JP2002109512A - Candidate shade abnormality detector, and recording medium therefor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce effects by a fine structure, noise, etc., such as mammary gland, blood vessel, etc., in detecting a candidate area, and to utilize a detailed fine structure when evaluating malignancy of the candidate area. SOLUTION: A low resolution image acquiring means 10 acquires the low resolution image data P' based on the input original image data P, a high resolution image acquiring means 30 acquires the high resolution image data P" based on the input original image data P, and a candidate area detecting means 20 detects the candidate area of abnormality in the image by inputting the low resolution image data P'. A malignancy evaluating means 40 inputs the position data on the low resolution image of the candidate area from the candidate area detecting means 20, and inputs the high resolution image data P" from the high resolution image acquiring means 30, and evaluates the malignancy of the candidate area based on the input high resolution image data P" and the position data on the high resolution image acquired by converting the position data in the candidate area on the low resolution image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an abnormal shadow candidate detecting apparatus, and more particularly, to an abnormal shadow candidate detecting apparatus in which processing for detecting an abnormal shadow candidate from a radiation image of a subject is improved.

[0002]

2. Description of the Related Art Conventionally, in the medical field, it has been necessary to interpret a radiographic image of a subject, find a lesion, and observe the state of the lesion to diagnose the presence or absence of a disease and the progress of the disease. Generally done. However, interpretation of radiographic images is influenced by the experience of the interpreter and the level of image interpretation ability, and is not always objective.

[0003] For example, in mammography (diagnosis radiation image of a breast as a subject) photographed for the purpose of examining breast cancer, a tumor shadow, a microcalcification shadow, or the like, which is one of the features of a cancerous portion, is obtained from the image. It is necessary to detect the abnormal shadow, but it is not always possible for the reader to specify the range of the abnormal shadow accurately. For this reason, it has been demanded to accurately detect abnormal shadows such as tumor shadows and microcalcification shadows without depending on the skill of the reader.

In order to meet this demand, abnormal shadow candidate detection is performed in which, based on image data of a subject acquired as a diagnostic image, abnormal shadow candidates in an image represented by the image data are automatically detected using a computer. A processing system (computer-assisted image diagnostic apparatus) has been proposed (Japanese Patent Laid-Open No. 8-29)
No. 4479, JP-A-8-287230, etc.). This abnormal shadow candidate detection processing system automatically detects abnormal shadow candidates using a computer based on the characteristics of the density distribution and morphological characteristics of abnormal shadows, and is mainly used for detecting tumor shadows. An abnormal shadow candidate area is detected by using an appropriate iris filter process or the like.

[0005] The iris filter process compares a iris filter output value representing the maximum value of the degree of concentration of the density gradient of the image signal with a predetermined threshold value, thereby obtaining a tumor, which is one of the characteristic forms of breast cancer in the image. This is an effective method for detecting a shadow candidate area.

On the other hand, in order to further detect an abnormal shadow having a high probability, which is a malignant abnormal shadow, from among the abnormal shadow candidate regions obtained by the iris filter processing or the like, the shape, the inside, and the side of the detected candidate region are detected. A method has been proposed in which the degree of malignancy of a candidate region is evaluated using a feature amount representing the feature of an edge, and a more definite abnormal shadow candidate is finally detected (JP-A-9-167238, etc.).

As a method of evaluating the degree of malignancy of the candidate area, a density histogram inside the candidate area is obtained for the candidate area of the tumor shadow detected by the iris filter processing, and a plurality of feature amounts based on the density histogram are obtained.
That is, a variance value, a contrast, an angular moment, and the like are calculated, each characteristic amount is defined by a predetermined weighting function, a new evaluation function value is calculated, and the degree of malignancy of the candidate area is calculated based on the calculated evaluation function value. There is a method of evaluating and detecting only candidate regions determined to be malignant. In addition, as the feature amount, in addition to the one related to the above-described density histogram,
The variance, the bias, the correlation value, the moment, the entropy, which is the edge information indicating the feature of the edge of the candidate area, the circularity indicating the feature of the shape of the candidate area, and the like are used.

The Mahalanobis distance can be used as the evaluation function value. The Mahalanobis distance means Dmi defined by the following equation (1), and is a distance measured from the center of the distribution by weighting a hyperellipsoid represented by a covariance matrix Σ.

