JP2005065855A - Mammary gland region threshold decision method, device and program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To calculate the threshold of the density of a mammary gland region capable of accurately separating the mammary gland region in the side image of breasts. <P>SOLUTION: A histogram preparation means 130 prepares the histogram Hs of the density value in a sampling region S on a pectoral muscle end region in the side face image P of the breast set by a sampling region setting means 120, and a threshold calculation means 140 calculates the threshold Th of the density value of the mammary gland region according to the following equation using the density value p1 equivalent to the boundary of a distribution Ha indicating pectoral muscles on the histogram Hs and a distribution Hb indicating fat and an average density value p2 in the distribution Ha indicating the pectoral muscles. In this case, the equation is expressed as Th=a×p1+(1-a)×p2 (where 0.2≤a≤0.5). <P>COPYRIGHT: (C)2005,JPO&NCIPI

Description

  The present invention relates to a mammary gland region threshold value determining method, apparatus and program for calculating and determining a density threshold value for extracting a mammary gland region in a side image of a breast, and an apparatus using the method.

  Conventionally, in the medical field, mammography is performed as one technique for detecting breast cancer. Mammography is a mammogram (diagnostic radiation image with the breast as a subject) taken for the purpose of diagnosing breast cancer, and abnormalities such as tumor shadows and microcalcification shadows that are one of the features of cancerous parts from the images This is a technique for detecting shadows.

  In the mammogram, since the mammary gland and the lesion appear at close concentrations, it is difficult to contrast the mammary gland and the lesion, and when there is a lesion in the mammary gland region, it is difficult to detect the lesion. Therefore, for example, grasping the distribution of the mammary gland region in the mammogram and knowing the type classified according to the mammary gland mass in the mammogram are useful for knowing the difficulty level of the lesion detection, and the detection performance of the lesion It leads to improvement.

For this reason, a method for automatically classifying mammograms by quantitatively analyzing the amount and distribution of mammary gland parenchyma in the breast using a computer has been proposed (Non-Patent Document 1). This method uses the variance histogram analysis and discriminant analysis method to divide the breast area into three areas according to the amount of mammary gland parenchyma, and the mammogram is divided into four types according to the ratio of the divided areas (specified in the mammography guidelines). 1) Fatty, 2. Mammal scattered, 3. Non-uniform high concentration, 4. High concentration.
Matsubara Tomoko, Yamazaki Daisuke, Kato Yudai, et al .: Automatic mammogram classification based on mammary gland concentration evaluation, Journal of the Institute of Electronics, Information and Communication Engineers, MI2000-79, 11-15, 2001-01.

  However, the automatic mammogram classification method can divide a breast region into a plurality of regions corresponding to the amount of mammary gland parenchyma, but cannot extract only the mammary gland region. In the classification method, the concentration threshold value used for extracting each region is calculated based only on the concentration of the pectoral muscle, or from the concentration histogram in the region where the mammary gland and fat that are likely to vary in concentration are mixed. Since the threshold is calculated by discriminant analysis, the calculated threshold value is easily affected by variations in the density distribution of the mammary gland, fat, or pectoral muscle, and the accuracy of the classification result of the extracted region and mammogram is not stable. There is a problem.

  In view of the above circumstances, the present invention is a method and apparatus for determining a threshold value of a mammary gland region that can determine a threshold value of density capable of accurately extracting a mammary gland region in a breast image, and a program therefor, and further uses this method. An object of the present invention is to provide an apparatus that generates information useful for detecting abnormal shadows in breast images.

The method for determining a threshold value of a mammary gland region according to the present invention sets a sampling region as a region substantially including pectoral muscles and fat in the side image represented by the side image of the breast comprising the mammary gland, fat and pectoral muscles. Based on the density distribution in the sampling area, a density p1 corresponding to the boundary between the pectoral muscle part and the fat part in the sampling area and an average density p2 of the pectoral muscle part are obtained, and the mammary gland in the side image A threshold value of a concentration for separating a mammary gland region representing fat and a fat region representing fat is expressed by the following formula using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
It is a method characterized by calculating and determining according to.

Then, the mammary gland region threshold value determining device of the present invention comprises image data input means for inputting image data carrying a side image of a breast made up of mammary gland, fat and pectoral muscles, and a substantially pectoral muscle and fat in the side image. A sampling area setting means for setting a region to be a sampling area, a density p1 corresponding to a boundary between a pectoral muscle part and a fat part in the sampling area based on a distribution of density in the sampling area, and the pectoral muscle part And using the obtained p1 and p2 as threshold values for the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
And a threshold value calculation means for calculating and determining according to the above.

  Here, the “thoracic muscle region” refers to the body side (trunk) portion (FIG. 3Pa) that appears in the side image of the breast, and the “breast region” refers to the breast from the nipple in the side image of the breast. It is a part (FIG. 3Pc) representing the mammary gland extending toward the center, and the “fat region” represents fat excluding the pectoral muscle region and the mammary gland region in the subject region (region other than the background) in the side image of the breast. This is a portion (FIG. 3Pb).

  Note that the shapes and ratios of the “mammary gland region” and “fat region” vary greatly depending on the breast as the subject, and FIG. 3 is a diagram schematically showing an example thereof.

  “Density” represents the degree of blackening in an image. For example, the density value (pixel value) of each pixel constituting the image can be considered. The correspondence relationship between the level of density value of a pixel and the degree of blackening of the pixel may be defined in any way. However, in this specification, the higher the density value, the blacker the color, and the smaller the density value, the whiter the level. Shall be.

  As the “region consisting of substantially pectoral muscles and fat”, for example, a region including the boundary between pectoral muscles and fat can be considered. Since such a region is not a region where the mammary gland and fat tend to vary in density, it is a stable region with little variation in density in the breast image. It is preferable that the “region consisting of substantially pectoral muscle and fat” does not include the mammary gland as much as possible, but some mammary glands are included as long as they do not adversely affect the above-described p1 and p2. It does not matter.

  In the mammary gland region threshold value determining method and apparatus of the present invention, the “region substantially consisting of pectoral muscle and fat” may be a pectoral muscle end region representing the end of the pectoral muscle.

  The “pectoralis end region” is a region near the end on the upper chest side (close to the clavicle) of the pectoral muscle line, which is a boundary line between the pectoral muscle region and the other breast region.

  At this time, in the mammary gland region threshold value determining method of the present invention, the side image is such that the pectoral muscle end region is arranged at a predetermined location in the side image, and the predetermined location is set as the sampling region. Good. Similarly, in the mammary gland region threshold value determining apparatus of the present invention, the side image is obtained by arranging the pectoral muscle end region at a predetermined location in the side image, and the sampling region setting means samples the predetermined location. It may be set as an area.

  The “predetermined location” can be, for example, the lower right part or the lower left part in the side image. It can be set as an area surrounded by a predetermined frame. Since the mammogram is usually photographed in accordance with a guideline that defines the arrangement of the breast, the pectoral muscle end region is arranged at a predetermined location in the image.

