CN109893100A - A kind of method that breast density quantification calculates in breast cancer risk assessment - Google Patents
A kind of method that breast density quantification calculates in breast cancer risk assessment Download PDFInfo
- Publication number
- CN109893100A CN109893100A CN201910313084.7A CN201910313084A CN109893100A CN 109893100 A CN109893100 A CN 109893100A CN 201910313084 A CN201910313084 A CN 201910313084A CN 109893100 A CN109893100 A CN 109893100A
- Authority
- CN
- China
- Prior art keywords
- breast
- cluster
- area
- value
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Abstract
The present invention relates to the methods that breast density quantification in a kind of breast cancer risk assessment calculates, specifically includes the following steps: pre-processing to breast molybdenum target original image, carry out histogram equalization, gaussian filtering, down-sampling;Divide breast area from breast molybdenum target original image, detects breast outer edge line and chest muscle inward flange line;Unsupervised clustering is made using fuzzy C-mean algorithm to the pixel in breast area;The feature for extracting clustering cluster region, the merging of the obtained clustering cluster is referred in body of gland and fatty two classes, makees two Classification and Identifications to clustering cluster;Two classification, training linear discriminant LDA classifier are carried out to breast area;Body of gland size in breast area is calculated according to classification results.Breast density quantification calculation method provided by the invention can be realized the full-automatic segmentation of breast tissue in breast, and the burden of image doctor is significantly reduced while providing objective results.
Description
Technical field
The present invention relates to breast density quantification in field of medical image processing more particularly to a kind of breast cancer risk assessment
The method of calculating.
Background technique
Breast cancer risk assessment model is divided into two classes, and one kind is empirical risk model such as Gail model, and another kind of is side
The IBIS model of gene mutation again.Wherein Gail model is the most frequently used, which is used to assess the personalized breast cancer of calculating and suffers from cancer
Value-at-risk, including multiple risks and assumptions relevant to breast cancer, model gives each risks and assumptions different weight coefficients, weighted
Combination be calculated one quantization suffer from cancer value-at-risk.Risks and assumptions in Gail model include weight, age, corpus mamma class
Type, menarche and menopause time, lactation feeds situation, BMI index, smokes or situation of drinking, stress situation etc., wherein
Corpus mamma type is mainly given by veteran image doctor according to Breast imaging reporting and data system (BI-RADS)
Mammary gland densification grade out.Corpus mamma type only gives the qualitative estimation of gland density, if quantifying for gland density can be provided
Calculated value will be so that assessment models be more accurate.
Most common gland density quantification calculation method is the semi-automation based on image doctor's manual interaction formula
Body of gland dividing method, image doctor passes through body of gland wheel in breast of sketching out by hand under image archiving and communication system (PACS)
Profile, calculated in system tool automatically in contour line the number of pixels of body of gland and, gland density value is equal to pixel in body of gland
Number and with whole breast number of pixels and ratio.
Semi-automatic body of gland dividing method based on image doctor's manual interaction formula is often by the subjectivity of image doctor
Factor is affected, as a result not objective enough;Although according to the guidance of Breast imaging reporting and data system (BI-RADS),
Doctor's image the case where there is also tired diagosis, visual fatigue frequently can lead to mammary gland contours segmentation and error occurs.Therefore,
The method for carrying out objective judgement it is necessary to propose a kind of pair of corpus mamma, to assist doctor to judge mammary cancer risk.
Summary of the invention
It is an object of the invention to overcome the body of gland dividing method based on image doctor's manual interaction formula semi-automation to exist
The defect of error provides a kind of full automatic corpus mamma dividing method, can quantitatively calculate the density value of body of gland in breast.
