 41a in the input training data 41, correcting for blur caused by the rotation of the subject 90°.

[0103] The learning model 40a is generated by machine learning from the input training data 41 and the output training data 42. The input training data 41 includes tomographic image data 41a and rotation information image data 41b of the subject 90 obtained by reconstruction processing based on multiple rotation projection image data 30. The output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blurring caused by the rotation of the subject 90.

[0104] The learning model 40a is pre-generated by a learning device 200 separate from the X-ray imaging apparatus 100. The learning device 200 is, for example, a machine learning computer including a CPU, GPU, ROM, and RAM. The learning device 200 is located externally to the X-ray imaging apparatus 100. Alternatively, the learning device 200 may be located within the X-ray imaging apparatus 100.

[0105] The tomographic image data 41a in the input training data 41 can be tomographic image data 31 acquired by the image processing unit 23 of the X-ray imaging apparatus 100 and stored in the storage unit 24, or it can be tomographic image data 31 stored in the storage unit 24 of another X-ray imaging apparatus 100 or the learning device 200. Similarly, the rotation information image data 41b in the input training data 41 can be rotation information image data 33 acquired by the image processing unit 23 of the X-ray imaging apparatus 100 and stored in the storage unit 24, or it can be rotation information image data 33 stored in the storage unit 24 of another X-ray imaging apparatus 100 or the learning device 200. The rotation information image data 41b in the input training data 41 can also be acquired (generated) by the learning device 200 to correspond to the tomographic image data 41a.

[0106] Furthermore, the output training data 42 is tomographic image data 42a (correct image data) that has been corrected for blurring caused by the rotation of the subject 90. However, the output training data 42 can also be tomographic image data of the subject 90 generated based on the 3D CAD (Computer Aided Design) data of the subject 90, or it can be tomographic image data of the actual cut-off subject 90. The output training data 42 can be stored in the storage unit 24 of the X-ray imaging device 100, or it can be stored in the storage unit 24 of another X-ray imaging device 100 or in the learning device 200.

[0107] The learning device 200 takes input training data 41, which includes tomographic image data 41a of the subject 90 and rotation information image data 41b of the subject 90, as input, and output training data 42, which includes tomographic image data 42a (correct image data) that corrects the blur caused by the rotation of the subject 90, as output. It then uses machine learning to generate a learning model 40a. That is, the learning device 200 uses a training image dataset 43, consisting of input training data 41 and output training data 42, as training data (training set), and uses machine learning to learn the learning model 40a.

[0108] In the first embodiment, the machine learning method for the learning model 40a uses U-Net++, a type of fully convolutional network (FCN). However, the machine learning method is not limited to U-Net++; any method such as U-Net, neural networks, support vector machines (SVM), or boosting can be used. The created learning model 40a is stored in the storage unit 24 of the X-ray imaging apparatus 100 via a network (not shown) or via a recording medium such as flash memory.

[0109] (Control for acquiring corrected tomographic image data using a learning model)

[0110] Regarding the acquisition of corrected tomographic image data 32 using the learning model 40a, the image processing unit 23 does not perform the operation if the rotational speed of the subject 90 in the rotational information image data 33 is less than a predetermined threshold, and performs the operation if the rotational speed of the subject 90 in the rotational information image data 33 is greater than or equal to the predetermined threshold. In other words, regarding the acquisition of corrected tomographic image data 32 using the learning model 40a, when acquiring rotational information image data 33, the image processing unit 23 does not perform the operation if the rotational speed of the acquired subject placement unit 3 is less than a predetermined threshold, and performs the operation if the rotational speed of the acquired subject placement unit 3 is greater than or equal to the predetermined threshold.

[0111] Furthermore, if the rotation speed of the acquired subject 90 (the rotation speed of the subject placement unit 3) is less than a predetermined threshold, the image processing unit 23 may not acquire rotation information image data 33. That is, if the rotation speed of the acquired subject 90 (the rotation speed of the subject placement unit 3) is less than a predetermined threshold, the image processing unit 23 may not generate rotation information image data 33.

[0112] Figure 4 The image above is an example of tomographic image data 42a (correct image data) corrected for blur caused by a 90° rotation of the subject. Furthermore, Figure 4 The middle image is an example of multiple tomographic image data 31 input to the learning model 40a, each with a different rotational speed (the rotation angle of the subject 90 among multiple rotational projection image data 30 corresponding to the rotational speed). Furthermore, Figure 4 The image below is an example of corrected tomographic image data 32, output from learning model 40a, which corrects for blurring caused by the 90° rotation of the subject.

[0113] like Figure 4As shown, it can be seen that when the rotation speed is greater than or equal to a predetermined threshold (i.e., when the rotation angle is greater than or equal to 0.75 degrees), the correction effect on the blur caused by the rotation of the subject 90 increases. Therefore, the image processing unit 23 does not perform the acquisition of corrected tomographic image data 32 using the learning model 40a when the rotation speed of the subject 90 is less than the predetermined threshold (i.e., when the rotation angle is less than 0.75 degrees), but performs the acquisition when the rotation speed of the subject 90 in the rotation information image data 33 is greater than or equal to the predetermined threshold (i.e., when the rotation angle is greater than or equal to 0.75 degrees). In addition, the predetermined threshold (rotation angle) in the rotation speed of the subject 90 is not limited to the rotation speed when the rotation angle is 0.75 degrees, and can be set appropriately.

[0114] Furthermore, when the rotational speed of the subject 90 is less than a predetermined threshold, the image processing unit 23 can perform known filtering processing on the tomographic image data 31 without using the learning model 40a to acquire the corrected tomographic image data 32. This filtering process reduces processing time and processing burden compared to the processing of outputting the corrected tomographic image data 32 using the learning model 40a. For example, a smoothing filter can be used as a known filtering process.

[0115] (CT image generation method)

[0116] Next, refer to Figure 5 This section explains the CT image generation method performed by the control unit 21 in the first embodiment. Furthermore, the order of the processing steps can be interchanged or performed simultaneously, provided they do not contradict each other.

[0117] In step S1, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by X-ray imaging performed by the imaging unit 5 while the subject 90 is rotated. Then, the processing proceeds to step S2.

[0118] In step S2, the image processing unit 23 performs reconstruction processing based on the acquired multiple rotational projection image data 30 to obtain tomographic image data 31. Then, the processing proceeds to step S3.

[0119] In step S3, the image processing unit 23 acquires rotation information image data 33, which reflects the rotation information of the subject 90 when acquiring multiple rotation projection image data 30. Then, the processing proceeds to step S4.

