JP7230645B2 - Mask generation device, three-dimensional reconstruction image generation device, mask generation method, and program - Google Patents

Description

The present disclosure relates to rendering technology for three-dimensional visualization using a plurality of two-dimensional tomographic images.

Conventionally, multiple images are taken continuously at predetermined slice intervals by medical imaging diagnostic equipment (modality) such as CT, MRI, and PET, and stored in a medical information system such as PACS (Picture Archiving and Communication Systems) in DICOM format. Based on the tomographic image, organs and the like are generated as a three-dimensional reconstructed image and three-dimensionally visualized.

In medical image diagnosis, there are cases where it is desired to obtain a three-dimensional reconstructed image suitable for observing a specific human tissue. For example, when observing predetermined human tissues such as internal organs and brain in the chest and head, the internal organs and brain are surrounded by bones, and the bone region rather hinders diagnosis. In a three-dimensional reconstructed image generated from a CT image, bones are generally depicted clearly, and internal organs and blood vessels may be hidden.

For example, a 3D mask is created by extracting the bone regions that are not to be observed using the region growing method (region growing method), and the bone regions that are not to be observed are removed by mask processing while referring to the 3D mask. A method for generating a volume rendering image is disclosed (see Patent Documents 2 and 3). The region expansion method is a method of specifying a pixel in a non-observation target region, using the pixel as a starting point (expansion starting point), and successively extracting neighboring pixels.

JP 2009-240569 A Japanese Patent No. 4087517 Japanese Patent No. 5257958

However, the creation of a three-dimensional mask using the region growing method requires the user to specify multiple extension start points, and the user also needs to specify the termination stage of the extension. There is a problem that variation occurs in In addition, not only the region growing method, but also the creation of a three-dimensional mask requires skill and experience of the user, and there is a problem that the creation takes time and effort.

Therefore, an object of the present disclosure is to provide a mask generation device that can easily generate mask data for visualizing a desired region.

The present disclosure includes multiple means for solving the above problems, but if one example is given,
an image acquisition means for acquiring a plurality of tomographic images;
subject area calculation means for calculating a subject area based on the signal value of each pixel for each tomographic image;
For each tomographic image, the calculated contour of the subject region is approximated by two predetermined geometric shapes, namely, a first geometric shape and a second geometric shape having an area smaller than that of the first geometric shape. subject shape calculation means for calculating a geometric shape and parameters of the second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value for each pixel is held. a mask data generating means for generating mask data for
and a mask generator.

Further, the present disclosure includes image acquisition means for acquiring a plurality of tomographic images,
subject area calculation means for calculating a subject area based on the signal value of each pixel for each tomographic image;
For each tomographic image, the contour of the calculated subject region is approximated by two predetermined first geometric shapes and a second geometric shape having a smaller area than the first geometric shape, and the first geometric shape and the subject shape calculation means for calculating a second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value of each pixel is held. a mask data generating means for generating mask data for
a three-dimensional reconstructed image generating means for generating a three-dimensional reconstructed image based on a three-dimensional voxel configured by arranging each pixel of the tomographic image in a voxel space and the mask data;
A three-dimensional reconstruction image generator is provided, comprising:

The present disclosure also provides a method for the computer to
an image acquisition step of acquiring a plurality of tomographic images;
a subject region calculation step of calculating a subject region based on the signal value of each pixel for each tomographic image;
For each tomographic image, the contour of the calculated subject region is approximated by two predetermined first geometric shapes and a second geometric shape having a smaller area than the first geometric shape, and the first geometric shape and the a subject shape calculation step of calculating a second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value of each pixel is held. a mask data generation step for generating mask data for
provides a mask generation method that performs

The present disclosure also provides a computer with
image acquisition means for acquiring a plurality of tomographic images;
subject area calculation means for calculating a subject area based on the signal value of each pixel for each tomographic image;
For each tomographic image, the contour of the calculated subject region is approximated by two predetermined first geometric shapes and a second geometric shape having a smaller area than the first geometric shape, and the first geometric shape and the object shape calculation means for calculating a second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value of each pixel is held. mask data generation means for generating mask data for
Provide a program that functions as

According to the present disclosure, it is possible to easily create mask data for visualizing a desired region.

FIG. 4 is a diagram showing an outline of a three-dimensional reconstructed image generation process; FIG. 1 is a hardware configuration diagram of a mask generation device 100 according to an embodiment of the present disclosure; FIG. 1 is a functional block diagram showing the configuration of a mask generation device according to this embodiment; FIG. 4 is a flow chart showing the processing operation of the mask generation device according to the embodiment; FIG. 3 is a diagram for explaining the relationship between subject candidate areas and subject areas; 4 is a flowchart showing details of subject area calculation processing in step S20. FIG. 10 is a diagram showing an overview of subject area calculation processing; It is a figure explaining calculation of a 2nd center of gravity. It is a figure explaining calculation of a 3rd center of gravity. 7 is a diagram showing a calculated rectangular area 72; FIG. is. (a) A rectangular area 72 when the signal value inside the area is relatively large, and (b) a rectangular area 72 when the signal value inside the area is relatively small. FIG. 10 is a diagram showing how the rectangular area 72 is corrected based on the area ratio with respect to the subject candidate area 60. FIG. FIG. 10 is a diagram showing a case where correction based on area ratio is applied to the upper chest; FIG. 10 is a diagram showing how the rectangular area 72 is corrected (reduced correction) based on the aspect ratio of the rectangular area 72; FIG. 10 is a flowchart showing the details of subject area correction processing in step S<b>30 ; FIG. FIG. 11 shows a deformed great ellipse which is the geometry KA; FIG. 10 is a flow chart showing details of correction of mask data in step S60. FIG. FIG. 10 is a diagram showing a state in which a deformed small ellipse, which is the geometric shape KB, is superimposed on a deformed large ellipse, which is the geometric shape KA; FIG. 10 is a diagram showing parallel processing when the mouse generation device 100 is realized by a computer; FIG. 4 is a diagram showing a software flow of parallel processing when the mouse generation device 100 is implemented by a computer. FIG. 4 is a diagram showing processing times of processing threads in each CPU core; FIG. 4 is a diagram showing a three-dimensional reconstructed image generated using mask data generated by a mask generation device;

Preferred embodiments of the present disclosure will be described in detail below with reference to the drawings.
The present disclosure relates to an apparatus and the like that generates mask data for non-rendering unnecessary portions such as vertebrae when generating rendering images obtained by three-dimensionally visualizing a plurality of tomographic images. FIG. 1 is a diagram showing an overview of the three-dimensional reconstructed image generation processing. FIG. 1(a) shows a plurality of tomographic images, FIG. 1(b) shows a single tomographic image, and FIG. 1(c) shows a three-dimensional reconstructed image. The tomographic images shown in FIGS. 1A and 1B are in DICOM format. FIG. 1(a) is a group of 370 tomographic images, and FIG. 1(b) is an enlarged view of one of them. In the mask generation apparatus of the embodiment described later, predetermined processing is performed on a plurality of tomographic images as shown in FIG. 1A to generate mask data, and predetermined Various three-dimensional reconstructed images are generated as shown in FIG. FIG. 1(c) shows an MPR (Multi-Planar Reconstruction) image, a 3D-MIP (Maximum Intensity Projection) image, and a volume rendering image. In the present embodiment, a state in which two-dimensional tomographic images are stacked in another dimension is treated as a three-dimensional voxel image. Each voxel in the three-dimensional voxel image corresponds to each pixel in a plurality of two-dimensional tomographic images on a one-to-one basis.