  In the mammary gland region threshold value determining method of the present invention, p1 and p2 may be obtained based on a density histogram in the sampling region. Similarly, in the mammary gland region threshold value determining device of the present invention, threshold value calculating means May determine p1 and p2 based on a histogram of density in the sampling region.

  As a method for obtaining p1 and p2, in addition to the above, a method using a cumulative histogram of density in the sampling region, or data corresponding to the sampling region in the image data is input, and p1 and p2 are output. A method using such a neural network can be considered.

  In addition, as a specific method for obtaining the above-described p1 and p2, for example, when obtaining based on a density histogram in the sampling region, the following method can be considered.

  The set “sampling area” includes the pectoral muscle part and the fat part. However, since the density value differs between the pectoral muscle part and the fat part, the “histogram” of the density in the sampling area The “distribution representing pectoral muscles” and the “distribution representing fat” appear at different concentrations. Note that which distribution corresponds to the pectoral muscle or fat can be determined from the level relationship of the density corresponding to the pectoral muscle and fat in the side image. For example, when the side image is a general X-ray image or the like, the density of the pectoral muscle region is low and the density of the fat region appears high, so the distribution with the lower concentration represents the pectoral muscle and the concentration is high. It can be judged that the distribution of the direction represents fat.

  “Density p1” is a method of adopting the least frequently used concentration between the distribution representing the pectoral muscle and the distribution representing fat on the histogram, or the intersection of two curves approximating the shapes of these two distributions. It can be obtained by a technique that employs a density corresponding to the position of.

  “Average density p2” is to calculate the average for the distribution representing the pectoral muscle on the histogram, that is, the density that contributes to the distribution in the region where the density is higher or lower in the predetermined direction from the density p1. It can ask for.

  The above formula has been determined based on the result of the applicant's finding that the threshold value of the density of the mammary gland region is within a predetermined range between the density p1 and the density p2 by the applicant. It is.

  In the experiment conducted by the applicant, the most suitable threshold value could be obtained when a = 0.4 in the above formula. Therefore, in the method and apparatus for determining a mammary gland region threshold value, it is desirable that the above equation is a = 0.4.

  In the mammary gland region threshold value determining method and apparatus, the side image may be a medical image.

  A “medical image” is an image used for diagnosis such as detection of a lesioned part or grasping of the progress of a disease, and an X-ray image, CT image, MRI image, or the like can be considered. As a side image of the breast that is a medical image, an X-ray image of a breast, which is generally taken for the purpose of examining breast cancer, a so-called mammogram can be considered. Mammograms usually have a front image taken from above with the breast sandwiched from the top and bottom and a side image taken from the side with the breast sandwiched from the left and right. The side image can be used as the “side image of the breast”.

The breast image region separation device of the present invention represents image data input means for inputting image data carrying a side image of a breast composed of mammary glands, fat and pectoral muscles, and represents a breast in the side image based on the image data. Subject region extraction means for extracting a subject region; and pectoral muscle region separation means for separating a pectoral muscle region representing the pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data. In the side image, sampling region setting means for setting a region substantially consisting of pectoral muscle and fat as a sampling region, and based on the distribution of concentration in the sampling region, A density p1 corresponding to the boundary of the breast muscle, and an average density p2 of the pectoral muscle portion, and a mammary gland region representing the mammary gland in the side image The threshold concentrations of separating the fat region representing the 肪 the following equation using the p1 and p2 obtained:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold calculating means for calculating and determining according to the above, and a mammary fat area separating means for separating a breast area obtained by removing the pectoral muscle area from the subject area into a mammary gland area and a fat area based on the threshold value. It is a feature.

  As the “subject area extraction means”, for example, when the side image has a high density in the background area and a low density in the subject area, a histogram of the density in the side image is created. The maximum density excluding the distribution corresponding to the background portion of the image can be considered as a threshold value, and an area having a density lower than the threshold value can be extracted as the subject area.

  The “pectoral muscle region separation means” includes boundary point extraction means for extracting one or more boundary points on the boundary line between the pectoral muscle region and the breast region in the side image based on the image data; It can be provided with a pectoral muscle line extracting means for extracting a pectoral muscle line by a predetermined method according to the number.

  The “boundary point extraction means” creates an edge image of the side image based on the image data, sets a plurality of scan lines in the upper breast region in the created edge image, and scans the set scan lines. Thus, edge points can be extracted, and the extracted edge points can be used as boundary points.

  In addition, when the number of extracted boundary points is 3, the “pectoral muscle line extracting means” uses a quadratic curve passing through the three boundary points as a pectoral muscle line, and the number of extracted boundary points is 2. When the number of boundary points extracted is one, the linear line passing through these two boundary points is a pectoral muscle line. When the number of extracted boundary points is one, A primary straight line passing through the lowest point of the lowered perpendicular line can be a pectoral muscle line.

  For details of the method of extracting the pectoral muscle line by the “boundary point extracting means” and the “pectoral muscle line extracting means”, refer to Japanese Unexamined Patent Publication No. 2003-115041.

The breast image classification apparatus according to the present invention includes an image data input unit that inputs image data carrying a side image of a breast composed of a mammary gland, a fat, and a pectoral muscle, and a subject that represents a breast in the side image based on the image data Subject region extraction means for extracting a region, and pectoral muscle region separation means for separating a pectoral muscle region representing the pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data; In the side image, sampling region setting means for setting a region consisting of approximately pectoral muscles and fat as a sampling region, and a boundary between the pectoral muscle portion and the fat portion in the sampling region based on the concentration distribution in the sampling region Is obtained, and the breast muscle region representing the mammary gland and the fat representing fat in the side image are obtained. Equation a threshold concentration to separate the region, with p1 and p2 obtained:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold calculating means for calculating and determining according to the above, a breast area extracting means for extracting a mammary gland area in the breast area excluding the pectoral muscle area from the subject area, and a mammary gland for the subject area or breast area based on the threshold value Image classification means for classifying the side images according to the proportion of the area is provided.

For example, four types proposed in the mammography guideline are considered as types to be “classified”, and “high concentration”, “non-uniform high concentration”, “mammary gland scattering”, Can be “fatty”. The above four types are defined in the guidelines as follows.
1) High concentration: Almost no fat is mixed in the mammary gland parenchyma, and the lesion detection rate is low.
2) Non-uniform high concentration: Fat is mixed in the mammary gland parenchyma and exhibits a non-uniform concentration. There is a risk that the lesion is hidden in the normal mammary gland.
3) Mammary gland scattering: Mammary gland parenchyma is scattered in the breast replaced with fat. Detection of the lesion is relatively easy.
4) Fatty: The breast is almost completely replaced by fat. If the lesion is within the imaging range, it is easy to detect the lesion.
“Density” in the above-mentioned types of “high density” and “nonuniform high density” does not indicate the degree of blackening on the image but indicates the amount of mammary glands in the breast.

