WO2014119412A1

WO2014119412A1 - Medical image processing device, and medical image capture device

Info

Publication number: WO2014119412A1
Application number: PCT/JP2014/050944
Authority: WO
Inventors: 正和岡部
Original assignee: 株式会社日立メディコ
Priority date: 2013-01-30
Filing date: 2014-01-20
Publication date: 2014-08-07
Also published as: JPWO2014119412A1

Abstract

The purpose of the present invention is to provide a technology which can perform noise elimination on a three-dimensional image by using an isotropic three-dimensional adaptive filter. With respect to a medical image including a three-dimensional reconstructed image (40) of a subject, the present invention sets three-dimensional regions of interest (41a-41d) centred on target pixels, and detects the travelling directions of structures (42a-42c) present in the three-dimensional regions of interest on the basis of the distribution of pixel values included in the three-dimensional regions of interest. The present invention sets three-dimensional spatial filters, the shapes of which are aligned in the travelling directions of the structures, or if there is no structure, sets a three-dimensional smoothing filter which does not take structural shapes into account, and performs noise elimination processing on the three-dimensional areas of interest by using the three-dimensional spatial filters.

Description

Medical image processing apparatus and medical image imaging apparatus

The present invention relates to a medical image processing apparatus and a medical image imaging apparatus, and more particularly to removal of noise components in a three-dimensional reconstructed image.

In Patent Document 1, two-dimensional X-ray images are subjected to two-dimensional adaptive filter processing that selectively reduces noise components without reducing linear shadows or edge-shaped shadow contrast. A diagnostic device is described.

Japanese Patent No. 4111762

However, the adaptive filter process described above is a two-dimensional adaptive filter process, and when trying to apply to a three-dimensional image, Cited Document 1 does not consider a three-dimensional structure, so appropriate noise An image that has been subjected to the removal process could not be acquired.

The present invention has been made in view of the above problems, and an object of the present invention is to provide a technique for acquiring an image obtained by performing appropriate noise removal processing on a three-dimensional image.

The present invention sets a three-dimensional region of interest centered on a pixel of interest on a medical image including a three-dimensional reconstructed image of a subject, and based on a distribution of pixel values included in the three-dimensional region of interest, Detecting the traveling direction of the structure included in the three-dimensional region of interest, setting a three-dimensional spatial filter having a shape along the traveling direction of the structure, and using the three-dimensional spatial filter, the three-dimensional region of interest A noise removal process is performed on the image.

According to the present invention, it is possible to provide a technique for acquiring a three-dimensional image in which appropriate noise removal processing has been performed on the three-dimensional image.

Functional block diagram of a medical image processing apparatus to which the present invention is applied The flowchart which shows the flow of a process of the medical image processing apparatus which concerns on 1st embodiment. Flowchart showing details of step S3 in FIG. Explanatory drawing which shows the process of the medical image processing apparatus which concerns on 1st embodiment. Explanatory drawing which shows an example of the operation screen which sets a line structure contrast and a surface structure contrast Explanatory view showing an example of an operation screen for setting the threshold TH ₂ for detecting the threshold value TH ₁ and planar structure for detecting a linear structure Functional block diagram of a cone beam X-ray CT apparatus (C-arm system) to which the present invention is applied

The medical image processing apparatus according to the present embodiment includes a region-of-interest setting unit that sets a three-dimensional region of interest around a pixel of interest on a medical image including a three-dimensional reconstructed image of a subject, and the three-dimensional Based on the distribution of pixel values included in the region of interest, a structure detection unit that detects the traveling direction of the structure included in the three-dimensional region of interest, and a three-dimensional shape having a shape along the traveling direction of the structure And a filter processing unit that sets a spatial filter and performs noise removal processing on the three-dimensional reconstructed image using the three-dimensional spatial filter.

Further, the structure detection unit can detect at least one traveling direction of the linear structure or the planar structure included in the three-dimensional region of interest. In addition, a planar structure is a structure spreading in a planar shape, and the traveling direction means a direction spreading in a planar shape.

The filter processing unit performs a convolution operation in a plane direction parallel to the traveling direction of the planar structure and a line filter that performs an axial convolution operation parallel to the traveling direction of the linear structure. The three-dimensional spatial filter can be generated by combining the surface filter with a predetermined intensity ratio.

In addition, the structure detection unit, based on the covariance using the product of the deviation of the coordinate value of the pixel included in the three-dimensional region of interest and the pixel value of the pixel, At least one traveling direction of the planar structure may be detected. At that time, the structure detection unit generates a cubic square matrix using the covariance, calculates three eigenvalues of the cubic square matrix and eigenvectors for each eigenvalue, and uses these three eigenvalues and eigenvectors. Then, at least one of the traveling direction of the linear structure and the traveling direction of the planar structure may be detected.

Further, the structure detection unit calculates the direction of the eigenvector corresponding to the eigenvalue having the largest absolute value among the three eigenvalues as the traveling direction of the linear structure included in the three-dimensional region of interest. May be. A plane orthogonal to the eigenvector corresponding to the eigenvalue having the smallest absolute value among the three eigenvalues may be calculated as the traveling direction of the planar structure included in the three-dimensional region of interest.

