JP2002157577A - Method and device for detecting abnormal shadow candidate - Google Patents

Abstract

PROBLEM TO BE SOLVED: To stably detect an abnormal shadow as an abnormal shadow candidate even when the abnormal shadow such as a tumor shadow exists in the neighborhood of the end edge of an image in an abnormal shadow candidate detecting device. SOLUTION: When the tumor shadow candidate is detected by using an iris filter processing, a reflection processing is performed, where a direction traverse is reflected at the end edge of the image by a reflection angle being the same as an incident angle to the image end edge as the gradient vector of a pixel after the image end edge concerning a direction towards the image end edge within the range of the maximum radius Rmax of the tumor shadow to be detected. The gradient vector of the pixel at a position corresponding to the pixel after the image end edge on the direction traverse after the reflection processing is folded with the image end edge as a linearly symmetrical axis so as to substitute a folding gradient vector which is obtained by this.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a method and an apparatus for detecting a candidate portion of an abnormal shadow such as a tumor shadow in an image.

[0002]

2. Description of the Related Art Conventionally, image processing such as gradation processing and frequency processing has been performed on an image signal representing an image obtained by various image acquisition methods to improve image observing and reading performance. Have been. Particularly in the field of medical images such as directional images of the human body as subjects, it is necessary for specialists such as doctors to accurately diagnose the presence or absence of a patient's disease or injury based on the obtained images. Image processing for improving image reading performance is indispensable.

In such image processing, the entire image may be processed, but if the purpose of inspection or diagnosis is clear to some extent, only a desired image portion suitable for the purpose is selectively selected. May be emphasized.

Normally, such an image portion is selected manually by an observer while observing an original image before image processing is performed, if necessary. The specified range is determined by the experience of the observer and the level of image reading ability, and may not be objective.

For example, in a direction line image photographed for the purpose of examining breast cancer, it is necessary to extract a tumor shadow, which is one of the features of a cancerous part, from the direction line image. It is not always possible to specify the range of the mass shadow. For this reason, it is required to accurately detect abnormal shadows such as tumor shadows without depending on the skill of the observer.

In order to meet this demand, Computer Aided Diagnosis of Medic (CADM; Computer Aided Diagnosis) which automatically detects abnormal shadow candidates using computer processing.
research on calimages) has recently been in progress. As one of the CADM technologies, iris filter processing has been proposed (Obata et al., "Detection of tumor shadow in DR image (iris filter)" IEICE Transactions D-II Vol.J75-D-II No.
3 P663-670, March 1992; hereinafter referred to as Reference 1).

The applicant of the present application also discloses, for example, JP-A-8-294479,
No. 9-167238, Japanese Patent Application No. 2000-038298, etc., automatically detect abnormal shadows suggesting the presence of breast cancer etc. using mammography (diagnostic breast images) and chest images using an iris filter. (This processing is referred to as abnormal shadow candidate detection processing), and an image having high diagnostic performance suitable for image interpretation is obtained by superimposing an image obtained by performing an emphasis processing, an enlargement processing, or the like on a detected abnormal shadow candidate part on a part of the entire image. And
An abnormal shadow candidate detection processing system (computer-assisted image diagnostic apparatus) has been proposed which outputs an index value or a certainty factor of the detection in the detection processing.

Here, abnormal shadow candidate detection processing using an iris filter (hereinafter also referred to as iris filter processing) represents a density gradient (or luminance gradient) in an image as a gradient vector, and the degree of concentration of the gradient vector is high. According to this method, an image part is detected as a candidate for an abnormal shadow. According to this, a candidate for a tumor shadow (a form of an abnormal shadow), which is a characteristic form of breast cancer, is automatically detected based on the degree of concentration of the gradient vector. Can be detected.

In this iris filter processing, in order to achieve shape adaptability, a large number of direction lines extending radially from the pixel of interest are set, the degree of concentration for each direction line is determined, and the concentration for each direction line is determined. The degree of concentration C of the gradient vector group for the pixel of interest is calculated based on the degree (specifically, the averaging is performed), and the degree of concentration C of the group of gradient vectors is compared with a predetermined threshold value as the degree of concentration for the pixel of interest. In some cases, detection of a candidate for a tumor shadow may be performed.

The iris filter processing has been studied as an effective method for detecting a tumor shadow, which is one of the characteristic features of breast cancer, but the target image is limited to a tumor shadow in mammography. However, if the gradient of an image signal (such as density) representing the image is concentrated, the present invention can be applied to any image portion.

Here, when iris filter processing is applied for detecting a tumor shadow, for example, a direction line image on an X-ray film (an image represented by an image signal of high density and high signal level) Shading is known to have a slightly lower density value than the surrounding image part,
The density value distribution has a density value gradient such that 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.

That is, the gradient vector at an arbitrary pixel (pixel of interest) in the tumor shadow points near the center of the tumor shadow, but the gradient vector concentrates at a specific point in an elongated shadow such as a blood vessel shadow or a mammary gland. Therefore, the evaluation is performed using the degree of concentration C indicating the degree of distribution of the direction of the local gradient vector, and if a region where the direction of the gradient vector is concentrated at a specific point is extracted, it is regarded as a tumor shadow. Become.

The gradient vector of each pixel near the tumor shadow is
Regardless of the magnitude of the contrast of the tumor shadow, it is directed substantially toward the center of the tumor shadow. Therefore, the target pixel having a large concentration value can be determined as 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 all the pixels constituting the image, and whether or not the value of the degree of concentration C exceeds a predetermined 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 and can efficiently detect a tumor shadow as compared with a normal difference filter.

Further, in the actual processing,
In order to achieve a detection power independent of the size and shape of a tumor, a technique for adaptively changing the size and shape of a filter has also been proposed as shown in Japanese Patent Application Laid-Open No. H11-157572.

By the way, in the iris filter processing, since the density information is referred to over the entire periphery of the region of interest, the calculation target range becomes small at the end of the photographed image, and as a result, the extraction accuracy decreases. With. Specifically, when a tumor shadow exists near the image edge of the original image, that is, the tumor shadow exists within a range from the image edge of the original image to the maximum value Rmax of the radius of the tumor shadow to be detected. In this case, the direction line set in the process reaches the image edge before reaching the maximum value Rmax of the radius of the tumor shadow, and the degree of concentration for each direction line is calculated within the range of the distance to the edge. Therefore, the degree of concentration of the pixel of interest finally obtained is smaller than the degree of concentration when the original image exists in a region other than the image edge of the original image. Therefore, the concentration of the gradient vector group is obtained using the concentration of the direction line,
Assuming that the degree of concentration C of the gradient vector group is the degree of concentration for the pixel of interest, the degree of concentration for this pixel of interest is also smaller than the value in the case where it exists in an area other than the end of the image. , It may not be detected as a tumor shadow candidate.

Although the end of the image of mammography is predicted to be a breast cancer-prone site, due to the above-mentioned drawbacks of the iris filter processing, detection errors at the end of the image and erroneous detection at other regions are likely to occur. In many cases, the candidate area is concentrated in the dense area.

As a method for solving a problem that may occur when a tumor shadow exists at the end of an image, the applicant of the present application disclosed in Japanese Patent Application Laid-Open No. H10-99305, when a tumor shadow exists at the end of an image. By duplicating the image with the edge of the image as the symmetry axis and wrapping the image, and applying the iris filter processing also using the duplicated image data of the duplicated image portion duplicated by this folding, abnormal shadows such as tumor shadows can be removed A method has been proposed in which an abnormal shadow candidate area, which is an area including an abnormal shadow candidate, can be stably extracted even when the abnormal shadow candidate area exists at an image edge.

[0018]

However, the method proposed in Japanese Patent Application Laid-Open No. H10-99305 has to generate a duplicate image by folding the image with the edge of the image as the axis of symmetry. There are drawbacks in that the overall throughput is reduced by the time required for generation, and that an extra memory for storing generated image data is required.

The present invention has been made in view of the above circumstances, is simple in processing, does not significantly reduce the overall throughput, and can be used even when an abnormal shadow such as a tumor shadow exists at the end of an image. It is an object of the present invention to provide an abnormal shadow candidate detection method and apparatus capable of stably detecting an abnormal shadow candidate.

[0020]

According to a first abnormal shadow candidate detecting method of the present invention, a gradient vector for each pixel based on the image data is obtained for all pixels of image data representing an image, and a predetermined pixel ( The predetermined pixel based on the gradient vector of each pixel on a number of directional lines extending radially from the predetermined pixel within a predetermined distance corresponding to the size of the abnormal shadow candidate to be detected centered on the target pixel) An abnormal shadow candidate in an image is detected by calculating a degree of concentration indicating a degree of distribution of a direction of a local gradient vector centered on a pixel, and comparing the value of the degree of concentration with a preset threshold. In the abnormal shadow candidate detection method, a pixel (virtual pixel) after an edge of a direction line reaching an edge of an image within a predetermined distance from a number of direction lines extending from a predetermined pixel. A folding gradient vector obtained by folding a gradient vector of a pixel at a position corresponding to a pixel subsequent to the edge on the direction line after reflecting the direction line at the edge as an arrangement vector, using the edge as an axis of line symmetry. Is used to determine the degree of concentration. That is, in the first method, the return gradient vector, which is a gradient vector obtained by folding the gradient vector of the real image using the image edge as the axis of line symmetry, is substituted for the gradient vector of the pixels subsequent to the edge and concentrated. It is to ask for degree.

Here, substituting the return gradient vector means using the direction and magnitude of the return gradient vector. In calculating the actual degree of concentration, the components of the gradient vectors in the direction of the line are used, as in the case of the process before the image edge arrives.

A pixel at a position corresponding to a pixel after the edge on the direction line after reflecting the direction line at the edge is, in short, a pixel on the direction line extending from a predetermined pixel as an axis whose edge is line-symmetrical. Is a pixel obtained by folding back. When there is a direction line reaching the edge of the image within a predetermined distance, it means that the predetermined pixel is a pixel existing within the predetermined distance from the edge of the image. The same applies hereinafter.

When the pixel at the edge of the image is set as the predetermined pixel, it is virtually impossible to take a direction line toward the edge. In this case, the predetermined pixel is outside the real image (virtual image space). In this case, the direction line extending inward) is immediately folded back with the image edge as the axis of line symmetry, and the folded gradient vector at the corresponding position in the real image space may be used. The same applies hereinafter.

A second method for detecting an abnormal shadow candidate according to the present invention is as follows.
For all the pixels of the image data representing the image, a gradient vector is determined for each pixel based on the image data, and a predetermined distance corresponding to the size of the abnormal shadow candidate to be detected centered on the predetermined pixel (pixel of interest) is determined. A degree of concentration indicating a degree of distribution of a direction of a local gradient vector centered on the predetermined pixel is obtained based on a gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within the range. In the abnormal shadow candidate detection method of detecting an abnormal shadow candidate in an image by comparing the value of the degree of concentration with a preset threshold, a predetermined distance of a number of direction lines extending from a predetermined pixel is determined. Pixels (virtual pixels) after the edge of the direction line reaching the edge of the image within the range
The concentration degree is obtained by substituting the gradient vector of a pixel corresponding to a pixel after the edge on the direction line after reflecting the direction line at the edge as the gradient vector of (i).

Here, "after the direction line is reflected at the edge" means that after the reflection process for reflecting the direction line at the edge at a reflection angle substantially equal to the incident angle to the edge. It is.

Here, substituting the gradient vector means using the direction and magnitude of the gradient vector. In the actual calculation of the degree of concentration, the components in the direction and line direction of each gradient vector are used in the same manner as in the process before the image edge arrives.

According to the investigation by the present inventor, in both of the first method and the second method, the positional relationship between the approximate center of the abnormal shadow candidate and the edge (disposition of the abnormal shadow at the edge) It has been found that, depending on the pattern, the effect may not always be sufficiently exerted. In other words, each method has a special arrangement pattern of abnormal shadows. The third abnormal shadow candidate detection method of the present invention is based on such knowledge.
That is, the third abnormal shadow candidate detection method of the present invention obtains a gradient vector for each pixel based on the image data for all pixels of the image data representing the image, and centers on a predetermined pixel (pixel of interest). Based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within a range of a predetermined distance corresponding to the size of the abnormal shadow candidate to be detected, a local area centered on the predetermined pixel In the abnormal shadow candidate detection method for detecting an abnormal shadow candidate in an image by obtaining a degree of concentration indicating the degree of distribution of the direction of the gradient vector, and comparing the value of the degree of concentration with a preset threshold value, When the distance between the pixel on the direction line and the edge when the direction line is reflected at the edge is longer than the distance between the predetermined pixel (pixel of interest) and the edge, the direction line is reflected at the edge. A direction that reaches the edge of the image within a predetermined distance within a predetermined distance within a range of a return gradient vector obtained by folding a gradient vector of a pixel at a position corresponding to a pixel subsequent to the edge on the direction line with the edge as an axis of line symmetry. While the degree of concentration is obtained by substituting the gradient vector of pixels (virtual pixels) after the edge of the line, the distance between the pixel on the direction line and the edge when the direction line is reflected at the edge is predetermined. Is smaller than the distance between the pixel of interest (the pixel of interest) and the edge, the gradient vector of the pixel at the position corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge is calculated as the pixel after the edge. This method is characterized in that the degree of concentration is obtained by substituting the gradient vector of (virtual pixel).

That is, in the third method, when the distance between the pixel on the direction line and the edge is greater than the distance between the pixel of interest and the edge when the direction line is reflected at the edge, the first method is used.
When the distance between the pixel on the direction line and the edge is less than the distance between the pixel of interest and the edge, the degree of concentration is calculated using the second method. In other words, in short, it depends on whether to use the return gradient vector at the corresponding pixel position or the gradient vector as it is, depending on the arrangement pattern of the abnormal shadow at the image edge.

When the direction line is reflected at the edge, the distance between the pixel on the direction line and the edge or the distance between the pixel of interest and the edge is the shortest distance between each pixel and the edge. Mean vertical distance.

In the first to third methods, when extracting the abnormal shadow candidate area by comparing the value of the degree of concentration with a preset threshold, for example, when the value of the degree of concentration is not less than the threshold, Determines that the predetermined pixel exists in the region of the abnormal shadow candidate, and determines that the predetermined pixel does not exist in the region of the abnormal shadow candidate when the predetermined pixel is less than the threshold value. It is good to extract a region. That is, pixels in a region where the direction of the gradient vector is concentrated at a specific point are detected as abnormal shadow candidates.

When the return gradient vector at the corresponding pixel position or the gradient vector itself is used as in the first to third methods, the gradient vector used as a substitute for the actual object far from the image edge is calculated. Since the image is different from the image indicated by the pixel data, the gradient vector used in the actual subject is different from the gradient vector.
In practice, accurate substitution (assuming) is difficult, and in some cases, the effect of improving the detection accuracy may be reduced.

In order to solve this problem, the first to fourth aspects of the present invention are described.
In the third abnormal shadow candidate detection method, weighting correction processing is performed according to the distance from a predetermined pixel, that is, the degree of substitution decreases as the distance from the predetermined pixel after the edge increases. It is desirable to calculate the degree of concentration by applying a weighting process to the data.

Here, "the weighting process is performed so as to be smaller as the distance from the predetermined pixel after the edge becomes larger" means that when the distance is large, the degree of substitution is made smaller than when the distance is small, Conversely, when the distance is small, it means that the degree of substitution is greater than when the distance is large.

On the other hand, as in the above-described first to third methods, the degree of concentration is obtained by using the return gradient vector at the corresponding pixel position or the gradient vector itself. When the abnormal shadow candidate area is extracted by comparing with a preset threshold value, the problem that the tumor shadow candidate present near the image edge is not detected can be solved. There is a possibility that a candidate area is not detected, and a tumor shadow candidate existing outside the vicinity of the image edge is not detected.

In order to solve this problem, the first to fourth aspects of the present invention are described.
In the third abnormal shadow candidate detection method, the image is divided into an image region near an edge of the image and an image region other than the vicinity of the edge according to the size of the predetermined distance, and each of the divided regions is divided into image regions. It is desirable to set a value corresponding to the maximum value of the number of abnormal shadow candidate areas that can be detected in the area for each time.

Here, "a value corresponding to the maximum value of the number of abnormal shadow candidate areas that can be detected in each of the divided areas is set", but the abnormal shadow candidate finally detected is set. What is necessary is just to make the number of areas fall within the maximum value, and the value of the maximum value actually set may be set according to the method of the determination processing. For example, when setting as the number of abnormal shadow candidate areas, the maximum value of the number of abnormal shadow candidate areas that can be detected may be the same as the maximum value itself, which is set when determining whether or not the pixel of interest is an abnormal shadow candidate pixel. In this case, the maximum value of the number of pixels of interest that can be determined to be a pixel of an abnormal shadow candidate corresponding to the maximum value of the number of abnormal shadow candidate areas that can be detected may be used.

The first abnormal shadow candidate detecting apparatus according to the present invention comprises:
A device for performing the first method, that is, a gradient vector calculating unit that calculates a gradient vector for each pixel based on the image data for all pixels of the image data representing the image, A local gradient vector centered on the predetermined pixel based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within a predetermined distance range corresponding to the size of the abnormal shadow candidate A degree-of-concentration calculating means for obtaining a degree of concentration indicating the degree of distribution of the orientation, a comparing means for comparing the obtained value of the degree of concentration with a preset threshold value, and the image based on an output from the comparing means. An abnormal shadow candidate detecting device for detecting the abnormal shadow candidate in the image processing apparatus, wherein the degree of concentration calculating means is arranged to reach an edge of the image within a predetermined distance range. As a gradient vector of a pixel after the edge of the line, a gradient vector of a pixel at a position corresponding to a pixel after the edge on the direction line after reflecting the direction line at the edge is defined as an axis symmetrical with the edge. In this case, the degree of concentration is obtained by using a folding gradient vector obtained by folding as a substitute.

The second abnormal shadow candidate detecting apparatus of the present invention comprises:
A device for performing the second method, that is, a gradient vector calculating unit that calculates a gradient vector for each pixel based on the image data for all pixels of image data representing an image; A local gradient vector centered on the predetermined pixel based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within a predetermined distance range corresponding to the size of the abnormal shadow candidate A degree-of-concentration calculating means for obtaining a degree of concentration indicating the degree of distribution of the orientation, a comparing means for comparing the obtained value of the degree of concentration with a preset threshold value, and the image based on an output from the comparing means. A candidate detecting means for detecting the abnormal shadow candidate in the abnormal shadow candidate detecting means. As a gradient vector of a pixel (virtual pixel) after the edge of the direction line reaching the edge of the image within the range of the distance, the edge after the edge on the direction line after reflecting the direction line at the edge It is characterized in that the degree of concentration is obtained by substituting the gradient vector of the pixel corresponding to the pixel.

The third abnormal shadow candidate detecting apparatus according to the present invention comprises:
A device for performing the third method, that is, a gradient vector calculating unit that calculates a gradient vector for each pixel based on the image data for all pixels of the image data representing the image, A local gradient vector centered on the predetermined pixel based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within a predetermined distance range corresponding to the size of the abnormal shadow candidate A degree-of-concentration calculating means for obtaining a degree of concentration indicating the degree of distribution of the orientation, a comparing means for comparing the obtained value of the degree of concentration with a preset threshold value, and the image based on an output from the comparing means. An abnormal shadow candidate detecting device for detecting the abnormal shadow candidate in the abnormal shadow candidate detecting device. When the distance between the element and the edge is equal to or greater than the distance between the predetermined pixel (target pixel) and the edge, the pixel at the position corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge Of the direction vector that reaches the edge of the image within a predetermined distance within a predetermined distance within a predetermined distance (virtual pixel) And the distance between the pixel on the direction line and the edge when the direction line is reflected at the edge is smaller than the distance between the predetermined pixel (pixel of interest) and the edge. Sometimes, a gradient vector of a pixel at a position corresponding to a pixel after the edge on the direction line after reflecting the direction line at the edge is substituted as a gradient vector of a pixel (virtual pixel) after the edge. Degree Out those characterized in that.

As the comparison means used in the first to third abnormal shadow candidate detecting devices of the present invention, for example, when the value of the degree of concentration is equal to or more than a threshold value, it is determined that a predetermined pixel exists in the abnormal shadow candidate area. If it is determined that it is less than the threshold, it may be determined that the predetermined pixel does not exist in the area of the abnormal shadow candidate.

In the first to third abnormal shadow candidate detecting devices according to the present invention, the concentration calculating means is superimposed such that the degree of substitution decreases as the distance from a predetermined pixel after the edge increases. It is desirable to perform the attaching process to obtain the degree of concentration.

In the first to third abnormal shadow candidate detecting apparatuses according to the present invention, the candidate detecting means includes an image area near an edge of the image and an area other than the edge near the edge according to the size of the predetermined distance. The image is divided into image regions, and a value corresponding to the maximum value of the number of regions including abnormal shadow candidates that can be detected in the region is set for each of the divided regions, and a range of the set value is set. It is desirable to detect the abnormal shadow candidate within the above.

[0043]

According to the first method and apparatus for detecting an abnormal shadow candidate of the present invention, when the direction line set at the time of iris filter processing reaches the edge of the image within the predetermined distance, As a gradient vector of the pixel after the edge of the direction line, a fold gradient vector obtained by folding the gradient vector for the corresponding pixel obtained by performing the reflection processing at the image edge with the edge as a line symmetric axis is defined as a gradient vector of the pixel. Instead, the degree of concentration is calculated using the folded gradient vector that is substituted, so that the calculation of the degree of concentration for each direction line calculated within the range up to the image edge is performed by calculating the radius of the tumor shadow. It is now possible to expand to the maximum value range. As a result, the degree of concentration for each direction line can take a larger value, and therefore, the value of the degree of concentration C for the target pixel increases and exceeds the threshold, and the target pixel exists in the abnormal shadow candidate area. Can be detected. Therefore, even when the tumor shadow exists near the image edge and the entire shadow is not reflected in the image, a tumor shadow candidate area corresponding to the tumor shadow can be extracted with higher accuracy than in the past.

According to the investigation by the inventor of the present invention, the arrangement pattern of the abnormal shadows at the edge portions where the above-mentioned effect of the first abnormal shadow candidate detecting method and apparatus can be sufficiently exerted is particularly effective. It turned out that this is the time when the center of the abnormal shadow candidate is located outside the edge of the image.

According to the second method and apparatus for detecting an abnormal shadow candidate of the present invention, when the direction line set at the time of the iris filter processing reaches the edge of the image within the range of the predetermined distance, the direction line As the gradient vector of the pixel after the edge of, the gradient vector for the corresponding pixel obtained by performing reflection processing at the image edge is substituted,
Since the degree of concentration is obtained by using this substitute gradient vector, the degree of concentration for each direction line calculated within the range up to the image edge is calculated in the same manner as in the first method and apparatus. Can be extended to the range of the maximum value of the radius of the tumor shadow, and even when the tumor shadow exists near the image edge and the entire shadow is not reflected in the image, the The corresponding tumor shadow candidate area can be extracted with higher accuracy than before.

According to the investigation by the inventor of the present invention, the arrangement pattern of the abnormal shadows at the edge portions where the above-described effect of the second abnormal shadow candidate detecting method and apparatus can be sufficiently exerted is particularly effective. It has been found that the center of the abnormal shadow candidate is located at the edge of the image, that is, within the original image area.

According to the third abnormal shadow candidate detecting method and apparatus of the present invention, when the direction line is reflected at the edge, the distance between the pixel on the direction line and the edge is equal to the distance between the target pixel and the edge. Depending on whether the distance is greater than or less than the distance, the alternative gradient vector at the corresponding pixel position can be substituted, or the gradient vector can be substituted as it is. It can be used properly according to the pattern, and each effect as the first or second method and device can be sufficiently exhibited, regardless of the arrangement pattern.
A tumor shadow candidate area corresponding to the tumor shadow located at the edge of the image can be extracted very accurately.

In any of the first to third methods and apparatuses, a memory for storing the duplicated image data is not required.

Further, weighting processing is performed so that the degree of each of the folded gradient vectors or the substitutes of the gradient vectors becomes smaller as the distance from the predetermined pixel after the edge increases, and the concentration degree is obtained. For example, even when it is necessary to use a return gradient vector or a gradient vector in a portion far from the image edge, the effect of improving the detection accuracy can be increased.

Further, the image is divided into an image region near the edge of the image and an image region other than the vicinity of the edge according to the size of the predetermined distance, and each divided region is divided into the image region. By setting a value corresponding to the maximum value of the number of regions including the abnormal shadow candidates that can be detected, excessively detecting a tumor shadow candidate region for one tumor shadow near the image edge, that is, an abnormal shadow candidate It is possible to prevent the position extracted as a region from being concentrated at the end of the image.

[0051]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS Specific embodiments of the abnormal shadow candidate detecting method and apparatus according to the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a first embodiment of an abnormal shadow candidate detecting apparatus according to the present invention. FIG. 2 is a schematic configuration of an example of a computer-aided image diagnostic apparatus including the abnormal shadow candidate detecting apparatus shown in FIG. FIG.

The computer-aided image diagnostic apparatus 1 shown in FIG.
Input image data (hereinafter referred to as whole image data) S
A whole image processing unit 4 for reading out the whole image data S stored in the storage unit 3 and performing image processing such as gradation processing and frequency emphasis processing; The present invention reads out the entire image data S and extracts image data (hereinafter referred to as local image data) representing an image of an abnormal shadow candidate area which is an area including an abnormal shadow (tumor shadow) candidate in the entire image data S. An abnormal shadow candidate detecting device 2 and a local image processing means 5 for emphasizing a tumor shadow candidate with respect to the extracted local image data
And a display means 6 for displaying, as a visible image, a whole image processed by the whole image processing means 4 and a tumor shadow candidate processed by the local image processing means 5.

Here, the whole image data S input to the computer-aided image diagnostic apparatus 1 is, for example, a stimulable fluorescent light obtained by storing and recording an image representing a patient's mammography (mammogram) as shown in FIG. This is image data (high-density high-signal level data) obtained by photoelectrically reading stimulated emission light generated by irradiating the body sheet with excitation light, and then performing digital conversion. A tumor shadow is recorded at the lower end of the image in the figure.

The display means 6 may display the whole image and the tumor shadow separately on the display surface, but in the present embodiment, while displaying the whole image, the image of the tumor shadow candidate in the whole image is displayed. It is assumed that only the part is replaced with the image of the tumor shadow candidate subjected to the image processing by the local image processing means 5 and displayed.

The abnormal shadow candidate detecting device 2 is described in detail in FIG.
As shown in FIG. 5, for all pixels of the input image data S, a gradient vector calculating unit 10 for calculating a density gradient vector for each pixel based on the image data, and a detection centering on a target pixel as a predetermined pixel A local area centered on the pixel of interest based on the gradient vector of each pixel on a number of directional lines radially extending from the pixel of interest within a predetermined distance range corresponding to the size (detection size) of the target abnormal shadow candidate Degree calculating section 20 for calculating the degree of concentration indicating the degree of distribution of the direction of the gradient vector, and comparing the calculated value of the degree of concentration C with a preset threshold value T. As a result of this comparison, the degree of concentration When the value of C is greater than or equal to the threshold T, it is determined that the pixel of interest exists in the area of the tumor shadow candidate, and when the value of the concentration C is less than the threshold T, the pixel of interest exists in the area of the abnormal shadow candidate. do not do And it determines the comparison determination unit 31, and a extraction means 32 for comparing based on the output from the determination unit 31 extracts the abnormal shadow candidate region is a region including the abnormal shadow candidate in the image.

The candidate detecting means 30 of the present invention is constituted by the comparing and judging means 31 and the extracting means 32. As described above, in the present embodiment, the candidate detection unit 30 of the present invention includes the part that functions after detecting an abnormal shadow candidate and extracting the abnormal shadow candidate region including the abnormal shadow candidate. The process of extracting a shadow candidate region is not always an essential requirement, and the candidate detection unit 30 of the present invention may include only the comparison determination unit 31.

The gradient vector calculating means 10 uses the image data (pixel value) of the outermost peripheral pixel of the mask having a size of 5 pixels × 5 pixels shown in FIG. ), The direction θ of the density gradient vector is obtained.

(Equation 1)

Here, f _{1 to} f ₁₆ are, as shown in FIG.
This is a pixel value (image data) corresponding to a pixel on the outer periphery of a mask having a size of, for example, 5 pixels × 5 pixels centered on the pixel j. The mask size is not limited to 5 pixels vertically × 5 pixels horizontally, and various sizes can be used.

By the way, as a general concentration degree calculation process, for all the pixels constituting the target image, for each pixel, the concentration degree C of the gradient vector having that pixel as the pixel of interest is expressed by the following equation (2). ).

[0061]

(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 This is the angle formed by the gradient vector calculated in (1) (see FIG. 5). 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 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, 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.

On the other hand, in the present embodiment, in order to achieve a detection power independent of the size and shape of the tumor, the abnormal shadow candidate detecting device 2 is adapted to change the size and shape of the filter adaptively. The configuration is such that the degree of concentration C is calculated. Specifically, a detection size setting unit 22 that sets a minimum size (radius Rmin) and a maximum size (radius Rmax) of a tumor shadow candidate to be detected as a detection size, and one of all pixels. A pixel of interest setting means 24 for sequentially setting as a pixel of interest, and a predetermined angular interval (2π / M) about the pixel of interest set by the pixel of interest setting means 24
Direction line setting means 23 for setting N radial direction lines (direction lines; see FIG. 6) adjacent at (degree intervals), and the maximum size of the tumor shadow candidate from the pixel of interest for each set direction line For each pixel on the direction line within the range of the distance Rmax according to the index value calculation means 25 for obtaining an index value cos θil based on the angle θil between the gradient vector of each pixel and the direction in which this direction line extends, , Starting from the pixel of interest,
By changing the end point from the distance Rmin according to the minimum size of the set tumor shadow candidate to the distance Rmax according to the maximum size, the index value cos of each pixel on the direction line within the range from the start point to the end point An average value Ci (n) of θil is obtained for each direction line according to a predetermined formula, and a maximum value calculating unit 26 for extracting a maximum value Cimax among the average values is provided in the concentration degree calculating unit 20.

In the gradient vector group concentration degree calculating means 27, the maximum value Cimax extracted for each direction line is averaged ((ΣCimax) / N) for all N direction lines.
Then, the concentration C of the gradient vector group for the pixel of interest
Is calculated.

Here, the angle θil is defined by the gradient vector of the l-th pixel from the target pixel and the extension of the i-th direction line on the i-th direction line among the N direction lines. Indicates the angle between the direction.

The detection size set by the detection size setting means 22, that is, the minimum size (radius Rmin) and the maximum size (radius Rmax) of the tumor shadow candidate to be detected are set by the operator in the detection size setting means 22. A method of inputting from an input means such as an attached keyboard (not shown) may be used, or an appropriate one is automatically selected from sizes stored in advance according to the type of input image. Alternatively, it may be set.

FIG. 6 shows a filter used in this embodiment. This filter differs from that shown in FIG.
The above-mentioned degree of concentration is evaluated only with pixels on radial direction lines in M types of directions (in FIG. 6, 32 directions every 11.25 degrees) adjacent at an angular interval of 2π / M degrees with respect to the target pixel. Things.

The number of direction lines set by the direction line setting means 23 is not limited to 32 shown in FIG.
If the number is set too large, the load required for the calculation processing increases rapidly. If the number is too small, the contour shape of the tumor shadow candidate cannot be detected with high accuracy. In addition, it is convenient in calculation processing or the like that the angular intervals between the direction lines are equal.

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

(Equation 3)

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

Further, for each direction line, the position of a predetermined pixel is set and changed according to the size of the tumor shadow to be detected, and a predetermined output value from the target pixel to the predetermined pixel is changed. For each set position, a predetermined output value based on the gradient vector is obtained, the maximum value of the predetermined output values for each set position is obtained, and the maximum value obtained for each direction line is calculated for all the direction lines. Average,
This average value is defined as the degree of concentration C of the gradient vector group for the target pixel.

Specifically, the maximum value calculating means 26 sets the starting point to the pixel of interest and sets the ending point n to the distance Rm corresponding to the minimum size of the tumor shadow candidate for each direction line.
The start point (target pixel) for the set end point n set within the range from in to the distance Rmax corresponding to the maximum size
The index value co of each pixel on the direction line within the range of the end point n from
The average value Ci (n) of s θil is determined by the following equation (5). This equation (5) means that the average value of the gradient vector in the target pixel direction component within the range from the start point to the end point n is obtained.

(Equation 4)

Then, the position of the end point is changed from the distance Rmin to the distance Rm.
The average value Ci (n) of the gradient vector at each end point is calculated in the same manner as in the above equation (5) while changing the value within the range up to ax.

Next, the maximum value calculating means 26 calculates the average values Ci obtained for the positions of the end points for each direction line.
The maximum value of (n) is extracted based on the following equation (6). This maximum value is defined as the degree of concentration Cimax for the direction line i.
And Then, in the gradient vector group concentration degree calculating means 27, a value obtained by averaging the concentration degrees Cimax for the direction lines i obtained by the expression (6) based on the expression (7) for all the direction lines, that is, for each direction line Of concentration C extracted in
imax is averaged over all 32 direction lines ((Σ
Cimax) / 32) to calculate the degree of concentration C of the gradient vector group for the target pixel. Note that the calculated value of the degree of concentration C is also referred to as an output value I of the iris filter processing.

(Equation 5)

The manner in which the size and shape of the area for evaluating the degree of concentration C of the gradient vector group of the equation (7) adaptively changes according to the distribution of the gradient vector is enlarged according to the brightness of the outside world. Since it resembles the state of the iris of the human eye that shrinks, the above-described method of detecting a candidate region of a tumor shadow using the degree of concentration of the gradient vector is called iris filter processing. .

Here, Cimax in the equation (6) is the maximum value of the average value Ci (n) of the component of the gradient vector for each direction line obtained in the equation (5), which is the pixel of interest. The region up to the pixel where the average value Ci (n) becomes the maximum value is a candidate region of the tumor shadow in the direction of the direction line, and the position of the pixel having the maximum value indicates the edge point of the region of the tumor shadow. Becomes Therefore, the calculation of Expression (6) is performed for all the radial direction lines, the edge points of the tumor shadow area on each direction line are obtained, and the adjacent edge points of the tumor shadow area on each direction line are determined. , 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.

Therefore, the comparing and judging means 31 compares a preset threshold value T suitable for judging whether or not the image is a tumor shadow and the value of the degree of concentration C (that is, the output value I of the iris filter processing). As a result of the comparison, when the value of the degree of concentration C is equal to or more than the threshold T (C ≧ T), it is determined that the target pixel exists in the area of the tumor shadow candidate, and the value of the degree of concentration C is less than the threshold T. If (C <T), it is determined that the pixel of interest does not exist in the abnormal shadow candidate area, and the result of this determination is input to the extraction means 32.

The threshold value T to be compared in the comparison / determination means 31 is set in advance. The term "in advance" means "by the time the final comparison is performed". Also in the embodiment, several different threshold values are prepared, and the threshold value of each level is experimentally applied, and the threshold value is such that the number of detected tumor shadow candidates is 7 to 10. Is finally adopted, and is used in the meaning including the final adoption.

The extracting means 32 determines, based on the judgment result input from the comparing and judging means 31, a pixel adjacent to the tumor shadow area on each direction line for the pixel of interest determined to be present in the abnormal shadow candidate area. By connecting the edge points with a straight line or a non-linear curve, the outline of a region that can be a candidate for a tumor shadow is specified. At this time, there may be a case where a large number of pixels of interest determined to be present in the region of the abnormal shadow candidate appear in close proximity. In this case, for example, the circumscribed region of each contour line by each pixel of interest is represented by one tumor shadow. What is necessary is just to make the outline of the area | region which becomes a candidate. Thereby, an abnormal shadow candidate area that is an area including the abnormal shadow candidate in the image is extracted. The extraction means 32 inputs all the pixel data of the extracted abnormal shadow candidate area to the local image processing means 30 as local image data for the abnormal shadow candidate area.

As shown in FIG. 3, when a tumor shadow exists at the end of the image of the original image, specifically, from the image edge of the original image, the tumor shadow of the tumor shadow to be detected in the iris filter processing is detected. When a tumor shadow exists within the range up to the maximum value Rmax of the radius, the direction line of the iris filter process reaches the image edge before reaching the maximum value Rmax of the radius of the tumor shadow. Since the average value Ci (n) shown in the expression (5A) described later is calculated within the range of the distance to the edge, the degree of concentration for the direction line, that is, the maximum value Cimax obtained by the expression (6) is the original value. The value is smaller than the maximum value when the image exists in an area other than the image edge of the image.

Since the iris filter processing is easily affected by the background image, if the direction line reaches the image edge before reaching the maximum value Rmax of the radius of the tumor shadow, the maximum value of the average value Ci (n) is obtained. The value Cimax may take a negative value.

That is, as shown in FIG. 7, for example, when a shadow of a blood vessel overlaps an abnormal shadow and a direction line overlaps the two shadows, the calculation result (Cimax) for this direction line is When the pixel reaches the distance Rmax from the pixel of interest, a positive value is used. However, when the pixel reaches the edge of the image before the pixel reaches the distance Rmax from the pixel of interest, a negative value may be used.

Therefore, if the degree of concentration C of the gradient vector group for the target pixel is obtained using the maximum value Cimax as the degree of concentration for each direction line, the degree of concentration in the case where the value also exists in an area other than the end of the image is obtained. It may be smaller than the threshold value T and may not be detected as a tumor shadow candidate.

Therefore, in order to solve this problem, the concentration calculating unit 20 of the tumor shadow candidate area extracting apparatus 2 according to the present invention includes the minimum size which is the information on the detection size inputted from the detection size setting means 22. Based on the (radius Rmin), the maximum size (radius Rmax), and information on each direction line input from the direction line setting means 23, a number of (32 in the above example) directions radially extending from the pixel of interest. It is determined whether or not there is a direction line that reaches the image edge within the range of the detection size, and the determination result is input to the maximum value calculation means 26.
Is provided.

The maximum value calculating means 26 determines the direction line reaching the image edge within the range of the detected size based on the judgment result inputted from the folding processing judging means 29. Subjected to reflection processing to reflect at the image edge at the same angle of reflection as the incident angle to
The gradient vector of the pixel on the direction line after the reflection processing is applied as a gradient vector for a virtual pixel in a portion beyond the image edge on the direction line, and the average value Ci is used.
The maximum value Cimax of (n) is obtained. This substitute gradient vector is hereinafter also referred to as a substitute gradient vector.

More specifically, since the range of the density gradient calculation in the ordinary iris filter processing is approximately 1 cm in radius, when the target pixel is present in an area within 1 cm from the edge of the image, the above-described reflection is used. By performing the processing, the concentration C of the density gradient vector group is obtained, and the concentration C is obtained in other regions as in the related art.

When the degree of concentration C of the density gradient vector group is obtained by using the substitute gradient vector obtained by performing the reflection processing as described above, the pixel data at a position far from the image edge and the actual pixel data are different. In contrast, since the actual gradient vector and the substitute gradient vector in the portion after the image edge are different from each other, it is actually difficult to substitute accurately. For this reason, the maximum value calculating means 26 performs weighting processing so that the degree of substitution of the gradient vector of the pixel after the image edge becomes smaller as the distance from the pixel of interest after the image edge increases, and the maximum value is calculated. Find Cimax.

On the other hand, by performing the above-described reflection processing to obtain the degree of concentration C instead of the gradient vector, the value of the degree of concentration C thus obtained is compared with a preset threshold T. When the abnormal shadow candidate region is extracted, the problem that the tumor shadow candidate existing near the image edge is not detected can be solved, but there is a possibility that a plurality of tumor shadow candidate regions are detected for one tumor shadow.

Therefore, the extraction means 32 of the present embodiment extracts an abnormal shadow candidate area which is an area including an abnormal shadow candidate in an image based on the comparison judgment result of each pixel inputted from the comparison judgment means 31. In doing so, the image is divided into an image region near the image edge and an image region other than near the edge according to the size of the tumor shadow to be detected, and for each of the divided regions, The maximum value of the number of abnormal shadow candidate areas that can be detected is set, and the abnormal shadow candidate areas are extracted within the range of the set maximum value. That is, a fixed number (up to the set maximum value) of candidate areas are independently detected from each area.

More specifically, an abnormal shadow candidate area is set so as to extract an abnormal shadow candidate area with an upper limit of two in a region up to the maximum radius Rmax of a tumor shadow to be detected from an image edge and five in other regions. I do.

As a result, a plurality of candidate areas are not detected for one shadow, and the position of the detected abnormal shadow candidate area can be prevented from being concentrated at the end of the image.

The image region is subdivided into three or more regions in accordance with the distance from the image edge to the nipple, and an abnormal shadow candidate that can be detected within the region is divided into each of the subdivided regions. May be set to the maximum value.

Further, the comparison / determination means 28 divides the area as described above, and sets the maximum value of the number of target pixels that can be detected as a pixel of the abnormal shadow candidate area in each of the divided areas. Alternatively, the number of determination results output from the comparison determination unit 28 to the extraction unit 32 as pixels in the abnormal shadow candidate region may be kept within the set maximum value.

Next, the operation of the abnormal shadow candidate detecting device 2 of the first embodiment will be described.

The whole image data S input from the storage means 3 to the abnormal shadow candidate detecting device 2 is first input to the gradient vector calculating means 10 and the target pixel setting means 24, and the gradient vector calculating means 10 performs the vertical 5 pixels x 5 horizontal
Using the image data (pixel value) of the outermost peripheral pixel of the pixel size mask, the direction θ of the density gradient vector is obtained for all the pixels. The obtained direction θ of the density gradient vector is input to the index value calculating means 25.

On the other hand, the pixel of interest setting means 24 sequentially sets one of all the pixels of the input image data as a pixel of interest, and inputs the set pixel of interest to the direction line setting means 23. The direction line setting means 23 sets 32 radial direction lines adjacent to each other at equal intervals of, for example, 11.25 degrees around the target pixel. Then, the set direction line is input to the index value calculating means 25.

The maximum size (radius Rmax) and the minimum value (radius Rmin) of the size of the tumor shadow candidate to be detected by the abnormal shadow candidate detecting device 2 are detected by the operator in the detection size setting means 22. Is entered. Then, the input maximum value Rmax and minimum value Rmin are input to the index value calculation unit 25, the maximum value calculation unit 26, and the return processing determination unit 29 as detection size information.

The index value calculating means 25 converts the gradient vector direction θ inputted from the gradient vector calculating means 12 into pixels which are defined in the same two-dimensional array as the image data (whole image data and duplicate image data). , 32 direction lines input from the direction line setting means 23 are superimposed,
Pixels that respectively overlap the book direction lines are extracted.

Further, the index value calculating means 25 calculates, for each direction line, for each pixel on the direction line within a range of a distance Rmax corresponding to the maximum size of the tumor shadow candidate from the target pixel.
An index value cos θil is obtained based on an angle θil between the gradient vector direction θ defined for each pixel and the direction in which the direction line extends.

Index values on each of the obtained direction lines
cos θil is input to the maximum value calculating means 26. The maximum value calculating means 26 sets the target pixel as a start point, sets an end point between a distance Rmin corresponding to the minimum size of the tumor shadow candidate and a distance Rmax corresponding to the maximum size, and changes the distance between the distances. The average value Ci (n) of the index value cos θil of each pixel within the range from the start point to the end point on the direction line is determined for each direction line, the maximum value Cimax of the average value Ci (n) is determined, and the maximum value Cimax is determined. Let the value Cimax be the degree of concentration for each direction line.

The gradient vector group concentration degree calculating means 27 performs averaging (simple averaging) on the maximum value Cimax obtained for each direction line based on the equation (7), and calculates the concentration of the gradient vector group for the pixel of interest. The value of the degree C is calculated. The calculated value of the degree of concentration C of the gradient vector group is input to the comparison determination unit 31.

By the way, there are various modes in the arrangement pattern of the tumor shadow when the tumor shadow appears near the edge of the image. However, there is a case where the center of the tumor shadow is located outside the image edge, that is, outside the original image area. The first arrangement pattern, the second arrangement pattern in which the center of the tumor shadow is located on the image edge, and the third arrangement pattern in which the center of the tumor shadow is located in the image edge, that is, in the original image area. Can be different. FIG.
(1) to (3) show the first to third arrangement patterns when the tumor shadow is assumed to be circular. FIG.
In the first arrangement pattern shown in (1), only an area less than half of the circular tumor shadow appears in the original image. On the other hand, in the second arrangement pattern shown in FIG. 8 (3), more than half the area of the circular tumor shadow appears in the original image.

Here, the average value Ci (n) becomes the maximum value Cimax
Is taken when the pixel at the end point is on the edge of the tumor shadow, as shown in FIG. 3 or FIG.
If present at the image edge, the end point of the direction line cannot be extended to the distance Rmax. For this reason, as shown in FIG.
The maximum value Cimax, that is, the degree of concentration for each direction line, is calculated within the range of the distance to the edge.
ax can only take a value smaller than the maximum value (original maximum value) when it exists in an area other than the image end of the original image. Also, the direction line is the maximum value R of the radius of the tumor shadow.
If the image edge is reached before reaching max, the maximum value Cimax of the average value Ci (n) may take a negative value.

In order to solve this problem, the abnormal shadow candidate detecting device 2 according to the first embodiment performs the following processing. First, as the gradient vector for the pixel of the portion a0 before the image edge in the direction line a, the gradient vector of each pixel is used as it is, while within the range of a predetermined distance according to the detection size of each direction line extending from the pixel of interest. , A direction line reaching the image edge, for example, FIG.
The reflection processing is performed on the direction line a extending in the lower left direction shown in FIG. 3 so that the direction line is reflected at the image edge at a reflection angle substantially equal to the incident angle to the image edge, and this reflection processing is performed. A gradient vector of a pixel (a pixel which overlaps with the pixel after the image edge when folded at the edge) at a position corresponding to the pixel after the image edge (virtual pixel) on the subsequent direction line a2 is calculated as the image edge. The edge is folded back as the axis of line symmetry, and the resulting folded gradient vector is used as a substitute gradient vector for the pixel of the portion a1 after the image edge in the direction line a to obtain the maximum concentration Ci (n). Value Cim
Find ax. That is, a folded gradient vector obtained by folding a gradient vector in the real image space into the virtual image space with the image edge as an axis of line symmetry is used as a gradient vector in the virtual image space.

In the first arrangement pattern, the true center of the tumor shadow does not appear in the real image space.
As shown in (1), since the center point of the tumor shadow on the edge of the image is the center of the tumor shadow candidate, when the pixel of interest is set at the center point, as shown in FIG. After setting the direction line a0 in the image space, it is virtually impossible to reflect the direction line a0 at the edge of the image, and the direction line extending into the virtual image space from the target pixel set at the center point a1 is immediately turned back with the image edge as the axis of line symmetry to take the direction line a2.

As a result, the calculation for obtaining the degree of concentration for each direction line, which has been calculated within the range up to the image edge, can be performed by expanding it to the range of the maximum value Rmax of the radius of the tumor shadow. The maximum value Cimax of the value Ci (n) can also take a larger value.

Also, the substitution degree of the substitution gradient vector is
The degree of substitution is smaller as the distance from the pixel of interest after the image edge increases, that is, when the distance is large, the degree of substitution is smaller than when the distance is small, and conversely, when the distance is small, the degree of substitution is smaller than when the distance is large. Weighting process so as to obtain the maximum value Cimax of the concentration Ci (n).
Ask for. Thereby, the value of the concentration C of the finally calculated density gradient vector group can be accurately obtained.

The same operation as described above is performed by sequentially changing the pixel of interest set by the pixel of interest setting means 24, and the value of the degree of concentration C for all the input pixels is input to the comparison determination means 31. .

The comparison / determination means 31 compares, for each pixel of interest (ie, all pixels), the value of the concentration C with the initially set threshold value T, and if C ≧ T, the pixel of interest Is determined to be a pixel existing in the region of the tumor shadow candidate,
If C <T, the pixel of interest is determined to be a pixel that does not exist in the region of the tumor shadow candidate, and the comparison determination result is input to the extraction unit 32.

The extracting means 32 extracts an abnormal shadow candidate area based on the comparison judgment result for each pixel inputted from the comparison judging means 31 and locally extracts the local image data which is the image data of the extracted abnormal shadow candidate area. Input to the image processing means 5. At this time, the image is divided into an image area near the image edge and an image area other than the area near the edge according to the size of the detection size, and the abnormal shadow that can be detected in the area for each area is divided. A maximum value of the number of candidate regions is set, and the abnormal shadow candidate region is extracted within the range of the set maximum value.

Here, in the case of the conventional abnormal shadow detecting device,
Since the tumor shadow existing near the image edge could not extend the end point of the direction line in the iris filter processing to the distance Rmax from the pixel of interest, the maximum value Cimax in that direction line became a small value. As a result, the iris filter There is a possibility that the value of the degree of concentration C, which is the output value, becomes a smaller value and falls below the threshold value T, and may not be detected as a tumor shadow candidate.
As described above, since the value of the degree of concentration C of the density gradient vector group is accurately determined by using the gradient vector of the corresponding pixel obtained by the reflection processing as described above, a tumor shadow exists near the image edge. Even if the entire shadow is not reflected in the image, it is possible to accurately determine the pixel of the tumor shadow candidate present near the image edge as a pixel of the tumor shadow candidate area, and as a result, It is possible to extract a tumor shadow candidate area with higher accuracy than before. Also, there is no need for a memory for storing the duplicated image data.

The local image data representing the tumor shadow candidate area accurately extracted by the extracting means 32 is displayed after the local image processing means 5 has performed an enlargement process or a frequency emphasis process which is optimal for observing the tumor shadow. It is input to the means 6.

On the other hand, a whole image signal which has been subjected to image processing suitable for observing the whole image, such as gradation processing and frequency emphasizing processing, by the whole image processing means 4 is also input to the display means 6. Means that while displaying the whole image, only the image portion of the tumor shadow candidate in the whole image is
The image is replaced with the image of the tumor shadow candidate subjected to the image processing and displayed, and is used for diagnosis of the tumor shadow by a doctor or the like who observes the image with image observation.

The arrangement pattern of abnormal shadows at the edge of the image which can sufficiently exhibit the above-mentioned effects of the apparatus of the first embodiment is, in particular, such that the center of the actual abnormal shadow candidate is located outside the edge of the image. It was found that this was the case of the first arrangement pattern shown in FIG. This will be described with reference to FIG.

FIG. 10 is a diagram showing the directions of the substitute gradient vectors when the first embodiment is applied to each of the first to third arrangement patterns.

For example, when an ideal abnormal shadow having a size of just Rmax is assumed, and the abnormal shadow of the third arrangement pattern shown in FIG.
/ 2 is the maximum, so that the above effects could not be obtained sufficiently.

This is because, as shown in FIG. 10 (3), for example, in the third arrangement pattern in which about 異常 of the abnormal shadows appear in the real image space, the virtual image space portion outside the image edge is removed. It has been found that the substitution gradient vector has a portion that is not directed to the center of the shadow and diverges with respect to the target pixel set at the center of the shadow, so that the degree of concentration does not become sufficiently high. However, even in this case, the degree of concentration does not become negative (minus) in each direction.

On the other hand, as shown in FIG. 10A, for example, in the second arrangement pattern in which approximately half of the first arrangement pattern in which about one-third of the abnormal shadow appears in the real image space appears. Such a problem does not occur because the substitute gradient vector in the virtual image space portion always faces the target pixel direction. Since the direction is slightly shifted, the component in the direction of the target pixel becomes small, but there is no problem.

As described above, according to the present invention, the return gradient vector folded at the end of the image is used as a substitute, and the degree of substitution of the gradient vector of the pixel after the edge is away from the pixel of interest after the edge. Since the maximum value Cimax is obtained by performing the weighting process so as to be smaller as much as possible, even in a case where an abnormal shadow exists at the image end and the entire shadow is not reflected in the image (particularly in the case of the first arrangement pattern), Improved extraction performance of abnormal shadow candidate area (detection rate TP
(Up, false detection rate FP down), and it is possible to improve the doctor's ability to assist diagnosis.

Further, since the image edge and the other image area are divided and the maximum value of the number of abnormal shadow candidates that can be detected is set for each area, the position extracted as an abnormal shadow candidate area is set. Can be prevented from being concentrated on the edge of the image.

It is to be noted that the maximum value calculating means 26 of the above-described embodiment sets the target pixel as the starting point. However, instead of setting the starting point as the target pixel, the maximum value calculating means 26 sets the minimum size of the set tumor shadow candidate. A pixel at the corresponding distance Rmin may be set as a start point.

In this case, the index value cos θil of each pixel on the direction line within the range from the start point to the end point, which is expressed by equation (5)
Is represented by equation (5A).

(Equation 6)

Equation (5A) shows that the degree of concentration Ci is 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.
(N) is calculated.

Although the first embodiment has been described with respect to the case where the target image is mammography,
In the method and apparatus for detecting each abnormal shadow candidate according to the present invention, the target image is not limited to mammography.

Next, an abnormal shadow candidate detecting apparatus according to the second embodiment will be described. The device configuration is the same as that of the abnormal shadow candidate detecting device of the first embodiment, but the processing method in the concentration degree calculating unit 20 is changed so as to be suitable for the abnormal shadow of the third arrangement pattern. Hereinafter, differences of the processing in the concentration degree calculation unit 20 of the second embodiment from the first embodiment will be described.

A portion a0 before the image edge in the direction line a
As in the first embodiment, the gradient vector of each pixel is used as it is as the gradient vector for the pixel. On the other hand, among the direction lines extending from the pixel of interest, a direction line reaching the image edge within a range of a predetermined distance according to the detection size, for example, a direction line a extending in the lower left direction shown in FIG. Is reflected at the image edge at a reflection angle substantially equal to the angle of incidence to the image edge, and the pixels after the image edge on the direction line a2 after the reflection process are performed ( A gradient vector of a pixel at a position corresponding to a virtual pixel) (a pixel overlapping with a pixel after the image edge when folded at the edge) is used as a substitute gradient for a pixel of a part a1 after the image edge in the direction line a. Using this as a vector, the maximum value Cimax of the degree of concentration Ci (n) is obtained. That is, the gradient vector in the real image space at the corresponding position where the image edge is the axis of line symmetry is used as it is as the gradient vector of the pixel in the virtual image space.

As a result, as shown in FIG. 12, in the case of an abnormal shadow of the third arrangement pattern, the substitute gradient vector within the processing target area outside the image edge (virtual image space portion) is set at the center of the shadow. Since it is oriented in the direction of the pixel of interest, the calculation for obtaining the degree of concentration for each direction line calculated within the range up to the image edge is performed in the same manner as in the first embodiment.
This can be performed by expanding to the range of the maximum value Rmax of the radius of the tumor shadow, and the maximum value Cimax of the average value Ci (n) can also take a larger value.

In the apparatus according to the second embodiment, the arrangement pattern of abnormal shadows at the edge of the image which can sufficiently exhibit the above-mentioned effects is, particularly, that the center of the actual abnormal shadow candidate is within the edge of the image. It was found that this was the case of the third arrangement pattern shown in FIG. This will be described with reference to FIG.

FIG. 13 shows the maximum value R of the radius of the tumor shadow to be processed when the second embodiment is applied to each of the first to third arrangement patterns.
It is a figure showing the direction of a substitute gradient vector on a predetermined direction line up to the range of max.

For example, as shown in FIG.
Assuming an ideal abnormal shadow of max, when the abnormal shadow of the third arrangement pattern shown in FIG.
As shown in FIG. 2 and FIG. 13 (3), since the substitute gradient vector in the processing target range of the virtual image space portion is directed to the center of the shadow, the above effect can be sufficiently enjoyed. In addition,
Since the direction is slightly shifted, the component in the direction of the pixel of interest becomes small, but there is no problem.

On the other hand, in the case of an abnormal shadow of the first arrangement pattern shown in FIG. 8A or the second arrangement pattern shown in FIG. 8B, as shown in FIGS. In the meantime, the substitute gradient vector in the processing target range of the virtual image space portion outside the image edge has a portion that is not directed to the shadow center, and becomes a direction that diverges with respect to the target pixel set at the shadow center, A problem that the above effects cannot be obtained may occur.

Next, an abnormal shadow candidate detecting device according to a third embodiment will be described.

The abnormal shadow candidate detecting device of the third embodiment is
The processing method in the degree-of-concentration calculation unit 20 is changed so as to be adapted to each abnormal shadow of each of the above arrangement patterns. Hereinafter, points of the abnormal shadow candidate detecting device according to the third embodiment that are different from the above embodiments will be described. There is no difference in the configuration itself.

As can be understood from the description of the first embodiment and the second embodiment, the processing of the degree of concentration calculation unit 20 in each embodiment includes an arrangement pattern of abnormal shadows that is good at sufficiently exhibiting effects. is there. Further, as can be understood from the above description, the problem can be solved if the substitute gradient vector in the processing target range in the virtual image space portion outside the image edge is directed to the center of the shadow.

Therefore, in the third embodiment, first, when the direction line is reflected at the edge, the distance between the pixel on the direction line and the edge is equal to or larger than the distance between the target pixel and the edge. It is determined whether or not there is, and when the determination result is “not less than”, the same processing as in the first embodiment is performed, and when the determination result is “less than”, the same processing as in the second embodiment is performed. In other words, this is a method in which vector folding is not performed only for a certain region (actually, a target pixel-image edge distance) from the image edge in the virtual image space. The division processing according to such a determination result is performed by the return processing determination means 29.

According to such a configuration, as shown in FIG. 14, even in the third arrangement pattern, it corresponds to the distance between the target pixel set at the center of the original tumor shadow and the image edge. Since the gradient vector of the region in the virtual image space is generally oriented in the direction of the pixel of interest, the substitute gradient vector can be used in accordance with the arrangement pattern of the abnormal shadow at the image edge, and as a result, the image edge Regardless of each arrangement pattern at the edge, each of the above-described effects as each device of the first embodiment or the second embodiment can be sufficiently exhibited. It should be noted that applying the processing method of the third embodiment to the first or second arrangement pattern does not cause any problem.

That is, depending on the arrangement pattern of the abnormal shadow, whether to use the return gradient vector at the corresponding pixel position (in the case of the first or second arrangement pattern) or whether to use the gradient vector as it is (third Can be used effectively, and a tumor shadow candidate area corresponding to a tumor shadow located at the edge of the image can be extracted with extremely high accuracy regardless of the layout pattern.

It is to be noted that any method may be used for generating the folded gradient vector. For example, it is also possible to use a method in which an image in the real image space is turned back using the image edge as a line symmetry axis to generate and obtain a duplicate image.

[Brief description of the drawings]

FIG. 1 is a block diagram showing a first embodiment of an abnormal shadow candidate detecting device according to the present invention;

FIG. 2 is a block diagram showing a schematic configuration of an example of a computer-aided image diagnostic apparatus including the abnormal shadow candidate detecting apparatus shown in FIG. 1;

FIG. 3 is a diagram showing a radiation image used in the computer-aided image diagnostic apparatus shown in FIG. 2;

FIG. 4 is a diagram showing a mask for calculating a gradient vector in iris filter processing.

FIG. 5 is a diagram showing the concept of the degree of concentration of a gradient vector for a target pixel.

FIG. 6 is a conceptual diagram showing an iris filter set so that a contour shape is adaptively changed.

FIG. 7 is a view showing an image in which a blood vessel shadow is superimposed on an abnormal shadow and recorded.

FIG. 8 is a diagram showing an arrangement pattern of a tumor shadow and a direction of a gradient vector when the tumor shadow is assumed to be circular;

FIG. 9 is a diagram illustrating the folding of the direction line in the first embodiment.

FIG. 10 is a diagram showing directions of gradient vectors when the first embodiment is applied to each of the first to third arrangement patterns shown in FIG. 8;

FIG. 11 is a diagram illustrating the folding of the direction line in the second embodiment.

FIG. 12 is a view for explaining the effect of the second embodiment when an abnormal shadow exists in a third arrangement pattern;

FIG. 13 is a diagram illustrating the direction of a gradient vector when the second embodiment is applied to each of the first to third arrangement patterns illustrated in FIG. 8;

FIG. 14 is a diagram illustrating a direction of a gradient vector when a third embodiment is applied to a third arrangement pattern;

[Explanation of symbols]

 DESCRIPTION OF SYMBOLS 1 Computer-assisted image diagnostic apparatus 2 Abnormal shadow candidate detecting apparatus 3 Storage means 4 Overall image processing means 5 Local image processing means 6 Display means 10 Gradient vector calculating means 20 Concentration calculating section 22 Detection size setting means 23 Direction line setting means 24 Attention Pixel setting means 25 Index value calculation means 26 Maximum value calculation means 27 Gradient vector group concentration degree calculation means 29 Loopback processing determination means 30 Candidate detection means 31 Comparison determination means 32 Extraction means

Continued on front page F term (reference) 4C093 AA01 CA02 CA21 FD05 FD09 FD11 FF11 FF17 FF20 FF22 FF27 5B057 AA08 CA02 CA08 CA12 CA16 CB02 CB08 CB12 CB16 CC02 CE06 CH09 DA08 DB02 DB05 DB09 DC22

Claims (10)

[Claims] 1. A gradient vector for each pixel based on the image data for all pixels of image data representing an image, and a predetermined distance corresponding to the size of an abnormal shadow candidate to be detected centered on the predetermined pixel The concentration degree indicating the degree of distribution of the direction of the local gradient vector centered on the predetermined pixel based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within the range of In the abnormal shadow candidate detecting method for detecting the abnormal shadow candidate in the image by comparing the value of the degree of concentration with a preset threshold value, the edge of the image within the range of the predetermined distance As the gradient vector of the pixel after the edge with respect to the direction line that reaches the pixel, corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge. The abnormal shadow candidate detecting method characterized by by substituting folded gradient vectors obtained by folding the gradient vector of the location of the pixel to said end edge as the axis of the line symmetry determining the degree of concentration. 2. For all pixels of image data representing an image, a gradient vector is determined for each pixel based on the image data, and a predetermined distance centered on the predetermined pixel and corresponding to the size of an abnormal shadow candidate to be detected is determined. The concentration degree indicating the degree of distribution of the direction of the local gradient vector centered on the predetermined pixel based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within the range of In the abnormal shadow candidate detecting method for detecting the abnormal shadow candidate in the image by comparing the value of the degree of concentration with a preset threshold value, the edge of the image within the range of the predetermined distance As the gradient vector of the pixel after the edge with respect to the direction line that reaches the pixel, corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge. The abnormal shadow candidate detecting method characterized by determining the degree of concentration by substituting the gradient vector of the location of the pixel. 3. A gradient vector for each pixel based on the image data is determined for all the pixels of the image data representing the image, and a predetermined distance centered on the predetermined pixel is determined according to the size of the abnormal shadow candidate to be detected. The concentration degree indicating the degree of distribution of the direction of the local gradient vector centered on the predetermined pixel based on the gradient vector of each pixel on a number of direction lines extending radially from the predetermined pixel within the range of In the abnormal shadow candidate detection method for detecting the abnormal shadow candidate in the image by comparing the value of the degree of concentration with a preset threshold, when the direction line is reflected at the edge When the distance between the pixel on the direction line and the edge is equal to or greater than the distance between the predetermined pixel and the edge, the edge on the direction line after reflecting the direction line at the edge A folded gradient vector obtained by folding a gradient vector of a pixel at a position corresponding to a subsequent pixel with the edge as an axis of line symmetry is used for a direction line reaching the edge of the image within the range of the predetermined distance. The concentration is obtained by substituting the gradient vector of the pixel after the edge with the gradient, and when the direction line is reflected at the edge, the distance between the pixel on the direction line and the edge is the predetermined pixel. When less than the distance from the edge, the gradient vector of a pixel at a position corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge is the gradient of the pixel after the edge. A method for detecting an abnormal shadow candidate, wherein the degree of concentration is obtained by substituting the vector as a vector. 4. The method according to claim 1, wherein a weighting process is performed so that the degree of substitution decreases as the distance from the predetermined pixel after the edge increases. 3. The method for detecting an abnormal shadow candidate according to claim 3. 5. The image is divided into an image region near an edge of the image and an image region other than the vicinity of the edge according to the size of the predetermined distance, and the image is divided for each of the divided regions. The abnormal shadow candidate detecting method according to claim 1, wherein a value corresponding to a maximum value of the number of regions including the abnormal shadow candidate that can be detected in the region is set. 6. A gradient vector calculating means for calculating a gradient vector for each pixel based on the image data for all pixels of image data representing an image, and a size of an abnormal shadow candidate to be detected centered on a predetermined pixel. Based on the gradient vectors of each pixel on a number of direction lines extending radially from the predetermined pixel within a predetermined distance range corresponding to the degree of distribution of the direction of the local gradient vector centered on the predetermined pixel A degree-of-concentration calculating means for obtaining a degree of concentration, a comparing means for comparing a value of the obtained degree of concentration with a preset threshold value, and the abnormal shadow in the image based on an output from the comparing means. An abnormal shadow candidate detecting device including candidate detecting means for detecting a candidate, wherein the concentration degree calculating means detects a direction line reaching an edge of the image within the range of the predetermined distance. As the gradient vector of the pixel after the edge, the gradient vector of a pixel at a position corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge is line-symmetric with respect to the edge. An abnormal shadow candidate detecting apparatus for obtaining the degree of concentration by using a folded gradient vector obtained by folding as an axis of. 7. A gradient vector calculating means for calculating a gradient vector for each pixel based on the image data for all pixels of image data representing an image, and a size of an abnormal shadow candidate to be detected centered on a predetermined pixel. Based on the gradient vectors of each pixel on a number of direction lines extending radially from the predetermined pixel within a predetermined distance range corresponding to the degree of distribution of the direction of the local gradient vector centered on the predetermined pixel A degree-of-concentration calculating means for obtaining a degree of concentration, a comparing means for comparing a value of the obtained degree of concentration with a preset threshold value, and the abnormal shadow in the image based on an output from the comparing means. An abnormal shadow candidate detecting device including candidate detecting means for detecting a candidate, wherein the concentration degree calculating means detects a direction line reaching an edge of the image within the range of the predetermined distance. As the gradient vector of the pixel after the edge, the gradient vector of the pixel at a position corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge is used as the gradient vector. An abnormal shadow candidate detection device, characterized in that it obtains the following. 8. A gradient vector calculating means for calculating a gradient vector for each pixel based on the image data for all pixels of image data representing an image, and a size of an abnormal shadow candidate to be detected centered on a predetermined pixel. Based on the gradient vectors of each pixel on a number of direction lines extending radially from the predetermined pixel within a predetermined distance range corresponding to the degree of distribution of the direction of the local gradient vector centered on the predetermined pixel A degree-of-concentration calculating means for obtaining a degree of concentration, a comparing means for comparing a value of the obtained degree of concentration with a preset threshold value, and the abnormal shadow in the image based on an output from the comparing means. An abnormal shadow candidate detecting device including candidate detecting means for detecting a candidate, wherein the concentration degree calculating means includes a pixel on the direction line when the direction line is reflected at the edge. When the distance to the edge is equal to or longer than the distance between the predetermined pixel and the edge, a pixel at a position corresponding to a pixel after the edge on the direction line after reflecting the direction line at the edge. A gradient vector obtained by folding the gradient vector with the edge as an axis of line symmetry is a gradient vector of a pixel after the edge with respect to a direction line reaching the edge of the image within the predetermined distance. When the degree of concentration is obtained in place of the above, when the distance between the pixel on the direction line and the edge is less than the distance between the predetermined pixel and the edge when the direction line is reflected at the edge, The gradient degree of the pixel at a position corresponding to the pixel after the edge on the direction line after reflecting the direction line at the edge is substituted for the gradient vector of the pixel after the edge, and the degree of concentration is calculated. An abnormal shadow candidate detection device, which is to be obtained. 9. The degree of concentration is obtained by performing weighting processing such that the degree of substitution becomes smaller as the distance of the pixel after the edge from the predetermined pixel increases. 7. The method according to claim 6, wherein
9. The abnormal shadow candidate detection device according to any one of items 1 to 8. 10. The candidate detecting means divides the image into an image area near an edge of the image and an image area other than near the edge according to the size of the predetermined distance, and For each region, set a value corresponding to the maximum value of the number of regions including the abnormal shadow candidate that can be detected in the region,
10. The apparatus according to claim 6, wherein the abnormal shadow candidate is detected within the set value range.
An abnormal shadow candidate detection device according to the item.