Method for generating CT images and method for generating trained models

Abstract

This invention provides a method for generating CT images and a method for generating a trained model that can reduce blurring caused by rotation of the subject in tomographic image data. [Solution] This CT image generation method includes the steps of: acquiring multiple rotational projection image data 30 by X-ray imaging while rotating the subject 90; acquiring tomographic image data 31 by performing reconstruction processing based on the acquired multiple rotational projection image data 30; and inputting the tomographic image data 31 as input image data 44 to a model 40a to acquire corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected.

Description

Technical Field

[0001] The present invention relates to a method for generating a CT image and a method for generating a learned model.

Background Art

[0002] Conventionally, a method for generating a CT image has been known (see, for example, Patent Document 1 and Patent Document 2).

[0003] In Patent Document 1, a method for generating a CT image by an industrial CT (Computed Tomography) scanner used for non-destructive inspection applications is disclosed. The industrial CT scanner includes an X-ray tube, a detector, a rotary table on which a subject is placed and which can be rotated by a rotation mechanism, and a CPU (Central Processing Unit). The X-ray tube irradiates X-rays toward the subject placed on the rotary table. The detector detects the X-rays irradiated from the X-ray tube. The detector acquires projection image data of the subject rotated once by the rotation mechanism. The CPU generates a reconstructed image (CT image) based on the projection image data acquired by the detector. Further, in Patent Document 2, it is disclosed that a filter for suppressing rotational blur is applied when generating volume data. In Patent Document 2, it is disclosed that a low-pass filter in the data region is modeled as a convolution of two filters, a Gaussian filter and a TopHat filter. The Gaussian filter is used to model the X-ray source and voxel blur, and the TopHat filter is used to model rotational blur. Also, in Patent Document 2, it is disclosed that rotational blur is caused by gantry movement during the integration time of the detection signal by the data acquisition circuit.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

[0005] Although not explicitly stated in Patent Document 1, it is practiced to acquire multiple projection image data (rotational projection image data) by irradiating an object placed on a rotary table with X-rays from an X-ray tube while rotating the object. This shortens the X-ray imaging time required to acquire projection image data compared to intermittent X-ray imaging, where the rotation of the rotary table is stopped at each imaging angle and X-rays are irradiated from the X-ray tube towards the object. However, in X-ray imaging where X-rays are irradiated while the object is rotating, blurring due to the rotation of the object occurs in the acquired projection image data. When reconstruction processing is performed based on projection image data containing such blurring, blurring due to the rotation of the object also appears in the acquired volume data and tomographic image data. Furthermore, the method in Patent Document 2 is insufficient in reducing blurring due to the rotation of the object in tomographic image data. Therefore, it is desirable to reduce blurring due to the rotation of the object in tomographic image data.

[0006] This invention was made to solve the above-mentioned problems, and one of its objectives is to provide a CT image generation method and a trained model generation method that can reduce blurring caused by rotation of the subject in tomographic image data. [Means for solving the problem]

[0007] A CT image generation method comprising the steps of: acquiring multiple rotational projection image data by taking X-ray images while rotating the subject; acquiring tomographic image data by performing a reconstruction process based on the acquired multiple rotational projection image data; and inputting tomographic image data as input image data to a model to acquire corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected as output image data. The CT image generation method further comprises the steps of: acquiring multiple rotational projection image data by X-ray imaging while rotating the subject; performing multiple reconstruction processes based on the acquired rotational projection image data and acquiring intermediate reconstructed image data through reconstruction processing in the intermediate stage; using a model to acquire corrected reconstructed image data in which blurring caused by the rotation of the subject has been corrected from the acquired intermediate reconstructed image data; and further performing reconstruction processing using the acquired corrected reconstructed image data to acquire final reconstructed image data, wherein the step of acquiring corrected reconstructed image data involves inputting intermediate reconstructed image data as input image data to a model, thereby acquiring corrected reconstructed image data in which blurring caused by the rotation of the subject in the intermediate reconstructed image data has been corrected as output image data. The method for generating a trained model comprises the steps of: obtaining a training image dataset comprising input training data including tomographic image data of a subject obtained by performing a reconstruction process based on multiple rotational projection image data, and output training data including tomographic image data in which blurring caused by the rotation of the subject has been corrected; and generating a trained model by machine learning based on the input training data and the output training data, which outputs corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected from the tomographic image data. [Effects of the Invention]

[0008] In the first CT image generation method described above, tomographic image data based on rotational projection image data obtained by X-ray imaging while rotating the subject is input to the model as input image data, thereby obtaining corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data is corrected as output image data. Therefore, even if blurring caused by the rotation of the subject occurs in the tomographic image data based on rotational projection image data, inputting tomographic image data into the model results in corrected tomographic image data in which blurring caused by the rotation of the subject is corrected. Thus, corrected tomographic image data in which blurring caused by the rotation of the subject is corrected can be obtained. As a result, blurring caused by the rotation of the subject can be reduced in tomographic image data at any tomographic plane. Furthermore, in the second CT image generation method described above, in the intermediate stage of the reconstruction process, which involves multiple reconstruction processes, a model is used to obtain corrected reconstructed image data from the acquired intermediate reconstructed image data in which blurring caused by the rotation of the subject has been corrected. This allows the reconstruction process to be performed on the corrected reconstructed image data in which blurring caused by the rotation of the subject has been corrected in the intermediate stage of the reconstruction process. Therefore, it is possible to obtain reconstructed image data in which blurring caused by the rotation of the subject has been accurately corrected as the final reconstructed image data. As a result, blurring caused by the rotation of the subject can be reduced in tomographic image data at any tomographic plane. Furthermore, in the method for generating a trained model, based on input training data including tomographic image data of the subject and output training data including tomographic image data in which blurring caused by the subject's rotation has been corrected, the model can be trained by machine learning to perform image processing that corrects blurring caused by the subject's rotation. Therefore, a trained model can be easily generated that outputs corrected tomographic image data in which blurring caused by the subject's rotation has been corrected from tomographic image data. [Brief explanation of the drawing]

[0009] [Figure 1]This is a schematic diagram showing the overall configuration of an X-ray imaging apparatus according to the first embodiment. [Figure 2] This figure shows rotation information image data according to the first embodiment. [Figure 3] This figure illustrates the generation of a trained model according to the first embodiment and the acquisition of corrected tomographic image data using the trained model. [Figure 4] This figure illustrates the control of acquiring corrected tomographic image data using a trained model according to the first embodiment. [Figure 5] This is a flowchart illustrating the processing of the CT image generation method according to the first embodiment. [Figure 6] This is a flowchart illustrating the process for generating a trained model according to the first embodiment. [Figure 7] This figure illustrates the comparison results between corrected tomographic image data obtained by the CT image generation method in the first embodiment and corrected tomographic image data obtained by the CT image generation method in a modified version of the first embodiment. [Figure 8] This is a schematic diagram showing the overall configuration of an X-ray imaging apparatus according to the second embodiment. [Figure 9] This figure illustrates the generation of a trained model according to the second embodiment and the acquisition of corrected tomographic image data using the trained model. [Figure 10] This is a flowchart illustrating the processing of the CT image generation method according to the second embodiment. [Figure 11] This is a flowchart illustrating the process for generating a trained model according to the second embodiment. [Figure 12] This figure illustrates the comparison results between corrected tomographic image data obtained by the CT image generation method in the second embodiment and corrected tomographic image data obtained by the CT image generation method in a modified example of the second embodiment. [Figure 13] This is a schematic diagram showing the overall configuration of an X-ray imaging apparatus according to the third embodiment. [Figure 14]It is a diagram for explaining the reconstruction process according to the third embodiment. [Figure 15] It is a flowchart for explaining the process of the CT image generation method according to the third embodiment. [Figure 16] It is a diagram for explaining the effect of the final reconstructed image data obtained by the CT image generation method in the third embodiment.

Modes for Carrying Out the Invention

[0010] Hereinafter, embodiments embodying the present invention will be described based on the drawings.

[0011] [First Embodiment] Referring to FIG. 1, the overall configuration of the X-ray imaging apparatus 100 according to the first embodiment will be described.

[0012] As shown in FIG. 1, the X-ray imaging apparatus 100 is an apparatus that takes an X-ray image of a subject 90 and generates a CT image. The X-ray imaging apparatus 100 of the first embodiment is used, for example, for non-destructive inspection purposes. The subject 90 to be inspected is not particularly limited. The X-ray imaging apparatus 100 acquires rotation projection image data 30 (X-ray image data) of the subject 90 from the entire circumference in the circumferential direction of the subject placement unit 3 on which the subject 90 is placed, and constructs a tomographic image (CT image) based on the acquired rotation projection image data 30.