The technical solution adopted by the present invention to solve the technical problems is:
A kind of method that breast density quantification calculates in breast cancer risk assessment, specifically includes the following steps:
Step 1 pre-processes breast molybdenum target original image, carries out histogram equalization, gaussian filtering, down-sampling;
Step 2 divides breast area from pretreated breast molybdenum target image, detects breast outer edge line and chest muscle inner edge
Edge line, and then obtain breast area size;
Step 3 makees Unsupervised clustering using fuzzy C-mean algorithm to the pixel in breast area, obtains clustering cluster;
Clustering cluster merging is referred in body of gland and fatty two classes, makees to clustering cluster by step 4, the feature for extracting clustering cluster region
Two Classification and Identifications;
Step 5 carries out two classification, training linear discriminant LDA classifier to breast area;
Step 6, the number of pixels that body of gland part in breast area is calculated according to classification results, i.e. body of gland size.
The breast molybdenum target original image of the step 1 is selected from MLO molybdenum target figure and CC molybdenum target as a preferred technical solution,
Figure.
The step 2 specifically includes as a preferred technical solution:
Step 2.1: seeking the gray-scale intensity histogram of pretreated molybdenum target image, intuitively distinguish background air region and cream
The pixel map in room region, Cong Tuzhong finds out the differentiation threshold value of background air and mammary region, to extract breast area
Edge;
Step 2.2: tentatively extracting the edge graph of breast area by Canny boundary operator, reuse Hough transformation to obtain cream
The straight line parameter histogram of the two-dimentional Polar coordinates in room region, and chest muscle edge line is located approximately at polar determining angle model
It encloses;
Step 2.3: using the outer edge line of the extraction as boundary, the region divided is breast, by binary conversion treatment
Afterwards, interior pixels number, as breast area size are counted.
The step 3 specifically includes as a preferred technical solution:
Step 3.1: cluster number is adaptively chosen first, and zero-mean normal distribution normalized is used to breast area, then
Using Gaussian filter to mammary region image smoothing filtering technique, using the peak value number of treated breast grey level histogram as poly-
Cluster number upper limit threshold is arranged in the number of class cluster, and when peak value number is more than upper limit threshold, capping threshold number is as cluster
Number;
Step 3.2: breast area pixel value being normalized and down-sampling processing, setting initial cluster center use Fuzzy C
Mean value is calculated initial centered value by pixel value equal proportion, find out each pixel in breast image to each center distance, with
Apart from minimum evaluation function, successive ignition calculates each pixel of breast area to each until cluster centre is constant
The distance at cluster center;
Step 3.3: then the pixel for belonging to each cluster in breast area being labeled, breast area is divided into the son of cluster number
Area image, and the energy spectrogram in each cluster region is drawn with the Euclidean distance at pixel each in breast to each cluster center, from energy
It measures in spectrogram substantially it can be seen that regional distribution chart of each cluster in breast.
The clustering cluster in the step 4 makees two Classification and Identifications as a preferred technical solution, and specific method is to use gray scale
Co-occurrence matrix extracts the whole textural characteristics with all subregion of breast, including contrast, the degree of correlation, energy, entropy and heterogeneity;Make
With region description operator extraction breast entirety and all subregion correlation attribute value, including area pixel number, minimum rectangle boundary, matter
The heart;Calculate third moment value, Fourth-order moment value, skewness value and the kurtosis value of breast image;Calculate each subregion pixel mean value, partially
Angle value and kurtosis value, the areal concentration that is connected be poor, compactness.
The method that breast area carries out two classification in the step 5 as a preferred technical solution, includes: first to training
It concentrates each clustering cluster region in each sample image to make label, needs veteran image doctor by hand in training set
Label is made in each clustering cluster region of each sample image respectively, reuses and marks the image data signed to train LDA classifier
Model finally makees two classification to breast image using the LDA classifier, distinguishes body of gland region and the fat region in breast.
Body of gland size can be used for calculating gland density value, calculation formula in the step 6 as a preferred technical solution,
Are as follows: gland density value=body of gland number of pixels/breast number of pixels.
By above technical scheme, compared with the existing technology, the invention has the following advantages: being calculated according to artificial intelligence
Method and big data technology analyze breast molybdenum target image, are partitioned into body of gland region automatically from breast image, close and calculate gland
Volume density value.Compared with semi-automatic interactive approach, the calculated result of this method is more objective, while also mitigating image doctor
The burden of a large amount of diagosis.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples.
Fig. 1 is the mammary gland tissue density quantification calculation method flow chart in the embodiment of the present invention;
Fig. 2 is certain breast molybdenum target MLO original image in the embodiment of the present invention;
Fig. 3 is same breast area binary map in the embodiment of the present invention;
Fig. 4 is the corresponding intensity histogram peak value figure of same breast initial clustering cluster number in the embodiment of the present invention;
Fig. 5 is the density energy spectrogram of same breast subregion cluster in the embodiment of the present invention;
Fig. 6 is same breast subregion tag energy spectrogram in the embodiment of the present invention;
Same breast breast contours (outside) and body of gland profile (inside) figure in Fig. 7 embodiment of the present invention.
Specific embodiment
In conjunction with the accompanying drawings, the present invention is further explained in detail.These attached drawings are simplified schematic diagram, only with
Illustration illustrates basic structure of the invention, therefore it only shows the composition relevant to the invention.
Fig. 1 is the mammary gland tissue density quantification calculation method flow chart of one embodiment of the invention.As shown,
It the described method comprises the following steps:
Step 1 pre-processes breast molybdenum target original image as shown in Figure 2, and it is dry to carry out histogram equalization removal original image
It disturbs, gaussian filtering, down-sampling.
Step 2 divides breast area from pretreated breast molybdenum target original image, detects breast outer edge line and chest
Intramuscular edge line, specific steps include:
2.1 seek the gray-scale intensity histogram of pretreated molybdenum target image, intuitively distinguish background air region and regio mammalis
The pixel map in domain, Cong Tuzhong find out the differentiation threshold value of background air and mammary region, to extract the side of breast area
Edge;
2.2 tentatively extract the edge graph of breast area by Canny boundary operator, reuse Hough transformation to obtain breast area
Two-dimentional Polar coordinates straight line parameter histogram, and chest muscle edge line is located approximately at polar determining angular range, meter
The number of pixels of breast area after calculating binaryzation, as breast area size.
Step 3 makees Unsupervised clustering using fuzzy C-mean algorithm to the pixel in breast area, and then breast is divided into more
A clustering cluster constitutes cluster gathering, specific steps are as follows:
3.1 adaptively choose cluster number first, use zero-mean normal distribution normalized to breast area, reuse height
This filter is to mammary region image smoothing filtering technique, using the peak value number of the grey level histogram of treated breast as clustering cluster
Center;
3.2 pairs of breast area pixel values are normalized and down-sampling processing, and initial cluster center is arranged, uses fuzzy C-mean algorithm
Initial centered value is acquired by the calculating of pixel value equal proportion, find out each pixel in breast image to each center distance, with
Apart from minimum evaluation function, successive ignition calculates each pixel of breast area to each cluster until cluster centre is constant
The distance at center;
Thus the pixel for belonging to each cluster in breast area is labeled, breast area is divided into the subregion figure of cluster number
Picture, and the energy spectrogram in each cluster region is drawn with the Euclidean distance at pixel each in breast to each cluster center, from energy spectrogram
In substantially it can be seen that regional distribution chart of each cluster in breast.
Step 4, the feature for extracting clustering cluster region, breast area inside are mainly two, fat by body of gland and non-body of gland
It is grouped as, the clustering cluster merging that step 3 obtains is referred in body of gland and fatty two classes, i.e., two Classification and Identifications are made to clustering cluster,
Suitable feature is extracted to each cluster, to the feature that each clustering cluster subregion extracts have statistics of histogram feature (such as curvature,
Kurtosis), gray scale textural characteristics (such as energy, entropy), morphologic description it is sub (such as compactness and perimeter).
Two classification of step 5, breast area, training linear discriminant LDA classifier, first have to sample each in training set
Label is made in each clustering cluster region in image, needs veteran image doctor by hand to each sample image of concentration training
Label is made in each clustering cluster region respectively, reuses and marks the image data signed to train LDA sorter model, finally using should
LDA classifier makees two classification to breast image, obtains body of gland region part.
Step 6, the number of pixels that body of gland part in breast area is calculated according to classification results, i.e. body of gland size, gland
Volume density value=body of gland number of pixels/breast number of pixels.
Breast molybdenum target image typically includes the left and right side breast of interior lateral oblique position (MLO) and two orientation cephalopodium position (CC)
Four striographs, CC images only contain breast portion, and MLO images also include chest muscle part, need chest first
Flesh segmentation removal, the present invention will be analyzed with more complex certain breast molybdenum target MLO original image 2.
Embodiment one
The mammary gland tissue density quantification calculation method of one embodiment of the invention, specifically includes the following steps:
Step 1 pre-processes molybdenum target MLO original image as shown in Figure 2, specific operation include pixel is taken logarithm, negate,
It is squared to make the image normalizing less standardized originally standardization;Reusing cubic spline interpolation is 2294* to pixel size
1914 image makees four times of down-sampled images for obtaining pixel size as 574*479, before guaranteeing that useful information is not blurred
Put the processing speed for improving subsequent algorithm.
Step 2 is partitioned into complete breast area from pretreated molybdenum target MLO image, image include chest muscle, breast and
Air background three parts region.Divide breast outer profile, the i.e. contour line of breast and air background, breast and air background first
Between there are apparent edge lines, seek pretreated molybdenum target image grayscale intensity histogram, intuitively find out threshold value and distinguish
Air background and breast area;Secondly segmentation breast Internal periphery, the i.e. contour line of breast and chest muscle, it is straight for defaulting contour line herein
Line, obtains the straight line parameter histogram of the two-dimentional Polar coordinates of breast area using Hough transformation, and chest muscle edge line approximation position
In polar angle 40o-80oBetween.Complete breast area binary map as shown in Figure 3, uses four times after finally being divided
Up-sampling obtains the breast area of original image, counts breast area number of pixels, is calculated as Tp.
Step 3 divides breast area using Unsupervised clustering algorithm, and the present invention uses adaptive fuzzy C mean algorithm point
Cut breast.Clustering cluster number is adaptively chosen first, and breast area pixel value is used at zero-mean normal distribution normalization
Reason, then Gaussian filter smothing filtering is used, thus acquisition treated breast grey level histogram, finally with the peak value in histogram
Number is the number of cluster, and the upper limit of setting cluster number is 15, obtains the grey level histogram at initial clustering cluster center as shown in Figure 4
Peak value figure;Secondly setting cluster central value, initial centered value takes the mean value of pixel region, ask each pixel to 15 cluster centers away from
From, with apart from minimum evaluation function, iteration is until cluster center is constant, so that breast is divided into 15 sub-regions, calculating
Each subregion to respective cluster center Euclidean distance, and make normalization obtain all subregion cluster density energy spectrogram 5, to each
Subregion label draws the tag energy spectrogram 6 of breast subregion.
Step 4 extracts feature to all subregion, extracts the correlated characteristic amount of each subregion, herein so as to right in next step
Breast classification.The whole textural characteristics with all subregion of breast, including contrast, correlation are extracted using gray level co-occurrence matrixes respectively
Degree, energy, entropy and heterogeneity;Using area describes operator extraction breast entirety and all subregion correlation attribute value, including region
Pixel number, minimum rectangle boundary, mass center;Calculate third moment value, Fourth-order moment value, skewness value and the kurtosis value of breast image;It calculates
Mean value, degree of bias value and the kurtosis value of each subregion pixel, the areal concentration that is connected be poor, compactness.