<1. Configuration of Mask Generation Device>
FIG. 2 is a hardware configuration diagram of the mask generation device 100 according to an embodiment of the present disclosure. The mask generating apparatus 100 according to the present embodiment can be realized by a general-purpose computer, and as shown in FIG. , a large-capacity storage device 3 such as a hard disk, SSD (Solid State Drive), flash memory, etc. for storing programs and data executed by the CPU 1, and an instruction input I / F (interface) 4 such as a keyboard, mouse, etc. , a data input/output I/F (interface) 5 for data communication with an external device such as a data storage medium, a display unit 6 which is a display device such as a liquid crystal display, and an arithmetic processing unit specializing in graphics. A GPU (Graphics Processing Unit) 7 and a frame memory 8 for holding an image to be displayed on the display unit 6 are provided, and are connected to each other via a bus. Since the calculation result by the GPU 7 is written to the frame memory 8, the GPU 7 and the frame memory 8 are often mounted on a video card having an interface to the display unit 6 and attached to a general-purpose computer via a bus.

In this embodiment, the CPU 1 is a multi-core CPU, has a plurality of CPU cores (physical cores), and is capable of parallel processing. Although only one RAM2 is shown in the example of FIG. 2, each CPU core of the CPU1 is configured to access one RAM2. Note that the number of CPUs 1 may be plural. A multi-core CPU may also be a CPU in which each CPU core logically accepts multiple threads of execution. That is, CPU 1 has a processor that accepts each of a plurality of threads of execution.

FIG. 3 is a functional block diagram showing the configuration of the mask generation device according to this embodiment. In FIG. 3, 10 is an image acquiring means, 20 is a subject area calculating means, 30 is a subject area correcting means, 40 is a subject shape calculating means, and 50 is a mask data generating means.

The image acquisition means 10 is means for reading and acquiring a plurality of tomographic images from PACS collected and stored by medical diagnostic imaging equipment such as CT, MRI, and PET via a storage medium and an input unit. A tomographic image is obtained by photographing an object at predetermined intervals, and is a two-dimensional tomographic image in which a signal value is assigned to each pixel. The object area calculation means 20 is means for calculating an object area based on the signal value of each pixel for each tomographic image. The object area correcting means 30 corrects the object area for each tomographic image using a correction method according to the ratio between the vertical size and the horizontal size of the object area calculated by the object area calculating means 20 . For each tomographic image, the object shape calculation means 40 calculates the outline of the calculated object region as two predetermined geometric shapes, namely a first geometric shape KA and a second geometric shape having a smaller area than the geometric shape KA. It is a means for approximating with the shape KB and calculating the parameters of the geometric shape KA and the geometric shape KB. The mask data generating means 50 calculates, for each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the geometric shape KA and outside the geometric shape KB, and generates a mask value for each pixel. It is a means of holding.

The image acquisition means 10 is realized mainly in the data input/output I/F 5 with the assistance of the CPU 1 . The subject area calculation means 20 , the subject area correction means 30 , the subject shape calculation means 40 and the mask data generation means 50 are implemented by the CPU 1 executing programs stored in the storage device 3 .

Each component shown in FIG. 3 is actually implemented by installing a dedicated program in hardware such as a computer and its peripherals, as shown in FIG. That is, the computer executes the contents of each means according to a dedicated program. In this specification, a computer means a device that has an arithmetic processing unit such as a CPU or GPU and is capable of data processing. It also includes computers embedded in mobile terminals and various devices.

<2. Processing Operation of Mask Generation Device (Mask Generation Method)>
Next, the processing operation of the mask generation apparatus shown in FIGS. 2 and 3 will be described together with the mask generation method of the present disclosure. FIG. 4 is a flowchart showing the processing operation of the mask generation device according to the present embodiment and an outline of mask generation processing. First, the image acquiring means 10 reads and acquires a plurality of tomographic images (step S10). The plurality of tomographic images are in the form shown in FIG. 1(a). In this embodiment, a DICOM format tomographic image group Do (x, y, z) (-32768≤Do (x, y, z)≤32767, 0≤x≤Sx-1, 0≤y≤Sy-1, 0≦z≦Sz−1; resolution: Rxy, Rz). Since one pixel of the tomographic image group Do(x, y, z) is recorded with 16 bits, signal values Do(x, y, z) of -32768 to 32767 are recorded. The tomographic image group Do (x, y, z) is obtained by stacking Sz tomographic images each having an x-coordinate size (number of pixels) of Sx and a y-coordinate size (number of pixels) of Sy. It is expressed as three-dimensional data. The horizontal direction of one tomographic image in FIG. 1(b) corresponds to the X axis, and the vertical direction corresponds to the Y axis. The lamination direction when superimposed at a predetermined slice interval (difference in Image Position tag information of adjacent tomographic images) corresponds to the z-axis. In the present disclosure, the X-axis corresponds to the left-right axis of the human body, the Y-axis corresponds to the dorsal-abdominal axis (front-back axis) of the human body, and the Z-axis corresponds to the craniocaudal axis (vertical axis) of the human body. The XY axes are set, for example, during imaging with a CT apparatus. Also, in the present disclosure, the X-axis direction may be called the horizontal direction, and the Y-axis direction may be called the vertical direction.

<2.1. Calculation Processing of Subject Area>
After reading a plurality of tomographic images in the DICOM format, the object area calculating means 20 calculates an object area for each tomographic image (step S20). FIG. 5 is a diagram for explaining the relationship between subject candidate areas and subject areas. In FIG. 5, 41 is a rectangular area. A rectangular area 71 is a rectangular area surrounding the outermost side of the subject candidate area 60 . The subject candidate region 60 is a pixel region in the tomographic image Do(z) whose signal value is greater than a predetermined threshold value (for example, a non-air signal value or more).

The rectangular area 72 (second rectangular area) is a rectangular area defined inside the rectangular area 71 as shown in FIG. Since the rectangular area 71 includes all pixels whose signal values are larger than the predetermined threshold, as shown in FIG. Therefore, in the present disclosure, a rectangular region 72 defined inside the rectangular region 71 is calculated in consideration of the signal value distribution of the subject candidate region 60 so that unnecessary regions such as a bed and a headrest are not included. Extract the subject area.

FIG. 6 is a flowchart showing the details of the subject area calculation process in step S20. FIG. 7 is a diagram illustrating an outline of subject area calculation processing. First, the subject area calculation means 20 calculates the barycentric coordinates weighted by the signal values of the pixels of the subject candidate area 60 as the central coordinates of the rectangular area 72 (step S21, FIG. 7A).

In step S21, the subject area calculation means 20 first sets the center coordinates of the rectangular area 72 to be calculated as (C(x), Cy(z)), the sum of the weights as Ws, and Cx(z)=Cy( z)=Ws=0. Then, Do(x, y, z)> If Vmin (for example, Vmin=−700) is satisfied, it is assumed to be a pixel in the subject candidate region 60, and Cx(z), Cy(z), and Ws are updated as follows.

[Formula 1]
As w=Do(x,y,z)−Vmin,
Cx(z)←Cx(z)+x・w
Cy(z)←Cy(z)+y・w
Ws←Ws+w

In [Equation 1], since the first equation is calculated only when Do(x, y, z)>Vmin, w>0 is always satisfied.

After executing the above processing for Sx×Sy pixels in the range of 0≦x≦Sx−1 and 0≦y≦Sy−1, the subject area calculating means 20 performs the following [ 2] is executed.

[Formula 2]
Cx(z)←Cx(z)/Ws
Cy(z)←Cy(z)/Ws

As described above, the barycentric coordinates weighted by the signal values of the pixels in the subject candidate area 60 are obtained, and the barycentric coordinates are calculated as the central coordinates (Cx(z), Cy(z)) of the rectangular area 72 .

Subsequently, the subject region calculation means 20 calculates the signal values within the range of a rectangular region 71 (first rectangular region) in four directions, up, down, left, and right, starting from the calculated center coordinates (Cx(z), Cy(z)). A weighted center (hereinafter referred to as a “second center of gravity”) is calculated using the signal values of pixels (pixels in the subject candidate region 60) for which is greater than a predetermined threshold (step S22, FIG. 7(b)).

FIG. 8 is a diagram illustrating calculation of the second center of gravity. In step S22, as shown in FIGS. 8(a) and 8(b), the subject area calculation means 20 calculates the subject from the center coordinates (Cx(z), Cy(z)) (the center of gravity of the subject candidate area 60). The candidate region 60 is divided into two in the X-axis direction, and X coordinates Gx1 and Gx2 (Gx1<Gx2) of the center of gravity (second center of gravity) weighted using the signal values of the pixels in each divided subject candidate region 60 are calculated. .

Similarly, as shown in FIGS. 8(c) and 8(d), the subject area calculator 20 calculates subject candidate areas starting from the central coordinates (Cx(z), Cy(z)) (the center of gravity of the subject candidate area 60). 60 is divided into two in the Y-axis direction, and Y coordinates Gy1 and Gy2 (Gy1<Gy2) of the center of gravity (second center of gravity) weighted using the signal values of the pixels in each divided subject candidate region 60 are calculated.

In the step S22 of obtaining the second center of gravity, the subject region calculating means 20 first sets the sum of the weights for the second center of gravity coordinates to Wx1, Wx2, Wy1, and Wy2, where Gx1=Gx2=Gy1=Gy2=Wx1=Wx2= Initialize to Wy1=Wy2=0.

Then, the subject area calculation means 20 calculates Do ( x, y, z)>Vmin (for example, Vmin=−700), it is assumed that the pixel is in the subject candidate region 60, and the processing according to the following [Equation 3] is executed to obtain Gx1, Gx2, Gy1 , Gy2, Wx1, Wx2, Wy1, Wy2.

[Formula 3]
As w=Do(x,y,z)−Vmin,
If x<Cx(z), then Gx1←Gx1+x·w, Wx1←Wx1+w
If x≧Cx(z), then Gx2←Gx2+x·w, Wx2←Wx2+w
If y<Cy(z), then Gy1←Gy1+y·w, Wy1←Wy1+w
If y≧Cy(z), then Gy2←Gy2+y·w, Wy2←Wy2+w

In [Equation 3], since the first equation is calculated only when Do(x, y, z)>Vmin, w>0 is always satisfied.

After performing the above processing on Sx×Sy pixels in the range of 0≦x≦Sx−1 and 0≦y≦Sy−1, the subject area calculation means 20 performs the following [Equation 4]. Execute the process.

[Formula 4]
Gx1←Gx1/Wx1
Gx2←Gx2/Wx2
Gy1←Gy1/Wy1
Gy2←Gy2/Wy2

As a result of executing the processing according to [Equation 4], the second barycentric coordinates Gx1, Gx2 (Gx1<Gx2), Gy1, Gy2 (Gy1<Gy2) in each of the divided subject candidate regions 60 are obtained.

Subsequently, the object region calculating means 20 calculates the signal value is smaller than a predetermined threshold in the range of a rectangular region 71 (first rectangular region) in four directions, up, down, left, and right, starting from the second barycentric coordinates calculated in step S22. A weighted center (hereinafter referred to as a “third center of gravity”) is calculated using the signal values of large pixels (pixels in the subject candidate region 60) (step S23, FIG. 7(c)).

FIG. 9 is a diagram explaining calculation of the third center of gravity. In step S23, as shown in FIGS. 9A and 9B, the subject area calculation means 20 divides the subject candidate area 60 into three parts in the X-axis direction starting from the second centers of gravity Gx1 and Gx2. X-coordinates Gtx1 and Gtx2 (Gtx1<Gtx2) of the center of gravity (third center of gravity) weighted using the signal values of the pixels in the two subject candidate regions 60 are calculated.

Similarly, as shown in FIGS. 9(c) and 9(d), the subject area calculation means 20 divides the subject candidate area 60 into three parts in the Y-axis direction starting from the second centers of gravity Gy1 and Gy2, and divides the object candidate area 60 into three parts at the upper and lower ends. Y coordinates Gty1 and Gty2 (Gty1<Gty2) of the center of gravity (third center of gravity) weighted using the signal values of the pixels in the subject candidate region 60 are calculated.

In the step S23 for obtaining the third center of gravity, the subject area calculating means 20 first sets the sum of the weights for the respective third center of gravity coordinates to Wx1, Wx2, Wy1, and Wy2, and Gtx1=Gtx2=Gty1=Gty2=Wx1=Wx2= Initialize to Wy1=Wy2=0.

Then, the subject area calculation means 20 calculates Do ( x, y, z)>Vmin (for example, Vmin=−700), it is assumed that the pixel is in the subject candidate region 60, and the processing according to the following [Equation 5] is executed to obtain Gtx1, Gtx2, Gty1 , Gty2, Wx1, Wx2, Wy1, Wy2.

[Formula 5]
As w=Do(x,y,z)−Vmin,
If x<Gx1, Gtx1←Gtx1+x·w, Wx1←Wx1+w
If x>Gx2, then Gtx2←Gtx2+x·w, Wx2←Wx2+w
If y<Gy1, then Gty1←Gty1+y·w, Wy1←Wy1+w
If y>Gy2, then Gty2←Gty2+y·w, Wy2←Wy2+w

In [Equation 5], since the first equation is calculated only when Do(x, y, z)>Vmin, w>0 is always satisfied.

After performing the above processing on Sx×Sy pixels in the range of 0≦x≦Sx−1 and 0≦y≦Sy−1, the subject area calculation means 20 performs the following [Equation 6]. Execute the process.

[Formula 6]
Gtx1←Gtx1/Wx1
Gtx2←Gtx2/Wx2
Gty1←Gty1/Wy1
Gty2←Gty2/Wy2

As a result of executing the process according to [Formula 6], the third barycentric coordinates Gtx1, Gtx2 (Gtx1<Gtx2), Gty1, Gty2 (Gty1<Gty2) in each of the divided subject candidate regions 60 are obtained.

Then, the subject area calculation means 20 performs the processing according to the following [Equation 7] based on the third barycentric coordinates calculated in step S23 to determine the horizontal size W(z) of the rectangular area 72. , the vertical size H(z) is calculated (step S34, FIG. 7(d)).

[Formula 7]
Horizontal size: W(z)=Cx(z)-Gtx1, W2(z)=Gtx2-Cx(z)
Vertical size: H(z)=Cy(z)-Gty1, H2(z)=Gty2-Cy(z)

FIG. 10 is a diagram showing the calculated rectangular area 72. As shown in FIG. As shown in the figure, the first horizontal size W(z) of the rectangular area 72 corresponds to the distance from the leftward (negative direction) third center of gravity Gtx1 to the center coordinates of the rectangular area 72, and the second is equivalent to the distance from the third center of gravity Gtx2 in the right direction (positive direction) to the center coordinates of the rectangular area 72 . The first vertical size H(z) corresponds to the distance from the third center of gravity Gty1 in the upward direction to the center coordinates of the rectangular area 72, and the second vertical size H2(z) corresponds to the distance in the downward direction. It corresponds to the distance from the third center of gravity Gty2 to the center coordinates of the rectangular area 72 .

By the above processing, the center coordinates (Cx(z), Cy(z)), the first lateral size W(z), the second lateral size W2(z), the first A rectangular area 72 defined by the vertical size H(z) and the second vertical size H2(z) is calculated.

In step S24, the subject area calculation means 20 calculates the first horizontal size W(z) and the second horizontal size W2 ( z), the first vertical size H(z), and the second vertical size H2(z).

Further, the subject area calculation unit 20 may calculate the fourth center of gravity in four directions, up, down, left and right, starting from the third center of gravity calculated in step S23. In this case, the subject area calculation means 20 calculates the first horizontal size W(z), the second horizontal size W2(z), and the first vertical size W(z) of the rectangular area 72 based on the fourth center of gravity. A size H(z) and a second vertical size H2(z) are calculated.