[0009]

(Equation 1) According to equation (1), a Mahalanobis distance Dm1 with a pattern class (i = 1) indicating a non-malignant shadow obtained in advance by experiment and a Mahalanobis distance Dm2 with a pattern class (i = 2) indicating a malignant shadow are calculated. Then, Dm1 and Dm2 are compared to determine whether the candidate area is malignant. For example, Mahalanobis distance D with a pattern class indicating a non-malignant shadow
When m1 is closer to the Mahalanobis distance Dm2 with the pattern class indicating a malignant shadow, that is, when Dm1 <Dm2, the pattern class is a non-malignant shadow and indicates a malignant shadow from the Mahalanobis distance Dm1 with the pattern class indicating the non-malignant shadow. If the Mahalanobis distance Dm2 is short, that is, Dm1>
In the case of Dm2, it is determined that the image is a malignant shadow.

[0010]

By the way, the iris filter used for detecting a candidate region of an abnormal shadow has a hemispherical structure having an ideal density gradient, that is, a density value smaller than that of a surrounding image portion. Is slightly lower and most strongly responds to a structure having a gradient of the concentration value in which the concentration value gradually decreases from the substantially circular peripheral portion toward the center portion.

However, in actual mammography, since a tumor shadow exists in structures such as the mammary gland and blood vessels, the shadow is affected by these local fine structures and noise, and the density of the tumor shadow is reduced. The gradient is unlikely to be an ideal hemispherical structure. For this reason, in the detection processing of the tumor shadow by the iris filter processing, there is a possibility that a detection omission may occur.

It is also known that malignant tumors have irregularities in the center and edges, but such features may also cause noise in the density gradient of the shadow of the tumor. In a detection process using an iris filter, a detection error may be caused.

On the other hand, when the degree of malignancy of the detected candidate area is evaluated, the fine structure such as the edge of the candidate area is an important factor. Therefore, the evaluation is performed based on image data including the detailed fine structure. Need to do.

In view of the above circumstances, the present invention reduces the influence of fine structures such as mammary glands and blood vessels, noise, and the like when performing candidate region detection processing and evaluates the malignancy of the candidate region. It is an object of the present invention to provide an abnormal shadow candidate detecting device which can use a detailed fine structure.

[0015]

SUMMARY OF THE INVENTION An abnormal shadow candidate detecting apparatus according to the present invention acquires low-resolution image data from radiation image data representing a radiation image of a subject, and acquires the low-resolution image data by low-resolution image acquisition means. Based on the obtained low-resolution image data, a candidate area detecting means for detecting a candidate area for an abnormal shadow in the radiographic image, a high-resolution image obtaining means for obtaining high-resolution image data from the radiographic image data, and a high-resolution image obtaining Malignancy evaluation means for evaluating the malignancy of the candidate area based on the high-resolution image data obtained by the means.

Here, the low-resolution image data is an image representing an image excluding a fine structure such as a mammary gland, a blood vessel, an edge portion of a candidate area, or the like, which appears on a radiation image (original image) of a subject. This means data, and reduced image data obtained by performing reduction processing such as averaging and thinning, smoothed image data obtained by performing smoothing processing, and the like can be used. The high-resolution image data may be image data of the original image itself, or image data having a resolution lower than that of the original image data as long as the image data has a resolution necessary for the malignancy evaluation by the malignancy evaluation means. May be used.

The abnormal shadow candidate area is an area detected from the radiation image of the subject by iris filter processing or the like used in the candidate area detecting means and becomes a candidate for a tumor shadow. Not only for shading,
This means that the image includes a shadow of a normal tissue erroneously detected by the above processing because the feature on the image shows the same feature as the tumor shadow, and also includes a benign tumor shadow.

The malignancy evaluation means for evaluating the malignancy of the candidate area may be a method of evaluating the malignancy of the candidate area based on the features of the fine structure such as the inside, shape, and edges of the candidate area. Any method may be used.For example, a method of evaluating based on a feature amount simply representing a feature such as a shape of a candidate region, a judgment method based on a Mahalanobis distance that is evaluated by combining the feature amounts, a judgment method by a neural network, and the like. Can be used.

