Medical image processing apparatus and medical image processing method

Abstract

To provide a medical image processing device and a medical image processing method capable of reducing deviation of noise in a corrected image in which high absorber artifacts contained in a reconfiguration image are corrected.SOLUTION: There is provided a medical image processing device equipped with an arithmetic unit for correcting high absorber artifacts. The arithmetic unit includes a projection data generation part for generating projection data corresponding to a high absorber region in a reconfiguration image in which high absorber artifacts are contained, a noise image generation part for generating a noise image using the projection data, and a weighted composition part for executing weighted composition of the noise image on a corrected image in which high absorber artifacts are corrected.SELECTED DRAWING: Figure 3

Description

Technical Field

[0001] The present invention relates to a medical image processing apparatus and a medical image processing method for handling medical images obtained by a medical imaging apparatus such as an X-ray CT (Computed Tomography) apparatus, and relates to a technique for correcting artifacts that occur when a high absorber such as metal is included in a subject.

Background Art

[0002] An X-ray CT apparatus, which is an example of a medical imaging apparatus, irradiates X-rays from around a subject to acquire projection data at a plurality of projection angles, and generates a reconstructed image of the subject used for image diagnosis by back-projecting the projection data. When a high absorber such as metal, for example, a plate used for bone fixation, is included in the subject, a high absorber artifact due to the influence of the high absorber occurs in the medical image, which hinders image diagnosis. Techniques for reducing metal artifacts are called MAR (Metal Artifact Reduction), and there are various methods such as a beam hardening correction method, a linear interpolation method, and a deep learning method.

[0003] Patent Document 1 discloses that a corrected image is obtained by weighted addition of a high-pass filter image of an original image and a high-pass filter image of a MAR image with reduced metal artifacts according to a weight depending on the proximity to metal, and further adding a low-pass filter image of the MAR image.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] However, Patent Document 1 does not take into consideration the noise bias in corrected images after high-absorbent artifacts have been corrected. In other words, near high-absorbent areas, noise is excessively reduced due to the correction of high-absorbent artifacts, resulting in a noise bias between areas far from the high-absorbent area and nearby areas, which hinders image diagnosis.

[0006] Therefore, the object of the present invention is to provide a medical image processing apparatus and a medical image processing method that can reduce the noise bias of a corrected image in which high-absorber artifacts contained in the reconstructed image have been corrected. [Means for solving the problem]

[0007] To achieve the above objective, the present invention provides a medical image processing apparatus comprising a calculation unit for correcting high-absorbent artifacts, wherein the calculation unit comprises a projection data generation unit that generates projection data corresponding to high-absorbent regions in a reconstructed image containing high-absorbent artifacts, a noise image generation unit that generates a noise image using the projection data, and a weighted synthesis unit that weightedly synthesizes the noise image with a corrected image in which high-absorbent artifacts have been corrected.

[0008] The present invention also relates to a medical image processing method for correcting high-absorbent artifacts, comprising: a projection data generation step of generating projection data corresponding to high-absorbent regions in a reconstructed image containing high-absorbent artifacts; a noise image generation step of generating a noise image using the projection data; and a weighted synthesis step of weightedly synthesizing the noise image with a corrected image in which high-absorbent artifacts have been corrected. [Effects of the Invention]

[0009] According to the present invention, it is possible to provide a medical image processing apparatus and a medical image processing method that can reduce the noise bias of a corrected image in which high-absorber artifacts contained in the reconstructed image have been corrected. [Brief explanation of the drawing]

[0010] [Figure 1] Overall configuration diagram of a medical image processing system [Figure 2] Overall configuration diagram of an X-ray CT scanner, an example of a medical imaging device. [Figure 3] A diagram showing an example of the processing flow in Example 1. [Figure 4] This diagram shows an example of the processing flow for S301 in Example 1. [Figure 5] This diagram shows an example of the processing flow for S302 in Example 1. [Figure 6] This diagram shows an example of the processing flow in S303 of Example 1. [Figure 7A] A diagram showing an example of a machine learning processing unit. [Figure 7B] A diagram showing an example of a machine learning processing unit. [Figure 7C] A diagram showing an example of a machine learning processing unit. [Modes for carrying out the invention]

[0011] Hereinafter, embodiments of the medical image processing apparatus and medical image processing method according to the present invention will be described with reference to the attached drawings. In the following description and attached drawings, components having the same functional configuration will be denoted by the same reference numerals to avoid redundant explanations. [Examples]

[0012] Figure 1 shows the hardware configuration of the medical image processing device 1. The medical image processing device 1 is configured with a processing unit 2, memory 3, storage device 4, and network adapter 5 connected via a system bus 6 in a signal transmission / reception manner. The medical image processing device 1 is also connected via a network 9 to a medical image acquisition device 10, a medical image database 11, and a machine learning processing device 12 in a signal transmission / reception manner. Furthermore, a display device 7 and an input device 8 are connected to the medical image processing device 1. Here, "signal transmission / reception manner" refers to a state in which signals can be transmitted and received between devices or from one to the other, electrically or optically, whether wired or wireless.

[0013] The arithmetic unit 2 is a device that controls the operation of each component, specifically a CPU (Central Processing Unit) or MPU (Micro Processor Unit). The arithmetic unit 2 loads programs and data necessary for program execution stored in the storage device 4 into memory 3 and executes them, performing various image processing on medical images. Memory 3 stores the programs executed by the arithmetic unit 2 and the intermediate results of the arithmetic processing. The storage device 4 is a device that stores programs executed by the arithmetic unit 2 and data necessary for program execution, specifically an HDD (Hard Disk Drive) or SSD (Solid State Drive). The network adapter 5 is for connecting the medical image processing device 1 to a network 9 such as a LAN, telephone line, or the Internet. Various data handled by the arithmetic unit 2 may be transmitted and received from outside the medical image processing device 1 via a network 9 such as a LAN (Local Area Network).

[0014] The display device 7 is a device that displays the processing results of the medical image processing device 1, and is specifically a liquid crystal display or the like. The input device 8 is an operating device that allows the operator to give operating instructions to the medical image processing device 1, and is specifically a keyboard, mouse, touch panel, or the like. The mouse may be replaced with other pointing devices such as a trackpad or trackball.

[0015] The medical imaging device 10 is, for example, an X-ray CT (Computed Tomography) device that acquires projection data of a subject and generates reconstructed images from the projection data, and will be described later with reference to Figure 2. The medical image database 11 is a database system that stores projection data and reconstructed images acquired by the medical imaging device 10, as well as corrected images obtained by image processing on the reconstructed images.

[0016] The machine learning processing device 12 is generated by machine learning to reduce high absorber artifacts included in the reconstructed image, and is configured using, for example, a CNN (Convolutional Neural Network). For the generation of the machine learning processing device 12, a reconstructed image that does not include a high absorber such as metal is used as a teacher image. In addition, for the input image, projection data including a high absorber is generated by forward projecting an image obtained by adding a high absorber region to the teacher image, and a reconstructed image including high absorber artifacts obtained by back projecting the projection data is used.

[0017] Here, a high absorber is typically a metal, bone, or contrast agent, and is a substance having a relatively high X-ray absorption rate compared to other tissues (for example, organs). In the following, examples will be described using metal and bone as examples, but in the present invention, the high absorber is not limited to metal and bone. For example, when comparing fat and muscle, muscle has a higher X-ray absorption rate, and in a region that does not include metal, bone, or contrast agent, muscle can be said to be a high absorber. That is, in the present invention, at a location where a region of interest is set by processing such as user selection or threshold extraction, artifacts caused by tissues having a relatively high X-ray absorption rate can be reduced.

