CN109893100B - Quantitative calculation method for breast density - Google Patents
Quantitative calculation method for breast density Download PDFInfo
- Publication number
- CN109893100B CN109893100B CN201910313084.7A CN201910313084A CN109893100B CN 109893100 B CN109893100 B CN 109893100B CN 201910313084 A CN201910313084 A CN 201910313084A CN 109893100 B CN109893100 B CN 109893100B
- Authority
- CN
- China
- Prior art keywords
- breast
- region
- cluster
- area
- 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.)
- Active
Links
Images
Abstract
The invention relates to a method for quantitatively calculating breast density in breast cancer risk assessment, which specifically comprises the following steps: preprocessing a mammary molybdenum target original image, and performing histogram equalization, Gaussian filtering and downsampling; segmenting a breast area from a breast molybdenum target original image, and detecting an outer edge line of a breast and an inner edge line of pectoralis muscles; unsupervised clustering of pixels in the breast region using the fuzzy C-means; extracting the characteristics of a cluster region, merging and classifying the obtained cluster into a gland and a fat, and performing two-classification identification on the cluster; carrying out secondary classification on the breast area, and training a linear discriminant LDA classifier; and calculating the size of the gland in the breast area according to the classification result. The quantitative calculation method for the breast density provided by the invention can realize the full-automatic segmentation of the breast tissue in the breast, and greatly reduces the burden of imaging doctors while giving objective results.
Description
Technical Field
The invention relates to the field of medical image processing, in particular to a method for quantitatively calculating breast density.
Background
The breast cancer risk assessment models are divided into two categories, one is an empirical risk model such as the Gail model, and the other is the IBIS model of the mutation of the genes with side weight. The Gail model is most commonly used, and is used for evaluating and calculating the risk value of individualized breast cancer, and comprises a plurality of risk factors related to the breast cancer, and the model gives different weight coefficients to the risk factors and obtains a quantitative risk value of cancer through weighted combination calculation. The risk factors in the Gail model include weight, age, mammary gland type, menstruation, menopause and menopause time, nursing condition, BMI index, smoking or drinking condition, mental stress condition, etc., wherein the mammary gland type is mainly classified by the breast image report and data system (BI-RADS) according to the breast image doctor with experience. The mammary gland type gives only a qualitative estimate of the gland density, and giving a quantitative calculation of the gland density will make the evaluation model more accurate.
The most common gland density quantitative calculation method at present is a semi-automatic gland segmentation method based on manual interaction of an imaging doctor, the imaging doctor manually draws a contour line of a gland in a breast under an image archiving and communication system (PACS), the sum of the number of pixels of the gland in the contour line is automatically calculated in a system tool, and the gland density value is equal to the ratio of the sum of the number of pixels in the gland to the sum of the number of pixels of the whole breast.
The semi-automatic gland segmentation method based on manual interaction of the imaging doctor is often greatly influenced by subjective factors of the imaging doctor, and the result is not objective; although the doctor images are subjected to fatigue radiographing according to the guidance of a breast image report and a data system (BI-RADS), the visual fatigue often causes errors in breast contour segmentation.
Disclosure of Invention
The invention aims to overcome the defect that an image doctor manual interactive semi-automatic gland segmentation method has errors, and provides a full-automatic mammary gland segmentation method which can quantitatively calculate the density value of a gland in a breast.
The technical scheme adopted by the invention for solving the technical problems is as follows:
a method for quantitatively calculating breast density specifically comprises the following steps:
step 4, extracting the characteristics of the cluster region, merging and classifying the cluster into a gland and a fat, and performing two-classification identification on the cluster;
and 6, calculating the number of pixels of the gland part in the breast area, namely the size of the gland according to the classification result.
As a preferred technical scheme, the mammary gland molybdenum target original image in the step 1 is selected from an MLO molybdenum target image and a CC molybdenum target image.