The abnormal shadow candidate detecting device of the present invention
This is particularly effective when the abnormal shadow is a tumor shadow and when the radiographic image is a breast radiographic image.

Further, a process for acquiring low-resolution image data from radiation image data representing a radiation image of a subject, and a process for detecting a candidate area for an abnormal shadow in a radiation image based on the acquired low-resolution image data. A computer-readable recording program for causing a computer to execute a process of acquiring high-resolution image data from radiation image data and a process of evaluating the degree of malignancy of a candidate region based on the acquired high-resolution image data It may be provided by recording it on a medium.

[0021]

According to the abnormal shadow candidate detecting apparatus of the present invention configured as described above, an abnormal shadow in a radiographic image is detected based on low-resolution image data obtained from radiographic image data representing a radiographic image of a subject. Since the candidate area is detected, it is possible to use an image in which fine structures and noise such as mammary glands and blood vessels, or irregularities such as margins of the candidate area have been removed or reduced in the candidate area detection processing. And a global density gradient vector can be calculated. That is, since it becomes possible to detect a candidate region of an abnormal shadow based on the concentration of the global density gradient vector, it becomes possible to detect a shadow that is difficult to detect due to the influence of the fine structure and the like, This leads to improved detection processing accuracy.

In particular, since irregularities in the central part, which are characteristic of malignant tumors, are alleviated and a smooth image is obtained, the center of the tumor shadow can be accurately detected.

Further, since the malignancy of the candidate area is evaluated based on the high-resolution image data obtained from the radiation image data representing the radiation image of the subject, a detailed microstructure which is an important finding in the evaluation of the malignancy is required. This can be used to perform highly accurate malignancy evaluation. That is, it is expected that the detection performance of the finally detected abnormal shadow candidate is improved.

[0024]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of an abnormal shadow candidate detecting apparatus according to the present invention will be described below with reference to the drawings. FIG.
It is a figure showing the concrete embodiment of the abnormal shadow candidate detecting device in the present invention.

The abnormal shadow candidate detecting device according to the present embodiment is a low-resolution image obtaining means 10 for inputting original image data P from an image reading device or the like and obtaining low-resolution image data P '.
A candidate area detecting means 20 for inputting the low-resolution image data P 'from the low-resolution image acquiring means 10 to detect a candidate area of an abnormal shadow (hereinafter simply referred to as a candidate area for simplicity) in the image; A high-resolution image obtaining means 30 for inputting data P to obtain high-resolution image data P ″, and position data of a candidate area on a low-resolution image from a candidate area detecting means 20 and a high-resolution image obtaining means 30 And a malignancy evaluation means 40 for inputting the resolution image data P "and evaluating the malignancy of the candidate area based on the input high-resolution image data P" and the position data of the candidate area on the low-resolution image. The position data on the low-resolution image is used after being converted from the resolution ratio to the coordinate system of the high-resolution image.

Next, the operation of the abnormal shadow candidate detecting device according to this embodiment configured as described above will be described.

The original image data P of the mammography (see FIG. 2A), which is a radiographic image of the breast, is input to the low-resolution image obtaining means 10 and the high-resolution image obtaining means 30 from an image reading device or the like.

The low-resolution image acquiring means 10 acquires low-resolution image data P 'based on the input original image data P. Various methods such as averaging and thinning can be considered as a method for acquiring the low-resolution image data P '.
/ 8 reduction processing is employed. That is, for example, when the number of pixels of the original image data P is 256 × 256 pixels,
The image is divided every 8 × 8 pixels, and the image signal values of each of the divided image parts are averaged to newly obtain low-resolution image data P ′ of 32 × 32 pixels.

The high-resolution image obtaining means 30 obtains high-resolution image data P ″ based on the input original image data P. Here, 1 is obtained by averaging the original image data P.
By performing / 4 reduction processing, high-resolution image data P ″ of 64 × 64 pixels is obtained.

The candidate area detecting means 20 receives the low-resolution image data P 'from the low-resolution image acquiring means 10 and performs an iris filter process on the low-resolution image data P' to detect a candidate area of a tumor shadow. .