[0120] In step S4, the image processing unit 23 determines whether the rotation speed of the subject 90 in the rotation information image data 33 is greater than or equal to a predetermined threshold. If the rotation speed of the subject 90 in the rotation information image data 33 is greater than or equal to the predetermined threshold ("Yes" in step S4), then proceed to step S5; if the rotation speed of the subject 90 in the rotation information image data 33 is less than the predetermined threshold ("No" in step S4), then proceed to step S7.

[0121] In step S5, the image processing unit 23 inputs tomographic image data 31 and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining corrected tomographic image data 32, which corrects the blurring caused by the rotation of the subject 90 in the tomographic image data 31, as output image data 45. Then, the processing proceeds to step S6.

[0122] In step S6, the image processing unit 23 stores the acquired corrected tomographic image data 32 in the storage unit 24. Afterward, the processing ends.

[0123] In step S7, the image processing unit 23 does not use the learning model 40a to acquire the corrected tomographic image data 32, but instead stores the acquired tomographic image data 31 in the storage unit 24. Afterwards, the processing ends.

[0124] (Methods for generating learning models)

[0125] Next, refer to Figure 6 This section explains the method for generating the learning model 40a by the learning device 200 in the first embodiment. Furthermore, the order of the processing steps can be interchanged or performed simultaneously, provided they do not contradict each other.

[0126] In step S11, the learning device 200 acquires a training image dataset 43 consisting of input training data 41 and output training data 42. The input training data 41 includes tomographic image data 41a of the subject 90 obtained by reconstruction processing based on multiple rotational projection image data 30, and rotational information image data 41b corresponding to the tomographic image data 41a of the subject 90. The output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blurring caused by the rotation of the subject 90. Afterward, the process proceeds to step S12.

[0127] In step S12, the learning device 200, based on the input training data 41 and the output training data 42, generates a learning model 40a for the corrected tomographic image data 32, which corrects the blurring caused by the rotation of the subject 90 in the tomographic image data 31, through machine learning. The processing then proceeds to step S13.

[0128] In step S13, the X-ray imaging apparatus 100 stores the learning model 40a generated by the learning device 200 in the storage unit 24 of the X-ray imaging apparatus 100. Afterwards, the process ends.

[0129] (Comparison between the first embodiment and its variations)

[0130] Reference Figure 3 as well as Figure 7 (a) ~ Figure 7 (d) illustrates the comparison results between the corrected tomographic image data 32 (CT image) obtained by the CT image generation method of the first embodiment and the corrected tomographic image data 32a (CT image) obtained by the CT image generation method of the modified example of the first embodiment.

[0131] The CT image generation method in the variation of the first embodiment and Figure 3 The CT image generation method of the first embodiment shown differs from that of the first embodiment. Instead of inputting rotation information image data 33, only tomographic image data 31 is input to the learning model as input image data 44, thereby obtaining corrected tomographic image data 32a (see reference). Figure 7 (d) is used as output image data 45. Furthermore, the learning model in the variation of the first embodiment and... Figure 3 The CT image generation method of the first embodiment shown differs from that of the first embodiment. It is generated through machine learning based on a training image dataset 43 consisting of input training data 41 (containing tomographic image data 41a of the subject 90 but not rotational information image data 41b) and output training data 42 (correct image data) of tomographic image data 42a that corrects for blurring caused by rotation of the subject 90). Furthermore, the learning model in the variation of the first embodiment is an example of the "model" used in the claims.

[0132] The subject 90 in the CT image generation method of the first embodiment is the same subject 90 as the subject 90 in the CT image generation method of the modified example of the first embodiment. The subject 90 is a cylindrical sample made of resin, and the interior of the subject 90 contains material or gaps with a low X-ray absorption coefficient 99.

[0133] Figure 7 The upper part of (a) is tomographic image data 42a (correct image data) of the first embodiment and its variations, which corrects for blur caused by the rotation of the subject 90. Figure 7 The following diagram is (a) Figure 7 A magnified view of part A in the upper part of (a). Figure 7The upper figure of (b) is based on multiple rotational projection image data 30 in the first embodiment and its variations (see reference). Figure 3 Tomographic image data 31 obtained by reconstructing the image (refer to) Figure 3 ), Figure 7 The image below (b) is Figure 7 (b) is an enlarged view of part B in the upper part of the diagram. Figure 7 The upper part of (c) is the corrected tomographic image data 32 of the first embodiment (see reference). Figure 3 ), Figure 7 The following diagram is (c). Figure 7 The enlarged view of part C in the upper part of (c). Figure 7 The upper part of (d) is the corrected tomographic image data 32a of a modified example of the first embodiment. Figure 7 The following diagram is (d). Figure 7 The enlarged view of part D in the upper part of (d).

[0134] Compare Figure 7 The following diagram of (c) and Figure 7 The following diagram (d) confirms that... Figure 7 The corrected tomographic image data 32 of the first embodiment shown in Figure (c) below is compared to Figure 7 The corrected tomographic image data 32a of the modified example of the first embodiment shown in Figure (d) better corrects the blurring caused by the rotation of the subject 90. Therefore, it can be confirmed that the CT image generation method according to the first embodiment can better reduce the blurring caused by the rotation of the subject 90 in tomographic image data of any cut surface than the CT image generation method of the modified example of the first embodiment.

[0135] [Second Implementation]

[0136] Next, refer to Figures 8-12 This section describes the CT image generation method and the generation method of the learning model 40b according to the second embodiment. In the second embodiment, an example is described where multiple tomographic image data 31 with varying noise levels are input to the learning model 40b as input image data 44, thereby obtaining corrected tomographic image data 32 as output image data 45. In the second embodiment, the same reference numerals are used for structures identical to those in the first embodiment described above, and descriptions are omitted. Furthermore, the learning model 40b is an example of the term "model" used in the claims.

[0137] like Figure 8As shown, in the second embodiment, the image processing unit 23 acquires multiple tomographic image data 31 with different noise levels during the reconstruction processing based on the acquired multiple rotational projection image data 30. In the second embodiment, for example, the image processing unit 23 acquires two types of tomographic image data 31 with different smoothness levels during the reconstruction processing based on the acquired multiple rotational projection image data 30.

[0138] Taking two types of tomographic image data 31 with different smoothness levels as an example, in the reconstruction process, the acquired tomographic image data 31 is processed by two smoothing filters with different smoothing intensities. As smoothing filters, known smoothing filters such as Gaussian filters can be used.

