CN101853376A - A computer-aided detection method for breast microcalcifications - Google Patents

Abstract

The invention discloses a computer-aided detection method for mammary gland microcalcification points. Firstly, the mammary gland image is grayscale corrected and transformed to obtain the grayscale corrected image; At the same time, the Top-hat transformation method in morphology is used to obtain another background-suppressed breast image; the two images are then segmented by double threshold to obtain a preliminary image of micro-calcification points, which are post-processed to form micro-calcification points Roughly inspect the image; extract features from the roughly detected microcalcification target area, use support vector machine to classify, remove the false microcalcification target area, and mark the rest into the breast image for doctors to read use. By adopting the method of the invention, the doctor can be freed from the tedious work of image reading and classification, and assist the doctor to better understand and judge the images, thereby reducing misdiagnosis and missed diagnosis, and achieving the purpose of improving the correct rate of diagnosis.

Description

A computer-aided detection method for breast microcalcifications

technical field

The invention belongs to the technical field of automatic analysis and processing of medical images, and in particular relates to a computer-aided detection method for breast microcalcification points.

Background technique

Breast cancer is one of the most common malignant tumors in women all over the world, which seriously threatens women's health and even life. In developed countries such as Northern Europe and the West, the incidence of breast cancer ranks first in the incidence of female malignant tumors. In my country, with the change of people's living and eating habits, the incidence of breast cancer has also shown an obvious upward trend, and the incidence of breast cancer among women aged 25-34 has also increased rapidly. In recent decades, with the development of medical technology, the diagnosis and treatment of breast cancer have made great progress. However, the mortality rate of breast cancer has not decreased significantly, which is mainly due to the fact that most breast cancers have already been diagnosed when they are discovered. Later, difficult to cure. Therefore, when the etiology and prevention of breast cancer are still uncertain, early detection is the key to reducing the mortality of patients.

At present, the main method used in the diagnosis of breast cancer is mammography soft X-ray examination, in which microcalcifications and masses are the most common imaging features of breast cancer. However, because the density of soft tissue such as glands, connective tissue, blood vessels, and fat in breast tissue is very close to the density of the lesion area, and the visual fatigue of the diagnostician and other factors, misdiagnosis and missed diagnosis of early cancer still often occur. With the rapid development of computer technology and digital image processing technology, it is possible to use computer to assist detection of breast microcalcifications. Using digital image processing technology to enhance the microcalcifications in mammography images and mark the lesion area can free doctors from the tedious work of reading and classifying images, and help doctors better understand the images and judgment, so as to reduce misdiagnosis and missed diagnosis, and achieve the purpose of improving the correct rate of diagnosis.

However, the currently used methods have the following problems in the automatic detection of micro-calcification points: First, the existing detection methods generally have unsatisfactory detection results, and the stability of the detection algorithm is not high, especially in some dense breast tissues. 1. In images with low image quality, it is difficult to detect the micro-calcification target; secondly, there are shortcomings such as insufficient extraction of the micro-calcification lesion area or too large extracted lesion area, which causes subsequent feature extraction. It is very difficult; finally, what kind of features and classifiers to use to describe and classify the microcalcification area can make the detection accuracy of positive samples the highest, and it has not been well resolved.

Contents of the invention

The purpose of the present invention is to provide a computer-aided detection method for breast microcalcifications, which solves the problems in the prior art that it is difficult to completely and correctly detect microcalcifications in dense mammograms and the false positive rate of detection results is too high question.

The technical scheme adopted in the present invention is a method for automatic detection of breast microcalcifications, comprising the following steps:

Step 1, perform grayscale correction on the original mammary gland image to improve the overall contrast of the image, and obtain the image F(x, y) after grayscale correction;

Step 2: Enhance the center position details and edge area details of the suspected microcalcification target area in the grayscale corrected image, and then superimpose the enhanced suspected microcalcification target area on the grayscale corrected image to obtain Microcalcification-enhanced breast image F ₁ (x, y);

Step 3, select a circular structural element slightly larger than the calcification point target, perform Top-hat transformation on the image after grayscale correction, and obtain the breast image F ₂ (x, y) after background suppression;