As a preferred technical solution, the step 2 specifically includes:
step 2.1: obtaining a gray intensity histogram of the preprocessed molybdenum target image, visually distinguishing pixel distribution maps of a background air region and a breast region, and finding out a distinguishing threshold value of the background air region and the breast region from the map so as to extract the edge of the breast region;
step 2.2: preliminarily extracting an edge image of the breast area through a Canny edge operator, and then obtaining a two-dimensional polar linear parameter histogram of the breast area by using Hough transform, wherein the edge linear of the breast muscle is approximately positioned in a determined angle range of polar coordinates;
step 2.3: and taking the extracted inner and outer edge lines as boundaries, obtaining a region obtained by segmentation as a breast, and counting the number of internal pixels after binarization processing, namely the size of the breast region.
As a preferred technical solution, the step 3 specifically includes:
step 3.1: firstly, adaptively selecting the number of clusters, carrying out zero-mean normal distribution normalization processing on a breast area, then carrying out smooth filtering on a breast area image by using a Gaussian filter, setting the upper limit threshold of the number of clusters by taking the number of peak values of a processed breast gray level histogram as the number of the cluster clusters, and taking the number of the upper limit threshold as the number of the clusters when the number of the peak values exceeds the upper limit threshold;
step 3.2: normalizing and downsampling the breast area pixel values, setting an initial clustering center, calculating the initial center value in an equal proportion according to the pixel values by using a fuzzy C mean value, solving the distance from each pixel to each center in the breast image, taking the minimum distance as an evaluation function, iterating for multiple times until the clustering center is unchanged, and calculating the distance from each pixel to each cluster center in the breast area;
step 3.3: and then labeling pixels belonging to each cluster in the breast area, dividing the breast area into sub-area images with the number of clusters, drawing an energy spectrogram of each cluster area according to Euclidean distance from each pixel in the breast to the center of each cluster, and roughly seeing the area distribution diagram of each cluster in the breast from the energy spectrogram.
As a preferred technical scheme, the cluster in step 4 is subjected to two-classification identification, and the specific method is to extract texture features of the whole breast and each sub-region by using a gray level co-occurrence matrix, wherein the texture features comprise contrast, correlation, energy, entropy and heterogeneity; extracting relevant attribute values of the whole breast and each sub-region by using a region description operator, wherein the attribute values comprise region pixel number, minimum rectangular boundary and centroid; calculating a third moment value, a fourth moment value, a skewness value and a kurtosis value of the breast image; and calculating the mean value, the skewness value and the kurtosis value of each sub-region pixel, the density difference of the associated region and the compactness.
As a preferred technical solution, the method for classifying the breast region in step 5 includes: firstly labeling each cluster region in each sample image in a training set, manually labeling each cluster region of each sample image in the training set by an experienced imaging doctor, then training an LDA classifier model by using image data labeled with the labels, and finally performing secondary classification on breast images by using the LDA classifier to distinguish a gland region and a fat region in a breast.
As a preferred technical solution, the size of the gland in the step 6 can be used for calculating the gland density value, and the calculation formula is as follows: gland density value = number of gland pixels/number of breast pixels.
Through the technical scheme, compared with the prior art, the invention has the following beneficial effects: and analyzing the molybdenum target image of the mammary gland according to an artificial intelligence algorithm and a big data technology, automatically segmenting a gland region from the breast image, and calculating the gland density value. Compared with a semi-automatic interactive method, the method has the advantages that the calculation result is more objective, and meanwhile, the burden of a large number of film reading of imaging doctors is relieved.
Drawings
The invention is further illustrated by the following examples in conjunction with the drawings.
FIG. 1 is a flow chart of a quantitative calculation method for breast glandular tissue density according to an embodiment of the present invention;
FIG. 2 is an MLO raw image of a breast molybdenum target according to an embodiment of the present invention;
FIG. 3 is a two-value map of the same breast area in an embodiment of the present invention;
FIG. 4 is a histogram of gray levels corresponding to the number of initial clusters in the same breast in the embodiment of the present invention;
FIG. 5 is a density energy spectrum of the same cluster of sub-regions of the breast in an embodiment of the present invention;
FIG. 6 is a plot of the energy spectrum of the same tag in the breast sub-region in an embodiment of the present invention;
FIG. 7 shows the same breast contour (outer) and gland contour (inner) for the present embodiment.