For example, in a radiographic image (an image represented by a high-density high-signal level image signal) on an X-ray film, it is known that the density value of the tumor shadow portion is slightly lower than that of the surrounding image portion. Therefore, inside the tumor shadow, a gradient of the density value is seen in which the density value decreases from the periphery of the substantially circular shape toward the center.
Therefore, in the tumor shadow, a local gradient of the density value is recognized, and the gradient line is concentrated toward the center of the tumor.

The iris filter calculates the gradient of the image signal represented by the density value as a gradient vector, and outputs the degree of concentration of the gradient vector. Based on this, a candidate region of a tumor shadow is detected.

Hereinafter, description will be made with reference to the mammography shown in FIG. Gradient vector at an arbitrary pixel inside the tumor pattern P ₁ in the original image data P is directed to the vicinity of the center of masses shadow P ₁ as shown in FIG. 2 (2), an elongated shadow P ₂ such as a blood vessel shadows and mammary gland 2 As shown in (3), the gradient vector does not concentrate on a specific point. Therefore, to evaluate the distribution of orientation of locally gradient vector, by detecting the area are concentrated in a specific point, it is the tumor pattern P ₁ in the candidate region. Also, shading P ₃ which is elongated shadows between the mammary gland and the like, as shown in FIG. 2 (4) crossed can be detected as a pseudo-candidate region because there is a tendency that the gradient vector is concentrated on a specific point.

A specific algorithm of the iris filter processing will be described below.

First, for all the pixels constituting the target image, the direction θ of the gradient vector of the image data is obtained for each pixel j based on the calculation formula shown in the following equation (2).

[0036]

(Equation 2) Here, as shown in FIG. 3, f _{1 to} f ₁₆ are pixel values (image data) corresponding to pixels on the outer periphery of a mask having a size of, for example, 5 pixels × 5 pixels around the pixel j. It is.

Next, with respect to all the pixels constituting the target image, the degree of concentration C of the gradient vector using the pixel as the target pixel is calculated according to the following equation (3).

[0038]

(Equation 3) Here, N is the number of pixels existing in a circle having a radius R around the pixel of interest, θj is a straight line connecting the pixel of interest and each pixel j in the circle, and the above equation (2) for each pixel j This is the angle formed by the calculated gradient vector (see FIG. 4). Therefore, the degree of concentration C represented by the above equation (3) becomes large when the direction of the gradient vector of each pixel j is concentrated on the target pixel.

By the way, the gradient vector of each pixel j in the vicinity of the tumor shadow is determined regardless of the contrast of the tumor shadow.
Since the pixel of interest is directed substantially toward the center of the tumor shadow, the pixel of interest in which the degree of concentration C takes a large value can be said to be the pixel at the center of the tumor shadow. On the other hand, the value of the degree of concentration C is small in an elongated shadow such as a blood vessel shadow because the direction of the gradient vector is biased in a certain direction. Therefore, for all the pixels that make up the image,
The value of the degree of concentration C for each pixel of interest is calculated,
By evaluating whether or not the value of the degree of concentration C exceeds a preset threshold, a tumor shadow can be detected. That is, this filter is less affected by blood vessels, mammary glands, and the like than a normal difference filter, and has a feature that a tumor shadow can be detected efficiently.

Further, in the actual processing, in order to achieve a detection power independent of the size and shape of the tumor, a method of adaptively changing the size and shape of the filter is used. FIG. 5 shows the filter. This filter is different from the filter shown in FIG.
The above-mentioned degree of concentration is evaluated only by pixels on a radial line in M types of directions (in FIG. 5, 32 directions every 11.25 degrees) adjacent at an angular interval of degrees.

Here, the coordinates ([x], [y]) of the n-th pixel on the i-th line and from the pixel of interest are given by the following equation if the coordinates of the pixel of interest are (k, l). (4), given by (5).

[0042]

(Equation 4) Here, [x] and [y] are the maximum integers not exceeding x and y.

Further, the output value up to the pixel at which the maximum concentration is obtained for each line on the line on the radiation is defined as the concentration Cimax in that direction, and the concentration Cimax is averaged in all directions. Let the value be the degree of concentration C of the gradient vector group for the pixel of interest.

Specifically, first, the degree of concentration Ci (n) obtained from the pixel of interest to the n-th pixel on the i-th radial line is obtained by the following equation (6).