<2.2. Subject Area Correction Processing>
By the way, in the case of chest CT, if the vertical and horizontal sizes of the rectangular area 72 are calculated by the second and third centroids in the vertical and horizontal directions from the center point, the rectangular area may vary depending on the signal value distribution of the organs in the bone area. 72 may not be of a size suitable for removing the bone region and visualizing organs and the like within the bone region.

FIG. 11 shows a rectangular area 72 when the signal value inside the area is relatively large, and a rectangular area 72 when the signal value inside the area is relatively small. For example, when the signal value inside the region is relatively large as shown in FIG. , the rectangular area 72 is calculated to be smaller. That is, the rectangular area 72 may be defined inside the bone area. If the mask data is created based on such a rectangular area 72, the visceral area, which is the body drawing target, may be removed.

In addition, when the signal value inside the region is relatively small as shown in FIG. be. That is, the rectangular area 72 may be defined outside the bone area. If mask data is created based on such a rectangular area 72, it may not be possible to accurately remove the bone area, which is a non-rendering target.

In order to solve the above problem, as shown in FIG. 12, the calculated horizontal and vertical sizes of the rectangular area 72 are multiplied by the area ratio between the subject candidate area 60 and the rectangular area 72 as a magnification. By doing so, the size of the rectangular area 72 is corrected. FIG. 12 is a diagram showing how the rectangular area 72 is corrected based on the area ratio with respect to the subject candidate area 60. As shown in FIG.

As a result, when the signal value inside the region is relatively large as shown in FIG. When the signal value of is relatively small (for example, in the case of the thoracic and lung region), the rectangular region 72 is reduced and corrected, and the rectangular region 72 is a size suitable for removing the bone region and visualizing the organs and the like in the bone region. is corrected to

However, when the above correction is performed, there arises another problem that the originally unnecessary upper chest rectangular region 72 (object region) is enlarged and corrected. FIG. 13 is a diagram showing a case where correction based on the area ratio as described above is applied to the upper chest. As shown in FIG. 13, the upper thoracic region is not surrounded by bone regions, and high-signal bone regions are distributed independently. Therefore, the area ratio of the subject candidate region 60 to the rectangular region 72 becomes as large as the abdomen, and the enlargement correction is performed in the same manner as the abdomen. However, since the upper chest is mostly bone regions that are not to be drawn, it is not desirable to enlarge the rectangular region 72 (object region) in the upper chest region.

By the way, since the bone region of the upper chest extends in the lateral direction, the rectangular region 72 is characterized by being flattened. Therefore, in the present disclosure, as shown in FIG. 14, the aspect ratio of the rectangular area 72 (vertical size {H(z)+H2(z)}/horizontal size {W(z)+W2(z)}) , and if this value is smaller than a predetermined threshold value (for example, 0.62), it is identified as the upper chest, and the rectangular region 72 and the subject candidate region Instead of multiplying the area ratio with 60, the aspect ratio is multiplied for correction (reduction correction). On the other hand, if the aspect ratio of the rectangular area 72 is equal to or greater than a predetermined threshold value (for example, 0.62), it is determined to be the chest or abdomen. and the subject candidate region 60 are multiplied to correct. FIG. 14 is a diagram showing how the rectangular area 72 is corrected (reduced correction) based on the aspect ratio of the rectangular area 72 .

In this way, the part of the subject area (upper chest/chest or abdomen) is determined based on the magnitude relationship between the value of the aspect ratio (vertical size/horizontal size) of the rectangular area 72 and the predetermined threshold, If a method of correcting the rectangular area 72 by switching either the aspect ratio or the area ratio as a correction magnification according to the part is adopted, a step may occur in the boundary area where the part changes. Therefore, in the present disclosure, a weight based on the difference between the aspect ratio and a predetermined threshold is used, and the ratio obtained by weighting and synthesizing the aspect ratio and the area ratio is used as a correction magnification. The correction magnification to be applied gradually changes from the aspect ratio to the area ratio.

FIG. 15 is a flow chart showing details of a method of using, as a correction magnification, a weighted combination of the aspect ratio and the area ratio in the subject area correction process in step S30 of FIG.
The subject area correction means 30 first calculates the aspect ratio Rvh of the rectangular area 72 as follows (step S31).

[Formula 8]
Aspect ratio Rvh={H(z)+H2(z)}/{W(z)+W2(z)}

A threshold value for the aspect ratio is set to Srvh (for example, 0.62), and the weight w is calculated by executing the processing according to [Equation 9] below (step S32).

[Formula 9]
w = (Rvh-Srvh)/Srvh

As a result of calculation by [Equation 9], if w<0.0, then w=0.0, and if w>1.0, then w=1.0. Therefore, the weight w is obtained as a real number between 0.0 and 1.0.

On the other hand, the subject area correction means 30 performs Do The number of pixels satisfying (x, y, z)>Vmin (for example, Vmin=−700) is counted, and the count value Cobj is calculated as the area of the subject candidate region 60 (step S33).

Subsequently, the subject region correcting means 30 executes processing according to [Formula 10] below to calculate an area ratio Robj between the count value Cobj and the rectangular region 72 (step S34).

[Formula 10]
Robj=Cobj/[[(H(z)+H2(z)]·[W(z)+W2(z)]]

Then, the subject region correcting means 30 executes processing according to the following [Equation 11] to calculate horizontal and vertical magnifications Rox and Roy (step S35).

[Formula 11]
Rox=Roy=(1.0−w)·Rvh+w·2.0·Robj

The upper limit of the scaling factor is set to Mox and Moy, and when Rox>Mox, Rox=Mox, and when Roy>Moy, Roy=Moy, and the subject area correction means 30 performs processing according to [Equation 12] below. is executed to correct the rectangular area 72 (step S36, correction in FIG. 12).

[Formula 12]
First lateral size: Wc(z)=W(z)Rox
Second lateral size: W2c(z)=W2(z)Rox
First vertical size: Hc(z)=H(z)Roy
Second vertical size: H2c(z)=H2(z)Roy

<2.3. Subject Shape Calculation Processing>
Subsequently, the subject shape calculation means 40 calculates the outline of the subject area (rectangular area 72) calculated in step S20 or the subject area (rectangular area 72) corrected in step S30 for each tomographic image Do(z). A predetermined geometric shape (object shape) is approximated, and a parameter (geometric parameter P(z)) defining the geometric shape is calculated (step S40).

In the present disclosure, two geometric shapes KA and a geometric shape KB having a smaller area than the geometric shape KA are approximated for each tomographic image. The object shape can be calculated by a method such as obtaining two geometric shapes in succession. After obtaining the second geometric shape, it is also possible to correct the created mask data. In the present disclosure, a method of obtaining a first geometric shape KA with a large area and once creating mask data, and then obtaining a second geometric shape with a small area and correcting the mask data will be described.

