JP2020188991A - Medical image processor, medical image processing program, x-ray diagnostic device, and x-ray diagnostic system - Google Patents

Abstract

To efficiently reduce an artifact accompanying multi-phase photographing using rotational photographing.SOLUTION: A medical image processor according to an embodiment includes: an acquisition unit that acquires a first projection data set generated by rotational photographing executed at first timing on a subject in which a contrast medium is injected and a second projection data set generated by rotational photographing executed at second timing differing from the first timing on the subject; a reconfiguration unit that generates a first reconfiguration image on the basis of the first projection data set and generates a second reconfiguration image on the basis of the second projection data set; and a reduction unit that on the basis of the first reconfiguration image and the second reconfiguration image, generates a third reconfiguration image corresponding to the second timing and having an artifact reduced more than that of the second reconfiguration image.SELECTED DRAWING: Figure 1

Description

Embodiments of the present invention relate to a medical image processing apparatus, a medical image processing program, an X-ray diagnostic apparatus, and an X-ray diagnostic system.

The X-ray circulatory organ diagnostic apparatus has an observation technique by polymorphic imaging using rotary imaging in order to diagnose cancer in the liver. In polyphasic imaging, the same site is photographed more than once, so the subject must hold his breath each time. When the subject is particularly elderly, breath-holding becomes a burden, and motion artifacts due to poor breath-holding may occur in the reconstructed image.

Japanese Unexamined Patent Publication No. 2017-22146 International Publication No. 2017/98888 Japanese Unexamined Patent Publication No. 2006-150080

An object to be solved by the present invention is to efficiently reduce artifacts associated with polymorphic imaging using rotary imaging.

The medical image processing apparatus according to the embodiment includes a first projection data set generated by rotational imaging performed at a first timing on a subject injected with a contrast agent, and the subject. An acquisition unit that acquires a second projection data set generated by rotation imaging executed at a second timing different from the first timing, and a first reconstruction based on the first projection data set. The reconstruction unit that generates an image and generates a second reconstruction image based on the second projection data set, and the first reconstruction image and the second reconstruction image. It includes a reduction unit that generates a third reconstructed image that corresponds to the second timing and has fewer artifacts than the second reconstructed image.

FIG. 1 is a diagram showing a configuration of an X-ray diagnostic apparatus according to the present embodiment. FIG. 2 is a diagram showing the appearance of the X-ray imaging unit of FIG. FIG. 3 is a diagram schematically showing a reconstructed image relating to the early phase. FIG. 4 is a diagram schematically showing a reconstructed image relating to the late phase. FIG. 5 is a diagram showing an example of input / output of the trained model used in the artifact reduction function. FIG. 6 is a diagram showing a typical flow of multiphase radiography by an X-ray diagnostic apparatus. FIG. 7 is a diagram showing an example of a display screen of an early phase image and a late phase image. FIG. 8 is a diagram showing an example of early phase position data. FIG. 9 is a diagram showing an example of late phase position data. FIG. 10 is a diagram showing an example of input / output of the trained model used in step S9. FIG. 11 is a diagram showing an example of input / output of a trained model for generating position data. FIG. 12 is a diagram showing an example of input / output of another trained model that outputs a late phase image with reduced motion artifacts. FIG. 13 is a diagram showing a configuration of an X-ray diagnostic system according to the present embodiment.

Hereinafter, the medical image processing apparatus, the medical image processing program, the X-ray diagnostic apparatus, and the X-ray diagnostic system according to the present embodiment will be described with reference to the drawings.

The medical image processing apparatus according to the present embodiment is a computer that processes data collected by the X-ray diagnostic apparatus. The medical image processing apparatus according to the present embodiment may be a computer included in the X-ray diagnostic apparatus or a computer separate from the X-ray diagnostic apparatus. Hereinafter, the medical image processing apparatus is assumed to be a computer included in the X-ray diagnostic apparatus.

FIG. 1 is a diagram showing a configuration of an X-ray diagnostic apparatus 1 according to the present embodiment. As shown in FIG. 1, the X-ray diagnostic apparatus 1 has an X-ray photographing unit 3 and a console 5. The console 5 functions as a medical image processing device. For example, the X-ray imaging unit 3 is installed in an imaging room, and the console 5 is installed in a control room adjacent to the imaging room.

As shown in FIG. 1, the X-ray photographing unit 3 has an X-ray tube holding device 10 and a sleeper device 30. As shown in FIG. 1, the X-ray tube holding device 10 is a mechanical device that holds the X-ray tube 11 and the X-ray detector 13. The X-ray tube holding device 10 may be either a ceiling-mounted type or a floor-standing type. As the floor-standing type, a cardiovascular radiography type that movably holds the C-arm 15 equipped with the X-ray tube 11 and the X-ray detector 13 is preferable. Further, the X-ray tube holding device 10 may be a single plane type including either a floor-standing type X-ray tube holding device or a ceiling-suspended type X-ray tube holding device, or a floor-standing type X-ray tube holding device 10. Any biplane method including both a wire tube holding device and a ceiling-suspended type X-ray tube holding device may be used. Hereinafter, the X-ray tube holding device is assumed to be a single plane type floor-standing type as an example.

FIG. 2 is a diagram showing the appearance of the X-ray imaging unit 3 of the X-ray diagnostic apparatus 1 according to the present embodiment. As shown in FIGS. 1 and 2, the X-ray tube holding device 10 includes, for example, an X-ray tube 11, an X-ray detector 13, a C arm 15, and a C arm holding system 17. The X-ray tube 11 is connected to the high voltage generator 12 via a high voltage cable. The high voltage generator 12 generates a high voltage to be applied to X-rays in response to a control signal from the system control circuit 56 of the console 5. The X-ray tube 11 generates X-rays by receiving a high voltage applied from the high voltage generator 12. The X-ray detector 13 detects the X-rays generated from the X-ray tube 11. Specifically, it is assumed that the X-ray detector 13 is an FPD (Flat Panel Detector) having a plurality of detector pixels arranged in a two-dimensional plane, a reading circuit, and an A / D conversion circuit. Each detector pixel detects X-rays generated from the X-ray tube 11 and generates an electric signal (detection signal) according to the intensity of the detected X-rays. The read circuit reads a detection signal from a plurality of detector pixels at a predetermined timing in frame units. The A / D conversion circuit A / D-converts the detection signals read from the plurality of detector pixels and generates projection data related to the subject in frame units. The generated projection data is transmitted to the storage device 52 of the console 5.

As shown in FIGS. 1 and 2, the X-ray tube 11 and the X-ray detector 13 are attached to the C arm 15. The C-arm 15 is a support structure equipped with an X-ray tube 11 and an X-ray detector 13. An X-ray tube 11 is attached to one end of the C arm 15, and an X-ray detector 13 is attached to the other end of the C arm 15 facing the X-ray tube 11. The C-arm 15 is movably held by the C-arm holding system 17. The C-arm holding system 17 holds the C-arm 15 three-dimensionally so as to be movable. The C-arm holding system 17 is designed according to the degree of freedom of movement of the C-arm 15, and is composed of a plurality of structures (hereinafter, referred to as holding structures) that are intertwined with each other. The C-arm holding system 17 rotatably supports the X-ray tube 11 and the X-ray detector 13 around the rotation axis AR via the C-arm 15 in order to perform at least rotational imaging. The axis connecting the focal point of the X-ray tube 11 and the center of the detection surface of the X-ray detector 13 is called the imaging axis AI. The intersection IC between the rotation axis AR and the imaging axis AI is called an isocenter.

The rotation of the X-ray tube 11 and the X-ray detector 13 around the rotation axis AR may be realized by the so-called propeller rotation of the C arm 15, or by the orbital rotation that slides along the C shape of the C arm 15. May be done.

As shown in FIG. 1, the X-ray tube holding device 10 has a C-arm drive system 19 that generates power for driving the X-ray tube holding device 10. Typically, the C-arm drive system 19 is mounted on the C-arm holding system 17. For example, the C-arm drive system 19 corresponding to the C-arm holding system 17 illustrated in FIG. 2 has a drive device for rotating the C-arm 15 around the rotation axis AR. The drive device receives, for example, a drive signal from the system control circuit 56 of the console 5 and drives the drive device. For each drive device, for example, an existing motor such as a servo motor is used.

As shown in FIGS. 1 and 2, a sleeper 31 is provided in the photographing room. The sleeper 31 has a base 37 and a top plate 39. The base 37 is installed on the floor surface and movably supports the top plate 39. The subject is placed on the top plate 39. A sleeper drive system 33 is mounted on the base 37. The sleeper drive system 33 has, for example, a drive device for sliding the top plate 39 in the longitudinal direction and a drive device for raising and lowering the top plate 39 in the vertical direction. Each drive device receives, for example, a drive signal from the system control circuit 56 of the console 5 and drives the device.

The X-ray imaging unit 3 according to the present embodiment executes multi-phase imaging using rotation imaging as described above. In the rotation imaging, the C-arm drive system 19 rotates the C-arm 15 at a predetermined speed according to the instruction of the system control circuit 56, and when the C-arm 15 is rotating, the high voltage generator 12 emits X-rays according to the instruction of the system control circuit 56. X-rays are generated from the tube 11. The C-arm 15 is rotated by the rotation angle range required for image reconstruction. The rotation angle range (hereinafter referred to as the reconstruction angle range) required for image reconstruction is, for example, 360 degrees, but any angle range may be used as long as it is 180 degrees + fan angle or more. The X-ray detector 13 detects the X-rays generated from the X-ray tube 11 and transmitted through the subject, and generates projection data according to the intensity of the detected X-rays. The projection data for the reconstruction angle range will be called the projection data set. A reconstructed image is generated by performing an image reconstruction process on the projection data set.

Polymorphic imaging is used, for example, to contrast and observe cancer that develops in an organ such as the liver with a contrast medium. In polymorphic imaging, after the contrast medium is injected into the subject, rotational imaging is performed at a plurality of timings. For example, in polyphase imaging in which the liver is the subject of examination, rotational imaging is performed at the timing of the early phase and the timing of the late phase. In the early phase, the contrast agent accumulates in the cancer via the feeding vessels and is visualized as a deep tumor stain, and in the late phase, the contrast agent flows out of the cancer and is visualized as a deep corona stain or washout.

FIG. 3 is a diagram schematically showing a reconstructed image I1 regarding the early phase. FIG. 4 is a diagram schematically showing a reconstructed image I2 relating to the late phase. 3 and 4 show, as an example, a reconstructed image of liver cancer as an imaging target. As shown in FIGS. 3 and 4, the reconstructed images I1 and I2 have an image region (hereinafter, referred to as a liver region) R1 relating to the liver. The liver region has an image region (hereinafter referred to as a cancer region) R2 relating to liver cancer. Since the reconstructed image I1 is collected in the early phase, the cancer region R2 in the reconstructed image I1 is more strongly imaged by the contrast agent than the cancer region R2 in the reconstructed image I2. The cancerous region in the reconstructed image I1 relating to the early phase is sometimes referred to as a tumor-dense region. Since the reconstructed image I2 is collected in the late phase, the cancer region R2 in the reconstructed image I2 is not contrasted with the cancer region R2 in the reconstructed image I1. The reconstructed image I2 has an image region related to corona deep staining (hereinafter referred to as corona deep staining region) R3 around the cancer region R2. Liver cancer and the like are diagnosed based on the contrast medium deep staining pattern by polymorphic imaging.

As shown in FIG. 1, the console 5 is a computer that controls an X-ray diagnostic apparatus. The console 5 has a processing circuit 51, a storage device 52, an input device 53, a communication device 54, and a display device 55, with the system control circuit 56 as the center. The processing circuit 51, the storage device 52, the input device 53, the communication device 54, the display device 55, and the system control circuit 56 are communicably connected to each other via a bus.

The processing circuit 51 has a processor as a hardware resource. The processing circuit 51 executes the processing program according to the present embodiment, and realizes the acquisition function 511, the reconstruction function 512, the position identification function 513, the artifact reduction function 514, and the display control function 515. It should be noted that each function 511 to 515 is not limited to the case where it is realized by a single processing circuit. A processing circuit may be configured by combining a plurality of independent processors, and each function 511 to 515 may be realized by executing a program by each processor.

In the acquisition function 511, the processing circuit 51 includes the first projection data set generated by the rotation imaging performed at the first timing on the subject into which the contrast medium is injected, and the first projection data set on the subject. Acquires the second projection data set generated by the rotation imaging executed at the second timing different from the timing of. For example, the first timing is the early phase and the second timing is the late phase. The second-timing projection dataset shall contain stronger and / or more artifact components than the first-timing projection dataset.

In the reconstruction function 512, the processing circuit 51 performs reconstruction processing on the projection data set to generate a reconstruction image. Specifically, the processing circuit 51 performs preprocessing such as gain correction and offset correction on the projection data set from the X-ray detector 13, and reconstructs the projection data set after the preprocessing. As the reconstruction process, a filter correction back projection method, a successive approximation reconstruction method, an image reconstruction using machine learning, or the like can be appropriately used. The processing circuit 51 generates a reconstructed image of the first timing based on the projection data set of the first timing, and generates a reconstructed image of the second timing based on the projection data set of the second timing. .. The reconstructed image of the second timing shall contain stronger and / or more artifacts than the reconstructed image of the first timing.

Unless otherwise specified, the reconstructed image according to the present embodiment is generated by performing three-dimensional image processing on the three-dimensional image data (volume data) generated by performing reconstruction processing on the projection data and the three-dimensional image data. It is assumed that the concept includes both the two-dimensional image data obtained. The three-dimensional image processing is an arbitrary process for converting three-dimensional image data into two-dimensional image data, such as MPR processing, volume rendering processing, surface rendering processing, and pixel value projection processing. It is assumed that the three-dimensional image processing is realized by the reconstruction function 512.

In the position specifying function 513, the processing circuit 51 identifies various image regions included in the reconstructed image by image processing. For example, the processing circuit 51 identifies the position of the tumor-developing tissue region in the reconstructed image of the first timing and the position of the tumor-developing tissue region in the reconstructed image of the second timing. The tumorigenic tissue area is an image area relating to the tissue in which the tumor has developed. For example, when cancer develops in the liver, the tumorigenic tissue region is the liver region. In addition, the processing circuit 51 specifies the position of the tumor deep-staining region included in the reconstructed image of the first timing and the position of the corona deep-staining region included in the reconstructed image of the second timing.

In the artifact reduction function 514, the processing circuit 51 corresponds to the second timing based on the reconstructed image of the first timing and the reconstructed image of the second timing, and is more than the reconstructed image of the second timing. Generate a reconstructed image with reduced artifacts. Hereinafter, the reconstructed image generated by the artifact reduction function 514 will be referred to as a corrected image. The processing circuit 51 determines the position of the tumor-developing tissue region in the reconstructed image of the first timing, the reconstructed image of the second timing, and the reconstructed image of the first timing, and the tumor development in the reconstructed image of the second timing. A corrected image for the second timing may be generated based on the position of the tissue area. The processing circuit 51 determines the position of the tumor-dense region in the reconstructed image of the first timing, the reconstructed image of the second timing, and the reconstructed image of the first timing, and the corona density in the reconstructed image of the second timing. A corrected image for the second timing may be generated based on the position of the dyed area.

In the display control function 515, the processing circuit 51 displays various data via the display device 55. For example, the processing circuit 51 displays the reconstructed image generated by the reconstructed function 512, the corrected image generated by the artifact reduction function 514, and the like via the display device 55.

The storage device 52 is a storage device such as an HDD (Hard Disk Drive), an SSD (Solid State Drive), or an integrated circuit storage device that stores various data. For example, the storage device 52 stores the projection data transmitted from the X-ray detector 13, the reconstructed image generated by the reconstructed function 512, and the corrected image generated by the artifact reduction function 514.

The input device 53 receives various input operations from the operator, converts the received input operations into electric signals, and outputs the received input operations. As the input device 53, for example, a mouse, a keyboard, a trackball, a switch, a button, a joystick, a touch pad, a touch panel display, or the like can be appropriately used. The input device 53 may be not only a physical device but also a computer terminal connected via the communication device 54, a tablet terminal capable of wireless communication, and the like.

The communication device 54 performs data communication with another computer via a wired or wireless device (not shown). For example, the communication device 54 communicates projection data, a reconstructed image, a corrected image, and the like.

The display device 55 displays various data by realizing the display control function 515 of the processing circuit 51. As the display device 55, for example, a CRT display, a liquid crystal display, an organic EL display, an LED display, a plasma display, or any other display known in the art can be appropriately used. Further, the display device 55 may be a projector. For example, the display device 55 displays a reconstructed image generated by the reconstructed function 512, a corrected image generated by the artifact reduction function 514, and the like.

The system control circuit 56 has a processor as a hardware resource. The system control circuit 56 controls the entire X-ray diagnostic apparatus 1 by executing a multiphase radiography program. The system control circuit 56 controls the high voltage generator 12 and the C-arm drive system 19 in conjunction with each other in order to perform rotational imaging. In rotary imaging, the system control circuit 56 supplies a control signal to the high voltage generator 12 so as to perform X-ray exposure corresponding to preset X-ray conditions. The high voltage generator 12 to which the control signal is supplied applies a high voltage corresponding to the control signal to the X-ray tube 11, and the X-ray tube 11 generates X-rays when the high voltage is applied. In the rotation imaging, the system control circuit 56 controls the C-arm drive system 19 in order to rotate the C-arm. Further, the system control circuit 56 may control the bed drive system 33 for positioning the top plate 39.

The configuration of the X-ray diagnostic apparatus 1 is an example, and the present invention is not limited to this, and various modifications are possible. For example, the system control circuit 56 and the processing circuit 51 may be configured by a single circuit instead of a separate circuit. Further, the processing circuit 51, the storage device 52, the input device 53, the communication device 54, the display device 55, and the system control circuit 56 do not all have to be mounted on a single computer, but are mounted on a plurality of computers in a distributed manner. May be done.

Next, an operation example of the X-ray diagnostic apparatus 1 will be described.

In the following operation example, it is assumed that the first timing in polyphase imaging is the early phase and the second timing is the late phase. The artifact to be reduced by the artifact reduction function 514 may be any artifact, but it is assumed that the artifact is a motion artifact caused by the body movement of the subject during rotational photography. Motion artifacts can occur in both the reconstructed image for the early phase and the reconstructed image for the late phase. Rotational imaging of the late phase is performed several seconds to several tens of seconds after the rotational imaging of the early phase. Therefore, the subject is more likely to move during the late phase rotary imaging than during the early phase rotary imaging. Therefore, in the following operation example, it is assumed that motion artifacts are significantly generated in the reconstructed image related to the late phase. Hereinafter, the reconstructed image relating to the early phase is referred to as an early phase image, and the reconstructed image relating to the late phase is referred to as a late phase image.

As described above, in the artifact reduction function 514, the processing circuit 51 generates a corrected image relating to the late phase in which motion artifacts are reduced as compared with the late phase image, based on the early phase image and the late phase image. Specifically, the processing circuit 51 does not include the correlation between the pre-generated early phase image containing no motion artifact and the late phase image not containing the motion artifact, and the late phase image including the motion artifact and the motion artifact. Correlation with the late phase image is used to generate a corrected image for the late phase from the early phase image and the late phase image for the subject. The correlation is learned by, for example, a machine learning method. As such a machine learning algorithm, it is assumed that a deep neural network (DNN), which is a multi-layer network model that imitates a neural circuit of a living organism's brain. DNN includes a composite function with parameters defined by a combination of multiple adjustable functions and parameters. A machine learning model or DNN in which parameters are trained will be referred to as a trained model. The trained model according to this embodiment can have various structures. For example, the trained model according to this embodiment includes an input layer, an output layer, and an intermediate layer. The intermediate layer includes at least one convolutional layer, a fully connected layer, a pooling layer, and the like. The trained model may be a network structure such as ResNet (Residual Network), DenseNet (Dense Convolutional Network), or U-Net. The term "trained model" does not preclude additional learning. For example, the adjusted parameters may be further updated by subjecting the trained model to additional training as appropriate.

FIG. 5 is a diagram showing an example of input / output of the trained model used in the artifact reduction function 514. As shown in FIG. 5, the trained model inputs an early phase image without motion artifacts and a late phase image with motion artifacts as input data. As described above, the early phase image depicts a deep tumor stain, and the late phase image depicts a deep corona stain. As output data, the trained model outputs a late phase image that does not include motion artifacts. In the following description, an image without artifacts means that the image contains weak and / or less artifacts as compared with other images, or contains weaker and / or less artifacts than a predetermined standard, and is an image with artifacts. Means that it contains strong and / or more artifacts than other images or contains stronger and / or more artifacts than a predetermined standard.

Here, the generation of the trained model of FIG. 5 will be described with reference to FIG. The trained model is generated by the processing circuit 51 or another computer. The processing circuit 51 or another computer that generates the trained model will be referred to as a learning device.

The learning device learns DNN parameters based on a plurality of learning samples. Each training sample includes a combination of an early phase image without motion artifacts and a late phase image with motion artifacts as input data, and a late phase image without motion artifacts as teacher data (hereinafter referred to as a correct late phase image). )including. These learning samples are generated by rotating the subject with an X-ray diagnostic apparatus. Each image included in the same learning sample is generated by rotating the same subject. It is not necessary for the plurality of learning samples to be the same subject, and the learning samples may be collected by rotating and photographing various subjects.