Detailed Description
The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic views illustrating only the basic structure of the present invention in a schematic manner, and thus show only the constitution related to the present invention.
Fig. 1 is a flowchart of a breast glandular tissue density quantification calculation method according to an embodiment of the invention. As shown, the method comprises the steps of:
2.1, obtaining a gray intensity histogram of the preprocessed molybdenum target image, visually distinguishing pixel distribution maps of a background air region and a breast region, and finding out a distinguishing threshold value of the background air region and the breast region from the image so as to extract the edge of the breast region;
2.2, preliminarily extracting an edge image of the breast area through a Canny edge operator, obtaining a two-dimensional polar linear parameter histogram of the breast area by using Hough transform, wherein the edge linear of the breast muscle is approximately positioned in a determined angle range of polar coordinates, and calculating the number of pixels of the binarized breast area, namely the size of the breast area.
3.1, firstly, adaptively selecting the cluster number, carrying out zero-mean normal distribution normalization processing on a breast area, then, carrying out smooth filtering on a breast area image by using a Gaussian filter, and taking the peak value number of the processed breast gray level histogram as a cluster center;
3.2, carrying out normalization and downsampling processing on the breast area pixel values, setting an initial clustering center, calculating the initial center value by using a fuzzy C mean value according to the pixel value in an equal proportion, calculating the distance from each pixel to each center in the breast image, taking the minimum distance as an evaluation function, carrying out multiple iterations until the clustering center is unchanged, and calculating the distance from each pixel to each cluster center in the breast area;
and marking pixels belonging to each cluster in the breast area, dividing the breast area into sub-area images with the number of clusters, and drawing an energy spectrogram of each cluster area according to Euclidean distance from each pixel in the breast to the center of each cluster, wherein the area distribution diagram of each cluster in the breast can be roughly seen from the energy spectrogram.
And 4, extracting the characteristics of a cluster region, wherein the interior of the breast region mainly comprises two parts, namely, glandular fat and non-glandular fat, the cluster obtained in the step 3 is merged and classified into glandular fat and fatty fat, namely, the cluster is subjected to binary identification, proper characteristics are extracted from each cluster, and the characteristics extracted from each cluster sub-region comprise gray histogram statistical characteristics (such as curvature and kurtosis), gray texture characteristics (such as energy and entropy) and morphological descriptors (such as compactness and perimeter).
And 5, secondary classification of breast areas and training of a linear discriminant LDA classifier, namely labeling each cluster area in each sample image in a training set, labeling each cluster area of each sample image in the training set manually by an experienced imaging doctor, training an LDA classifier model by using image data labeled with labels, and performing secondary classification on the breast images by using the LDA classifier to obtain a gland area part.
And 6, calculating the number of pixels of the gland part in the breast area according to the classification result, namely the size of the gland, wherein the gland density value = the number of gland pixels/number of breast pixels.
The breast molybdenum target image usually comprises left and right breast four-image images of two orientations of medial oblique position (MLO) and head-foot position (CC), the CC image only comprises the breast part, while the MLO image also comprises the pectoral muscle part, the pectoral muscle needs to be segmented and removed, and the invention analyzes with a more complex MLO original image 2 of a certain breast molybdenum target.
Example one
The breast gland tissue density quantitative calculation method provided by the embodiment of the invention specifically comprises the following steps:
And 2, segmenting a complete breast area from the preprocessed molybdenum target MLO image, wherein the image comprises three areas of pectoralis muscle, breast and air background. Firstly, segmenting the outer contour of a breast, namely contour lines of the breast and an air background, wherein an obvious edge line exists between the breast and the air background, solving a gray intensity histogram of a preprocessed molybdenum target image, and visually finding out a threshold value to distinguish the air background from a breast area; the intramammary contours, i.e. the contours of the breast and pectoral muscles, are then segmented, where the default contours are straight lines, and a two-dimensional polar-coordinated histogram of straight-line parameters of the breast area is obtained using the Hough transform, with the straight lines at the edge of the pectoral muscles approximately lying at a polar angle of 40o-80oIn the meantime. Finally, a complete segmented breast area binary image as shown in fig. 3 is obtained, the breast area of the original image is obtained by four times of up-sampling, and the number of pixels in the breast area is counted and counted as Tp.
And 3, segmenting the breast area by using an unsupervised clustering algorithm, and segmenting the breast by using a self-adaptive fuzzy C-means algorithm. Firstly, adaptively selecting the number of clustering clusters, carrying out zero-mean normal distribution normalization processing on the pixel values of the breast area, and then carrying out smooth filtering by using a Gaussian filter, thereby obtaining a processed breast gray level histogram, and finally setting the upper limit of the number of clusters to be 15 by taking the number of peak values in the histogram as the number of the clusters, so as to obtain a gray level histogram peak value diagram of the initial clustering cluster center as shown in FIG. 4; setting a cluster center value, taking the average value of a pixel area as an initial center value, calculating the distance from each pixel to 15 cluster centers, taking the minimum distance as an evaluation function, iterating until the cluster centers are unchanged, thereby dividing the breast into 15 sub-areas, calculating the Euclidean distance from each sub-area to the respective cluster centers, normalizing to obtain a density energy spectrogram 5 of each sub-area cluster, and drawing a label energy spectrogram 6 of the breast sub-area for each sub-area label.
And 4, extracting characteristics of each sub-region, wherein the related characteristic quantity of each sub-region is extracted, so that the breast can be classified in the next step. Extracting texture characteristics of the whole breast and each sub-region respectively by using the gray level co-occurrence matrix, wherein the texture characteristics comprise contrast, correlation, energy, entropy and heterogeneity; extracting relevant attribute values of the whole breast and each sub-region by using a region description operator, wherein the attribute values comprise region pixel number, minimum rectangular boundary and centroid; calculating a third moment value, a fourth moment value, a skewness value and a kurtosis value of the breast image; and calculating the mean value, the skewness value and the kurtosis value of each sub-region pixel, the density difference of the associated region and the compactness.
And 5, classifying the breast area II by using a linear discriminant classifier LDA, and identifying a dense gland part and a non-dense fat part in the breast. Manually labeling 15 sub-regions as glands or fat by a senior image doctor, extracting features of each sub-region of each sample in the training sample set, training an LDA classifier model by using the feature set, and finally identifying the category of each sub-region of the sample in the test set by using a well-trained LDA classifier.
Step 6, identifying the gland part in the breast area by using an LDA classifier, and calculating the number of pixels of the gland part to be counted as Bp; breast density = Bp/Tp 100%; the contours of the breast (red lines) and the glands (green lines) are finally drawn, as shown in fig. 7.
The existing imaging doctors can only carry out qualitative grade division on the breast density according to a breast image report and a data system, the given grade result is very subjective, and meanwhile, errors caused by fatigue radiograph reading exist. The invention can realize the full-automatic segmentation of the mammary tissue in the breast, realize the precise quantitative calculation of the glandular tissue, further calculate the density value of the mammary gland, and greatly reduce the burden of imaging doctors while giving objective results.
In light of the foregoing description of the preferred embodiment of the present invention, many modifications and variations will be apparent to those skilled in the art without departing from the spirit and scope of the invention. The technical scope of the present invention is not limited to the content of the specification, and must be determined according to the scope of the claims.