Step 5 classifies to breast area two using linear discrimination classification device LDA, identifies the body of gland of compactness in breast
Part and un-densified fats portion.Training sample set by senior image doctor by hand to 15 sub-regions labels be body of gland or
Fat is extracted feature to all subregion of sample each in training set, LDA sorter model is trained using this feature collection, finally
The classification of all subregion of the sample in test set is identified using the mature LDA classifier of training.
Step 6 identifies the body of gland part in breast area using LDA classifier, calculates the pixel of body of gland part
Number, is calculated as Bp;Breast density=Bp/Tp*100%;Breast contours (red line) and body of gland profile (green line) finally are drawn out, such as
Shown in Fig. 7.
Existing image doctor can only make qualitatively grade classification to breast density according to Breast imaging reporting and data system, give
Level results out often have very strong subjectivity, while there is also tired diagosis bring mistakes.The present invention can be realized
The full-automatic segmentation of breast tissue, realizes that the precision of gland tissue quantitatively calculates, it is close to further calculate out mammary gland in breast
Angle value significantly reduces the burden of image doctor while providing objective results.
Taking the above-mentioned ideal embodiment according to the present invention as inspiration, through the above description, relevant staff is complete
Various changes and amendments can be carried out without departing from the scope of the technological thought of the present invention' entirely.The technology of this invention
Property range is not limited to the contents of the specification, it is necessary to which the technical scope thereof is determined according to the scope of the claim.
Claims (7)
1. a kind of method that breast density quantification calculates in breast cancer risk assessment, which is characterized in that the method is specifically wrapped
Include following steps:
Step 1 pre-processes breast molybdenum target original image, carries out histogram equalization, gaussian filtering, down-sampling;
Step 2 divides breast area from pretreated breast molybdenum target image, detects breast outer edge line and chest muscle inner edge
Edge line, and then obtain breast area size;
Step 3 makees Unsupervised clustering using fuzzy C-mean algorithm to the pixel in breast area, obtains clustering cluster;
Clustering cluster merging is referred in body of gland and fatty two classes, makees to clustering cluster by step 4, the feature for extracting clustering cluster region
Two Classification and Identifications;
Step 5 carries out two classification, training linear discriminant LDA classifier to breast area;
Step 6, the number of pixels that body of gland part in breast area is calculated according to classification results, i.e. body of gland size.
2. the method that breast density quantification calculates in breast cancer risk assessment according to claim 1, which is characterized in that
The step 2 specifically includes:
Step 2.1: seeking the gray-scale intensity histogram of pretreated molybdenum target image, intuitively distinguish background air region and cream
The pixel map in room region, Cong Tuzhong finds out the differentiation threshold value of background air and mammary region, to extract breast area
Edge;
Step 2.2: tentatively extracting the edge graph of breast area by Canny boundary operator, reuse Hough transformation to obtain cream
The straight line parameter histogram of the two-dimentional Polar coordinates in room region, and chest muscle edge line is located approximately at polar determining angle model
It encloses;
Step 2.3: using the outer edge line of the extraction as boundary, the region divided is breast, by binary conversion treatment
Afterwards, interior pixels number, as breast area size are counted.
3. the method that breast density quantification calculates in breast cancer risk assessment according to claim 1, which is characterized in that
The step 3 specifically includes:
Step 3.1: cluster number is adaptively chosen first, and zero-mean normal distribution normalized is used to breast area, then
Using Gaussian filter to mammary region image smoothing filtering technique, using the peak value number of treated breast grey level histogram as poly-
Cluster number upper limit threshold is arranged in the number of class cluster, and when peak value number is more than upper limit threshold, capping threshold number is as cluster
Number;
Step 3.2: breast area pixel value being normalized and down-sampling processing, setting initial cluster center use Fuzzy C
Mean value is calculated initial centered value by pixel value equal proportion, find out each pixel in breast image to each center distance, with
Apart from minimum evaluation function, successive ignition calculates each pixel of breast area to each until cluster centre is constant
The distance at cluster center;
Step 3.3: then the pixel for belonging to each cluster in breast area being labeled, breast area is divided into the son of cluster number
Area image, and the energy spectrogram in each cluster region is drawn with the Euclidean distance at pixel each in breast to each cluster center, from energy
It measures in spectrogram substantially it can be seen that regional distribution chart of each cluster in breast.