Step 4, perform dual-threshold segmentation on the image F ₁ (x, y) obtained in the above step 2 and the image F ₂ (x, y) obtained in the step 3 jointly with the threshold T ₁ and the threshold T ₂ , where the threshold T ₁ is the image 85% of the maximum gray level of F ₁ (x, y), and the threshold T ₂ is 80% of the maximum gray level of image F ₂ (x, y); the target points formed after the double-threshold segmentation are used as suspected microcalcification points initially detected target area; then remove part of the false calcification point target area, and complete the rough inspection of the calcification point target area;

Step 5, for each microcalcification target area roughly detected in step 4, extract a 4-dimensional feature vector consisting of circularity, contrast, mean and variance of each target area at the position corresponding to the original breast image;

Step 6, submit the extracted 4-dimensional feature vector to the SVM classifier for judgment, and judge whether the roughly detected microcalcification point is the real microcalcification point target area;

Step 7: mark the microcalcification point target area that is judged to be real on the original mammary gland image, that is, complete the automatic detection of mammary gland microcalcification point.

The beneficial effects of the present invention are: 1. When the present invention detects microcalcifications in low-quality mammary gland images, it can accurately detect microcalcifications in the image and there are few false positive areas, which improves the detection accuracy. 2. It can free doctors from the tedious work of image reading and classification, reduce the workload of doctors, and improve the automation and speed of detection.

Description of drawings

Fig. 1 is the flowchart of detection method of the present invention;

Fig. 2 is an original mammary gland image with microcalcifications in step 1 of the embodiment of the present invention;

Fig. 3 is the image after Gamma transform correction in step 1 of the embodiment of the present invention;

Fig. 4 is the image after micro-calcification point enhancement using double structural elements in step 2 of the embodiment of the present invention;

Fig. 5 is the image after Top-hat transformation in step 3 of the embodiment of the present invention;

Fig. 6 is the binary image after the double-threshold segmentation in step 4 of the embodiment of the present invention;

Fig. 7 is the binary image after post-processing in step 4 of the embodiment of the present invention;

Fig. 8 is an image of microcalcifications marked after classification in step 7 of the embodiment of the present invention.

Detailed ways

The present invention will be described in detail below through specific embodiments.

As shown in Figure 1, a kind of breast microcalcification point computer aided detection method provided by the present invention comprises the following steps:

Step 1, because the overall contrast of the original mammogram image is not high, so the Gamma gray scale correction method is used to perform gray scale correction on the original breast image to improve the overall contrast of the image;

其中I(x，y)为输入的原始乳腺图像，F(x，y)为Gamma灰度校正后的图像，γ为Gamma值；The method of Gamma gray scale correction is:

Wherein I(x, y) is the input original mammary image, F(x, y) is the image after Gamma grayscale correction, and γ is the Gamma value;

Step 2, because the micro-calcification points in the image usually present the characteristics of a near-circular shape and a small area, using dual structural elements, through the gray-scale gradient operation in the morphology, to enhance the gray-scale corrected image F(x, y) The center position details and edge area details of the suspected microcalcification target area, and then the enhanced suspected microcalcification target area is superimposed on the grayscale corrected image F(x, y) to obtain a microcalcification enhanced breast image F ₁ (x, y) is convenient for detecting calcification points; the specific method is:

及

其中，B₁为内中心对称结构，用于强化靠近目标中心位置的细节；B₂为外中心对称结构，用于强化目标边缘区域的细节；1) Utilize the double structuring element as

and

Among them, B ₁ is an inner centrosymmetric structure, which is used to enhance the details near the center of the target; B ₂ is an outer centrosymmetric structure, which is used to enhance the details of the target edge area;

2) Using the structural element B ₁ , perform the gray-scale morphological gradient operation on the image F(x, y) after Gamma gray-scale correction to obtain the image G ₁ (x, y), namely

3) Using the structural element B ₂ , perform the gray-scale morphological gradient operation on the image F(x, y) after Gamma gray-scale correction to obtain the image G ₂ (x, y), namely

4) Superimpose the formed images G ₁ (x, y) and G ₂ (x, y) on the Gamma-corrected image F (x, y) to form a breast image F ₁ ( x, y), namely: F ₁ (x, y) = F (x, y) + G ₁ (x, y) + G ₂ (x, y);

Step 3: In order to further highlight the calcification point target area and suppress the influence of the breast tissue background, select a circular structural element slightly larger than the calcification point target, and perform Top-hat transformation on the Gamma-corrected image to obtain the background-suppressed breast image , in which the target area of microcalcification point appears initially; the specific method is:

1) Utilize circular structural elements

2) Perform morphological Top-hat transformation on the image F(x, y) after Gamma grayscale correction to form a background-suppressed breast image F ₂ (x, y), that is, F ₂ (x, y)=F(x ,y)-(F(x,y)оB ₃ (x,y));

In step 4, in the image obtained in step 2, while enhancing the microcalcification target area, the area in the background that is close to the microcalcification target shape characteristics is also enhanced, resulting in a "noise interference area"; and step 3 is obtained In the image, the background is well suppressed. To this end, the two images obtained in step 2 and step 3 are separately segmented using different thresholds, and the formed target point is used as the preliminary calcification point target area; further, considering the size and distribution range of the micro-calcification point target Limit, and then perform post-processing to remove part of the false calcification point target area, and complete the rough inspection of the calcification point target area; the specific method is;

1) Obtained through statistical data of many experiments, set 85% of the maximum grayscale of the image F ₁ (x, y) obtained in step 2 as the threshold T ₁ , and set the threshold T 1 of the image F ₂ (x, y) obtained in step 3 80% of the maximum gray scale is set as the threshold T ₂ ;

2) Carry out double-threshold segmentation on image F ₁ (x, y) and image F ₂ (x, y) jointly with threshold T ₁ and threshold T ₂ , that is, to satisfy the condition at the same time: F ₁ (x, y) > T ₁ And the pixel points with F ₂ (x, y)>T ₂ are determined as the target areas of suspected microcalcifications initially detected;

3) Finally, remove the target area with an area smaller than 2 pixels or larger than 20 pixels, remove the target area near the upper, lower, left, and right boundaries in the image, and form a roughly detected micro-calcification target area;

Step 5, in the coarsely detected calcification point target image in step 4, there are many false calcification point targets, which are called false positive regions. In order to further remove these false positive areas, extract the circularity, contrast, mean and variance of each calcification target area roughly detected in step 4 on the original breast image, and use it as the typical feature of the calcification target area, consisting of 4-dimensional feature vector; the extracted 4-dimensional feature vector is:

1) Circularity of the target area: Y=P ² /(4πS), where P is the perimeter of the target area, and S is the area of the target area;

其中f和b分别表示目标区域和背景区域的平均灰度，具体为：其中N_f表示粗检出的一个微钙化点目标区域的像素个数，Ω_f表示粗检出的一个微钙化点目标区像素组成的集合；

其中Ω_b表示包围微钙化点的外界矩形区域中非微钙化点组成的像素集合，而外界矩形区域是由包含一个微钙化点目标区的最小外接矩形分别向上、下、左、右各外扩一个像素组成的区域，N_b表示Ω_b中的像素个数；2) Target area contrast:

where f and b represent the average grayscale of the target area and the background area, respectively, specifically: Wherein N _f represents the number of pixels of a microcalcification target area detected roughly, and Ω _f represents a set of pixels in a micro calcification target area detected roughly;

Among them, Ω _b represents the pixel set composed of non-micro-calcification points in the outer rectangular area surrounding the micro-calcification point, and the outer rectangular area is expanded upward, downward, left, and right respectively by the smallest circumscribed rectangle containing a micro-calcification point target area. An area composed of one pixel, N _b represents the number of pixels in Ω _b ;

其中N_f表示粗检出的一个微钙化点目标区域的像素个数，Ω_f表示粗检出的一个微钙化点目标区像素组成的集合；3) Mean value of the target area:

Wherein N _f represents the number of pixels of a microcalcification target area detected roughly, and Ω _f represents a set of pixels in a micro calcification target area detected roughly;

其中N_f表示粗检出的一个微钙化点目标区域的像素个数，Ω_f表示粗检出的一个微钙化点目标区像素组成的集合；4) Target area variance:

Wherein N _f represents the number of pixels of a microcalcification target area detected roughly, and Ω _f represents a set of pixels in a micro calcification target area detected roughly;

Step 6, using the feature vector extracted in step 5, using positive samples and negative samples, to train the support vector machine; then extracting the feature vector of the rough detected calcification point target area and inputting it into the trained vector machine, Classifying the target area to obtain whether it is a real calcification point target area;

Using the method of support vector machine, the method for classifying the target area of microcalcification point detected roughly is:

1) The kernel function used in the support vector machine is a Gaussian radial basis kernel function;

2) Compared with the marked real microcalcification area, the microcalcification target detected by rough inspection is divided into positive samples and negative samples, and the samples are randomly divided into 5 size subsets. For each model parameter set, SVM classification is trained device;

3) Extract the 4-dimensional feature vector of the microcalcification point target area that is roughly detected, and use the trained SVM classifier to classify the target to obtain whether it is a real microcalcification point target area;

Step 7: mark the target area of the microcalcification point that is judged to be real on the original mammary gland image, that is, complete the automatic detection of the microcalcification point of the mammary gland.