[0139] The smoothing intensity can be increased or decreased by changing the kernel size (filter size) of the filter function used. One of two smoothing filters with different smoothing intensities is the first smoothing filter, which increases the smoothing intensity by increasing the kernel size. Increasing the smoothing intensity makes pixel value changes between pixels smoother. The other of the two smoothing filters with different smoothing intensities is the second smoothing filter, which decreases the smoothing intensity compared to the first smoothing filter by decreasing the kernel size. Furthermore, smoothing filters are not limited to Gaussian filters; for example, they can also be low-pass filters.

[0140] In the reconstruction process, the image processing unit 23 performs smoothing processing on the acquired tomographic image data 31 based on a first smoothing filter, thereby obtaining first tomographic image data 31a with a high smoothing intensity. Furthermore, in the reconstruction process, the image processing unit 23 performs smoothing processing on the acquired tomographic image data 31 based on a second smoothing filter, thereby obtaining second tomographic image data 31b with a low smoothing intensity.

[0141] like Figure 9As shown, the image processing unit 23 inputs multiple tomographic image data 31 with different noise levels to the learning model 40b stored in the storage unit 24 as input image data 44, thereby obtaining corrected tomographic image data 32 that corrects the blur caused by the rotation of the subject 90 in the tomographic image data 31 with different noise levels, and uses it as output image data 45. In the second embodiment, the image processing unit 23 inputs multiple tomographic image data 31 with different smoothness levels to the learning model 40b as input image data 44, thereby obtaining corrected tomographic image data 32 that corrects the blur caused by the rotation of the subject 90 in the tomographic image data 31 with different smoothness levels, and uses it as output image data 45. Specifically, the image processing unit 23 inputs a first tomographic image data 31a with a high smoothing intensity and a second tomographic image data 31b with a low smoothing intensity into the learning model 40b as input image data 44, thereby obtaining corrected tomographic image data 32 that corrects the blur caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b, and uses it as output image data 45.

[0142] That is, the image processing unit 23 inputs the first tomographic image data 31a and the second tomographic image data 31b into the learning model 40b as input image data 44, thereby obtaining the corrected tomographic image data 32 as the output result (inference result).

[0143] Furthermore, in the second embodiment, the multiple tomographic image data 31 with varying noise levels may, for example, consist of three or more tomographic image data 31 with varying smoothness. In this case, the image processing unit 23 inputs the three or more tomographic image data 31 with varying smoothness as input image data 44 to the learning model 40b, thereby obtaining corrected tomographic image data 32 as output image data 45. Additionally, in the second embodiment, the multiple tomographic image data 31 with varying noise levels may, for example, consist of multiple tomographic image data 31 with varying noise levels obtained by changing the reconstruction parameters in the reconstruction process.

[0144] (Learning models and methods for generating learning models)

[0145] The learning model 40b takes multiple tomographic image data 31 with varying noise levels as input image data 44, and outputs corrected tomographic image data 32, which corrects the blurring caused by the subject's rotation 90° in the tomographic image data 31 with varying noise levels, as output image data 45. In the second embodiment, the learning model 40b takes first tomographic image data 31a and second tomographic image data 31b as input image data 44, and outputs corrected tomographic image data 32 as output image data 45.

[0146] The methods for generating learning model 40b include:

[0147] The step of obtaining a training image dataset 43 consisting of input training data 41 and output training data 42, wherein the input training data 41 includes multiple tomographic image data (41c, 41d) of a subject 90 with varying noise levels, obtained in reconstruction processing based on multiple rotationally projected image data 30; and the output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blurring caused by the rotation of the subject 90; and

[0148] Based on the input training data 41 and the output training data 42, the step of generating a learning model 40b for correcting the blur caused by the rotation of the subject 90 in the tomographic image data 31 by means of machine learning is to generate a learning model 40b for the tomographic image data 32 that corrects the blur caused by the rotation of the subject 90 in the tomographic image data 31.

[0149] That is, in the second embodiment, the method for generating the learning model 40b includes:

[0150] The step of obtaining a training image dataset 43 consisting of input training data 41 and output training data 42, wherein the input training data 41 includes a first tomographic image data 41c with high smoothing intensity and a second tomographic image data 41d with low smoothing intensity, obtained by reconstruction processing based on multiple rotationally projected image data 30; and the output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blur caused by the rotation of the subject 90.

[0151] Based on input training data 41 and output training data 42, the step of generating a learning model 40b for corrected tomographic image data 32 through machine learning, based on the first tomographic image data 31a and the second tomographic image data 31b, is to correct the blur caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b.

[0152] The first tomographic image data 41c and the second tomographic image data 41d in the input training data 41 comprise a set of multiple first tomographic image data 41c and second tomographic image data 41d. The multiple first tomographic image data 41c and second tomographic image data 41d contain blur caused by the rotation of the subject 90°. Furthermore, the tomographic image data 42a (correct image data) in the output training data 42, which corrects for blur caused by the rotation of the subject 90°, comprises multiple tomographic image data 42a (correct image data) corresponding to each of the multiple sets of first tomographic image data 41c and second tomographic image data 41d in the input training data 41, correcting for blur caused by the rotation of the subject 90°.

[0153] The learning model 40b is generated by machine learning from the input training data 41 and the output training data 42. The input training data 41 includes a first tomographic image data 41c with high smoothing intensity and a second tomographic image data 41d with low smoothing intensity, which are obtained by reconstruction processing based on multiple rotational projection image data 30. The output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blur caused by the rotation of the subject 90.

[0154] The first tomographic image data 41c and the second tomographic image data 41d in the input training data 41 can be the first tomographic image data 31a and the second tomographic image data 31b acquired by the image processing unit 23 of the X-ray imaging device 100 and stored in the storage unit 24, or they can be the first tomographic image data 31a and the second tomographic image data 31b stored in the storage unit 24 of another X-ray imaging device 100 or the learning device 200.

[0155] The learning device 200 takes input training data 41, which includes first tomographic image data 41c and second tomographic image data 41d of the subject 90, as input, and output training data 42, which includes tomographic image data 42a (correct image data) that corrects the blur caused by the rotation of the subject 90, as output, and learns through machine learning to generate a learning model 40b.

[0156] (CT image generation method)

[0157] Next, refer to Figure 10 The CT image generation method performed by the control unit 21 in the second embodiment will be explained. Furthermore, the order of the processing steps can be interchanged or performed simultaneously, as long as they do not contradict each other.

[0158] In step S21, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by X-ray imaging performed by the imaging unit 5 while the subject 90 is rotated. Then, the processing proceeds to step S22.

