CN114693672A - A method for removing skin glands and nipples from mammography images based on image processing - Google Patents

Abstract

The invention discloses a mammary gland molybdenum target image skin gland and nipple removing method based on image processing, which comprises the following steps: acquiring and preprocessing an original mammary gland molybdenum target image data set; binarizing the preprocessed image based on an Otsu algorithm, and denoising to obtain a binarized mammary molybdenum target image; sketching the skin gland area of the preprocessed image, and calculating the average thickness of the skin gland; determining the skin gland boundary of each binary mammary gland molybdenum target image, and acquiring a corresponding skin gland mask image by using morphological operation; judging the type of each skin gland mask image, correspondingly translating according to the type and fusing the interior of the boundary according to the translation difference to obtain a nipple mask image; normalizing the nipple mask image after reversing the color; and multiplying the normalized image with the corresponding preprocessed image to obtain a corresponding target mammary gland molybdenum target image. The method can accurately remove the skin gland and nipple area, avoid errors and excessive cutting caused by human factors, and has higher cutting efficiency.

Description

A method for removing skin glands and nipples from mammography images based on image processing

technical field

The invention belongs to the technical field of image processing, and particularly relates to a method for removing skin glands and nipples from mammography images based on image processing.

Background technique

Quantification of glandular content in mammography images is one of the important means for the correct diagnosis of early breast cancer. Accurate quantification of glandular content in mammography images is of great significance for the diagnosis of breast cancer. In the 1970s, some scholars put forward the theory of the correlation between breast gland content in mammography images and breast cancer, and then some scholars proposed that breast gland density can be one of the independent risk factors for breast cancer. However, since the gray levels of the skin glands and nipple areas in the mammography images are close to or exceed those of the gland areas, the skin glands and nipple areas need to be removed before accurate measurement of the gland content in mammography. Accurate removal of skin glands and nipple areas in mammography images is of great significance for quantifying breast gland content and breast cancer risk.

At this stage, the method of removing skin glands and nipple areas in mammography images is mainly by manual cropping. Because the removal of skin glands and nipple areas by manual cropping is time-consuming and inefficient, and the same patient's mammography The image will be biased due to subjective factors under the cropping of different physicians. Especially when removing the skin glands, the original breast contour will be affected by manual cutting, and there will be errors in the subsequent quantification of the proportion of glands. Considering the large error in the quantification of glands by the above methods, a method for removing skin glands and nipples from mammography images based on image processing is proposed.

SUMMARY OF THE INVENTION

The purpose of the present invention is to solve the above problems, and propose a method for removing skin glands and nipples from mammography images based on image processing, which can accurately remove skin glands and nipple areas, avoid errors and excessive cropping caused by human factors, and Compared with the traditional manual cutting method, it has higher cutting efficiency.

To achieve the above object, the technical scheme adopted by the present invention is:

A method for removing skin glands and nipples from mammography images based on image processing proposed by the present invention includes the following steps:

S1. Acquire the original mammography image data set and perform preprocessing. The preprocessing is to sequentially perform format conversion and shooting information label removal operations on each original mammography image;

S2. Binarize the preprocessed image based on the Otsu algorithm, and use median filtering to denoise the binarized image to obtain a binarized mammography image;

S3. Use Labelme software to delineate the skin gland area of the preprocessed image, count the number of pixels occupied by all the preprocessed image skin gland areas, and take the average number of pixels occupied by skin glands as the average thickness of the skin glands;

S4. Determine the skin gland boundary of each binarized mammography image based on the Canny edge detection algorithm, and obtain the corresponding skin gland mask image by morphological operation according to the average thickness of the skin gland, and mark the skin gland area in the skin gland mask image. is white, the rest of the area is marked in black, and the morphological operation adopts the corrosion operation;

S5. Determine the type of each skin gland mask image. If it is a left breast image, move the skin gland area in the skin gland mask image to the right by at least half the image width, and fuse the inside of the boundary according to the translation difference. The area is marked with white, the rest of the area is marked with black, and the nipple mask image is obtained. If it is the right breast image, the skin gland area in the skin gland mask image is shifted to the left by at least half the image width, and fused according to the translation difference. Inside the border, mark the fusion area as white and the rest as black to obtain a nipple mask image;

S6, invert the color of the nipple mask image and perform normalization processing;

S7. Multiply the normalized nipple mask images and the corresponding preprocessed images to remove the skin glands and nipples to obtain a corresponding target mammography target image.