Claims (5)
1. A method for quantitatively calculating breast density is characterized by comprising the following steps:
step 1, preprocessing a mammary molybdenum target original image, and performing histogram equalization, Gaussian filtering and downsampling; the preprocessing comprises the operations of taking logarithm, negation and squaring of pixels to normalize an originally less-standard image; then, carrying out quadruple down-sampling on the image by using cubic spline interpolation to obtain an image with non-blurred useful information, and improving the processing speed of a subsequent algorithm;
step 2, segmenting a breast area from the preprocessed breast molybdenum target image, detecting an outer edge line of the breast and an inner edge line of the pectoral muscle, and further obtaining the size of the breast area;
step 2.1: obtaining a gray intensity histogram of the preprocessed molybdenum target image, visually distinguishing pixel distribution maps of a background air region and a breast region, and finding out a distinguishing threshold value of the background air region and the breast region from the map so as to extract the edge of the breast region;
step 2.2: preliminarily extracting an edge image of the breast area through a Canny edge operator, and then obtaining a two-dimensional polar linear parameter histogram of the breast area by using Hough transform, wherein the edge linear of the breast muscle is approximately positioned in a determined angle range of polar coordinates;
step 2.3: taking the extracted inner and outer edge lines as boundaries, obtaining a region obtained by segmentation as a breast, and counting the number of internal pixels after binarization processing, namely the size of the breast region;
step 3, carrying out unsupervised clustering on the pixels in the breast area by using the fuzzy C mean value to obtain a clustering cluster;
step 3.1: firstly, adaptively selecting the number of clusters, carrying out zero-mean normal distribution normalization processing on a breast area, then carrying out smooth filtering on a breast area image by using a Gaussian filter, setting the upper limit threshold of the number of clusters by taking the number of peak values of a processed breast gray level histogram as the number of the cluster clusters, and taking the number of the upper limit threshold as the number of the clusters when the number of the peak values exceeds the upper limit threshold;
step 3.2: normalizing and downsampling the pixel values of the breast area, setting an initial clustering center, calculating the initial center value in an equal proportion according to the pixel values by using a fuzzy C mean value, solving the distance from each pixel to each center in the breast image, taking the minimum distance as an evaluation function, iterating for multiple times until the clustering center is unchanged, and calculating the distance from each pixel to each cluster center in the breast area;
step 3.3: marking pixels belonging to each cluster in the breast area, dividing the breast area into sub-area images with the number of clusters, iterating until the cluster center is unchanged by taking the minimum distance as an evaluation function, obtaining Euclidean distance from each pixel in the breast to the cluster center, drawing an energy spectrogram of each cluster area, and approximately seeing the area distribution diagram of each cluster in the breast from the energy spectrogram;
step 4, extracting the characteristics of the cluster region, merging and classifying the cluster into a gland and a fat, and performing two-classification identification on the cluster;
step 5, carrying out secondary classification on the breast area, and training a linear discriminant LDA classifier;
and 6, calculating the number of pixels of the gland part in the breast area, namely the size of the gland according to the classification result.
2. The method for quantitative calculation of breast density according to claim 1, wherein the clustering in step 4 is used for classification identification, and the method is to extract texture features of the whole breast and each sub-region, including contrast, correlation, energy, entropy and heterogeneity, by using gray level co-occurrence matrix; extracting relevant attribute values of the whole breast and each sub-region by using a region description operator, wherein the attribute values comprise region pixel number, minimum rectangular boundary and centroid; calculating a third moment value, a fourth moment value, a skewness value and a kurtosis value of the breast image; and calculating the mean value, the skewness value and the kurtosis value of each sub-region pixel, the density difference of the associated region and the compactness.
3. The method for quantitative calculation of breast density as claimed in claim 1, wherein the step 5 method for classifying the breast area twice comprises the following steps: firstly labeling each cluster region in each sample image in a training set, manually labeling each cluster region of each sample image in the training set by an experienced imaging doctor, then training an LDA classifier model by using image data labeled with the labels, and finally performing secondary classification on breast images by using the LDA classifier to distinguish a gland region and a fat region in a breast.
4. The method of claim 1, wherein the size of the gland in step 6 can be used to calculate the density of the gland by the following formula: gland density value = number of gland pixels/number of breast pixels.
5. The method for quantitative calculation of breast density according to claim 1, wherein the breast molybdenum target raw image of step 1 is selected from MLO molybdenum target map and CC molybdenum target map.