[0159] In step S22, the image processing unit 23 acquires first tomographic image data 31a and second tomographic image data 31b, which are two types of tomographic image data 31 with different smoothness, in the reconstruction processing based on the acquired multiple rotational projection image data 30. After that, the processing proceeds to step S23.

[0160] In step S23, the image processing unit 23 inputs the first tomographic image data 31a and the second tomographic image data 31b as input image data 44 to the learning model 40b, thereby obtaining corrected tomographic image data 32, which corrects the blurring caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b, as output image data 45. Then, the processing proceeds to step S24.

[0161] In step S24, the image processing unit 23 stores the acquired corrected tomographic image data 32 in the storage unit 24. Afterward, the processing ends.

[0162] (Methods for generating learning models)

[0163] Next, refer to Figure 11 This section explains the method for generating the learning model 40b by the learning device 200 in the second embodiment. Furthermore, the order of the processing steps can be interchanged or performed simultaneously, provided they do not contradict each other.

[0164] In step S31, the learning device 200 acquires a training image dataset 43 consisting of input training data 41 and output training data 42. The input training data 41 includes a first tomographic image data 41c with high smoothing intensity and a second tomographic image data 41d with low smoothing intensity, acquired in the reconstruction processing based on multiple rotationally projected image data 30. The output training data 42 includes tomographic image data 42a (correct image data) that has been corrected for blurring caused by the rotation of the subject 90. The processing then proceeds to step S32.

[0165] In step S32, the learning device 200, based on the input training data 41 and the output training data 42, generates a learning model 40b for corrected tomographic image data 32 by machine learning, based on the first tomographic image data 31a and the second tomographic image data 31b. This model corrects the blurring caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b. The process then proceeds to step S33.

[0166] In step S33, the X-ray imaging apparatus 100 stores the learning model 40b generated by the learning device 200 in the storage unit 24 of the X-ray imaging apparatus 100. Afterwards, the process ends.

[0167] (Comparison between the second embodiment and its variations)

[0168] Reference Figure 9 as well as Figure 12 (a) ~ Figure 12 (e) illustrates the comparison results between the corrected tomographic image data 32 (CT image) obtained by the CT image generation method of the second embodiment and the corrected tomographic image data 32b (CT image) obtained by the CT image generation method of the modified example of the second embodiment.

[0169] The CT image generation method in the variation of the second embodiment is the same as Figure 9 The CT image generation method of the second embodiment shown differs from that of the second embodiment. As input image data 44, only the first tomographic image data 31a is input to the learning model, without inputting the second tomographic image data 31b, thereby obtaining the corrected tomographic image data 32b (see reference). Figure 12 (e) is used as output image data 45. Furthermore, the learning model in the variation of the second embodiment is... Figure 9 The CT image generation method of the second embodiment shown differs from that of the second embodiment. It is generated through machine learning based on a training image dataset 43 consisting of input training data 41 that includes first tomographic image data 41 but not second tomographic image data 41d, and output training data 42 that corrects the blurring caused by the rotation of the subject 90° (correct image data). Furthermore, the learning model in the variation of the second embodiment is an example of the "model" used in the claims.

[0170] The subject 90 in the CT image generation method of the second embodiment is the same subject 90 as the subject 90 in the CT image generation method of the modified example of the second embodiment. The subject 90 is a cylindrical sample made of resin, and the interior of the subject 90 contains material or gaps with a low X-ray absorption coefficient 99.

[0171] Figure 12 The upper image of (a) is tomographic image data 42a (correct image data) of the second embodiment and its variations, after correcting for blur caused by the 90° rotation of the subject (see reference). Figure 9 ), Figure 12 The following diagram is (a) Figure 12 A magnified view of part E in the upper part of (a). Figure 12The upper part of (b) is the first tomographic image data 31a with high smoothing intensity in the second embodiment and its variations (see reference). Figure 9 ), Figure 12 The image below (b) is Figure 12 (b) is an enlarged view of part F in the upper part of the diagram. Figure 12 The upper part of (c) is the second tomographic image data 31b with low smoothing intensity in the second embodiment (see reference). Figure 9 ), Figure 12 The following diagram is (c). Figure 12 The enlarged view of part G in the upper part of (c). Figure 12 The upper part of (d) is the corrected tomographic image data 32 of the second embodiment (see reference). Figure 9 ), Figure 12 The following diagram is (d). Figure 12 The enlarged view of part H in the upper part of (d). Figure 12 The upper part of (e) is the corrected tomographic image data 32b of a modified example of the second embodiment. Figure 12 The image below (e) is Figure 12 An enlarged view of part I in the upper part of (e).

[0172] Compare Figure 12 The following diagram of (d) and Figure 12 The image below (e) confirms that... Figure 12 The corrected tomographic image data 32 of the second embodiment shown in Figure (d) below is compared to Figure 12 The corrected tomographic image data 32b of the modified example of the second embodiment shown in Figure (e) better corrects the blurring caused by the subject's 90° rotation. Furthermore, it can be confirmed that... Figure 12 The corrected tomographic image data 32 of the second embodiment shown in Figure (d) below is compared to Figure 12 The corrected tomographic image data 32b of the modified example of the second embodiment shown in Figure (e) extracts the feature points 91 of the subject 90 more accurately. Therefore, it can be confirmed that the CT image generation method according to the second embodiment, compared with the CT image generation method of the modified example of the second embodiment, can reduce the blurring caused by the rotation of the subject 90 in the tomographic image data on any cross-section, and at the same time extract the feature points 91 of the subject 90 more accurately.

[0173] [Third Implementation Method]

[0174] Next, refer to Figures 13-16This section describes a CT image generation method according to a third embodiment. In this third embodiment, an example is described where, during the reconstruction process, intermediate reconstructed image data 31c is obtained through an intermediate-stage reconstruction process. Corrected reconstructed image data 35 is then obtained from the obtained intermediate reconstructed image data 31c using a learning model 40a. Simultaneously, reconstruction processing is performed on the obtained corrected reconstructed image data 35 to obtain the final reconstructed image data 31d. In this third embodiment, the same reference numerals are used for structures identical to those in the first embodiment described above, and descriptions are omitted.

[0175] Image processing unit 23 performs reconstruction processing using an iterative approximation method. In the reconstruction processing using the iterative approximation method, as an example, calculation processing including forward projection, back projection, comparison and update is repeatedly performed in the intermediate stage. Specifically, in an example of reconstruction processing using the iterative approximation method, the following calculation processing (reconstruction processing) is repeatedly performed in the intermediate stage of reconstruction processing: (1) Calculate and generate the k-th projection (forward projection) from the k-th image (intermediate reconstructed image data 31c), (2) Calculate the ratio of the k-th forward projection to the measured projection, (3) Perform back projection on the calculated ratio, (4) Multiply the k-th image by the back-projected image and update it to the (k+1)-th image (intermediate reconstructed image data 31c).

