Breast cancer image processing method and system

Abstract

The invention belongs to the field of image processing, and discloses a breast cancer image processing method and system, and the method comprises the steps: S1, obtaining an MRI image of the chest of a breast cancer patient through employing a nuclear magnetic resonance technology; s2, partitioning the MRI image, and dividing the MRI image into a plurality of local images with consistent areas; s3, segmenting the MRI image by adopting a preset single-threshold segmentation algorithm to obtain a segmentation threshold TH1 of the MRI image; s4, respectively calculating a segmentation threshold TH2 of each local image by adopting a preset single threshold segmentation algorithm; s5, segmenting the local image based on TH1 and TH2, and obtaining a set of pixel points of which the gray values are greater than TH2 in the local image; and S6, forming the image of the region of interest by the pixel points in the set of the pixel points of which the gray values are greater than TH2 in all the local images. According to the method, the more accurate region of interest can be effectively obtained.

Description

Technical Field

[0001] The present invention relates to the field of image processing, and in particular to a breast cancer image processing method and system. Background Art

[0002] In the prior art, nuclear magnetic resonance imaging (NMR) technology is usually used to obtain MRI images of the chest of breast cancer patients at different time points, and the pathological changes of breast cancer patients are tracked based on the changes (such as the size of the region) of the region of interest (such as a mass) in the MRI image.

[0003] In the process of segmenting MRI images, the prior art usually performs segmentation by first dividing the MRI image into multiple regions and then performing threshold segmentation on each region. However, this segmentation method does not take into account the different distribution of grayscale values ​​of images in different regions. For pixels in an area where all pixels do not actually belong to the background, if such an area is segmented using a threshold, the number of pixels that do not originally belong to the region of interest is increased, resulting in the region of interest obtained after segmentation containing more pixels that do not actually belong to the region of interest, thereby reducing the accuracy of the obtained region of interest and affecting the accuracy of subsequent pathological change tracking based on the characteristics of the region of interest. Summary of the invention

[0004] The purpose of the present invention is to disclose a breast cancer image processing method and system to solve the technical problem of how to obtain more accurate segmentation results in the process of segmenting MRI images of breast cancer patients.

[0005] In order to achieve the above object, the present invention adopts the following technical solution:

[0006] In one aspect, the present invention provides a method for processing breast cancer images, comprising:

[0007] S1, MRI images of the chest of a breast cancer patient are obtained using nuclear magnetic resonance technology;

[0008] S2, partitioning the MRI image into multiple local images with the same area;

[0009] S3, using a preset single threshold segmentation algorithm to segment the MRI image to obtain a segmentation threshold TH1 of the MRI image;

[0010] S4, using a preset single threshold segmentation algorithm to calculate the segmentation threshold TH2 of each local image;

[0011] S5, segmenting the local image based on TH1 and TH2, and obtaining a set of pixel points in the local image whose grayscale value is greater than TH2;

[0012] S6, forming an image of the region of interest from the pixel points in the set of all the pixel points whose grayscale values ​​in the local images are greater than TH2;

[0013] The local image is segmented based on TH1 and TH2 to obtain a set of pixels in the local image whose grayscale value is greater than TH2, including:

[0014] S51, use p to represent the local image, and use U p Represents the set of pixels in p whose grayscale value is greater than TH2;

[0015] S52, calculate the segmentation judgment value coef of p based on TH1 and TH2 p ;

[0016] S53, if coef p is greater than the adaptive judgment value threshold, then U p is an empty set, otherwise, store the pixels in p whose grayscale value is greater than TH2 in U p .

[0017] Preferably, the MRI image is partitioned into a plurality of local images with the same area, including:

[0018] Perform noise reduction on the MRI image to obtain the image wn;

[0019] Calculate the partition eigenvalues ​​of the image wn;

[0020] The image wn is partitioned based on the partition feature values ​​to obtain multiple local images with the same area.