Further, the structure detection unit generates a cubic square matrix using the covariance, calculates three eigenvalues of the cubic square matrix and eigenvectors for each eigenvalue, and uses these three eigenvalues and eigenvectors. Calculating at least one of a contrast of the linear structure with respect to a background region in the three-dimensional region of interest and a contrast of the planar structure with respect to a background region in the three-dimensional region of interest, and the filtering process The unit may increase or decrease a weighting factor by which the pixel value in the three-dimensional spatial filter is multiplied according to the contrast of the linear structure and the contrast of the planar structure.

Further, the structure detection unit calculates a ratio between the eigenvalue having the largest absolute value and the square root of the product of the second largest eigenvalue and the smallest eigenvalue among the three eigenvalues. The contrast of the linear structure included in the region of interest may be calculated. Further, the structure detection unit is configured to calculate a ratio of a square root of a product of the largest eigenvalue and the second largest eigenvalue among the three eigenvalues and a eigenvalue having the smallest absolute value to the three-dimensional eigenvalue. The contrast of the planar structure in the region of interest may be calculated.

Furthermore, the filter processing unit may generate a surface filter having directivity based on the eigenvalue having the largest absolute value and the eigenvalue having the second largest absolute value among the three eigenvalues.

Further, the structure detection unit may obtain the traveling direction of the three-dimensional region of interest by weighted averaging using the traveling direction calculated from the region of interest located around the three-dimensional region of interest. .

Further, the structure detection unit calculates a linear structure contrast or a planar structure contrast calculated from the three-dimensional region of interest from a region of interest located around the three-dimensional region of interest. It may be obtained by weighted averaging using the contrast of the linear structure or the contrast of the planar structure.

In addition, the medical imaging apparatus according to the present embodiment generates a three-dimensional reconstructed image of a subject by performing a reconstruction operation based on the imaging unit that captures the subject and generates image data, and the image data. A region of interest setting unit for setting a three-dimensional region of interest around the pixel of interest on a medical image including the three-dimensionally reconstructed image, and a pixel included in the three-dimensional region of interest Based on the distribution of values, a structure detection unit that detects the traveling direction of the structure included in the three-dimensional region of interest, and a three-dimensional spatial filter having a shape along the traveling direction of the structure, Using the three-dimensional spatial filter, a filter processing unit that performs noise removal processing on the three-dimensional reconstruction image, and a display control unit that performs control for displaying the three-dimensional reconstruction image after the noise removal processing, It is characterized by having.

The imaging unit includes an X-ray source that generates X-rays, and an X-ray flat panel detector configured by arranging detection elements that detect the X-rays in a two-dimensional array, and the region of interest setting unit includes the The three-dimensional region of interest having the same shape as the shape of the reconstruction field formed by rotating the image receiving surface of the detection element by 360 ° may be set. Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings.

<First embodiment>
The first embodiment detects a traveling direction of a structure within a three-dimensional region of interest centered on a pixel of interest on a three-dimensional reconstruction image, sets a three-dimensional spatial filter along the traveling direction, and generates noise. It is embodiment which performs a removal process. Hereinafter, the first embodiment will be described with reference to FIGS. FIG. 1 is a functional block diagram of the medical image processing apparatus according to the present embodiment. FIG. 2 is a flowchart showing a process flow of the medical image processing apparatus according to the first embodiment. FIG. 3 is a flowchart showing details of the process in step S3 of FIG. FIG. 4 is an explanatory diagram showing the contents of processing of the medical image processing apparatus according to the first embodiment.

A medical image processing apparatus 1 shown in FIG. 1 includes a control calculation unit 20 that performs noise removal processing on a three-dimensional reconstructed image by setting a three-dimensional spatial filter, and a mouse and a keyboard for inputting various parameters and instructions. Or an information input device 30 composed of a trackball or the like, and a display device 40 for displaying an image. The control calculation unit 20 includes a reconstructed image acquisition unit 21 that acquires a three-dimensional reconstructed image that is a target of noise removal processing, and a three-dimensional region of interest (hereinafter referred to as a target pixel on the three-dimensional reconstructed image). A region of interest setting unit 22 for setting a 3D region of interest), a structure detecting unit 23 for detecting a traveling direction of a structure included in the three-dimensional region of interest, and a detected traveling direction of the structure A filter processing unit 24 that sets a three-dimensional spatial filter and performs noise removal processing, and a display control unit 25 that performs display control of the three-dimensional reconstructed image after the filter processing. The reconstructed image acquisition unit 21 acquires image data from the imaging unit 10 that captures an image of a subject and generates image data, performs a reconstruction operation based on the acquired image data, and generates an image reconstruction image that generates a three-dimensional reconstructed image. It may be configured as a configuration unit, or read or receive a 3D reconstructed image to be subjected to noise removal from an image storage unit (or image database) 11 that stores a 3D reconstructed image generated in advance. Alternatively, the image reading unit may be configured.

Next, the processing contents of the medical image processing apparatus 1 will be described along the steps of FIG. 2 and FIG.

(Step S1)
The reconstructed image acquisition unit 21 reads or generates a three-dimensional reconstructed image that is a target of noise removal processing (S1).