4. the method that breast density quantification calculates in breast cancer risk assessment according to claim 1, which is characterized in that
Clustering cluster in the step 4 makees two Classification and Identifications, and specific method is that breast entirety and each son are extracted using gray level co-occurrence matrixes
The textural characteristics in region, including contrast, the degree of correlation, energy, entropy and heterogeneity;It is whole that using area describes operator extraction breast
With all subregion correlation attribute value, including area pixel number, minimum rectangle boundary, mass center;The third moment value of calculating breast image,
Fourth-order moment value, skewness value and kurtosis value;Mean value, degree of bias value and the kurtosis value of calculating each subregion pixel, be connected areal concentration
Difference, compactness.
5. the method that breast density quantification calculates in breast cancer risk assessment according to claim 1, which is characterized in that
The method that breast area carries out two classification in the step 5 includes: first to each cluster in each sample image in training set
Label is made in cluster region, needs veteran image doctor by hand to each clustering cluster area of each sample image in training set
Label is made in domain respectively, reuses and marks the image data signed to train LDA sorter model, finally uses the LDA classifier pair
Breast image makees two classification, distinguishes body of gland region and the fat region in breast.
6. the method that breast density quantification calculates in breast cancer risk assessment according to claim 1, which is characterized in that
Body of gland size can be used for calculating gland density value, calculation formula in the step 6 are as follows: and gland density value=body of gland number of pixels/
Breast number of pixels.
7. the method that breast density quantification calculates in breast cancer risk assessment according to claim 1, which is characterized in that
The breast molybdenum target original image of the step 1 is selected from MLO molybdenum target figure and CC molybdenum target figure.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910313084.7A CN109893100B (en) | 2019-04-18 | 2019-04-18 | Quantitative calculation method for breast density |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910313084.7A CN109893100B (en) | 2019-04-18 | 2019-04-18 | Quantitative calculation method for breast density |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109893100A true CN109893100A (en) | 2019-06-18 |
CN109893100B CN109893100B (en) | 2022-05-10 |
Family
ID=66954068
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910313084.7A Active CN109893100B (en) | 2019-04-18 | 2019-04-18 | Quantitative calculation method for breast density |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109893100B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112998651A (en) * | 2021-02-10 | 2021-06-22 | 中国医学科学院北京协和医院 | Application of photoacoustic imaging in breast tumor scoring system and scoring system |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100124364A1 (en) * | 2008-11-19 | 2010-05-20 | Zhimin Huo | Assessment of breast density and related cancer risk |
CN103020978A (en) * | 2012-12-14 | 2013-04-03 | 西安电子科技大学 | SAR (synthetic aperture radar) image change detection method combining multi-threshold segmentation with fuzzy clustering |
KR101258814B1 (en) * | 2011-11-09 | 2013-04-26 | 서울여자대학교 산학협력단 | Nonrigid registration method and system with density correction of each tissue and rigidity constraint of tumor in dynamic contrast-enhanced breast mr images |
CN103700085A (en) * | 2012-09-28 | 2014-04-02 | 深圳市蓝韵实业有限公司 | Cutting method of pectoral muscle region in mammary gland X-ray image |
CN105139414A (en) * | 2015-09-29 | 2015-12-09 | 盐城工学院 | Clustering integration method for image data of X-ray films |