[0021] Preferably, the calculation formula of the partition characteristic value is:

[0022]

[0023] feadpt represents the partition feature value, area i Represents the area of ​​the connected domain i, area MRI represents the area of ​​the MRI image, dpareu represents the set of connected domains obtained by the preset acquisition method; Ni represents the total number of connected domains in dpareu; MRIU represents the set of all pixels in the image wn, and grad j represents the gradient value of pixel j, Nj represents the total number of all pixels in image wn, w1 and w2 are area weight and gradient value comparison weight respectively, and midp represents the maximum gradient value of all pixels in wn.

[0024] Preferably, the process of obtaining dpareu includes:

[0025] Binarize the image wn to obtain a binary image;

[0026] Get the set tempU of connected domains in the binary image;

[0027] Calculate the adaptive ratio apro;

[0028] Store the top apro% of connected domains with the largest area in tempU into dpareu.

[0029] Preferably, performing binarization processing on the image wn to obtain a binary image includes:

[0030] Use Otsu's method to calculate the image wn, and use the threshold calculated by Otsu's method as the binarization threshold of the image wn;

[0031] In the image wn, the grayscale values ​​of the pixels whose grayscale values ​​are greater than or equal to the binary threshold are set to 255, and the grayscale values ​​of the pixels whose grayscale values ​​are less than the binary threshold are set to 0, thereby obtaining a binary image.

[0032] Preferably, performing noise reduction processing on the MRI image to obtain the image wn includes:

[0033] The MRI image is denoised using a preset denoising algorithm to obtain the image wn.

[0034] Preferably, the image wn is partitioned based on the partition feature values ​​to obtain a plurality of local images with the same area, including:

[0035] LA and LB are used to represent the length and width of the image wn respectively;

[0036] The calculation formula of the length dpLA of the local image is as follows:

[0037]

[0038] The calculation formula of the width dpLB of the local image is as follows:

[0039]

[0040] prs is the preset control coefficient; feadpt represents the partition characteristic value;

[0041] The image wn is divided into multiple local images with a length of dpLA and a width of dpLB.

[0042] Preferably, the segmentation judgment value coef of p is calculated based on TH1 and TH2 p ,include:

[0043] Use the following formula to calculate coef p ,

[0044]

[0045] stadev represents the standard deviation of the gradient value of the pixel in the local image, mrgad represents the maximum gradient value of the pixel in the MRI image, η represents the threshold weight, and λ represents the gradient value weight.

[0046] Preferably, the preset single threshold segmentation algorithm includes Otsu's method.

[0047] In another aspect, the present invention provides a breast cancer image processing system, comprising an imaging device, a partitioning device, a first segmentation device, a calculation device, a second segmentation device, and a generation device;

[0048] The imaging device is used to obtain MRI images of the chest of a breast cancer patient using nuclear magnetic resonance technology;

[0049] The partitioning device is used to partition the MRI image into a plurality of local images with the same area;

[0050] The first segmentation device is used to segment the MRI image using a preset single threshold segmentation algorithm to obtain a segmentation threshold TH1 of the MRI image;

[0051] The calculation device is used to calculate the segmentation threshold TH2 of each local image respectively by using a preset single threshold segmentation algorithm;

[0052] The second segmentation device is used to segment the local image based on TH1 and TH2, and obtain a set of pixel points in the local image whose grayscale value is greater than TH2;

[0053] The generating device is used to form an image of the region of interest by combining pixel points in a set of pixel points having grayscale values ​​greater than TH2 in all local images;

[0054] The local image is segmented based on TH1 and TH2 to obtain a set of pixels in the local image whose grayscale value is greater than TH2, including:

[0055] Let p represent the local image and U p Represents the set of pixels in p whose grayscale value is greater than TH2;

[0056] Calculate the segmentation judgment value coef of p based on TH1 and TH2 p ;

[0057] If coef p is greater than the adaptive judgment value threshold, then U p is an empty set, otherwise, store the pixels in p whose grayscale value is greater than TH2 in U p .