[0018] The overall configuration of an X-ray CT apparatus 100, which is an example of the medical imaging apparatus 10, will be described using FIG. 2. In FIG. 2, the horizontal direction is the X-axis, the vertical direction is the Y-axis, and the direction perpendicular to the paper surface is the Z-axis. The X-ray CT apparatus 100 includes a scanner 200 and an operation unit 250. The scanner 200 includes an X-ray tube 211, a detector 212, a collimator 213, a drive unit 214, a central control unit 215, an X-ray control unit 216, a high voltage generation unit 217, a scanner control unit 218, a bed control unit 219, a collimator control unit 221, a preamplifier 222, an A / D converter 223, a bed 240, and the like.

[0019] The X-ray tube 211 is a device that irradiates an object 210 placed on the bed 240 with X-rays. X-rays are irradiated from the X-ray tube 211 to the object by applying a high voltage generated by the high voltage generation unit 217 according to a control signal transmitted from the X-ray control unit 216.

[0020] The collimator 213 is a device that limits the irradiation range of X-rays emitted from the X-ray tube 211. The X-ray irradiation range is set according to a control signal transmitted from the collimator control unit 221.

[0021] The detector 212 is a device that measures the spatial distribution of transmitted X-rays by detecting X-rays that have passed through the subject 210. The detector 212 is positioned opposite the X-ray tube 211, and a large number of detection elements are arranged in two dimensions within the plane opposite the X-ray tube 211. The signal measured by the detector 212 is amplified by the preamplifier 222 and then converted into a digital signal by the A / D converter 223. After that, various correction processes are performed on the digital signal, and projection data is acquired.

[0022] The drive unit 214 rotates the X-ray tube 211 and detector 212 around the subject 210 according to control signals transmitted from the scanner control unit 218. As the X-ray tube 211 and detector 212 rotate, X-ray irradiation and detection occur, acquiring projection data from multiple projection angles. The data acquisition unit for each projection angle is called a view. The arrangement of each detection element in the two-dimensionally arranged detector 212 is called a channel in the direction of rotation of the detector 212, and a column in the direction perpendicular to the channel. The projection data is identified by view, channel, and column.

[0023] The bed control unit 219 controls the movement of the bed 240, either keeping it stationary or moving it at a constant velocity in the Z-axis direction, which is the body axis direction of the subject 210, while X-ray irradiation and detection are performed. Scanning while the bed 240 is stationary is called an axial scan, and scanning while the bed 240 is moving is called a helical scan.

[0024] The central control unit 215 controls the operation of the scanner 200 as described above, according to instructions from the operation unit 250. Next, the operation unit 250 will be described. The operation unit 250 includes a reconstruction processing unit 251, an image processing unit 252, a storage unit 254, a display unit 256, an input unit 258, and the like.

[0025] The reconstruction processing unit 251 generates a reconstructed image by back-projecting the projection data acquired by the scanner 200. The image processing unit 252 performs various image processing to make the reconstructed image suitable for diagnosis. The storage unit 254 stores the projection data, reconstructed image, and image after image processing. The display unit 256 displays the reconstructed image and image after image processing. The input unit 258 is used by the operator to set the acquisition conditions for projection data (tube voltage, tube current, scan speed, etc.) and the reconstruction conditions for the reconstructed image (reconstruction filter, FOV size, etc.).

[0026] The operation unit 250 may also be the medical image processing device 1 shown in Figure 1. In that case, the reconstruction processing unit 251 and the image processing unit 252 would correspond to the calculation unit 2, the storage unit 254 to the storage device 4, the display unit 256 to the display device 7, and the input unit 258 to the input device 8.

[0027] Using Figure 3, we will explain step by step an example of the processing flow performed in Example 1.

[0028] (S301) The calculation unit 2 generates projection data corresponding to the high-absorbent regions in the reconstructed image that contain high-absorbent artifacts. In other words, the calculation unit 2 functions as a projection data generation unit that generates projection data corresponding to the high-absorbent regions in the reconstructed image that contain high-absorbent artifacts.

[0029] An example of the processing flow of S301 will be explained step by step using Figure 4.

[0030] (S401) The calculation unit 2 acquires a reconstructed image that includes high-absortide artifacts. The reconstructed image that includes high-absortide artifacts may be generated by back-projecting projection data of a subject containing high-absortide materials such as metal or bone, or it may be read from the storage device 4 or the medical image database 11.

[0031] (S402) The calculation unit 2 extracts high-absorbent regions from the reconstructed image containing high-absorbent artifacts. High-absorbent regions are extracted, for example, by thresholding. That is, pixels with pixel values ​​higher than a predetermined threshold are extracted from the reconstructed image as high-absorbent regions.

[0032] (S403) The calculation unit 2 forward-projects the regions other than the high-absorbent regions to generate projection data for the regions other than the high-absorbent regions. More specifically, the reconstructed image in which the pixel values ​​of the high-absorbent regions extracted in S402 are replaced with zeros is forward-projected to generate projection data for the regions other than the high-absorbent regions.

[0033] (S404) The calculation unit 2 obtains projection data for the high-absorbent region by taking the difference between the projection data for regions other than the high-absorbent region and the original projection data. More specifically, the projection data for regions other than the high-absorbent region generated in S403 is subtracted from the projection data corresponding to the reconstructed image obtained in S401, thereby generating projection data for the high-absorbent region.

[0034] The processing flow illustrated in Figure 4 generates projection data corresponding to the high-absorbent regions in the reconstructed image containing high-absorbent artifacts. According to the processing flow in Figure 4, projection data including noise due to the influence of high-absorbent regions can be generated. Returning to the explanation of Figure 3.

[0035] (S302) The calculation unit 2 generates a noise image due to the influence of high-absorbent regions. In other words, the calculation unit 2 functions as a noise image generation unit that generates a noise image using projection data that includes noise due to the influence of high-absorbent regions.

[0036] An example of the processing flow in S302 will be explained step by step using Figure 5.

[0037] (S501) The calculation unit 2 performs an even-odd division of the projection data corresponding to the high-absorbent region. That is, the projection data of the high-absorbent region generated in S301 is divided into odd views and even views, and odd projection data is generated from multiple odd views, and even projection data is generated from multiple even views.

[0038] (S502) The calculation unit 2 reconstructs the odd projection data and the even projection data, respectively. That is, the odd-reconstructed image is generated by reconstructing the odd projection data generated in S301, and the even-reconstructed image is generated by reconstructing the even projection data. Since the odd-reconstructed and even projection data are generated by dividing adjacent views, the odd-reconstructed image and the even-reconstructed image generated by their respective reconstructions contain equivalent high-absorbent regions.

[0039] (S503) The calculation unit 2 generates a difference image between the odd-reconstructed image and the even-reconstructed image. Since the odd-reconstructed image and the even-reconstructed image contain equivalent high-absorbent regions, the high-absorbent regions are removed from the difference image between the two, leaving only noise. In other words, the noise caused by the influence of the high-absorbent regions is visualized by the difference between the odd-reconstructed image and the even-reconstructed image generated in S502. Note that the amount of noise is multiplied by √2 due to the difference processing, so a noise image may also be generated by dividing the difference image by √2.

[0040] The processing flow illustrated in Figure 5 generates a noisy image due to the influence of high-absorbent regions. According to the processing flow in Figure 5, it is possible to generate an image that does not include high-absorbent regions and only retains the noise due to the influence of high-absorbent regions, thus making it easier to reduce the noise bias in the corrected image. Return to the explanation of Figure 3.

[0041] (S303) The calculation unit 2 weights and combines the corrected image, in which high-absorber artifacts have been corrected, with the noise image generated in S302. In other words, the calculation unit 2 functions as a weighted blending unit that weights and combines the noise image with the corrected image, in which high-absorber artifacts have been corrected.

[0042] An example of the processing flow in S303 will be explained step by step using Figure 6.

[0043] (S601) The calculation unit 2 calculates the noise distribution of the reconstructed image, including high-absorber artifacts. More specifically, the standard deviation calculated using the pixel values ​​of the target pixel and the surrounding pixels in the reconstructed image is determined as the noise amount of the target pixel. In other words, the noise distribution is calculated by determining the noise amount for each pixel in the reconstructed image. Furthermore, the calculation unit 2 generates a noise coefficient image in which the calculated noise distribution is normalized. More specifically, the noise distribution is normalized by dividing each noise amount in the noise distribution by the maximum value in the noise distribution, and a noise coefficient image is generated.

[0044] (S602) The calculation unit 2 generates a multiplication image of the noise image of the high-absorber region and the noise coefficient image. That is, the noise image generated in S302 is multiplied by the noise coefficient image generated in S601 to generate a noise adjustment image used to reduce the noise bias of the corrected image.

[0045] (S603) The calculation unit 2 generates an additive image of the corrected image, in which high-absorber artifacts have been corrected, and the noise-adjusted image. That is, the noise-adjusted image generated in S602 is added to the corrected image, thereby generating a noise-added image, which is an image in which the noise bias of the corrected image has been reduced.

[0046] The processing flow illustrated in Figure 6 generates a noise-added image with reduced noise bias in the corrected image. According to the processing flow in Figure 6, a noise-added image can be generated simply by adding a noise-adjusted image to a corrected image in which high-absorber artifacts have been corrected, thus easily reducing noise bias between areas far from high-absorbers and nearby areas.

[0047] As explained above, in Example 1, a noise image generated from projection data corresponding to a high-absorbent region and a corrected image in which high-absorbent artifacts have been corrected are weighted and combined to generate a noise-added image in which the noise bias of the corrected image has been reduced. In the noise-added image, the noise bias that occurs as a result of correcting high-absorbent artifacts is reduced, so that there is no impediment to image diagnosis between regions far from the high-absorbent region and nearby regions.

[0048] Corrected images with high-absorbent artifacts removed are generated using various MAR methods such as beam hardening correction, linear interpolation, and deep learning. An example of a machine learning processing unit that generates corrected images is shown in Figure 7A. The machine learning processing unit shown in Figure 7A is generated by machine learning a large number of input image-training image pairs, with a reconstructed image without high-absorbent artifacts as the training image and a reconstructed image with high-absorbent artifacts added to the training image as the input image. The input image, which is the reconstructed image with high-absorbent artifacts added, is generated by back-projecting projection data, which is generated by forward-projecting an image with high-absorbent regions added to the training image.

[0049] In the machine learning processing unit shown in Figure 7A, noise may be excessively reduced in the vicinity of high absorbers, so it is necessary to execute the processing flow shown in Figure 3. Therefore, by performing machine learning on the noisy image or the noisy image generated by the processing flow shown in Figure 3 as input images, a machine learning processing unit may be generated that outputs an image in which high absorber artifacts are corrected and the noise bias is reduced.

[0050] Using Figure 7B, an example of a machine learning processing unit that generates a corrected image in which high-absorbent artifacts are corrected and noise bias is reduced will be explained. The machine learning processing unit exemplified in Figure 7B is generated by machine learning on a large number of input image-training image pairs, with a reconstructed image that does not contain high-absorbent artifacts as the training image, and a reconstructed image with high-absorbent artifacts added to the training image and a noise-added image as input images. The image generated in S303 is used as the noise-added image. By machine learning with the noise-added image as input images along with the reconstructed image containing high-absorbent artifacts, a machine learning processing unit is generated that can output a corrected image in which high-absorbent artifacts are corrected and noise bias is reduced.