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 CN109893100A (en) | 2019-06-18 |
CN109893100B true 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) |
Families Citing this family (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
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 |
Family Cites Families (8)
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 |
CN103020978B (en) * | 2012-12-14 | 2015-07-15 | 西安电子科技大学 | SAR (synthetic aperture radar) image change detection method combining multi-threshold segmentation with fuzzy 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 |
CN107958453B (en) * | 2017-12-01 | 2022-01-28 | 深圳蓝韵医学影像有限公司 | Method and device for detecting lesion region of mammary gland image and computer storage medium |
CN107798679B (en) * | 2017-12-11 | 2021-04-27 | 福建师范大学 | Breast region segmentation and calcification detection method for mammary gland molybdenum target image |
CN109102510B (en) * | 2018-08-03 | 2022-08-26 | 东北大学 | Breast cancer pathological tissue image segmentation method based on semi-supervised k-means algorithm |
CN109300137B (en) * | 2018-09-20 | 2021-05-07 | 北京航空航天大学 | Two-type fuzzy clustering magnetic resonance brain image segmentation method for multi-surface estimation interval |
CN109598709B (en) * | 2018-11-29 | 2023-05-26 | 东北大学 | Mammary gland auxiliary diagnosis system and method based on fusion depth characteristic |
-
2019
- 2019-04-18 CN CN201910313084.7A patent/CN109893100B/en active Active
Patent Citations (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
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 |
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 |
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 |
Non-Patent Citations (2)
Title |
---|
A modified fuzzy c-means algorithm for breast tissue density segmentation in mammograms;Zhili Chen等;《Proceedings of the 10th IEEE International Conference on Information Technology and Applications in Biomedicine》;20101105;全文 * |
超像素有偏观测模糊聚类的乳腺超声图像分割;孟爽,等;《中国医学物理学杂志》;20170731;第34卷(第7期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109893100A (en) | 2019-06-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Raba et al. | Breast segmentation with pectoral muscle suppression on digital mammograms | |
JP4634418B2 (en) | Automated method and apparatus for detecting masses and parenchymal tissue deformation in medical images using a computer | |
EP3432263A1 (en) | Semantic segmentation for cancer detection in digital breast tomosynthesis | |
US9959617B2 (en) | Medical image processing apparatus and breast image processing method thereof | |
EP2847738B1 (en) | Method and apparatus for image scoring and analysis | |
El Atlas et al. | Computer-aided breast cancer detection using mammograms: A review | |
Maji et al. | An automated method for counting and characterizing red blood cells using mathematical morphology | |
WO2013019856A1 (en) | Automated malignancy detection in breast histopathological images | |
Sarwar et al. | Segmentation of cervical cells for automated screening of cervical cancer: a review | |
Alam et al. | Pectoral muscle elimination on mammogram using K-means clustering approach | |
CN113160185A (en) | Method for guiding cervical cell segmentation by using generated boundary position | |
Saltanat et al. | An efficient pixel value based mapping scheme to delineate pectoral muscle from mammograms | |
CN109893100B (en) | Quantitative calculation method for breast density | |
CN117274278B (en) | Retina image focus part segmentation method and system based on simulated receptive field | |
CN110782451B (en) | Suspected microcalcification area automatic positioning method based on discriminant depth confidence network | |
Rajkumar et al. | Automated mammogram segmentation using seed point identification and modified region growing algorithm | |
CN111062909A (en) | Method and equipment for judging benign and malignant breast tumor | |
Suhail et al. | Histogram-based approach for mass segmentation in mammograms | |
CN113870194A (en) | Deep layer characteristic and superficial layer LBP characteristic fused breast tumor ultrasonic image processing device | |
Lakshmanan et al. | Pectoral Muscle Boundary detection-A preprocessing method for early breast cancer detection | |
Boujelben et al. | Automatic application level set approach in detection calcifications in mammographic image | |
Tatjana et al. | Computer-aided Analysis and Interpretation of HRCT Images of the Lung | |
CN117576127B (en) | Liver cancer area automatic sketching method based on pathological image | |
Fazilov et al. | Improvement of Image Enhancement Technique for Mammography Images | |
Feudjio et al. | Automatic Extraction of breast region in raw mammograms using a combined strategy |
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 |