[0058] Beneficial effects:

[0059] Compared with the prior art, in the process of segmenting the MRI images of breast cancer patients, the present invention does not directly summarize the segmentation results of all local images to obtain the image of the region of interest. Instead, after segmenting each local image, the segmentation judgment value of each local image is calculated respectively; only when the segmentation judgment value is less than or equal to the adaptive judgment value threshold, the pixel points with grayscale values ​​greater than TH2 in the local image are taken as the pixel points in the region of interest, so that there is a greater probability to avoid segmenting the local image of the pixel points that do not belong to the region of interest, and directly take the pixel points greater than TH2 as the pixel points of the region of interest, so as to effectively obtain a more accurate region of interest. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0061] Figure 1 A schematic diagram of a breast cancer image processing method of the present invention.

[0062] Figure 2 The figure is a schematic diagram of a breast cancer image processing system of the present invention. DETAILED DESCRIPTION

[0063] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. The components of the embodiments of the present invention generally described and shown in the drawings here can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present invention provided in the drawings is not intended to limit the scope of the claimed invention, but merely represents the selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without making creative work belong to the scope of protection of the present invention.

[0064] like Figure 1 In one embodiment shown, the present invention provides a method for processing breast cancer images, comprising:

[0065] S1, MRI images of the chest of a breast cancer patient are obtained using nuclear magnetic resonance technology;

[0066] S2, partitioning the MRI image into multiple local images with the same area;

[0067] S3, using a preset single threshold segmentation algorithm to segment the MRI image to obtain a segmentation threshold TH1 of the MRI image;

[0068] S4, using a preset single threshold segmentation algorithm to calculate the segmentation threshold TH2 of each local image;

[0069] S5, segmenting the local image based on TH1 and TH2, and obtaining a set of pixel points in the local image whose grayscale value is greater than TH2;

[0070] S6, forming an image of the region of interest from the pixel points in the set of all the pixel points whose grayscale values ​​in the local images are greater than TH2;

[0071] The local image is segmented based on TH1 and TH2 to obtain a set of pixels in the local image whose grayscale value is greater than TH2, including:

[0072] S51, use p to represent the local image, and use U p Represents the set of pixels in p whose grayscale value is greater than TH2;

[0073] S52, calculate the segmentation judgment value coef of p based on TH1 and TH2 p ;

[0074] S53, if coef p is greater than the adaptive judgment value threshold, then U p is an empty set, otherwise, store the pixels in p whose grayscale value is greater than TH2 in U p .

[0075] In the process of segmenting the MRI image of a breast cancer patient, the above embodiment does not directly take the pixel points greater than the segmentation threshold in the segmentation results of all local images as the pixel points of the region of interest. Instead, after performing segmentation calculations on each local image and obtaining TH2, the segmentation judgment value of each local image is calculated respectively; only when the segmentation judgment value is less than or equal to the adaptive judgment value threshold, the pixel points with grayscale values ​​greater than TH2 in the local image are taken as the pixel points in the region of interest. In this way, there is a greater probability of avoiding the segmentation of the local image where all the pixels do not belong to the pixel points of the region of interest, and directly taking the pixel points greater than TH2 as the pixel points of the region of interest, which effectively improves the probability of obtaining a more accurate region of interest.

[0076] Preferably, the MRI image is partitioned into a plurality of local images with the same area, including:

[0077] Perform noise reduction on the MRI image to obtain the image wn;

[0078] Calculate the partition eigenvalues ​​of the image wn;

[0079] The image wn is partitioned based on the partition feature values ​​to obtain multiple local images with the same area.

[0080] Preferably, the calculation formula of the partition characteristic value is:

[0081]

[0082] feadpt represents the partition feature value, area i Represents the area of ​​the connected domain i, area MRI represents the area of ​​the MRI image, dpareu represents the set of connected domains obtained by the preset acquisition method; Ni represents the total number of connected domains in dpareu; MRIU represents the set of all pixels in the image wn, and grad j represents the gradient value of pixel j, Nj represents the total number of all pixels in image wn, w1 and w2 are area weight and gradient value comparison weight respectively, and midp represents the maximum gradient value of all pixels in wn.