Preferably, the original mammography image dataset includes an equal number of left breast images and right breast images, and the number of cranio-caudal images and lateral oblique images in all images is equal.

Preferably, the format conversion is to convert DICOM format to PNG format.

Preferably, the operation of removing the photographing information label is as follows:

Based on the maximum contour detection algorithm, the maximum contour of the original mammography image is extracted, and the maximum contour is multiplied by the format-converted original mammary target image to obtain the image with the shooting information label removed.

Preferably, the median filter uses a 5×5 area.

Preferably, the inner core of the erosion operation is elliptical and 30×30 in size.

Preferably, the Canny operator of the Canny edge detection algorithm adopts a weak boundary threshold of 100 and a strong boundary threshold of 200.

Compared with the prior art, the beneficial effects of the present invention are:

1) According to the thickness and position characteristics of skin glands and nipple areas in mammography images, the method counts the number of pixel values of all skin gland thicknesses in the existing data set to obtain the average thickness of skin glands, and sets the corresponding morphological characteristics according to the average thickness of skin glands. The core structure and size of the operation, accurately remove the skin glands and nipple area, avoid errors and over-cropping problems caused by human factors, and provide more accurate calculation of gland content in mammography images and prediction of the risk of breast cancer for patients basis;

2) Compared with the traditional manual cropping method, it has higher cropping efficiency. For example, when the dataset includes 500 image samples, the average time for this method to remove the skin gland area of a single image is 300 milliseconds. The average time is 325 milliseconds; the time for manually delineating a single skin gland area is 2 to 3 minutes, and the time for manually delineating a single nipple area is 0.5 minutes to 1 minute, which is far more efficient in removing skin glands and nipple areas. Compared with the manual cutting method, the work efficiency is greatly improved.

Description of drawings

1 is a flowchart of a method for removing skin glands and nipples from mammography images of the present invention;

FIG. 2 is a schematic diagram of the processing process of the original mammary mammography target image of the present invention through steps S1 and S2;

3 is a schematic diagram of the skin gland removal process of the present invention;

FIG. 4 is a schematic diagram of the nipple removal process of the present invention.

Detailed ways

The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, but not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

It should be noted that, unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the technical field of the present application. The terms used herein in the specification of the present application are for the purpose of describing specific embodiments only, and are not intended to limit the present application.

As shown in Figure 1-4, a method for removing skin glands and nipples from mammography images based on image processing includes the following steps:

S1. Acquire an original mammography image data set and perform preprocessing. The preprocessing is to sequentially perform format conversion and shooting information label removal operations on each original mammography image.

In one embodiment, the original mammography image dataset includes an equal number of left breast images and right breast images, and the number of cranio-caudal images and lateral oblique images is equal in all images. The original mammography images can be cranio-caudal (CC) or lateral oblique (MLO), and the dataset is balanced for subsequent median filtering and morphological manipulation.

In one embodiment, the format conversion converts DICOM format to PNG format. The pixel value is mapped to 0 to 255 and saved as a high-quality PNG mammography image.

In one embodiment, the operation of removing the photographing information tag is as follows:

Based on the maximum contour detection algorithm, the maximum contour of the original mammography image is extracted, and the maximum contour is multiplied by the format-converted original mammary target image to obtain the image without the shooting information labels (such as location, hospital, time, etc.). The original mammography image after format conversion, that is, PNG mammary target image, when using the maximum contour detection algorithm, searches all outer contours and sorts them, and extracts the largest contour area as the breast area. The target image is multiplied to obtain the image with the shooting information label removed.

S2. Binarize the preprocessed image based on the Otsu algorithm, and use median filtering to denoise the binarized image to obtain a binarized mammography image.

