JP4982091B2 - Radiation tomography equipment - Google Patents

Description

  The present invention relates to a radiation tomography apparatus and an image processing apparatus that realize real-time labeling processing or image feedback processing to image reconstruction processing based on the result of labeling processing in a medical X-ray CT (Computed Tomography) apparatus or the like. .

  Conventionally, an X-ray CT apparatus, which is one of the radiation tomography apparatuses, can independently handle each part, each organ, or each tissue (hereinafter, abbreviated as each part) in a tomographic image of a reconstructed scan. The labeling process is performed as follows. For example, Patent Document 1 discloses a three-dimensional labeling process. In Patent Document 1, as shown in FIG. 47, the range from the z-direction start coordinate position zs to the z-direction end coordinate position ze for X-ray CT imaging by conventional scan (axial scan) or helical scan is from time t1 to time t2. After that, a continuous area numbering process, a labeling process, or a segmentation process (hereinafter referred to as a labeling process) of a three-dimensional image composed of tomographic images continuous in the z direction was performed from time t3 to t4. . For this reason, it is disadvantageous and inconvenient in the diagnosis workflow because it is a process that takes time after imaging to perform image reconstruction again based on the result of the 3D labeling process after all imaging is completed. Met.

At this time, t3 <t2 was not satisfied. That is, the labeling processing result is not fed back to the image reconstruction. That is, the result of the three-dimensional labeling process cannot be fed back to the image reconstruction, that is, the image feedback process to the tomographic image reconstruction process cannot be performed in real time. For this reason, there is a limit to the image quality optimization processing for each segment area. Further, in the labeling process in the conventional image reconstruction process and image determination process, the binarization is performed from the grayscale image to form a binary image, and then the labeling process and the relabeling process are performed. Furthermore, since image measurement was performed for each continuous area thereafter, it was not efficient.
JP 2003-141548 A

  In a radiation tomography apparatus having a multi-row X-ray detector, as the cone angle of the X-ray cone beam increases, the imaging time is also shortened. Even if the shooting time is shortened, the image quality is not compromised, but rather the improvement and optimization of the image quality is desired. As the imaging time is shortened, what is desired in the workflow is optimization of image quality for each part, as represented by a hybrid kernel or a segmented hybrid kernel. In other words, it is desired that the image quality is optimized for each region of the region simultaneously with the photographing.

  As a labeling technique in a conventional radiation tomography apparatus, also in three-dimensional labeling, three-dimensional labeling is performed after the entire three-dimensional image is first prepared by image reconstruction. However, if, for example, a three-dimensional labeling process is performed after a long helical scan, the workflow is slowed down and the rate is determined. Further, the result of the three-dimensional labeling process is not fed back to the image reconstruction process, and so-called image feedback process cannot be performed in real time. This applies not only to the helical scan, but also to a variable pitch helical scan, a helical shuttle scan, or an axial scan that performs scanning at a plurality of coordinate positions in the z direction continuously in the z direction.

  Therefore, in order to solve these problems, in the local region of the three-dimensional image composed of tomographic images continuous in the z direction, the image reconstruction processing and pipeline, and the three-dimensional labeling processing as image reconstruction and parallel processing The need to do it comes out. There is a new technical problem in deriving final three-dimensional labeling result information from the result information of the local parallel labeling process in the three-dimensional local region. In particular, there is a new technical problem in editing information of each continuous area in the relabeling process. When the image feature quantity is measured simultaneously with the labeling process, it is necessary to edit the measured value of the image feature quantity based on the result of the continuity of each continuous area during the re-labeling process. There is a new technical problem in this respect as well.

  Normally, a 3D grayscale image is binarized, and a 3D labeling process is performed as a 3D binary image. However, a 3D labeling process is performed directly from the 3D grayscale image to measure 3D labeling and image feature values. Even in the simultaneous processing, it is necessary to edit the measurement result of the image feature amount during the relabeling processing. There is a new technical problem in this respect as well. In addition, there is a new technical problem in the case where the adaptive noise filter in the z direction is applied simultaneously with the three-dimensional labeling.

  Accordingly, an object of the present invention is to perform various scans of a radiation tomography apparatus having a multi-row X-ray detector and real-time labeling processing in an image processing apparatus, or image feedback processing to image reconstruction processing based on the result of labeling processing. An object of the present invention is to provide a radiation tomography apparatus and an image processing apparatus.

  Hereinafter, in order to facilitate understanding, description will be made with three-dimensional labeling. According to the present invention, in various scans, image reconstruction of one tomographic image or a plurality of tomographic images is completed, and these are used as a part of a three-dimensional image composed of continuous tomographic images in the z direction, for example, three-dimensional Start labeling. Here, for example, three-dimensional labeling of 64 tomographic images is performed. Although it was an independent three-dimensional continuous region up to the middle, it may be found that as the three-dimensional labeling proceeds in the z direction, the three-dimensional continuous regions that seem to be independent are connected.

  FIG. 1A is a diagram illustrating a state in which the 3D labeling process is performed partway through the 3D image, and B illustrates a case where the 3D labeling process of the 3D image is further advanced. For example, as shown in FIG. 1A, when three-dimensional labeling is performed to the middle of the three-dimensional image, the three-dimensional continuous region “1” and the three-dimensional continuous region “2” are independent and not continuous. However, when the three-dimensional labeling process is further advanced, as shown in FIG. 1B, it can be seen that the three-dimensional continuous region “1” and the three-dimensional continuous region “2” are connected after a certain tomographic image. There is. At this time, information of three-dimensional continuous regions having different connected numbers is stored in the database. That is, the re-labeling process is only a process of information on this database, and the continuous area number is not rewritten. It is assumed that such re-labeling processing on the database is performed.

  FIG. 2 is a diagram illustrating timing of the three-dimensional labeling process and the three-dimensional relabeling process. As shown in FIG. 2, it is assumed that a three-dimensional labeling process is performed on a three-dimensional image composed of 256 tomographic images. At this time, measurement of the image feature amount is performed simultaneously with the three-dimensional labeling process. When the three-dimensional labeling process for the 64 tomographic images has been completed, the three-dimensional relabeling process is performed as a process on the three-dimensional continuous information database. Specifically, in the three-dimensional relabeling process, image feature values are edited, for example, combined or merged. Assume that scanning as a radiation tomography apparatus starts at time 0. After this, the three-dimensional labeling process starts from time t3. At time t5, the re-labeling process is started simultaneously with the completion of the three-dimensional labeling process for 64 tomographic images. At time t4, the three-dimensional labeling process ends. After this, the three-dimensional relabeling process ends at time t6. In this way, imaging of the radiation tomography apparatus, 3D labeling processing, and 3D relabeling processing can be performed in parallel.

  Compared with the conventional FIG. 47, the understanding of the parallel processing of the present invention is deepened. FIG. 3 is a diagram showing the labeling process of the present invention. From time t1 to time t4 is the same time as in FIG. In FIG. 3, the labeling process of the three-dimensional image is performed from the middle from the z-direction start coordinate position zs to the z-direction end coordinate position ze. In this way, imaging of the X-ray CT apparatus and three-dimensional labeling processing are performed in parallel.

  Further, the processing speed can be increased by performing the three-dimensional labeling process and the image feature amount measurement directly from the grayscale image without performing the binarization process. At this time, the adaptive noise filter processing in the z direction is also performed on the grayscale image, so that the processing speed can be further increased, and the three-dimensional labeling processing result and the image feature value measurement result can be stabilized.

  Using each image feature quantity of each three-dimensional continuous area extracted by three-dimensional labeling as described above, it is recognized as a part, an organ, or a tissue (hereinafter collectively referred to as a part). In this way, the image quality is optimized with each three-dimensional continuous region recognized as each part region. In optimizing the image quality, optimization by an image reconstruction function and optimization by an image filter in an image space are performed for each region.

  In this way, the present invention realizes real-time labeling by performing image reconstruction, labeling processing, and relabeling processing as pipeline processing in synchronization with data collection such as helical scanning, or labeling. Based on the processing result, the above problem is solved by applying image feedback to the image reconstruction processing.

A radiation tomography apparatus according to a first aspect collects projection data irradiated with radiation and transmitted through a predetermined region of a subject, and performs image reconstruction to display a tomographic image. The image processing apparatus includes a labeling unit that performs a labeling process on the tomographic image that has been reconstructed, and an image reconstruction unit that optimizes the image quality for each continuous area extracted by the labeling unit.
In the radiation tomography apparatus according to the first aspect, tomographic images reconstructed by various scans are subjected to labeling processing, and a reconstruction function of image reconstruction processing and an image space image filter are performed for each of the extracted regions. , The image quality can be optimized.

A radiation tomography apparatus according to a second aspect collects projection data irradiated with radiation and transmitted through a predetermined region of a subject, and performs image reconstruction to display a tomographic image. For each continuous area extracted by the re-labeling means, the labeling means for labeling the tomographic image reconstructed, the re-labeling means for re-labeling based on the information of the continuous areas labeled by the labeling means, and Image reconstruction means for optimizing the image quality.
In the radiation tomography apparatus according to the second aspect, when performing a labeling process on tomographic images continuously taken in the z direction by various scans, three-dimensional labeling is advanced in the z direction to find a three-dimensional continuous region. When you go, you will also find a 3D continuous area where you can see the connection. For this reason, by correcting the connection information of the continuous area by the re-labeling process, the labeling process is appropriately performed, and the three-dimensional continuous area can be detected correctly. Thus, the image quality can be optimized for each region of each part that is the extracted three-dimensional continuous region.

A radiation tomography apparatus according to a third aspect collects projection data irradiated with radiation and transmitted through a predetermined region of a subject, and performs image reconstruction to display a tomographic image. Then, while labeling the tomographic image that has been reconstructed, a labeling unit that obtains an image feature amount for each continuous region that has been labeled and a re-labeling process based on information on the continuous region that has been labeled by the labeling unit And a re-labeling unit that edits the image feature amount for each continuous region, and an image reconstruction unit that optimizes the image quality of the continuous region edited by the re-labeling unit.
In the radiation tomography apparatus according to the third aspect, when a three-dimensional labeling process is performed on a tomographic image continuously taken in the z direction by various scans, the area and volume are determined by counting pixels for each continuous region. The circumference or surface area is obtained by counting the pixels of the contour or the surface, the circularity is obtained from the area and the circumference, and the sphericity is obtained from the volume and the surface area. In this way, the image feature amount for each continuous region is obtained in the scanning of the labeling process. Also, when relabeling each labeled continuous region, it can be seen that an independent continuous region that is not connected up to a certain z coordinate is connected at a certain z coordinate, and that continuous region is edited. . This process is a re-labeling process. At the time of this re-labeling process, the image feature amount of each continuous area is also edited (combined / combined). For example, when two continuous areas are edited, the area values and the volume values of two continuous areas may be added to each other and the volume values may be added to each other. Similarly, editing can be performed by adding the perimeter and surface area. The circularity and the sphericity may be newly calculated from the area value, volume value, perimeter length value, and surface area value newly obtained when two continuous regions are edited. As described above, in the re-labeling process, the continuous region is edited to give a new continuous region number, and the image feature amount of the two continuous regions is edited to obtain the image feature amount of the edited new continuous region. be able to.