Example 1

A method for computer-aided detection of mammary gland microcalcifications is implemented through the following steps:

Step 1, read an original mammary gland image as shown in Figure 2, and then use the Gamma grayscale correction method to perform grayscale correction on the original mammary gland image:

The Gamma grayscale correction method is: Wherein I(x, y) is the input original image, F(x, y) is the image after Gamma grayscale correction, and the value of γ is 3, and the image after grayscale correction is obtained as shown in Figure 3;

Step 2, use the background overlay method based on double structural elements to enhance the center position details and edge area details of the suspected microcalcification target area in Figure 3:

及

对图3进行下述形态学处理，得到图像G₁(x，y)和G₂(x，y)：Using double structuring elements:

and

Perform the following morphological processing on Fig. 3 to obtain images G ₁ (x, y) and G ₂ (x, y):

${\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CirclePlus;</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; \&Theta;</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; 1</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$

${\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; \&CirclePlus;</span>{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span><span\; class="notranslate">\; -</span><span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; f</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; \&Theta;</span>}{\mathrm{<span\; class="notranslate">\; B</span>}}_{\mathrm{<span\; class="notranslate">\; 2</span>}}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; the\; y</span>}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; )</span>$

Then, images G ₁ (x, y) and G ₂ (x, y) are superimposed on Fig. 3 to form a breast image F ₁ (x, y) enhanced by microcalcifications, namely: F ₁ (x, y )=F(x, y)+G ₁ (x, y)+G ₂ (x, y), as shown in Figure 4;

对图3进行Top-hat变换，得到如图5所示背景抑制后的乳腺图像F₂(x，y)，即F₂(x，y)＝F(x，y)-(F(x，y)оB₃(x，y))；Step 3, use the Top-hat transformation in morphology to form a breast image after background suppression: use structural elements

Perform Top-hat transformation on Fig. 3 to obtain the breast image F ₂ (x, y) after background suppression as shown in Fig. 5 , that is, F ₂ (x, y)=F(x, y)-(F(x, y) о B ₃ (x, y));

Step 4, use the method of double threshold and post-processing to roughly detect the target area of microcalcification points:

Select threshold _T1 to be 85% of the maximum gray scale of the image shown in Figure 4, and threshold _T2 to be 80% of the maximum gray scale of the image shown in Figure 5, respectively for the image shown in Figure 4 (the image after the enhancement of microcalcification points) and the image shown in Figure 5 (image after Top-hat transformation) combined with threshold T ₁ and threshold T ₂ to perform dual-threshold segmentation to obtain a binary image after dual-threshold segmentation, as shown in Figure 6, which is the suspected Microcalcification target area; finally, remove the target area with an area of less than 2 pixels or greater than 20 pixels, and the target area near the upper, lower, left, and right boundaries in Figure 6 to form a coarsely detected microcalcification target area, such as As shown in Figure 7;

Step 5, for each microcalcification point target area roughly detected in Figure 7, extract the circularity, contrast, histogram mean and histogram variance 4-dimensional feature vector of each target area at the corresponding position in Figure 2:

1) Circularity of the target area: Y=P ² /(4πS), where P is the perimeter of the target area, and S is the area of the target area;

其中f和b分别表示目标区域和背景区域的平均灰度，具体为：

其中N_f表示粗检出的一个微钙化点目标区域的像素个数，Ω_f表示粗检出的一个微钙化点目标区像素组成的集合；其中Ω_b表示包围微钙化点的外界矩形区域中非微钙化点组成的像素集合，而外界矩形区域是由包含一个微钙化点目标区的最小外接矩形分别向上、下、左、右各外扩一个像素组成的区域，N_b表示Ω_b中的像素个数；2) Target area contrast:

where f and b represent the average grayscale of the target area and the background area, respectively, specifically:

Wherein N _f represents the number of pixels of a microcalcification target area detected roughly, and Ω _f represents a set of pixels in a micro calcification target area detected roughly; Among them, Ω _b represents the pixel set composed of non-micro-calcification points in the outer rectangular area surrounding the micro-calcification point, and the outer rectangular area is expanded upward, downward, left, and right respectively by the smallest circumscribed rectangle containing a micro-calcification point target area. An area composed of one pixel, N _b represents the number of pixels in Ω _b ;

其中N_f表示粗检出的一个微钙化点目标区域的像素个数，Ω_f表示粗检出的一个微钙化点目标区像素组成的集合；3) Mean value of the target area:

Wherein N _f represents the number of pixels of a microcalcification target area detected roughly, and Ω _f represents a set of pixels in a micro calcification target area detected roughly;

其中N_f表示粗检出的一个微钙化点目标区域的像素个数，Ω_f表示粗检出的一个微钙化点目标区像素组成的集合；4) Target area variance:

Wherein N _f represents the number of pixels of a microcalcification target area detected roughly, and Ω _f represents a set of pixels in a micro calcification target area detected roughly;

Step 6, using the support vector machine method to classify the coarsely detected microcalcification target area: the 4-dimensional feature vector extracted from each microcalcification target area in Figure 7 is handed over to the SVM classifier for judgment , to get whether it is the real microcalcification target area;

Step 7: mark the microcalcification target area that is judged to be real on the original mammary image in Figure 2, and the marking result is shown in Figure 8, that is, the automatic detection of breast microcalcification is completed.

Claims (6)

1. the method that the mammary gland microcalcifications detects automatically is characterized in that, comprises following operation steps:

Step 1 is carried out gray correction to original galactophore image, to improve the overall contrast of image, obtain after the gray correction image F (x, y);

Step 2, center details and fringe region details to doubtful microcalcifications target area in the image after the gray correction strengthen, doubtful microcalcifications target area after will strengthening then is added on the image after the gray correction, obtains the galactophore image F that microcalcifications strengthens ₁(x, y);

Step 3 is selected the circular configuration element more bigger than calcification point target for use, and the image after the gray correction is carried out the Top-hat conversion, obtains the galactophore image F after background suppresses ₂(x, y);

Step 4 is to the image F of above-mentioned steps 2 gained ₁(x, y) and the image F of step 3 gained ₂(x, y) associating threshold value T ₁With threshold value T ₂Carry out dual threshold and cut apart, wherein, threshold value T ₁Be image F ₁(x, y) 85% of maximum gray scale, threshold value T ₂Be image F ₂(x, y) 80% of maximum gray scale; The doubtful microcalcifications target area that the impact point that forms after dual threshold is cut apart goes out as Preliminary detection; Remove the false calcification point of part target area then, finish the Rough Inspection of calcification point target area;

Step 5, each microcalcifications target area that Rough Inspection goes out at step 4 extracts the circularity, contrast, average of each target area and 4 dimensional feature vectors that variance is formed in the position of original galactophore image correspondence;

Step 6 with 4 dimensional feature vectors that extract, is transferred to the svm classifier device and is judged, judges whether the microcalcifications that Rough Inspection goes out is real microcalcifications target area;

Step 7 is thought real microcalcifications target area to judgement, and its mark to original galactophore image, is promptly finished the automatic detection to the mammary gland microcalcifications.

2. the method that a kind of mammary gland microcalcifications according to claim 1 detects automatically is characterized in that the concrete operations step of described step 1 is:

Adopt Gamma gray correction method that original galactophore image is carried out gray correction, to improve the overall contrast of image;

Described Gamma gray correction method is:

Wherein (x y) is the original galactophore image of input to I, and (x y) is the image after the gray correction to F, and γ is the Gamma value.

3. the method that a kind of mammary gland microcalcifications according to claim 1 detects automatically is characterized in that the concrete operations step of described step 2 is:

Adopt the double structure element, by shade of gray computing in the morphology, to the image F (x after the gray correction, y) the center details of doubtful microcalcifications target area and fringe region details strengthen in, doubtful microcalcifications target area after will strengthening the then image F (x after the gray correction that is added to, y) on, obtain the galactophore image F that microcalcifications strengthens ₁(x, y); Its concrete grammar is:

1) adopts the double structure element

And

Wherein, B ₁Be interior centrosymmetric structure, be used to strengthen details, B near the target's center position ₂Be outer centrosymmetric structure, be used to strengthen the details in object edge zone;

2) utilize structural element B ₁, (x y) carries out the computing of gray scale morphology gradient and obtains image G to the image F after the gray correction ₁(x, y), promptly

3) utilize structural element B ₂, (x y) carries out the computing of gray scale morphology gradient and obtains image G to the image F after the gray correction ₂(x, y), promptly

4) with the image G that forms ₁(x, y) and G ₂(x, y) be added to image F after the gray correction (x, y) on, form the galactophore image F after microcalcifications strengthens ₁(x, y), that is: F ₁(x, y)=F (x, y)+G ₁(x, y)+G ₂(x, y).

4. the method that a kind of mammary gland microcalcifications according to claim 1 detects automatically is characterized in that the concrete operations step of described step 3:

The circular configuration element that utilization is more bigger than calcification point target carries out the Top-hat conversion to the image after the gray correction, obtains the galactophore image F after background suppresses ₂(x, y), microcalcifications target area wherein tentatively displays; Its concrete grammar is:

1) selects the circular configuration element for use $B_3=\left[\begin{array}{ccccc}0& 0& 1& 0& 0\\ 0& 1& 1& 1& 0\\ 1& 1& 1& 1& 1\\ 0& 1& 1& 1& 0\\ 0& 0& 1& 0& 0\end{array}\right];$

2) (x y) carries out morphology Top-hat conversion, forms the galactophore image F that background suppresses to the image F after the gray correction ₂(x, y), i.e. F ₂(x, y)=F (x, y)-(F (x, y) о B ₃(x, y)).

5. the method that a kind of mammary gland microcalcifications according to claim 1 detects automatically is characterized in that the concrete operations step of described step 4:

Two width of cloth images that obtain in step 2 and the step 3 are adopted different threshold values respectively, unite and cut apart, the impact point of formation removes the calcification point target area of part falseness then as preliminary calcification point target area, finishes the Rough Inspection of calcification point target area; Its concrete grammar is:

1) the image F that obtains for step 2 ₁(x, y), with image F ₁(x y) 85% of maximum gray scale is made as threshold value T ₁The image F that obtains for step 3 ₂(x, y), with image F ₂(x y) 80% of maximum gray scale is made as threshold value T ₂

2) again to image F ₁(x is y) with image F ₂(x, y) associating threshold value T ₁With threshold value T ₂Carry out dual threshold and cut apart, promptly satisfy T simultaneously ₁(x, y)＞T ₁And T ₂(x, y)＞T ₂The doubtful microcalcifications target area that goes out as Preliminary detection of pixel;

3) remove area at last less than 2 pixels or greater than 20 pixel target areas, be positioned at the target area of boundary vicinity up and down in the removal image, form the microcalcifications target area that Rough Inspection goes out.

6. the method that a kind of mammary gland microcalcifications according to claim 1 detects automatically is characterized in that the concrete operations step of described step 5:

Each calcification point target area that Rough Inspection goes out at step 4 extracts circularity, contrast, average and the variance of each target area in the position of original galactophore image correspondence, with its characteristic feature as the calcification point target area, form 4 dimensional feature vectors; 4 dimensional feature vectors that extract are:

1) target area circularity: Y=P ²/ (4 π S), wherein, P is the girth of target area, S is the area of target area;

2) target area contrast:

Wherein f and b represent the average gray of target area and background area respectively; Be specially:

N wherein _fThe number of pixels of the microcalcifications target area that the expression Rough Inspection goes out, Ω _fThe set that microcalcifications target area pixel is formed that the expression Rough Inspection goes out;

Ω wherein _bThe collection of pixels that non-microcalcifications is formed in the extraneous rectangular area of microcalcifications is surrounded in expression, and extraneous rectangular area by the minimum boundary rectangle that comprises a microcalcifications target area respectively up and down, the left and right zone that a pixel is formed, the N of respectively extending out _bExpression Ω _bIn number of pixels;

3) target area average:

N wherein _fThe number of pixels of the microcalcifications target area that the expression Rough Inspection goes out, Ω _fThe set that microcalcifications target area pixel is formed that the expression Rough Inspection goes out;

4) target area variance: N wherein _fThe number of pixels of the microcalcifications target area that the expression Rough Inspection goes out, Ω _fThe set that microcalcifications target area pixel is formed that the expression Rough Inspection goes out.

Images