[0013] The X-ray imaging apparatus 100 includes an X-ray tube 1, a detector 2, a subject placement unit 3, a rotation mechanism 4, and a control device 20. The X-ray tube 1 and the detector 2 constitute an imaging unit 5 that takes an X-ray image.

[0014] The X-ray tube 1 is configured to irradiate an object 90 placed on the object placement unit 3 with X-rays 99. Specifically, the X-ray tube 1 continuously irradiates the object 90 with X-rays 99 as the object placement unit 3, on which the object 90 is placed, rotates. The X-ray tube 1 is configured to generate X-rays 99 when a high voltage is applied. The X-ray tube 1 faces the detector 2 via the object placement unit 3. The X-ray tube 1, object placement unit 3, and detector 2 are arranged horizontally.

[0015] 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 enter the detection surface of detector 2. Detector 2 is configured to convert the detected X-rays 99 into an electrical signal. This provides an X-ray image that reflects the transmission of X-rays 99 through the subject 90. Detector 2 is, for example, an FPD (Flat Panel Detector). Detector 2 is composed of a plurality of conversion elements (not shown) and pixel electrodes (not shown) arranged on the plurality of conversion elements. The plurality of conversion elements and pixel electrodes are arranged in a matrix within the detection surface at a predetermined period (pixel pitch). The detection signal (image signal) from detector 2 is sent to image processing unit 23.

[0016] The subject placement section 3 is positioned between the X-ray tube 1 and the detector 2 and is configured to place the subject 90 on it. The subject placement section 3 consists of a subject stage on which the subject 90 is placed.

[0017] The rotation mechanism 4 rotates either the imaging unit 5, which includes the X-ray tube 1 and the detector 2, or the subject placement unit 3. The rotation mechanism 4 rotates either the imaging unit 5 or the subject placement unit 3 around the rotation axis 4a. In the first embodiment, the rotation mechanism 4 rotates the subject placement unit 3 in the horizontal plane around the rotation axis 4a. The rotation mechanism 4 does not rotate the imaging unit 5. The rotation axis 4a passes through the subject placement unit 3 and is aligned vertically. The rotation axis 4a is perpendicular to the straight line (representative line of the X-ray beam) that goes from the X-ray tube 1 through the subject 90 on the subject placement unit 3 to the detector 2. The rotation mechanism 4 includes a motor (not shown) and a reduction gear (not shown) for rotating the subject placement unit 3.

[0018] The rotation mechanism 4 does not stop the rotation of the subject mounting section 3 on which the subject 90 is placed while the X-ray tube 1 is continuously irradiating the subject 90 with X-rays 99. In other words, as the subject mounting section 3 on which the subject 90 is placed is rotated by the rotation mechanism 4 and the X-ray tube 1 is continuously irradiating the subject 90 with X-rays 99, an X-ray image of the subject 90 is taken while the subject 90 is rotating.

[0019] The control device 20 includes a control unit 21, a storage unit 24, and an input / output unit 25. The control device 20 is configured, for example, by a PC (personal computer). The control device 20 is connected to a display device 26 and an input device 27.

[0020] 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 performs predetermined control when the CPU executes a predetermined program 70. Functionally, the control unit 21 includes an image capture control unit 22 and an image processing unit 23. That is, the control unit 21 functions as the image capture control unit 22 and the image processing unit 23 when the CPU executes a predetermined program 70.

[0021] The imaging control unit 22 executes the program 70 stored in the memory unit 24 to set imaging conditions in the X-ray imaging apparatus 100 and to control the start and stop of X-ray imaging. In other words, the imaging control unit 22 controls the operation of the X-ray tube 1. The imaging control unit 22 also controls the operation of the rotation mechanism 4.

[0022] 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 from the detection signal (image signal) of the detector 2. As described above, the subject 90 is X-ray photographed 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 by X-ray photographing the subject 90 while it is rotating.

[0023] The image processing unit 23 acquires multiple rotational projection image data 30 from the detector 2 for each of multiple shooting angles, which are set based on the rotation speed of the subject placement unit 3 by the rotation mechanism 4 and the frame rate (number of frames per second). The rotational projection image data 30 is X-ray image data acquired for each shooting angle. The acquisition of rotational projection image data 30 for each shooting angle is performed over a predetermined angle range that has been set in advance. The predetermined angle range that has been set in advance is 360 degrees (one rotation). In addition, the number of rotational projection image data 30 acquired is corresponding to the predetermined frame rate that has been set in advance. Note that the predetermined angle range that has been set in advance is not limited to 360 degrees (one rotation), but is not particularly limited as long as it is 180 degrees (half rotation) or more.

[0024] The image processing unit 23 obtains tomographic image data 31 by performing a reconstruction process based on the acquired rotational projection image data 30. The tomographic image data 31 may be obtained by cutting the rotational projection image data 30 at an arbitrary position. In this embodiment, preferably, the tomographic image data 31 is a tomographic image obtained by cutting in a direction perpendicular to the rotation axis of the subject 90. This is because the blurring effect of the present invention is particularly large in this tomographic image due to the rotation of the subject 90. In addition, 3D volume data can also be constructed by stacking multiple tomographic image data 31. The image processing unit 23 generates tomographic image data 31 by performing a reconstruction process on a set of rotational projection image data 30 for each of the 360 ​​degrees of shooting angles (referred to as a projection dataset).

[0025] The image processing unit 23 performs a reconstruction process using a successive approximation method, as an example. However, the reconstruction process is not limited to a successive approximation method; any known reconstruction process can be performed. For example, the reconstruction process may be an analytical method using the FDK method, or it may be an analytical method other than the FDK method.

[0026] The image processing unit 23 acquires rotation information image data 33 (see Figure 2). The 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. Details of the rotation information image data 33 will be described later.

[0027] As shown in Figure 3, the image processing unit 23 inputs tomographic image data 31 as input image data 44 to the trained model 40a stored in the memory unit 24, thereby obtaining corrected tomographic image data 32 as output image data 45. The corrected tomographic image data 32 is image data in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected. 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 trained model 40a, thereby obtaining corrected tomographic image data 32 as output image data 45 in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected. 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 trained model 40a, thereby obtaining corrected tomographic image data 32 as output result (inference result). The details of the trained model 40a will be described later. The trained model 40a is also an example of the "model" in the claims.

[0028] As shown in Figure 1, the storage unit 24 is configured to include a volatile storage device and a non-volatile storage device. The storage unit 24 stores the program 70, various setting information (not shown) related to X-ray image acquisition of the X-ray imaging device 100, rotation information image data 33, and a trained model 40a. The storage unit 24 also stores multiple acquired rotation projection image data 30, tomographic image data 31 generated based on the rotation projection image data 30, and corrected tomographic image data 32 generated using the trained model 40a.

[0029] The input / output unit 25 is composed of various interfaces for inputting and outputting signals to and from 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 detection signals (image signals) from the detector 2 via the input / output unit 25.

[0030] (Rotation information image data) Referring to Figure 2, the rotation information image data 33 will be explained. 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.

[0031] Specifically, the rotation information image data 33 is image data in which the rotation information of the subject 90 is reflected for each pixel of the tomographic image data 31 after reconstruction processing when acquiring multiple rotation projection image data 30. The rotation information of the subject 90 includes the rotation speed of the subject 90 and the rotation angle of the subject 90. In other words, the rotation information image data 33 is displacement image data that shows the amount of movement of each pixel in the tomographic image data 31, based on the rotation speed of the subject 90 as the rotation information of the subject 90 and the rotation angle of the subject 90 between the multiple rotation projection image data 30.

[0032] The rotation speed of the subject 90 is the rotation speed of the subject 90 when acquiring multiple rotational projection image data 30. In other words, the rotation speed of the subject 90 is the amount of rotation per unit time of the subject mounting section 3 by the rotation mechanism 4 when acquiring multiple rotational projection image data 30. The rotation speed of the subject mounting section 3 by the rotation mechanism 4 may be set before acquiring multiple rotational projection image data 30, or it may be set as a fixed value of the X-ray imaging apparatus 100.

[0033] The rotation angle of the subject 90 is the rotation angle of the subject 90 between the multiple rotation projection image data 30, based on the frame rate set before acquiring the multiple rotation projection image data 30. In other words, the rotation angle of the subject 90 is the rotation angle of the subject mounting unit 3 around the rotation axis 4a between the rotation projection image data 30 that is acquired and the rotation projection image data 30 that is acquired next.