The radiation tomography apparatus according to the fourth aspect performs labeling processing from a tomographic image reconstructed as a grayscale image without using a binary image.
In the radiation tomography apparatus according to the fourth aspect, each pixel such as a tomographic image or a three-dimensional image obtained by various scans has a CT value (HU: Hounsfield proportional to the X-ray absorption coefficient).
A gray-scale image and a gray-scale three-dimensional image. Even if the grayscale image is not binarized at this time, by performing a labeling process by determining whether or not it is within the binarization threshold range, the binarization process is performed without using the binarized binary image. Labeling can be performed directly from the image.

In the fifth aspect, at least one of the labeling unit and the re-labeling unit is performed in synchronization with the collection of projection data.
In the radiation tomography apparatus according to the fifth aspect, in order to perform a real-time labeling process on a three-dimensional image of tomographic images continuously taken in the z direction obtained by various scans, it is necessary to immediately start the labeling process. . For this reason, when image reconstruction of one or several tomographic images is performed after the projection data, after reconstruction of the first tomographic image in synchronization with the projection data and image reconstruction, or a predetermined plurality of images Immediately after the reconstruction of the tomographic image, the real-time labeling process is started. Therefore, real-time labeling can be realized by performing labeling processing in synchronization with these during scanning and image reconstruction.

The radiation tomography apparatus according to the sixth aspect performs synchronously during a helical scan, a variable pitch helical scan, a helical shuttle scan, an axial scan at a plurality of positions continuous in the z direction, or a cine scan.
In the radiation tomography apparatus according to the sixth aspect, in the fifth aspect, when tomographic images continuous in the z direction are reconstructed one after another at substantially constant time intervals, a predetermined amount is used for scan data collection and image reconstruction. The labeling process can be synchronized and moved with a delay of the number of tomographic images. Further, the real-time labeling process can be similarly realized by synchronizing with a delay of a fixed time interval instead of a delay corresponding to a predetermined number of tomographic images.

In the seventh aspect, the labeling unit starts the labeling process after the image reconstruction of one or a plurality of predetermined tomographic images is completed.
In the radiation tomography apparatus according to the seventh aspect, when synchronizing the labeling process, it is necessary to wait for at least one tomographic image reconstructed for the synchronization delay. For this reason, the real-time labeling process can be realized by starting the labeling process after waiting for at least one tomographic image or a predetermined number of sheets.

In the eighth aspect, the re-labeling means starts the re-labeling process after the labeling process for one sheet or a predetermined plurality of sheets is completed.
In the radiation tomography apparatus according to the eighth aspect, the relabeling process is performed after the labeling process according to the seventh aspect. It is necessary to take a certain interval between the labeling process and the relabeling process. The re-labeling process is a process for correcting the connection of a plurality of three-dimensional continuous areas detected by the labeling process by the re-labeling process and re-assigning the area number (label number) of the three-dimensional continuous area to be corrected. That is, in the re-labeling process, the connection information detected in the labeling process is fed back. For this reason, correction may not be possible unless the delay between the labeling process and the re-labeling process is too large. For this reason, in the re-labeling process, a sufficient delay can be obtained in the real-time labeling process by taking a sufficient delay for a plurality of tomographic images after the labeling process, and a more appropriate labeling process can be performed.

In the ninth aspect, the labeling means includes at least one of a two-dimensional labeling process, a three-dimensional labeling process, and a four-dimensional labeling process.
In the radiation tomography apparatus according to the ninth aspect, when performing three-dimensional labeling as a three-dimensional image and a tomographic image continuously photographed in the z direction, the three-dimensional labeling is based on the information of the two-dimensional labeling. I do. Further, in the helical shuttle scan, it is possible to collect a three-dimensional image composed of tomographic images continuously taken in the z direction in time series. That is, a four-dimensional image can be collected. In this case, a four-dimensional labeling process can be performed in synchronization with data collection and image reconstruction.

In a tenth aspect, the labeling unit performs a labeling process based on the image feature amount of each pixel of the tomographic image or an image feature amount in the vicinity of the pixel.
In the radiation tomography apparatus according to the tenth aspect, when labeling is performed, each pixel and its surroundings are improved in order to increase accuracy, rather than binarizing or labeling by determining only the CT value of the pixel of the tomographic image using a threshold condition. It is also possible to determine the continuity of the pixel based on the standard deviation of the CT value of the pixel and other conditions based on the image feature amount and perform the labeling process. Thereby, a more accurate labeling process can be performed.

In an eleventh aspect, the image reconstruction unit recognizes a labeled region or a re-labeled region as a part according to an image feature amount of the labeled region or the re-labeled region, and the part Optimize image quality.
In the radiation tomography apparatus according to the eleventh aspect, when performing the labeling process or the relabeling process, not only the CT value of the pixel but also the standard deviation of the CT value of each pixel and its surrounding pixels, and other image feature quantities It is also possible to determine the region by determining the continuity of the pixels by the condition determination according to.

In the twelfth aspect, the image feature amount includes area, image density sum, perimeter length, ferret diameter, ferret diameter ratio, circularity, area ratio, first moment, second moment, ellipse approximated major and minor diameters, volume. , Volume concentration sum, surface area, three-dimensional ferret diameter, three-dimensional ferret diameter ratio, sphericity, volume fraction, three-dimensional primary moment, three-dimensional secondary moment, and ellipsoid approximate diameter.
In the radiation tomography apparatus according to the twelfth aspect, these can be measured or edited as image feature amounts. Further, there may be a case where a further effect can be obtained by adding those similar to these image feature amounts. In addition, among these image feature amounts, those having little effect may be deleted.

The thirteenth image processing apparatus includes: an image input unit that inputs an image; a labeling unit that obtains an image feature amount of a continuous region that has been labeled while performing a labeling process on the input image; and a labeling process that is performed by the labeling unit. Based on the information of the continuous area, re-labeling processing is performed, and the image features for each re-labeled continuous area are edited, and the image that optimizes the image quality of the continuous area edited by the re-labeling means Optimization means.
In the image processing apparatus according to the thirteenth aspect, after inputting a three-dimensional grayscale image, the three-dimensional grayscale image is labeled, and image reprojection processing and backprojection processing or image space is performed for each labeled continuous region. The image quality can be optimized by the image filter.

In a fourteenth aspect, the image input unit inputs a tomographic image that is a grayscale image, and the labeling unit performs a labeling process from the tomographic image without using a binary image.
In the image processing apparatus according to the fourteenth aspect, for example, each pixel of a three-dimensional gray image is a gray image or a gray three-dimensional image having a continuous value. Even if the grayscale image is not binarized at this time, the grayscale image can be obtained without performing the binarization processing and the binarized binary image by performing the labeling process by determining whether or not the binarization threshold value is within the range. The labeling process can be performed directly. The image quality can be optimized for each region of the region which is the labeled three-dimensional continuous region thus obtained.

In the fifteenth aspect, at least one of the labeling unit or the re-labeling unit performs the labeling process or the re-labeling process in synchronization with the image input of the image input unit.
In the image processing apparatus according to the fifteenth aspect, starting the labeling process immediately in synchronization with the image input leads to the real-time labeling process. Therefore, real-time labeling can be realized by performing labeling processing in synchronization with image input.

In the sixteenth aspect, the labeling unit starts the labeling process after the image reconstruction of one or a predetermined plurality of tomographic images is completed.
In the image processing apparatus according to the sixteenth aspect, when synchronizing the labeling process, it is necessary to wait for at least one tomographic image whose image has been reconstructed for the synchronization delay. For this reason, the real-time labeling process can be realized by starting the labeling process after waiting for at least one tomographic image or a predetermined number of sheets.

In the seventeenth aspect, the re-labeling means starts the re-labeling process after the labeling process for one sheet or a predetermined plurality of sheets is completed.
In the image processing apparatus according to the seventeenth aspect, the relabeling process performs a correction process at the time of the relabeling process based on the connection of a plurality of three-dimensional continuous areas detected by the labeling process. This is a process of reassigning the area number (label number). That is, the connection information detected by the labeling process is fed back to the three-dimensional image by the relabeling process. In the re-labeling process, after the labeling process, a delay corresponding to a plurality of image inputs can be taken to allow sufficient correction in the real-time labeling process, and a more optimal labeling process can be performed.

In an eighteenth aspect, the labeling means includes at least one of a two-dimensional labeling process, a three-dimensional labeling process, and a four-dimensional labeling process.
In the image processing apparatus according to the eighteenth aspect, when performing a three-dimensional labeling process by synchronizing a three-dimensional image composed of two-dimensional images continuous in the z direction during input, the three-dimensional labeling is based on the information of the two-dimensional labeling. To perform 3D labeling. In addition, it is possible to collect 4D images by collecting 3D images consisting of 2D images continuous in the z direction in time series. In this case, 4D images can be synchronized with 4D image input. Labeling can be performed.

In a nineteenth aspect, the labeling means performs a labeling process based on the image feature amount of each pixel of the tomographic image or an image feature amount in the vicinity of the pixel.
In the image processing apparatus according to the nineteenth aspect, in order to further improve the accuracy of labeling, pixel continuity is determined by determining the standard deviation of the pixel values of each pixel and its surrounding pixels and other conditions based on image features. Labeling can be performed. Thereby, a more accurate labeling process can be performed.

In the twentieth aspect, the image feature amount includes area, image density sum, perimeter length, ferret diameter, ferret diameter ratio, circularity, area ratio, first moment, second moment, ellipse approximated major and minor diameters, volume. , Volume concentration sum, surface area, three-dimensional ferret diameter, three-dimensional ferret diameter ratio, sphericity, volume fraction, three-dimensional primary moment, three-dimensional secondary moment, and ellipsoid approximate diameter.
In the image processing apparatus according to the twentieth aspect, these can be measured or edited as image feature amounts. Further, there may be a case where a further effect can be obtained by adding those similar to these image feature amounts. In addition, among these image feature amounts, those having little effect may be deleted.

  According to the radiation tomography apparatus or the image processing apparatus of the present invention, there is an effect that the real-time labeling process, or the real-time labeling process, or the image feedback process to the image reconstruction process based on the result of the labeling process can be realized.

  Hereinafter, the present invention will be described in more detail with reference to embodiments shown in the drawings. Note that the present invention is not limited thereby.

<Overall configuration of X-ray CT apparatus>
FIG. 4 is a configuration block diagram of the X-ray CT apparatus 100 according to the embodiment of the present invention. The X-ray CT apparatus 100 includes an operation console 1, an imaging table 10, and a scanning gantry 20.

  The operation console 1 collects X-ray detector data collected by the input device 2 that receives input from the operator, the central processing device 3 that executes pre-processing, image reconstruction processing, post-processing, and the like, and the scanning gantry 20. A data acquisition buffer 5, a monitor 6 for displaying a tomographic image reconstructed from projection data obtained by preprocessing the X-ray detector data, a program, X-ray detector data, projection data, and X-ray tomogram And a storage device 7 for storing. The photographing condition is input from the input device 2 and stored in the storage device 7. FIG. 5 shows an example of the photographing condition input screen 13A displayed on the monitor 6. An input button 13a for performing a predetermined input is displayed on the screen shooting condition input screen 13A. FIG. 5 shows a screen in which a scan tab is selected. When P-Recon is selected as the tab, the display for input is switched as shown in the lower part of FIG. A tomographic image 13b is displayed above the input button 13a, and a reconstruction area 13c is displayed below. If necessary, a biological signal may be displayed as displayed on the upper right.

  Returning to FIG. 4, the imaging table 10 includes a cradle 12 on which a subject is placed and taken in and out of the opening of the scanning gantry 20. The cradle 12 is moved up and down and linearly moved by the motor built in the imaging table 10.

  The scanning gantry 20 includes an X-ray tube 21, an X-ray controller 22, a collimator 23, a beam forming X-ray filter 28, a multi-row X-ray detector 24, and a data acquisition device (DAS: Data Acquisition System) 25. A rotation unit controller 26 that controls the X-ray tube 21 rotating around the body axis of the subject, and a control controller 29 that exchanges control signals and the like with the operation console 1 and the imaging table 10. . The beam forming X-ray filter 28 has the thinnest filter thickness in the X-ray direction toward the center of rotation, which is the imaging center, and the filter thickness increases toward the periphery so that X-rays can be absorbed more. X-ray filter. For this reason, exposure of the body surface of the subject having a cross-sectional shape close to a circle or an ellipse can be reduced. The scanning gantry tilt controller 27 can tilt the scanning gantry 20 forward and backward in the z direction by about ± 30 degrees.

  The X-ray tube 21 and the multi-row X-ray detector 24 rotate around the rotation center IC. When the vertical direction is the y direction, the horizontal direction is the x direction, and the table and cradle traveling direction perpendicular to these are the z direction, the rotation plane of the X-ray tube 21 and the multi-row X-ray detector 24 is the xy plane. is there. The moving direction of the cradle 12 is the z direction.

  FIG. 6 is a diagram showing the geometric arrangement of the X-ray tube 21 and the multi-row X-ray detector 24 as viewed from the xy plane, and FIG. 7 shows the geometry of the X-ray tube 21 and the multi-row X-ray detector 24. It is the figure which looked at the target arrangement from the yz plane. The X-ray tube 21 generates an X-ray beam called a cone beam CB. When the central axis direction of the X-ray beam is parallel to the y direction, the view angle is 0 degree. The multi-row X-ray detector 24 has X-ray detector rows of J rows, for example, 256 rows in the z direction. Each X-ray detector row has I-channels, for example, 1024 channels of X-ray detector channels in the channel direction.

  In FIG. 6, the X-ray beam emitted from the X-ray focal point of the X-ray tube 21 is subjected to more X-rays at the center of the reconstruction area P by the beam forming X-ray filter 28 and more at the periphery of the reconstruction area P. Less X-rays are irradiated. After the X-ray dose is thus spatially controlled, X-rays are absorbed by the subject existing inside the reconstruction region P, and the transmitted X-rays are converted into X-ray detector data by the multi-row X-ray detector 24. Collected as. In FIG. 7, the X-ray beam emitted from the X-ray focal point of the X-ray tube 21 is controlled by the X-ray collimator 23 in the slice thickness direction of the tomogram, that is, the X-ray beam width becomes D at the rotation center axis IC. Thus, X-rays are absorbed by the subject existing in the vicinity of the rotation center axis IC, and the transmitted X-rays are collected by the multi-row X-ray detector 24 as X-ray detector data. Projection data acquired by irradiating the subject with X-rays is A / D converted from the multi-row X-ray detector 24 by the data acquisition device 25 and input to the data acquisition buffer 5 via the slip ring 30. . Data input to the data collection buffer 5 is processed by the central processing unit 3 according to a program in the storage device 7, reconstructed into a tomographic image, and displayed on the monitor 6. In this embodiment, the multi-row X-ray detector 24 is applied. However, a two-dimensional X-ray area detector having a matrix structure represented by a flat panel X-ray detector can be applied. A row X-ray detector can be applied.

<Operation flowchart of X-ray CT apparatus>
FIG. 8 is a flowchart showing an outline of the operation of the X-ray CT apparatus 100 of the present embodiment.

  In step P1, the subject is placed on the cradle 12 and aligned. The subject placed on the cradle 12 aligns the center position of the slice light of the scanning gantry 20 with the reference point of each part.

  In step P2, a scout image (also called a scano image or a fluoroscopic image) is collected. Scout images can capture two types of scout images for adults or children depending on the size of the body of the subject, and can usually be captured at 0 degrees and 90 degrees. Depending on the part, for example, the head may be a 90-degree scout image only. In scout imaging, the X-ray tube 21 and the multi-row X-ray detector 24 are fixed, and the data collection operation of X-ray detector data is performed while the cradle 12 is moved linearly. Details of scout image shooting will be described later with reference to FIG.

  In step P3, shooting conditions are set while displaying the position and size of the tomographic image to be shot on the scout image. In the present embodiment, a plurality of scan patterns such as a conventional scan (axial scan), a helical scan, a variable pitch helical scan, and a helical shuttle scan are provided. Conventional scanning is a scanning method in which projection data is acquired by rotating the X-ray rod 21 and the X-ray detector 24 each time the cradle 12 is moved by a predetermined pitch in the Z-axis direction. The helical scan is a scanning method for acquiring projection data by moving the cradle 12 at a predetermined speed while the X-ray tube 21 and the X-ray detection unit 24 are rotating. The variable pitch helical scan is a scan method in which the projection data is acquired by changing the speed of the cradle 12 while rotating the X-ray rod 21 and the X-ray detector 24 as in the helical scan. The helical shuttle scan is a scanning method for acquiring projection data by reciprocating the cradle 12 in the Z-axis direction or the −Z-axis direction while rotating the X-ray rod 21 and the X-ray detection unit 24 as in the helical scan. . When these multiple scans are set, X-ray dose information as a whole is displayed. In the cine scan, when the number of rotations or time is entered, X-ray dose information for the input number of rotations or the input time in the region of interest is displayed.

  In step P4, tomographic imaging is performed. Details of tomographic imaging and image reconstruction will be described later with reference to FIG. In step P5, the tomographic image reconstructed is displayed. In Step P6, a three-dimensional image is displayed as shown in FIG. 9, using tomographic images continuously taken in the z direction as a three-dimensional image.

  FIG. 9 shows a volume rendering 3D image display method 40, a MIP (Maximum Intensity Projection) image display method 41, and an MPR (Multi Plain Reformat) image display method 42 as the 3D image display method. Although there are various image display methods, any display method can be applied in the present invention, and can be properly used depending on the purpose of diagnosis.

<Operation flowchart of tomography and scout imaging>
FIG. 10 is a flowchart showing an outline of tomographic and scout imaging operations of the X-ray CT apparatus 100 of the present invention.

  In step S1, the helical scan rotates the X-ray tube 21 and the multi-row X-ray detector 24 around the subject and moves the cradle 12 on the imaging table 10 in a straight line while the X-ray detector data data. Perform the collection operation. X-ray detector data D0 (view, j, i) (j = 1 to ROW, i = 1 to CH) represented by the view angle view, detector row number j, and channel number i, and z-direction position Ztable. (View) is added and data is collected within a certain range of speed. In the variable pitch helical scan or helical shuttle scan, in addition to collecting data in a constant speed range, data collection is also performed during acceleration and deceleration. In the conventional scan (axial scan) or cine scan, the data acquisition system is rotated one or more times while the cradle 12 on the imaging table 10 is fixed at a certain z-direction position to collect data of X-ray detector data. Do. If necessary, after moving to the next position in the z direction, the data acquisition system is rotated once or a plurality of times to collect data of X-ray detector data. In scout imaging, the X-ray tube 21 and the multi-row X-ray detector 24 are fixed, and the X-ray detector data is collected while the cradle 12 on the imaging table 10 is linearly moved. .

  In step S2, the X-ray detector data D0 (view, j, i) is preprocessed and converted into projection data. FIG. 11 shows specific processing for the preprocessing in step S2 of FIG. In step T21, offset correction is performed, logarithmic conversion is performed in step T22, X-ray dose correction is performed in step T23, and sensitivity correction is performed in step T24. In the case of scout imaging, if preprocessed X-ray detector data is displayed in accordance with the pixel size in the channel direction and the pixel size in the z direction, which is the linear movement direction of the cradle 12, in accordance with the display pixel size of the monitor 6. Completed as a scout statue.

Returning to FIG. 10, in step S3, beam hardening correction is performed on the preprocessed projection data D1 (view, j, i). In the beam hardening correction in step S3, the projection data subjected to the sensitivity correction in step T24 of the preprocessing S2 is D1 (view, j, i), and the data after the beam hardening correction in step S3 is D11 (view , J, i), the beam hardening correction is expressed, for example, in a polynomial form as shown in the following (Equation 1). In this specification, the multiplication operation is represented by “●”.
... (Formula 1)
At this time, independent beam hardening correction can be performed for each j column of the detector, so if the tube voltage of each data acquisition system differs depending on the imaging conditions, the difference in the X-ray energy characteristics of the detector for each column is corrected. it can.

In step S4, a z-filter convolution process for applying a filter in the z direction (column direction) is performed on the projection data D11 (view, j, i) subjected to beam hardening correction. That is, the projection of the multi-row X-ray detector D11 (view, j, i) (i = 1 to CH, j = 1 to ROW) subjected to beam hardening correction after preprocessing in each view angle and each data acquisition system. For example, a filter having a column direction filter size of 5 columns as shown in the following (Equation 2) and (Equation 3) is applied to the data in the column direction.
(W1 (i), w2 (i), w3 (i), w4 (i), w5 (i)) (Formula 2)

However,
... (Formula 3)
The corrected detector data D12 (view, j, i) is expressed by the following (Formula 4).
... (Formula 4)
It becomes. The maximum channel value is CH,
If the maximum value of the column is ROW,
The following (Formula 5) and (Formula 6) are obtained.
... (Formula 5)
... (Formula 6)

  Further, when the column direction filter coefficient is changed for each channel, the slice thickness can be controlled according to the distance from the image reconstruction center. In general, in the tomogram, the slice thickness is thicker in the peripheral part than in the reconstruction center. For this reason, the filter coefficient can be changed between the central part and the peripheral part, and the slice thickness can be made uniform in both the peripheral part and the image reconstruction central part. For example, if the width of the column direction filter coefficient is changed widely in the vicinity of the central channel and the width of the column direction filter coefficient is changed narrow in the vicinity of the peripheral channel, the slice thickness may be changed in the central portion of the image reconstruction even in the peripheral portion. But it can be uniform.

  In this way, by controlling the column direction filter coefficients of the central channel and the peripheral channel of the multi-row X-ray detector 24, the slice thickness can also be controlled in the central portion and the peripheral portion. When the slice thickness is slightly reduced with the row direction filter, both artifacts and noise are greatly improved. Thereby, artifact improvement and noise improvement can also be controlled. That is, it is possible to control the tomographic image reconstructed, that is, the image quality in the xy plane. As another embodiment, a thin slice thickness tomographic image can be realized by using a column direction (z direction) filter coefficient as a deconvolution filter.

In step S5, reconstruction function superimposition processing is performed. That is, a fast Fourier transform (FFT) that transforms projection data into the frequency domain is performed, a reconstruction function is applied, and an inverse Fourier transform is performed. In the reconstruction function superimposing process S5, assuming that the projection data after the z filter convolution process is D12, the projection data after the reconstruction function convolution process is D13, and the reconstruction function to be superimposed is Kernel (j), the reconstruction function convolution process Is expressed as (Equation 7) below. In the present specification, the convolution operation is represented by “*”.
... (Formula 7)
That is, the reconstruction function Kernel (j) can perform an independent reconstruction function convolution process for each j column of the detector, so that differences in noise characteristics and resolution characteristics for each column can be corrected.

  In step S6, three-dimensional backprojection processing is performed on the projection data D13 (view, j, i) subjected to reconstruction function superimposition processing to obtain backprojection data D3 (x, y, z). The image to be reconstructed is a three-dimensional image reconstructed on a plane perpendicular to the z axis and on the xy plane. The following reconstruction area P is assumed to be parallel to the xy plane. This three-dimensional backprojection process will be described later with reference to FIG.

In step S7, post-processing such as image filter superimposition and CT value conversion is performed on the backprojection data D3 (x, y, z) to obtain a tomographic image D31 (x, y, z). In post-processing image filter superimposition processing, a tomographic image after three-dimensional backprojection is set to D31 (x, y, z), data after image filter superimposition is D32 (x, y, z), and xy that is a tomographic image plane. If the two-dimensional image filter superimposed on the plane is Filter (z), the following (Formula 8) is obtained.
... (Formula 8)
That is, since independent image filter superimposition processing can be performed for each tomographic image at each z coordinate position, differences in noise characteristics and resolution characteristics for each column can be corrected.

  Further, after the two-dimensional image filter convolution process, an image space z-direction filter convolution process described below may be performed. The image space z-direction filter convolution process may be performed before the two-dimensional image filter convolution process. Furthermore, a three-dimensional image filter convolution process may be performed to produce an effect that serves as both the two-dimensional image filter convolution process and the image space z-direction filter convolution process.

In the image space z-direction filter convolution process, the tomographic image subjected to the image space z-direction filter convolution process is D33 (x, y, z), and the tomographic image subjected to the two-dimensional image filter convolution process is D32 (x, y, z). Then, the following (Formula 9) is obtained. However, v (i) is an image space z-direction filter coefficient whose width in the z direction is 21 + 1, and is a coefficient sequence as shown in the following (Equation 10).
... (Formula 9)
... (Formula 10)

  In the helical scan, the image space filter coefficient v (i) may be an image space z-direction filter coefficient that does not depend on the z-direction position. However, when using a multi-row X-ray detector 24 or a two-dimensional X-ray area detector having a wide detector width in the z direction and performing a conventional scan (axial scan) or a cine scan, the image space z-direction filter coefficient It is preferable to use an image space z-direction filter coefficient depending on the position of the X-ray detector column in the z-direction for v (i). This is because it is effective because detailed adjustment depending on the column position of each tomographic image can be performed.

(Three-dimensional backprojection processing flowchart)
FIG. 12 shows details of step S6 in FIG. 11, and is a flowchart of the three-dimensional backprojection process. In the present embodiment, the image to be reconstructed is reconstructed into a three-dimensional image on a plane perpendicular to the z axis and an xy plane. The following reconstruction area P is assumed to be parallel to the xy plane.

  In step S61, each pixel in the reconstruction area P is focused on one view among all the views necessary for image reconstruction of the tomographic image (ie, a view for 360 degrees or a view for 180 degrees and a fan angle). Projection data Dr corresponding to is extracted.

  Here, the projection data Dr will be described with reference to FIGS. FIG. 13 is a conceptual diagram showing a state in which lines on the reconstruction area are projected in the X-ray transmission direction, where A indicates the xy plane and B indicates the yz plane. FIG. 14 is a conceptual diagram showing lines projected on the X-ray detector surface. As shown in FIG. 13, a 512 × 512 pixel square region parallel to the xy plane is used as a reconstruction region P, and pixel rows L0, y = 63, which are parallel to the x axis at y = 0, y = 63. 127 pixel row L127, y = 191 pixel row L191, y = 255 pixel row L255, y = 319 pixel row L319, y = 383 pixel row L383, y = 447 pixel row L447, y = 511 The pixel column L511 is taken as a column. Then, if projection data on lines T0 to T511 as shown in FIG. 14 obtained by projecting these pixel rows L0 to L511 onto the surface of the multi-row X-ray detector 24 in the X-ray transmission direction is extracted, they are extracted into the pixel row L0. To projection data Dr (view, x, y) of L511. However, x and y correspond to each pixel (x, y) of the tomographic image.

  The X-ray transmission direction is determined by the X-ray focal point of the X-ray tube 21 and the geometric position of each pixel and the multi-row X-ray detector 24, but z of the X-ray detector data D0 (view, j, i). Since the coordinate z (view) is attached to the X-ray detector data as the table linear movement z-direction position Ztable (view), the X-ray detector data D0 (view, j, i) during acceleration / deceleration is also known. In the data acquisition geometric system of the X-ray focus and multi-row X-ray detector, the X-ray transmission direction can be accurately obtained.

  For example, a part of the line goes out of the channel direction of the multi-row X-ray detector 24, such as a line T0 in which the pixel row L0 is projected on the surface of the multi-row X-ray detector 24 in the X-ray transmission direction. In this case, the corresponding projection data Dr (view, x, y) is set to “0”. Further, if the projection is out of the z direction, the projection data Dr (view, x, y) is extrapolated.

  In this way, projection data Dr (view, x, y) corresponding to each pixel in the reconstruction area P shown in FIG. 15 can be extracted.

Returning to FIG. 12, in step S62, the projection data Dr (view, x, y) is multiplied by the cone beam reconstruction weighting coefficient to create projection data D2 (view, x, y) as shown in FIG. Here, the cone beam reconstruction weighting coefficient w (i, j) is as follows. In the case of fan beam image reconstruction, generally, when view = βa, a straight line connecting the focal point of the X-ray tube 21 and the pixel g (x, y) on the reconstruction area P (on the xy plane) is the center of the X-ray beam. When the angle formed with respect to the axis Bc is γ and the opposite view is view = βb, the following (Formula 11) is obtained.
βb = βa + 180 ° −2γ (Expression 11)

If the angles formed by the X-ray beam passing through the pixel g (x, y) on the reconstruction area P and the opposite X-ray beam to the reconstruction plane P are αa and αb, the cone beam reconstruction weighting coefficient depending on these angles ωa and ωb are multiplied and added to obtain backprojected pixel data D2 (0, x, y). In this case, (Formula 12) is obtained.
D2 (0, x, y) = ωa · D2 (0, x, y) _a + ωb · D2 (0, x, y) _b (Equation 12)
However, D2 (0, x, y) _a is back projection data of the view βa, and D2 (0, x, y) _b is back projection data of the view βb.
Note that the sum of the cone beam reconstruction weighting coefficients between the opposed beams is expressed by (Equation 13).
ωa + ωb = 1 (Formula 13)
Cone angle artifacts can be reduced by multiplying and adding cone beam reconstruction weighting coefficients ωa and ωb.

For example, the cone beam reconstruction weighting coefficients ωa and ωb can be obtained by the following equations. Note that ga is a weighting coefficient for the view βa, and gb is a weighting coefficient for the view βb. When ½ of the fan beam angle is γmax, the following (Expression 14) to (Expression 19) are obtained.
(For example, q = 1)
For example, as an example of ga and gb, when max [] is a function that takes the larger value, the following (Equation 20) and (Equation 21) are obtained.

  In the case of fan beam image reconstruction, each pixel on the reconstruction area P is further multiplied by a distance coefficient. The distance coefficient corresponds to the distance from the focus of the X-ray tube 21 to the detector row j and the channel i of the multi-row X-ray detector 24 corresponding to the projection data Dr, r0, and corresponds to the projection data Dr from the focus of the X-ray tube 21. (R1 / r0) 2 where r1 is the distance to the pixel on the reconstruction area P. In the case of parallel beam image reconstruction, each pixel on the reconstruction area P may be multiplied by only the cone beam reconstruction weight coefficient w (i, j).

  In step S63, the projection data D2 (view, x, y) is added in correspondence with the pixels to the backprojection data D3 (x, y) that has been cleared in advance. FIG. 17 shows an image in which projection data D2 (view, x, y) is added in correspondence with pixels. In step S64, steps S61 to S63 are repeated for all views necessary for image reconstruction of tomographic images (that is, views for 360 degrees or “180 degrees + fan angle”), and are necessary for image reconstruction. When all the views are added, back projection data D3 (x, y) at the left end in FIG. 17 can be obtained.

  As described above, the flowchart of the three-dimensional backprojection process in FIG. 12 describes the reconstruction area P shown in FIG. 13 as a square 512 × 512 pixels. However, it is not limited to this. FIG. 18 is a conceptual diagram showing a state in which a line on a circular reconstruction area is projected in the X-ray transmission direction, where A is the xy plane and B is the yz plane. As shown in FIG. 18, the reconstruction area P may not be a square area of 512 × 512 pixels, but a circular area having a diameter of 512 pixels.

<Continuous area labeling and image quality optimization for each part>
FIG. 19 is a flowchart of this embodiment regarding the labeling of continuous regions and the optimization of image quality for each part. In the embodiment shown in FIG. 19, first, each continuous region labeled by the image feature amount of each pixel or a neighboring region of each pixel is recognized as each part by the image feature amount for each continuous region. Thereafter, the image quality is optimized for each region by a reconstruction function optimized for each region or an image filter in the image space so as to obtain an optimal image quality as the region.

  Up to step S7 in FIG. 10, a tomographic image of one two-dimensional image or a three-dimensional image composed of tomographic images continuous in the z direction, or a tomographic image continuous in the z direction in time series by a helical shuttle scan. A four-dimensional image is obtained. The following description is based on the assumption of a three-dimensional image, but a two-dimensional image or a four-dimensional image can also be applied in the same manner.

  In step S12, the three-dimensional labeling process is performed before obtaining the images from all the imaging regions up to step S7 in FIG. A three-dimensional continuous region is extracted from the three-dimensional image obtained in step S7 of FIG. 10 by a three-dimensional labeling process. When performing the labeling process, the tomographic image composed of CT values or the grayscale image which is a three-dimensional image is subjected to the labeling process by any of the following methods. The labeling process and image feature amount extraction / measurement will be described later.

  In step S14, the image feature quantity is measured simultaneously with the three-dimensional labeling process. In step S16, a three-dimensional relabeling process is performed. The reason why the three-dimensional relabeling process is performed is that the three-dimensional labeling process is performed before images from all imaging regions are obtained. In step S18, it is confirmed whether or not the three-dimensional continuous regions are combined. If the three-dimensional continuous area is not connected, recognition as one part is performed in step S22. If the three-dimensional continuous regions are combined, in step S20, the image feature amount is edited, that is, the image feature amount is merged. Thus, in step S22, recognition as one part is performed.