In the present disclosure, a special ellipse (hereinafter also referred to as a “deformed ellipse”) that can set different diameters in each of the following regions (four quadrants) is approximated. The parameters (ellipse parameters P1(z) to P4(z)) that define the ellipse of the region (4 quadrants) are calculated by executing the processing according to [Equation 13] below. As noted above, in this disclosure, two geometries are sought, the larger area geometry being also referred to as the "deformed large ellipse" and the smaller area geometry being referred to as the "deformed small ellipse".
(Region 1) x<Cx(z) and y<Cy(z)
(Region 2) x≧Cx(z) and y<Cy(z)
(Region 3) x<Cx(z) and y≧Cy(z)
(Region 4) x≧Cx(z) and y≧Cy(z)

[Formula 13]
(Ellipse parameter P1(z) in region 1)
Ellipse central coordinates Cx(z), Cy(z)
2 division size of the ellipse in the horizontal direction rx = Wc (z) · a (horizontal diameter)
2 division size of ellipse in vertical direction ry = Hc (z) · b (longitudinal diameter)
(Ellipse parameter P2(z) in region 2)
Ellipse central coordinates Cx(z), Cy(z)
2 division size of the ellipse in the horizontal direction rx2 = W2c (z) · a2 (horizontal diameter)
2 division size of ellipse in vertical direction ry = Hc (z) · b (longitudinal diameter)
(Ellipse parameter P3(z) in region 3)
Ellipse central coordinates Cx(z), Cy(z)
2 division size of the ellipse in the horizontal direction rx = Wc (z) · a (horizontal diameter)
2 division size of ellipse in vertical direction ry2=H2c(z)·b2 (longitudinal diameter)
(Ellipse parameter P4(z) in region 4)
Ellipse central coordinates Cx(z), Cy(z)
2 division size of the ellipse in the horizontal direction rx2 = W2c (z) · a2 (horizontal diameter)
2 division size of ellipse in vertical direction ry2=H2c(z)·b2 (longitudinal diameter)

In [Equation 13], a, a2, b, and b2 are correction coefficients (real value correction 1.0 if not), left (negative) (aWc(z)), right (positive) (a2, W2c(z)), front (b, Hc(z)), back It is set for each of the four directions (b2, H2c(z)).

For example, the correction coefficient is set as follows. a=a2=0.95, b=1.05, b2=0.8

FIG. 16 is a diagram showing a deformed great ellipse, which is the geometric shape KA calculated by [Equation 14]. As shown in the figure, in the region inside the deformed great ellipse including the center (Cx(z), Cy(z)) of the deformed great ellipse, the mask value is set to 1 in step S50 to be described later, and the deformed great ellipse In the outer regions, the mask value is set to 0.

The object shape calculation means 40 calculates the distance r(x, y, z) in each area by executing the processing according to [Equation 14] below.

[Formula 14]
r(x, y, z)=[(x−Cx(z))/rx] ² +[(y−Cy(z))/ry] ² Region 1
r(x, y, z)=[(x−Cx(z))/rx2] ² +[(y−Cy(z))/ry] ² Region 2
r(x, y, z)=[(x−Cx(z))/rx] ² +[(y−Cy(z))/ry2] ² Region 3
r(x, y, z)=[(x−Cx(z))/rx2] ² +[(y−Cy(z))/ry2] ² Region 4

<2.4. Generation of mask data>
Next, the process of generating the mask data Mask executed in step S50 of FIG. 4 will be described. This process is performed for all pixels (x, y, z).

The mask data generating means 50 obtains the distance r (x, y, z) of each pixel (x, y, z), and determines that the distance r (x, y, z)<threshold Th2 (for example, Th2=1.0 ), the mask value corresponding to the pixel (x, y, z) is set to 1, and the distance r (x, y, z) <threshold value Th2 is not satisfied. For (x, y, z), set the mask value corresponding to the pixel (x, y, z) to zero. As described above, the distance r (x, y, z) of each pixel (x, y, z) is binarized with a predetermined threshold, and the mask value Mask (x, y , z) is generated.

<2.5. Correction of mask data>
After the mask data Mask is generated, the subject area calculation means 20, the subject area correction means 30, the subject shape calculation means 40, and the mask data generation means 50 cooperate to correct the mask data Mask. FIG. 17 is a flow chart showing details of correction of mask data in step S60. The processing of FIG. 17 is performed for all pixels (x, y, z).

First, the mask generation device determines whether mask data Mask(x, y, z)=1 is satisfied (step S61). If the mask data Mask (x, y, z) = 1 is not satisfied, that is, if the mask data Mask (x, y, z) = 0, the pixel (x, y, z) is not corrected. , the mask data correction process ends. That is, in the mask data correction process in step S60, correction is performed only for the pixels with mask data Mask(x, y, z)=1.

If the mask data Mask(x, y, z)=1 is satisfied, the mask data Mask is corrected for the pixel (x, y, z). The purpose of modifying the mask data Mask is to mask the vertebral bone region. Therefore, it is necessary to identify the position of the deformed small ellipse for estimating the vertebral region. First, the X-coordinate of the center and the lateral diameter in the X-axis direction, which are the parameters in the X-axis direction of the deformed small ellipse, are specified (step S62).

Specifically, first, the X coordinate of the center of the deformed small ellipse (the center coordinate is lateral component) is slightly shifted to the left (negative direction) with respect to the X-coordinate Cx(z) of the center of gravity. Even if rx=rx2, offx>offx2, resulting in a leftward shift of about 5% of rx. Then, the lateral diameter of the deformed small ellipse in the X-axis direction is set to 30% of the deformed large ellipse for each of the left direction (negative direction) and the right direction (positive direction), rrx=rx·0.3, rrx2=rx2·0. .3, the processing according to the following [Equation 15] is executed, and the X-axis direction of the pixel (x, y, z) satisfying Mask (x, y, z)=1 and the deformed small ellipse is Calculate the distance ratio xx.

[Formula 15]
If x<Cx(z): xx=(x−Cx(z)+offx)/rrx
If x≧Cx(z): xx=(x−Cx(z)+offx2)/rrx2

Once the parameters of the deformed small ellipse in the X-axis direction are specified, the center Y-coordinate (vertical component of the center coordinate) and the vertical diameter in the Y-axis direction, which are the parameters of the deformed small ellipse in the Y-axis direction, are specified. Since there is a remarkable difference for each part, it is divided into three parts and specified. Then, the distance ratio xx is calculated for each part as in [Formula 15], and it is determined whether or not any part is outside the deformed small ellipse. First, the Y coordinate of the center of the deformed small ellipse and the longitudinal diameter in the Y direction are specified on the upper chest side, and the outside of the deformed small ellipse is determined together with the previously specified X coordinate of the center of the deformed small ellipse and the lateral diameter in the X direction. (step S63). Specifically, the Y coordinate of the center of the deformed small ellipse is shifted from the Y coordinate of the center of gravity Cy(z) of the deformed large ellipse by the vertical diameter ry2 in the back direction (positive direction), and is offset by offy in the back direction ( positive direction). A point shifted from the barycentric Y-coordinate Cy(z) of the deformed great ellipse in the backward direction (positive direction) by the vertical diameter ry2 indicates the lower end of the geometric shape KA. Then, the process according to [Formula 16] below is executed to calculate the vertical diameter rry of the deformed small ellipse. In [Equation 16], Offy is a front-back position correction ratio adjusted for each CT image, and is set to 1.0 to 1.5.

[Formula 16]
offy=1.5*ry2*offy
rry=0.95·b2·{H2(z)} ² /H2c(z)

After the Y-direction position and vertical diameter of the deformed small ellipse on the upper chest side are calculated by executing the processing according to [Equation 16], the calculated offy and rry are used to perform the following [Equation 17]. Then, the distance ratio yy in the Y-axis direction between the pixel (x, y, z) that satisfies Mask (x, y, z)=1 and the deformed small ellipse is calculated.

[Formula 17]
yy=(y-Cy(z)-ry2-off)/rry

(y-Cy(z)-ry2-offy) in [Equation 17] is at a position where the vertical component of the center coordinates of the geometric shape KB is shifted downward from the lower end of the geometric shape KA by the vertical diameter of the geometric shape KA. It is shown that. Further, the process according to [Formula 18] below is executed to determine whether or not the pixel (x, y, z) that satisfies Mask (x, y, z)=1 is outside the deformed small ellipse. do.

[Formula 18]
xx ² +yy ² >1.0

If [Formula 18] is satisfied, it is outside the deformed small ellipse on the upper chest side, and if [Formula 18] is not satisfied, it is inside the deformed small ellipse on the upper chest side.