[0083] When calculating the partition characteristic value, the present invention comprehensively considers the average area of ​​the connected domain and the variation range of the gradient value of the image, so that when the average area is larger and the variation range of the gradient of the image is smaller, the partition characteristic value is larger and the area of ​​the local image is larger; when the average area is smaller and the variation range of the gradient of the image is larger, the partition characteristic value is smaller and the area of ​​the local image is smaller. This partition characteristic value calculation method can make the size of the local image change with the change of the size of the connected domain, and can effectively avoid setting an overly large local image when the variation range of the gradient value is large, resulting in the calculated segmentation threshold being interfered by too many other pixels that do not belong to the region of interest during the segmentation process of the local image, and can make the distribution of the pixels belonging to the region of interest and the pixels not belonging to the region of interest in the local image more concentrated, that is, there is a simpler edge distribution between the region composed of the pixels belonging to the region of interest and the region composed of the pixels not belonging to the region of interest, thereby improving the accuracy of the segmentation result. At the same time, it can also avoid setting an overly small local image when the variation range of the gradient value is small, resulting in too many local images, and increasing the calculation amount of calculating the segmentation threshold for the local image, affecting the overall segmentation efficiency.

[0084] Preferably, the area weight and the gradient value comparison weight are and

[0085] The above two weights can be determined by conducting multiple tests on a large number of MRI images, comparing the relationship between the size of the local image after partitioning and the size of the region of interest. In the determination process, the values ​​of the two weights are continuously adjusted so that during the processing of all MRI images, the ratio between the size of the local image and the size of the region of interest is as close to 0.1 as possible.

[0086] Preferably, the large number of MRI images are TotalSegmentator MRI dataset or QIN Beast breast cancer MRI image dataset.

[0087] Preferably, the process of obtaining dpareu includes:

[0088] Binarize the image wn to obtain a binary image;

[0089] Get the set tempU of connected domains in the binary image;

[0090] Calculate the adaptive ratio apro;

[0091] Store the top apro% of connected domains with the largest area in tempU into dpareu.

[0092] By calculating the adaptive ratio to screen the connected domains, it is possible to avoid the connected domains with too small an area affecting the final calculated average area and the area gap between the regions of interest being too large, which requires calculating the segmentation threshold for too many local images, thereby improving the overall segmentation efficiency.

[0093] Preferably, performing binarization processing on the image wn to obtain a binary image includes:

[0094] Use Otsu's method to calculate the image wn, and use the threshold calculated by Otsu's method as the binarization threshold of the image wn;

[0095] In the image wn, the grayscale values ​​of the pixels whose grayscale values ​​are greater than or equal to the binary threshold are set to 255, and the grayscale values ​​of the pixels whose grayscale values ​​are less than the binary threshold are set to 0, thereby obtaining a binary image.

[0096] Preferably, the calculation formula of the adaptive ratio is:

[0097]

[0098] apro represents the adaptive ratio, spro represents the set basic ratio, barea represents the maximum area of ​​the connected domain in the set tempU, and area k represents the area of ​​the connected domain k, and Nk represents the total number of connected domains in tempU.

[0099] The adaptive ratio of the present invention can be adaptively changed based on the change of the fluctuation amplitude of the area of ​​the connected domain in tempU, so that when the fluctuation amplitude is larger, a smaller adaptive ratio is set, thereby reducing the number of connected domains in dpareu, improving the efficiency of the calculation process of the partition eigenvalue, and thus improving the overall segmentation efficiency; and when the fluctuation amplitude is smaller, a larger adaptive ratio is set to avoid excluding the region of interest from the set dpareu. Because the region of interest generally occupies a larger area in breast cancer MRI and is not the connected domain with the smallest area.

[0100] Preferably, the base ratio is set to 0.1.

[0101] In another embodiment, the base ratio may also be 0.15.

[0102] Preferably, performing noise reduction processing on the MRI image to obtain the image wn includes:

[0103] The MRI image is denoised using a preset denoising algorithm to obtain the image wn.

[0104] Benefits of denoising MRI images:

[0105] Improve image quality: Noise reduction can significantly reduce random noise in images, thereby improving image clarity and contrast, making details clearer.

[0106] Enhance the visibility of lesion areas: Noise may mask or blur the details of the lesion area. By reducing noise, the lesion area can be made more obvious, which helps to accurately identify the lesion and evaluate its characteristics.

[0107] Improving the performance of image processing algorithms: Noise reduction can improve the stability and accuracy of segmentation algorithms.

[0108] In one embodiment, the preset noise reduction algorithms include a Gaussian filtering algorithm, a wavelet transform noise reduction algorithm, and an adaptive filtering algorithm.

[0109] Preferably, the image wn is partitioned based on the partition feature values ​​to obtain a plurality of local images with the same area, including:

[0110] LA and LB are used to represent the length and width of the image wn respectively;

[0111] The calculation formula of the length dpLA of the local image is as follows:

[0112]

[0113] The calculation formula of the width dpLB of the local image is as follows:

[0114]

[0115] prs is the preset control coefficient; feadpt represents the partition characteristic value;

[0116] The image wn is divided into multiple local images with a length of dpLA and a width of dpLB.

[0117] In the process of partitioning, the length and width of the local image of the present invention are not fixed values, but can change with the change of the partition characteristic value. The advantage of this setting is that the size of the local image changes with the change of the partition characteristic value, reducing the probability of the local image being too large or too small, thereby obtaining a more reasonable number of local images and improving the efficiency of segmentation.

[0118] Preferably, the preset control coefficient is 20.

[0119] Preferably, the segmentation judgment value coef of p is calculated based on TH1 and TH2 p ,include:

[0120] Use the following formula to calculate coef p ,

[0121]

[0122] stadev represents the standard deviation of the gradient value of the pixel in the local image, mrgad represents the maximum gradient value of the pixel in the MRI image, η represents the threshold weight, and λ represents the gradient value weight.

[0123] The segmentation judgment value of the present invention is obtained by weighted calculation of data from two aspects: the numerical difference between TH1 and TH2 and the standard deviation of the gradient value of the pixel points in the local image. The segmentation judgment value obtained by such calculation can be such that the smaller the difference between the gradient values ​​of the pixel points in the local image and the larger the numerical difference between TH1 and TH2, the smaller the probability that the pixel points with gray values ​​greater than the threshold TH in the local image are considered as the pixel points of the region of interest, so that when the pixel points contained in the local image are not the pixel points of the region of interest, it is possible to avoid forced segmentation thereof, which is conducive to obtaining a more accurate segmentation result.

[0124] Preferably, the threshold weight and gray value weight are and

[0125] Preferably, the calculation formula of the adaptive judgment value threshold is:

[0126]

[0127] stadevMRI represents the standard deviation of the gradient value of the pixel in the MRI image, and mgry represents the median value of the grayscale value of the pixel in the MRI image. Represents the threshold control coefficient.

[0128] In the local image, since the number of pixels involved in the segmentation calculation is much smaller than that of the entire MRI image, the calculated area has a greater probability of being slightly larger than the segmentation threshold calculated based on all the pixels in the MRI. However, if the difference between the thresholds calculated by the two different methods is too large, it means that there is a high probability that the pixels in the local area do not belong to the region of interest.

[0129] The adaptive judgment threshold of the present invention adopts a calculation formula similar to the segmentation judgment value, but the object of calculation is the entire MRI image, thereby realizing the adaptive change of the adaptive judgment threshold. If the gap between the segmentation judgment value calculated in the local image and the adaptive judgment value threshold is larger, it means that the gap between the pixel points belonging to the region of interest and the pixel points not belonging to the region of interest in the local image is smaller, and the probability that neither of them belongs to the region of interest is greater, thereby realizing effective screening of the pixel points greater than the threshold TH2 in the local image, which is conducive to obtaining more accurate segmentation results.

[0130] Preferably, the threshold control coefficient is 1.1.

[0131] Preferably, the preset single threshold segmentation algorithm includes Otsu's method.

[0132] As another embodiment, the single threshold algorithm may also be any one of a maximum entropy method, a median threshold algorithm and a mean threshold algorithm.

[0133] The median threshold algorithm sorts the grayscale values ​​of the image and takes the middle value as the threshold. Pixels greater than the median value are classified as foreground, and pixels less than the median value are classified as background.

[0134] The mean threshold algorithm uses the average grayscale value of the image as the threshold, setting pixels greater than the mean as the foreground and pixels less than the mean as the background.

[0135] like Figure 2 In one embodiment shown, the present invention provides a breast cancer image processing system, comprising an imaging device, a partitioning device, a first segmentation device, a calculation device, a second segmentation device and a generation device;

[0136] The imaging device is used to obtain MRI images of the chest of a breast cancer patient using nuclear magnetic resonance technology;

[0137] The partitioning device is used to partition the MRI image into a plurality of local images with the same area;

[0138] The first segmentation device is used to segment the MRI image using a preset single threshold segmentation algorithm to obtain a segmentation threshold TH1 of the MRI image;

[0139] The calculation device is used to calculate the segmentation threshold TH2 of each local image respectively by using a preset single threshold segmentation algorithm;

[0140] The second segmentation device is used to segment the local image based on TH1 and TH2, and obtain a set of pixel points in the local image whose grayscale value is greater than TH2;

[0141] The generating device is used to form an image of the region of interest by combining pixel points in a set of pixel points having grayscale values ​​greater than TH2 in all local images;

[0142] The local image is segmented based on TH1 and TH2 to obtain a set of pixels in the local image whose grayscale value is greater than TH2, including:

[0143] Let p represent the local image and U p Represents the set of pixels in p whose grayscale value is greater than TH2;

[0144] Calculate the segmentation judgment value coef of p based on TH1 and TH2 p ;

[0145] If coef p is greater than the adaptive judgment value threshold, then U p is an empty set, otherwise, store the pixels in p whose grayscale value is greater than TH2 in U p .

[0146] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A breast cancer image processing method, characterized in that: include: S1, MRI images of the chest of a breast cancer patient are obtained using nuclear magnetic resonance technology; S2, partitioning the MRI image into multiple local images with the same area; S3, using a preset single threshold segmentation algorithm to segment the MRI image to obtain a segmentation threshold TH1 of the MRI image; S4, using a preset single threshold segmentation algorithm to calculate the segmentation threshold TH2 of each local image; S5, segmenting the local image based on TH1 and TH2, and obtaining a set of pixel points in the local image whose grayscale value is greater than TH2; S6, forming an image of the region of interest from the pixel points in the set of all the pixel points whose grayscale values ​​in the local images are greater than TH2; The local image is segmented based on TH1 and TH2 to obtain a set of pixels in the local image whose grayscale value is greater than TH2, including: S51, use p to represent the local image, and use U p Represents the set of pixels in p whose grayscale value is greater than TH2; S52, calculate the segmentation judgment value coef of p based on TH1 and TH2 p ; S53, if coef p is greater than the adaptive judgment value threshold, then U p is an empty set, otherwise, store the pixels in p whose grayscale value is greater than TH2 in U p .

2. A breast cancer image processing method according to claim 1, characterized in that: Partition the MRI image into multiple local images with the same area, including: Perform noise reduction on the MRI image to obtain the image wn; Calculate the partition eigenvalues ​​of the image wn; The image wn is partitioned based on the partition feature values ​​to obtain multiple local images with the same area.