  That is, in the three-dimensional labeling process in step S12 and the three-dimensional relabeling process in step S16, “a three-dimensional labeling process is performed in the three-dimensional continuous region subjected to the one-dimensional three-dimensional labeling process. When the connection of the original area is detected, the re-labeling process performs a correction based on the connection of the new three-dimensional area. "

  In step S22, since the three-dimensional continuous region is recognized as one part, in step S24, image optimization is performed using a reconstruction function suitable for the part. In step S26, image optimization using an image filter suitable for the part is performed. Subsequently, post-processing is performed in step S28.

<Measurement of labeling processing and image features>
There are the following methods for the labeling process and the measurement of the image feature amount in step S12 and step S14 in FIG.

<Labeling process 1>
FIG. 20 shows the labeling process 1 and is a flowchart for performing the labeling process by binarizing the grayscale image to create a binary image. In this labeling process 1, in step L1, a three-dimensional image composed of tomographic images continuous in the z direction is input as a grayscale image and stored in a memory buffer (hereinafter referred to as a buffer) of the central processing unit 3.

In step L2, for each pixel G (x, y, z) of the grayscale three-dimensional image stored in this buffer, the conditions shown in the following (Equation 22), (Equation 23) or (Equation 24) are satisfied. A binarization process that satisfies any one of them is performed and output to a buffer. In step L3, each pixel B (x, y, z) of the binary three-dimensional image output to the buffer is stored in the buffer. Note that Th1 and Th2 are threshold values, G (x, y, z) is one pixel with a grayscale three-dimensional image, and B (x, y, z) is one pixel with a binary three-dimensional image.
In step L4, a three-dimensional labeling process is performed on the binary three-dimensional image stored in this buffer.

<Labeling process 2>
FIG. 21 shows the labeling process 2, and is a flowchart for performing the labeling process by comparing the pixel values of the continuous pixels of the grayscale image with threshold values. In the labeling process 2, as in the labeling process 1, a three-dimensional image composed of tomographic images continuous in the z direction is input as a grayscale image in step L11.

In step L12, for each pixel G (x, y, z) of the input grayscale three-dimensional image, any one of the conditions shown in (Equation 25), (Equation 26), and (Equation 27) is set. The threshold processing to satisfy is performed, and “1” or “0” is input to the labeling processing. Note that Th1 and Th2 are threshold values, G (x, y, z) is one pixel with a grayscale three-dimensional image, and B (x, y, z) is one pixel with a binary three-dimensional image.
In step L13, a three-dimensional labeling process is performed based on the value input from the threshold process.

<Labing process 3>
FIG. 22 is a flowchart showing the labeling process 3. The labeling process 3 performs a binarization process on a pixel value of a continuous pixel of interest in a grayscale image, a measured value of an image feature amount of the target pixel, and a measured value of an image feature amount of a region near the target pixel. Create a value image and perform a labeling process.

  In the labeling process 3, as in the labeling process 1, in step L21, a three-dimensional image composed of continuous tomographic images in the z direction is input as a grayscale image and stored in a buffer. In step L22, the following processing is performed for all pixels while scanning each pixel G (x, y, z) of the grayscale three-dimensional image stored in this buffer as a pixel of interest in the order of, for example, the x direction, the y direction, and the z direction. I do. As the processing of step L22, the measured value of the image feature amount of the target pixel, for example, the CT value in the tomographic image, and the image feature amount of the neighboring region of the target pixel, that is, the image feature amount or the local region image feature amount (hereinafter referred to as the neighborhood) This is referred to as an image feature amount of a region).

  FIG. 23 shows the x-direction, y-direction, and z-direction scanning of the pixel of interest in the three-dimensional image and the state of the neighboring area. FIG. 23 shows a pixel of interest and its neighboring region 3 × 3 × 3 pixels that are first scanned in the x direction, then scanned in the y direction, and then scanned in the z direction. Here, the neighboring pixels are defined as 3 × 3 × 3 pixels, but may be 5 × 5 × 5 pixels.

When measuring the image feature amount of the nearby region, for example, the following image feature amount of the nearby region is used for the three-dimensional image G (x, y, z).
(1) 5 × 5 × 5 standard deviation in the vicinity of the target pixel (2) 3 × 3 × 3 average pixel value (CT value) in the vicinity of the target pixel
(3) Median value of 5 × 5 × 5 in the vicinity of the target pixel (4) When the maximum difference absolute value target pixel in the vicinity 3 × 3 × 3 of the target pixel is P (x, y, z), the following ( In Equation 28), the maximum value is the maximum difference absolute value.
... (Formula 28)

However, a, b, c = ± 1. That is, it becomes the maximum value of the absolute value difference between the neighboring pixel within the range of the neighboring region 3 × 3 × 3 pixels and the target pixel. In addition, the thing other than the above, such as a primary differential value and a secondary differential value, may be added to the image feature amount in the vicinity region, or a part of the above may be used.
As described above, image feature values F1 (G (x, y, z)), F2 (G (x, y, z)), F3 (G (x, y, z)),... N types of (G (x, y, z)) are obtained. It should be noted that the same effect can be obtained even if the present embodiment is modified in terms of the type of image feature quantity in the neighborhood area, the size of the neighborhood pixel size, and the number of dimensions of the neighborhood pixels. In particular, the number of dimensions of neighboring pixels may not be three-dimensional, but may be two-dimensional or one-dimensional.

  In step L23, binarization processing is performed. In step L23, the binarization process that satisfies the condition shown in the following (Equation 29) is performed on each pixel G (x, y, z) of the grayscale three-dimensional image stored in the buffer, and the result is output to the buffer. To do. In step L24, each pixel B (x, y, z) of the binary three-dimensional image output to the buffer is stored in the buffer.

  In addition, a spatial region (Mahalanobis space region) of an image feature amount that should satisfy the image feature amount of the target pixel is PR1, and a spatial position of the image feature amount of the target pixel is R1 (G (x, y, z)). Also, the image feature quantity space area (Mahalanobis space area) to be satisfied by the image feature quantity in the neighborhood area of the target pixel is PR2, and the spatial position of the image feature quantity in the neighborhood area of the target pixel is R2 (G (x, y, z )).

In addition, Th1 and Th2 are threshold values, G (x, y, z) is one pixel with a grayscale three-dimensional image, and B (x, y, z) is one pixel with a binary three-dimensional image.
... (Formula 29)
In step L25, a three-dimensional labeling process is performed on the binary three-dimensional image stored in this buffer.

<Labeling process 4>
FIG. 24 is a flowchart showing the labeling process 4. Threshold processing is performed on the pixel value of the target pixel of the continuous value of the grayscale image in the labeling process 4, the measured value of the image feature amount of the target pixel, and the measured value of the image feature amount in the vicinity region of the target pixel to perform the labeling process. .

  In step L31, a grayscale image is input. As in the labeling process 2, input is performed as a three-dimensional gray image composed of continuous tomographic images in the z direction.

  In step L32, the image feature amount of the pixel of interest is measured, and the image feature amount of a region near the pixel of interest is measured. For example, processing is performed on each pixel of the grayscale three-dimensional image input while scanning as a target pixel in the order of x direction, y direction, and z direction. At this time, as in step L22 of the labeling process 3, the image feature value of the target pixel and the image feature amount in the vicinity region of the target pixel are also measured.

In step L33, threshold processing is performed. Threshold processing is performed for pixels that satisfy the conditions shown in (Equation 30) below. It is assumed that the image feature amount space region (Mahalanobis space region) to be satisfied by the image feature amount of the target pixel is PR1, and the spatial position of the image feature amount of the target pixel is R1 (G (x, y, z)). Also, the image feature quantity space area (Mahalanobis space area) to be satisfied by the image feature quantity in the neighborhood area of the target pixel is PR2, and the spatial position of the image feature quantity in the neighborhood area of the target pixel is R2 (G (x, y, z )). Further, Th1 and Th2 are threshold values, and G (x, y, z) is one pixel having a grayscale three-dimensional image.
... (Formula 30)
In step L34, a three-dimensional labeling process is performed based on the value input from the threshold process.

  As described above, the three-dimensional labeling process is performed in step L4 of the labeling process 1, step L13 of the labeling process 2, step L25 of the labeling process 3, and step L34 of the labeling process 4. As shown in FIG. 23, in the three-dimensional labeling process, each pixel input from the binary image or each pixel input from the threshold value process is stored in a buffer. In the three-dimensional labeling process, the neighboring mask region including the target pixel as shown in FIG. 23 is scanned in the order of x direction, then y direction, and then z direction.

  Then, “the 3D continuous area“ 1 ”and the 3D continuous area“ 2 ”are connected at the z coordinate position z1” is stored in the connection information table ”. Note that a neighborhood mask as shown in FIG. 25 is used to see the connection with neighboring pixels in the three-dimensional labeling process. FIG. 25 shows a neighborhood mask area in the vicinity of 26 (3 × 3 × 3-1 = 26). As other examples of the neighborhood mask, 18 neighborhood (5 + 8 + 5) neighborhood mask regions with different neighborhood conditions are shown in FIG. 26A, and 6 neighborhood (1 + 4 + 1) neighborhood mask regions are shown in FIG. 26B.

  A neighboring pixel considered in the vicinity of the target pixel in the vicinity of the pixel 26 is an area in which the coordinate is small in the z direction, the area in which the coordinate is small in the y direction, and the direction in which the coordinate is small in the x direction. A region satisfying the region is set as a three-dimensional neighborhood mask region. When scanning the pixel of interest in the x, y, and z directions, this 3D neighborhood mask is also scanned, and whether or not a 3D continuous region exists in this 3D neighborhood mask region, It can be determined that the three-dimensional continuous area detected in the three-dimensional neighborhood mask area is connected. By scanning this process in the x direction, y direction, and z direction, the numbering process of the three-dimensional continuous region in the three-dimensional image can be performed. The above description is made into a flowchart as follows.

FIG. 27 is a flowchart of the three-dimensional labeling process in step L4, step L13, step L25, and step L34.
In step M1, x = 1, y = 1, and z = 1.
In step M2, the value of the target pixel G (x, y, z) is input.
In Step M3, it is determined whether or not the target pixel is “1”. If YES, the process goes to Step M4, and if NO, the process goes to Step M10.
In step M4, it is determined whether there is a three-dimensional area number other than “0” in the neighborhood mask area. If YES, the process goes to step M5, and if NO, the process goes to step M8.
In Step M5, it is determined whether there are two or more. If YES, go to Step M6, and if NO, go to Step M9.
In step M6, the smallest three-dimensional area number is assigned to the target pixel.
In step M7, two or more three-dimensional area numbers are written in the three-dimensional connection information table as being connected. After this, go to step M10.
In step M8, the latest three-dimensional area number is assigned to the target pixel. After this, go to step M10.
In step M9, one existing three-dimensional region number is assigned to the target pixel. After this, go to step M10.
In step M10, it is determined whether x = xe. If YES, go to step M11, and if NO, go to step M15.
In step M11, x = 1.
In step M12, it is determined whether y = ye. If YES, go to step M13, and if NO, go to step M16.
In step M13, y = 1.
In step M14, it is determined whether z = ze. If YES, the process ends. If NO, the process goes to step M17.
In step M15, x = x + 1. After this, go to step M2.
In step M16, y = y + 1. After this, go to step M2.
In step M17, z = z + 1 is set. After this, go to step M2.