CN106023239A (en) * | 2016-07-05 | 2016-10-12 | 东北大学 | Breast lump segmentation system and method based on mammary gland subarea density clustering |
WO2017054775A1 (en) * | 2015-09-30 | 2017-04-06 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
CN107798679A (en) * | 2017-12-11 | 2018-03-13 | 福建师范大学 | Breast molybdenum target image breast area is split and tufa formation method |
CN107958453A (en) * | 2017-12-01 | 2018-04-24 | 深圳蓝韵医学影像有限公司 | Detection method, device and the computer-readable storage medium of galactophore image lesion region |
US10037601B1 (en) * | 2017-02-02 | 2018-07-31 | International Business Machines Corporation | Systems and methods for automatic detection of architectural distortion in two dimensional mammographic images |
CN108550150A (en) * | 2018-04-17 | 2018-09-18 | 上海联影医疗科技有限公司 | Acquisition methods, equipment and the readable storage medium storing program for executing of breast density |
CN109102510A (en) * | 2018-08-03 | 2018-12-28 | 东北大学 | A kind of breast cancer pathology organization chart picture dividing method based on semi-supervised k-means algorithm |
CN109300137A (en) * | 2018-09-20 | 2019-02-01 | 北京航空航天大学 | A kind of two type fuzzy clustering magnetic resonance brain image segmentation method of more curved surface estimation intervals |
CN109598709A (en) * | 2018-11-29 | 2019-04-09 | 东北大学 | Mammary gland assistant diagnosis system and method based on fusion depth characteristic |
-
2019
- 2019-04-18 CN CN201910313084.7A patent/CN109893100B/en active Active
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20100124364A1 (en) * | 2008-11-19 | 2010-05-20 | Zhimin Huo | Assessment of breast density and related cancer risk |
KR101258814B1 (en) * | 2011-11-09 | 2013-04-26 | 서울여자대학교 산학협력단 | Nonrigid registration method and system with density correction of each tissue and rigidity constraint of tumor in dynamic contrast-enhanced breast mr images |
CN103700085A (en) * | 2012-09-28 | 2014-04-02 | 深圳市蓝韵实业有限公司 | Cutting method of pectoral muscle region in mammary gland X-ray image |
CN103020978A (en) * | 2012-12-14 | 2013-04-03 | 西安电子科技大学 | SAR (synthetic aperture radar) image change detection method combining multi-threshold segmentation with fuzzy clustering |
CN105139414A (en) * | 2015-09-29 | 2015-12-09 | 盐城工学院 | Clustering integration method for image data of X-ray films |
WO2017054775A1 (en) * | 2015-09-30 | 2017-04-06 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
CN108471995A (en) * | 2015-09-30 | 2018-08-31 | 上海联影医疗科技有限公司 | The system and method for determining breast area in medical image |
CN106023239A (en) * | 2016-07-05 | 2016-10-12 | 东北大学 | Breast lump segmentation system and method based on mammary gland subarea density clustering |
US10037601B1 (en) * | 2017-02-02 | 2018-07-31 | International Business Machines Corporation | Systems and methods for automatic detection of architectural distortion in two dimensional mammographic images |
CN107958453A (en) * | 2017-12-01 | 2018-04-24 | 深圳蓝韵医学影像有限公司 | Detection method, device and the computer-readable storage medium of galactophore image lesion region |
CN107798679A (en) * | 2017-12-11 | 2018-03-13 | 福建师范大学 | Breast molybdenum target image breast area is split and tufa formation method |
CN108550150A (en) * | 2018-04-17 | 2018-09-18 | 上海联影医疗科技有限公司 | Acquisition methods, equipment and the readable storage medium storing program for executing of breast density |
CN109102510A (en) * | 2018-08-03 | 2018-12-28 | 东北大学 | A kind of breast cancer pathology organization chart picture dividing method based on semi-supervised k-means algorithm |
CN109300137A (en) * | 2018-09-20 | 2019-02-01 | 北京航空航天大学 | A kind of two type fuzzy clustering magnetic resonance brain image segmentation method of more curved surface estimation intervals |
CN109598709A (en) * | 2018-11-29 | 2019-04-09 | 东北大学 | Mammary gland assistant diagnosis system and method based on fusion depth characteristic |