[0034] Here, when rotational projection image data 30 is acquired while continuously irradiating a rotating subject 90 with X-rays 99, blurring (shaking of the subject 90) due to the rotation of the subject 90 occurs in the acquired rotational projection image data 30. In other words, the movement of the subject 90 during the acquisition of one rotational projection image data 30 appears as blurring (shaking of the subject 90) in the acquired rotational projection image data 30. Furthermore, when reconstruction processing is performed based on the rotational projection image data 30 which includes blurring (shaking of the subject 90) due to the rotation of the subject 90, blurring (shaking of the subject 90) due to the rotation of the subject 90 also appears in the tomographic image data 31 at an arbitrary cross-section acquired by the reconstruction processing. In other words, even in the tomographic image data 31 acquired by the reconstruction processing, a part of the subject is blurred due to the rotation of the subject 90.

[0035] The rotation information image data 33 shows the amount of movement of the subject 90 at each pixel of the tomographic image data 31 acquired by the reconstruction process. In other words, the rotation information image data 33 shows the amount of movement of each pixel per unit time. In the rotation information image data 33, areas with a darker color (dark areas, areas close to black) indicate a small amount of pixel movement, while areas with a lighter color (bright areas, areas close to white) indicate a large amount of pixel movement.

[0036] In the rotation information image data 33, the central part is the rotation center of the subject 90, so there is almost no pixel movement. In other words, the central part of the rotation information image data 33 is dark in color. In contrast, as you move away from the central part of the rotation information image data 33 and approach the outer periphery of the subject 90, the amount of pixel movement increases. In other words, as you move away from the central part of the rotation information image data 33 and approach the outer periphery of the subject 90, the color becomes lighter. Note that the part of the rotation information image data 33 that corresponds to the outside of the subject 90 is black because there is no pixel movement as it is not rotated.

[0037] The rotation information image data 33 shows the amount of movement of the subject 90 at each pixel of the tomographic image data 31 acquired by the reconstruction process, but does not include the direction of movement of the subject 90 at each pixel of the tomographic image data 31 acquired by the reconstruction process. However, since the central part of the rotation information image data 33 is the rotation center of the subject 90, the direction of movement of the subject 90 at each pixel is automatically determined based on the rotation direction of the subject 90.

[0038] The rotation speed of the subject 90 and the rotation angle of the subject 90 between multiple rotational projection image data 30 are set before the start of X-ray imaging by the imaging unit 5 to acquire multiple rotational projection image data 30. The image processing unit 23 acquires rotational information image data 33 based on the set rotation speed of the subject 90 and the rotation angle of the subject 90 between multiple rotational projection image data 30. For each X-ray imaging of the subject 90, the image processing unit 23 acquires rotational information image data 33 based on the set rotation speed of the subject 90 and the rotation angle of the subject 90 between multiple rotational projection image data 30.

[0039] The acquisition of rotation information image data 33 by the image processing unit 23 can occur at any time, as long as it is before the acquisition of corrected tomographic image data 32 using the trained model 40a. The acquisition of rotation information image data 33 by the image processing unit 23 can occur, for example, before the start of X-ray imaging by the imaging unit 5, after the acquisition of multiple rotation projection image data 30 by the image processing unit 23, or after the acquisition of tomographic image data 31 by the execution of reconstruction processing by the image processing unit 23.

[0040] (Trained models and methods for generating trained models) As shown in Figure 3, the trained model 40a receives tomographic image data 31 and rotation information image data 33 as input image data 44, and outputs corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected.

[0041] The method for generating a trained model 40a comprises the steps of: obtaining a training image dataset 43 consisting of input training data 41 including tomographic image data 41a of a subject 90 obtained by performing a reconstruction process based on multiple rotational projection image data 30, and output training data 42 including tomographic image data 42a (ground truth image data) in which blurring caused by the rotation of the subject 90 has been corrected; and generating a trained model 40a by machine learning based on the input training data 41 and the output training data 42, which outputs corrected tomographic image data 32 in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected.

[0042] In other words, in the first embodiment, the method for generating a trained model 40a comprises the steps of: acquiring a training image dataset 43 comprising input training data 41 including tomographic image data 41a of a subject 90 obtained by performing a reconstruction process based on a plurality of rotational projection image data 30, and rotation information image data 41b of the subject 90, and output training data 42 including tomographic image data 42a (ground truth image data) in which blurring caused by the rotation of the subject 90 has been corrected; and generating a trained model 40a by machine learning based on the input training data 41 and the output training data 42, which outputs corrected tomographic image data 32 in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected.

[0043] The tomographic image data 41a in the input training data 41 includes multiple tomographic image data 41a. These multiple tomographic image data 41a include blurring caused by the rotation of the subject 90. The rotation information image data 41b in the input training data 41 includes multiple rotation information image data 41b corresponding to each of the multiple tomographic image data 41a in the input training data 41. The tomographic image data 42a (ground truth image data) in the output training data 42, which has been corrected for blurring caused by the rotation of the subject 90, includes multiple tomographic image data 42a (ground truth image data) corresponding to each of the multiple tomographic image data 41a in the input training data 41, which has been corrected for blurring caused by the rotation of the subject 90.

[0044] The trained model 40a is generated by machine learning using input training data 41, which includes tomographic image data 41a of the subject 90 obtained by reconstructing it based on multiple rotational projection image data 30, and rotation information image data 41b of the subject 90, and output training data 42, which includes tomographic image data 42a (ground truth image data) from which blurring caused by the rotation of the subject 90 has been corrected.

[0045] The trained model 40a is pre-generated by a separate learning device 200, which is independent of the X-ray imaging device 100. The learning device 200 is a computer for machine learning, including, for example, a CPU, GPU, ROM, and RAM. The learning device 200 is located outside the X-ray imaging device 100. However, the learning device 200 may also be located within the X-ray imaging device 100.

[0046] The tomographic image data 41a in the input training data 41 may 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 may be tomographic image data 31 stored in the storage unit 24 of another X-ray imaging apparatus 100 or in the learning device 200. Similarly, the rotation information image data 41b in the input training data 41 may 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 may be rotation information image data 33 stored in the storage unit 24 of another X-ray imaging apparatus 100 or in the learning device 200. The rotation information image data 41b in the input training data 41 may be acquired (generated) by the learning device 200 to correspond to the tomographic image data 41a.

[0047] Furthermore, the output training data 42 is tomographic image data 42a (ground truth image data) from which blurring caused by rotation of the subject 90 has been corrected. However, the output training data 42 may also be tomographic image data of the subject 90 generated based on 3D CAD (Computer Aided Design) data of the subject 90, or it may be tomographic image data of the actual subject 90 that has been cut. The output training data 42 may be stored in the memory unit 24 of the X-ray imaging device 100, or it may be stored in the memory unit 24 of another X-ray imaging device 100 or in the learning device 200.

[0048] The learning device 200 takes input training data 41, which includes tomographic image data 41a and rotation information image data 41b of the subject 90, as input, and outputs output training data 42, which includes tomographic image data 42a (ground truth image data) corrected for blurring caused by the rotation of the subject 90, as output, and performs learning by machine learning to generate a trained model 40a. In other words, the learning device 200 uses a training image dataset 43, which is composed of the input training data 41 and the output training data 42, as training data (training set) to train the trained model 40a by machine learning.

[0049] In the first embodiment, U-Net++, a type of fully convolutional neural network (FCN), is used as the machine learning method for the trained model 40a. Note that 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 trained 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.

[0050] (Controlling the acquisition of corrected tomographic image data using a pre-trained model) The image processing unit 23 does not perform the acquisition of corrected tomographic image data 32 using the trained model 40a if the rotation speed of the subject 90 in the rotation information image data 33 is below a predetermined threshold, but does so if the rotation speed of the subject 90 in the rotation information image data 33 is above a predetermined threshold. In other words, when acquiring the rotation information image data 33, the image processing unit 23 does not perform the acquisition of corrected tomographic image data 32 using the trained model 40a if the rotation speed of the acquired subject mounting unit 3 is below a predetermined threshold, but does so if the rotation speed of the acquired subject mounting unit 3 is above a predetermined threshold.

[0051] Furthermore, the image processing unit 23 does not need to acquire rotation information image data 33 if the rotation speed of the acquired subject 90 (the rotation speed of the subject mounting unit 3) is less than a predetermined threshold. In other words, the image processing unit 23 does not need to generate rotation information image data 33 if the rotation speed of the acquired subject 90 (the rotation speed of the subject mounting unit 3) is less than a predetermined threshold.

[0052] The top image in Figure 4 shows an example of tomographic image data 42a (ground truth image data) with blurring caused by the rotation of the subject 90 corrected. The middle image in Figure 4 shows an example of multiple tomographic image data 31 input to the trained model 40a, each with a different rotation speed (the rotation angle of the subject 90 between multiple rotation projection image data 30 corresponding to the rotation speed). The bottom image in Figure 4 shows an example of corrected tomographic image data 32 output from the trained model 40a, with blurring caused by the rotation of the subject 90 corrected.