As a result of the processing in step S63, if it is determined that the area is outside the deformed small ellipse on the upper chest side, the Y coordinate of the center of the deformed small ellipse and the vertical diameter in the Y-axis direction are specified at the boundary, It is determined whether or not it is outside the deformed small ellipse along with the X coordinate of the center of the deformed small ellipse and the lateral diameter in the X-axis direction (step S64). Specifically, the Y coordinate of the center of the deformed small ellipse is offset from the Y coordinate of the center of gravity Cy(z) of the deformed large ellipse in the backward direction (positive direction) by the vertical diameter ry2, and the point is offset by offy in the backward direction ( positive direction). Then, the process according to [Formula 19] below is executed to calculate the vertical diameter rry of the deformed small ellipse. In [Equation 19], Offy is a front-back position correction ratio adjusted for each CT image, and is set to 1.0 to 1.5.

[Formula 19]
offy=ry2.offy
rry=1.52.ry2

After the Y-direction position and vertical diameter of the deformed small ellipse at the boundary are calculated by executing the processing according to [Formula 19], the calculated offy and rry are used to perform the processing according to [Formula 17] above. to calculate the distance ratio yy in the Y-axis direction between the pixel (x, y, z) that satisfies Mask (x, y, z)=1 and the deformed small ellipse. Furthermore, the process according to the above [Formula 18] is executed to determine whether or not the pixel (x, y, z) that satisfies Mask (x, y, z)=1 is outside the deformed small ellipse. .

As a result of the processing in step S64, if it is determined that the area is outside the deformed small ellipse at the boundary, the Y coordinate of the center of the deformed small ellipse and the vertical diameter in the Y-axis direction are specified on the ventral side. Along with the X-coordinate of the center of the deformed small ellipse and the lateral diameter in the X-axis direction, it is determined whether or not they are outside the deformed small ellipse (step S65). Specifically, the Y coordinate of the center of the deformed small ellipse is offset from the Y coordinate of the center of gravity Cy(z) of the deformed large ellipse in the backward direction (positive direction) by the vertical diameter ry2, and the point is offset by offy in the backward direction ( positive direction). Then, the process according to [Formula 20] below is executed to calculate the vertical diameter rry of the deformed small ellipse. In [Formula 20], Offy is a front-back position correction ratio adjusted for each CT image, and is set to 1.0 to 1.5.

[Formula 20]
offy=0.5*ry2*offy
rry=0.95·b2·{H2c(z)} ² /H2(z)

After the Y-direction position and vertical diameter of the deformed small ellipse on the abdominal side are calculated by executing the processing according to [Formula 20], the calculated offy and rry are used to perform the processing according to [Formula 17] above. to calculate the distance ratio yy in the Y-axis direction between the pixel (x, y, z) that satisfies Mask (x, y, z)=1 and the deformed small ellipse. Furthermore, the process according to the above [Formula 18] is performed to determine whether the pixel (x, y, z) satisfying Mask (x, y, z)=1 is outside the deformed small ellipse on the ventral side. judge.

FIG. 18 shows a deformed small ellipse that is the geometric shape KB that satisfies the condition that the sum of the squares in the X-axis direction and the Y-axis direction calculated by [Formula 15] and [Formula 17] becomes 1, and that is the deformed small ellipse that is the geometric shape KA. It is a figure which shows the state superimposed on a large ellipse. As shown in FIG. 18, in the area inside the deformed small ellipse, the mask value is set to 0 in step S66, which will be described later, even if it is inside the deformed large ellipse.

As a result of the processing in step S65, if it is determined that the area is outside the deformed small ellipse on the abdominal side, the value of the mask data Mask (x, y, z) remains 1 and the process proceeds to step S67. On the other hand, in any one of steps S63, S64, and S65, if it is determined that the pixel (x, y, z) is inside the deformed small ellipse, mask data Mask (x, y, z)= change to 0.

As a result of the processing in steps S63 to S66, if it is inside the deformed small ellipse on either the upper chest side, the boundary part, or the abdominal side, it is judged to be in the vertebrae region, and the mask data Mask (x, y, z) is changed to 0, and if it is outside the deformed small ellipse on the upper chest side, boundary part, and abdomen side, it is judged to be outside the vertebral bone region, and mask data Mask (x, y, z) = It is kept as 1. However, there is a possibility that the vertebra remains in the peripheral region outside the deformed small ellipse determined to be outside the vertebrae region. A vascular region also exists around the remaining vertebral region, and the CT signal value of both the bone region and the vascular region is high and covers each other. It's hard to judge by yourself. However, since the bone region has a characteristic that the signal value changes more than the blood vessel region, a method of determining whether or not the pixels in the peripheral region are included in the bone region is proposed based on the gradient of the signal value. Therefore, for a pixel satisfying mask data Mask (x, y, z)=1 in the peripheral region outside the deformed small ellipse on the ventral side, the gradient vector of the signal value of the pixel adjacent to the pixel in each of the XYZ directions is used to determine whether it is a bone region or not (step S67). The deformed small ellipse has a different shape for each part, but in order to facilitate the judgment process, in the present disclosure, judgment is limited to the area around the deformed small ellipse on the abdominal side, which has the largest area.

Specifically, the region satisfying the following [Formula 21] is defined as the inside of the peripheral region including the outside of the deformed small ellipse on the abdominal side.

[Formula 21]
^xx2 + ^yy2 <3.0

In [Formula 18], the inside and outside of the deformed small ellipse on the upper chest side, the boundary part, and the abdomen side are determined by comparing with 1.0. is included in the peripheral region of the area of . The peripheral area of the deformed small ellipse is obtained by multiplying the geometric shape KB by a predetermined magnification larger than 1 (in this embodiment, the root 3 times), which is an enlarged geometric shape KB (a second geometric shape KB whose horizontal and vertical diameters are enlarged by the root 3 times). geometry).

As a result of the processing according to [Equation 21], if it is inside the peripheral region of the deformed small ellipse, the processing according to [Equation 22] below is performed, and the tomographic image Do (x, y, z) A gradient vector Gr (x, y, z) is calculated.

[Formula 22]
Gr (x, y, z) = [Do (x + 1, y, z) - Do (x - 1, y, z), Do (x, y + 1, z) - Do (x, y - 1, z), Do(x,y,z+1)−Do(x,y,z−1)]

As shown in [Equation 22], when the gradient vector is calculated based on the signal values of the neighboring pixels of the pixel (x, y, z) for each tomographic image (z), the pixel is included as a neighboring pixel. The signal values of the pixels of the same two-dimensional coordinates (x, y) as the pixels included in the two tomographic images (z−1) and (z+1) adjacent to the tomographic image are also included. Further, a process according to [Equation 23] below is executed to determine whether or not the magnitude of the gradient vector Gr(x, y, z) is greater than the bone region lower limit value Sbone, which is a predetermined threshold.

[Formula 23]
|Gr(x,y,z)|>Sbone

A value of 30 to 300 is set as the bone region lower limit value Sbone. In this embodiment, Sbone=60. If the magnitude |Gr(x, y, z)| of the gradient vector Gr(x, y, z) is greater than the bone region lower limit value Sbone, it is determined to be in the vertebrae region, and the mask data Mask(x, y, z)=0.

In the above-described embodiment, the subject area calculation means, the subject area correction means, the subject shape calculation means, and the mask data generation means generate the mask data for the geometric shape KA, and then correct the mask data for the geometric shape KB. A mask value defining whether or not the pixel is included inside the shape KA and outside the geometric shape KB was calculated, and mask data holding the mask value for each pixel was generated. Not limited to the above embodiment, after approximating with a geometric shape KC that is inside the geometric shape KA and outside the geometric shape KB, a mask value that defines whether or not it is included inside the geometric shape KC is calculated, and each It is also possible to generate mask data holding mask values for pixels.

As described above, in the mask generation device 100 according to the present embodiment, the CPU 1, which is a multi-core CPU, executes programs to generate the subject area calculation means 20, the subject area correction means 30, the subject shape calculation means 40, and the mask data generation. Means 50 are implemented. This program divides the read and acquired tomographic images into a plurality of (for example, N) tomographic image groups, assigns a processing thread for each tomographic image group to each CPU thread, and performs each process in parallel. A case in which the mask generating apparatus 100 is realized by a personal computer, which is a general-purpose computer, will be described. FIG. 19 is a diagram showing parallel processing when the mask generation device 100 is implemented by a computer. As shown in FIG. 19, which shows an example of division into N=8 tomographic image groups and executed, subject area calculation means 20, subject area correction means 30, subject shape calculation means 40, and mask data generation are performed. Means 50 are implemented by an application program. For example, when dividing into N=8 tomographic image groups, the application program divides the processing for each of the 8 tomographic image groups into eight processing threads #n1 to #n8. A job scheduler of a multitasking operating system in which subject area calculation means 20, subject area correction means 30, subject shape calculation means 40, and mask data generation means 50 (application programs) operate operates a plurality of CPU cores (physical cores). CPU thread #1 to CPU thread #8.

FIG. 20 is a diagram showing a software flow of parallel processing when the mask generating apparatus 100 is implemented by a computer. FIG. 20 shows an example of division into N=8 tomographic image groups for execution. As shown in FIG. 20, an application program that implements the subject area calculation means 20, the subject area correction means 30, the subject shape calculation means 40, and the mask data generation means 50 performs eight processes on each of the eight tomographic image groups. After dividing into thread #n1 to processing thread #n8, each parallel processing function is activated. Then, the job scheduler of the multitasking operating system allocates eight threads #n1 to #n8 to the CPU threads #1 to #8 of the plurality of CPU cores, respectively. In this way, the operating system allocates the processing threads to be processed to the CPU threads of the CPU cores that are free, thereby enabling parallel processing. In the example of FIG. 20, the application program waits for end messages from each CPU thread, and when the end messages are obtained from all CPU threads, the next process is performed.

Instead of physically allocating each processing thread to each CPU core as shown in FIG. 19 and FIG. FIG. 21 is a diagram showing processing times of processing threads in each CPU core. In FIG. 21, the width (horizontal direction of the drawing) of the rectangle shown on the right side of each CPU core indicates time. When eight processing threads are physically assigned to four CPU cores in the order of thread activation, the processing time of each CPU core varies as shown in FIG. 21(a). In FIG. 21(a), the processing time by the CPU core #1 is significantly longer than the processing time by the CPU core #4, and the processing by the CPU core #1 is rate-determining, extending the total processing time. In this embodiment, the job scheduler considers the availability of four CPU cores and allocates them to eight CPU threads. As a result, as shown in FIG. 21(b), the processing time of each CPU core can be leveled, and the total processing time is shortened compared to FIG. 21(a).

<3. Three-dimensional reconstruction image generation processing>
As an application example of the mask data Mask generated by the mask generation device, a three-dimensional reconstruction image generation device that generates a three-dimensional reconstruction image using the mask data Mask will be described. The three-dimensional reconstructed image generating apparatus of the present disclosure is realized by the hardware configuration as shown in FIG. 2, like the mask generating apparatus, and is realized by the CPU 1 executing a program. The three-dimensional reconstructed image generation apparatus of the present disclosure includes a known three-dimensional reconstructed image generation processing section and mask generation section. The three-dimensional reconstructed image generation processing unit generates a three-dimensional reconstructed image based on three-dimensional voxels configured by arranging pixels of a tomographic image in a voxel space and mask data Mask.

The three-dimensional reconstruction image generation processing unit and the mask generation unit included in the three-dimensional reconstruction image generation apparatus of the present disclosure are both implemented by the CPU 1 executing programs. The mask generator has the same functions as the mask generator of the present disclosure. The three-dimensional reconstruction image generation processing unit provided in the three-dimensional reconstruction image generation device of the present disclosure generates three-dimensional reconstruction images such as volume rendering images, MIP (Maximum Intensity Projection) images, and MPR (Multi-Planar Reconstruction) images. Generate. A known technique can be applied to the three-dimensional reconstructed image generation processing unit, so a detailed description thereof will be omitted. In the three-dimensional reconstruction image generation apparatus of the present disclosure, when generating a three-dimensional reconstruction image as a two-dimensional image from a plurality of tomographic images Do (x, y, z), mask data Mask (x, y, While referring to whether the value of z) is 0 (non-rendering) or 1 (rendering), using the values of the pixels (x, y, z) of the corresponding tomographic image Do, a three-dimensional reconstructed image to generate By generating a three-dimensional reconstructed image in this way, when generating a three-dimensional reconstructed image in which a plurality of tomographic images obtained by imaging the human body are three-dimensionally visualized, no vertebrae or the like are required. It is possible to mask such areas and make them non-drawing.

FIG. 22 is a diagram showing a three-dimensional reconstructed image generated using mask data generated by a mask generation device. 22(a) and 22(b) both show the state of the human body from the side. FIG. 22(a) is an example using mask data generated by a conventional mask generation device that does not use a deformed small ellipse, and FIG. This is an example using generated mask data. As is clear from a comparison of FIGS. 22(a) and 22(b), in the example shown in FIG. 22(b), the upper right vertebrae are removed to generate a three-dimensional reconstructed image.

Although the preferred embodiments of the present disclosure have been described above, the present disclosure is not limited to the above embodiments, and various modifications are possible. For example, in the above embodiment, a multi-core CPU having a plurality of CPU cores and capable of parallel processing is used as the CPU 1, but a single-core CPU in which one CPU core only processes one processing thread is used. may be used.

1 CPU (Central Processing Unit)
2 RAM (random access memory)
3... Storage device 4... Instruction input I/F
5 Data input/output I/F
6... Display unit 7... GPU
8 frame memory 10 image acquisition means 20 object area calculation means 30 object area correction means 40 object shape calculation means 50 mask data generation means 60 object Candidate area 71 Rectangular area (first rectangular area)
72 Rectangular area (second rectangular area)
100... mask generation device

Claims (19)

an image acquisition means for acquiring a plurality of tomographic images;
subject area calculation means for calculating a subject area based on the signal value of each pixel for each tomographic image;
For each tomographic image, the calculated contour of the subject region is approximated by two predetermined geometric shapes, namely, a first geometric shape and a second geometric shape having an area smaller than that of the first geometric shape. subject shape calculation means for calculating a geometric shape and parameters of the second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value for each pixel is held. a mask data generating means for generating mask data for
A mask generator comprising: The mask data generating means arranges the horizontal component of the center coordinates of the second geometric shape near the center of the first geometric shape, and arranges the vertical component of the center coordinates of the second geometric shape to the positioned below the lower end of the first geometric shape, wherein the lateral size of the second geometric shape is the lateral size of the first geometric shape multiplied by a predetermined scaling factor of less than one; 2. The mask generator of claim 1, wherein the vertical size of the second geometric shape is the vertical size of the first geometric shape multiplied by a predetermined scaling factor. The mask data generating means defines, for each tomographic image, an enlarged second geometric shape obtained by multiplying the second geometric shape by a predetermined magnification larger than 1, and the inside of the first geometric shape And for a pixel included outside the second geometric shape and included in the expanded second geometric shape, a gradient vector is calculated based on the signal values of neighboring pixels of the pixel, and the calculated gradient vector is If the magnitude of the gradient vector is greater than a predetermined threshold, the pixel is considered not included inside the first geometric shape and outside the second geometric shape, and the mask value is calculated. A mask generation apparatus according to claim 1. For each tomographic image, the mask data generating means, when calculating a gradient vector based on a signal value of a pixel in the neighborhood of the pixel, uses two tomographic images adjacent to the tomographic image including the pixel as the neighborhood pixel. 4. The mask generation device according to claim 3, wherein the signal values of pixels having the same two-dimensional coordinates as the included pixels are also included. 2. The mask generation according to claim 1, wherein said subject area calculating means calculates, as said subject area, a first rectangular area which is a rectangular area enclosing an outermost area of an area in which pixels having a signal value greater than a predetermined threshold exist. Device. The object area calculating means calculates, as the object area, a second rectangular area defined inside a first rectangular area which is a rectangular area surrounding an outermost area of an area in which pixels having a signal value larger than a predetermined threshold exist. to calculate
A mask generation apparatus according to claim 1. The object area calculating means calculates the center coordinates of the second rectangular area so that the center coordinates are weighted by the signal values of the pixels in the first rectangular area whose signal values are greater than the predetermined threshold value. calculating a second rectangular region;
7. A mask generation device according to claim 6. The subject area calculation means is
A second centroid which is barycentric coordinates weighted by the signal values of pixels whose signal values are greater than the predetermined threshold value within the range of the first rectangular area in four directions, up, down, left, and right, starting from the center coordinates of the second rectangular area. is calculated at 4 points,
calculating the horizontal size of the second rectangular area based on the calculated second center of gravity in the left direction and the second center of gravity in the right direction;
calculating the vertical size of the second rectangular area based on the calculated second center of gravity in the upward direction and the second center of gravity in the downward direction;
8. The mask generation device according to claim 7. The subject area calculation means is
A third barycentric coordinate is weighted by the signal value of a pixel whose signal value is greater than the predetermined threshold within the range of the first rectangular area in four directions, up, down, left, and right, starting from the four calculated second barycenters. Calculate the center of gravity at four points,
calculating the horizontal size of the second rectangular area based on the calculated third center of gravity in the left direction and the calculated third center of gravity in the right direction;
calculating the vertical size of the second rectangular area based on the calculated third center of gravity in the upward direction and the third center of gravity in the downward direction;
9. A mask generation device according to claim 8. The subject area calculation means is
calculating a first horizontal size, which is the distance from the calculated second or third center of gravity in the left direction to the center coordinates of the second rectangular area;
calculating a second horizontal size, which is the distance from the calculated second or third center of gravity in the right direction to the center coordinates of the second rectangular area;
calculating a first vertical size, which is the distance from the calculated second or third center of gravity in the upward direction to the center coordinates of the second rectangular area;
calculating a second vertical size, which is the distance from the calculated second or third center of gravity in the downward direction to the center coordinates of the second rectangular area;
10. The mask generation device according to claim 8 or 9. The first geometry has a center coordinate, a first horizontal size that is the distance from the center coordinate to the left edge, a second horizontal size that is the distance from the center coordinate to the right edge, and a center coordinate to the top edge. and a second vertical size that is the distance from the center coordinate to the bottom edge;
The object shape calculation means is
calculating the center coordinates of the second rectangular area as the center coordinates of the first geometric shape;
calculating a first lateral size of the second rectangular region as the first lateral size of the first geometric shape;
calculating a second lateral size of the second rectangular region as the second lateral size of the first geometric shape;
calculating a first vertical size of the second rectangular region as the first vertical size of the first geometric shape;
calculating a second vertical size of the second rectangular region as the second vertical size of the first geometric shape;
11. The mask generation device according to claim 10. The object shape calculation means is
The calculated first lateral size and second lateral size of the first geometric shape are multiplied by scaling factors for different lateral sizes to obtain the first lateral size and the second lateral size. Correct the horizontal size,
The calculated first longitudinal size and second longitudinal size of the first geometric shape are multiplied by scaling factors for different longitudinal sizes to obtain the first longitudinal size and the second longitudinal size. to correct the vertical size,
12. The mask generation apparatus according to claim 11. The second geometry has a center coordinate, a first lateral size that is the distance from the center coordinate to the left edge, a second lateral size that is the distance from the center coordinate to the right edge, a center coordinate to the top edge and A geometric shape defined by a vertical size that is the distance to the bottom edge,
The object shape calculation means is
calculating the horizontal component of the center coordinates of the second rectangular area as the horizontal component of the center coordinates of the second geometric shape;
As the vertical component of the center coordinates of the second geometric shape, the second vertical size of the second rectangular area is calculated based on the value obtained by calculating the vertical component of the center coordinates of the second rectangular area. Calculate the value obtained by adding the value obtained by
calculating, as the first horizontal size of the second geometric shape, a value obtained by multiplying the calculated value of the first horizontal size of the second rectangular area by a predetermined scaling factor of less than 1;
calculating, as the second horizontal size of the second geometric shape, a value obtained by multiplying the calculated value of the second horizontal size of the second rectangular area by the scaling factor of less than 1;
Calculate a value obtained by multiplying the value obtained by calculating the second vertical size of the second rectangular area by a predetermined magnification as the vertical size of the second geometric shape;
11. The mask generation device according to claim 10. The object shape calculation means is
adding a predetermined offset value to the vertical component of the calculated center coordinates of the second geometric shape to correct the vertical component of the center coordinates;
The calculated first lateral size and second lateral size of the second geometric shape are multiplied by scaling factors for different lateral sizes to obtain the first lateral size and the second lateral size. Correct the horizontal size,
correcting the vertical size by multiplying the calculated first vertical size of the second geometric shape by a scaling factor for the vertical size;
14. A mask generation apparatus according to claim 13. subject area correction means for correcting the subject area by a correction method according to the ratio of the vertical size and the horizontal size of the subject area calculated by the subject area calculation means for each tomographic image,
15. The mask generation device according to any one of claims 6 to 14. The subject area correction means calculates an aspect ratio by dividing the vertical size of the second rectangular area calculated by the subject area calculation means by the horizontal size,
calculating an area ratio obtained by dividing the number of pixels whose signal value is greater than the predetermined threshold within the first rectangular region by the area of the second rectangular region;
calculating a weight value based on the difference between the aspect ratio and a predetermined threshold, and using the calculated weight value to calculate a composite ratio obtained by synthesizing the aspect ratio and the area ratio;
correcting each of the horizontal size and the vertical size of the second rectangular area calculated by the subject area calculation means by multiplying the synthesis ratio;
16. A mask generation apparatus according to claim 15. an image acquisition means for acquiring a plurality of tomographic images;
subject area calculation means for calculating a subject area based on the signal value of each pixel for each tomographic image;
For each tomographic image, the contour of the calculated subject region is approximated by two predetermined first geometric shapes and a second geometric shape having a smaller area than the first geometric shape, and the first geometric shape and the subject shape calculation means for calculating a second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value of each pixel is held. a mask data generating means for generating mask data for
a three-dimensional reconstructed image generating means for generating a three-dimensional reconstructed image based on a three-dimensional voxel configured by arranging each pixel of the tomographic image in a voxel space and the mask data;
A three-dimensional reconstruction image generating device comprising: the computer
an image acquisition step of acquiring a plurality of tomographic images;
a subject region calculation step of calculating a subject region based on the signal value of each pixel for each tomographic image;
For each tomographic image, the contour of the calculated subject region is approximated by two predetermined first geometric shapes and a second geometric shape having a smaller area than the first geometric shape, and the first geometric shape and the a subject shape calculation step of calculating a second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value of each pixel is held. a mask data generation step for generating mask data for
A mask generation method that performs the computer,
image acquisition means for acquiring a plurality of tomographic images;
subject area calculation means for calculating a subject area based on the signal value of each pixel for each tomographic image;
For each tomographic image, the contour of the calculated subject region is approximated by two predetermined first geometric shapes and a second geometric shape having a smaller area than the first geometric shape, and the first geometric shape and the object shape calculation means for calculating a second geometric shape;
For each tomographic image, a mask value defining whether each pixel of the tomographic image is included inside the first geometric shape and outside the second geometric shape is calculated, and the mask value of each pixel is held. mask data generation means for generating mask data for
A program that acts as a

Images