  In this way, when the pixel of interest is a pixel input from the binary image or the pixel input from the threshold processing is “1”, there is no 3D continuous region number in the neighboring mask region including the pixel of interest. A new three-dimensional continuous region number is taken, and when there is one three-dimensional continuous region number, the three-dimensional continuous region number is assigned to the target pixel. In addition, when there are a plurality of three-dimensional continuous region numbers, the three-dimensional continuous region number having the smallest number is assigned to the pixel of interest, and the plurality of three-dimensional continuous region numbers are assumed to be connected in a three-dimensional space, and the connection information table indicates that Write information. Based on the information in the connection information table, a three-dimensional relabeling process is performed.

<Re-labeling and image feature editing>
As described above, after the three-dimensional labeling process is performed, the three-dimensional relabeling process is performed. There are several methods for the three-dimensional relabeling process in step S16 in FIG.

  FIG. 28 is a diagram illustrating feedback of the labeling process to the re-labeling process. As shown in FIG. 28, the three-dimensional relabeling process proceeds with a delay of a certain interval. When the three-dimensional labeling process proceeds in the positive z direction, in the tomographic image G (z2) at the z-coordinate position z2 in the three-dimensional image, the three-dimensional continuous region “1” intersects at the cross section saz2, and the three-dimensional continuous region “ 2 ″ intersect at the cross section sbz2. When the three-dimensional labeling process further proceeds in the positive z direction, in the tomographic image G (z1) at the z coordinate z1, the three-dimensional continuous region “1” intersects with the cross section saz1, and the three-dimensional continuous region “2” has the cross section. Intersect with sbz1. The cross section saz1 and the cross section sbz1 are in contact at a point C. That is, it can be seen that the three-dimensional continuous region “1” and the three-dimensional continuous region “2” are connected at the point C in the tomographic image G (z1).

  In FIG. 28, the three-dimensional relabeling process is performed with a delay of the distance (Z1-Z2). On the other hand, when viewed from the three-dimensional relabeling process, only the connection information of the three-dimensional continuous area in the z direction is known by the distance (Z1-Z2). For this reason, the longer this distance is taken, the more the connection information in the z direction is known, and more accurate three-dimensional relabeling processing can be performed, but the start timing of the three-dimensional relabeling processing is delayed by that amount. The entire time will be longer. In this way, the accuracy of the connection information in the labeling process and the overall processing time of the labeling process including the re-labeling process cannot be compatible and have a trade-off relationship. Actually, it is preferable to change the distance between the labeling process and the relabeling process in accordance with the change in the shape of the subject in the z direction. That is, it is preferable to determine the distance between the labeling process and the relabeling process for each part based on the expected shape change for each part. There are several methods for 3D relabeling.

<Re-labeling processing method 1>
In the example of the three-dimensional relabeling process shown in FIG. 29, when the three-dimensional labeling process is performed once and it is detected that the three-dimensional continuous area “1” and the three-dimensional continuous area “2” are connected, In the labeling process, all the pixels written as “2” in the three-dimensional continuous region “2” are converted to “1”. According to this method, at the time of the three-dimensional relabeling process, it is necessary to scan and process the same amount of pixels as the three-dimensional labeling process. That is, it is assumed that the number of pixels of the three-dimensional image is LX / LY / LZ.

  At this time, if the 3D labeling process is performed while scanning the LX, LY, and LZ pixels in the 3D image, the LX, LY, and LZ pixels in the 3D image are scanned during the 3D relabeling process. The 3D image area number whose connection is found is converted into the original 3D image area number connected. In other words, if the 3D labeling process takes 1T, the 3D relabeling process also takes 1T. Here, LX is the number of pixels in the x direction of the three-dimensional image, LY is the number of pixels in the y direction of the three-dimensional image, and LZ is the number of pixels in the z direction of the three-dimensional image.

<Re-labeling processing method 2>
In the example of the three-dimensional relabeling process shown in FIG. 30-1, the region information of the three-dimensional continuous region “1” and the three-dimensional continuous region “2” and the image feature amount measured simultaneously with the three-dimensional labeling processing are entered. This is an example of recognizing that these two three-dimensional continuous areas are connected by associating files.

  According to this method, during the 3D labeling process, the 3D labeling process is performed while scanning the LX, LY, and LZ pixels in the 3D image. However, in the 3D relabeling process, there is no need to scan the pixels. Can process. FIG. 31 shows a flowchart of the three-dimensional relabeling process at this time.

In step R1, I is initialized to I = 1.
In step R2, each connected three-dimensional continuous area number in the first row of the connection information table is extracted. The connection area number in each row is extracted from the connection information table shown in FIG. As a result, it is possible to know what number region and what number region of the three-dimensional continuous region obtained by the three-dimensional labeling process are connected.

  In step R3, the area information table of the extracted connection area number is read out. The area information tables of a plurality of areas that are found to be connected are read out. As illustrated in FIG. 30A, the region information table of the three-dimensional continuous region “1” and the three-dimensional continuous region “2” includes a continuous region number, a connected region number, and an image feature amount. In addition to this, information on run coordinates indicating the coordinates constituting the three-dimensional continuous region (the run length-coded start point and the length of the line segment, or the start point and the end point are referred to as run coordinates) may be added. . An example of run coordinate information is shown in FIG. FIG. 33A illustrates run coordinate information on a binarized tomographic image at each z coordinate position. For example, as shown in FIG. 33-2, in the binarized tomographic image that is a part of the three-dimensional continuous region at the z-direction coordinate position z = z1, the run coordinate information of y = y11 is the run coordinate start point ( The x coordinate start point is xs11 and the run coordinate end point (x coordinate end point) is xe11.

  In step R4, the region information tables of the three-dimensional continuous regions with the extracted connection region numbers are merged and combined. At the same time, the image features are combined and combined. In order to merge the region information tables of the three-dimensional continuous regions with the extracted connection region numbers, the region numbers connected to the region information tables of the three-dimensional continuous regions are written and correlated.

  In step R5, it is determined whether the connected three-dimensional continuous area in all connection information tables has been completed. If YES, the process ends. If NO, the process goes to step R6. In step R6, I = I + 1. To step R2.

  As described above, when obtaining an image feature amount across a plurality of associated three-dimensional continuous regions, the image feature amount may be calculated while following this association. In FIG. 30-2, the area information tables of the three-dimensional continuous area “1” and the three-dimensional continuous area “2” are combined to form a new three-dimensional continuous area “1” area information table. At this time, the image feature amount is also obtained by combining.

Specifically, the volume and the surface area are obtained by adding each of two three-dimensional continuous regions “1” and three-dimensional continuous regions “2”. This may be the volume or surface area of the new three-dimensional continuous region “1”. Further, the center-of-gravity position x-coordinate and y-coordinate can obtain the center-of-gravity coordinates (g3x, g3y) of the newly combined three-dimensional continuous area as shown in the following (Equation 31). The center of gravity coordinates of the three-dimensional continuous area “1” is (g1x, g1y), the center of gravity coordinates of the three-dimensional continuous area “2” is (g2x, g2y), and the volume of the three-dimensional continuous area “1” is V1. The volume of the three-dimensional continuous region “2” is V2, and the volume of the newly merged three-dimensional continuous region is V3.
... (Formula 31)
In this way, the area information table including the image feature amount can be combined. In addition, when run coordinate information is attached, run coordinate information can also be combined.

FIG. 33-3 is a diagram illustrating an example of merged run coordinates. Here, for the sake of simplicity, a description is given using a two-dimensional continuous region instead of a three-dimensional continuous region, but the same concept applies to a three-dimensional continuous region. In FIG. 33-3, there is run coordinate information of the continuous region 1 and run coordinate information of the continuous region 2. When combining these, check if there is a run coordinate that can be combined for each y coordinate. That is, in this case, it is checked whether or not there is a portion where the continuous region 1 and the continuous region 2 overlap. That is, it is examined whether or not the following (Expression 32) or (Expression 33) may be satisfied. However, i is an integer.
If it is found that this condition is met, the two run coordinate information are merged. In the case of FIG. 33-3, when y = y6 and ae6 = bs6, the run coordinate information of y = y6 was (as6, ae6) and (bs6, be6) merged into (as6, be6). Yes.

A flowchart of this process is shown in FIG. However, yN is the maximum coordinate in the y direction. In step B1, i = 1. In step B2, the run coordinates of y = yi are read. In step B3, it is determined whether or not there is a run coordinate that satisfies (Expression 32) or (Expression 33). If YES, the process goes to Step B4, and if NO, the process goes to Step B5.
In step B4, the run coordinates are merged. In step B5, it is determined whether i = N. If YES, the process ends. If NO, the process goes to step B6. In step B6, i = i + 1 is set, and the process returns to step B2.

  As described above, the three-dimensional continuous region subjected to the three-dimensional labeling process and the three-dimensional relabeling process in steps S12 to S20 in FIG. 19 is recognized as a part of the three-dimensional continuous region in step S22.

<Recognition of part of 3D continuous area>
FIG. 35 shows a flowchart of 3D continuous area recognition of a 3D image.
In step D11, each three-dimensional continuous area is extracted. Each three-dimensional continuous area that is three-dimensionally labeled (with a continuous area number) is extracted one by one, and the following processing is performed for all three-dimensional continuous areas.

In step D12, the image feature amount of each three-dimensional continuous area is measured. At least one of the following image feature amounts is obtained for each three-dimensional continuous region extracted in step D11.
Volume (number of pixels), surface area, average pixel value, pixel value sum (CT value sum, density sum), pixel value standard deviation, x, y, z direction ferret diameter, ellipsoid ratio, sphericity, xy plane area ratio, yz plane area ratio, xz plane area ratio, three-dimensional first moment, three-dimensional second moment.
It should be noted that a part of the above-described image feature values of each three-dimensional continuous region may be used, or the same effect can be obtained even if added.

  In step D13, each three-dimensional continuous area is recognized as each part. Based on the value of the image feature value of each three-dimensional continuous area, it is determined whether the condition for each selected part is satisfied, and the part is recognized. FIG. 36 shows an example of each area recognized as each part.

  In this way, a three-dimensional continuous region that is a candidate region of each part is extracted from the three-dimensional image using image feature amounts of a plurality of neighboring regions, and each part is extracted from the image feature amount as the three-dimensional continuous region. Can be judged and recognized. In this embodiment, in step L2 for binarization processing or step L12 for threshold processing, etc., a logical expression of a plurality of image feature quantities is obtained and binarization processing or threshold processing is performed, and then three-dimensional labeling is performed. Although processing is performed, the following may be performed.

  Binarization is performed on the image feature amount of each neighboring region within a predetermined threshold range. Next, a three-dimensional labeling process is performed on the binarized or threshold-processed three-dimensional image of each neighboring region image feature amount (local region image feature amount) to obtain an image feature amount of each neighboring region. A logical expression (for example, logical product) may be obtained between the three-dimensional continuous regions subjected to the three-dimensional labeling process. As described above, three-dimensional continuous region recognition of a three-dimensional image can be performed based on image feature amounts of a plurality of neighboring regions. This can be performed more accurately and correctly than three-dimensional continuous region recognition of a three-dimensional image using only normal CT values.

After recognition as a three-dimensional continuous area, the image quality of each three-dimensional continuous area is optimized as in the following two embodiments. The image optimization by the reconstruction function in step S24 in FIG. 19 and the image optimization by the image filter in step S28 correspond to this.
Example 1: Example in which image quality is optimized by a reconstruction function and an image filter.
Example 2: An example in which the image quality is optimized by a reconstruction function and an image filter, and the boundary of a three-dimensional continuous region is continuously changed.

  FIG. 37 shows a flowchart of the first embodiment which is an embodiment in which the image quality is optimized by the reconstruction function and the image filter. In the first embodiment, steps H1 to H5 are the standard image reconstruction described in FIG. Steps H1 to H5 in this embodiment are the standard image reconstruction described in FIG. Although the beam hardening correction in step S3 and the Z filter convolution process in step S4 in FIG. 10 are omitted here, the beam hardening correction and Z filter convolution process may be included in the flowchart in FIG.

  In addition, the three-dimensional continuous area division in step H6 and the recognition of each three-dimensional continuous area in H7 as a part of the three-dimensional labeling process and the three-dimensional relabeling process from step S18 to step S22 in step FIG. , And image feature value measurement and editing. In step H8, the reconstruction function is superimposed on the three-dimensional continuous region. In this case, the fast Fourier transform (FFT) X-ray projection data obtained in step H3 can be reused as if the dotted line is connected from step H3 to step H8 in FIG. In the reconstruction step H9, image filtering processing of the three-dimensional continuous area is performed. FIG. 38 shows an example of a flowchart for reconstructing a tomographic image of each segment area with an optimal reconstruction function and an optimal image filter. FIG. 39 is a conceptual diagram for helping understanding of this flowchart. FIG. 39 shows an example when the lung field is imaged. In this case, there are segment regions of the heart, lung field, soft tissue, and bone, which are recognized as the heart, lung field, soft tissue, and bone, respectively. Each of these is reconstructed using an optimum reconstruction function and an optimum image filter. Reconstruction function for the heart and image filter for heart in the region of the heart, Reconstruction function for the lung field and image filter for the lung field in the region of the lung field, Reconstruction function for the soft tissue and soft tissue in the region of the soft tissue The image reconstruction filter and the bone reconstruction function and the bone image filter are optimal for the bone region.

  In step R1, optimum reconstruction functions K1 to KN and optimum image filters F1 to FN are examined for each of N segment areas S1 to SN. In step R2, i = 1. In step R3, each segment area Si is used as an image reconstruction area, image reconstruction is performed with an optimum reconstruction function Ki for each segment area Si, and an optimum image filter Fi is superimposed to reconstruct a tomographic image ISi. To do. In step R4, it is determined whether i = N. If YES, the process goes to step R6, and if NO, the process goes to step R5. In step R5, i = i + 1 is set, and by returning to step R3, steps R3 to R5 are repeated.

In step R6, a tomographic image I reconstructed using the optimum reconstruction function Ki and optimum image filter Fi for each three-dimensional continuous region is obtained. At this time, the tomographic image I is expressed by the following (Equation 34).
... (Formula 34)

Note that ∪ is a logical sum (OR). That is, in FIG. 39, only the heart segment region S1 is reconstructed with the cardiac reconstruction function K1, and the tomographic image IS1 is obtained by superimposing the image filter F1. Further, only the lung field segment region S2 is reconstructed with the lung field reconstruction function K2, and the tomographic image IS2 is obtained by superimposing the image filter F2. Also, only the soft tissue segment region S3 is reconstructed with the soft tissue reconstruction function K3, and the image filter F3 is superimposed to obtain the tomographic image IS13. Further, only the bone segment region S4 is reconstructed with the bone reconstruction function K4, and the tomographic image IS4 is obtained by superimposing the image filter F4. These logical sums become a tomographic image I in which each segment region shown in the right column of FIG. 39 is reconstructed with an optimum reconstruction function and an optimum image filter. A tomographic image I reconstructed with an optimum reconstruction function for each segment region, that is, each part is obtained by the following (Equation 35).
... (Formula 35)

  The point that the image quality of the boundary of the three-dimensional continuous region and the vicinity of the boundary in Step R6 of FIG. However, since the image quality of each segment region is optimized by an optimal reconstruction function or optimal image filter, the image quality may change discontinuously at and near the boundary of the segment region of each region, There is a possibility that unnaturalness remains in image quality. Therefore, there is a possibility that the image quality needs to be continuously changed.

  In the second embodiment, an embodiment in which the image quality at the boundary of the three-dimensional continuous region is continuously changed will be described. The overall flowchart of this embodiment is the same as that of the first embodiment shown in FIG. In addition, although the beam hardening correction | amendment of step S3 of FIG. 10 and the z filter convolution process of step S4 are omitted, these may be included.

  In this embodiment as well, as in the first embodiment, steps H1 to H5 are the standard image reconstruction described in FIG. In addition, the three-dimensional continuous area division in step H6 and the recognition of each three-dimensional continuous area in H7 as a part of the three-dimensional labeling process and the three-dimensional relabeling process from step S18 to step S22 in FIG. , And image feature value measurement and editing. In step H8, the reconstruction function is superimposed on the three-dimensional continuous area, and in step H9, the image filter processing of the three-dimensional continuous area is performed.

  FIG. 40 is a flowchart for reconstructing a tomographic image with an optimum reconstruction function and an optimum image filter at the boundary of the three-dimensional continuous region, and reconstructing each three-dimensional continuous region at step H8 in FIG. The function superimposition and image filter processing of each three-dimensional continuous area in step H9 will be described. FIG. 41 is a conceptual diagram for helping understanding of this flowchart.

  In step R11 in FIG. 40, each three-dimensional continuous region is expanded by L pixels. By expanding each three-dimensional continuous area by L pixels, each three-dimensional continuous area having a width of 2 L pixels is overlapped.

  In step R12, optimum reconstruction functions K1 to KN and optimum image filters F1 to FN are examined for each of the N three-dimensional continuous regions S1 to SN. FIG. 41 shows an example when the lung field is imaged. In this case, there are three-dimensional continuous regions of the heart, lung field, soft tissue, and bone, which are recognized as the heart, lung field, soft tissue, and bone, respectively. Each of these is reconstructed using an optimum reconstruction function and an optimum image filter.

Reconstruction function for the heart and image filter for heart in the region of the heart, Reconstruction function for the lung field and image filter for the lung field in the region of the lung field, Reconstruction function for the soft tissue and soft tissue in the region of the soft tissue The bone reconstruction function and the bone image filter are optimal for the bone image filter and the bone part. In this case, four kinds of reconstruction functions and image filters are prepared.
In step R13, i = 1 is initialized.

In step R14, each three-dimensional continuous area Si expanded by L pixels is used as an image reconstruction area, image reconstruction is performed with the optimum reconstruction function Ki for each three-dimensional continuous area Si, and an optimum image filter Fi is applied. The tomographic image ISi is reconstructed.
In step R15, it is determined whether i = N. If YES, the process goes to step R17. If NO, go to step R16.

  In step R16, i = i + 1. Steps R13 to R16 are repeated. The processing from step R11 to step R16 in FIG. 40 is as follows with reference to FIG. Using each of the four types of reconstruction functions and image filters prepared in step R12, only the cardiac segment region S1 is reconstructed with the cardiac reconstruction function K1, the image filter F1 is superimposed, and the tomographic image IS1 is superimposed. Ask. Further, only the lung field segment region S2 is reconstructed with the lung field reconstruction function K2, and the tomographic image IS2 is obtained by superimposing the image filter F2. Also, only the soft tissue segment region S3 is reconstructed with the soft tissue reconstruction function K3, and the image filter F3 is superimposed to obtain the tomographic image IS13. Further, only the bone segment region S4 is reconstructed with the bone reconstruction function K4, and the tomographic image IS4 is obtained by superimposing the image filter F4.

Next, in step R17 in FIG. 40, a weighted addition is performed by applying a weighted addition coefficient to the tomographic image ISi that has been reconstructed using the optimum image filter Fi that is the optimum reconstruction function Ki for each segment region. Find the image I. At this time, the tomographic image I is expressed by the following (Equation 36).
... (Formula 36)
Here, ∪ is a logical sum (OR), and Wi is a weighted addition coefficient for each part.

  Here, the weighted addition in step R17 in FIG. 40 will be described with reference to FIG. A tomographic image 71 of the lung field is displayed at the top of FIG. Below that, a tomographic image 72 in which the heart is enlarged is displayed. In the enlarged tomographic image 72, the segmental region of the heart is displayed with a solid line, the segmented region of the heart is displayed with a chain line, and the segmental region of the lung field is displayed with a solid line. A region obtained by expanding the lung field segment region by L pixels is indicated by a chain line. Each segment region and each weighted addition coefficient are displayed below the enlarged tomographic image 72 in accordance with the profile P crossing the enlarged tomographic image.

  Relationship between the segment area of the expanded heart and the weighted addition coefficient of the heart, the relationship between the segment area of the soft tissue and the weighted addition coefficient of the soft tissue, the segment area of the expanded lung field and the weighted addition coefficient of the lung field You will understand the relationship. On the profile P crossing the enlarged tomographic image, the expanded heart region is multiplied by the weighted addition coefficient for heart, the expanded soft part region is multiplied by the weighted addition coefficient for soft tissue, and the expanded lung region is lung tissue. By multiplying the weighted addition coefficients for use and weighting and adding them, the image quality on the profile P continuously changes as lung field → soft tissue → heart → soft tissue → lung field. What is necessary is just to extend the example on the line of this one-dimensional profile P to a two-dimensional tomographic image, a three-dimensional three-dimensional image, and a four-dimensional time series three-dimensional image.

That is, in FIG. 41, the tomographic image IS1 of the heart is multiplied by the weighted addition coefficient W1 for the heart. Further, the lung field tomographic image IS2 is multiplied by a lung field weighted addition coefficient W2. Further, the soft tissue tomographic image IS13 is multiplied by the soft tissue weighted addition coefficient W3. Further, the bone tomographic image IS4 is multiplied by the bone weighted addition coefficient W4. An optimal reconstruction function with continuous image quality at the boundary of each segment area, and a tomographic image I reconstructed with an optimal image filter are obtained by the following (Equation 37).
... (Formula 37)

  An example of the expansion three-dimensional logic filter used in Example 2 is shown in FIGS. 43, 44-1, and 44-2. FIG. 43 is a diagram showing the structure of the three-dimensional logic filter, the position of the target pixel, and the scanning method of the three-dimensional image. FIGS. 44A and 44B are diagrams illustrating an example of the expansion three-dimensional logic filter. In this expansion three-dimensional logic filter, when the target pixel is “0”, if there is at least one “1” pixel in the neighboring pixels 3 × 3 × 3 of the target pixel, the target pixel is “ It is made so as to expand a binary three-dimensional region by setting it to 1 ″. Note that the target pixel is scanned and moved for all the pixels in the three-dimensional region.

  In the first embodiment and the second embodiment, the example in which the image quality is optimized by the reconstruction function and the image filter for each three-dimensional continuous region is shown. When the subject is actually imaged, the process of data collection, image reconstruction, three-dimensional labeling, and image optimization shown in FIG. 2 is not easy to use unless it is processed smoothly in real time. However, it takes time to store, read, and preprocess X-ray projection data (scan data) on a disk between scan data collection and image reconstruction processing. In addition, between the image reconstruction process, the 3D labeling process, the 3D relabeling process, and the image optimization process, it is necessary to see the connection information of a certain distance in the z direction. And you need a certain time delay.

In FIG. 45, the processing of the first and second embodiments is divided into the following three, and the timing at which the processing is performed is shown.
1. Scan and data collection2. 2. Image reconstruction processing and three-dimensional labeling processing Three-dimensional relabeling processing and image optimization processing By performing processing timing control in this way, the processing time of the processing of the first and second embodiments is optimized. That is, it is possible to perform the three-dimensional labeling process by the pipeline process and the three-dimensional relabeling process based on the result in synchronization with the photographing of the three-dimensional image composed of the tomographic images continuous in the z direction.

From the viewpoint of control, the control flowchart of the first and second embodiments is as shown in FIG.
Process 1 includes data collection at step F1, preprocessing at step F2, beam hardening correction at step F3, and z filter convolution process at step F4. Process 2 includes the reconstruction function convolution process in step F5, the three-dimensional backprojection process in step F6, and the post-process in step F7. Further, the three-dimensional continuous area in step F8 is recognized as each part, and optimization of the image reconstruction condition is set as process 3. Then, the process 3 is a feedback process of the process 2. That is, the image feedback process of the image reconstruction process of the X-ray CT apparatus 100 is used. Thus, a three-dimensional image composed of tomographic images reconstructed by applying image feedback processing or tomographic images continuous in the z direction can be converged to better image quality.

  That is, in the X-ray CT apparatus 100 having the multi-row X-ray detector 24, there is an effect that real-time labeling processing in various scans or image feedback processing to image reconstruction processing based on the result of labeling processing can be realized.

  The image reconstruction method in the present embodiment may be a three-dimensional image reconstruction method by a conventionally known Feldkamp method. Furthermore, other three-dimensional image reconstruction methods may be used. Alternatively, two-dimensional image reconstruction may be used.

  This embodiment is written in the case of a helical scan, but the same effect is also obtained in the case of a variable pitch helical scan, a helical shuttle scan, a conventional scan (axial scan) at a plurality of positions continuous in the z direction, or a cine scan. Can be put out.

  Although the present embodiment is written when the scanning gantry 20 is not tilted, the same effect can be obtained even in the case of so-called tilt scanning in which the scanning gantry 20 is tilted. Furthermore, although this embodiment is written in the case where it does not synchronize with a biological signal, the same effect can be produced even if it synchronizes with a biological signal, especially a heartbeat signal. In this embodiment, an X-ray CT apparatus having a multi-row X-ray detector such as a two-dimensional X-ray area detector having a matrix structure represented by a flat panel X-ray detector is described. The same effect can be obtained in the X-ray CT apparatus of the row X-ray detector.

  Also, in this embodiment, column-direction (z-direction) filters with different coefficients are superimposed on each column to adjust image quality variation, and to achieve uniform slice thickness, artifact, and noise image quality in each column. However, various z-direction filter coefficients are conceivable for this, and any of them can produce the same effect.

  In this embodiment, it is written based on a medical X-ray CT apparatus, but can be used in an X-ray CT-PET apparatus, an X-ray CT-SPECT apparatus, etc. combined with an industrial X-ray CT apparatus or another apparatus.

A is a diagram showing a case where the three-dimensional labeling process is performed halfway through the three-dimensional image, and B is a diagram showing a case where the three-dimensional labeling process of the three-dimensional image is further advanced. It is a figure which shows a three-dimensional labeling process and a three-dimensional re-labeling process. It is a figure which shows the labeling process of this invention. It is a block diagram which shows the X-ray CT apparatus concerning one Embodiment of this invention. It is a figure which shows the imaging condition input screen of a X-ray CT apparatus. It is explanatory drawing which looked at the X-ray plane of the X-ray generator (X-ray tube) and the multi-row X-ray detector. It is explanatory drawing which looked at X-ray generator (X-ray tube) and the multi-row X-ray detector in yz plane. It is a flowchart which shows the flow of subject imaging. It is a figure which shows the example of a three-dimensional image display. It is a flowchart which shows schematic operation | movement of the image reconstruction of the X-ray CT apparatus which concerns on one Embodiment of this invention. It is a flowchart which shows the detail of pre-processing. It is a flowchart which shows the detail of a three-dimensional image reconstruction process. It is the state which projected the line on a reconstruction area | region to the X-ray transmissive direction, A is a conceptual diagram which shows xy plane and B is yz plane. It is a conceptual diagram which shows the line projected on the X-ray detector surface. It is a conceptual diagram which shows the state which projected the projection data Dr (view, x, y) on the reconstruction area. It is a conceptual diagram which shows the backprojection pixel data D2 of each pixel on a reconstruction area. It is explanatory drawing which shows the state which obtains backprojection data D3 by adding all the views to backprojection pixel data D2 corresponding to a pixel. It is the state which projected the line on a circular reconstruction area | region to the X-ray transmissive direction, A is a conceptual diagram which shows xy plane and B shows yz plane. It is a flowchart which optimizes an image quality for every labeling and site | part of a continuous area | region. It is a flowchart of the labeling process 1. It is a flowchart of the labeling process 2. FIG. It is a flowchart of the labeling process 3. It is a figure which shows the attention pixel and neighborhood area | region of a three-dimensional image. It is a flowchart of the labeling process 4. It is a figure which shows the attention pixel and 26 vicinity mask area | region of 26 vicinity of a three-dimensional labeling process. A is a diagram showing a pixel of interest and a neighborhood mask region in the vicinity of 18 in the three-dimensional labeling process, and B is a diagram showing a pixel of interest and a neighborhood mask region in the vicinity of 6. FIG. It is a flowchart which shows the flow of a labeling process. It is a figure which shows the feedback to a re-labeling process after a labeling process. It is a figure which shows an example of a three-dimensional relabeling process. It is a figure which shows another example of a three-dimensional relabeling process. It is a figure which shows the other example of a three-dimensional relabeling process, and unites an image feature-value. It is a flowchart which shows the 3D re-labeling using the connection information table and the area | region information table of the connected 3D continuous area | region. It is a figure which shows the example of a connection information table. It is a figure which shows the example of run coordinate information. It is a figure which shows the start point and end point of a run coordinate. It is a figure which shows merge | combination of a run coordinate. FIG. 34 is a flow chart of merging of the run coordinates in FIG. It is a flowchart of 3D continuous area recognition of a 3D image. It is a figure which shows the three-dimensional continuous area | region recognized by each site | part. It is a flowchart of a process in the case of optimizing the image quality of each three-dimensional continuous area | region with a reconstruction function and an image filter. 10 is a flowchart for reconstructing a tomographic image with an optimal reconstruction function and an optimal image filter for each three-dimensional continuous region. It is a figure which shows the tomogram which image-reconstructed each three-dimensional continuous area | region with the optimal reconstruction function and the optimal image filter. It is a flowchart for reconstructing a tomographic image with an optimum reconstruction function and an optimum image filter at the boundary of each three-dimensional continuous region with continuous image quality. It is a figure which shows the tomogram which reconfigure | reconstructed the image of the boundary of each three-dimensional continuous area | region with the optimal reconstruction function and the optimal image filter by continuous image quality. It is a figure which shows the tomographic image by which each three-dimensional continuous area | region was weighted and added and image reconstruction was carried out with the optimal image quality. It is a figure which shows the structure of a three-dimensional logic filter, the position of a focused pixel, and the scanning method of a three-dimensional image. It is a figure which shows an example of the three-dimensional logic filter for expansion | swelling. It is a figure which shows an example of the three-dimensional logic filter for expansion | swelling. It is a figure which shows the example of optimization of the timing of each process. It is a figure which recognizes a three-dimensional continuous area | region as each site | part, and shows the image feedback by optimization of an image reconstruction condition. It is a figure which shows the conventional labeling process.

Explanation of symbols

1 Operation Console 2 Input Device 3 Central Processing Unit 5 Data Collection Buffer 6 Monitor 7 Storage Device 10 Imaging Table 12 Cradle 15 Rotating Unit 20 Scanning Gantry 21 X-ray Tube 22 X-ray Controller 23 Collimator 24 Multi-row X-ray Detector 25 Data Collection Equipment (DAS)
26 Rotating section controller 27 Scanning gantry tilt controller 28 Beam forming X-ray filter 29 Control controller 30 Slip ring

dP X-ray detector plane P Image reconstruction area PP Projection plane
IC rotation center (ISO)
CB X-ray beam
BC beam center axis
D Multi-row X-ray detector width on the rotation axis

Claims (8)

A projection data collecting means for collecting projection data irradiated with radiation from around the subject and transmitted through a predetermined area of the subject;
Image reconstruction means for performing image reconstruction using the projection data and obtaining a plurality of tomographic images continuous in the body axis direction of the subject;
In a radiation tomography apparatus comprising:
The image reconstruction means sequentially reconstructs the plurality of tomographic images during the collection of the projection data,
Labeling means for sequentially labeling the plurality of tomographic images during the sequential image reconstruction.
Radiation tomography apparatus characterized by further comprising a. The radiation tomography apparatus according to claim 1, wherein the labeling unit performs a re-labeling process based on information on continuous regions during the sequential labeling process. The labeling means, not through the binary image, radiological according to any one of claims 1 or claim 2, characterized in that performing the labeling process from tomogram image reconstruction is gray image Shooting device. 4. The labeling unit according to claim 1, wherein the labeling unit performs a labeling process based on an image feature amount of each pixel of the tomographic image or an image feature amount in the vicinity of the pixel. 5. Radiation tomography equipment 5. The radiation tomography according to claim 1, wherein the labeling unit obtains an image feature amount for each continuous region subjected to the labeling process while performing the labeling process. Shooting device. The radiation tomography apparatus according to claim 5, wherein the relabeling process includes editing of an image feature amount for each continuous region. The image reconstruction unit recognizes a labeled region or a re-labeled region as a part in accordance with an image feature amount of the labeled region or the re-labeled region, and optimizes the image quality of the part. radiation tomography apparatus according to any one of claims 5 or claim 6, characterized in that it comprises means for. The image feature amount includes area, image density sum, perimeter length, ferret diameter, ferret diameter ratio, circularity, area ratio, first moment, second moment, ellipse approximated major axis / minor axis, volume, volume density sum, The surface area, the three-dimensional ferret diameter, the three-dimensional ferret diameter ratio, the sphericity, the volume ratio, the three-dimensional first moment, the three-dimensional second moment, and an ellipsoid approximate diameter are included. The radiation tomography apparatus according to any one of claims 5 to 7 .