In one embodiment, median filtering uses a 5x5 region. The Otsu algorithm (OTSU) will automatically calculate the segmentation threshold and perform a binarization operation on the breast region, distinguish the target from the background, and use a median filtering operation on a 5×5 area in the image to remove noise. As shown in Figure 2, from left to right are the original mammography image (DICOM mammography image), the format-converted original mammography image (PNG mammography image), and the image after the shooting information label has been removed (i.e. pre- processed images), Otsu binarized images, and median filtered images (ie, binarized mammography images).

S3. Use Labelme software to outline the skin gland area of the preprocessed image, count the number of pixels occupied by all the preprocessed image skin gland areas, and take the average number of pixels occupied by skin glands as the average thickness of the skin glands.

S4. Determine the skin gland boundary of each binarized mammography image based on the Canny edge detection algorithm, and obtain the corresponding skin gland mask image by morphological operation according to the average thickness of the skin gland, and mark the skin gland area in the skin gland mask image. are in white, the rest of the regions are marked in black, and the morphological operation adopts the erosion operation.

In one embodiment, the inner core of the erosion operation is elliptical and 30×30 in size. The kernel size of the erosion operation is determined according to the average thickness of the skin glands. The more pixels in the thickness of the skin glands, the larger the kernel size of the erosion operation.

In one embodiment, the Canny operator of the Canny edge detection algorithm adopts a weak boundary threshold of 100 and a strong boundary threshold of 200. After setting the weak boundary threshold of the Canny operator to 100 and the strong boundary threshold to 200, the Canny operator was used to find the skin gland boundary in the binarized mammography image, and the boundary coordinate points were recorded.

As shown in Figure 3, from left to right are the median filtered image (ie, the binarized mammography image), the boundary mask image (ie, the skin gland boundary map determined by the Canny edge detection algorithm), and the corroded boundary map ( i.e. skin gland mask image), color inversion map, and skin gland removal map. It should be noted that for the color inversion map and the skin gland removal map, when the skin gland area needs to be removed independently, the skin gland mask image needs to be color-reversed and then normalized, and then the normalized Each skin gland mask image obtained by multiplying the corresponding preprocessed image can be multiplied to remove the skin glands.

S5. Determine the type of each skin gland mask image. If it is a left breast image, move the skin gland area in the skin gland mask image to the right by at least half the image width, and fuse the inside of the boundary according to the translation difference. The area is marked with white, the rest of the area is marked with black, and the nipple mask image is obtained. If it is the right breast image, the skin gland area in the skin gland mask image is shifted to the left by at least half the image width, and fused according to the translation difference. Inside the border, the fused area is marked white and the rest of the area is marked black to obtain a nipple mask image.

S6, invert the color of the nipple mask image and perform normalization processing. After the color is flipped, the white area pixel is 1, which is the reserved area, and the black area pixel is 0, which is the skin gland and nipple area, that is, the area to be removed.

S7. Multiply the normalized nipple mask images and the corresponding preprocessed images to remove the skin glands and nipples to obtain a corresponding target mammography target image.

As shown in Figure 4, the image is the left breast image, and in the order indicated by the arrows, the median filter image (ie, the binarized mammography image) and the boundary mask image (ie, the skin gland boundary map determined by the Canny edge detection algorithm) , Boundary map after erosion (i.e., skin gland mask image, not shown), Boundary translation map, Boundary fusion map (white is the fusion area), color inversion map, and removal of skin glands and nipple map (i.e. target mammography image).

It should be noted that the operator can also adjust according to actual needs, such as choosing to remove only the nipple area or the skin gland area, which can be achieved by simply multiplying the corresponding images.

According to the thickness and position characteristics of skin glands and nipple areas in mammography images, the method counts the number of pixel values of all skin gland thicknesses in the existing data set to obtain the average thickness of skin glands, and sets the corresponding morphological operation according to the average thickness of skin glands. The structure and size of the inner core, accurately remove the skin glands and nipple area, avoid errors and excessive cropping caused by human factors, and provide a more accurate basis for accurately calculating the gland content in mammography images and predicting the risk of breast cancer in patients ; Compared with the traditional manual cropping method, it has higher cropping efficiency. For example, when the data set includes 500 image samples, the average time for this method to remove the skin gland area of a single image is 300ms, and the average time of removing the nipple area of a single image is 300ms. The average time is 325 milliseconds; while the time for manually delineating a single skin gland area is 2 to 3 minutes, and the time for manually delineating a single nipple area is 0.5 minutes to 1 minute, which is much more efficient in removing skin glands and nipple areas. Manual cutting method greatly improves work efficiency.