[0053] As shown in Figure 4, it can be seen that when the rotation speed is above a predetermined threshold (i.e., when the rotation angle is 0.75 degrees or more), the effect of correcting blur caused by the rotation of the subject 90 becomes greater. Therefore, the image processing unit 23 does not acquire corrected tomographic image data 32 using the trained model 40a when the rotation speed of the subject 90 is below a predetermined threshold (i.e., when the rotation angle is less than 0.75 degrees or more), but only when the rotation speed of the subject 90 in the rotation information image data 33 is above a predetermined threshold (i.e., when the rotation angle is 0.75 degrees or more). Note that the predetermined threshold (rotation angle) for the rotation speed of the subject 90 is not limited to the rotation speed at a rotation angle of 0.75 degrees, but can be set as appropriate.

[0054] Furthermore, if the rotation speed of the subject 90 is below a predetermined threshold, the image processing unit 23 may, instead of acquiring corrected tomographic image data 32 using the trained model 40a, perform a known filtering process on the tomographic image data 31 that reduces processing time and processing burden compared to the output processing of corrected tomographic image data 32 using the trained model 40a. As a known filtering process, for example, a smoothing filter may be used.

[0055] (CT image generation method) Next, with reference to Figure 5, the CT image generation method by the control unit 21 in the first embodiment will be described. Note that the order of the processing steps can be reversed or executed simultaneously, as long as they do not contradict each other.

[0056] In step S1, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by X-ray imaging with the imaging unit 5 while rotating the subject 90. The processing then proceeds to step S2.

[0057] In step S2, the image processing unit 23 obtains tomographic image data 31 by performing a reconstruction process based on the acquired rotational projection image data 30. The process then proceeds to step S3.

[0058] In step S3, the image processing unit 23 acquires rotation information image data 33 that reflects the rotation information of the subject 90 when acquiring multiple rotation projection image data 30. The process then proceeds to step S4.

[0059] 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), the process proceeds 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), the process proceeds to step S7.

[0060] 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 trained model 40a, thereby obtaining corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 is corrected. The process then proceeds to step S6.

[0061] In step S6, the image processing unit 23 stores the acquired corrected tomographic image data 32 in the storage unit 24. After that, the processing is terminated.

[0062] In step S7, the image processing unit 23 stores the acquired tomographic image data 31 in the storage unit 24 without acquiring corrected tomographic image data 32 using the trained model 40a. After that, the processing ends.

[0063] (Method for generating a pre-trained model) Next, with reference to Figure 6, the method for generating a trained model 40a by the learning device 200 in the first embodiment will be described. Note that the order of the processing steps can be reversed or executed simultaneously, as long as they do not contradict each other.

[0064] In step S11, the learning device 200 acquires a training image dataset 43, which consists of input training data 41 including tomographic image data 41a of the subject 90 obtained by performing reconstruction processing based on a plurality of rotational projection image data 30, and rotational information image data 41b corresponding to the tomographic image data 41a of the subject 90, and output training data 42 including tomographic image data 42a (ground truth image data) from which blurring caused by the rotation of the subject 90 has been corrected. The process then proceeds to step S12.

[0065] In step S12, the learning device 200 generates a trained model 40a by machine learning based on the input training data 41 and the output training data 42, which outputs corrected tomographic image data 32 from the tomographic image data 31, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected. The process then proceeds to step S13.

[0066] In step S13, the X-ray imaging apparatus 100 stores the trained model 40a generated by the learning device 200 in the storage unit 24 of the X-ray imaging apparatus 100. After that, the process ends.

[0067] (Comparison of the first embodiment and a modified example of the first embodiment) Referring to Figures 3 and 7(a) to 7(d), the comparison results between the corrected tomographic image data 32 (CT image) acquired by the CT image generation method in the first embodiment and the corrected tomographic image data 32a (CT image) acquired by the CT image generation method in a modified example of the first embodiment will be explained.

[0068] The modified version of the first embodiment differs from the CT image generation method in the first embodiment shown in Figure 3 in that, instead of inputting rotation information image data 33 as input image data 44 to the trained model, only tomographic image data 31 is input, thereby obtaining corrected tomographic image data 32a (see Figure 7(d)) ​​as output image data 45. Furthermore, the trained model in the modified version of the first embodiment differs from the CT image generation method in the first embodiment shown in Figure 3 in that, without including rotation information image data 41b, is generated by machine learning based on a training image dataset 43 composed of input training data 41 including tomographic image data 41a of the subject 90, and output training data 42 which is tomographic image data 42a (ground truth image data) in which blurring caused by the rotation of the subject 90 has been corrected. Note that the trained model in the modified version of the first embodiment is an example of the "model" in the claims.

[0069] The subject 90 in the CT image generation method of the first embodiment and the subject 90 in the modified CT image generation method of the first embodiment are the same subject 90. The subject 90 is a cylindrical sample made of resin, and contains a material or gap with a low X-ray absorption coefficient 99 inside the subject 90.

[0070] The upper part of Figure 7(a) is tomographic image data 42a (ground truth image data) in which blurring caused by rotation of the subject 90 in the first embodiment and modified example of the first embodiment has been corrected, and the lower part of Figure 7(a) is an enlarged view of part A of the upper part of Figure 7(a). The upper part of Figure 7(b) is tomographic image data 31 (see Figure 3) obtained by performing reconstruction processing based on a plurality of rotation projection image data 30 (see Figure 3) in the first embodiment and modified example of the first embodiment, and the lower part of Figure 7(b) is an enlarged view of part B of the upper part of Figure 7(b). The upper part of Figure 7(c) is corrected tomographic image data 32 (see Figure 3) in the first embodiment, and the lower part of Figure 7(c) is an enlarged view of part C of the upper part of Figure 7(c). The upper part of Figure 7(d) is corrected tomographic image data 32a in modified example of the first embodiment, and the lower part of Figure 7(d) is an enlarged view of part D of the upper part of Figure 7(d).

[0071] Comparing the lower diagram of Figure 7(c) with the lower diagram of Figure 7(d), it can be confirmed that the corrected tomographic image data 32 of the first embodiment shown in the lower diagram of Figure 7(c) corrects the blurring caused by the rotation of the subject 90 more effectively than the corrected tomographic image data 32a of the modified example of the first embodiment shown in the lower diagram of Figure 7(d). Therefore, it can be confirmed that the CT image generation method of the first embodiment can reduce the blurring caused by the rotation of the subject 90 in tomographic image data at any cross-section more effectively than the CT image generation method of the modified example of the first embodiment.

[0072] [Second Embodiment] Next, with reference to Figures 8 to 12, a CT image generation method and a method for generating a trained model 40b according to the second embodiment will be described. In the second embodiment, an example will be described in which a plurality of tomographic image data 31 with different noise levels are input to the trained 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 components similar to those in the first embodiment, and their descriptions are omitted. Note that the trained model 40b is an example of the "model" in the claims.

[0073] As shown in Figure 8, in the second embodiment, the image processing unit 23 acquires multiple tomographic image data 31 with different noise levels in the reconstruction process based on the acquired multiple rotational projection image data 30. In the second embodiment, as an example, the image processing unit 23 acquires two types of tomographic image data 31 with different smoothness levels in the reconstruction process based on the acquired multiple rotational projection image data 30.

[0074] Two types of tomographic image data 31 with different smoothing intensities can be obtained, for example, in the reconstruction process by processing the acquired tomographic image data 31 with two smoothing filters having different smoothing intensities. Known smoothing filters such as Gaussian filters can be used as smoothing filters.

[0075] The strength of smoothing can be increased or decreased by varying the kernel size (filter size) of the filter function used. One of two smoothing filters with different strengths is a first smoothing filter in which the smoothing strength is increased by increasing the kernel size. Increasing the smoothing strength makes the change in pixel values ​​between pixels smoother. The other of the two smoothing filters with different strengths is a second smoothing filter in which the smoothing strength is reduced compared to the first smoothing filter by decreasing the kernel size. Note that the smoothing filter is not limited to a Gaussian filter; for example, a low-pass filter may also be used.

[0076] In the reconstruction process, the image processing unit 23 applies a smoothing process to the acquired tomographic image data 31 using a first smoothing filter to obtain a first tomographic image data 31a with a high smoothing intensity. In the reconstruction process, the image processing unit 23 also applies a smoothing process to the acquired tomographic image data 31 using a second smoothing filter to obtain a second tomographic image data 31b with a low smoothing intensity.

[0077] As shown in Figure 9, the image processing unit 23 inputs a plurality of tomographic image data 31 with different noise levels as input image data 44 to the trained model 40b stored in the storage unit 24, thereby obtaining corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 with different noise levels has been corrected. In the second embodiment, the image processing unit 23 inputs a plurality of tomographic image data 31 with different smoothness levels as input image data 44 to the trained model 40b, thereby obtaining corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 with different smoothness levels has been corrected. 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 as input image data 44 to the trained model 40b, thereby obtaining corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b has been corrected.

[0078] In other words, 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 trained model 40b, and obtains corrected tomographic image data 32 as the output result (inference result).

[0079] In the second embodiment, the multiple tomographic image data 31 with different noise levels may be, for example, three or more types of tomographic image data 31 with different smoothness levels. In this case, the image processing unit 23 inputs three or more types of tomographic image data 31 with different smoothness levels as input image data 44 to the trained model 40b, thereby obtaining corrected tomographic image data 32 as output image data 45. Also, in the second embodiment, the multiple tomographic image data 31 with different noise levels may be, for example, multiple tomographic image data 31 with different noise levels obtained by using different reconstruction parameters in the reconstruction process.

[0080] (Trained models and methods for generating trained models) In the first embodiment, the trained model 40b receives multiple tomographic image data 31 with varying degrees of noise as input image data 44, and outputs corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 with varying degrees of noise has been corrected.

[0081] The method for generating a trained model 40b comprises the steps of: obtaining a training image dataset 43 consisting of input training data 41 including multiple tomographic image data (41c, 41d) of a subject 90 with different degrees of noise, obtained in a reconstruction process based on multiple rotational projection image data 30, and output training data 42 including tomographic image data 42a (ground truth image data) in which blurring caused by the rotation of the subject 90 has been corrected; and generating a trained model 40b by machine learning based on the input training data 41 and the output training data 42, which outputs corrected tomographic image data 32 in which blurring caused by the rotation of the subject 90 in the tomographic image data 31 has been corrected.

[0082] In other words, in the second embodiment, the method for generating a trained model 40b comprises the steps of: acquiring a training image dataset 43 comprising input training data 41 including a first tomographic image data 41c with a high smoothing intensity and a second tomographic image data 41d with a low smoothing intensity, which are obtained by performing a reconstruction process based on a plurality of rotational projection image data 30, and output training data 42 including tomographic image data 42a (ground truth image data) in which blurring caused by the rotation of the subject 90 has been corrected; and generating a trained model 40b by machine learning based on the input training data 41 and the output training data 42, which outputs corrected tomographic image data 32 in which blurring caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b has been corrected.

[0083] The first tomographic image data 41c and the second tomographic image data 41d in the input training data 41 include sets 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 include blurring caused by the rotation of the subject 90. Furthermore, the tomographic image data 42a (ground truth image data) in the output training data 42, which has been corrected for blurring caused by the rotation of the subject 90, includes multiple tomographic image data 42a (ground truth image data) which have been corrected for blurring caused by the rotation of the subject 90 corresponding to each of the sets of multiple first tomographic image data 41c and second tomographic image data 41d in the input training data 41.

[0084] The trained model 40b is generated by machine learning using input training data 41, which includes a first tomographic image data 41c with a high smoothing intensity and a second tomographic image data 41d with a low smoothing intensity, obtained by performing a reconstruction process based on multiple rotational projection image data 30, and output training data 42, which includes tomographic image data 42a (ground truth image data) from which blurring caused by the rotation of the subject 90 has been corrected.

[0085] The first tomographic image data 41c and the second tomographic image data 41d in the input training data 41 may 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 may 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 in a learning device 200.

[0086] 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 outputs output training data 42, which includes tomographic image data 42a (ground truth image data) from which blurring caused by the rotation of the subject 90 has been corrected, as output, and performs learning by machine learning to generate a trained model 40b.

[0087] (CT image generation method) Next, with reference to Figure 10, the CT image generation method by the control unit 21 in the second embodiment will be described. Note that the order of the processing steps can be reversed or executed simultaneously, as long as they do not contradict each other.

[0088] In step S21, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by X-ray imaging with the imaging unit 5 while rotating the subject 90. The processing then proceeds to step S22.

[0089] In step S22, the image processing unit 23 obtains a first tomographic image data 31a and a second tomographic image data 31b, which are two types of tomographic image data 31 with different smoothness levels, in a reconstruction process based on the acquired multiple rotational projection image data 30. The process then proceeds to step S23.

[0090] 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 trained model 40b, thereby obtaining corrected tomographic image data 32 as output image data 45, in which blurring caused by the rotation of the subject 90 in the first tomographic image data 31a and the second tomographic image data 31b is corrected. The processing then proceeds to step S24.

[0091] In step S24, the image processing unit 23 stores the acquired corrected tomographic image data 32 in the storage unit 24. After that, the processing is completed.

[0092] (Method for generating a pre-trained model) Next, with reference to Figure 11, the method for generating the trained model 40b by the learning device 200 in the second embodiment will be described. Note that the order of the processing steps can be reversed or executed simultaneously, as long as they do not contradict each other.

[0093] In step S31, the learning device 200 acquires a training image dataset 43, which consists of input training data 41 including a first tomographic image data 41c with a high smoothing intensity and a second tomographic image data 41d with a low smoothing intensity, obtained in a reconstruction process based on multiple rotational projection image data 30, and output training data 42 including tomographic image data 42a (ground truth image data) from which blurring caused by the rotation of the subject 90 has been corrected. The process then proceeds to step S32.

[0094] In step S32, the learning device 200 generates a trained model 40b by machine learning based on the input training data 41 and the output training data 42. This model outputs corrected tomographic image data 32 from the first tomographic image data 31a and the second tomographic image data 31b, correcting 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.

[0095] In step S33, the X-ray imaging apparatus 100 stores the trained model 40b generated by the learning device 200 in the storage unit 24 of the X-ray imaging apparatus 100. After that, the process ends.

[0096] (Comparison of the second embodiment and a modified example of the second embodiment) Referring to Figures 9 and 12(a) to 12(e), the comparison results between the corrected tomographic image data 32 (CT image) obtained by the CT image generation method in the second embodiment and the corrected tomographic image data 32b (CT image) obtained by the CT image generation method in a modified example of the second embodiment will be explained.

[0097] The CT image generation method in the modified version of the second embodiment differs from the CT image generation method in the second embodiment shown in Figure 9. In this method, instead of inputting the second tomographic image data 31b as input image data 44 to the trained model, only the first tomographic image data 31a is input, thereby obtaining corrected tomographic image data 32b (see Figure 12(e)) as output image data 45. Furthermore, the trained model in the modified version of the second embodiment differs from the CT image generation method in the second embodiment shown in Figure 9. It is generated by machine learning based on a training image dataset 43, which consists of input training data 41 including the first tomographic image data 41c but not the second tomographic image data 41d, and output training data 42 which is tomographic image data 42a (ground truth image data) with blurring caused by the rotation of the subject 90 corrected. Note that the trained model in the modified version of the second embodiment is an example of the "model" in the claims.

[0098] The subject 90 in the CT image generation method of the second embodiment and the subject 90 in the modified CT image generation method of the second embodiment are the same subject 90. The subject 90 is a cylindrical sample made of resin, and contains a material or gap with a low X-ray absorption coefficient 99 inside the subject 90.

[0099] The upper part of Figure 12(a) is tomographic image data 42a (ground truth image data) (see Figure 9) in which blurring caused by rotation of the subject 90 in the second embodiment and modified example of the second embodiment has been corrected, and the lower part of Figure 12(a) is an enlarged view of part E in the upper part of Figure 12(a). The upper part of Figure 12(b) is first tomographic image data 31a (see Figure 9) in the second embodiment and modified example of the second embodiment with a high degree of smoothing, and the lower part of Figure 12(b) is an enlarged view of part F in the upper part of Figure 12(b). The upper part of Figure 12(c) is second tomographic image data 31b (see Figure 9) in the second embodiment with a low degree of smoothing, and the lower part of Figure 12(c) is an enlarged view of part G in the upper part of Figure 12(c). The upper part of Figure 12(d) is corrected tomographic image data 32 (see Figure 9) in the second embodiment, and the lower part of Figure 12(d) is an enlarged view of part H in the upper part of Figure 12(d). The upper part of Figure 12(e) is corrected tomographic image data 32b of a modified example of the second embodiment, and the lower part of Figure 12(e) is an enlarged view of part I of the upper part of Figure 12(e).

[0100] Comparing the lower diagram of Figure 12(d) with the lower diagram of Figure 12(e), it can be confirmed that the corrected tomographic image data 32 of the second embodiment shown in the lower diagram of Figure 12(d) corrects blurring caused by the rotation of the subject 90 more effectively than the corrected tomographic image data 32b of the modified example of the second embodiment shown in the lower diagram of Figure 12(e). Furthermore, it can be confirmed that the corrected tomographic image data 32 of the second embodiment shown in the lower diagram of Figure 12(d) extracts the feature points 91 of the subject 90 with greater accuracy than the corrected tomographic image data 32b of the modified example of the second embodiment shown in the lower diagram of Figure 12(e). Therefore, it can be confirmed that the CT image generation method in the second embodiment reduces blurring caused by the rotation of the subject 90 in tomographic image data at any tomographic plane and extracts the feature points 91 of the subject 90 with greater accuracy than the CT image generation method in the modified example of the second embodiment.

[0101] [Third Embodiment] Next, a CT image generation method according to the third embodiment will be described with reference to Figures 13 to 16. In the third embodiment, in the reconstruction process, intermediate reconstructed image data 31c is acquired by an intermediate reconstruction process, corrected reconstructed image data 35 is acquired from the acquired intermediate reconstructed image data 31c using a trained model 40a, and the final reconstructed image data 31d is obtained by performing a reconstruction process on the acquired corrected reconstructed image data 35. In the third embodiment, the same reference numerals are used for components similar to those in the first embodiment, and their descriptions are omitted.

[0102] The image processing unit 23 performs a reconstruction process using the iterative approximation method. In the reconstruction process using the iterative approximation method, as an example, calculation processes including forward projection, back projection, comparison, and updating are repeatedly performed in the intermediate stages. Specifically, in an example of the reconstruction process using the iterative approximation method, the following calculation processes (reconstruction process) are repeatedly performed in the intermediate stages of the reconstruction process: (1) the k-th projection is calculated from the k-th image (intermediate reconstructed image data 31c) (forward projection); (2) the ratio between the k-th forward projection and the measured projection is found; (3) the found ratio is back-projected; and (4) the k+1-th image is multiplied by the back-projected image to update the image (intermediate reconstructed image data 31c).

[0103] In the reconstruction process, the image processing unit 23 performs reconstruction processing multiple times based on the acquired rotational projection image data 30, and also acquires a first intermediate reconstruction image data 31c through reconstruction processing in the intermediate stage.

[0104] Furthermore, during the reconstruction process, the image processing unit 23 uses the trained model 40a to obtain the first corrected reconstructed image data 35 from the acquired first intermediate reconstructed image data 31c. Specifically, when obtaining 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 trained model 40a, thereby obtaining the first corrected reconstructed image data 35 as output image data 45, in which the blurring caused by the rotation of the subject 90 in the first intermediate reconstructed image data 31c has been corrected.

[0105] Furthermore, in the reconstruction process, the image processing unit 23 obtains a second intermediate reconstruction image data 31c by performing a reconstruction process on the acquired first corrected reconstruction image data 35. In the third embodiment, the image processing unit 23 obtains a second intermediate reconstruction image data 31c by performing a calculation process in the reconstruction process using the iterative approximation method on the acquired first corrected reconstruction image data 35.

[0106] Furthermore, the image processing unit 23 receives the second intermediate reconstructed image data 31c and the rotation information image data 33 as input image data 44 to the trained model 40a, thereby obtaining the second corrected reconstructed image data 35 as output image data 45. The image processing unit 23 then performs calculation processing using the successive approximation method on the obtained second corrected reconstructed image data 35 to obtain the final reconstructed image data 31d.

[0107] In the third embodiment, the trained model is, for example, the same trained model 40a as the trained model 40a in the first embodiment. However, the trained model in the third embodiment is not limited to the same trained model 40a as the trained model 40a in the first embodiment. The trained model in the third embodiment may be, for example, the trained model 40b in the second embodiment. When the trained model is the trained model 40a in the first embodiment, the trained model 40a is generated by the method for generating the trained model 40a described in the first embodiment. When the trained model is the trained model 40b in the second embodiment, the trained model 40b is generated by the method for generating the trained model 40b described in the second embodiment.

[0108] The trained model in the third embodiment is not limited to the trained model 40a in the first embodiment and the trained model 40b in the second embodiment. The trained model in the third embodiment may be a trained model configured to output corrected reconstructed image data 35 from the intermediate reconstructed image data 31c, in which blurring caused by rotation of the subject 90 in the intermediate reconstructed image data 31c has been corrected, using supervised learning, unsupervised learning, or reinforcement learning. Furthermore, the machine learning method for the trained model in the third embodiment is not particularly limited. The trained model in the third embodiment may also be a generative AI (Artificial Intelligence). The trained model in the third embodiment is an example of the "model" in the claims.

[0109] (CT image generation) Next, with reference to Figure 14, the CT image generation by the control unit 21 in the third embodiment will be described. Specifically, the reconstruction process using the trained model 40a by the control unit 21 and the iterative approximation method will be described. The image processing unit 23 performs the calculation process of the reconstruction process using the iterative approximation method n times.

[0110] As shown in Figure 14, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by the detector 2 by X-ray imaging while rotating the subject 90.

[0111] The image processing unit 23 obtains the first intermediate reconstructed image data 31c by performing the first calculation in the reconstruction process using the iterative approximation method. Specifically, as the first calculation in the reconstruction process using the iterative approximation method, the image processing unit 23 obtains the first intermediate reconstructed image data 31c from the initial image data 34 by performing forward projection, back projection, comparison, and update.

[0112] The image processing unit 23 receives the first intermediate reconstructed image data 31c and the rotation information image data 33 as input image data 44 to the trained model 40a, thereby obtaining the first corrected reconstructed image data 35 as output image data 45.

[0113] The image processing unit 23 obtains the second intermediate reconstructed image data 31c from the first corrected reconstructed image data 35 by performing forward projection, back projection, comparison, and update as the second calculation process of the reconstruction process using the successive approximation method.

[0114] The image processing unit 23 receives the second intermediate reconstructed image data 31c and the rotation information image data 33 as input image data 44 to the trained model 40a, thereby obtaining the second corrected reconstructed image data 35 as output image data 45.

[0115] The image processing unit 23 then repeatedly performs the calculation process for reconstruction using the successive approximation method, and the input of input image data 44 to the trained model 40a and the acquisition of output image data 45. The image processing unit 23 repeatedly performs the calculation process for reconstruction and the acquisition of output image data 45 using the trained model 40a n-1 times.

[0116] The image processing unit 23 performs forward projection, back projection, comparison, and update as the nth calculation step of the reconstruction process using the successive approximation method, thereby obtaining the nth intermediate reconstruction image data 31c as the final reconstruction image data 31d from the (n-1)th corrected reconstruction image data 35. The image processing unit 23 stores the final reconstruction image data 31d, which is the nth intermediate reconstruction image data 31c, in the storage unit 24.

[0117] (CT image generation method) Next, with reference to Figure 15, the CT image generation method by the control unit 21 in the third embodiment will be described. Note that the order of the processing steps can be reversed or executed simultaneously, as long as they do not contradict each other. The image processing unit 23 performs the reconstruction processing calculation using the iterative approximation method n times.

[0118] In step S51, the image processing unit 23 acquires multiple rotational projection image data 30 obtained by X-ray imaging with the imaging unit 5 while rotating the subject 90. The processing then proceeds to step S52.

[0119] In step S52, the image processing unit 23 obtains a first intermediate reconstructed image data 31c from the initial image data 34 by performing the first calculation in the reconstruction process using the iterative approximation method. The process then proceeds to step S53.

[0120] In step S53, the image processing unit 23 inputs the first intermediate reconstructed image data 31c and the rotation information image data 33 as input image data 44 to the trained model 40a, thereby obtaining the first corrected reconstructed image data 35 as output image data 45. The process then proceeds to step S54.

[0121] In step S54, the image processing unit 23 performs the second (n-1) calculation in the reconstruction process using the iterative approximation method, thereby obtaining the second (n-1) intermediate reconstruction image data 31c from the first (n-2) corrected reconstruction image data 35. The process then proceeds to step S55.

[0122] In step S55, the image processing unit 23 inputs the second (n-1) intermediate reconstructed image data 31c and rotation information image data 33 as input image data 44 to the trained model 40a, thereby obtaining the second (n-1) corrected reconstructed image data 35 as output image data 45. The process then proceeds to step S56.

[0123] In step S56, the image processing unit 23 determines whether the calculation process of the reconstruction process using the successive approximation method and the acquisition of input image data 44 and output image data 45 for the trained model 40a have been repeated n-1 times. If the calculation process of the reconstruction process and the acquisition of output image data 45 using the trained model 40a have been performed n-1 times (Yes in step S56), the process proceeds to step S57. If the calculation process of the reconstruction process and the acquisition of output image data 45 using the trained model 40a have been performed less than n-1 times (No in step S56), the process proceeds to step S54.

[0124] 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 successive approximation method, thereby obtaining the nth intermediate reconstruction image data 31c as the final reconstruction image data 31d from the (n-1)th corrected reconstruction image data 35. The process then proceeds to step S58.

[0125] 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.

[0126] (Effect of the final reconstructed image data according to the third embodiment) Referring to Figures 3, 14, and 16(a) to 16(c), the effect of the final reconstructed image data 31d obtained by the CT image generation method in the third embodiment will be explained.

[0127] Figure 16(b) shows tomographic image data 42a (ground truth image data) in which blurring caused by rotation of the subject 90a in the third embodiment and a modified example of the third embodiment has been corrected. Figure 16(c) is a magnified portion of the final reconstructed image data 31d, which is the nth intermediate reconstructed image data 31c of the third embodiment.

[0128] The final reconstructed image data 31d, which is the nth intermediate reconstructed image data 31c of the third embodiment shown in Figure 16(c), can be seen to have the blurring caused by the rotation of the subject 90 corrected. Therefore, it can be seen that the CT image generation method of the third embodiment can reduce blurring caused by the rotation of the subject 90 in tomographic image data at any tomographic plane.

[0129] [Differentiation] It should be noted that the embodiments disclosed herein are illustrative and not restrictive in all respects. The scope of the present invention is defined by the claims rather than by the description of the embodiments above, and further includes all modifications (exceptions) within the meaning and scope equivalent to the claims. For example, in each of the first to third embodiments, instead of using a pre-trained model, unsupervised learning may be used to output an image in which the blurring caused by rotation has been corrected. Furthermore, in the first and second embodiments, volume data with blur correction may be obtained and output using the multiple corrected tomographic image data output. Furthermore, an embodiment combining the first and second embodiments can also be adopted. That is, by inputting multiple tomographic image data with different noise levels and rotation information image data into the model, it is possible to output corrected tomographic image data in which blurring caused by the rotation of the subject is further corrected. Similarly, in the third embodiment, an embodiment combining the first and / or second embodiments may 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 in which blurring caused by the rotation of the subject is corrected may be output, (2) by inputting a plurality of intermediate reconstructed image data with different noise levels into the model, corrected reconstructed image data in which blurring caused by the rotation of the subject is corrected may be output, or (3) by inputting a plurality of intermediate reconstructed image data with different noise levels and rotation information image data into the model, corrected reconstructed image data in which blurring caused by the rotation of the subject is corrected may be output. Furthermore, for example, in the first embodiment described above, the system may be configured to acquire corrected tomographic image data even when the rotation speed of the subject in the rotation information image data is less than a predetermined threshold. Furthermore, the system may be configured to acquire rotational projection image data while pulsed X-rays are irradiated onto a rotating subject using an X-ray tube.

[0130] [Pattern] Those skilled in the art will understand that the exemplary embodiments described above are specific examples of the following embodiments.

[0131] (Item 1) The process involves acquiring multiple rotational projection image data by taking X-ray images while rotating the subject, The steps include obtaining tomographic image data by performing a reconstruction process based on the acquired plurality of rotational projection image data, A CT image generation method comprising the steps of: inputting the tomographic image data as input image data to a model, thereby obtaining corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected as output image data. By inputting tomographic image data based on rotational projection image data obtained by X-ray imaging while rotating the subject as input image data to the model, it is possible to obtain corrected tomographic image data as output image data in which blurring caused by the subject's rotation in the tomographic image data has been corrected. Therefore, even if blurring caused by the subject's rotation occurs in the tomographic image data based on rotational projection image data, inputting tomographic image data into the model will result in corrected tomographic image data from which the blurring caused by the subject's rotation has been corrected. Thus, it is possible to obtain corrected tomographic image data in which blurring caused by the subject's rotation has been corrected. As a result, blurring caused by the subject's rotation can be reduced in tomographic image data at any given fault plane. (Item 2) The process further 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 CT image generation method according to item 1, wherein the step of acquiring the corrected tomographic image data is to input the tomographic image data and the rotation information image data as input image data to the model, thereby acquiring the corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected as output image data. By inputting tomographic image data and rotation information image data that reflects the rotation of the subject as input image data to the model, it is possible to obtain corrected tomographic image data as output image data in which blurring caused by the rotation of the subject has been corrected. Therefore, compared to performing deconvolution or filtering on tomographic image data, it is possible to obtain corrected tomographic image data in which blurring caused by the rotation of the subject has been more effectively corrected. As a result, blurring caused by the rotation of the subject can be further reduced in tomographic image data at any given fault plane. (Item 3) The CT image generation method according to item 2, wherein the step of acquiring rotation information image data is to acquire image data that reflects the rotation speed of the subject and the rotation angle of the subject between the plurality of rotation projection image data, as the rotation information image data. The rotational information of a known subject, including the subject's rotation speed and the subject's rotation angle between multiple rotational projection image data, is known subject rotation information that was set before acquiring the multiple rotational projection image data, or set as a fixed value in the X-ray imaging device. Therefore, rotational information image data can be easily acquired based on the rotational information of the known subject. (Item 4) The CT image generation method according to item 3, wherein the step of acquiring rotation information image data is to acquire displacement image data, which indicates the amount of movement of each pixel in the tomographic image data, based on the rotation speed of the subject and the rotation angle of the subject between the plurality of rotation projection image data, as the rotation information image data. By inputting displacement data, which shows the amount of movement of each pixel in tomographic image data, as the input image to the model, it is possible to accurately obtain corrected tomographic image data as output image data, in which blurring caused by the rotation of the subject has been corrected. (Item 5) The CT image generation method according to any one of items 2 to 4, wherein the step of acquiring the corrected tomographic image data is not performed when the rotation speed of the subject is less than a predetermined threshold, and is performed when the rotation speed of the subject is equal to or greater than the predetermined threshold. When the blurring caused by the rotation of the subject is relatively small and the rotation speed of the subject is below a predetermined threshold, the processing load can be reduced by not performing the acquisition of corrected tomographic image data using the model. Conversely, when the blurring caused by the rotation of the subject is relatively large and the rotation speed of the subject is above a predetermined threshold, the acquisition of corrected tomographic image data using the model can be performed to obtain corrected tomographic image data in which the blurring caused by the rotation of the subject has been effectively corrected. (Item 6) The step of acquiring the tomographic image data involves acquiring a plurality of tomographic image data having different noise levels from each other. The CT image generation method according to any one of items 1 to 5, wherein the step of acquiring the corrected tomographic image data is to input a plurality of tomographic image data, each having a different degree of noise, as the input image data to the model, thereby acquiring the corrected tomographic image data, in which blurring caused by the rotation of the subject in the tomographic image data has been corrected, as the output image data. By inputting multiple tomographic image data with varying degrees of noise into the model, it is possible to obtain corrected tomographic image data as output image data, which corrects blurring caused by subject rotation. Compared to inputting a single tomographic image data, this method allows for more effective correction of blurring caused by subject rotation. Therefore, blurring caused by subject rotation can be further reduced in tomographic image data at any given fault plane. (Item 7) The step of acquiring the tomographic image data involves acquiring a plurality of tomographic image data having different degrees of smoothness from each other. The CT image generation method according to item 6, wherein the step of acquiring the corrected tomographic image data is to input a plurality of tomographic image data with different smoothness levels to the model as the input image data, thereby acquiring the corrected tomographic image data as the output image data. By inputting multiple tomographic image data with different smoothness levels as input image data for training, it is possible to obtain corrected tomographic image data as output image data in which blurring caused by the rotation of the subject is accurately corrected and feature points of the subject are accurately extracted. Therefore, blurring caused by the rotation of the subject can be reduced and feature points of the subject can be accurately extracted in tomographic image data at any tomographic plane. (Item 8) The aforementioned model is a trained model, The step further comprises generating the aforementioned trained model, The step of generating the aforementioned trained model is: A CT image generation method according to any one of items 1 to 7, comprising generating a trained model by machine learning based on input training data including the tomographic image data and output training data including tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected, thereby outputting corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected from the tomographic image data. By inputting tomographic image data as input image data to a pre-trained model generated by machine learning, corrected tomographic image data, in which blurring caused by subject rotation in the tomographic image data has been corrected, can be easily obtained as output image data. (Item 9) The process involves acquiring multiple rotational projection image data by taking X-ray images while rotating the subject, The steps include performing multiple reconstruction processes based on the acquired multiple rotational projection image data, and acquiring intermediate reconstruction image data through the reconstruction process in the intermediate stage, The steps include: using a model to obtain corrected reconstructed image data from the acquired intermediate reconstructed image data, in which blurring caused by the rotation of the subject has been corrected; The process includes the step of obtaining the final reconstructed image data by further performing a reconstruction process using the acquired corrected reconstructed image data, The step of acquiring the corrected reconstructed image data is a CT image generation method in which, by inputting the intermediate reconstructed image data as input image data to the model, the corrected reconstructed image data is obtained as output image data in which blurring caused by the rotation of the subject in the intermediate reconstructed image data has been corrected. In reconstruction processes that involve multiple reconstruction steps, a model can be used to obtain corrected reconstruction image data from the acquired intermediate reconstruction image data, which has been corrected for blurring caused by the rotation of the subject. This allows the reconstruction process to be performed on the corrected reconstruction image data, which has already been corrected for blurring caused by the rotation of the subject, during the intermediate reconstruction steps. As a result, the final reconstruction image data can be obtained with high accuracy in correcting for blurring caused by the rotation of the subject. Therefore, blurring caused by the rotation of the subject can be further reduced in tomographic image data at any given tomographic plane. (Item 10) The step of obtaining the intermediate reconstructed image data involves obtaining the intermediate reconstructed image data by performing the calculation process in the reconstruction process using the successive approximation method, The CT image generation method according to item 9, wherein the step of obtaining the final reconstructed image data is to obtain the final reconstructed image data by performing the calculation process by the iterative approximation method on the acquired corrected reconstructed image data. When performing multiple successive reconstruction calculations using the successive approximation method, a model can be used to obtain corrected reconstruction image data from intermediate reconstruction image data acquired through the calculations, thereby correcting blur caused by subject rotation. This allows subsequent successive reconstruction calculations using the successive approximation method to be performed on the corrected reconstruction image data, which has already corrected blur caused by subject rotation. As a result, it is easy to obtain reconstruction image data with accurately corrected blur caused by subject rotation as the final reconstruction image data. Consequently, blur caused by subject rotation can be more easily reduced in tomographic image data at any given fault plane. (Item 11) A step of obtaining a training image dataset comprising input training data including tomographic image data of a subject obtained by performing a reconstruction process based on multiple rotational projection image data, and output training data including the tomographic image data from which blurring caused by the rotation of the subject has been corrected, A method for generating a trained model, comprising the steps of generating a trained model by machine learning based on the input training data and the output training data, which outputs corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected from the tomographic image data. Based on input training data containing tomographic image data of the subject and output training data containing tomographic image data corrected for blurring caused by the subject's rotation, machine learning can be used to train an image processing model that corrects blurring caused by the subject's rotation. Therefore, a trained model can be easily generated that outputs corrected tomographic image data in which the blurring caused by the subject's rotation in the tomographic image data has been corrected. (Item 12) The method for generating a trained model according to item 11, wherein the step of acquiring the training image dataset is to acquire the training image dataset comprising input training data including tomographic image data of the subject and rotation information image data reflecting the rotation information of the subject, and output training data. Based on input training data including tomographic image data and rotational information image data that reflects the rotation of the subject, and output training data, machine learning can be used to train an image processing model that corrects blurring caused by the rotation of the subject. Therefore, it is possible to easily generate a trained model that outputs corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data is effectively corrected. (Item 13) The method for generating a trained model according to item 11, wherein the step of obtaining the training image dataset is to obtain the training image dataset comprising input training data including a plurality of tomographic image data of the subject having different degrees of noise, and output training data. Based on input training data containing multiple tomographic image data of a subject with varying degrees of noise, and output training data, machine learning can be used to train an image processing model that corrects blurring caused by the rotation of the subject. Therefore, compared to training an image processing model using machine learning based on input training data consisting of a single tomographic image, it is possible to generate a trained model that outputs corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data is effectively corrected. [Explanation of Symbols]

[0132] 30 Rotation Projection Image Data 31, 41a Fault image data 31c Intermediate Reconstructed Image Data 31d Final reconstructed image data 32 Corrected tomographic image data 33, 41b Rotation information image data 35 Corrected and reconstructed image data 40a, 40b Pre-trained models (models) 41 Input training data 42 Output training data 42a Tomographic image data (ground truth image data) with blurring caused by subject rotation corrected. 43 Training Image Datasets 44 Input image data 45 Output image data 90 Subjects

Claims

1. The process involves acquiring multiple rotational projection image data by taking X-ray images while rotating the subject, The steps include: obtaining tomographic image data by performing a reconstruction process based on the acquired plurality of rotational projection image data, A CT image generation method comprising the steps of inputting the tomographic image data as input image data to a model, thereby obtaining corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected as output image data.

2. The process further 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 CT image generation method according to claim 1, wherein the step of acquiring the corrected tomographic image data is to input the tomographic image data and the rotation information image data as input image data to the model, thereby acquiring the corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected as output image data.

3. The CT image generation method according to claim 2, wherein the step of acquiring rotation information image data is to acquire image data as the rotation information image data that reflects the rotation speed of the subject and the rotation angle of the subject between the plurality of rotation projection image data.

4. The CT image generation method according to claim 3, wherein the step of acquiring rotation information image data is to acquire, as the rotation information image data, displacement image data indicating the amount of movement of each pixel in the tomographic image data, based on the rotation speed of the subject and the rotation angle of the subject between the plurality of rotation projection image data.

5. The CT image generation method according to claim 2, wherein the step of acquiring the corrected tomographic image data is not performed when the rotation speed of the subject is less than a predetermined threshold, and is performed when the rotation speed of the subject is equal to or greater than the predetermined threshold.

6. The step of acquiring the tomographic image data involves acquiring a plurality of tomographic image data having different noise levels from each other. The CT image generation method according to claim 1 or 2, wherein the step of acquiring the corrected tomographic image data is to input a plurality of tomographic image data, each having a different degree of noise, to the model as the input image data, thereby acquiring the corrected tomographic image data, in which blurring caused by the rotation of the subject in the tomographic image data has been corrected, as the output image data.

7. The step of acquiring the tomographic image data involves acquiring a plurality of tomographic image data having different degrees of smoothness from each other, The CT image generation method according to claim 6, wherein the step of acquiring the corrected tomographic image data is to input a plurality of tomographic image data with different smoothness levels to the model as the input image data, thereby acquiring the corrected tomographic image data as the output image data.

8. The aforementioned model is a trained model, The step of generating the aforementioned trained model further comprises: The step of generating the aforementioned trained model is: A CT image generation method according to claim 1 or 2, comprising generating a trained model by machine learning based on input training data including the tomographic image data and output training data including tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected, thereby outputting corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected from the tomographic image data.

9. The process involves acquiring multiple rotational projection image data by taking X-ray images while rotating the subject, The steps include performing multiple reconstruction processes based on the acquired multiple rotational projection image data, and acquiring intermediate reconstruction image data through the reconstruction process in the intermediate stage, The steps include: using a model to obtain corrected reconstructed image data from the acquired intermediate reconstructed image data, in which blurring caused by the rotation of the subject has been corrected; The process includes the step of obtaining the final reconstructed image data by further performing a reconstruction process using the acquired corrected reconstructed image data, A CT image generation method in which the step of acquiring the corrected reconstructed image data is performed by inputting the intermediate reconstructed image data as input image data to the model, thereby acquiring the corrected reconstructed image data as output image data in which blurring caused by the rotation of the subject in the intermediate reconstructed image data has been corrected.

10. The step of obtaining the intermediate reconstructed image data involves obtaining the intermediate reconstructed image data by performing the calculation process in the reconstruction process using the successive approximation method. The CT image generation method according to claim 9, wherein the step of obtaining the final reconstructed image data is to obtain the final reconstructed image data by performing the calculation processing by the successive approximation method on the acquired corrected reconstructed image data.

11. A step of obtaining a training image dataset comprising input training data including tomographic image data of a subject obtained by performing a reconstruction process based on multiple rotational projection image data, and output training data including the tomographic image data from which blurring caused by the rotation of the subject has been corrected, A method for generating a trained model, comprising the steps of generating a trained model by machine learning based on the input training data and the output training data, which outputs corrected tomographic image data in which blurring caused by the rotation of the subject in the tomographic image data has been corrected from the tomographic image data.

12. The method for generating a trained model according to claim 11, wherein the step of acquiring the training image dataset is to acquire the training image dataset comprising input training data including tomographic image data of the subject and rotation information image data reflecting the rotation information of the subject, and output training data.

13. The method for generating a trained model according to claim 11, wherein the step of acquiring the training image dataset is to acquire the training image dataset comprising input training data including a plurality of tomographic image data of the subject having different degrees of noise, and output training data.

Images