The mammary fat region average concentration calculating apparatus according to the present invention includes an image data input unit that inputs image data carrying a side image of a breast composed of a mammary gland, a fat, and a pectoral muscle, and a breast in the side image based on the image data. A subject region extracting means for extracting a subject region representing the pectoral muscle, and a pectoral muscle region separation for separating a pectoral muscle region representing the pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data Means, a sampling area setting means for setting an area substantially consisting of pectoral muscles and fat in the side image as a sampling area, and a pectoral muscle portion and fat in the sampling area based on a concentration distribution in the sampling area Milk representing the mammary gland in the side image is obtained by calculating a density p1 corresponding to the boundary with the part and an average density p2 of the pectoral muscle part. The threshold concentrations of separating the fat region representing the region and fat, the following formula using the p1 and p2 obtained:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold calculating means for calculating and determining according to the above, a mammary fat area separating means for separating a breast area obtained by removing the pectoral muscle area from the subject area into a mammary gland area and a fat area based on the threshold value, and the image data On the basis of this, it is provided with an average concentration calculation means for calculating an average concentration in each of the separated mammary gland region and fat region.

The mammary gland distribution map creation device of the present invention represents image data input means for inputting side data of a breast composed of mammary glands, fat and pectoral muscles, and represents a breast in the side image based on the image data. Subject region extraction means for extracting a subject region; and pectoral muscle region separation means for separating a pectoral muscle region representing the pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data. In the side image, sampling region setting means for setting a region substantially consisting of pectoral muscle and fat as a sampling region, and based on the distribution of concentration in the sampling region, the pectoral muscle portion and the fat portion in the sampling region A mammary gland region representing a mammary gland in the side image by obtaining a density p1 corresponding to the border of the breast and an average density p2 of the pectoral muscle portion The threshold concentrations of separating the fat region representing the fat, the following formula using the p1 and p2 obtained:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold calculating means for calculating and determining according to the above, a mammary fat area separating means for separating a breast area obtained by removing the pectoral muscle area from the subject area into a mammary gland area and a fat area based on the threshold value, and the image data On the basis of the average concentration calculating means for calculating the average concentration in the separated mammary gland region and fat region, respectively, and by replacing the concentration in the mammary gland region and fat region with the average concentration calculated for each region. And a mammary gland distribution map creating means for creating a mammary gland distribution map of an image.

The program according to the present invention includes a procedure for setting, on a computer, a region composed of substantially pectoral muscle and fat as a sampling region in the side image represented by image data carrying a breast side image composed of mammary gland, fat and pectoral muscle. A procedure for obtaining a density p1 corresponding to a boundary between a pectoral muscle part and a fat part in the sampling area and an average density p2 of the pectoral muscle part based on the density distribution in the sampling area; The following formula using the obtained p1 and p2 as the threshold value of the concentration that separates the mammary gland region representing the mammary gland and the fat region representing fat:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
The program for executing the procedure to calculate and determine according to the above.

  This program may be supplied by being recorded on a computer-readable recording medium, or may be stored in a server connectable to the computer and supplied by downloading.

  According to the mammary gland region threshold value determining method and apparatus of the present invention, in the side image of the breast, a region consisting of approximately pectoral muscles and fat is set as a sampling region, and based on the concentration distribution in the set sampling region, A density p1 corresponding to the boundary between the pectoral muscle part and the fat part in the sampling region and an average density p2 of the pectoral muscle part are obtained, and the side image is obtained according to a predetermined calculation formula using the obtained p1 and p2. The concentration threshold value that separates the mammary gland region from the fat region is calculated, so the concentration distribution of the mammary gland and fat regions that tend to vary in concentration is not used, and not only the pectoral muscle concentration. By using the fat concentration, the above stable threshold value can be calculated accurately, and the concentration threshold value of the mammary gland region that can accurately extract the mammary gland region can be determined. .

  In the method and apparatus for determining a mammary gland region threshold, if a region consisting of approximately pectoral muscles and fat is a pectoral muscle end region representing the end of the pectoral muscle, a more stable region with less variation in density is obtained as a sampling region. And a more accurate threshold value for density can be calculated.

  Further, in the mammary gland region threshold value determining method, if the side image has the pectoral muscle end region arranged at a predetermined location in the side image and the predetermined location is set as a sampling region, Even if the muscle end region is not detected, the sampling region can be easily set on the pectoral muscle end region.

  In the mammary gland region threshold value determining device, the side image has a pectoral muscle end region arranged at a predetermined location in the side image, and the sampling region setting means sets the predetermined location as a sampling region. Then, similarly, it is possible to easily set the sampling region on the pectoral muscle end region without detecting the pectoral muscle end region.

  In the mammary gland region threshold value determination method, if the above p1 and p2 are obtained based on the density histogram in the sampling region, the pectoral muscle portion and the fat portion in the sampling region are separated relatively clearly. Therefore, p1 and p2 can be obtained with high accuracy by a simple method.

  Similarly, in the mammary gland region threshold value determination device, if the threshold value calculation means obtains the p1 and p2 based on the density histogram in the sampling region, the distribution of the pectoral muscle portion and the fat portion in the sampling region Since the mountains appear relatively clearly separated, the above-described p1 and p2 can be obtained with high accuracy by a simple method.

  In the mammary gland region threshold value determination method and apparatus, if the formula for calculating the threshold value is a = 0.4, the threshold value can be calculated according to the calculation formula optimized from the experimental results. The mammary gland region can be extracted with higher accuracy.

  According to the breast image region separation device of the present invention, the breast region is determined as the breast region based on the threshold value of the density of the breast region that can be extracted with high accuracy, which is determined by the breast region threshold determination method. Since it is separated from the fat region, the subject region in the side image of the breast can be accurately separated into the pectoral muscle region, the mammary gland region, and the fat region. In addition, since the correct distribution of each region can be known from the separation result, in particular, in the detection process of abnormal shadows, it is possible to locally know the degree of difficulty of detection, and the detection performance can be improved.

  According to the breast image classification apparatus of the present invention, a mammary gland region representing the mammary gland itself, which is shown in a state closer to a human sense when grasping the mammary gland distribution, is extracted, and the side surface of the breast is determined by the proportion of the mammary gland region Since the images are classified, they can be automatically classified according to the standards of the mammography guidelines. Moreover, since the amount of mammary gland parenchyma is known from the classification result, in particular, in the detection process of abnormal shadows, the degree of detection difficulty can be known for each breast image, and the detection performance can be improved.

  According to the mammary fat region average concentration calculation apparatus of the present invention, since the average concentration is calculated in the mammary gland region and the fat region that are accurately separated, the concentration of each region is useful information when detecting a lesion. The outline can be grasped correctly, and the detection performance of abnormal shadows can be improved.

  According to the mammary gland distribution map creation device of the present invention, the mammary gland distribution map is created by replacing the density of the mammary gland region and the fat region, which are accurately separated, with the average density for each region, which is useful when detecting lesions Therefore, it is possible to simultaneously grasp the two-dimensional distribution and the density outline in each region, which can be accurate information, and improve the detection performance of abnormal shadows. In addition, the created mammary gland distribution map is useful for detecting a left-right asymmetric shadow by comparing the mammary gland distribution maps of the left and right breasts.