[0045]

(Equation 5) That is, in equation (6), the starting point is set as the target pixel and the ending point is set as Rmi.
n within the range from n to Rmax,
(N) is calculated.

Here, Rmin and Rmax are the minimum and maximum values of the radius of the tumor shadow to be detected.

Next, the concentration C of the gradient vector group is calculated by the following equations (7) and (8).

[0048]

(Equation 6) Here, Cimax in Expression (7) is the maximum value of the concentration Ci (n) for each radial direction line obtained in Expression (6).
The area from the pixel of interest to the pixel at which the degree of concentration Ci (n) is the maximum value is the area of the tumor shadow in the direction of the line.

Equation (7) is calculated for all radial direction lines to determine the edge points of the tumor shadow area on each line, and the adjacent edge points of the tumor shadow area on each line are calculated as follows: By connecting with a straight line or a non-linear curve, it is possible to specify the contour of a candidate area that can be a candidate for a tumor shadow.

In the equation (8), the maximum value Cimax of the degree of concentration given by the equation (7) in this area is calculated in all directions of the radial direction line (the equation (8) exemplifies the case of 32 directions). An average value is determined for. The obtained value is an iris filter output value I, and this output value I is compared with a predetermined threshold value T suitable for determining whether or not the shadow is a tumor shadow, and I ≧ T (or If I> T), a region centered on the target pixel is a candidate region, and I <T
(Or ≦ T), it is determined that the area is not a candidate area,
Detect candidate areas.

The calculation of the degree of concentration Ci (n) is as follows.
The following equation (6 ′) may be used instead of equation (6).

[0052]

(Equation 7) That is, equation (6 ') shows that the concentration corresponding to the minimum value Rmin of the radius of the tumor shadow to be detected is set as the starting point, and the concentration point Ci (n) is set within the range from the end point Rmin to Rmax.
Is calculated.

Here, the candidate area detecting means 20 obtains position data indicating the position of the detected candidate area on the low resolution image from the low resolution image data P '. That is, when there are a plurality of candidate areas, the position data is acquired for each candidate area.

The malignancy evaluation means 40 inputs the high-resolution image data P ″ from the high-resolution image acquisition means 30 and further inputs the position data indicating the position of the candidate area on the low-resolution image from the candidate area detection means 20. Then, the degree of malignancy of the candidate area is evaluated based on the high-resolution image data P ″. The details will be described below.

First, the feature amount of the candidate area is calculated. Here, the malignancy evaluation unit 40 converts the position data of the candidate area on the low-resolution image input from the candidate area detection unit 20 into a coordinate system of a high-resolution image based on the resolution ratio, and Get location data. That is, based on the high-resolution image data P ″ input from the high-resolution image obtaining means 30 while referring to the position data on the high-resolution image obtained by being converted into the coordinate system of the high-resolution image, Calculate the feature values of the inside and the edge.

As the first feature amount of the candidate area, the circularity Sp is used because the outline of the tumor shadow is a shape close to a circle. As shown in FIG. 6, the area A of the candidate region and its center of gravity AO are obtained, and a virtual circle having a radius R centered on the center of gravity AO and having the same area as the area A is assumed, and candidates included in the virtual circle are included. The circularity is calculated as the occupation ratio of the area to the area A. That is, assuming that the area of the portion where the virtual circle and the candidate area overlap is A ', the circularity is calculated by the following equation (9).

[0057]

(Equation 8) Next, the following three feature values are used as feature values inside the candidate area. That is, a histogram of the density value S of the candidate area is created, and the frequency of the density value S is defined as P (S), and the second feature value (10) representing the variance var and the third feature representing the contrast con are obtained from the following equation. Features (11), angular moment asm
Is calculated.

[0058]

(Equation 9) In addition, IFED (Iris Fil
ter Edge) Five feature amounts calculated from the image are used. Hereinafter, the calculation method will be described in detail.

The candidate area detected by the iris filter processing, that is, the tumor shadow P representing the breast cancer in the radiation image
About ₁ and pseudo-abnormal pattern P _3, it detects an image portion including the vicinity thereof, for example, as a square area, creating a peripheral edge image using the iris filter processing for the detected square area (IFED image).