[0051] Using Figure 7C, another example of a machine learning processing unit that generates a corrected image in which high-absorbent artifacts are corrected and noise bias is reduced will be described. The machine learning processing unit illustrated in Figure 7C is generated by machine learning a large number of input image-training image pairs, with a reconstructed image without high-absorbent artifacts as the training image, and a reconstructed image with high-absorbent artifacts added to the training image and a noise image as input images. The noise image used is the image generated in S302. By machine learning with the noise image as input images along with the reconstructed image containing high-absorbent artifacts, a machine learning processing unit is generated that can output a corrected image in which high-absorbent artifacts are corrected and noise bias is reduced.

[0052] The embodiments of the present invention have been described above. However, the present invention is not limited to the embodiments described above, and the components can be modified and implemented without departing from the spirit of the invention. Furthermore, the components disclosed in the above embodiments may be combined as appropriate. In addition, some components may be removed from all the components shown in the above embodiments. [Explanation of symbols]

[0053] 1: Medical image processing device, 2: Processing unit, 3: Memory, 4: Storage device, 5: Network adapter, 6: System bus, 7: Display device, 8: Input device, 10: Medical image acquisition device, 11: Medical image database, 12: Machine learning processing device, 100: X-ray CT device, 200: Scanner, 210: Subject, 211: X-ray tube, 212: Detector, 213: Collimator, 214: Drive unit, 215: Central control unit, 216: X-ray control unit, 217: High voltage generation unit, 218: Scanner control unit, 219: Patient bed control unit, 221: Collimator control unit, 222: Preamplifier, 223: A / D converter, 240: Patient bed, 250: Operation unit, 251: Reconstruction processing unit, 252: Image processing unit, 254: Storage unit, 256: Display unit, 258: Input unit

Claims

1. A medical image processing apparatus comprising a calculation unit for correcting high-absorption artifacts, The calculation unit includes a projection data generation unit that generates projection data corresponding to high-absorbent regions in a reconstructed image containing high-absorbent artifacts, A noise image generation unit that generates a noise image using the aforementioned projection data, A medical image processing apparatus characterized by having a weighted synthesis unit that weights the noise image onto a corrected image from which high-absorber artifacts have been corrected.

2. A medical image processing apparatus according to claim 1, The medical image processing apparatus is characterized in that the projection data generation unit generates projection data corresponding to a high-absorbent region by the difference between projection data corresponding to the reconstructed image and projection data of regions in the reconstructed image other than the high-absorbent region.

3. A medical image processing apparatus according to claim 1, The noise image generation unit generates odd projection data and even projection data by dividing projection data corresponding to a high-absorbent region into even and odd divisions, generates odd-reconstructed images and even-reconstructed images by reconstructing the odd-reconstructed image and the even-reconstructed image, respectively, and generates the noise image based on the difference between the odd-reconstructed image and the even-reconstructed image.

4. A medical image processing apparatus according to claim 1, The weighted synthesis unit is characterized by adding to the corrected image a product of a noise coefficient image generated based on the noise distribution of the reconstructed image and the noise image.

5. A medical image processing apparatus according to claim 1, The medical image processing apparatus is characterized in that the calculation unit further comprises a machine learning processing unit which is generated by machine learning using a reconstructed image that does not contain high-absorbent artifacts as a training image, a reconstructed image in which high-absorbent artifacts are added to the training image, and a noise-added image generated by the weighted synthesis unit as input images.

6. A medical image processing apparatus according to claim 1, The medical image processing apparatus is characterized in that the calculation unit further comprises a machine learning processing unit which is generated by machine learning using a reconstructed image that does not contain high-absorbent artifacts as a training image, a reconstructed image in which high-absorbent artifacts are added to the training image, and a noise image generated by the noise image generation unit as input images.

7. A medical image processing method for correcting high-absorption artifacts, A projection data generation step that generates projection data corresponding to high-absor regions in a reconstructed image containing high-absor artifacts, A noise image generation step of generating a noise image using the aforementioned projection data, A medical image processing method characterized by comprising a weighted synthesis step of weighting the noise image into a corrected image in which high-absorption artifacts have been corrected.

Images