The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity, all possible combinations of the technical features in the above-described embodiments are not described. However, as long as there is no contradiction between the combinations of these technical features, All should be regarded as the scope described in this specification.

The above-mentioned embodiments only represent relatively specific and detailed embodiments described in the present application, but should not be construed as a limitation on the scope of the patent application. It should be pointed out that for those skilled in the art, without departing from the concept of the present application, several modifications and improvements can be made, which all belong to the protection scope of the present application. Therefore, the scope of protection of the patent of the present application shall be subject to the appended claims.

Claims (7)

1. A mammary gland molybdenum target image skin gland and nipple removing method based on image processing is characterized in that: the method for removing the mammary gland molybdenum target image skin gland and nipple based on image processing comprises the following steps:

s1, acquiring a data set of original breast molybdenum target images and preprocessing, wherein the preprocessing comprises the steps of sequentially carrying out format conversion and shooting information label removal on each original breast molybdenum target image;

s2, binarizing the preprocessed image based on an Otsu algorithm, and denoising the binarized image by adopting median filtering to obtain a binarized mammary molybdenum target image;

s3, delineating the skin gland area of the preprocessed image by using Labelme software, counting the number of pixel points occupied by the skin gland area of the preprocessed image, and taking the number of the pixel points occupied by the average skin gland as the average thickness of the skin gland;

s4, determining the skin gland boundary of each binary breast molybdenum target image based on a Canny edge detection algorithm, and acquiring a corresponding skin gland mask image by using morphological operation according to the average thickness of the skin glands, wherein the skin gland area in the skin gland mask image is marked as white, the rest areas are marked as black, and the morphological operation adopts corrosion operation;

s5, judging the type of each skin gland mask image, if the skin gland mask image is a left breast image, translating a skin gland area in the skin gland mask image to the right by at least one half of image width, fusing the interior of a boundary according to translation difference, marking the fused area as white, marking the rest areas as black, obtaining a nipple mask image, if the skin gland mask image is a right breast image, translating the skin gland area in the skin gland mask image to the left by at least one half of image width, fusing the interior of the boundary according to translation difference, marking the fused area as white, marking the rest areas as black, and obtaining the nipple mask image;

s6, reversing the color of the nipple mask image and then carrying out normalization processing;

and S7, multiplying each nipple mask image after normalization processing by the corresponding preprocessed image to remove the skin gland and the nipple, and obtaining a corresponding target breast molybdenum target image.

2. The image processing-based breast molybdenum target image skin gland and nipple removal method of claim 1, wherein: the original mammary gland molybdenum target image data set comprises a left mammary gland image and a right mammary gland image which are equal in number, and the head and tail position images and the lateral oblique position images in all the images are equal in number.

3. The image processing-based breast molybdenum target image skin gland and nipple removal method of claim 1, wherein: the format conversion converts the DICOM format to the PNG format.

4. The image processing-based breast molybdenum target image skin gland and nipple removal method of claim 1, wherein: the shooting information tag removing operation is specifically as follows:

and extracting the maximum contour of the original mammary gland molybdenum target image based on a maximum contour detection algorithm, and multiplying the maximum contour with the original mammary gland molybdenum target image after format conversion to obtain an image without the shooting information label.

5. The image processing-based breast molybdenum target image skin gland and nipple removal method of claim 1, wherein: the median filtering takes a 5 x 5 region.

6. The image processing-based breast molybdenum target image skin gland and nipple removal method of claim 1, wherein: the kernel of the etching operation is oval and 30 × 30 in size.

7. The image processing-based breast molybdenum target image skin gland and nipple removal method of claim 1, wherein: the Canny operator of the Canny edge detection algorithm adopts a weak boundary threshold value of 100 and a strong boundary threshold value of 200.

Images