That is, in equation (7) of the iris filter processing, the position of the point giving the maximum value of the concentration Ci (n) on the i-th line extending radially from the pixel of interest is obtained. However, this process does not limit the value of n that gives the maximum value.

As a result, if the pixel of interest is the candidate area P ₁ or P ₃
When n is within the range, n when the equation (7) takes the maximum value
Indicates the pixel to which the i-th line intersects the edge B of the candidate regions P ₁ and P _3. For example, as shown in FIG. 7, pixels B1, B2, B3, and B4 are designated for the target pixel 1, and pixels B2, B5, B6, and B7 are provided for the target pixel 2.
Instruct.

On the other hand, when the target pixel is outside the candidate areas P ₁ and P ₃ , the maximum value of the equation (7) is obtained when the target pixel itself is specified. That is, in FIG. 7, for the target pixel 3 outside the candidate regions P ₁ and P ₃ , the value of Expression (7) becomes the maximum when the target pixel 3 itself is specified.

As described above, all the pixels in the square area including the candidate area are sequentially set as the target pixel, and the pixel having the maximum value of the equation (7) is counted. This is illustrated schematically in FIG.

That is, the count values for pixels outside the candidate areas P ₁ and P ₃ are all “1”, and the count values for pixels inside the candidate areas P ₁ and P ₃ are all “0”. count value for the pixels on the edge B of the area P ₁ and P ₃ is an image composed of all value of 1 or more is obtained. The image of this count value is defined as an IFED image.

Next, the following processing is performed on this IFED image to create a simultaneous generation matrix.

That is, as shown in FIG.
The center of gravity AO of ₁ and P ₃ is obtained, a radial line is extended from the center of gravity AO, an arbitrary point on this line is set as i, and a point perpendicular to this line and separated from the point i by two pixels is defined as j.

The count value in the IFED image at point i and the count value at point j are counted up in a matrix as shown in FIG. Specifically, when the point i is outside the candidate areas P ₁ and P ₃ , the count value of the point i in the IFED image is “1”, and at that time the point j is also in the candidate areas P ₁ and P ₃ . If it is outside, the count value at point j is also “1”. In this case, in the matrix of FIG. 10, “1” is counted in the column where “1” in the vertical direction i and “1” in the horizontal direction j intersect. Is done.

On the other hand, the point i is the candidate area P₁ , P_Three Inside
Yes, and point j is also a candidate area P₁ , P _Three If inside
Since the count value is 0 at both the i point and the j point,
In the column where “0” in direction i and “0” in horizontal direction j intersect
Count "1".

Further, the point i is the candidate area P₁ , P_Three Margin of
B, j point is also a candidate area P₁ , P _Three On the edge B of
If the count value at point i is “5” and the count value at point j is
When the event value is “3”, “5” in the vertical direction i and the horizontal direction
“1” is counted in the column where “3” in the direction j intersects.
The count value that counts up in this matrix is cumulative
Is what you do. That is, the count value is “5” again.
When scanning the i-point and the j-point whose count value is “3”,
“5” in the vertical direction i and “3” in the horizontal direction j of the matrix
"2" which added "1" to the original "1" in the crossing column
Is stored.

Since the point i is an arbitrary point in the IFED image, the matrix is completed by scanning a radial line so that all the pixels of the IFED image become the point i and scanning the point i on the line. Let it. The matrix of the IFED image is called a simultaneous generation matrix Pg (x, y).

Here, when the candidate area is a tumor shadow, the shape characteristics of the tumor shadow that the periphery of the tumor shadow is substantially circular, and the fact that the points i and j are extremely close to each other are considered. , I point is on the periphery (the count value of the IFED image has a large value of 1 or more), and the j point may be on the periphery (the count value of the IFED image has a large value of 1 or more). Is extremely high.

On the other hand, when the candidate region is a pseudo abnormal shadow, it is extremely rare that the pseudo abnormal shadow has a circular edge like the above-mentioned intersection of two blood vessels.
Even if the point and the point j are close to each other, just because the point i is on the edge does not necessarily mean that the point j is also on the edge. Rather, the possibility that the point j is on the edge is extremely low. Become.

Therefore, the simultaneous generation matrix Pg (x, y)
Is significantly different depending on whether the candidate area is a tumor shadow or a pseudo abnormal shadow. The characteristic value of this simultaneous generation matrix is edge information, and this edge information is used as a feature value. That is, from the following expression, the fifth feature value (1
3), a sixth feature (14) representing the deviation dfe (difference entropy), a seventh feature (15) representing the correlation value cor (correlation), and a moment idm (inverse difference m).
oment), the eighth feature (16), entropy se
A ninth feature (17) representing (sum entropy) is obtained.

[0074]

(Equation 10) Note that the malignancy evaluation means 40 calculates the above nine feature amounts in association with the position data on the high-resolution image. That is, when a plurality of candidate areas are detected by the candidate area detecting means 20, the position data of each candidate area on the low-resolution image is input, and each is converted into position data on the high-resolution image. Nine feature values are calculated for each region in association with the position data on the high-resolution image.

Next, the malignancy evaluation means 40 evaluates the malignancy of the candidate area by using the calculated nine feature amounts. In the present embodiment, a determination method based on the Mahalanobis distance is used as an example of malignancy evaluation.

First, according to the following equation (1), a pattern class (i =
The Mahalanobis distance Dm1 from 1) and the Mahalanobis distance Dm2 from the pattern class (i = 2) indicating the malignant shadow are calculated.

[0077]

[Equation 11] The nine feature amounts correspond to x1 to x9 in the above equation (1), and represent a nine-dimensional space of (x1, x2, x3,..., X9). The Mahalanobis distance between the pattern of the candidate area expressed on the nine-dimensional pattern space and the pattern of the non-malignant shadow is Dm1, and the Mahalanobis distance of the pattern of the malignant shadow is Dm2.

Here, the non-malignant shadow pattern and the malignant shadow pattern are defined as a vector x for each non-malignant shadow and each malignant shadow, which is set in advance based on the result of an experimental investigation of a large number of abnormal shadow candidates. Means the defined pattern space. For example, a pattern class w1 formed by an average of the above-mentioned vector x for a non-malignant shadow, and a pattern class w formed by an average of the above-mentioned vector x for a non-malignant shadow, respectively.
Indicated by 2.

Next, when the candidate area is a malignant shadow, the Mahalanobis distance with the pattern class of the malignant shadow is short (Dm2 shows a low value), and the Mahalanobis distance with the pattern class of the non-malignant shadow varies. When the candidate area is a non-malignant shadow, the Mahalanobis distance with the pattern class of the non-malignant shadow is short (Dm1 shows a low value), and the Mahalanobis distance with the pattern class of the malignant shadow varies. Therefore, according to this tendency, a likelihood ratio that can significantly distinguish a malignant shadow from a non-malignant shadow is calculated for each candidate region.

The likelihood ratio is represented by Dm1 / Dm2, and is shown in FIG.
1 shows an inclination on a coordinate plane. That is, it can be evaluated that the larger the likelihood ratio, the higher the possibility of a malignant shadow, and the smaller the likelihood ratio, the higher the possibility of a non-malignant shadow. Therefore, for example, a threshold value is set to 2, and when the likelihood ratio is 2 or more, Is determined to be non-malignant when less than 2.

The malignancy evaluation means 40 acquires the result of the evaluation in association with the position data of the candidate area on the high-resolution image. When it is desired to obtain the position data of the candidate area on the original image, based on the position data on the low-resolution image or the high-resolution image, each of the resolution ratios is converted into the coordinate system of the original image, and the original image is easily converted. The upper position data can be obtained.

According to the abnormal shadow candidate detecting apparatus of the present embodiment, the candidate area is detected based on the low resolution image obtained by performing the 1/8 reduction processing on the original image. The influence of noise or the like on the detection processing can be reduced, and the detection processing of the candidate area can be performed more accurately. In addition, when evaluating the degree of malignancy of the detected candidate area, a high-resolution image obtained by performing 1/4 reduction processing on the original image, that is, a fine structure that is regarded as important in the degree of malignancy evaluation Can be used, and accurate malignancy evaluation can be performed. Furthermore, since a low-resolution image is used for the detection process and a reduced image having sufficient accuracy for the evaluation of the degree of malignancy is also used as the high-resolution image, it is possible to expect a higher processing speed.

The method of acquiring a low-resolution image and a high-resolution image in the abnormal shadow candidate detecting apparatus according to the present invention is not limited to the above-described mode in which the reduction processing is performed by averaging. Is also possible. In addition, of course, the reduction ratio can be variously changed in consideration of the image quality of the original image, the size of the tumor shadow to be detected, the processing time, and the like, and the original image may be used for the high-resolution image. For the low-resolution image, for example, a smoothed image obtained by performing a smoothing process on the original image to remove the fine structure and noise of the image may be used, that is, removing the fine structure and noise of the original image. Various images can be used.

The detection processing by the candidate area detection means is not limited to the above-described method of detecting a tumor shadow by the iris filter processing, but may be a processing of detecting an abnormal shadow by looking at the density gradient of an image globally. If there is, various use is possible.

Further, the method of evaluating the degree of malignancy and the type of the characteristic amount are not limited to the above-described embodiments, and various evaluation methods and characteristic amounts based on the fine structure of the candidate area may be adopted.

When the simultaneous generation matrix is created, the point j is defined as a point perpendicular to the radial line where the point i exists and separated from the point i by two pixels (see FIG. 9). ,
The angle of intersection with the radial line, the pixel separated from point i, and the like can be appropriately changed according to the size of the tumor shadow to be detected.

The present invention is not limited to breast CAD, but can be applied to chest CAD and the like.

[Brief description of the drawings]

FIG. 1 is a configuration diagram of an abnormal shadow candidate detection device according to an embodiment of the present invention;

FIG. 2 is a diagram illustrating an operation of an iris filter process.

FIG. 3 is a diagram illustrating a mask having a size of 5 pixels vertically and 5 pixels horizontally centered on a target pixel j;

FIG. 4 is a view for explaining an angle between a target pixel and a gradient vector at each pixel j.

FIG. 5 is a conceptual diagram showing an iris filter set so that a contour shape is adaptively changed.

FIG. 6 is a diagram showing a virtual circle having an area equal to the area A of a candidate area;

FIG. 7 is a diagram illustrating an operation of forming an IFED image;

FIG. 8 is a conceptual diagram showing an IFED image.

FIG. 9 is a diagram illustrating an operation of creating a simultaneous generation matrix based on an IFED image;

FIG. 10 is a diagram showing a simultaneous generation matrix.

FIG. 11 is a diagram showing the operation of a judgment method based on the Mahalanobis distance.

[Explanation of symbols]

 10 Low-resolution image acquisition means 20 Candidate area detection means 30 High-resolution image acquisition means 40 Malignancy evaluation means P Original image data P 'Low-resolution image data P "High-resolution image data

Continued on the front page F term (reference) 4C093 AA01 AA26 CA21 FD05 FF07 FF17 FF18 FF28 FH07 5B057 AA07 BA03 CA02 CA08 CA12 CA16 CB02 CB08 CB12 CB16 CC01 CE03 CE05 CE06 CH09 DA07 DA08 DA17 DB02 DB05 DB09 FA06 GA55

Claims (4)

[Claims] 1. A low-resolution image acquiring unit for acquiring low-resolution image data from radiation image data representing a radiation image of a subject, and the radiation based on the low-resolution image data acquired by the low-resolution image acquiring unit. Candidate area detecting means for detecting a candidate area of an abnormal shadow in an image; high-resolution image obtaining means for obtaining high-resolution image data from the radiation image data; and the high-resolution image obtained by the high-resolution image obtaining means An abnormal shadow candidate detection device, comprising: a malignancy evaluation means for evaluating the malignancy of the candidate area based on the data. 2. The abnormal shadow candidate detecting device according to claim 1, wherein the abnormal shadow is a tumor shadow. 3. The abnormal shadow candidate detecting device according to claim 1, wherein the radiation image is a breast radiation image. 4. A process for acquiring low-resolution image data from radiation image data representing a radiation image of a subject, and detecting an abnormal shadow candidate region in the radiation image based on the acquired low-resolution image data. A program for causing a computer to execute a process, a process of acquiring high-resolution image data from the radiation image data, and a process of evaluating the degree of malignancy of the candidate region based on the acquired high-resolution image data. A recorded computer-readable recording medium.