As each learning sample, for example, data when re-imaging is performed can be used. That is, when the rotation imaging of the early phase and the late phase is performed and then the rotation imaging of the late phase is performed again, the early phase image without motion artifact, the late phase image with motion artifact, and the late phase without motion artifact are performed. A phase image is obtained. These sets can be used as learning samples.

It should be noted that not all the learning samples need to be collected by rotary photography. For example, a learning sample may be collected by an X-ray computed tomography apparatus that performs CT imaging, which has substantially the same imaging principle as rotation imaging. Further, not all the learning samples need to be actually measured learning samples, and may be a calculated learning sample calculated by a computer simulation simulating rotation imaging or CT imaging. Further, a semi-actually measured learning sample generated by subjecting the actually measured learning sample to arbitrary image processing may be used.

The early phase image, the late phase image, and the correct late phase image included in the training sample may be three-dimensional image data or two-dimensional image data. In the case of two-dimensional image data, the cross section may be in any direction of an axial cross section, a sagittal cross section, a coronal cross section, and an oblique cross section. In addition, two-dimensional image data of one cross section intersecting the tumor deep staining region or corona deep staining region may be used for input and output, or intersects the tumor deep staining region or corona deep staining region at various positions and orientations. Two-dimensional image data of a plurality of cross sections may be used for input and output. Pixel value projection images such as maximum value projection images and average value projection images based on two-dimensional image data of a plurality of cross sections may be used for input and output.

In the process of generating the trained model, the learning device applies the early phase image without motion artifacts and the late phase image with motion artifacts to the machine learning model, performs forward propagation processing, and outputs the estimated late phase image. Next, the learning device applies the machine learning model to the difference (error) between the estimated late phase image and the correct late phase image, performs back propagation processing, and calculates the gradient vector. The learning device then updates parameters such as the weighted matrix and bias of the machine learning model based on the gradient vector. The trained model is completed by repeating the forward propagation process and the back propagation process for a large number of training samples and updating the parameters. The trained model after completion is stored in the storage device 52.

If there are no motion artifacts in both the early and late phase images, the difference between the early and late phase images is ideally due to the flow of contrast agent during the period from early to late phase. This is the difference in pixel values. Further, the difference between the late phase image with motion artifacts and the late phase image without motion artifacts is ideally the difference in pixel values due to the presence or absence of motion artifacts. Therefore, by performing machine learning based on the combination of the early phase image without motion artifact and the late phase image with motion artifact and the late phase image without motion artifact, contrast in the period from the early phase to the late phase is performed. It is possible to make the DNN learn the difference in the pixel value due to the flow of the agent and the difference in the pixel value due to the presence or absence of the motion artifact. By using the trained model thus generated, the processing circuit 51 maintains the contrast agent region contained in the late phase image and is a motion artifact caused by the body movement in the rotational imaging performed in the late phase. Can be reduced.

By the way, it is also conceivable to create a trained model that outputs a late phase image without motion artifacts by inputting only a late phase image with motion artifacts without inputting an early phase image without motion artifacts. Even in this case, it is possible to reduce motion artifacts, but there is a risk that processing artifacts will occur in the trained model. On the other hand, the trained model according to the present embodiment can reduce the risk of processing artifacts in the trained model by further inputting an early phase image without motion artifacts.

The method of generating the trained model is not limited to supervised learning, and the trained model may be generated using a method such as a hostile generation network (GAN: Generative Adversarial Nets) or pix2pix.

The trained model of FIG. 5 is an example, and the input / output relationship and the like are not limited to those shown in FIG. For example, the trained model may input a late phase image without motion artifacts and an early phase image with motion artifacts, and output an early phase image without motion artifacts (with reduced motion artifacts). Further, each of the input and output images may be three-dimensional image data or two-dimensional image data.

In addition, the input of the trained model is not limited to the early phase image and the late phase image, but the position of the tumor deeply stained region, the position of the corona deeply stained region, and the position of the tumor-developing tissue region specified by the position identification function 513. Information about at least one of them may be supplementarily entered. As a result, a region to be substantially matched between the early phase image and the corrected late phase image, which is not affected by the contrast medium, is specified, and more accurate correction of artifacts becomes possible.

Next, multiphase radiography with the X-ray diagnostic apparatus 1 will be described with liver cancer as a clinical example.

FIG. 6 is a diagram showing a typical flow of multiphase radiography by the X-ray diagnostic apparatus 1. In the multiphase imaging shown in FIG. 6, the trained model inputs the early phase image, the late phase image, the position of the tumor deeply stained region, the position of the corona deeply stained region, and the position of the tumor-developing tissue region. The details of the trained model will be described later. When the subject is placed on the bed 31 and the contrast medium is injected and it is determined that the contrast medium is in the early phase timing, the user who is a medical worker presses the imaging switch. When the imaging switch is pressed, the system control circuit 56 executes the polyphase imaging program and starts the process shown in FIG.

First, the system control circuit 56 controls the high voltage generator 12 and the C-arm drive system 19, performs rotational imaging in the early phase, and collects a projection data set related to the early phase via the X-ray detector 13 (step). S1). The collected projection data set is stored in the storage device 52, and is acquired by the processing circuit 51 when the acquisition function 511 of the processing circuit 51 is realized.

When step S1 is performed, the processing circuit 51 generates an early phase image based on the projection data set regarding the early phase by realizing the reconstruction function 512 (step S2). More specifically, in step S2, the processing circuit 51 generates three-dimensional image data based on the projection data set for the early phase, extracts the tumor-dense region from the three-dimensional image data by image processing, and extracts the three-dimensional image data. Generate a cross-sectional image of the cross-section including the tumor-dense area. The cross section may be any of an axial cross section, a coronal cross section and a sagittal cross section, and may be an oblique cross section. In this embodiment, the cross section includes the concept of both a slice having a thickness of 1 voxel and a slab having a thickness of 2 voxels or more, unless otherwise specified. When the cross section is a slab, it is preferable that the maximum value projection image generated by performing the maximum value projection process only on the slab is generated as the cross section image. The maximum value projection process is not limited to the maximum value projection process, and the average value projection process, the intermediate value device process, and the minimum value projection process may be performed depending on the properties of the contrast medium and the like. Hereinafter, the early phase image is assumed to be a two-dimensional image relating to a slice containing a tumor-dense region.

The system control circuit 56 controls the high voltage generator 12 and the C-arm drive system 19 again after a certain time interval from the rotation imaging of the early phase, performs the rotation imaging in the late phase, and performs the rotation imaging via the X-ray detector 13. A projection dataset for the early phase is collected (step S3). The certain time interval may be a value empirically determined as the time interval from the early phase to the late phase, a value calculated by simulation, or a value determined in real time by monitoring the contrast medium concentration. Good. The collected projection data set is stored in the storage device 52, and is acquired by the processing circuit 51 when the acquisition function 511 of the processing circuit 51 is realized.

When step S3 is performed, the processing circuit 51 generates a late phase image based on the projection data set for the late phase by realizing the reconstruction function 512 (step S4). More specifically, in step S4, the processing circuit 51 generates three-dimensional image data based on the projection data set for the late phase, extracts the corona dark-stained region from the three-dimensional image data by image processing, and extracts the three-dimensional image data. Generate a cross-sectional image of the cross section including the corona dark dyed area. Hereinafter, the late phase image is assumed to be a two-dimensional image relating to a slice including a corona deep-stained region.

When step S4 is performed, the processing circuit 51 displays the early phase image generated in step S2 and the late phase image generated in step S4 by realizing the display control function 515 (step S5).

FIG. 7 is a diagram showing an example of the display screen IS1 of the early phase image I1 and the late phase image I2. As shown in FIG. 7, the early phase image I1 and the late phase image I2 are arranged side by side on the display screen IS1. The user makes a diagnosis of liver cancer by comparing the early phase image I1 and the late phase image I2. On the display screen IS1, a "moving" button B11 and a "no movement" button B12 for the early phase image I1 and a "moving" button B21 and a "no movement" button B22 for the late phase image I2 are arranged. Buttons B11, B12, B21 and B22 are graphical user interface (GUI) buttons that can be pressed via an input device 53 such as a mouse. When the user observes the early phase image I1 and determines that a motion artifact has occurred in the early phase image I1, the user presses the "moving" button B11 and determines that no motion artifact has occurred. , Press the "no movement" button B12. Similarly, when the user observes the late phase image I2 and determines that a motion artifact has occurred in the late phase image I2, he presses the "moving" button B21 and says that no motion artifact has occurred. If it is determined, the "no movement" button B22 is pressed.

The processing circuit 51 determines that the motion artifact of the early phase image I1 is reduced when the "moving" button B11 and the "no moving" button B22 are pressed, and the "moving" button B21 and the "no movement" button B12. When is pressed, it is determined that the motion artifact of the late phase image I2 is reduced (step S6: YES). On the other hand, the processing circuit 51 means that when the "no movement" button B12 and the "no movement" button B22 are pressed, no motion artifact occurs in both the early phase image I1 and the late phase image I2. It is determined that the motion artifact is not reduced (step S6: NO).

On the other hand, the processing circuit 51 means that when the "moving" button B11 and the "moving" button B21 are pressed, motion artifacts occur in both the early phase image I1 and the late phase image I2. It is determined that the motion artifact is not reduced (step S6: NO). In this case, the user causes the system control circuit 56 to move the C arm 15 to the initial position again and execute steps S1 to S4 to perform rotational imaging for the early phase and the late phase. As shown in FIG. 7, the "re-shooting" button B3 may be arranged on the display screen IS2. The "re-shooting" button B3 is a GUI button that can be pressed via an input device 53 such as a mouse. Even when the "reshoot" button B3 is pressed, the system control circuit 56 rotates the C arm 15 to the initial position again and executes steps S1 to S4 to rotate the early phase and the late phase. Take a picture. At this time, the early phase image I1 and the late phase image I2 obtained by the rotation imaging before the re-imaging and the reconstructed image obtained by the re-imaging are stored in the storage device 52 as learning samples, or a communication device. It may be output to another device such as a server through 54.

In the following, it is assumed that the early phase image has no motion artifact and the late phase image has motion artifact.

When it is determined in step S6 to reduce the motion artifact (step S6: YES), the processing circuit 51 realizes the position identification function 513, and the position data of the liver region and the tumor deeply stained region from the early phase image (hereinafter, early stage). (Called phase position data) is generated (step S7).

FIG. 8 is a diagram showing an example of early phase position data I3. In step S7, the processing circuit 51 performs image processing on the early phase image to identify the position of the liver region and the position of the tumor deeply stained region, and determines the position of the liver region and the position of the tumor deeply stained region in the early phase image. Image data showing the spatial distribution is generated as early phase position data. For example, "2" is assigned to the pixel corresponding to the tumor deep staining region, "1" is assigned to the pixel corresponding to the liver region, and "0" is assigned to the pixel corresponding to the other background region. Be done. The early phase position data I3 is stored in the storage device 52. The image processing is not particularly limited as long as each area can be specified, and any method such as template matching, threshold value processing, image recognition, and machine learning may be used. The matrix size of the early phase position data I3 is typically the same as the matrix size of the early phase image.

The pixel values "2", "1" and "0" are examples, and any value may be used as long as the tumor dark staining region, the liver region and the background region can be distinguished. Further, the matrix size of the early phase position data I3 is not limited to the same as the matrix size of the early phase image, and may be reduced or enlarged by a predetermined magnification. Further, the early phase position data I3 does not have to be image data, and may be numerical data indicating the coordinates of the tumor-dense region, the liver region, and the background region.

When step S7 is performed, the processing circuit 51 generates position data (hereinafter, referred to as late phase position data) of the liver region and the corona deep-stained region from the late phase image by realizing the position identification function 513 (step S8). ..

FIG. 9 is a diagram showing an example of late phase position data I4. In step S8, the processing circuit 51 performs image processing on the late phase image to specify the position of the liver region and the position of the corona deep-stained region, and determines the position of the liver region and the position of the corona deep-stained region in the late-phase image. Image data showing the spatial distribution is generated as late phase position data. For example, "3" is assigned to the pixel corresponding to the corona dark dyed region, "1" is assigned to the pixel corresponding to the liver region, and "0" is assigned to the pixel corresponding to the other background region. Be done. The late phase position data I4 is stored in the storage device 52. The image processing in step S8 is the same as the image processing in step S7. The matrix size of the late phase position data I4 is typically the same as the matrix size of the late phase image.

The pixel values "3", "1", and "0" are examples, and any value may be used as long as the corona deep-stained region, the liver region, and the background region can be distinguished. Further, the matrix size of the late phase position data I4 is not limited to the same as the matrix size of the late phase image, and may be reduced or enlarged by a predetermined magnification. Further, the early phase position data I3 does not have to be image data, and may be numerical data indicating the coordinates of the corona deep-stained region, the liver region, and the background region.

When step S8 is performed, the processing circuit 51 applies the trained model to the early phase image, the late phase image, the early phase position data, and the late phase position data by realizing the artifact reduction function 514, and the motion artifact is reduced. A corrected image relating to the late phase is generated (step S9).

FIG. 10 is a diagram showing an example of input / output of the trained model used in step S9. As shown in FIG. 10, the trained model inputs early phase images without motion artifacts, late phase images with motion artifacts, early phase position data, and late phase position data as input data. As output data, the trained model outputs a late phase image without motion artifacts (with reduced motion artifacts).

Here, the generation of the trained model of FIG. 10 will be described with reference to FIG. The method of generating the trained model of FIG. 10 is the same as that of the trained model of FIG. 5, except for the contents of the training sample. For the sake of simplicity, only the difference from the trained model of FIG. 5 will be described.

The learning device learns DNN parameters based on a plurality of learning samples. Each training sample contains a combination of an early phase image without motion artifacts, a late phase image with motion artifacts, an early phase position data, and a late phase position data as input data, and a late phase without motion artifacts as teacher data. Includes images (correct late phase images). In the process of generating the trained model, the learning device applies the early phase image without motion artifact, the late phase image with motion artifact, the early phase position data, and the late phase position data to the machine learning model to perform forward propagation processing. , Output the estimated late phase image. Next, the learning device applies the machine learning model to the difference (error) between the estimated late phase image and the correct late phase image, performs back propagation processing, and calculates the gradient vector. The learning device then updates parameters such as the weighted matrix and bias of the machine learning model based on the gradient vector. The trained model is completed by repeating the forward propagation process and the back propagation process for a large number of training samples and updating the parameters. The trained model after completion is stored in the storage device 52.

Compared to the trained model of FIG. 5, the trained model of FIG. 10 further includes early phase position data indicating the positions of the liver region and the tumor deeply stained region as input data, and the positions of the liver region and the corona deeply stained region. Enter the late phase position data indicating. As a result, the trained model can easily distinguish the positions of the liver region and the tumor deeply stained region in the early phase from the positions of the liver region and the corona deeply stained region in the late phase, thereby improving the accuracy of motion artifact reduction. , Learning efficiency can be improved. By using the trained model thus generated, the processing circuit 51 maintains the contrast agent region contained in the late phase image, while maintaining the motion artifacts caused by the body movements in the rotational imaging performed in the late phase. Can be reduced with high accuracy.

When step S9 is performed, the processing circuit 51 causes the display device 55 to display the corrected image relating to the late phase by realizing the display control function 515 (step S10). In step S10, the display device 55 displays, for example, the early phase image generated in step S2 and the corrected image related to the late phase generated in step S9 side by side. Since the corrected image generated in step S9 has less motion artifacts than the late phase image generated in step S4, it is expected that liver cancer can be diagnosed more accurately and efficiently. ..

This completes the polymorphic imaging shown in FIG.

The processing flow shown in FIG. 6 is an example, and is not limited to this. For example, step S7 may be performed after step S2, and similarly step S8 may be performed after step S4. Further, step S7 may be performed after step S8.

In FIG. 6, the early phase position data is data showing the positions of both the liver region and the tumor-dense region, but it is the data showing the positions of only one of the liver region and the tumor-dense region. May be good. Similarly, although the late phase position data is data indicating the positions of both the liver region and the corona deeply stained region, it may be data indicating the positions of only one of the liver region and the corona deeply stained region. ..

In the above embodiment, two-dimensional image data is input to the trained model, but three-dimensional image data may be input. Further, even when the two-dimensional image data is input to the trained model, the two-dimensional image data relating to a plurality of cross sections used for diagnosis may be input to the trained model one by one. Alternatively, two-dimensional image data relating to a single cross section used for diagnosis may be input to the trained model. This makes it possible to make a diagnosis efficiently.

In the above embodiment, it is assumed that the early phase position data is generated from the early phase image and the late phase position data is generated from the late phase image. However, both late phase position data and early phase position data may be generated from the early phase image. This process is realized by, for example, a machine learning process.

FIG. 11 is a diagram showing an example of input / output of a trained model for generating position data. As shown in FIG. 11, the trained model inputs an early image in which the tumor deep staining is visualized as input data. As output data, the trained model outputs early phase position data and late phase position data. The output early phase position data corresponds to the positions of the liver region and the tumor-dense region included in the input early phase image, and the output late phase position data is the late phase liver inferred from the input early phase image. Corresponds to the location of the area and the corona dark dyed area.

Here, the generation of the trained model of FIG. 11 will be described with reference to FIG. The method of generating the trained model of FIG. 11 is the same as that of the trained model of FIGS. 5 and 10 except for the contents of the training sample. For the sake of simplicity, only the differences from the trained models of FIGS. 5 and 10 will be described.

The learning device learns DNN parameters based on a plurality of learning samples. Each learning sample includes an early phase image as input data, and as teacher data, early phase position data (hereinafter referred to as correct early phase position data) and late phase position data (hereinafter referred to as correct late phase position data). Including combinations with. The early phase image of the training sample is, for example, an early phase image obtained by polyphase imaging using rotary imaging. The early phase image of the training sample may have motion artifacts or no motion artifacts. The early phase position data of the training sample is generated from the input early phase image according to image processing or user instruction. The late phase position data of the training sample is a late phase image obtained by polyphase imaging using rotation imaging, and it is desirable that the data is collected following the input early phase image. The early phase image and the late phase image of the learning sample do not have to belong to the same polyphase imaging, and may be collected by separate polyphase imaging.

In the process of generating the trained model, the learning device applies the early phase image to the machine learning model, performs forward propagation processing, and outputs the estimated early phase position data and the estimated late phase position data. Next, the learning device applies the machine learning model to the difference (error) between the combination of the estimated early phase position data and the estimated late phase position data and the combination of the correct early phase position data and the correct late phase position data, and back-propagates. Perform processing and calculate the gradient vector. The learning device then updates parameters such as the weighted matrix and bias of the machine learning model based on the gradient vector. The trained model is completed by repeating the forward propagation process and the back propagation process for a large number of training samples and updating the parameters. The trained model after completion is stored in the storage device 52.

Since the early phase image includes the liver region and the tumor deeply stained region, it can be said that there is a correlation between the early phase image and the early phase position data. The difference between the liver region and the tumor-dense region included in the early phase image and the late phase position data is that the position of the liver region is ideally immobile, so that the contrast medium in the period from the early phase to the late phase It is a change in the contrast agent region due to the flow. By using the trained model generated in this way, the processing circuit 51 can output the early phase position data and the late phase position data from the early phase image.

The trained model for generating position data may be configured to input an early phase image and output late phase position data. In this case, the early phase position data may be generated in the early phase image according to image processing such as threshold processing or a user instruction via the input device 53. Further, the trained model for generating position data may be configured to input a late phase image and output late phase position data and early phase position data, or input a late phase image. It may be configured to output early phase position data. In addition, the results obtained by an application for segmentation of tumors and blood vessels may be used for the position of the tumor-dense region in the early phase.

In the trained model that outputs a late phase image without motion artifacts (with reduced motion artifacts), additional information may be input as input data instead of position data.

FIG. 12 is a diagram showing an example of input / output of another trained model that outputs a late phase image without motion artifacts (with reduced motion artifacts). As shown in FIG. 12, the trained model inputs an early phase image without motion artifacts, a late phase image with motion artifacts, and additional information as input data. As described above, the early phase image depicts a deep tumor stain, and the late phase image depicts a deep corona stain. As additional information, any information that affects the flow of the contrast medium is applicable. For example, additional information includes anatomical information, contrast agent information and blood flow information. Examples of anatomical information include morphological images such as CT images and MR images of subjects, and human body models relating to standard anatomical structures. As the anatomical information input to the trained model, it is preferable to use the same imaging site as the early phase image and the late phase image. The contrast medium information is information on the characteristics of the contrast medium injected into the subject, and specifically, the concentration, type, flow velocity, and the like. The blood flow information is information on the blood flow of the subject, and specifically, it is measurement information of blood pressure and ultrasonic Doppler. It is not necessary to input all the anatomical information, the contrast medium information, and the blood flow information as additional information to the trained model, but at least one kind of information may be input. As output data, the trained model outputs a late phase image without motion artifacts.

Here, the generation of the trained model of FIG. 12 will be described with reference to FIG. The method of generating the trained model of FIG. 12 is the same as that of the trained model of FIGS. 5 and 10 except for the contents of the training sample. For the sake of simplicity, only the differences from the trained models of FIGS. 5 and 10 will be described.

The learning device learns DNN parameters based on a plurality of learning samples. Each training sample contains a combination of an early phase image without motion artifacts, a late phase image with motion artifacts, and additional information as input data, and a late phase image without motion artifacts as teacher data (correct late phase image). including. In the process of generating the trained model, the learning device applies the early phase image without motion artifacts, the late phase image with motion artifacts, and additional information to the machine learning model, performs forward propagation processing, and outputs the estimated late phase image. To do. Next, the learning device applies the machine learning model to the difference (error) between the estimated late phase image and the correct late phase image, performs back propagation processing, and calculates the gradient vector. The learning device then updates parameters such as the weighted matrix and bias of the machine learning model based on the gradient vector. The trained model is completed by repeating the forward propagation process and the back propagation process for a large number of training samples and updating the parameters. The trained model after completion is stored in the storage device 52.

The trained model of FIG. 12 inputs additional information as compared with the trained model of FIG. By inputting additional information, it becomes possible to incorporate information that cannot be obtained from the early phase image and the late phase image into the trained model. By using the trained model thus generated, the processing circuit 51 maintains the contrast agent region contained in the late phase image, while maintaining the motion artifacts caused by the body movements in the rotational imaging performed in the late phase. Can be reduced with high accuracy.

In the above embodiments, the trained model may be generated depending on the size of the tumor. That is, the storage device 52 stores a plurality of learned models according to the size of the tumor. Typically, a trained model is generated for each tumor size segment. The size of the tumor may be divided into two categories, for example, a small tumor less than 2 cm and a large tumor 2 cm or more. By properly using the trained model according to the size of the tumor, the trained model can more accurately identify the tumor region, contrast agent region, and artifact, which is expected to improve the image quality of the corrected image. ..

When using the trained model, the processing circuit 51 identifies the size of the tumor region from the early phase image, the late phase image, the early phase position data or the late phase position data. For example, the size of the tumor densely stained area is calculated from the early phase image or the early phase position data, and the size is specified as the size of the tumor area. In addition, the size of the image region surrounded by the corona deep-stained region is calculated from the late phase image or the late phase position data, and the size is specified as the size of the tumor region. The size may be defined by the number of pixels of the tumor region, or may be defined by the length of the minor axis or the major axis of the tumor region. Alternatively, information on the size of the tumor described in the electronic medical record of the subject or the like may be used. Then, the processing circuit 51 discriminates the above-mentioned division corresponding to the specified size, reads out the trained model associated with the division, and applies the read-out learned model to the early phase image and the late phase image. Generate a corrected image.

The trained model may be generated according to the body shape of the subject, contrast agent information, or blood flow information. Specifically, the body shape of the subject is the height, weight, sex, age, etc. of the subject. The contrast medium information is information on the characteristics of the contrast medium injected into the subject, and specifically, the concentration, type, flow velocity, and the like. The blood flow information is information on the blood flow of the subject, and specifically, it is measurement information of blood pressure and ultrasonic Doppler. By properly using the trained model according to this information, the trained model can more accurately identify the tumor region, the contrast agent region, and the artifact, and it is expected that the image quality of the corrected image will be improved.

As described above, in the X-ray diagnostic apparatus 1 according to the present embodiment, the processing circuit 51 having the processing circuit 51 realizes the acquisition function 511, the reconstruction function 512, and the artifact reduction function 514. The acquisition function 511 includes a first projection data set generated by rotation imaging performed at the first timing on the subject injected with the contrast medium, and the first timing on the subject. Acquires a second projection data set generated by rotational imaging performed at different second timings. The reconstruction function 512 generates a first reconstruction image based on the first projection data set and a second reconstruction image based on the second projection data set. The artifact reduction function 514 corresponds to the second timing based on the first reconstructed image and the second reconstructed image, and the artifacts are reduced as compared with the second reconstructed image. Generate the reconstructed image of 3.

According to the above configuration, for example, a late phase image without motion artifacts for the subject is generated based on an early phase image with reduced motion artifacts for the same subject and a late phase image with motion artifacts. Therefore, when an artifact such as a motion artifact occurs in the multi-phase imaging using the rotation imaging, it is possible to avoid the re-imaging. Therefore, it is possible to improve the workflow of polymorphic imaging using rotary imaging for patients with poor breath holding and to support accurate diagnosis.

In the above description, the medical image processing apparatus according to the present embodiment is assumed to be mounted on the X-ray diagnostic apparatus. However, the present embodiment is not limited to this, and may be mounted on an X-ray computed tomography apparatus capable of collecting a plurality of projection data sets relating to a plurality of angles as in the X-ray diagnostic apparatus.

The medical image processing apparatus according to the present embodiment is a computer that processes data collected by the X-ray diagnostic apparatus. The medical image processing apparatus according to the present embodiment may be a computer included in the X-ray diagnostic apparatus or a computer separate from the X-ray diagnostic apparatus. Hereinafter, the medical image processing apparatus is assumed to be a computer included in the X-ray diagnostic apparatus.

In the above embodiment, it is assumed that the X-ray diagnostic apparatus 1 performs rotation imaging, image reconstruction, artifact reduction, and image display. However, this embodiment is not limited to this. Rotational imaging, image reconstruction, artifact reduction, and image display may be performed in a distributed manner among a plurality of devices constituting the X-ray diagnostic system.

FIG. 13 is a diagram showing a configuration of an X-ray diagnostic system 100 according to the present embodiment. As shown in FIG. 13, it has an X-ray photographing unit 3, a console 5, an image processing computer 7, and a display device 9. The X-ray imaging unit 3 and the console 5 form an X-ray diagnostic apparatus. The console 5, the image processing computer 7, and the display device 9 are communicably connected via a network. As a network communication standard, for example, DICOM (Digital Imaging and Communications and in Medicine) is used.

As described above, the X-ray imaging unit 3 performs rotational imaging, and the projection data set for the early phase and the projection data set for the late phase are collected and transmitted to the console 5. The console 5 is equipped with a reconstruction function 512. The console 5 performs image reconstruction processing on the projection data set for the early phase to generate an early phase image, and image reconstruction processing for the projection data set for the late phase to generate a late phase image. The early phase image and the late phase image are transmitted to the image processing computer 7 via the network. The image processing computer 7 is a computer such as a workstation. The trained model is stored in the image processing computer 7, and the artifact reduction function 514 is installed. The image processing computer 7 applies the trained model to the early phase image and the late phase image to generate a corrected image with reduced artifacts. The corrected image is transmitted to the display device 9. The display device 9 is a computer called a viewer. The display device 9 is equipped with a display control function 515. The display device 9 displays the corrected image. A medical worker such as an image interpreting doctor observes the displayed corrected image and diagnoses liver cancer or the like.

According to the above configuration, since the artifact reduction function 514 is mounted on the image processing computer 7, the X-ray diagnostic apparatus can be made into an existing configuration. In FIG. 13, one X-ray diagnostic device including the X-ray imaging unit 3 and the console 5 is connected to the network, but a plurality of X-ray diagnostic devices may be connected.

According to the above embodiment, the trained model inputs the reconstructed image corresponding to the first timing and the reconstructed image corresponding to the second timing, and corresponds to the second timing and is an artifact. Is supposed to be output as a reconstructed image with reduced. However, this embodiment is not limited to this. The trained model inputs a projection data set corresponding to the first timing and a projection data set corresponding to the second timing, and outputs a reconstructed image corresponding to the second timing and with reduced artifacts. It may be learned as follows. According to this trained model, for example, it is possible to input a projection data set of an early phase and a projection data set of a late phase, and output a corrected image corresponding to the late phase and with reduced motion artifacts. .. Therefore, in the process of generating the corrected image, it is possible to omit the step of generating the reconstructed image from the projection data set.

According to at least one embodiment described above, it is possible to efficiently reduce the artifacts associated with the polymorphic imaging using the rotation imaging.

The term "processor" used in the above description refers to, for example, a CPU, a GPU, or an application specific integrated circuit (ASIC), a programmable logic device (for example, a simple programmable logic device (Simple Programmable Logic Device)). : SPLD), Complex Programmable Logic Device (CPLD), and Field Programmable Gate Array (FPGA)). The processor realizes the function by reading and executing the program stored in the storage circuit. Instead of storing the program in the storage circuit, the program may be directly embedded in the circuit of the processor. In this case, the processor realizes the function by reading and executing the program embedded in the circuit. Further, instead of executing the program, the function corresponding to the program may be realized by a combination of logic circuits. It should be noted that each processor of the present embodiment is not limited to the case where each processor is configured as a single circuit, and a plurality of independent circuits may be combined to form one processor to realize its function. Good. Further, the plurality of components in FIGS. 1 and 13 may be integrated into one processor to realize the function.

Although some embodiments of the present invention have been described, these embodiments are presented as examples and are not intended to limit the scope of the invention. These embodiments can be implemented in various other forms, and various omissions, replacements, and changes can be made without departing from the gist of the invention. These embodiments and modifications thereof are included in the scope and gist of the invention, as well as in the scope of the invention described in the claims and the equivalent scope thereof.

1 X-ray diagnostic device 3 X-ray imaging unit 5 Console 7 Image processing computer 9 Display device 10 X-ray tube holding device 11 X-ray tube 12 High voltage generator 13 X-ray detector 15 C arm 17 C arm holding system 19 C arm Drive system 30 Sleeper device 31 Sleeper 33 Sleeper drive system 37 Base 39 Top plate 51 Processing circuit 52 Storage device 53 Input device 54 Communication device 55 Display device 56 System control circuit 100 X-ray diagnostic system 511 Acquisition function 512 Reconstruction function 513 Position Specific function 514 Artist reduction function 515 Display control function

Claims (15)

A first projection data set generated by rotation imaging performed at the first timing on a subject injected with a contrast medium and a second projection data set different from the first timing on the subject. An acquisition unit that acquires a second projection data set generated by rotational photography executed at the timing,
A reconstruction unit that generates a first reconstruction image based on the first projection data set and a second reconstruction image based on the second projection data set.
Based on the first reconstructed image and the second reconstructed image, a third reconstructed image corresponding to the second timing and having less artifacts than the second reconstructed image is obtained. The reduction part to be generated and
A medical image processing device comprising. The reduction unit maintains the contrast medium region included in the second reconstructed image, and as the artifact, a motion artifact caused by the body movement of the subject in the rotation imaging performed at the second timing. The medical image processing apparatus according to claim 1. Reconstruction without artifacts corresponding to the first timing generated in advance and reconstruction images without artifacts corresponding to the second timing are input as reconstructed images with artifacts corresponding to the second timing. The medical image processing apparatus according to claim 1, wherein the trained model that outputs an image is applied to the first reconstructed image and the second reconstructed image to generate the third reconstructed image. .. The trained model has a plurality of models according to the size of the tumor.
The reduction unit selects a model corresponding to the size of the tumor contained in the subject from the plurality of models, and uses the selected model as the first reconstructed image and the second reconstruction image. The third reconstructed image is generated by applying to the reconstructed image.
The medical image processing apparatus according to claim 3. The trained model further inputs at least one additional information of anatomical information, contrast agent information, and blood flow information.
The reduction unit applies the trained model to the first reconstructed image, the second reconstructed image, and additional information about the subject to generate the third reconstructed image.
The medical image processing apparatus according to claim 3. The first timing is any one of the early phase and the late phase in polyphase imaging.
The second timing is the other phase of the early phase and the late phase.
The medical image processing apparatus according to claim 1. The first reconstructed image includes any one of a tumor deeply stained region and a corona deeply stained region by a contrast medium.
The second reconstructed image includes the tumor-dense region or the other region of the corona-dense region.
The medical image processing apparatus according to claim 1. Further comprising a specific portion for identifying the first position of the tumorigenic tissue region in the first reconstructed image and the second position of the tumorigenesis tissue region in the second reconstructed image.
The reduction unit generates the third reconstructed image based on the first reconstructed image, the second reconstructed image, the first position, and the second position.
The medical image processing apparatus according to claim 1. Further provided with a specific portion for identifying the first position of the tumor deep-stained area included in the first reconstructed image and the second position of the corona deep-stained area included in the second reconstructed image.
The reduction unit generates the third reconstructed image based on the first reconstructed image, the second reconstructed image, the first position, and the second position.
The medical image processing apparatus according to claim 1. The medical image processing apparatus according to claim 9, wherein the specific unit identifies the second position of the corona deep-stained region from the first reconstructed image. A display unit for displaying the second reconstructed image and the first display element for instructing the generation of the third reconstructed image is further provided.
The reduction unit generates the third reconstructed image when the first display element is operated.
The medical image processing apparatus according to claim 1. On the computer
A function to generate a first reconstructed image based on a first projection data set generated by a rotation imaging performed at a first timing on a subject injected with a contrast medium.
A function of generating a second reconstructed image based on a second projection data set generated by a rotation imaging performed on the subject at a second timing different from the first timing.
Based on the first reconstructed image and the second reconstructed image, a third reconstructed image corresponding to the second timing and having less artifacts than the second reconstructed image is obtained. The function to generate and
A medical image processing program that realizes. A support mechanism that rotatably supports the X-ray tube and the X-ray detector,
The X-ray tube, the X-ray detector, and the support mechanism are controlled, and a rotation image is executed at the first timing on the subject into which the contrast medium is injected to collect the first projection data set. An imaging control unit that executes rotary imaging on the subject at a second timing different from the first timing and collects a second projection data set.
A reconstruction unit that generates a first reconstruction image based on the first projection data set and a second reconstruction image based on the second projection data set.
Based on the first reconstructed image and the second reconstructed image, a third reconstructed image corresponding to the second timing and having less artifacts than the second reconstructed image is obtained. The reduction part to be generated and
An X-ray diagnostic apparatus comprising. A first rotational imaging is performed on the subject injected with the contrast medium at the first timing to collect the first projection data set, and the subject is different from the first timing. An imaging unit that executes a second rotation imaging at the timing of 2 and collects a second projection data set,
A reconstruction unit that generates a first reconstruction image based on the first projection data set and a second reconstruction image based on the second projection data set.
Based on the first reconstructed image and the second reconstructed image, a third reconstructed image corresponding to the second timing and having less artifacts than the second reconstructed image is obtained. The reduction part to be generated and
A display unit that displays the third reconstructed image and
An X-ray diagnostic system comprising. The first projection data set generated by the first rotational imaging performed at the first timing on the subject injected with the contrast medium is different from the first timing on the subject. An acquisition unit that acquires a second projection data set generated by the second rotation imaging executed at the second timing, and an acquisition unit.
A reduction unit that generates a reconstructed image with reduced artifacts corresponding to the second timing based on the first projection data set and the second projection data set.
A medical image processing device comprising.