  Hereinafter, embodiments of the present invention will be described with reference to the drawings.

(Embodiment 1)
FIG. 1 is a diagram showing a schematic configuration according to an embodiment of the mammary gland region threshold value determining apparatus of the present invention.

  A mammary gland region threshold value determining apparatus 100 shown in FIG. 1 includes an image data input unit 110 that inputs image data P carrying a side image P of a breast composed of mammary gland, fat, and pectoral muscles. Sampling region setting means 120 that sets a pectoral muscle end region that represents the end of the pectoral muscle composed of fat as the sampling region S, and a histogram creation means 130 that creates a density histogram Hs in the sampling region S based on the image data P. Based on the generated histogram Hs, a density p1 corresponding to the boundary between the pectoral muscle part and the fat part in the sampling region S and an average density p2 of this pectoral muscle part are obtained, and the mammary gland in the side image P The concentration threshold Th for separating the mammary gland region Pc representing the fat and the fat region Pb representing the fat is expressed by the following equation (1) using the obtained p1 and p2 (1 And a threshold calculating means 140 which calculates and determines in accordance with.

Th = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5) (1)
In the present embodiment, a mammogram of the right breast, which is a medical image, is used as the side image of the breast, and Equation (1) is expressed as a coefficient a as the most suitable condition based on the experimental results so far. = 0.4. This mammogram is an image composed of a plurality of pixels having density values, and the higher the density value, the more black the image becomes. In this mammogram, the concentration (value) of the pectoral muscle and the mammary gland is expressed as low, and the concentration (value) of fat is expressed as high.

  Next, the operation of the mammary gland region threshold value determining apparatus configured as described above will be described. FIG. 2 is a diagram showing a processing flow in the mammary gland region threshold value determining apparatus 100.

  The image data input means 110 inputs image data P representing a mammogram P, which is a side image of the breast, from a CR system capable of acquiring a digital image, a database storing digital images, or the like (step S11). FIG. 3 is a diagram schematically showing the mammogram P. As shown in FIG. In the figure, Pa, Pb, Pc, and Pd indicate a pectoral muscle region, a fat region, a mammary gland region, and a background region, respectively.

  When the image data P is input by the image data input means 110, the sampling area setting means 120 sets the pectoral muscle end area in the mammogram P as the sampling area S (step S12). FIG. 4 is a diagram showing a state in which a sampling area is set on the mammogram P. Since the mammogram is usually taken in accordance with mammography guidelines, the arrangement of the breast on the image is almost determined, and therefore a predetermined location can be set as the sampling region. In the case of a mammogram of the right breast side taken along the above guidelines, the pectoral muscle end region is located in the lower right part of the image. Therefore, an area specified by a frame of a predetermined size and shape at a predetermined coordinate position in the lower right part of the image can be set as the sampling area S. Instead of setting the predetermined location as the sampling region in this way, the pectoral muscle end region may be searched and set each time. For example, a location located at a predetermined distance from the edge of the image on the extracted pectoral muscle line may be set as the sampling region.

  When the sampling area S is set by the sampling area setting means 120, the histogram creation means 130 creates a histogram Hs of density values of pixels corresponding to the sampling area S (step S13). FIG. 5 is a diagram showing an example of the created histogram Hs. The set sampling region S includes a pectoral muscle portion that is a part of the pectoral muscle region Pa and a fat portion that is a part of the fat region Pb. Since the density values are different, the created histogram Hs has a distribution peak Ha representing the pectoral muscles and a distribution peak Hb representing the fat. In the mammogram, as described above, the concentration level of the pectoralis muscle is usually low and the concentration value of fat is high, so the mountain with the lower concentration value corresponds to the pectoral muscle part, and the mountain with the higher concentration value It can be determined to correspond to the fat portion.

  When the histogram Hs is created by the histogram creation unit 130, the threshold calculation unit 140 calculates a threshold value Th of the density value that separates the mammary gland region Pc and the fat region Pb in the mammogram P according to the above equation (1). Determine (step S14). This equation (1) is derived based on the experiment result by the present applicant, and the threshold value Th is determined by the concentration of the boundary between the pectoral muscle region Pa and the fat region Pb and the average concentration of the pectoral muscle region Pa. It means that a predetermined density is set in between. Here, as the density value p1, a density value having the lowest frequency is adopted in the vicinity of the distribution peak Ha representing the pectoral muscle on the histogram Hs and the distribution peak Hb representing the fat. The average density value p2 is obtained by calculating an average for the distribution Ha representing the pectoral muscle on the histogram Hs, that is, the density value contributing to the distribution in the low density side region from the density value p1. The density value p1 may be a density value corresponding to the position of the intersection of the two curves approximating the peaks Ha and Hb of the two distributions.

  According to such a mammary gland region threshold value determining apparatus 100, a region consisting of approximately pectoral muscles and fat in a side image of a breast is set as a sampling region, and sampling is performed based on the concentration distribution in the set sampling region. A density p1 corresponding to the boundary between the pectoral muscle part and the fat part in the region and an average density p2 of the pectoral muscle part are obtained, and in the side image according to a predetermined calculation formula using the obtained p1 and p2. Since the concentration threshold value that separates the mammary gland region from the fat region is calculated, the concentration distribution of the mammary gland and fat region where the concentration tends to vary is not used, and not only the pectoral muscle concentration but also the fat The above-mentioned stable threshold value can be calculated with high accuracy and the concentration threshold value of the mammary gland region from which the mammary gland region can be accurately extracted can be determined.

(Embodiment 2)
FIG. 6 is a diagram showing a schematic configuration according to an embodiment of the breast image region separation apparatus of the present invention.

  A breast image region separating apparatus 200 shown in FIG. 6 has an image data input means 210 for inputting image data P carrying a side image P of a breast composed of mammary glands, fat and pectoral muscles, and a side image based on the image data P. A subject region extraction means 220 for extracting a subject region Pabc representing a breast in P, and a pectoral muscle region Pa representing a pectoral muscle from the subject region Pabc by extracting a pectoral muscle line in the side image P based on the image data P. A breast muscle region separation means 230 for separating the breast muscle region, and a breast region threshold value determination means 240 for calculating and determining a concentration threshold Th for separating the breast region Pc representing the mammary gland and the fat region Pb representing fat in the side image P; Based on the threshold Th, the mammary fat region separating means 25 for separating the breast region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc into the mammary gland region Pc and the fat region Pb. It is equipped with a door.

  The pectoral muscle region separation unit 230 is further extracted with a boundary point extraction unit 231 that extracts one or more boundary points on the boundary line between the pectoral muscle region and the breast region in the side image P based on the image data P. And a pectoral muscle line extracting means 232 for extracting a pectoral muscle line by a predetermined method corresponding to the number of boundary points.

  The mammary gland region threshold determining unit 240 further includes a sampling region setting unit 241, a histogram creating unit 242, and a threshold calculating unit 243. These units 241 to 243 are the mammary gland region threshold determining apparatus according to the first embodiment. 100, the sampling region setting means 120, the histogram creation means 130, and the threshold value calculation means 140 have the same effects.

  The other conditions in the present embodiment are the same as those in the first embodiment.

  Next, the operation of the breast image region separation device 200 configured as described above will be described. FIG. 7 is a diagram illustrating a processing flow in the breast image region separation device 200.

  The image data input means 210 inputs image data P representing a mammogram P as shown in FIG. 3, which is a side image of a breast, from a CR system capable of acquiring digital images, a database storing digital images, or the like. (Step S21).

  When the mammogram P is input, the subject extraction unit 220 creates a histogram of the density values of the pixels constituting the mammogram P, and calculates the maximum density value pmax excluding the distribution corresponding to the background region Pd (the portion having a high density value). The mammogram P is binarized using the maximum density value pmax as a threshold value (see FIG. 8). That is, a region having a density lower than this threshold is extracted as the subject region Pabc (step S22). This area is a combination of the pectoral muscle area Pa, the fat area Pb, and the mammary gland area Pc.

The boundary point extraction means 231 uses the intensity of the density gradient vector because the density of the pectoral muscle area Pa is lower than that of the breast area Pbc and the density gradient of the boundary tends to be higher than the surrounding area. An edge image P ′ is created (step S23). In this edge image P ′, the boundary line between the pectoral muscle region Pa and the breast region Pbc is shown as a low density (white) shadow. The boundary point extraction unit 231 further extracts a boundary point on the boundary line between the pectoral muscle region Pa and the breast region Pbc using the subject region Pabc extracted by the subject extraction unit 220 and the information of the edge image P ′. (Step S24). Hereinafter, the boundary point extraction method will be described using the schematic diagram of the edge image P ′ shown in FIG. 9. In the actual edge image P ′, each edge (the boundary line P _L between the pectoral muscle area Pa and the breast area Pbc and the boundary line between the breast area Pbc and the background area Pd) has a low density (white) shadow. Although shown, this schematic diagram is shown by a black line for simplicity.

First, three scanning lines L ₁ , L ₂ , L ₃ having a vertical directionality are set in the right region (upper breast region) in the subject region in the created edge image P ′. At this time, the rightmost first scanning line L ₁ is set at a position of 200 pixels from the right end of the image, and further, the second scanning line L ₂ and the third scanning line L ₃ are set at intervals of 300 pixels. Each image data in this embodiment is assumed to be 10 bits, 10 pixels / mm.

Next, based on the information representing the range of the extracted subject area Pabc, a range of 20 to 80% within the subject area on each of the scanning lines L ₁ , L ₂ and L ₃ is obtained as shown in FIG. A directional edge search is performed on each scanning line L ₁ , L ₂ , L ₃ within the range. The directional edge search means to search only the edge of density falling when scanning from the top to the bottom, and the directional edge search satisfies a predetermined threshold (from the predetermined threshold). When an edge point is found, the edge point is extracted as a boundary point (marked with x in FIG. 9). Note that when no edge point satisfying the predetermined threshold is found, no boundary point exists on the scan line. That is, as a result of scanning the _three scanning lines L ₁ , L ₂ and L ₃ , 0 to 3 boundary points are extracted.

  When the boundary point is extracted by the boundary point extracting means 231, the pectoral muscle line extracting means 232 inputs the number of extracted boundary points and position information, and extracts the pectoral muscle line (step S25). Here, when there are three extracted boundary points, a quadratic curve that passes through the three boundary points is drawn into a pectoral muscle line, and when there are two extracted boundary points, the primary that passes through the two boundary points. Draw a straight line to make the pectoral muscle line. When the number of extracted boundary points is one, a perpendicular is drawn from the highest point of the breast in the image to obtain the lowest point on the perpendicular (the point at the lower end position of the image), and this point and the extracted 1 A pectoral muscle line is drawn by drawing a linear line passing through the boundary points (see FIG. 10). When the extracted boundary points are 0, it is assumed that no pectoral muscle line exists.

  When the pectoral muscle line is extracted by the pectoral muscle line extraction means 232, the sampling area setting means 241 sets the pectoral muscle end area in the side image P as the sampling area S (step S26), and the histogram creation means 242 is set to the sampling area. A histogram Hs of density values corresponding to S is created (step S27), and the threshold value calculation means 243 calculates a threshold value Th of density values for separating the mammary gland region Pc and the fat region Pb in the mammogram P based on the histogram Hs. (Step S28). In addition, since the process of step S26-28 is the same as that of step S12-14 of Embodiment 1, detailed description is abbreviate | omitted.

  When the threshold Th is determined by the threshold calculation unit 243, the mammary fat region separation unit 50 determines, based on the threshold Th, the breast region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc, the mammary region Pc, and the fat region Pb. (Step S29). That is, in the breast region Pbc, a region composed of pixels having density values exceeding the threshold Th is extracted as a fat region Pb, and a region composed of pixels having density values below the threshold Th is extracted as a mammary gland region Pc (see FIG. 11). . As a result, the subject area Pabc can be separated into the mammary gland area Pc, the fat area Pb, and the pectoral muscle area Pa.

  According to the breast image region separation apparatus 200 as described above, the breast region is determined based on the threshold value of the density of the mammary gland region that can be accurately extracted, which is determined by the mammary gland region threshold value determining method of the present invention. Since the mammary gland region and the fat region are separated, the subject region in the side image of the breast can be accurately separated into the pectoral muscle region, the mammary gland region, and the fat region. In addition, since the correct distribution of each region can be known from the separation result, in particular, in the detection process of abnormal shadows, it is possible to locally know the degree of difficulty of detection, and the detection performance can be improved.

(Embodiment 3)
FIG. 12 is a diagram showing a schematic configuration according to an embodiment of the breast image classification apparatus of the present invention.

  A breast image classification apparatus 300 shown in FIG. 12 includes an image data input unit 310 for inputting image data P carrying a side image P of a breast composed of mammary glands, fat and pectoral muscles, and a side image P based on the image data P. The subject region extraction means 320 for extracting the subject region Pabc representing the breast in FIG. 5 and the pectoral muscle region Pa representing the pectoral muscle from the subject region Pabc by extracting the pectoral muscle line in the side image P based on the image data P. A pectoral muscle region separating unit 330 for separating, a mammary region threshold value determining unit 340 for calculating and determining a concentration threshold Th for separating the mammary gland region Pc representing the mammary gland and the fat region Pb representing fat in the side image P; Based on the threshold Th, a mammary gland region extracting means 350 for extracting a mammary gland region Pc in the mammary region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc, and a breast region Pb An image classification unit 360 that classifies the side image P according to the ratio of the mammary gland region Pc to c is provided.

  The pectoral muscle region extracting unit 330 further includes a boundary point extracting unit 331 and a pectoral muscle line extracting unit 332, and the mammary region threshold value determining unit 340 further includes a sampling region setting unit 341 and a histogram creation. Means 342 and threshold value calculation means 343 are provided.

  The above means 310 to 340 have the same effects as the respective means 210 to 240 in the breast image region separation apparatus 200 of the second embodiment.

  The other conditions in the present embodiment are the same as those in the first embodiment.

  Next, the operation of the breast image classification apparatus 300 configured as described above will be described. FIG. 13 is a diagram illustrating a processing flow in the breast image classification apparatus 300.

  The image data input means 310 inputs image data P representing the mammogram P (step S31), and the subject extraction means 320 extracts the subject area Pabc in the mammogram P (step S32).

  The boundary point extraction means 331 creates an edge image P ′ of the mammogram P (step S33), and uses the information on the subject area Pabc and the edge image P ′ to create a boundary on the boundary line between the pectoral muscle area Pa and the breast area Pbc. A point is extracted (step S34).

  The pectoral muscle line extraction means 332 extracts a pectoral muscle line by a predetermined method corresponding to the number of extracted boundary points (step S35).

  On the other hand, the sampling area setting means 341 sets the pectoral muscle end area in the mammogram P as the sampling area S (step S36), and the histogram creation means 342 creates a histogram Hs of density values corresponding to the sampling area S (step S36). In step S37, the threshold value calculation unit 343 calculates and determines a threshold value Th for separating the mammary gland region Pc and the fat region Pb in the side image P based on the histogram Hs (step S38). In addition, since the process of step S31-38 is the same as that of step S21-28 of Embodiment 2, detailed description is abbreviate | omitted.

  The mammary gland region extracting means 350 extracts, as the mammary gland region Pc, a region whose density value is lower than the threshold Th in the breast region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc (step S39).

  When the mammary gland region Pc is extracted by the mammary gland region extracting means 350, the image classification means 360 calculates the ratio R occupied by the mammary gland area Pc with respect to the breast area Pbc, and the mammogram P according to the size of the ratio R. Are classified into four types as follows (step S3A). The ratio R can be calculated as R = (area of the mammary gland region Pc (number of pixels)) / (area of the breast region Pbc (number of pixels)).

1) R ≧ 0.7: high concentration 2) 0.7> R ≧ 0.5: heterogeneous high concentration 3) 0.5> R ≧ 0.2: scattered mammary gland 4) 0.2> R: fatty Incidentally, each of these types is defined as follows in the mammography guidelines.

  1) High concentration: Almost no fat is mixed in the mammary gland parenchyma, and the lesion detection rate is low.

  2) Non-uniform high concentration: Fat is mixed in the mammary gland parenchyma and exhibits a non-uniform concentration. There is a risk that the lesion is hidden in the normal mammary gland.

  3) Mammary gland scattering: Mammary gland parenchyma is scattered in the breast replaced with fat. Detection of the lesion is relatively easy.

  4) Fatty: The breast is almost completely replaced by fat. If the lesion is within the imaging range, it is easy to detect the lesion.

  “Density” here does not represent the degree of blackening on the image but the amount of mammary glands in the breast.

  According to such a breast image classification apparatus 300, a mammary gland region representing the mammary gland itself, which is shown in a state closer to a human sense when grasping the mammary gland distribution, is extracted, and the mammary gland region accounts for the ratio of the breast. Since the side images are classified, they can be automatically classified according to the standards of the mammography guidelines. Moreover, since the amount of mammary gland parenchyma is known from the classification result, in particular, in the detection process of abnormal shadows, the degree of detection difficulty can be known for each breast image, and the detection performance can be improved.

  When the ratio of the pectoral muscle area Pa to the subject area Pabc in the side image P is substantially constant in each image, the ratio R is the ratio of the mammary gland area Pc to the subject area Pabc. It is good.

  Further, the mammogram may be classified based not only on the ratio R but also on the contour shape of the mammary gland region Pc.

(Embodiment 4)
FIG. 14 is a diagram showing a schematic configuration according to an embodiment of the mammary fat region average concentration calculating apparatus of the present invention.

  The mammary fat region average concentration calculating apparatus 400 shown in FIG. 14 is based on image data input means 410 for inputting image data P carrying a side image P of a breast composed of mammary gland, fat and pectoral muscles, and based on the image data P. The subject region extraction means 420 for extracting the subject region Pabc representing the breast in the side image P and the pectoral muscle line Pa are extracted from the side image P based on the image data P, thereby separating the pectoral muscle region Pa from the subject region Pabc. A pectoral muscle region separating unit 430, a mammary region threshold value determining unit 440 for calculating and determining a concentration threshold Th for separating the mammary gland region Pc representing the mammary gland and the fat region Pb representing fat in the side image P, and a threshold value Mammary fat region separating means 45 for separating the breast region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc into the breast region Pc and the fat region Pb based on Th. If, based on the image data P, and a mean density calculation unit 460 for calculating an average density, respectively, in the separated breast region Pc and fat region Pb.

  The pectoral muscle region separating unit 430 further includes a boundary point extracting unit 431 and a pectoral muscle line extracting unit 432, and the mammary gland region threshold determining unit 440 further includes a sampling region setting unit 441 and a histogram creation. Means 442 and threshold value calculation means 443 are provided.

  Each of the above-described means 410 to 450 has the same effect as each of the means 210 to 250 in the breast image region separation apparatus 200 of the second embodiment.

  The other conditions in the present embodiment are the same as those in the first embodiment.

  Next, the operation of the mammary gland fat region average concentration calculating apparatus 400 configured as described above will be described. FIG. 15 is a diagram showing a processing flow in the mammary fat region average concentration calculating apparatus 400.

  The image data input means 410 inputs the image data P representing the mammogram P (step S41), and the subject extraction means 420 extracts the subject area Pabc in the mammogram P (step S42).

  The boundary point extraction means 431 creates an edge image P ′ of the mammogram P (step S43), and uses the information on the subject area Pabc and the edge image P ′ to create a boundary on the boundary line between the pectoral muscle area Pa and the breast area Pbc. A point is extracted (step S44).

  The pectoral muscle line extraction means 432 extracts a pectoral muscle line by a predetermined method corresponding to the number of extracted boundary points (step S45).

  On the other hand, the sampling area setting means 441 sets the pectoral muscle end area in the mammogram P as the sampling area S (step S46), and the histogram creation means 442 creates a histogram Hs of density values corresponding to the sampling area S (step S46). In step S47, the threshold value calculation means 443 calculates and determines a threshold value Th for separating the mammary gland region Pc and the fat region Pb in the side image P based on the histogram Hs (step S48).

  The mammary fat region separating means 450 separates the breast region Pbc into the mammary region Pc and the fat region Pb based on the threshold Th (step S49). In addition, since the process of step S41-49 is the same as that of step S21-29 of Embodiment 2, detailed description is abbreviate | omitted.

  When the breast region Pbc is separated into the mammary gland region Pc and the fat region Pb by the mammary fat region separating means 450, the average density calculating means 460 determines the separated mammary gland region Pc and fat region Pb based on the image data P. Average density values Ac and Ab of the pixels in each region are calculated (step S4A).

  According to the mammary fat region average concentration calculating apparatus 400 as described above, since the average concentration is calculated in the mammary gland region and the fat region that are accurately separated, the concentration in each region becomes useful information when detecting a lesion. Therefore, it is possible to improve the detection performance of abnormal shadows.

(Embodiment 5)
FIG. 16 is a diagram showing a schematic configuration according to an embodiment of the mammary gland distribution map creation device of the present invention.

  A mammary gland distribution map creating apparatus 500 shown in FIG. 16 has an image data input means 510 for inputting image data P carrying a breast side image P, and a subject area Pabc representing a breast in the side image P based on the image data P. The subject region extracting means 520 for extracting the pectoral muscle and the pectoral muscle region separation for separating the pectoral muscle region Pa representing the pectoral muscle from the subject region Pabc by extracting the pectoral muscle line in the side image P based on the image data P. Based on the threshold value Th, means 530, mammary region threshold value determining means 540 for calculating and determining a concentration threshold Th for separating the mammary gland region Pc representing the mammary gland in the side image P and the fat region Pb representing fat, Mammary fat region separating means 550 for separating a breast region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc into a mammary region Pc and a fat region Pb; Then, the average concentration calculating means 560 for calculating the average concentrations Ac and Ab in the separated mammary gland region Pc and fat region Pb, respectively, and the concentrations calculated in the respective mammary gland region Pc and fat region Pb are averaged. A mammary gland distribution map creating unit 570 that creates a mammary gland distribution map M of the side image P by replacing with the concentrations Ac and Ab is provided.

  The pectoral muscle region separating unit 530 further includes a boundary point extracting unit 531 and a pectoral muscle line extracting unit 532, and the mammary region threshold value determining unit 540 further includes a sampling region setting unit 541 and a histogram creation. Means 542 and threshold value calculation means 543 are provided.

  The means 510 to 550 described above correspond to the means 210 to 250 in the breast image region separation apparatus 200 according to the second embodiment, respectively, and provide the same action.

  The other conditions in this embodiment are the same as those in the first embodiment.

  Next, the operation of the mammary gland distribution map creation device 500 configured as described above will be described. FIG. 17 is a diagram showing a processing flow in the mammary gland distribution map creation device 500.

  The image data input means 510 inputs the image data P representing the mammogram P (step S51), and the subject extraction means 520 extracts the subject area Pabc in the mammogram P (step S52).

  The boundary point extraction means 531 creates an edge image P ′ of the mammogram P (step S 53), and uses the information on the subject area Pabc and the edge image P ′ to create a boundary on the boundary line between the pectoral muscle area Pa and the breast area Pbc. A point is extracted (step S54).

  The pectoral muscle line extraction means 532 extracts a pectoral muscle line by a predetermined method according to the number of extracted boundary points (step S55).

  On the other hand, the sampling area setting means 541 sets the pectoral muscle end area in the mammogram P as the sampling area S (step S56), and the histogram creation means 542 creates a histogram Hs of density values corresponding to the sampling area S (step S56). S57) Based on the histogram Hs, the threshold value calculation means 543 calculates and determines a threshold value Th for the density value used for threshold processing for extracting the mammary gland region in the breast region Pbc (step S58).

  The mammary fat region extracting means 550 extracts the mammary gland region Pc and the fat region Pb in the mammogram P by performing threshold value processing of the density value using the threshold Th in the breast region Pbc obtained by removing the pectoral muscle region Pa from the subject region Pabc. (Step S59). In addition, since the process of step S51-59 is the same as that of step S21-29 of Embodiment 2, detailed description is abbreviate | omitted.

  When the breast region Pbc is separated into the mammary gland region Pc and the fat region Pb by the mammary fat region separating means 550, the average density calculating means 560 includes the average density value Ac of pixels in the separated mammary gland region Pc and fat region Pb, Ab is calculated (step S5A).

  Then, the mammary gland distribution map creating means 570 creates a mammary gland distribution map M by binarizing the density values of the pixels in the mammary gland region Pc and the fat region Pb with the average density values Ac and Ab, respectively (step). S5B). FIG. 18 is a diagram showing an example of the mammary gland distribution map M created in this way.

  According to such a mammary gland distribution map creation device 500, the mammary gland distribution map is created by replacing the density of the mammary gland region and the fat region, which are accurately separated, with the average density for each region. It is possible to simultaneously grasp the two-dimensional distribution in each region and the outline of the concentration, which are useful information, and improve the detection performance of abnormal shadows. In addition, the created mammary gland distribution map is useful for detecting a left-right asymmetric shadow by comparing the mammary gland distribution maps of the left and right breasts.

  Note that the program of the present invention may be recorded on a computer-readable recording medium and supplied, or may be stored in a server connectable to the computer and supplied by downloading. According to such a program, it is described that the computer executes the same procedure as the method for determining the mammary gland region threshold of the present invention. The same effect as the determination method can be obtained.

The figure which shows schematic structure by one Embodiment of the mammary gland area | region threshold value determination apparatus of this invention. The figure which shows the processing flow in a mammary gland area | region threshold value determination apparatus Diagram showing mammogram P The figure which showed the state by which the sampling area was set on the mammogram P The figure which showed an example of histogram Hs of the density value in a sampling area The figure which shows schematic structure by one Embodiment of the breast image area | region separation apparatus of this invention. The figure which shows the processing flow in a breast image area | region separation apparatus. A diagram representing a histogram of density values in a mammogram The figure explaining the method of extracting a boundary point in an edge image The figure which shows the pectoral muscle line extracted when the extracted boundary point is one The figure which shows the mammogram of the state where the mammary gland region and the fat region were extracted The figure which shows schematic structure by one Embodiment of the breast image classification device of this invention The figure which shows the processing flow in a breast image classification device The figure which shows schematic structure by one Embodiment of the mammary gland fat area | region average density | concentration calculation apparatus of this invention. The figure which shows the processing flow in the average density calculation device The figure which shows schematic structure by one Embodiment of the mammary gland distribution map preparation apparatus of this invention. The figure which shows the processing flow in a mammary gland distribution map preparation apparatus Figure showing an example of a mammary gland distribution map

Explanation of symbols

DESCRIPTION OF SYMBOLS 100 Breast region threshold value determination apparatus 110 Image data input means 120 (241, 341, 441, 541) Sampling area setting means 130 (242, 342, 442, 542) Histogram creation means 140 (243, 343, 444, 543) Threshold calculation Means 200 Breast image region separation device 210 (310, 410, 510) Image data input means 220 (320, 420, 520) Subject region extraction means 230 (330, 430, 530) Pectoral muscle region separation means 231 (331, 431) 531) Boundary point extraction means 232 (332, 432, 532) pectoral muscle line extraction means 240 (340, 440, 540) breast area threshold value determination means 250 (450, 550) mammary fat area separation means 300 breast image classification device 350 mammary gland Area extraction means 360 image Classification means 400 Mammary fat region average concentration calculation device 460 (560) Average concentration calculation means 500 Breast distribution map creation device 570 Breast distribution map creation means

Claims (18)

In the side image represented by the image data carrying the side image of the breast consisting of mammary gland, fat and pectoral muscle, an area consisting of approximately pectoral muscle and fat is set as a sampling area,
Based on the concentration distribution in the sampling region, a concentration p1 corresponding to the boundary between the pectoral muscle portion and the fat portion in the sampling region and an average concentration p2 of the pectoral muscle portion are obtained,
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
A method for determining a mammary gland region threshold value, comprising calculating and determining according to: 2. The mammary gland region threshold value determining method according to claim 1, wherein the region composed of substantially pectoral muscles and fat is a pectoral muscle end region representing the end of the pectoral muscle. The side image is the pectoral muscle end region arranged at a predetermined location in the side image,
The mammary gland region threshold value determining method according to claim 2, wherein the predetermined location is set as the sampling region. 4. The mammary gland region threshold value determining method according to claim 1, wherein p1 and p2 are obtained based on a histogram of density in the sampling region. 5. The mammary gland region threshold value determining method according to claim 1, wherein the formula is a = 0.4. 6. The method of determining a mammary gland region threshold according to claim 1, wherein the side image is a medical image. Image data input means for inputting image data carrying a side image of the breast consisting of mammary gland, fat and pectoral muscles;
In the side image, sampling region setting means for setting a region consisting of approximately pectoral muscle and fat as a sampling region;
Based on the concentration distribution in the sampling region, a concentration p1 corresponding to the boundary between the pectoral muscle portion and the fat portion in the sampling region and an average concentration p2 of the pectoral muscle portion are obtained,
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
A mammary gland region threshold value determining device comprising threshold value calculating means for calculating and determining according to the above. 8. The mammary gland region threshold value determining device according to claim 7, wherein the region composed of substantially pectoral muscles and fat is a pectoral muscle end region representing the end of the pectoral muscle. The side image is the pectoral muscle end region arranged at a predetermined location in the side image,
9. The mammary gland region threshold value determining device according to claim 8, wherein the sampling region setting means sets the predetermined location as the sampling region. The mammary gland region threshold value determination device according to claim 7, 8 or 9, wherein the threshold value calculation means calculates the p1 and p2 based on a density histogram in the sampling region. The mammary gland region threshold value determination device according to any one of claims 7 to 10, wherein the equation is a = 0.4. The mammary gland region threshold value determination device according to any one of claims 7 to 11, wherein the side image is a medical image. Image data input means for inputting image data carrying a side image of the breast consisting of mammary gland, fat and pectoral muscles;
Subject region extraction means for extracting a subject region representing a breast in the side image based on the image data;
A pectoral muscle region separating means for separating a pectoral muscle region representing a pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data;
In the side image, sampling region setting means for setting a region consisting of approximately pectoral muscle and fat as a sampling region;
Based on the concentration distribution in the sampling region, a concentration p1 corresponding to the boundary between the pectoral muscle portion and the fat portion in the sampling region and an average concentration p2 of the pectoral muscle portion are obtained,
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold value calculating means for calculating and determining according to
Breast image region separation comprising: a mammary fat region separation means for separating a breast region obtained by removing the pectoral muscle region from the subject region into the breast region and the fat region based on the threshold value apparatus. The pectoral muscle region separating means extracts one or more boundary points on a boundary line between the pectoral muscle region and the breast region in the side image based on the image data;
14. The breast image region separating apparatus according to claim 13, further comprising a pectoral muscle line extracting unit that extracts the pectoral muscle line by a predetermined method corresponding to the number of extracted boundary points. Image data input means for inputting image data carrying a side image of the breast consisting of mammary gland, fat and pectoral muscles;
Subject region extraction means for extracting a subject region representing a breast in the side image based on the image data;
A pectoral muscle region separating means for separating a pectoral muscle region representing a pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data;
In the side image, sampling region setting means for setting a region consisting of approximately pectoral muscle and fat as a sampling region;
Based on the concentration distribution in the sampling region, a concentration p1 corresponding to the boundary between the pectoral muscle portion and the fat portion in the sampling region and an average concentration p2 of the pectoral muscle portion are obtained,
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold value calculating means for calculating and determining according to
Based on the threshold value, a mammary gland region extracting means for extracting the mammary gland region in a breast region excluding the pectoral muscle region from the subject region;
A breast image classification apparatus comprising: an image classification unit that classifies the side images according to a ratio of the breast region to the subject region or the breast region. Image data input means for inputting image data carrying a side image of the breast consisting of mammary gland, fat and pectoral muscles;
Subject region extraction means for extracting a subject region representing a breast in the side image based on the image data;
A pectoral muscle region separating means for separating a pectoral muscle region representing a pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data;
In the side image, sampling region setting means for setting a region consisting of approximately pectoral muscle and fat as a sampling region;
Based on the concentration distribution in the sampling region, a concentration p1 corresponding to the boundary between the pectoral muscle portion and the fat portion in the sampling region and an average concentration p2 of the pectoral muscle portion are obtained,
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold value calculating means for calculating and determining according to
A mammary fat region separating means for separating a breast region obtained by removing the pectoral muscle region from the subject region into the mammary gland region and the fat region based on the threshold value;
A mammary fat region average concentration calculating apparatus, comprising: an average concentration calculating means for calculating an average concentration in each of the separated mammary gland region and fat region based on the image data. Image data input means for inputting image data carrying a side image of the breast consisting of mammary gland, fat and pectoral muscles;
Subject region extraction means for extracting a subject region representing a breast in the side image based on the image data;
A pectoral muscle region separating means for separating a pectoral muscle region representing a pectoral muscle from the subject region by extracting a pectoral muscle line in the side image based on the image data;
In the side image, sampling region setting means for setting a region consisting of approximately pectoral muscle and fat as a sampling region;
Based on the concentration distribution in the sampling region, a concentration p1 corresponding to the boundary between the pectoral muscle portion and the fat portion in the sampling region and an average concentration p2 of the pectoral muscle portion are obtained,
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
Threshold calculation means for calculating and determining according to
A mammary fat region separating means for separating a breast region obtained by removing the pectoral muscle region from the subject region into the mammary gland region and the fat region based on the threshold value;
Based on the image data, average concentration calculating means for calculating an average concentration in the separated mammary gland region and the fat region, respectively,
Breast gland distribution map creating means for creating a mammary gland distribution map of the side image by replacing the density in the mammary gland region and the fat region with the average concentration calculated for each region. To create a mammary gland distribution map. On the computer,
In the side image represented by the image data carrying the side image of the breast consisting of the mammary gland, fat and pectoral muscle, a procedure for setting a region consisting of substantially pectoral muscle and fat as a sampling region;
A procedure for obtaining a density p1 corresponding to a boundary between a pectoral muscle part and a fat part in the sampling area and an average density p2 of the pectoral muscle part based on a distribution of density in the sampling area;
The threshold value of the concentration for separating the mammary gland region representing the mammary gland and the fat region representing fat in the side image is expressed by the following equation using the obtained p1 and p2:
Threshold = a × p1 + (1−a) × p2 (0.2 ≦ a ≦ 0.5)
A program for executing the procedure of calculating and determining according to the above.

Images