[0176] In the reconstruction process, the image processing unit 23 performs multiple reconstruction processes based on the acquired multiple rotational projection image data 30, and simultaneously acquires first intermediate reconstructed image data 31c through intermediate stage reconstruction processing.

[0177] Furthermore, in the reconstruction process, the image processing unit 23 uses the learning model 40a to obtain the first corrected reconstructed image data 35 from the acquired first intermediate reconstructed image data 31c. Specifically, when acquiring the first corrected reconstructed image data 35, the image processing unit 23 inputs the first intermediate reconstructed image data 31c and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining the first corrected reconstructed image data 35, which corrects the blur caused by the rotation of the subject 90 in the first intermediate reconstructed image data 31c, as output image data 45.

[0178] Furthermore, in the reconstruction process, the image processing unit 23 performs reconstruction processing on the acquired first corrected reconstructed image data 35, thereby obtaining second intermediate reconstructed image data 31c. In the third embodiment, the image processing unit 23 performs calculation processing on the acquired first corrected reconstructed image data 35 in the reconstruction process based on the iterative approximation method, thereby obtaining second intermediate reconstructed image data 31c.

[0179] Furthermore, the image processing unit 23 inputs the second intermediate reconstructed image data 31c and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining the second corrected reconstructed image data 35 as output image data 45. Then, the image processing unit 23 performs calculation processing based on the iterative approximation method on the obtained second corrected reconstructed image data 35, thereby obtaining the final reconstructed image data 31d.

[0180] In the third embodiment, the learning model, as an example, is the same as the learning model 40a in the first embodiment described above. However, the learning model in the third embodiment is not limited to the same learning model 40a as in the first embodiment. For example, the learning model in the third embodiment could also be the learning model 40b in the second embodiment described above. When the learning model is the learning model 40a in the first embodiment, it is generated using the method for generating the learning model 40a described in the first embodiment. When the learning model is the learning model 40b in the second embodiment, it is generated using the method for generating the learning model 40b described in the second embodiment.

[0181] Furthermore, the learning model in the third embodiment is not limited to the learning model 40a in the first embodiment and the learning model 40b in the second embodiment. The learning model in the third embodiment can also be a learning model that constructs corrected reconstructed image data 35 from the intermediate reconstructed image data 31c by any of the following machine learning methods: supervised learning, unsupervised learning, and reinforcement learning. Furthermore, the machine learning method used in the learning model in the third embodiment is not particularly limited. Additionally, the learning model in the third embodiment can also be generative AI (Artificial Intelligence). Moreover, the learning model in the third embodiment is an example of the "model" used in the claims.

[0182] (CT image generation)

[0183] Next, refer to Figure 14 The CT image generation performed by the control unit 21 in the third embodiment will be explained. Specifically, the reconstruction process using the learning model 40a of the control unit 21 and the iterative approximation method will be explained. In addition, the image processing unit 23 performs the calculation process using the iterative approximation method n times.

[0184] like Figure 14 As shown, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by the detector 2 while rotating the subject 90 to perform X-ray imaging.

[0185] The image processing unit 23 performs the first calculation process in the reconstruction process using the iterative approximation method, thereby obtaining the first intermediate reconstructed image data 31c. Specifically, the image processing unit 23 performs forward projection, back projection, comparison, and update, thereby obtaining the first intermediate reconstructed image data 31c from the initial image data 34 as the first calculation process in the reconstruction process using the iterative approximation method.

[0186] The image processing unit 23 inputs the first intermediate reconstructed image data 31c and rotation information image data 33 into the learning model 40a as input image data 44, thereby obtaining the first corrected reconstructed image data 35 as output image data 45.

[0187] The image processing unit 23 performs forward projection, back projection, comparison, and update, thereby obtaining second intermediate reconstructed image data 31c from the first corrected reconstructed image data 35 as the second calculation process of the reconstruction process using the iterative approximation method.

[0188] The image processing unit 23 inputs the second intermediate reconstructed image data 31c and rotation information image data 33 into the learning model 40a as input image data 44, thereby obtaining the second corrected reconstructed image data 35 as output image data 45.

[0189] Then, the image processing unit 23 repeatedly performs the reconstruction processing using the iterative approximation method, and acquires the input image data 44 and output image data 45 of the learning model 40a. The image processing unit 23 repeatedly performs the reconstruction processing and acquires the output image data 45 using the learning model 40a a total of n-1 times.

[0190] The image processing unit 23 performs forward projection, back projection, comparison, and update as the nth calculation process of the reconstruction process using the iterative approximation method, thereby obtaining the nth intermediate reconstructed image data 31c from the (n-1)th corrected reconstructed image data 35, which is used as the final reconstructed image data 31d. The image processing unit 23 stores the final reconstructed image data 31d, which is the nth intermediate reconstructed image data 31c, in the storage unit 24.

[0191] (CT image generation method)

[0192] Next, refer to Figure 15 The CT image generation method performed by the control unit 21 in the third embodiment will be explained. Furthermore, the order of the processing steps can be interchanged or performed simultaneously, as long as they do not contradict each other. Additionally, the image processing unit 23 performs n calculations using the iterative approximation method for reconstruction processing.

[0193] In step S51, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by X-ray imaging performed by the imaging unit 5 while the subject 90 is rotated. Then, the processing proceeds to step S52.

[0194] In step S52, the image processing unit 23 performs the first calculation process in the reconstruction process using the iterative approximation method, thereby obtaining the first intermediate reconstructed image data 31c from the initial image data 34. After that, the process proceeds to step S53.

[0195] In step S53, the image processing unit 23 inputs the first intermediate reconstructed image data 31c and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining the first corrected reconstructed image data 35 as output image data 45. Then, the processing proceeds to step S54.

[0196] In step S54, the image processing unit 23 performs the second (n-1) calculation process in the reconstruction process using the iterative approximation method, thereby obtaining the second (n-1) intermediate reconstructed image data 31c from the first (n-2) corrected reconstructed image data 35. After that, the process proceeds to step S55.

[0197] In step S55, the image processing unit 23 inputs the second (n-1)th intermediate reconstructed image data 31c and rotation information image data 33 as input image data 44 to the learning model 40a, thereby obtaining the second (n-1)th corrected reconstructed image data 35 as output image data 45. After that, the processing proceeds to step S56.

[0198] In step S56, the image processing unit 23 determines whether the reconstruction processing using the iterative approximation method, and the acquisition of input and output image data 45 of the input image data 44 of the learning model 40a, have been repeated (repeatedly) n-1 times. If the reconstruction processing and the acquisition of output image data 45 using the learning model 40a have been repeated n-1 times ("Yes" in step S56), then proceed to step S57; if the reconstruction processing and the acquisition of output image data 45 using the learning model 40a have not been repeated n-1 times ("No" in step S56), then return to step S54.

[0199] In step S57, the image processing unit 23 performs forward projection, back projection, comparison, and update as the nth calculation process of the reconstruction process using the iterative approximation method, thereby obtaining the nth intermediate reconstructed image data 31c from the (n-1)th corrected reconstructed image data 35 as the final reconstructed image data 31d. After that, the process proceeds to step S58.

[0200] In step S58, the image processing unit 23 stores the final reconstructed image data 31d, which is the nth intermediate reconstructed image data 31c, in the storage unit 24. After that, the processing ends.

[0201] (The effect of the final reconstructed image data according to the third embodiment)

[0202] Reference Figure 3 , Figure 14 as well as Figure 16 (a) ~ Figure 16 (c) illustrates the effect of the final reconstructed image data 31d obtained by the CT image generation method of the third embodiment.

[0203] Figure 16 (b) is tomographic image data 42a (correct image data) of the third embodiment and its variations, which corrects for blurring caused by the rotation of the subject 90a. Furthermore, Figure 16 (c) is a partial enlarged view of the final reconstructed image data 31d, which is the nth intermediate reconstructed image data 31c, in the third embodiment.

[0204] It can be confirmed. Figure 16 The final reconstructed image data 31d, which is the nth intermediate reconstructed image data 31c, shown in (c) of the third embodiment, corrects for the blur caused by the rotation of the subject 90. Therefore, it can be confirmed that the CT image generation method according to the third embodiment can reduce the blur caused by the rotation of the subject 90 in tomographic image data at any cross-sectional plane.

[0205] [Variation Example]

[0206] Furthermore, the embodiments disclosed herein should be considered exemplary in all respects and not restrictive. The scope of the invention is defined not by the description of the above embodiments, but by the claims, and includes all modifications (variations) within the scope and equivalent meaning of the claims.

[0207] For example, in each of the first to third embodiments, unsupervised learning can be used instead of a learning model to output an image that corrects for blur caused by rotation.

[0208] In addition, in the first or second embodiment, multiple corrected tomographic image data can be used to acquire and output volume data that has been corrected for blur.

[0209] Alternatively, an implementation combining the first and second embodiments can be used. That is, by inputting multiple tomographic image data with varying noise levels and rotation information image data into the model, corrected tomographic image data that better corrects for blurring caused by subject rotation can be output.

[0210] Similarly, in the third embodiment, an embodiment combining the first and / or second embodiments can also be adopted. That is, in the third embodiment, (1) by inputting intermediate reconstructed image data and rotation information image data into the model, corrected reconstructed image data that corrects the blur caused by the rotation of the subject can be output; (2) by inputting multiple intermediate reconstructed image data with different noise levels into the model, corrected reconstructed image data that corrects the blur caused by the rotation of the subject can be output; (3) by inputting multiple intermediate reconstructed image data with different noise levels and rotation information image data into the model, corrected reconstructed image data that corrects the blur caused by the rotation of the subject can be output.

[0211] Furthermore, for example, in the first embodiment described above, it is also possible to acquire corrected tomographic image data even when the rotational speed of the subject in the rotational information image data is less than a predetermined threshold.

[0212] Alternatively, for example, it can be configured to acquire rotational projection image data while the rotating subject is pulsed with X-rays through an X-ray tube.

[0213] [aspect]

[0214] Those skilled in the art will understand that the above exemplary embodiments are specific examples of the following aspects.

[0215] (Project 1)

[0216] A method for generating CT images, comprising:

[0217] The step of rotating the subject while taking X-ray images to obtain multiple rotational projection image data;

[0218] The step of reconstructing the multiple rotational projection image data to obtain tomographic image data; and

[0219] The step of inputting the tomographic image data into the model as input image data, thereby obtaining corrected tomographic image data that corrects the blur caused by the rotation of the subject in the tomographic image data as output image data.

[0220] By inputting tomographic image data—based on rotational projection image data acquired while rotating the subject during X-ray imaging—as input image data to the model, corrected tomographic image data that corrects for blurring caused by subject rotation can be obtained as output image data. Therefore, even if blurring due to subject rotation occurs in tomographic image data based on rotational projection image data, by inputting tomographic image data to the model, the model outputs corrected tomographic image data that corrects for blurring caused by subject rotation. Thus, it is possible to obtain corrected tomographic image data that corrects for blurring caused by subject rotation. Therefore, blurring caused by subject rotation can be reduced in tomographic image data at any cross-sectional plane.

[0221] (Project 2)

[0222] A CT image generation method as described in Project 1 further includes:

[0223] The step of acquiring rotation information image data that reflects the rotation information of the subject when acquiring the plurality of rotation projection image data;

[0224] The step of acquiring the corrected tomographic image data involves inputting the tomographic image data and the rotation information image data into the model as input image data, thereby acquiring the corrected tomographic image data, which corrects the blurring caused by the rotation of the subject in the tomographic image data, as the output image data.

[0225] By inputting tomographic image data and rotation information image data reflecting the rotation of the subject into the model, corrected tomographic image data that corrects for blurring caused by subject rotation can be obtained as output image data. Therefore, compared to deconvolution (inverse convolution) or filtering of tomographic image data, corrected tomographic image data with better correction for blurring caused by subject rotation can be obtained. Thus, blurring caused by subject rotation can be reduced more effectively in tomographic image data at any tomographic plane.

[0226] (Project 3)

[0227] A CT image generation method as described in Project 2, wherein,

[0228] The step of acquiring the rotation information image data involves acquiring image data that reflects the rotation speed of the subject and the rotation angle of the subject between the multiple rotation projection image data, and using this data as the rotation information image data.

[0229] Regarding the known rotational information of the subject, the rotational speed of the subject and the rotational angle of the subject between multiple rotational projection image data are set before acquiring multiple rotational projection image data, or are set as fixed values ​​of the X-ray imaging device. Therefore, rotational information image data can be easily acquired based on the known rotational information of the subject.

[0230] (Project 4)

[0231] A CT image generation method as described in Project 3, wherein,

[0232] The step of acquiring the rotation information image data involves acquiring motion amount image data as the rotation information image data. The motion amount image data is based on the rotation speed of the subject and the rotation angle of the subject between the plurality of rotation projection image data, and represents the motion amount of each pixel in the tomographic image data.

[0233] Since the motion image data, representing the amount of movement of each pixel in the tomographic image data, is used as the input image to the model, it is possible to obtain the corrected tomographic image data, which corrects for the blur caused by the rotation of the subject, as the output image data with high precision.

[0234] (Project 5)

[0235] A CT image generation method as described in any one of items 2 to 4, wherein,

[0236] The step of acquiring the corrected tomographic image data is not performed if the rotational speed of the subject is less than a predetermined threshold, and is performed if the rotational speed of the subject is greater than or equal to the predetermined threshold.

[0237] When the blur caused by subject rotation is relatively small and the rotation speed of the subject is less than a specified threshold, the processing burden can be reduced by not performing the acquisition processing of the corrected tomographic image data using the model. However, when the blur caused by subject rotation is relatively large and the rotation speed of the subject is greater than or equal to the specified threshold, the corrected tomographic image data that effectively corrects the blur caused by subject rotation can be obtained by performing the acquisition processing of the corrected tomographic image data using the model.

[0238] (Project 6)

[0239] A CT image generation method as described in any one of items 1 to 5, wherein,

[0240] The step of acquiring the tomographic image data involves acquiring multiple tomographic image data with different levels of noise.

[0241] The step of obtaining the corrected tomographic image data involves inputting multiple tomographic image data with different noise levels into the model as input image data, thereby obtaining the corrected tomographic image data, which corrects the blur caused by the rotation of the subject in the tomographic image data, as the output image data.

[0242] By inputting multiple tomographic images with varying noise levels into the model, corrected tomographic images that correct for blurring caused by subject rotation can be obtained as output images. Therefore, compared to inputting a single tomographic image, corrected tomographic images with better correction for blurring caused by subject rotation can be obtained. Consequently, blurring caused by subject rotation can be reduced more effectively in tomographic images across any tomographic plane.

[0243] (Project 7)

[0244] A CT image generation method as described in Project 6, wherein,

[0245] The step of acquiring the tomographic image data involves acquiring multiple tomographic image data with different smoothness levels.

[0246] The step of obtaining the corrected tomographic image data involves inputting multiple tomographic image data with different smoothnesses into the model as input image data, thereby obtaining the corrected tomographic image data as the output image data.

[0247] By using multiple tomographic image data with varying smoothness as input to the learning image data, it is possible to obtain corrected tomographic image data as output image data that has been precisely corrected for blurring caused by subject rotation and has also precisely extracted the subject's feature points. Therefore, it is possible to reduce blurring caused by subject rotation in tomographic image data on any tomographic plane while simultaneously extracting the subject's feature points with high precision.

[0248] (Project 8)

[0249] A CT image generation method as described in any one of items 1 to 7, wherein,

[0250] The model described is a learning model;

[0251] The CT image generation method further includes the step of generating the learning model;

[0252] The step of generating the learning model is based on input training data and output training data. The learning model is generated by machine learning according to the tomographic image data. The input training data includes the tomographic image data, and the output training data includes tomographic image data that has been corrected for blurring caused by the rotation of the subject. The learning model outputs the corrected tomographic image data that has been corrected for blurring caused by the rotation of the subject.

[0253] By inputting tomographic image data into a learning model generated through machine learning, it is possible to easily obtain corrected tomographic image data, which corrects for blurring caused by subject rotation, as output image data.

[0254] (Project 9)

[0255] A method for generating CT images, comprising:

[0256] The step of rotating the subject while taking X-ray images to obtain multiple rotational projection image data;

[0257] The steps include performing multiple reconstruction processes based on the acquired multiple rotational projection image data, and obtaining intermediate reconstructed image data through the intermediate stage of the reconstruction process.

[0258] The steps of using a model to obtain corrected reconstructed image data from the acquired intermediate reconstructed image data, which corrects for blurring caused by subject rotation; and

[0259] The step of further reconstructing the obtained corrected and reconstructed image data to obtain the final reconstructed image data;

[0260] The step of obtaining the corrected and reconstructed image data involves inputting the intermediate reconstructed image data into the model as input image data, thereby obtaining the corrected and reconstructed image data, which corrects the blur caused by the rotation of the subject in the intermediate reconstructed image data, as output image data.

[0261] In the intermediate stage of the reconstruction process involving multiple reconstruction steps, corrected reconstructed image data, which corrects for blurring caused by subject rotation, is obtained from the acquired intermediate reconstructed image data using a model. This allows for reconstruction processing of the corrected reconstructed image data, which corrects for blurring caused by subject rotation, during the intermediate stage of the reconstruction process. Therefore, it is possible to obtain reconstructed image data with high-precision correction for blurring caused by subject rotation as the final reconstructed image data. Consequently, it is possible to better reduce blurring caused by subject rotation in tomographic image data at any cross-sectional plane.

[0262] (Project 10)

[0263] A CT image generation method as described in Project 9, wherein,

[0264] The step of obtaining the intermediate reconstructed image data is to obtain the intermediate reconstructed image data by performing calculations in the reconstruction process based on the iterative approximation method.

[0265] The step of obtaining the final reconstructed image data involves performing the calculation processing based on the iterative approximation method on the obtained corrected reconstructed image data, thereby obtaining the final reconstructed image data.

[0266] In reconstruction processes involving multiple iterations based on approximation methods, a model is used to extract corrected reconstructed image data from intermediate reconstructed image data obtained through computational processing. This corrected reconstructed image data, which corrects for blurring caused by subject rotation, is then used in subsequent iterations based on the approximation method. Therefore, reconstructed image data with high-precision correction for blurring caused by subject rotation can be easily obtained as the final reconstructed image data. Consequently, blurring caused by subject rotation can be more easily reduced in tomographic image data at any cross-sectional level.

[0267] (Project 11)

[0268] A method for generating a learning model, comprising:

[0269] The step of obtaining a training image dataset, wherein the training image dataset consists of input training data and output training data, wherein the input training data includes tomographic image data of the subject obtained by reconstructing multiple rotational projection image data, and the output training data includes the tomographic image data after correcting for blurring caused by the rotation of the subject; and

[0270] Based on the input training data and the output training data, a learning model is generated from the tomographic image data through machine learning. The learning model outputs corrected tomographic image data that corrects the blurring caused by the rotation of the subject in the tomographic image data.

[0271] Based on input training data containing tomographic image data of the subject and output training data containing tomographic image data that has corrected for blurring caused by subject rotation, this method can learn image processing techniques to correct for blurring caused by subject rotation through machine learning. Therefore, it is possible to easily generate a learning model for corrected tomographic image data that outputs corrections for blurring caused by subject rotation in tomographic image data, based on the tomographic image data itself.

[0272] (Project 12)

[0273] A method for generating a learning model as described in Project 11, wherein,

[0274] The step of obtaining the training image dataset is to obtain the training image dataset composed of the input training data and the output training data, wherein the input training data includes the tomographic image data of the subject and rotation information image data reflecting the rotation information of the subject.

[0275] Based on input and output training data containing tomographic image data and rotation information image data reflecting the subject's rotation, this system can learn image processing techniques through machine learning to correct blur caused by subject rotation. Therefore, it can easily generate a learning model for corrected tomographic image data that effectively corrects blur caused by subject rotation in the tomographic image data.

[0276] (Project 13)

[0277] A method for generating a learning model as described in Project 11, wherein,

[0278] The step of obtaining the training image dataset is to obtain the training image dataset, which consists of the input training data and the output training data, which contain multiple tomographic images of the subject with different levels of noise.

[0279] This invention enables the learning of image processing techniques, using machine learning to correct blur caused by subject rotation, based on input and output training data containing multiple tomographic images of a subject with varying levels of noise. Therefore, compared to learning image processing techniques using machine learning on input training data consisting of a single tomographic image, this invention provides a learning model for corrected tomographic image data that generates outputs that effectively correct blur caused by subject rotation in tomographic image data.

Claims

1. A method for generating computed tomography images, characterized in that, include: The step of rotating the subject while taking X-ray images to obtain multiple rotational projection image data; The step of reconstructing based on the acquired multiple rotational projection image data to obtain tomographic image data; as well as The step of inputting the tomographic image data into the model as input image data, thereby obtaining corrected tomographic image data that corrects the blur caused by the rotation of the subject in the tomographic image data as output image data.

2. The method for generating computed tomography images according to claim 1, characterized in that, It also includes the step of acquiring rotation information image data that reflects the rotation information of the subject when acquiring the plurality of rotation projection image data; The step of obtaining the corrected tomographic image data involves inputting the tomographic image data and the rotation information image data into the model as the input image data, thereby obtaining the corrected tomographic image data, which corrects the blurring caused by the rotation of the subject in the tomographic image data, as the output image data.

3. The method for generating computed tomography images according to claim 2, characterized in that, The step of acquiring the rotation information image data involves acquiring image data that reflects the rotation speed of the subject and the rotation angle of the subject between the multiple rotation projection image data, and using this data as the rotation information image data.

4. The method for generating computed tomography images according to claim 3, characterized in that, The step of acquiring the rotation information image data involves acquiring motion amount image data as the rotation information image data. The motion amount image data is based on the rotation speed of the subject and the rotation angle of the subject between the plurality of rotation projection image data, and represents the motion amount of each pixel in the tomographic image data.

5. The method for generating computed tomography images according to claim 2, characterized in that, The step of acquiring the corrected tomographic image data is not performed if the rotational speed of the subject is less than a predetermined threshold, and is performed if the rotational speed of the subject is greater than or equal to the predetermined threshold.

6. The method for generating computed tomography images according to claim 1 or 2, characterized in that, The step of acquiring the tomographic image data involves acquiring multiple tomographic image data with different levels of noise. The step of obtaining the corrected tomographic image data involves inputting multiple tomographic image data with different noise levels into the model as input image data, thereby obtaining the corrected tomographic image data, which corrects the blur caused by the rotation of the subject in the tomographic image data, as the output image data.

7. The method for generating computed tomography images according to claim 6, characterized in that, The step of acquiring the tomographic image data involves acquiring multiple tomographic image data with different smoothness levels. The step of obtaining the corrected tomographic image data involves inputting multiple tomographic image data with different smoothnesses into the model as input image data, thereby obtaining the corrected tomographic image data as the output image data.

8. The method for generating computed tomography images according to claim 1 or 2, characterized in that, The model described is a learning model; The computed tomography image generation method further includes the step of generating the learning model; The step of generating the learning model is based on input training data and output training data. The learning model is generated by machine learning according to the tomographic image data. The input training data includes the tomographic image data, and the output training data includes tomographic image data that has been corrected for blurring caused by the rotation of the subject. The learning model outputs the corrected tomographic image data that has been corrected for blurring caused by the rotation of the subject.

9. A method for generating computed tomography images, characterized in that, include: The step of rotating the subject while taking X-ray images to obtain multiple rotational projection image data; The steps include performing multiple reconstruction processes based on the acquired multiple rotational projection image data, and obtaining intermediate reconstructed image data through the intermediate stage of the reconstruction process. The steps of using a model to obtain corrected reconstructed image data from the acquired intermediate reconstructed image data, which corrects for blurring caused by subject rotation; and The step of further reconstructing the obtained corrected and reconstructed image data to obtain the final reconstructed image data; The step of obtaining the corrected and reconstructed image data involves inputting the intermediate reconstructed image data into the model as input image data, thereby obtaining the corrected and reconstructed image data, which corrects the blurring caused by the rotation of the subject in the intermediate reconstructed image data, as output image data.

10. The method for generating computed tomography images according to claim 9, characterized in that, The step of obtaining the intermediate reconstructed image data involves performing computational processing in the reconstruction process based on the iterative approximation method to obtain the intermediate reconstructed image data. The step of obtaining the final reconstructed image data involves performing the calculation processing based on the iterative approximation method on the obtained corrected reconstructed image data, thereby obtaining the final reconstructed image data.

11. A method for generating a learning model, characterized in that, include: The step of obtaining a training image dataset, wherein the training image dataset consists of input training data and output training data, wherein the input training data includes tomographic image data of the subject obtained by reconstructing multiple rotational projection image data, and the output training data includes the tomographic image data that has been corrected for blurring caused by rotation of the subject. as well as Based on the input training data and the output training data, a learning model is generated from the tomographic image data through machine learning. The learning model outputs corrected tomographic image data that corrects the blurring caused by the rotation of the subject in the tomographic image data.

12. The method for generating a learning model according to claim 11, characterized in that, The step of obtaining the training image dataset is to obtain the training image dataset consisting of the input training data and the output training data, wherein the input training data includes the tomographic image data of the subject and rotation information image data reflecting the rotation information of the subject.

13. The method for generating a learning model according to claim 11, characterized in that, The step of obtaining the training image dataset is to obtain the training image dataset, which consists of the input training data and the output training data, which contain multiple tomographic images of the subject with different levels of noise.

Images