3. A breast cancer image processing method according to claim 2, characterized in that: The calculation formula of the partition characteristic value is: feadpt represents the partition feature value, area i Represents the area of ​​the connected domain i, area MRI represents the area of ​​the MRI image, dpareu represents the set of connected domains obtained by the preset acquisition method; Ni represents the total number of connected domains in dpareu; MRIU represents the set of all pixels in the image wn, and grad j represents the gradient value of pixel j, Nj represents the total number of all pixels in image wn, w1 and w2 are area weight and gradient value comparison weight respectively, and midp represents the maximum gradient value of all pixels in wn.

4. A breast cancer image processing method according to claim 3, characterized in that: The process of obtaining dpareu includes: Binarize the image wn to obtain a binary image; Get the set tempU of connected domains in the binary image; Calculate the adaptive ratio apro; Store the top apro% of connected domains with the largest area in tempU into dpareu.

5. A breast cancer image processing method according to claim 4, characterized in that: The image wn is binarized to obtain a binary image, including: Use Otsu's method to calculate the image wn, and use the threshold calculated by Otsu's method as the binarization threshold of the image wn; In the image wn, the grayscale values ​​of the pixels whose grayscale values ​​are greater than or equal to the binary threshold are set to 255, and the grayscale values ​​of the pixels whose grayscale values ​​are less than the binary threshold are set to 0, thereby obtaining a binary image.

6. A breast cancer image processing method according to claim 2, characterized in that: The MRI image is subjected to noise reduction processing to obtain the image wn, including: The MRI image is denoised using a preset denoising algorithm to obtain the image wn.

7. The method for processing breast cancer images according to claim 2, characterized in that: The image wn is partitioned based on the partition feature values ​​to obtain multiple local images with the same area, including: LA and LB are used to represent the length and width of the image wn respectively; The calculation formula of the length dpLA of the local image is as follows: The calculation formula of the width dpLB of the local image is as follows: prs is the preset control coefficient; feadpt represents the partition characteristic value; The image wn is divided into multiple local images with a length of dpLA and a width of dpLB.

8. The method for processing breast cancer images according to claim 1, characterized in that: Calculate the segmentation judgment value coef of p based on TH1 and TH2 p ,include: Use the following formula to calculate coef p , stadev represents the standard deviation of the gradient value of the pixel in the local image, mrgad represents the maximum gradient value of the pixel in the MRI image, η represents the threshold weight, and λ represents the gradient value weight.

9. The method for processing breast cancer images according to claim 1, characterized in that: The preset single threshold segmentation algorithms include Otsu's method.

10. A breast cancer image processing system, characterized in that: It includes an imaging device, a partitioning device, a first segmenting device, a calculating device, a second segmenting device and a generating device; The imaging device is used to obtain MRI images of the chest of a breast cancer patient using nuclear magnetic resonance technology; The partitioning device is used to partition the MRI image into a plurality of local images with the same area; The first segmentation device is used to segment the MRI image using a preset single threshold segmentation algorithm to obtain a segmentation threshold TH1 of the MRI image; The calculation device is used to calculate the segmentation threshold TH2 of each local image respectively by using a preset single threshold segmentation algorithm; The second segmentation device is used to segment the local image based on TH1 and TH2, and obtain a set of pixel points in the local image whose grayscale value is greater than TH2; The generating device is used to form an image of the region of interest by combining pixel points in a set of pixel points having grayscale values ​​greater than TH2 in all local images; The local image is segmented based on TH1 and TH2 to obtain a set of pixels in the local image whose grayscale value is greater than TH2, including: Let p represent the local image and U p Represents the set of pixels in p whose grayscale value is greater than TH2; Calculate the segmentation judgment value coef of p based on TH1 and TH2 p ; If coef p is greater than the adaptive judgment value threshold, then U p is an empty set, otherwise, store the pixels in p whose grayscale value is greater than TH2 in U p .

Images