(Step S2)
The region-of-interest setting unit 22 sets a three-dimensional region of interest centered on the pixel of interest on the three-dimensional reconstruction of the subject read or generated in step S1 (S2). In the following description, the coordinates of the pixel of interest are represented as (I, j, k), and the pixel value of the pixel of interest is represented as v (I, j, k). Then, a cubic three-dimensional region of interest consisting of 5 × 5 × 5 pixels centered on the pixel of interest is set, but the number of pixels included in the three-dimensional region of interest and the shape of the three-dimensional region of interest are not limited to these. For example, when the medical image processing apparatus 1 according to this embodiment is mounted on an X-ray CT apparatus, the shape of a three-dimensional region of interest is determined by an X-ray image of a detection element of an X-ray detector mounted on the X-ray CT apparatus. You may comprise so that the shape of the reconstruction visual field obtained by rotating a surface 360 degrees may be the same shape. For example, when the X-ray image receiving surface is rectangular, the three-dimensional region of interest may be configured in a cylindrical shape configured by rotating the rectangular surface. In addition, when an image intensifier is used as the X-ray detector and the X-ray receiving surface is circular, the 3D reconstruction area is spherical, so the 3D region of interest is also configured in a spherical shape. May be.

(Step S3)
The structure detection unit 23 detects the traveling direction of the structure in the three-dimensional region of interest based on the distribution of pixel values in the three-dimensional region of interest (S3). In the present embodiment, an example of processing for detecting a distribution of pixel values in a three-dimensional region of interest using eigenvalues and eigenvectors of a covariance matrix will be described. The processing of this step will be described along the order of steps in FIG.

(Step S3-1)
The structure detection unit 23 uses a product of the deviation (moment) of the coordinate value of each pixel included in the three-dimensional region of interest and the pixel value, and weights the deviation (moment) of the coordinate value with the pixel value. The covariance of the region of interest is obtained, and a 3 × 3 covariance matrix is generated (S3-1). The following equation (1) shows a 3 × 3 covariance matrix.

It becomes.
Therefore, the following covariance matrix (1) is expressed by equation (1) ′.

(Step S3-2)
The structure detection unit 23 solves the eigen equation of the 3 × 3 covariance matrix and obtains the eigen value and the eigen vector corresponding to each eigen value (S3-2). Therefore, first, an eigen equation shown in the following equation (2) is obtained from the covariance matrix (1) ′.

Develop the eigen equation of equation (2) to obtain a cubic equation (see equation (3) below) for the three eigenvalues λ = (α, β, γ).

The solution λ = (α, β, γ) of the cubic equation a · λ ³ + b · λ ² + c · λ + d = 0 is given by the following equation (4), where the cube root of 1 is ω.

In the formula of λ in the above (4), the inner root ^{^{(R 2 -Q 3) (-2}} 2 · 3 3) multiplied by ones, discriminant D of the cubic equation is defined by the following equation (5) It has become.

In the discriminant of the cubic equation (formula (5)), a, b, c, and d are defined as in the proviso of the above formula (4) and are positive (0 in special cases). Therefore, the inside of the two cube roots in the formula of the cubic equation solution (see equation (4)) is a conjugate complex number (equal value R in the special case), and is obtained by solving the eigen equation of equation (2) above The three eigenvalues λ (= α, β, γ) are all real numbers regardless of the value of the covariance matrix. Eigenvalues The three eigenvalues λ (= α, β, γ) can be written as in the following equation (6).

Next, eigenvectors for the three eigenvalues α, β, and γ are obtained.
The eigenvector

far. The three eigenvectors of Equation (7) are

Meet.
When the 3 × 3 orthogonal matrix U of the following equation (9), which has three eigenvectors of the equation (7) in a column, is configured,

The covariance matrix (1) is diagonalized using an orthogonal matrix U and an inverse matrix U ⁻¹ obtained by transposing U (formula (10)).

The orthogonal matrix U in Equation (9) has four parameters e ₀ , e ₁ , e ₂ , e ₃ (where e ₀ ² + e ₁ ² + e ₂ ² + e ₃ ² = 1) called Euler parameters. And is represented by Formula (11).

The orthogonal matrix U is a rotation matrix having an angle φ with a vector (l, m, n) as a rotation axis in a three-dimensional space. The Euler parameters e ₀ , e ₁ , e ₂ , e _{3 of the} equation (11), the rotation axis (l, m, n), and the rotation angle φ are linked by the following equation (12).

The three eigenvectors of Equation (7) are orthogonal. In the following, consider a case where three eigenvalues λ = (α, β, γ) are arranged so that their absolute values are | α |> | β |> | γ |. According to Equation (10), the eigenvector for the eigenvalue α is the unit vector in the X-axis direction rotated by the orthogonal matrix U, the eigenvector for the eigenvalue β is the unit vector in the Y-axis direction rotated by the orthogonal matrix U, and the eigenvector for the eigenvalue γ Is obtained by rotating the unit vector in the Z-axis direction by the orthogonal matrix U.

Through the above calculation, three eigenvalues and eigenvectors are obtained from the covariance matrix (1), and the traveling direction of the structure in the three-dimensional region of interest set in step S2 can be determined.

(Step S3-3)
The structure detection unit 23 detects the traveling direction of the structure using the three eigenvalues and eigenvectors obtained in step S3-2. The structure here includes a linear structure and a planar structure. Examples of the linear structure include a blood vessel, a catheter, and a guide wire that guides the catheter. Examples of the planar structure include a bone, an organ surface, and a soft tissue boundary surface. Hereinafter, it should be noted that the planar structure means a boundary surface between organs and tissues and does not necessarily indicate a flat structure. In the present embodiment, first, the presence or absence of a linear structure is determined (S3-3), and then the presence or absence of a planar structure is determined, but this order is changed to determine the presence or absence of a planar structure, Thereafter, the presence or absence of the linear structure may be determined.

In order to determine the presence or absence of a linear structure, the structure detection unit 23 compares the largest value and the second largest value among the absolute values of the three eigenvalues λ (= α, β, γ). and, the ratio is determined whether the first threshold value TH ₁ or more. As the first threshold TH ₁ is smaller, it becomes easier to recognize more linear structures. On the other hand, as the first threshold TH ₁ is larger, it becomes easier to recognize only the linear structures that are captured more clearly. If affirmative, the process proceeds to step S3-4, and if negative, the process proceeds to step S3-5.

(Steps S3-4, S3-5)
The presence / absence of the planar structure is determined (S3-4, S3-5). The structure detection unit 23 has the ratio of the largest and smallest of the absolute values of the three eigenvalues λ (= α, β, γ), and the ratio of the second largest to the smallest. determining whether both are the second threshold value TH ₂ or more. As the second threshold TH ₂ is smaller, more planar structures are more easily recognized. On the other hand, as the second threshold TH ₂ is larger, it is easier to recognize planar structures that have been captured more clearly. If the determination is affirmative, the process proceeds to step S3-6 (in the case of S3-5, the process proceeds to S3-8). If the determination is negative, the process proceeds to step S3-7 (in the case of S3-5, the process proceeds to S3-9).

(Step S3-6)
It is determined that both the linear structure and the planar structure are included (S3-6). The structure detection unit 23 travels the direction of the eigenvector corresponding to the eigenvalue having the largest absolute value among the above three eigenvalues λ (= α, β, γ) of the linear structure in the three-dimensional region of interest. Calculate as direction. Furthermore, the structure detection unit 23 calculates a plane structure orthogonal to the eigenvector corresponding to the eigenvalue having the smallest absolute value among the three eigenvalues λ (= α, β, γ) in the planar structure in the three-dimensional region of interest. Detect three types of travel directions.

(Step S3-7)
The structure detection unit 23 determines that only the linear structure is included, and similarly calculates the traveling direction of the linear structure (S3-7).

(Step S3-8)
The structure detection unit 23 determines that only the planar structure is included, and calculates the traveling direction of the planar structure as described above (S3-8).

(Step S3-9)
The structure detection unit 23 determines that no structure is included in the three-dimensional region of interest (S3-9).

(Step S4)
The filter processing unit 24 generates a three-dimensional spatial filter having a shape along the traveling direction of the structure (S4). Based on FIG. 4, the contents of the filtering process in this step will be described. It is assumed that three-dimensional regions of

interest

41a, 41b, 41c, and 41d are set in the three-dimensional reconstructed image 40 before the filter processing in FIG. Only the linear structure 42a is included in the 3D region of interest 41a, the linear structure and the planar structure 42b are included in the 3D region of interest 41b, and the planar shape is included in the 3D region of interest 41c. Only the structure 42c is included. In addition, no structure is included in the three-dimensional region of interest 41d. When the process of step S3 is performed on the three-dimensional region of

interest

41a, 41b, 41c, 41d, step S3-7 is performed on the three-dimensional region of interest 41a, and step S3-6 is performed on the three-dimensional region of interest 41b. The determination result of step S3-8 is obtained for the three-dimensional region of interest 41c, and step S3-9 is obtained for the three-dimensional region of interest 41d.

Therefore, the filter processing unit 24 is a line filter for the three-dimensional region of interest 41a, a line filter and a surface filter for the three-dimensional region of interest 41b, a surface filter for the three-dimensional region of interest 41c, A three-dimensional smoothing filter that does not consider the shape of the structure is set for the three-dimensional region of interest 41d.

The 3D spatial filter is a function that determines the weighting coefficient to be set for each pixel when the pixel value of the pixel of interest is determined by convolution calculation of the pixel value of the pixel included in the 3D region of interest. Then, the weighting factor is determined so as to perform the convolution calculation in the axial direction parallel to the traveling direction of the linear structure, and the surface filter performs the convolution calculation in the surface direction parallel to the traveling direction of the planar structure. Thus, the weighting factor is determined. Specifically, the line filter is such that among the pixels included in the region of interest, the pixel value of the pixels arranged in the direction parallel to the traveling direction of the linear structure from the target pixel is relatively heavy. A function in which a weighting factor w ₁ (w ₁ > 0) is set and a weighting factor w ₂ (w ₁ > w ₂ or w ₂ = 0) is set so that the weight is relatively small for other pixels. It is. The surface filter is such that, among the pixels included in the region of interest, the weight is relatively heavy with respect to the pixel value of the pixel located in the traveling direction (surface spreading direction) of the planar structure around the pixel of interest. This is a function in which a weighting coefficient is set and a weighting coefficient is set so that the weight is relatively small for other pixels.

Next, the filter processing unit 24 performs a convolution operation by multiplying each pixel included in the three-dimensional region of

interest

41a, 41b, 41c, 41d by the weighting factor of the set three-dimensional spatial filter. The value obtained as a result is calculated as a filtered pixel value of the pixel of interest P. If it is determined that the region includes both a linear structure and a planar structure (like many tissues in this case), as shown in the region of interest 41b in FIG. 4, a line filter and Apply both area filters. In this case, it is possible to give priority to one of the line filters and the surface filters with different strengths depending on the inspection target. The setting of the filter strength will be described in another embodiment.

(Step S5)
The filter processing unit 24 determines whether or not the processing from steps S2 to S4 has been completed for all the pixels of the three-dimensional reconstructed image. If the determination is affirmative, the process proceeds to step S6. If the determination is negative, the process returns to step S2, and the subsequent processing is repeated (S5). That is, the processing from step S2 to step S4 is executed while scanning the three-dimensional region of interest over the entire region of the three-dimensional reconstructed image.

(Step S6)
The display control unit 25 displays the filtered three-dimensional reconstructed image on the screen of the display device 80 (S6).

According to the present embodiment, by performing noise removal processing using a spatial filter in a three-dimensional direction having a shape along the traveling direction of the structure in the three-dimensional region of interest, the background region (in the three-dimensional region of interest Only the noise in the area (excluding the structure) is selectively reduced. Therefore, it is possible to selectively reduce the noise component of the three-dimensional reconstructed image without reducing the contrast of the linear / planar structure. In other words, the ability to extract linear structures and planar structures can be relatively enhanced.

In the flowchart of FIG. 2, when all the pixels are processed in step S5, the reconstructed image after the filter processing is displayed. However, after all the pixels are processed, the process returns to step S1 again, Multiple filtering processes may be performed on the three-dimensional reconstructed image. Also in this case, according to the present embodiment, noise is selectively removed from the background region, so that the ability to extract a structure can be improved even if filter processing is performed a plurality of times.

<Second embodiment>
The second embodiment is an embodiment in which the weight coefficient w ₁ is changed in a three-dimensional spatial filter. The the way of change of the weighting coefficients w _1, there are mainly the following three aspects.

(2-1) A mode in which the contrast between the structure in the three-dimensional region of interest and the background region is used The structure detector 23 has the largest absolute value among the three eigenvalues of the covariance matrix of the three-dimensional region of interest. Calculate the ratio of the large eigenvalue to the square root of the product of the second largest eigenvalue and the smallest eigenvalue as the contrast C ₁ of the linear structure in the 3D region of interest (hereinafter referred to as “line structure contrast”) (Refer to the following formula (13)).

However, | α |> | β |> | γ |

Further, the structure detection unit 23, among the three eigenvalues of the covariance matrix of the three-dimensional region of interest, the square root of the product of the eigenvalue having the largest absolute value and the second largest eigenvalue, the eigenvalue having the smallest absolute value, Is calculated as the contrast C ₂ of the planar structure in the three-dimensional region of interest (hereinafter referred to as “plane structure contrast”) (see the following equation (14)).

However, | α |> | β |> | γ |

The larger the line structure contrast C ₁ and the surface structure contrast C ₂ , the larger the difference between the pixel value of the structure in the three-dimensional region of interest and the pixel value of the background region pixel. Therefore, the filter processing unit 24, so that the difference between the line structures contrast C ₁ larger the weighting coefficients w ₁ and the weight coefficient w ₂ is large, it sets the weighting factor. For example, the weight coefficient w ₁ is set to a value relatively close to “1”, and the weight coefficient w ₂ is set to a value relatively close to “0” or 0. The same applies to the foliation contrast C _2. Thereby, in the three-dimensional region of interest where the line structure contrast C ₁ and the surface structure contrast C ₂ are large, the pixel value of the pixel of interest after the convolution calculation is increased by increasing the contribution ratio of the pixel values of the pixels constituting the structure. The value can be calculated, and noise can be removed while suppressing a decrease in the definition of the structure.

(2-2) A mode in which the weighting coefficient w ₁ is changed in the planar structure The surface filter has directivity based on the eigenvalue having the largest absolute value and the eigenvalue having the second largest absolute value. Good. For example, the weight value w ₁₁ (w ₁₁ > 0) is used for the pixel value of the pixel located along the eigenvector for the eigenvalue having the largest absolute value, and the pixel value of the pixel located along the eigenvector for the eigenvalue having the second largest absolute value is used. The pixel value may be multiplied by a weighting factor w ₁₂ (0 <w ₁₂ <w ₁₁ ). Thereby, even in the same plane, the pixel value of the pixel of interest after the convolution calculation can be calculated while relatively increasing the contribution ratio of the pixel value having higher directivity.

(2-3) aspect weighting coefficients w ₁ to change the weighting coefficients w _1, depending on the distance from the target pixel, depending on the distance from the target pixel may be changed. For example, it may be relatively high weighting factor w ₁ for multiplying the pixel value of the pixel near the target pixel. As a result, the pixel value of the target pixel after the convolution calculation can be calculated by increasing the contribution ratio of the pixel closer to the target pixel in the three-dimensional spatial filter, and the blur of the image due to the noise removal processing is suppressed. be able to. The mode shown in (2-3) can be applied alone or in combination with any of (2-1) and (2-2).

According to this embodiment, the weighting coefficient to be multiplied with respect to the pixels in the three-dimensional spatial filter can be changed according to the difference from the background area, the directionality of the surface filter, and the distance from the target pixel. Thus, noise removal processing can be performed without causing blur.

<Third embodiment>
The third embodiment is an embodiment in which the strength of a line filter and a surface filter is set using an operation GUI. Hereinafter, the present embodiment will be described based on FIG. FIG. 5 is an explanatory diagram illustrating an example of an operation screen for setting the strengths of the line filter and the surface filter.

The operation screen 50 in FIG. 5 is a screen for setting the strength of the line filter and the surface filter for each imaging region and imaging condition. In the operation screen 50, there are four tags, “head” 51, “chest” 52, “abdomen” 53, and limb “54” as an imaging part selection unit, and “imaging part for adding an imaging part” An “add” button 55 is provided, and “abdomen” 53 is selected in FIG. Further, a “standard reconstructed image” column 61 and a “DSA reconstructed image” column 62 are provided for each photographing condition.

In each of the “standard reconstructed image” column 61 and the “DSA reconstructed image” column 62, setting bars 62 and 65 for setting the line filter strength, and setting

bars

63 and 66 for setting the surface filter strength, There is. The example of FIG. 5 shows a case where the minimum value of the settable line filter strength is 0 and the maximum value is 1.0, and the minimum value of the settable surface filter strength is 0 and the maximum value is 1.0. On each setting bar,

sliders

62a, 63a, 65a, 66a that can move along the axial direction of each bar are provided, and setting values corresponding to the positions of the sliders are displayed in setting

value display fields

62b, 63b, 64b, Displayed in 65b. Note that the strength of the line filter and the strength of the surface filter may be set independently, or may be set so as to have a correlation such that when one is strong, the other is weakened.

Slider 62a is for setting the first coefficient multiplying against lineation contrast C ₁ of three-dimensional spatial filter to be used for standard reconstructed image, slider 63a is used for standard reconstructed image The second coefficient to be multiplied to the surface structure contrast C ₂ of the three-dimensional spatial filter to be obtained is set. When the first coefficient is relatively large, the line structure contrast C ₁ also increases, and the weight coefficient w ₁ multiplied by the pixel located along the traveling direction of the linear structure around the target pixel is large. As a result, the strength of the line filter can be increased. When the strength of the line filter and the strength of the surface filter are correlated, when the first coefficient related to the line filter is large, the second coefficient is relatively small, the surface structure contrast C ₂ is also small, and the surface structure results weighting coefficients w ₁ to be multiplied by the three-dimensional space pixels in the filter which travel is set along the direction of the smaller, the strength of the surface filter is weakened. The same applies to the first coefficient and the second coefficient of the DSA reconstructed image 64.

In general, in a DSA reconstructed image, there is a demand for highlighting only contrast-enhanced blood vessels, so the structure (contrast-enhanced blood vessels) is clearer than the non-structure (contrast-enhanced blood vessel region) compared to the standard reconstructed image. It is desirable to display on. Further, since the contrast blood vessel is closer to the linear structure, it is desirable to set the weight coefficient of the linear structure larger than the weight coefficient of the planar structure. In addition, depending on the imaging region, it may be different which of the linear structure or the planar structure is to be emphasized. For example, in general, it may be desired to emphasize the linear structure more strongly in the extremities and more strongly the planar structure in the chest or abdomen. And the intensity ratio of the surface filter can be set, and noise removal processing can be performed according to the shape of the part necessary for diagnosis.

<Fourth embodiment>
The fourth embodiment is an embodiment in which the structure detection threshold (the first threshold TH ₁ and the second threshold TH _{2 described above} ) is set using the operation GUI. Hereinafter, the present embodiment will be described with reference to FIG. FIG. 6 is an explanatory diagram illustrating an example of an operation screen for setting a structure detection threshold.

Operation screen 70 in FIG. 6, the type of the eigenvalues one axis, the graph 71 shows the absolute value of each eigenvalue to the other shaft, and a slider 72 for setting the first threshold value TH ₁ on the graph, the It includes a slider 73 for setting two threshold TH _2, the. The value of the first threshold TH ₁ can be increased or decreased by moving the slider 72 up and down, and the value of the second threshold TH ₂ can be increased or decreased by moving the slider 73 up and down. When the operator wants to draw a linear structure more clearly than a planar structure like a DSA image, the operator sets the first threshold value TH ₁ relatively high. Also, for example, when you want to draw a planar structure more clearly, such as a standard reconstructed image of the abdomen or chest, the second threshold TH ₂ is set relatively low, and the surface shape is drawn more finely. It is possible to generate a three-dimensional reconstruction image drawn on a simple surface.

In the above embodiment, the traveling direction and the contrast are obtained for each set three-dimensional region of interest. However, the weighted average is calculated with the traveling direction and the contrast calculated from the surrounding three-dimensional region of interest, and the set three-dimensional region of interest is set. The traveling direction and contrast may be obtained. For example, in the case of a three-dimensional region of interest including a linear structure (e.g., the three-dimensional region of interest 42a in FIG. 4), neighboring three-dimensional regions of interest (e.g., in FIG. 4) that are adjacent along the traveling direction of the linear structure. The value obtained by weighted average of the traveling direction and contrast of the three-dimensional region of interest 42a centered on the three-dimensional region of interest 42a in the vertical direction in FIG. 4) is obtained as the traveling direction and contrast of the three-dimensional region of interest. Also good.

In addition, in the case of a three-dimensional region of interest including a planar structure (for example, the three-dimensional region of interest 42c in FIG. 4), neighboring peripheral three-dimensional regions of interest (for example, in FIG. 4) along the traveling direction of the planar structure. The three-dimensional region of interest 42c is the center of the three-dimensional region of interest adjacent to the left-right direction and the depth direction in FIG. You may ask as. Thereby, the change of the running direction and contrast between adjacent regions of interest can be smoothed.

Next, an embodiment in which the medical image processing apparatus according to the above embodiment is mounted on a medical image imaging apparatus will be described. In the present embodiment, a C-arm type X-ray CT apparatus will be described as an example of a medical imaging apparatus. However, a gantry type X-ray CT apparatus, an MRI apparatus, an ultrasonic diagnostic apparatus, or the like 3 The present invention can be applied to any medical imaging apparatus that generates a three-dimensional reconstructed image regardless of the type. Hereinafter, based on FIG. 7, the X-ray diagram 7 according to the present embodiment is a functional block diagram of a cone beam X-ray CT apparatus (C-arm system) to which the present invention is applied.

The cone beam X-ray CT apparatus 200 in FIG. 7 irradiates the subject 2 with X-rays and controls the imaging unit 10 that captures the projection data 111 of the subject 2 and each component of the imaging unit 10. A control calculation unit 20 for reconstructing a three-dimensional CT image of the subject 2 based on the projection data 111, and an information input device 30 including a mouse, a keyboard, a trackball, etc. for inputting various parameters and instructions And a display device 40 for displaying an image.

The imaging unit 10 includes a bed 17, an X-ray source 11 that irradiates the subject 2 lying on the bed 17 with X-rays, and an X-ray that is installed opposite to the X-ray source 11 and passes through the subject 2 A two-dimensional X-ray detector 12 that outputs projection data 111 by detecting the X-ray, a C-type arm 13 that mechanically connects the X-ray source 11 and the two-dimensional X-ray detector 12, and the C-type arm 13 A C-type arm holding body 14 to be held, a ceiling support 15 for attaching the C-type arm holding body 14 to the ceiling, and a ceiling for supporting the ceiling support 15 so as to be movable in the two-dimensional direction of front, rear, left and right in the illustrated state. Rail 16 and injector 18 for injecting contrast medium into subject 2 are provided.

The X-ray source 11 includes an X-ray tube 11t that generates X-rays, and a collimator 11c that controls the direction of X-ray irradiation from the X-ray tube 11t to be a cone, a quadrangular pyramid, or a multi-sided pyramid.

For the two-dimensional X-ray detector 12, for example, a flat panel detector (hereinafter referred to as “FPD”) using a TFT element is used. Further, as another example of the two-dimensional X-ray detector 12, an X-ray image intensifier that converts an X-ray transmission image into a visible light image, an optical lens that forms an image of the X-ray image intensifier, and You may use the two-dimensional X-ray detector comprised from the combination of the CCD television camera etc. which image | photograph the visible light image of the X-ray image intensifier imaged by the optical lens. The imaging field of view of the two-dimensional X-ray detector 12 may be any shape such as a circle or a rectangle.

The C-arm 13 rotates around the rotation center axis 4 at every predetermined projection angle when the subject 2 is imaged. As a result, while the X-ray source 11 and the two-dimensional X-ray detector 12 are arranged to face each other, the X-ray imaging is performed by rotating and moving on a circular orbit in substantially the same plane.

The control calculation unit 20 includes an imaging unit control unit 100 that controls the imaging unit 10, an image collection unit 110 that collects and stores the projection data 111 output from the imaging unit 10, and 3 based on the collected projection data 111. An image reconstruction unit 21a that reconstructs the two-dimensional reconstruction image 211, and a region of interest setting unit that sets a three-dimensional region of interest centered on the pixel of interest on the three-dimensional reconstruction image 211 generated by the image reconstruction unit 21a 22, a structure detection unit 23 that detects the traveling direction of the structure included in the three-dimensional region of interest, and a filter that performs noise removal processing by setting a three-dimensional spatial filter based on the detected traveling direction of the structure A processing unit 24 and a display control unit 25 that performs display control for displaying the filtered three-dimensional reconstructed image 241 on the multiplication device 40 are provided.

Further, the information input device 30 has a structure detection threshold (the first threshold TH ₁ and the second threshold TH _{2 described above} ) used when the structure detection unit 23 detects a structure in the three-dimensional region of interest. 301 and a first coefficient to be multiplied by the line structure contrast C ₁ and a second coefficient to be multiplied by the surface structure contrast C ₂ (hereinafter referred to as “contrast coefficient”) 302 in order to set the enhancement ratio between the line filter and the surface filter. Accept input. The structure detection threshold 301 is delivered to the structure detection unit 23. The contrast coefficient 302 is delivered to the filter processing unit 24.

The processing performed by the region-of-interest setting unit 22, the structure detection unit 23, the filter processing unit 24, and the display control unit 25 is the same as the processing described in the first embodiment and the second embodiment for the control calculation unit 20 illustrated in FIG. Therefore, a duplicate description is omitted here.

The imaging unit control unit 100 controls the position of the C-type arm 13 on the ceiling rail 16 and the imaging system rotation control unit 101 that controls the rotational movement of the C-arm 13 around the rotation center axis 4 to control the C-type arm 13. An imaging system position control unit 102 that two-dimensionally controls the position of the mold arm 13 with respect to the subject 2, an X-ray irradiation control unit 103 that controls ON / OFF of a tube current flowing through the X-ray tube 11t, and an injector 18 Injector control unit 104 that controls the injection amount and injection timing of contrast medium to be injected into subject 2, bed control unit 105 for adjusting the position of subject 2 by controlling the position of bed 17, and two-dimensional And a detection system control unit 107 that controls imaging of the projection data 111 by the X-ray detector 12.

According to the medical imaging apparatus according to the present embodiment, a three-dimensional reconstructed image 211 is generated based on the projection data 111 obtained by the imaging unit 10, and a noise is reduced by applying a three-dimensional spatial filter to this. Can be processed. Therefore, a three-dimensional reconstructed image obtained by performing noise reduction processing without reducing linear and planar contrast is obtained. In addition, by using the three-dimensional reconstructed image obtained in this way, even with a low-dose X-ray imaging device, three-dimensional image processing, extraction of thin blood vessels and catheters is possible, and the amount of X exposure of the patient is kept low. A line imaging apparatus can be provided.

1 Medical image processing device, 10 imaging unit, 20 control calculation unit, 30 information input device, 40 display device, 200 cone beam X-ray CT device (medical image imaging device)

Claims

On a medical image including a three-dimensional reconstructed image of a subject, a region of interest setting unit that sets a three-dimensional region of interest centered on a pixel of interest;
Based on the distribution of pixel values included in the three-dimensional region of interest, a structure detection unit that detects the traveling direction of the structure included in the three-dimensional region of interest;
Set a three-dimensional spatial filter having a shape along the traveling direction of the structure, and using the three-dimensional spatial filter, a filter processing unit that performs noise removal processing on the three-dimensional reconstructed image;
A medical image processing apparatus comprising:
The structure detection unit detects at least one traveling direction of a linear structure and a planar structure included in the three-dimensional region of interest;
2. The medical image processing apparatus according to claim 1, wherein:
The filter processing unit includes: a line filter that relatively increases a weight of a pixel value of pixels arranged along the running direction of the linear structure; and a pixel value of pixels arranged along the running direction of the planar structure. A surface filter that relatively increases the weight of the image, and performs a convolution operation on a pixel value included in the three-dimensional region of interest using at least one of the line filter and the surface filter. The medical image processing apparatus according to claim 2.
4. The medical image processing apparatus according to claim 3, wherein the filter processing unit generates the three-dimensional spatial filter by combining the line filter and the surface filter at a predetermined intensity ratio.
5. The medical image processing apparatus according to claim 4, further comprising input means for setting strengths of the line filter and the surface filter.
The structure detection unit, based on a covariance using a product of a deviation of a coordinate value of a pixel included in the three-dimensional region of interest and a pixel value of the pixel, the linear structure and the surface Detecting at least one traveling direction of the structure,
3. The medical image processing apparatus according to claim 2, wherein
The structure detection unit generates a cubic square matrix using the covariance, calculates three eigenvalues of the cubic square matrix and eigenvectors for each eigenvalue, and uses these three eigenvalues and eigenvectors, Detecting at least one of a traveling direction of the linear structure and a traveling direction of the planar structure;
7. The medical image processing apparatus according to claim 6, wherein
The structure detection unit calculates the direction of the eigenvector corresponding to the eigenvalue having the largest absolute value among the three eigenvalues as the traveling direction of the linear structure included in the three-dimensional region of interest.
8. The medical image processing apparatus according to claim 7, wherein
The structure detection unit calculates a plane orthogonal to the eigenvector corresponding to the eigenvalue having the smallest absolute value among the three eigenvalues as the traveling direction of the planar structure included in the three-dimensional region of interest. ,
8. The medical image processing apparatus according to claim 7, wherein
8. The medical device according to claim 7, further comprising input means for setting a threshold value for detecting a traveling direction of the linear structure and / or the planar structure using the three eigenvalues. Image processing device.
The structure detection unit generates a cubic square matrix using the covariance, calculates three eigenvalues of the cubic square matrix and eigenvectors for each eigenvalue, and uses these three eigenvalues and eigenvectors, Calculating at least one of a contrast of the linear structure with respect to a background region in a three-dimensional region of interest, and a contrast of the planar structure with respect to a background region in the three-dimensional region of interest;
The filter processing unit increases or decreases a weighting factor by which a pixel value in the three-dimensional spatial filter is multiplied according to the contrast of the linear structure and the contrast of the planar structure.
7. The medical image processing apparatus according to claim 6, wherein
The structure detection unit is configured to calculate a ratio of the eigenvalue having the largest absolute value among the three eigenvalues to the square root of the product of the second largest eigenvalue and the smallest eigenvalue, and the three-dimensional region of interest. Calculated as the contrast of the linear structure included in
12. The medical image processing apparatus according to claim 11, wherein:
The structure detecting unit is configured to calculate a ratio of a square root of a product of the largest eigenvalue and the second largest eigenvalue among the three eigenvalues to a eigenvalue having the smallest absolute value, and the three-dimensional region of interest. Calculating as the contrast of the planar structure in
12. The medical image processing apparatus according to claim 11, wherein:
The filter processing unit generates a surface filter having directivity based on the eigenvalue having the largest absolute value and the eigenvalue having the second largest absolute value among the three eigenvalues.
14. The medical image processing apparatus according to claim 13, wherein
An imaging unit that images a subject and generates image data;
An image reconstruction unit that performs a reconstruction calculation based on the image data and generates a three-dimensional reconstruction image of the subject;
On the medical image including the three-dimensional reconstructed image, a region-of-interest setting unit that sets a three-dimensional region of interest centered on the pixel of interest;
Based on the distribution of pixel values included in the three-dimensional region of interest, a structure detection unit that detects the traveling direction of the structure included in the three-dimensional region of interest;
Set a three-dimensional spatial filter having a shape along the traveling direction of the structure, and using the three-dimensional spatial filter, a filter processing unit that performs noise removal processing on the three-dimensional reconstructed image;
A display control unit that performs control for displaying the three-dimensional reconstructed image after the noise removal processing;
A medical image imaging apparatus comprising: