JP2001008923A - Method and device for detecting abnormal shade - Google Patents

Abstract

PROBLEM TO BE SOLVED: To change a detection level according to a detection region a method for detecting abnormal shade with an iris filter treatment or the like. SOLUTION: When a closed area which obtaines an iris filter output value I is existed at an area corresponding to an outer upper areas for breasts RC, RF, i.e., an outer area RA or an upper area RC which is defined by an area regulating means, a detectability changing means 30 multiplies an extra coefficient α (=1.2) which is set in advance on the iris filter output value I (I'=Iα). If a compared value between the output value I' after when the extra value is multiplied and a threshold T is I'>T, an abnormal shade detecting means 20 detects the closed area which obtains the output value I' as a tumor shadow candidate PJ.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and apparatus for detecting abnormal shadows such as tumor shadows and microcalcification shadows in a radiographic image, and more particularly to an improvement in the accuracy of detecting abnormal shadows. .

[0002]

2. Description of the Related Art Conventionally, in the medical field, it has been common to detect a diseased part using a radiographic image of a subject and observe the state of the diseased part to diagnose the presence or absence of a disease and the progress of the disease. It is being done. Conventionally, the interpretation of the radiographic image is influenced by the observer's experience and the level of the ability to interpret the image, and is not always objective.

[0003] For example, in mammography (or mammogram: a diagnostic radiation image of a breast as a subject) taken for the purpose of examining breast cancer, the image shows a tumor shadow or microcalcification which is one of the features of a cancerous part. Although it is necessary to extract the shadow, the range of the tumor shadow cannot always be specified accurately. Therefore, it is required to accurately detect abnormal shadows such as tumor shadows and microcalcification shadows without depending on the skill of the observer.

In response to this demand, Computer Aided Diagnosis (CADM; Computer Aided Diagnosis) which automatically detects abnormal shadow candidates using computer processing.
f Medical images) has recently been researched.

[0005] That is, the CADM technique uses a computer to perform enormous detection processing based on the density distribution characteristics and morphological characteristics of abnormal shadows such as the above-mentioned tumor shadows and microcalcification shadows. The abnormal shadow candidate detection process is mainly suitable for detecting a tumor shadow candidate, and is suitable mainly for detecting microcalcification shadow candidates. Morphological filter processing etc. (Obata et al. "Detection of tumor shadow in DR image (iris filter)" IEICE Transactions on Information and Communication D-II Vol.J75-
D-II No.3 P663-670 March 1992, "Extraction of microcalcification image by morphological filter using multiple structural elements" IEICE Transactions D-II Vol.J75-D-II No.7
P1170-1176 July 1992, "Basics of Morphology and Its Application to Mammogram Processing" MEDICAL IMAGING TECHNO
LOGY Vol.12 No.1 January 1994 etc.).

[0006] The iris filter processing 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 to obtain a tumor, which is one of the characteristic forms of breast cancer in the image. It has been studied as an effective method for detecting shadow candidates. However, the target image is not limited to such a tumor shadow in mammography, and gradients of image signals (density, etc.) representing the image are concentrated. Can be applied to any image part.

On the other hand, the morphology filter process compares an output value of a morphology operation process using a structural element having a size larger than a microcalcification shadow to be detected with a predetermined threshold value for an image signal. It has been studied as an effective method for detecting the microcalcification shadow candidate, which is one of the characteristic forms of breast cancer in an image.

Further, in order to exclude those abnormal shadows which are erroneously detected from among the abnormal shadow candidates obtained by the above-mentioned iris filter processing, the abnormal shadows are further classified into a malignant abnormal shadow and a benign abnormal shadow or a normal shadow. By evaluating a plurality of feature values that can be significantly distinguished from each other by using an evaluation value (Maharanobis distance, etc.) for evaluating malignancy, which is associated with a predetermined function, a more probable abnormal shadow candidate that is a malignant tumor is accurately determined. It can be detected well.

[0009]

However, according to the various methods for detecting abnormal shadows proposed above, there is a possibility that omission of detection of a true abnormal shadow that should be originally detected may occur. It is necessary to prevent omission of detection for abnormal shadows that are experimentally or statistically known to occur relatively frequently in the area.

SUMMARY OF THE INVENTION The present invention has been made in view of the above circumstances, and has as its object to provide an abnormal shadow detection method and an abnormal shadow detection device that improve the detection accuracy of abnormal shadows as compared with conventional detection methods and devices. It is assumed that.

[0011]

SUMMARY OF THE INVENTION The abnormal shadow detecting method and abnormal shadow detecting apparatus according to the present invention provide a method for detecting a tumor shadow or microcalcification shadow, which is a characteristic form of breast cancer, by the present inventors. in rather than randomly generated over, statistically, a region close to the region R _C and the arm when the outer upper (nipple region near to the R _E, as shown in FIG. 4 R
_F ) is based on the finding that it frequently occurs, and the detection accuracy of abnormal shadows is improved by changing the detectability of abnormal shadow candidates for each subject area.

That is, according to the abnormal shadow detecting method of the present invention, an abnormal shadow detecting method for detecting a candidate for an abnormal shadow in an image based on an image signal representing the radiation image of the object. The detection capability of the abnormal shadow candidate is changed for each region.

As a result, for a specific area of a subject where it is known experimentally or statistically that abnormal shadows occur relatively frequently, the detectability in the specific area is increased to enable detection. Can be increased, and omission of detection of an abnormal shadow can be prevented. On the other hand, with respect to a specific region in which no abnormal shadow is generated or the occurrence of the shadow is relatively small is experimentally or statistically known in advance, the detection process is performed by lowering the detectability in the specific region. Thus, detection of a pseudo abnormal shadow can be suppressed as much as possible.

The detection ability (detection level) means the detection ability, and can detect an abnormal shadow which has not been detected conventionally and / or a normal shadow which has been erroneously detected conventionally. Etc. means the ability to prevent detection.

[0015] The change of the detectability can be achieved by detecting the abnormal shadow candidate more preferentially than other regions in an area of the subject in which there is a high possibility that the abnormal shadow candidate is present experimentally or statistically. Often, it is possible to increase the possibility of detection in an area of a subject in which there is a high possibility that an abnormal shadow candidate exists.

For example, when the radiation image is mammography, the abnormal shadow candidate is detected in a region corresponding to the outer side or the upper part of the breast as the subject in the mammography more than the area corresponding to the inner side or the lower part. In this way, it is desirable to relatively increase the detection ability. In mammography, abnormal shadows occur more specifically in the area corresponding to the outer or upper part of the breast than in the area corresponding to the inner or lower part. This is because the possibility can be increased.

The outer and upper parts indicate relative positional relationships when the human body is upright, respectively.
For the left breast shown in FIG. 4, the outside is the region _RC ,
R _B and R _F are applicable, upper, region R _D, R _C and R _F corresponds.

As a specific method of relatively increasing the detectability, (1) an iris filter output value representing the maximum value of the concentration gradient of the density gradient of the image signal for each pixel of the radiation image, a predetermined threshold value, When detecting a tumor shadow candidate that is the abnormal shadow candidate by iris filter processing to compare the iris filter output value in a direction to relatively increase the iris filter output value for the region corresponding to the outside or the region corresponding to the upper portion Weighting and / or weighting in a direction to relatively lower the threshold for the region corresponding to the outside or the region corresponding to the upper side, (2) the image signal for each pixel of the radiation image Of the morphological operation using a structural element having a size larger than the abnormal shadow candidate to be detected. When detecting a microcalcification shadow candidate that is the abnormal shadow candidate by a morphology filter process that compares a force value with a predetermined threshold, the morphology calculation is performed on the region corresponding to the outside or the region corresponding to the upper side. A method of weighting in a direction of relatively increasing the output value, and / or a method of weighting in a direction of relatively lowering the threshold value for the region corresponding to the outside or the region corresponding to the upper side, or (3) A) detecting an abnormal shadow candidate as a tumor shadow candidate by performing an iris filter process of comparing an iris filter output value representing a maximum value of a concentration degree of a density gradient of the image signal with respect to each pixel of the radiation image with a predetermined threshold value; Performing a predetermined malignancy evaluation value for the detected tumor shadow candidate, When performing detection of a more probable tumor shadow candidate based on the evaluation value from among the shadow candidates, weighting is performed in a direction to relatively increase the evaluation value for the region corresponding to the outside or the region corresponding to the upper side. It is preferable to apply a method.

Here, the iris filter output value and the threshold value to be weighted as described in the above (1) will be described together with the outline of the iris filter processing and the detection processing of a tumor shadow candidate as an example.

For example, in a radiographic image (an image represented by a high-density high-signal level image signal) on an X-ray film, the tumor shadow has a slightly lower density value (image signal value) than the surrounding image portion. It is known that the density value distribution has a density value gradient such that the density value decreases from the substantially circular peripheral portion 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. It is to detect a tumor shadow on the basis.

[0022] That is, for example, FIG. 2 (1) of predetermined irradiation field stop (P _S is squeezed for representing the contour) breast P _M with
While a captured mammogram P, the gradient vector at an arbitrary pixel in the emerged tumor pattern P _J of this mammography P is directed to the vicinity of the center of masses shadow P _J as shown in FIG. (2), vascular shadow and since the gradient vector as shown in the elongated shadow P _K Fig (3) as such mammary gland does not concentrate on a specific point, to assess the distribution of orientation of locally gradient vector, a particular point be extracted regions are concentrated in, it is the tumor pattern P _J. In addition, as for the shadow P _L where elongated shadows such as the mammary gland intersect each other as shown in FIG. 4D, the gradient vector tends to concentrate on a specific point, and may be a pseudo abnormal shadow.

The above is the basic concept of the iris filter processing. The specific algorithm steps are shown below.

(Step 1) Calculation of Gradient Vector For all the pixels constituting the target image, the direction θ of the gradient vector of the image data is calculated for each pixel j based on the following formula (1). Ask.

[0025]

(Equation 1)

Here, f _{1 to} f ₁₆ are, as shown in FIG.
This is a pixel value (image signal value) corresponding to a pixel on the outer periphery of the mask having a size of, for example, 5 pixels by 5 pixels around the pixel j.

(Step 2) Calculation of Concentration of Gradient Vector Next, for all the pixels constituting the target image,
For each pixel, the degree of concentration C of the gradient vector with that pixel as the pixel of interest is calculated according to the following equation (2).

[0028]

(Equation 2)

Here, N is the number of pixels existing in a circle having a radius R around the pixel of interest, and θj is a straight line connecting the pixel of interest and each pixel j in the circle, and the above equation for each pixel j. The angle formed by the gradient vector calculated in (1) (see FIG. 6). Therefore, the concentration C expressed by the above equation (2)
Has a large value 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 having a large value of the degree of concentration C can be said to be the pixel at the center of the tumor shadow. On the other hand, in the shadow of a linear pattern such as a blood vessel, the value of the degree of concentration C is small because the direction of the gradient vector is biased in a certain direction. Therefore, the value of the degree of concentration C with respect to the pixel of interest is calculated for each of the pixels constituting the image, and whether or not the value of the degree of concentration C exceeds a preset threshold T is evaluated. Can be detected. That is, this filter has a feature that it is less affected by blood vessels and mammary glands than a normal difference filter, and can efficiently detect a tumor shadow.

Further, in actual processing, in order to achieve a detection power which is not affected by the size or shape of the tumor, a device for adaptively changing the size and shape of the filter is used. FIG. 7 shows the filter. This filter is
Unlike the one shown in FIG. 6, 2π /
The above-mentioned degree of concentration is evaluated only by pixels on radial direction lines in M types of directions (in FIG. 7, 32 directions every 11.25 degrees) adjacent at an angular interval of M degrees.

Here, the coordinates ([x], [y]) of the n-th pixel on the i-th direction line and from the pixel of interest are
Assuming that the coordinates of the pixel of interest are (k, l), the coordinates are given by the following equations (3) and (4).

[0033]

(Equation 3)

Here, [x] and [y] are the maximum integers not exceeding x and y.

Further, the output value up to the pixel where the maximum degree of concentration is obtained for each direction line on the radial direction line is defined as the degree of concentration Cimax in that direction, and the degree of concentration Cimax is averaged for all the direction lines. This average value is defined as the degree of concentration C of the gradient vector group for the target pixel.

More specifically, first, the degree of concentration Ci (n) obtained from the target pixel to the n-th pixel on the i-th radial line is obtained by the following equation (5).

[0037]

(Equation 4)

That is, equation (5) calculates the degree of concentration Ci (n) by changing the starting point within the range of the pixel of interest and the ending point within the range from Rmin to Rmax.

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

Next, the concentration C of the gradient vector group is calculated by the following equations (6) and (7).

[0041]

(Equation 5)

Here, Cimax in equation (6) is the maximum value of the degree of concentration Ci (n) for each radial direction line obtained in equation (5). The region up to the pixel where is the maximum value is the candidate region for the tumor shadow in the direction of the direction line.

Equation (6) is calculated for all the radial direction lines to determine the outline (edge point) of the region of the tumor shadow on each direction line, and the outline of the region of the tumor shadow on each direction line is determined. By connecting the edge points to be performed with a straight line or a non-linear curve, it is possible to specify the outline of a region that can be a candidate for a tumor shadow.

In equation (7), the maximum value Cimax of the degree of concentration given by equation (6) in this area is calculated in all directions of the radial direction line (expression (7) 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. The region centered on the pixel of interest is an abnormal shadow candidate (tumor shadow candidate) and I <T
If so, it is determined that the candidate is not a tumor shadow candidate.

Here, in the present invention, when the target pixel exists in the region outside or above the breast, the iris filter output value I is weighted in a direction to make it relatively high, and / or the threshold value T is made relatively high. By weighting in the direction of lowering the target pixel, the detectability is relatively increased as compared with the case where the target pixel exists in another region. This allows
In a region corresponding to the outside or upper part of the breast where a large number of tumor shadows are specifically generated, the possibility of detecting the tumor shadow can be increased more than in other regions.

It should be noted that the weighting of the output value I is set to 1
For example, multiplication by a value α (α> 1) exceeding (Iα) or adding a positive value β (β> 0) (I + β) can be applied. The values of α and β are obtained statistically or experimentally, but it is preferable to set values that increase the iris filter output value I by about 5 to 30%. On the other hand, when weighting in the direction of relatively lowering the threshold T while leaving the iris filter output value I as it is, the threshold T is multiplied by a positive value α ′ (0 <α ′ <1) smaller than 1 (Iα ′). Or a process of adding a negative value β ′ (β ′ <0) (I + β ′) can be applied. Also, both weighting may be performed to relatively lower the threshold value while relatively increasing the iris filter output value I.

The calculation of the above-mentioned concentration degree Ci (n) is expressed by the following equation (5).
May be used instead of the following equation (5 ′).

[0048]

(Equation 6)

That is, equation (5 ') calculates the degree of concentration Ci (n) within the range from the pixel corresponding to the minimum value Rmin of the radius of the tumor shadow to be extracted to the start point and the end point from Rmin to Rmax. Is what you do.

Next, regarding the morphological (Morphology) operation output value and the threshold value to be weighted, which are described in (2) above, the outline of the morphological filter processing and the detection processing of the microcalcification shadow candidate Will be described as an example.

The morphological filter processing is a technique for detecting a candidate for a microcalcification image which is a characteristic form in breast cancer together with a tumor shadow, and uses a multi-scale λ and a structural element (mask) B to obtain (A) lime. It has the following features: (1) it is effective for extracting the calcified image itself; (B) it is hardly affected by complicated background information; and (C) the extracted calcified image is not distorted. In other words, this method is different from general differential processing in that the size, shape,
It is possible to perform detection while better maintaining geometric information such as a density distribution.

(Basic Operation of Morphological Filter) Morphology is generally developed as a set theory in an N-dimensional space. However, for intuitive understanding, a description will be given of a two-dimensional gray image.

The point of coordinates (x, y) is represented by a density value f
Consider a space having a height corresponding to (x, y), and consider a one-dimensional function f (x) corresponding to this cross section. The structuring element g used in the morphological operation processing is a symmetric function symmetric about the origin as shown in the following equation (8).

[0054]

(Equation 7)

It is assumed that the value is 0 in the domain and the domain G is represented by the following equation (9).

[0056]

(Equation 8)

At this time, the basic form of the morphological operation is a very simple operation as shown in equations (10) to (13).

[0058]

(Equation 9)

That is, the dilation processing is performed in a range of ± m (a value determined according to the structural element B and corresponding to the mask size in FIG. 8) around the target pixel. (A) in the figure, while erosion processing is a processing for searching for a minimum value within a range of ± m around the pixel of interest. (See FIG. 2B). The opening process is a process of performing a dilation process after the erosion process, that is, a process of searching for a maximum value after searching for a minimum value.
ing) The process corresponds to a process of performing an erosion process after a dilation process, that is, a process of searching for a minimum value after searching for a maximum value.

That is, in the opening process, the density curve f (x) is smoothed from the low density side, and a convex density variation portion (a portion having a higher density than the surrounding portion) which fluctuates in a space narrower than the mask size of 2 m. (See FIG. 3C).

On the other hand, in the closing processing, the density curve f (x) is smoothed from the high density side, and a concave density variation portion (a portion having a density lower than the surrounding portion) that fluctuates in a space narrower than the mask size 2 m. (See FIG. 3D).

Here, in the case of a signal having a high density and a high signal level, the higher the density, the higher the density, the higher the density value f
Since the magnitude relationship is reversed with respect to the case where the image signal value of (x) is at the high luminance and high signal level, the dilation processing for the high density and high signal level signal and the erosion processing for the high luminance and high signal level (FIG. )), The erosion process for the high-density high-signal level signal and the dilation process for the high-brightness high-signal level (FIG. 10A) match the opening process for the high-density high-signal level signal. The closing process for a high-brightness high-signal level and the opening process for a high-brightness / high-signal level (FIG. 10C) match with the closing process for the high-brightness / high-signal level. .

In this section, the case of an image signal (luminance value) at a high luminance and high signal level will be described.

(Application to detection of microcalcification shadow) Detection of microcalcification shadow is based on the following formula (1) based on the opening process.
4) applies.

[0065]

(Equation 10)

Here, Bi (i = 1, 2, 3, 4) corresponds to FIG.
Are four linear structural elements B shown in FIG. If the structural element B is set to be larger than the microcalcification shadow to be detected, the lime that is a convex signal change portion (an image portion that fluctuates in a narrow spatial range) finer than the structural element B in the opening process. The image is removed. On the other hand, if the length of the elongated non-calcified shadow is longer than the structural element B, and its inclination (extending direction) matches any one of the four structural elements Bi, the opening process (the equation (14)) Even if the two-term operation is performed, it remains as it is. Therefore, by subtracting the smoothed image (image from which micro-calcification shadows have been removed) obtained by the opening process from the original image f, an image containing only small calcification image candidates is obtained.

As described above, in the case of a signal having a high density and a high signal level, the calcified shadow has a lower density value than the surrounding image portion, and the calcified shadow has a concave signal change with respect to the surrounding portion. Therefore, the closing process is applied instead of the opening process, and the expression (15) is applied instead of the expression (14).

[0068]

[Equation 11]

In this way, a microcalcification shadow candidate can be detected. However, a noncalcification shadow having the same size as the microcalcification shadow is erroneously detected as a microcalcification shadow candidate. In order to remove the non-calcified image, the non-calcified image is further removed using differential information based on the morphological operation of the following equation (16).

[0070]

(Equation 12)

Here, the larger the value of Mgrad, the greater the possibility of a true microcalcification shadow, so the microcalcification candidate image Cs can be obtained by the following equation (17).

[0072]

(Equation 13)

Note that T1 and T2 are preset thresholds defined experimentally or statistically, and P (i, j)
And Mgrad are morphology operation output values.

However, non-calcified shadows different from the size of the calcified shadow can be removed only by comparing P (i, j) in the equation (14) with the threshold value T1, so that the non-calcified shadow is equivalent to the microcalcified shadow. In the case where a non-calcified shadow having a size does not remain, the condition (P (i, j) ≧ T
1) only needs to be satisfied. In this case, the threshold value is only T1, and the morphological operation output value is only P (i, j).

Here, in the present invention, when the target pixel exists in the region outside or above the breast, the morphology operation output value P (i, j) (and Mgrad) is weighted in a direction to make it relatively high. And / or threshold T
By weighting 1 (and T2) in a direction to make it relatively lower, the detectability is relatively higher than when the target pixel exists in another region. This makes it possible to increase the detectability of a tumor shadow in a region corresponding to the outside or the upper part of the breast where microcalcification shadows occur more specifically than in other regions.

The morphological operation output value P (i,
For the weighting in the direction of relatively increasing j), the above-described method of weighting in the direction of relatively increasing the output value of the iris filter can be applied, and in the direction of relatively decreasing the threshold value T1. For the weighting, a method of weighting the threshold T in the iris filter processing in a direction of relatively lowering the threshold T can be applied. In addition, the morphology operation output value is increased by about 5 to 30%, and / or
Alternatively, it is preferable to increase the detectability by reducing the threshold value by about 5 to 30%.

Next, the iris filter processing for comparing the iris filter output value representing the maximum value of the concentration degree of the density gradient of the image signal with respect to each pixel of the radiation image and a predetermined threshold value described in (3) above. Detecting a tumor shadow candidate that is the abnormal shadow candidate, obtaining an evaluation value for a predetermined malignancy evaluation for the detected tumor shadow candidate, based on the evaluation value from among the detected tumor shadow candidates, A description will be given of a method of weighting the evaluation value of the region corresponding to the outside or the region corresponding to the upper side in the direction of relatively increasing the detection value when a likely tumor shadow candidate is detected.

This method is applied to a case where a further refinement is performed to detect a tumor shadow candidate so that only a more probable tumor shadow candidate is detected from among the tumor shadow candidates detected by the iris filter processing. It is a method. That is, the iris filter can detect benign tumor shadows other than malignant tumor shadows such as breast cancer, so that only the malignant tumor shadows are accurately extracted from the malignant tumor shadows and benign tumor shadows. In this method, a significant feature amount (one or more types) between a benign tumor shadow and a benign tumor shadow is used as an evaluation value for malignancy evaluation, and only the malignant tumor shadow is detected as a tumor shadow candidate according to the evaluation value. (See JP-A-9-167238).

In determining an evaluation value based on two or more types of feature values, these two or more types of feature values may be associated with each other by a predetermined evaluation function to form one evaluation value. It may be.

As the feature amount, for example, a first index value representing the variance var represented by the histogram equation (18), a second index value representing the contrast con represented by the histogram equation (19), and the histogram At least one of the third index values representing the angular moment asm expressed by Expression (20) can be used.

[0081]

[Equation 14]

As an evaluation function that associates them into one evaluation value, for example, a discrimination function of a Mahalanobis distance can be used. The Mahalanobis distance means Dmi defined by the following equation (21), and is a distance measured from the center of the distribution by weighting a hyperellipsoid represented by a covariance matrix Σ.

[0083]

(Equation 15)

That is, when the feature amount of the malignancy evaluation for the detected tumor shadow candidate region is set to n-th order (the feature amount x
1, x2, x3, ..., xn) and Mahalanobis distance x
It is expressed in an n-dimensional space of the form (x1, x2, x3,..., Xn), and the pattern of the tumor shadow candidate expressed in this n-order pattern space is obtained experimentally or statistically in advance. The respective distances Dm1 and Dm2 between the benign tumor pattern and the malignant tumor pattern are obtained. If Dm1> Dm2, the pattern is close to the malignant tumor pattern. , And is detected as a tumor shadow candidate. If Dm1 <Dm2, the tumor shadow candidate has a pattern close to the benign tumor pattern. Not detected as a shadow candidate.

In the present invention, the evaluation value (pattern in the n-order space) obtained in this manner is used as a candidate for a tumor shadow candidate detected in a region corresponding to the outside of the breast as a subject or a region corresponding to the upper side. By weighting in the direction closer to the pattern of the malignant mass,
What is necessary is just to improve detectability. The weighting method and the degree of weighting can be applied in the same manner as in the cases (1) and (2) described above. This makes it possible to increase the detectability of the tumor shadow in a region corresponding to the outside or upper part of the breast where a large number of tumor shadows occur specifically, compared to other regions.

Instead of weighting in the direction closer to the pattern of the malignant mass, in addition to the n-th feature described above,
Whether the region where the tumor shadow candidate is detected is a region corresponding to the outside or upper side of the breast, which is a subject, is defined as one feature amount, and a pattern based on the (n + 1) -th order feature amount In the pattern of the corresponding malignant mass represented by the following formula, a feature amount indicating that the pattern exists in a region corresponding to the outside of the breast, which is the subject, or a region corresponding to the upper side may be added in advance. In this case, since the tumor shadow candidate detected in the region corresponding to the outside of the breast or the region corresponding to the upper side matches the pattern of the malignant mass in terms of the feature amount of the detected region, the Mahalanobis distance is equal to the malignant mass. This has the same effect as being weighted in a direction in which it is more likely to be detected.

The following describes how these areas serving as a basis for determining whether or not the above-mentioned pixel of interest, a detected candidate for a tumor shadow, and the like are present in an area corresponding to the outside or upper part of the breast which is the subject. For example, the arrangement of the breasts in mammography is first determined based on an image signal, and the breasts appearing on any edge of the mammography (for example, a plan view shown in FIG. CC-view: also referred to as a front view) or line connecting with a straight line K1 and the K2 in the body-side edge (FIG. (1) of P _M) in the side view shown in FIG. (2) (MLO-view) And the midpoint position (KM in FIG. 1A, KN in FIG. 1A) of the midpoint of the breast (line segment connecting K3 and K4 with a straight line in FIG. Figures (1) and (2)
, A nipple position (e.g., a nipple position) is obtained, and the midpoint position (KM, KN) and the tip position (N) are obtained.
The straight line connecting the bets, the breast P _M the two regions (region P _A and P _B in FIG. (1), region P _C and P _D in Fig. (2)) by dividing the, to the outer A corresponding area (the area P _A in FIG. _1A ) or an area corresponding to the upper portion (the area P _C in FIG. 2B) may be defined.

The arrangement of the breasts in the mammography (P1, P2) based on the image signals may be obtained by analyzing the histogram of the image signals, or may be obtained by the breast image recording disclosed in Japanese Patent Application Laid-Open No. 4-8351. A position recognition technique may be applied.

When a nipple is applied as the tip position of the breast, the nipple may be detected by utilizing the characteristic that the output value of the iris filter is large.

In the above detailed description, mainly,
Although the case where the present invention is applied to mammography with a breast as a subject has been described, the abnormal shadow detection method of the present invention is not limited to the mammography, and the target image is likely to occur in a specific region of the subject, or vice versa. As long as abnormal shadows having characteristics (experimental or statistical) such as hardly occurring in a specific area are to be detected, not only breasts but also other subjects are subject to the abnormal shadow detection method of the present invention. It becomes.

An abnormal shadow detecting apparatus according to the present invention is an apparatus for implementing the above-described abnormal shadow detecting method, and is a candidate for an abnormal shadow in an image based on an image signal representing a radiation image of a subject. The abnormal shadow detection device further comprising: an abnormal shadow detecting device for detecting the abnormal shadow candidate, for each of the regions of the subject appearing in the radiation image, The means is for detecting an abnormal shadow candidate based on the changed detection ability.

Here, the detection capability changing means may detect the abnormal shadow candidate in an area of the subject where the abnormal shadow is likely to exist experimentally or statistically more than any other area. It is preferable to apply the one that changes the detectability.

When the radiographic image is a mammogram, the detection capability changing means may be configured such that the area corresponding to the outside or the top of the breast as the subject is smaller than the area corresponding to the inside or the bottom in the mammography. What is necessary is just to make the said detection capability relatively high so that the said abnormal shadow candidate may be detected predominantly.

Here, the arrangement of the breast in the mammography is obtained based on the image signal, and the midpoint position of the body side edge of the breast appearing at any one of the edges of the mammography is obtained. Area defining means for determining a tip position and dividing the breast into two areas by a straight line connecting the midpoint position and the tip position, thereby defining an area corresponding to the outside or an area corresponding to the upper part. It is preferable that the detection capability changing unit increases the detection capability in accordance with the region defined by the region defining unit.

As the abnormal shadow detecting means, the abnormal iris filter processing is performed by comparing an iris filter output value representing the maximum value of the concentration of the density gradient of the image signal for each pixel of the radiation image with a predetermined threshold value. A method for detecting a tumor shadow candidate that is a shadow candidate is applied, and as the detection capability changing means, a direction in which the iris filter output value for an area corresponding to the outside of the breast or an area corresponding to the upper part is relatively high. And / or by weighting in the direction of relatively lowering the threshold value for the region corresponding to the outside or the region corresponding to the upper side, thereby enhancing the detectability. it can.

Further, as the abnormal shadow detecting means, the image signal of each pixel of the radiation image is
Detecting the microcalcification shadow candidate, which is the abnormal shadow candidate, by morphological filter processing that compares an output value of a morphological operation process using a structural element having a size larger than the abnormal shadow candidate to be detected with a predetermined threshold value And applying a weight to the area corresponding to the outside of the breast or the area corresponding to the upper side in a direction to relatively increase the morphological operation output value, and / or By weighting in a direction in which the threshold value for the region corresponding to the region or the region corresponding to the upper side is relatively lowered, a device that enhances the detectability can be applied.

Further, as an abnormal shadow detecting means, an iris filter process for comparing an iris filter output value representing the maximum value of the degree of concentration of the density gradient of the image signal for each pixel of the radiation image with a predetermined threshold value is used. Perform detection of a tumor shadow candidate that is an abnormal shadow candidate, obtain an evaluation value for a predetermined degree of malignancy evaluation for the detected tumor shadow candidate, and more accurately determine the evaluation value based on the evaluation value from among the detected tumor shadow candidates. A method for detecting a candidate tumor shadow candidate is applied, and the detection capability is changed by weighting the evaluation value for the region corresponding to the outside or the region corresponding to the upper side in a direction of relatively increasing the evaluation value. Something that enhances performance can also be applied.

[0098]

According to the method and the apparatus for detecting an abnormal shadow of the present invention, the detection capability of the abnormal shadow candidate is changed for each area of the object appearing in the radiation image, so that the area can be compared with a specific area of the object. For abnormal shadows that are known experimentally or statistically to occur most often, the detectability of that specific area can be increased to increase the possibility of detection, preventing the omission of abnormal shadows from being detected As a result, the detection accuracy can be improved.

[0099]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, a specific embodiment of the method for detecting an abnormal shadow according to the present invention will be described with reference to the drawings.

FIG. 1 is a diagram showing a specific embodiment of an abnormal shadow detecting apparatus for implementing the abnormal shadow detecting method of the present invention.

[0102] The abnormal shadow detecting apparatus shown unshown image reading object input from the device such as a (breast) P _M of the radiographic image; based on the image data P representing a (mammography see FIG 2 (1)) P, It determined the placement of the breast P _M, region defining means 10 for dividing the breast P _M, in the outer region and the inner region or the upper and lower regions depending on the shooting Positions
Abnormal shadow detection means 20 for performing iris filter processing on the image data P to detect an image portion (tumor shadow candidate) P _J representing breast cancer in the mammography P;
Depending on the area of the breast P _M that abnormal shadow detecting means 20 detects the tumor shadow candidate P _J (area defined by the area defining means 10), change the iris filter output value by the iris filter processing of the abnormal shadow detecting means 20 And a detection capability changing unit 30 that performs the detection.

First, the mammography P input from an image reading device (not shown) will be described.

The mammography P generally includes a plane image (conventionally referred to as a front image; CC-view) which is photographed as a transmission image in the vertical direction with the breast sandwiched from above and below.
There is a lateral image (MLO-view) taken as a transmission image in the left and right direction with the breast sandwiched from the left and right. Planar image P1 is 3 (1), a side image P2 are formed as shown in FIG. (2), the outline P _S appearing in each of these images P1, P2, unnecessary as image information, shows a directly to the image portion radiation is irradiated (background-portion) outline of the stop irradiation field was used during shooting as P _X is less without passing through the breast as the object.

The area defining means 10 determines whether the input image data P represents the left or right breast and which imaging position (based on the header information attached to the image data P input from the image reading device or the like). Front (flat) or side)
The image data P
, The arrangement of the breast P _M is first determined, and one of the edges of the mammography P (the left side in FIG. 3 (1))
Determine the midpoint position of the body side edge of the breast P _M, which appeared in. That is, the image data P1 representing a planar image P1 shown in FIG. 3 (1) is input, the bodyside edge of the breast P _M (K1
Midpoint position KM is determined (a line segment connecting the line and K2 with a straight line). Further region defining means 10, the tip position of the breast (e.g. nipple position) determined in N, the line connecting the midpoint KM and tip position N previously obtained, breast P _M of the two regions P _A and P It is divided into and _B, the outer region P of the breast P _M
Defining the _A and the inner region P _B.

[0105] On the other hand, the midpoint figure when the image data P2 representing the side view image P2 shown is input to the (2), (a line segment connecting the K3 and K4 in a straight line) bodyside edge of the breast P _M Position K
N, and further, the tip position N of the breast P _M is determined,
Defined by the straight line connecting the midpoint previously determined KN and tip position N, divides the breast P _M into two regions P _C and P _D, an upper region P _C and the lower region P _D of the breast I do.

[0106] Here, the left breast of the human body 200 as shown in FIG. 4, to represent the nipple and the vicinity thereof in the region R _E, outer region R _C, R _B, R _F is the area P in FIG. 3 (1) _A , and the inner regions _RA and _RD are shown in FIG.
Corresponding to the region P _B in the plane image P1 of. On the other hand, in the upper regions R _C , R _D , and R _F , the side surface of FIG.
Corresponds to the region P _C in the lower region R _A, R _B corresponds to the region P _D in FIG. 3 (2).

[0107] Note that detection of the placement of the breast P _M based on the image data P may be determined by analyzing the histogram of the image data P, the breast image recording position recognition disclosed by JP-A 4-8351 Patent, etc. May be applied. Further, the nipple position N as the tip position of the breast may be obtained by applying iris filter processing by the abnormal shadow detecting means 20.

The image data P input from the image reading device or the like is also input to the abnormal shadow detecting means 20. The abnormal shadow detecting means 20 performs iris filter processing on the input image data P and outputs the iris filter output. The value I is obtained, and the obtained iris filter output value I is compared with a preset threshold value T. If I> T, the closed area where the output value I is obtained is regarded as a tumor shadow candidate P _J (FIG. 1) and (2)). On the other hand, if I <T, it does not detect the closed region to give the output value I as a tumor mass shadow candidate P _J.

However, since actual tumor shadows often occur in the outer upper regions R _C and R _F shown in FIG. 4, it is easier to detect the tumor shadow candidate P _J in the outer upper regions R _C and R _F. In the present embodiment, the closed region (candidate shadow candidate) from which the iris filter output value I is obtained is a region corresponding to the outer upper regions R _C and R _{F of} the breast (FIG. 4), that is, The external region _RA (FIG. 3A) or the upper region R defined by the region defining means 10.
_C (FIG. 3 (2)), the detection capability changing means 30
Is obtained by multiplying the iris filter output value I by a preset premium coefficient α (for example, α = 1.2) (I ′ =
Iα), the abnormal shadow detection means 20 compares the extra output value I ′ with the threshold value T, and if I ′> T, the closed area having the output value I ′ is regarded as a tumor shadow candidate P Detected as _J.

As described above, according to the abnormal shadow detecting apparatus of the present embodiment, by individually increasing the detectability of the tumor shadow candidate in the region where the actual tumor shadow specifically occurs, the detection in the other region is performed. Without changing the accuracy
The detection omission of the tumor shadow candidate in the region can be reduced, and the detection accuracy of the abnormal shadow candidate can be improved as a whole.

In the abnormal shadow detecting apparatus according to the present embodiment, the processing for detecting a tumor shadow candidate using an iris filter has been described. However, the processing for detecting a microcalcification shadow candidate using a morphological filter is not limited to the iris filter processing. It can also be applied to In this case, the morphological operation output value P in the aforementioned equation (17)
By multiplying (i, j) and Mgrad by a preset premium coefficient α (for example, α = 1.2), the abnormal shadow detecting means 20 outputs the output values P (i, j) and Mgrad to the output value P
Equation (17) may be applied instead of (i, j) 'and Mgrad'.

In addition, in the n-th space defined by the n-th feature amount of the malignancy evaluation, each Mahalanobis pattern of a benign tumor pattern and a malignant tumor pattern obtained experimentally or statistically in advance is used. In the comparison between the distances Dm1 and Dm2, the distance Dm1 to the pattern of the benign tumor shadow is:
The present invention can also be applied so that the comparison is performed after multiplying by a preset premium coefficient α (for example, α = 1.2).

[Brief description of the drawings]

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

FIG. 2 is a diagram for explaining the operation of iris filter processing;

FIG. 3 is a view for explaining the definition of areas for mammography P1 and P2.

FIG. 4 is a diagram showing an area of a breast on a human body;

FIG. 5 is a diagram showing a mask having a size of 5 pixels vertically × 5 pixels horizontally centered on a target pixel j;

FIG. 6 is a diagram illustrating an angle formed between a target pixel and a gradient vector at each pixel j.

FIG. 7 is a conceptual diagram showing an iris filter set so that a contour shape is adaptively changed.

FIG. 8 is a diagram showing a basic operation of a morphological filter.

FIG. 9 shows four linear structural elements B.

[Explanation of symbols]

 10 Area defining means 20 Abnormal shadow detecting means 30 Detectability changing means P Image data

Claims (14)

[Claims] 1. An abnormal shadow detection method for detecting a candidate for an abnormal shadow in an image based on an image signal representing a radiation image of the object, wherein the abnormal shadow candidate is determined for each region of the object appearing in the radiation image. A method for detecting abnormal shadows, characterized by changing the detectability of a shadow. 2. The method according to claim 1, wherein the detection capability is changed so that the abnormal shadow candidate is detected more preferentially than other regions in an area of the subject where the abnormal shadow is likely to be present experimentally or statistically. The method for detecting abnormal shadows according to claim 1, wherein: 3. The radiographic image is a mammogram, and in the mammography, an abnormal shadow candidate is detected in a region corresponding to an outer side or an upper side of the breast as the subject more than a region corresponding to an inner side or a lower side. 3. The abnormal shadow detection method according to claim 2, wherein the detection performance is relatively increased. 4. A tumor which is an abnormal shadow candidate by iris filter processing for comparing an iris filter output value representing a maximum value of a concentration degree of a density gradient of the image signal for each pixel of the radiation image with a predetermined threshold. In detecting a shadow candidate, the area corresponding to the outside or the area corresponding to the top is weighted in a direction in which the output value of the iris filter is relatively increased, and / or the area corresponding to the outside. 4. The abnormal shadow detection method according to claim 3, wherein the detectability is increased by weighting in a direction in which the threshold value for the region corresponding to the upper side is relatively reduced. 5. 5. An output value of a morphological operation using a structuring element having a size larger than the abnormal shadow candidate to be detected and a predetermined threshold value for the image signal of each pixel of the radiation image. A direction for relatively increasing the morphology calculation output value for the area corresponding to the outside or the area corresponding to the upper side when detecting the microcalcification shadow candidate that is the abnormal shadow candidate by the morphology filter processing to be compared. And / or increasing the detectability by weighting in a direction in which the threshold value for the region corresponding to the outside or the region corresponding to the upper side is relatively reduced. Item 3. The abnormal shadow detection method according to Item 3. 6. A tumor which is an abnormal shadow candidate by performing an iris filter process of comparing an iris filter output value representing a maximum value of a degree of concentration of a density gradient of the image signal for each pixel of the radiation image with a predetermined threshold value. Detect shadow candidates, determine an evaluation value for a predetermined degree of malignancy evaluation for the detected tumor shadow candidate, and detect a more probable tumor shadow candidate based on the evaluation value from among the detected tumor shadow candidates 4. The abnormal shadow according to claim 3, wherein, when performing the detection, the detectability is increased by weighting the evaluation value of the area corresponding to the outside or the area corresponding to the upper side in a direction of relatively increasing the evaluation value. Detection method. 7. An arrangement of the breast in the mammography based on the image signal, a midpoint position of a body side edge of the breast appearing at any one of the edges of the mammography, and Finding the tip position, dividing the breast into two regions by a straight line connecting the midpoint position and the tip position, defining a region corresponding to the outside or a region corresponding to the top. The abnormal shadow detection method according to any one of claims 3 to 6. 8. An abnormal shadow detecting apparatus comprising an abnormal shadow detecting means for detecting a candidate for an abnormal shadow in an image based on an image signal representing a radiation image of the object. Each of which further comprises a detectability changing means for changing the detectability of the abnormal shadow candidate, wherein the abnormal shadow detecting means detects the abnormal shadow candidate based on the changed detectability. Shadow detection device. 9. The method according to claim 1, wherein the detection capability changing unit detects the abnormal shadow candidate in an area of the subject where the abnormal shadow is likely to be experimentally or statistically higher than other areas. 9. The abnormal shadow detecting device according to claim 8, wherein said detecting ability is changed. 10. The radiographic image is a mammography, and the detection capability changing means is superior to a region corresponding to an outside or an upper portion of the breast as the subject in the mammography, than a region corresponding to an inside or a lower portion. 10. The abnormal shadow detection device according to claim 9, wherein the detection capability is relatively increased so as to detect the abnormal shadow candidate. 11. The iris filter processing in which the abnormal shadow detecting means compares an iris filter output value representing the maximum value of the concentration gradient of the density gradient of the image signal for each pixel of the radiation image with a predetermined threshold value. Detecting a tumor shadow candidate that is the abnormal shadow candidate, wherein the detectability changing means relatively increases the iris filter output value for the region corresponding to the outside or the region corresponding to the top. By weighting
11. The detection performance is enhanced by weighting the threshold value for a region corresponding to the outside or the region corresponding to the upper side in a direction of relatively lowering the threshold. Abnormal shadow detection device. 12. An output of a morphological operation using the structural element having a size larger than the abnormal shadow candidate to be detected, for the image signal of each pixel of the radiographic image. A microcalcification shadow candidate, which is the abnormal shadow candidate, by a morphology filter process of comparing a value with a predetermined threshold value, wherein the detectability changing means corresponds to the area corresponding to the outside or the area corresponding to the upper side. And / or weighting in a direction to relatively lower the threshold value for the region corresponding to the outside or the region corresponding to the upper side by weighting the morphology operation output value for the region to be relatively increased. 11. The abnormal shadow detection device according to claim 10, wherein the detection capability is enhanced by performing the operation. . 13. An iris filter process in which the abnormal shadow detection means compares an iris filter output value representing the maximum value of the concentration gradient of the density gradient of the image signal for each pixel of the radiation image with a predetermined threshold value. Perform detection of a tumor shadow candidate that is the abnormal shadow candidate, determine an evaluation value for a predetermined degree of malignancy evaluation for the detected tumor shadow candidate,
It is for detecting a more probable tumor shadow candidate from the detected tumor shadow candidates based on the evaluation value, and the detectability changing means is an area corresponding to the outer side or an area corresponding to the upper side. 11. The abnormal shadow detection device according to claim 10, wherein the detection performance is enhanced by weighting the evaluation value of in a direction in which the evaluation value is relatively increased. 14. An arrangement of the breast in the mammography based on the image signal, a midpoint position of a body side edge of the breast appearing at any one of the edges of the mammography, and Finding the tip position, by dividing the breast into two regions by a straight line connecting the midpoint position and the tip position, an area defining means for defining an area corresponding to the outside or an area corresponding to the upper part The abnormal shadow according to any one of claims 10 to 13, further comprising: the detectability changing means for increasing the detectability in accordance with an area defined by the area defining means. Detection device.