Non-Patent Citations (4)
Title |
---|
KELLER B , 等: "Adaptive multi-cluster fuzzy C-means segmentation of breast parenchymal tissue in digital mammography", 《INTERNATIONAL CONFERENCE ON MEDICAL IMAGE COMPUTING & COMPUTER-ASSISTED INTERVENTION. SPRINGER-VERLAG》 * |
KELLER B M, 等: "Estimation of breast percent density in raw and processed full field digital mammography images via adaptive fuzzy c‐means clustering and support vector machine segmentation", 《MEDICAL PHYSICS》 * |
ZHILI CHEN等: "A modified fuzzy c-means algorithm for breast tissue density segmentation in mammograms", 《PROCEEDINGS OF THE 10TH IEEE INTERNATIONAL CONFERENCE ON INFORMATION TECHNOLOGY AND APPLICATIONS IN BIOMEDICINE》 * |
孟爽,等: "超像素有偏观测模糊聚类的乳腺超声图像分割", 《中国医学物理学杂志》 * |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112998651A (en) * | 2021-02-10 | 2021-06-22 | 中国医学科学院北京协和医院 | Application of photoacoustic imaging in breast tumor scoring system and scoring system |
CN112998651B (en) * | 2021-02-10 | 2021-08-27 | 中国医学科学院北京协和医院 | Application of photoacoustic imaging in breast tumor scoring system and scoring system |
Also Published As
Publication number | Publication date |
---|---|
CN109893100B (en) | 2022-05-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Raba et al. | Breast segmentation with pectoral muscle suppression on digital mammograms | |
CN109191457B (en) | Pathological image quality validity identification method | |
EP2847738B1 (en) | Method and apparatus for image scoring and analysis | |
EP2888718B1 (en) | Methods and systems for automatic location of optic structures in an image of an eye, and for automatic retina cup-to-disc ratio computation | |
CN109190567A (en) | Abnormal cervical cells automatic testing method based on depth convolutional neural networks | |
CN107506770A (en) | Diabetic retinopathy eye-ground photography standard picture generation method | |
CN108537751B (en) | Thyroid ultrasound image automatic segmentation method based on radial basis function neural network | |
US9898824B2 (en) | Method for volume evaluation of penumbra mismatch in acute ischemic stroke and system therefor | |
CN109785293B (en) | Ultrasonic image focus segmentation method based on global and local active contour models | |
KR102155381B1 (en) | Method, apparatus and software program for cervical cancer decision using image analysis of artificial intelligence based technology | |
CN111192251B (en) | Follicle ultrasonic processing method and system based on level set image segmentation | |
CN111476754B (en) | Bone marrow cell image artificial intelligence auxiliary grading diagnosis system and method | |
CN111882561A (en) | Cancer cell identification and diagnosis system | |
CN111681230A (en) | System and method for scoring high-signal of white matter of brain | |
CN111079620A (en) | Leukocyte image detection and identification model construction method based on transfer learning and application | |
Mohammad et al. | Texture analysis for the segmentation of optic disc in retinal images | |
CN113012093A (en) | Training method and training system for glaucoma image feature extraction | |
CN109893100A (en) | A kind of method that breast density quantification calculates in breast cancer risk assessment | |
CN109816667A (en) | A kind of Lung neoplasm dividing method based on CT images | |
CN114862799B (en) | Full-automatic brain volume segmentation method for FLAIR-MRI sequence | |
CN111292285B (en) | Automatic screening method for diabetes mellitus based on naive Bayes and support vector machine | |
CN115272333B (en) | Cup-disk ratio data storage system | |
JP2023143875A (en) | Medical image analysis method | |
CN111062909A (en) | Method and equipment for judging benign and malignant breast tumor | |
CN112150406A (en) | CT image-based pneumothorax lung collapse degree accurate calculation method |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |