CN101103924A - Breast cancer computer-aided diagnosis method and system based on mammography - Google Patents
Breast cancer computer-aided diagnosis method and system based on mammography Download PDFInfo
- Publication number
- CN101103924A CN101103924A CNA2007100527471A CN200710052747A CN101103924A CN 101103924 A CN101103924 A CN 101103924A CN A2007100527471 A CNA2007100527471 A CN A2007100527471A CN 200710052747 A CN200710052747 A CN 200710052747A CN 101103924 A CN101103924 A CN 101103924A
- Authority
- CN
- China
- Prior art keywords
- region
- mass
- suspicious
- suspicious mass
- interest
- 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
- 238000000034 method Methods 0.000 title claims abstract description 58
- 206010006187 Breast cancer Diseases 0.000 title claims abstract description 33
- 208000026310 Breast neoplasm Diseases 0.000 title claims abstract description 33
- 238000004195 computer-aided diagnosis Methods 0.000 title claims abstract description 22
- 238000009607 mammography Methods 0.000 title claims description 10
- 238000000605 extraction Methods 0.000 claims abstract description 28
- 238000003745 diagnosis Methods 0.000 claims abstract description 26
- 230000011218 segmentation Effects 0.000 claims abstract description 25
- 210000000481 breast Anatomy 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 17
- 238000012216 screening Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 4
- 230000000717 retained effect Effects 0.000 claims 1
- 238000012545 processing Methods 0.000 abstract description 10
- 206010028980 Neoplasm Diseases 0.000 description 33
- 210000005075 mammary gland Anatomy 0.000 description 15
- 238000002601 radiography Methods 0.000 description 15
- 238000001514 detection method Methods 0.000 description 12
- 238000004458 analytical method Methods 0.000 description 10
- 238000012549 training Methods 0.000 description 8
- 208000004434 Calcinosis Diseases 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 206010006272 Breast mass Diseases 0.000 description 5
- 210000001519 tissue Anatomy 0.000 description 5
- 238000010586 diagram Methods 0.000 description 4
- 125000000205 L-threonino group Chemical group [H]OC(=O)[C@@]([H])(N([H])[*])[C@](C([H])([H])[H])([H])O[H] 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 201000011510 cancer Diseases 0.000 description 3
- 208000030270 breast disease Diseases 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000013399 early diagnosis Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 230000000877 morphologic effect Effects 0.000 description 2
- ZOKXTWBITQBERF-UHFFFAOYSA-N Molybdenum Chemical compound [Mo] ZOKXTWBITQBERF-UHFFFAOYSA-N 0.000 description 1
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000002308 calcification Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000000762 glandular Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 230000003902 lesion Effects 0.000 description 1
- 229910052750 molybdenum Inorganic materials 0.000 description 1
- 239000011733 molybdenum Substances 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 238000003909 pattern recognition Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Landscapes
- Apparatus For Radiation Diagnosis (AREA)
Abstract
本发明公开了一个基于乳腺x线摄片的乳腺癌计算机辅助诊断方法及其系统。本发明首先给系统输入一待诊断的乳腺x线摄片，通过感兴趣区域提取模块、可疑肿块区域分割模块和可疑肿块区域的特征提取模块的处理，得到有关于可疑肿块区域的一系列相关特征值，然后将特征值输入已训练的分类器对可疑肿块区域进行分类识别，最后将计算机自动检测的最终可疑肿块区域的分割结果定位在输入的待诊断的乳腺x线摄片上，并将计算得到的区域相关特征值根据需要显示给放射科医师，从而提示放射科医师需要重点关注的区域和区域相关重点参数。本发明能在一定程度上提高放射科医师对乳腺癌诊断的准确性和效率，更客观有效地辅助放射科医师提出诊断意见和治疗方案。
The invention discloses a mammogram-based computer-aided diagnosis method and system for breast cancer. The present invention firstly inputs a mammogram to be diagnosed into the system, and obtains a series of related features about the suspicious mass area through the processing of the region of interest extraction module, the suspicious mass area segmentation module and the feature extraction module of the suspicious mass area value, and then input the feature value into the trained classifier to classify and identify the suspicious mass area, and finally locate the segmentation result of the final suspicious mass area automatically detected by the computer on the input mammogram to be diagnosed, and calculate the obtained The region-related eigenvalues of are displayed to the radiologist as needed, thereby prompting the radiologist to focus on the region and region-related key parameters. The invention can improve the accuracy and efficiency of radiologists in diagnosing breast cancer to a certain extent, and more objectively and effectively assist radiologists to put forward diagnosis opinions and treatment plans.
Description
Technical Field
The invention belongs to the field of application of computer analysis technology of medical images, and particularly relates to a breast cancer computer-aided diagnosis method and system based on a breast X-ray radiography.
Background
Breast cancer is one of the most common malignant tumors of middle-aged and old women, and the breast cancer is the first malignant tumor of women living in China and tends to rise year by year. At present, no good strategy exists for preventing breast cancer, and early diagnosis is the most effective way for reducing morbidity and mortality and improving the cure rate of breast cancer. Studies have shown that mammograms (breast molybdenum target x-ray technique) are an effective method for screening early clinical asymptomatic breast cancer, a general screening method for breast disease that has been approved by the Food and Drug Administration (FDA).
The imaging principle limits that high-density normal breast tissues in a mammography can also have high brightness similar to abnormal tissues, the focus of young Asian women with dense breast tissues is not easy to find, missed diagnosis and misdiagnosis are easy to cause, the early diagnosis of the breast cancer depends on the experience, professional ability and fatigue degree of a radiologist to a great extent, the subjective factors have great influence, and particularly, the diagnosis efficiency and accuracy are difficult to ensure in the general investigation activity of the breast diseases.
The computer-aided diagnosis (CAD) method and the system of the breast cancer combine the strong computing power of a computer in the automatic analysis of the x-ray radiography of the breast, effectively provide valuable 'second opinion' for radiologists, and have incomparable advantages in the aspects of reducing the diagnosis workload and improving the diagnosis efficiency and objectivity. The results of the Study by T.W.Freer et al show that the clinical use of Breast cancer Computer-aided diagnosis and its system significantly improves the Detection rate of early malignant tumors (see T.W.Freer, et al, "Screening Mammgraphics with Computer-aided Detection: prospectral Study of 12,860 Patents in a Community Breast Center," Radiology,220 (3): 781-786 (2001)). As such, computer-aided diagnosis methods of breast cancer and systems thereof are being increasingly accepted by medical institutions worldwide and applied to clinical practice.
Generally, computer-aided diagnosis methods and systems for breast cancer include the aided diagnosis of breast masses and the aided diagnosis of microcalcifications (see, in particular, david Gur., bin Zheng, yuan Hsiang Chang, computerized detection of tumors and microcalcifications in digital models, united States Patent No.5,627,907). The auxiliary diagnosis is to automatically analyze and process the mammary gland image through a computer, identify suspicious masses or lesion regions of suspicious microcalcification foci in the image, acquire a series of relevant characteristic parameters of the marked suspicious masses or suspicious microcalcification foci regions, return and display the parameters to a radiologist according to needs, and provide reference ideas for assisting diagnosis. Currently, computer-aided diagnosis methods and systems thereof have the characteristics of high sensitivity and low false positive rate in the detection of suspicious areas of microcalcifications, and have been accepted and adopted by radiologists (see, in particular, r.f. brem, j.w. hoffmeister, g.zisman, et al, "a computer-aided detection system for evaluation of breast cancer by breast pathological application and delivery size," AJR, am.j.rogentonol 184 (2005)), but the detection effect in breast mass-aided diagnosis is low and is not relied on by radiologists. Therefore, how to design a computer-aided diagnosis method and system for breast cancer to accurately determine the suspicious breast cancer mass region in the x-ray mammogram, and extract a series of effective region-related characteristic parameters, which accurately and effectively assist the doctor in breast cancer diagnosis becomes the focus of research.
Disclosure of Invention
The invention aims to provide a breast cancer computer-aided diagnosis method based on a breast X-ray radiography, which provides reference opinions for diagnosis of radiologists and is beneficial to improving the efficiency and the accuracy of positioning and diagnosing breast cancer masses; the invention also provides a system for realizing the method.
The invention provides a breast cancer computer-aided diagnosis method based on a breast X-ray radiography, which comprises the following steps:
inputting a mammary gland x-ray radiography to be diagnosed;
step (2) extracting a region of interest from the input mammography x-ray photograph to obtain the position of an initial suspicious lump region;
step (3) segmenting suspicious masses in the region of interest and determining the boundaries of the suspicious masses;
step (4) calculating the relevant characteristic value of the segmented suspicious lump area;
inputting the calculated characteristic values into a classifier, analyzing the initial suspicious tumor area and determining the final suspicious tumor area;
and (6) positioning the final segmentation result of the suspicious tumor area on the X-ray mammogram to be diagnosed input in the step (1), and displaying the characteristic value of the tumor area calculated in the step (4) to a user according to the requirement.
The invention provides a breast cancer computer-aided diagnosis system based on a breast x-ray radiography, which comprises an input module, a region-of-interest extraction module, a suspicious mass segmentation module, a suspicious mass region characteristic extraction module, a classification diagnosis module and an output module, wherein the input module is used for inputting a breast x-ray radiography image;
the input module is used for receiving a mammary gland x-ray radiography to be diagnosed input by a user and transmitting the mammary gland x-ray radiography to the interest region extraction module;
the interesting region extraction module is used for extracting an interesting region in an input mammary gland X-ray photograph to obtain initial suspicious tumor region position information and transmitting the position information to the suspicious tumor segmentation module;
the suspicious tumor segmentation module is used for segmenting the suspicious tumor in the region of interest extracted by the region of interest extraction module to obtain boundary information of the suspicious tumor and transmitting the boundary information to the suspicious tumor region characteristic extraction module;
the suspicious tumor region feature extraction module calculates to obtain a related feature value according to the received boundary information of the suspicious tumor and transmits the related feature value to the classification diagnosis module;
the classification diagnosis module is used for inputting the calculated characteristic value of each suspicious lump region into the classifier, automatically classifying and identifying the initial suspicious lump region by a computer, determining the final suspicious lump region, transmitting the result to the output module, positioning the segmentation result of the detected final lump region on the input X-ray mammary gland photograph to be diagnosed by the output module, and displaying the calculated region-related characteristic value to a user according to the requirement.
Firstly, inputting a mammary gland x-ray photograph to be diagnosed into a system, obtaining a series of related characteristic values of a suspicious mass through region-of-interest extraction, suspicious mass segmentation and characteristic extraction of a suspicious mass region (a region containing the suspicious mass), then inputting the characteristic values into a classifier to classify and diagnose an initial suspicious mass region, finally positioning the segmentation result of the final suspicious mass region automatically detected by a computer on the input mammary gland x-ray photograph to be diagnosed, and displaying the series of related characteristic values obtained by calculation to a user according to needs. In conclusion, the method automatically detects the suspicious breast mass through a breast X-ray radiography-based breast cancer computer-aided diagnosis method and a breast X-ray radiography-based breast cancer computer-aided diagnosis system, provides the position and the shape of the suspicious mass, and provides a series of relevant characteristic parameters of the suspicious mass according to the requirements, thereby prompting the radiologist to focus on the area needing attention and the relevant key parameters of the area, and improving the accuracy and the efficiency of the radiologist in breast cancer diagnosis to a certain extent.
Drawings
FIG. 1 is a flow chart of a breast cancer computer-aided diagnosis method based on mammography according to the present invention;
FIG. 2 is a schematic structural diagram of a breast cancer computer-aided diagnosis system based on mammography according to the present invention;
FIG. 3 is a schematic diagram illustrating a process of obtaining candidate boundary points according to an embodiment of the present invention;
FIG. 4 is a list of features extracted from a suspicious mass region according to an embodiment of the present invention;
FIG. 5 is a diagram illustrating a classifier training process according to an embodiment of the present invention.
Detailed Description
The present invention is described in further detail below with reference to the attached drawings and examples.
As shown in fig. 1, the method of the present invention comprises the steps of:
(1) Inputting a mammary gland X-ray photograph to be diagnosed.
(2) And extracting the region of interest on the input mammary gland x-ray photograph to obtain the initial suspicious lump region position.
The analysis of the whole image not only has a large amount of redundant information, but also is easy to introduce errors. In order to improve the speed and accuracy of the processing, the processing object needs to be reduced from the whole image to a plurality of small areas, namely, regions of interest, and the positions of the regions of interest are the positions of the initial suspicious lump areas of the subsequent processing.
For region of interest extraction, a left and right breast image comparison method has been studied (see in particular F.F. Yin, M.L. Giger, doi Kunio, and et al, "Computerized detection of massages in digital mammograms: analysis of biological events," Medical Physics, 18.
Considering the characteristics of high brightness, approximately circular shape, and contrast due to gray level difference with surrounding tissues, the method of the present invention uses a template matching method to locate the region of interest, i.e., the initial suspicious tumor region. The concrete implementation is as follows:
(2.1) obtaining a template T by utilizing a two-dimensional hyperbolic secant (sech) function, and calculating the correlation between the input mammary gland x-ray radiography and the template T to obtain a correlation image;
a two-dimensional hyperbolic secant (sech) function is adopted to generate a template T with the size of (2L + 1) x (2L + 1)
The center of the template is used as an origin, x and y represent horizontal and vertical coordinates in the template, the value ranges are [ -L, L ], and L is a positive integer. α = input mammogram maximum gray level-1, β = ln (2 × α)/L × L.
Moving the template T on the input mammary gland x-ray photograph pixel by pixel, and calculating the correlation cor (T, S) between the template T and the subimage S covered by the template T by using the formula (2)
Wherein, mu st Is the average value, mu, of the gray-scale products of the corresponding pixels of the sub-image S and the template T t And mu s Is the average of the pixel gray levels, σ, within the template T and the sub-image S t And σ s Is the variance value of the pixel gray levels in the template T and the sub-image S. By applying the method, each pixel point in the input mammary gland X-ray radiography can be calculated to obtain a correlation value related to the template T, and the correlation value range is [ -1,1]. Then all the correlation degrees are calibrated to obtain a correlation degree graphLike this. Setting the correlation degree smaller than 0 as 0; otherwise, the input mammography maximum gray level is multiplied by the correlation value.
(2.2) carrying out binarization processing on the correlation degree image by using a selection threshold, extracting all connected regions from the binarization correlation degree image and determining the center of each connected region;
selecting a suitable threshold T low Performing binarization on the correlation image obtained in the step (2.1)That is, if the correlation value of a certain pixel point of the correlation image is smaller than T low Then the point is assigned a value of 0, otherwise it is assigned a value of 1. And extracting a plurality of connected regions from the correlation degree binary image. And (3) taking the point with the maximum correlation degree in the corresponding region in the correlation degree image obtained in the step (2.1) as the center of each connected region.
(2.3) redefining a template with a size different from that of the template in the step (2.1) by using the template defining method in the step (2.1), and performing multi-scale analysis on each connected region extracted in the step (2.2);
according to the template defining method described in step (2.1), 3 templates are redefined according to 200%,33% and 66% of the original template size 2L + 1. For each connected region extracted in the step (2.2), moving the center of each scale template to the position of the center of the connected region on the original input mammogram, respectively calculating the correlation degree of each scale template and the corresponding sub-image in the original input mammogram image, taking the maximum value of the calculated 3 correlation degrees and the correlation degree obtained in the step (2.1) as the final correlation degree of the connected region, and selecting a proper threshold value T high (T high Greater than T low ) The final degree of correlation is less than T high The connected regions of (c) are excluded as false positive regions (non-tumor regions considered by the computer).
(2.4) screening the residual connected regions processed in the step (2.3) by using area and shape rules;
and (4) further screening the connected regions which are not excluded in the multi-scale analysis method in the step (2.3) by adopting simple area and shape rules. In general, a connected region with a small area (usually 5-30 pixels) is usually corresponding to a mammary calcification focus, while a long-strip-shaped region is usually corresponding to normal glandular tissue, and the connected region satisfying the two conditions is excluded as a false positive region.
(2.5) extracting a region of interest;
after the region screening in step (2.4), a square (usually 125 × 125 pixels) is cut out from the original input breast image for the remaining connected region by taking its geometric center as the center, and the extracted square region is the region of interest or referred to as the initial suspicious mass region.
(3) And (3) segmenting the suspicious masses in the region of interest extracted in the step (2), and determining the boundaries of the suspicious masses.
After the region of interest is extracted, suspicious masses contained in the region of interest need to be segmented, and the boundaries of the suspicious masses need to be accurately determined.
Multi-layer topographic region growing has been studied for suspicious mass Segmentation (see in particular b.zheng, y.h. chang, d.gu. "Computerized Detection of massages in differential Mammograms using single-Image Segmentation and a multi-layer topographic visualization analysis," ad.radio., 2: 259-266 (2001)), segmentation based on multi-resolution analysis (see in particular Liu, C.babbs, and E.Delp, "Multiresolution Detection of partitioned versions in Digital algorithms," IEEE Transactions on Image Processing,10 (6): 874 884 (2001)), threshold Segmentation based on Fuzzy entropy (see in particular S.Amr.abdel-Daym, mahmoud r.El-Sakka, "Fuzzy entry based segmented Detection of partitioned versions in Digital algorithms," Images Processing of the 2005 Engineering and biological analysis number reference, 4017-4022 (Septer 2005, 2005), and so on.
The method adopts a segmentation method based on image gradient and dynamic programming method. The concrete implementation is as follows:
(3.1) determining candidate boundary points on the boundary of the suspicious lump by using the gradient correlation characteristics of the image of the region of interest;
as shown in the left diagram of fig. 3, for the step (2) of obtaining a region of interest and a given gray threshold Thres 1 The region of interest can be binarized. Obtaining a corresponding contour line 1 in the binarized region of interest by using a boundary tracking method; using a threshold value Thres 2 Repeating the above process to obtain the contour line 2; to meet i Repeating the above process to obtain a contour line I; say, adopt threshold Thres n Repeating the above process to obtain the contour line N. And selecting a plurality of thresholds to obtain a plurality of contour lines (contour line groups), wherein the density of the contour lines is related to the image gradient, the image gradient at the dense part of the contour lines is larger, and the image gradient at the sparse part of the contour lines is smaller.
And (4) taking the center of the region of interest as an end point, sequentially leading R rays outwards in an anticlockwise direction at equal angle intervals from a zero-degree angle, and solving the intersection point of each ray and the contour line group. If the Euclidean distance between two adjacent intersection points on the same ray is less than D min Then the two intersections are said to be connected. For the existence of more than S min And the connected point set is represented by taking the center of the connected point set as a candidate boundary point. There may or may not be multiple candidate boundary points on a ray. The right image in fig. 3 is the candidate boundary points marked in the original region of interest.
(3.2) obtaining a plurality of candidate boundary lines from the candidate boundary points on each ray obtained in the step (3.1), and selecting the optimal candidate boundary line by using a dynamic planning method, namely determining the boundary line of the suspicious lump;
ideally, if there is only one candidate boundary point on each ray in step (3.1), the candidate boundary points on each ray are connected in sequence to form a unique candidate boundary line, i.e. the suspicious lump boundary line. In practical situations, however, a plurality of candidate boundary points may be provided on one ray, one candidate boundary point is selected on each ray each time, the candidate boundary points selected on each ray are sequentially connected to form a candidate boundary line, so that a plurality of candidate boundary lines can be formed, according to the characteristics that the actual boundary line passes through the position where the image gray value changes greatly (namely the image gradient value is large) and has certain smoothness, the method of the invention determines the cost of each candidate boundary line by using a dynamic programming method and setting a cost function, and selects a candidate boundary line with the optimal cost from all candidate boundary lines as the final suspicious lump boundary line. The concrete implementation is as follows:
setting a candidate boundary line S: { n 1 ，n 2 ，…n R }, variable n i Representing the candidate boundary points selected on the ith ray. If the ith ray has no candidate boundary point, carrying out interpolation according to the distance between the candidate boundary point on the (i-1) th ray and the (i + 1) th ray and the center of the region of interest. The cost function C of the candidate boundary line S is the sum of local costs corresponding to all candidate boundary points on the candidate boundary line, that is:
local cost C (n) i ) From an internal cost C int (n i ) And an external cost C ext (n i ) Consists of the following components:
C(n i )＝αC int (n i )+C ext (n i ) (4)
where alpha is a constant value for adjusting the smoothness of the boundary line.
Internal cost C int (n i ) Defined as candidate boundary points n i And n i-1 Normalized distance between：
Wherein dist (n) i ，n i-1 )、dist(O，n i )、dist(O，n i-1 ) Respectively represent candidate boundary points n i And n i-1 Center of interest O and n i Center of interest O and n i-1 The smaller the normalized distance, i.e., the smoother the candidate boundary line, the smaller the cost.
External cost C ext (n i ) Defined as candidate boundary points n i The more connected points represented by a candidate point, the larger the gradient of the candidate point, the more likely the candidate point is to be a real boundary point.
C ext (n i )＝-(n i Number of communication points) (6)
And obtaining the cost of each candidate boundary line according to the cost function, and determining an optimal candidate boundary line in all the candidate boundary lines by adopting dynamic planning as a final suspicious lump boundary line.
(4) The serial related characteristic values of the segmented suspicious mass are calculated, and the selected characteristics can be generally divided into geometric characteristics, morphological characteristics, gray level characteristics, texture characteristics and the like.
The selected eigenvalues should follow several characteristics:
(1) identifiability: the characteristic values of different types of objects are obviously different;
(2) reliability: similar characteristic values are applied to the similar objects;
(3) independence: strong correlation should not exist among the characteristic values;
according to the above rules, 25 features of the suspicious tumor region are extracted, and the detailed description is shown in the table of fig. 4.
(5) And inputting the calculated characteristic values into a classifier, and automatically analyzing the initial suspicious tumor region by a computer to determine the final suspicious tumor region.
Suspicious mass classification is the last stage of automatic detection of breast masses. After extracting the characteristic values reflecting the characteristics of the tumor from the suspicious tumor with the boundary, the classifier is used to determine whether the suspicious tumor is positive (the true tumor area considered by the computer) or false positive. The selection and design of the classifier largely determines the accuracy of the tumor detection. Classification is an important component of pattern recognition theory, and can generally use methods such as linear classification, heuristic rules, statistical classification, fuzzy classification, artificial neural networks, and the like to classify features. The method of the invention adopts an improved k nearest neighbor method to classify the extracted features in the step (4).
The basic rule of the k-nearest neighbor method is: and (4) finding k samples closest to (or most similar to) the feature vector of the test sample from all the samples (except the test sample), voting the samples, and classifying the test sample into the category with the largest sample voting number. The k-nearest neighbor classifier adopted by the method firstly defines a feature vector similarity function and a decision function (DI for short).
(1) Definition of similarity function
Test specimen Y Q Is denoted as V (Y) Q ) Sample X divided by test sample) is denoted as V (X), the similarity function is defined as the inverse of the squared euclidian distance between two feature vectors, i.e.
(2) Definition of decision function
In pair with the test sample Y Q In the decision making process, in principle, the influence of the sample participation decision with the feature vector closer to the feature vector should be larger, and in an original k-nearest neighbor classifier, the influence of the vector distance difference is difficult to be reflected by a simple voting methodAnd (6) defining a row. The following takes decision test samples as the tumor class and the normal class as examples.
Calculating the test sample Y by the first k samples with the similarity arranged from large to small as shown in the formula (8) Q The decision value of (c). In and test sample Y Q Among the first k recent samples, mass was designated as Mass, normal as Norm, and N, sim (Y) were designated as Normal Q X) is the test specimen Y defined in (1) Q Similarity to sample X, rnk (X) represents the order of sample X in the similarity arrangement, X j Mass Represents the jth lump sample, and the value range of j is [1, M']，X l Norm Represents the l normal class sample, and the value range of l is [1, N ]]。
The decision value calculation method considers the classes of samples adjacent to the test sample like a simple voting method, also considers the sequence of the similarity of the samples and the test sample, and experiments prove that the decision function is superior to the calculation method of the original k neighbor decision function.
For the application of the classifier, training data is firstly used to train the classifier to obtain classifier parameters suitable for specific problems, and the classifier training process includes two steps of collecting classifier training data and obtaining classifier parameters, as shown in fig. 5:
(5.1) collecting classifier training data;
firstly, inputting a group of mammary X-ray radiographs with known diagnosis results, and applying the steps of the region-of-interest extraction, the suspicious tumor region segmentation and the feature extraction of the suspicious tumor region to obtain the segmentation results of the suspicious tumor and a series of related feature values, thereby completing the collection of classifier training data.
(5.2) obtaining classifier parameters;
training the designed classifier by using the calculated related characteristic value of the suspicious mass and the actual diagnosis result of the suspicious mass to obtain classification parameters, and writing the classification parameters into a classifier parameter file until the training process of the classifier is finished.
(6) And (3) positioning the final segmentation result of the suspicious tumor region on the X-ray mammary gland photograph to be diagnosed input in the step (1), and displaying the characteristic value of the tumor region calculated in the step (4) to a user according to the requirement.
As shown in fig. 2, the diagnosis assistance system of the present invention includes an input module 100, a region of interest extraction module 200, a suspicious mass segmentation module 300, a suspicious mass region feature extraction module 400, a classification diagnosis module 500, and an output module 600.
The input module 100 is used for receiving the x-ray radiograph of the breast to be diagnosed input by the user and transmitting the x-ray radiograph to the region of interest extraction module 200.
The region of interest extraction module 200 extracts the region of interest in the input mammography, obtains the initial suspicious mass region position information according to the steps described in the step (2), and transmits the position information to the suspicious mass segmentation module 300.
The suspicious tumor segmentation module 300 segments the suspicious tumor in the region of interest extracted by the region of interest extraction module 200 according to the process described in the step (3), so as to obtain boundary information of the suspicious tumor, and transmit the boundary information to the suspicious tumor region feature extraction module 400.
The suspicious mass region feature extraction module 400 calculates a series of region-related feature values, such as geometric features, morphological features, gray-scale features, texture features, etc., according to the received boundary information of the suspicious mass, and transmits the result to the classification diagnosis module 500.
The classification diagnosis module 500 inputs the calculated feature value of each suspicious mass region into a classifier, performs computer automatic classification and identification on the initial suspicious mass region, determines the final suspicious mass region, and transmits the result to the output module 600, the output module 600 positions the segmentation result of the final suspicious mass region automatically detected by the computer on the input x-ray mammogram of the breast to be diagnosed, and displays the calculated region-related feature value to the user as required.
Example (c):
the invention provides a breast cancer computer-aided diagnosis method based on galactophore X-ray radiography and a system thereof, which relate to a plurality of parameters, the parameters are comprehensively adjusted and set aiming at the data characteristics of specific processing so as to achieve the good performance of the whole system, and the parameters set aiming at the data set processing of the invention are listed:
step (2.1), obtaining an initial template size related parameter L =25 by using a two-dimensional hyperbolic secant (sech) function;
step (2.2) selecting threshold value T for binarization processing of correlation degree image low =0.5 × input mammary x-ray radiograph maximum gray level;
the threshold value T selected in the multi-scale analysis of the step (2.3) high =0.6 × input mammography maximum gray level;
and (3.1) sequentially leading R =64 rays outwards along the counterclockwise direction at equal angular intervals from a zero-degree angle by taking the center of the region of interest as an end point, and obtaining the intersection point of each ray and the contour line group. If the Euclidean distance between two adjacent intersection points on the same ray is less than D min And =3, the two intersection points are said to be connected. For the existence of more than S min =10 connected sets of points;
step (3.2) local cost C (n) i ) From an internal cost C int (n i ) And an external cost C ext (n i ) Consists of the following components:
C(n i )＝αC int (n i )+C ext (n i )
where the constant parameter a =110 for adjusting the smoothness of the boundary line.
The method automatically analyzes and processes the suspicious breast lump area in the mammary X-ray radiograph through a mammary X-ray radiograph-based breast cancer computer-aided diagnosis system, provides the lump position and the lump shape, and provides a series of characteristic parameters related to the area according to the requirement, thereby prompting the radiologist to focus on the area needing important attention and the area-related important parameters, and improving the accuracy and the efficiency of the radiologist for breast cancer diagnosis to a certain extent. The implementation of the present invention is not limited to the scope disclosed in the above examples, and the technical solutions described above may be implemented in a manner different from the above examples.
Claims (5)
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007100527471A CN100484474C (en) | 2007-07-13 | 2007-07-13 | Galactophore cancer computer auxiliary diagnosis system based on galactophore X-ray radiography |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2007100527471A CN100484474C (en) | 2007-07-13 | 2007-07-13 | Galactophore cancer computer auxiliary diagnosis system based on galactophore X-ray radiography |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101103924A true CN101103924A (en) | 2008-01-16 |
| CN100484474C CN100484474C (en) | 2009-05-06 |
Family
ID=38998148
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2007100527471A Expired - Fee Related CN100484474C (en) | 2007-07-13 | 2007-07-13 | Galactophore cancer computer auxiliary diagnosis system based on galactophore X-ray radiography |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100484474C (en) |
Cited By (21)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102429679A (en) * | 2011-09-09 | 2012-05-02 | 华南理工大学 | Computer-assisted emphysema analysis system based on chest CT (Computerized Tomography) image |
| CN103345753A (en) * | 2013-07-10 | 2013-10-09 | 深圳先进技术研究院 | Method system for brain image processing |
| CN103425986A (en) * | 2013-08-31 | 2013-12-04 | 西安电子科技大学 | Breast lump image feature extraction method based on edge neighborhood weighing |
| CN104732213A (en) * | 2015-03-23 | 2015-06-24 | 中山大学 | Computer-assisted lump detecting method based on mammary gland magnetic resonance image |
| CN104771228A (en) * | 2015-03-23 | 2015-07-15 | 中山大学 | Method and device for judging whether breast mass is benign or malignant |
| CN104794426A (en) * | 2015-01-13 | 2015-07-22 | 宁夏医科大学 | Method for improving prostate tumor MRI (Magnetic Resonance Imaging) image identification rate based on CAD (Computer-Aided Diagnosis) system |
| CN104915961A (en) * | 2015-06-08 | 2015-09-16 | 北京交通大学 | Lump image region display method and system based on mammary X-ray image |
| CN105354405A (en) * | 2014-08-20 | 2016-02-24 | 中国科学院上海高等研究院 | Machine learning based immunohistochemical image automatic interpretation system |
| CN105374025A (en) * | 2015-09-30 | 2016-03-02 | 上海联影医疗科技有限公司 | Breast image obtaining method and device and mammography system |
| WO2017092615A1 (en) * | 2015-11-30 | 2017-06-08 | 上海联影医疗科技有限公司 | Computer aided diagnosis system and method |
| CN107277528A (en) * | 2017-05-12 | 2017-10-20 | 慧影医疗科技（北京）有限公司 | Big image transmission optimization display methods |
| CN107492099A (en) * | 2017-08-28 | 2017-12-19 | 京东方科技集团股份有限公司 | Medical image analysis method, medical image analysis system and storage medium |
| 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 |
| CN109146864A (en) * | 2018-08-10 | 2019-01-04 | 华侨大学 | The method that galactophore image is split based on the differential evolution algorithm of fuzzy entropy |
| CN110033456A (en) * | 2019-03-07 | 2019-07-19 | 腾讯科技（深圳）有限公司 | A kind of processing method of medical imaging, device, equipment and system |
| CN110349662A (en) * | 2019-05-23 | 2019-10-18 | 复旦大学 | The outliers across image collection that result is accidentally surveyed for filtering pulmonary masses find method and system |
| CN110603567A (en) * | 2017-05-30 | 2019-12-20 | 富士胶片富山化学株式会社 | Dispensing monitoring support device and dispensing monitoring support method |
| CN111281430A (en) * | 2018-12-06 | 2020-06-16 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method, device and readable storage medium |
| WO2020133510A1 (en) * | 2018-12-29 | 2020-07-02 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and device |
| CN111508590A (en) * | 2020-04-03 | 2020-08-07 | 上海理工大学 | Efficient identification detection method for ribs and vertebrae in liver CT perfusion image |
| CN111936989A (en) * | 2018-03-29 | 2020-11-13 | 谷歌有限责任公司 | Similar Medical Image Search |
-
2007
- 2007-07-13 CN CNB2007100527471A patent/CN100484474C/en not_active Expired - Fee Related
Cited By (39)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102429679A (en) * | 2011-09-09 | 2012-05-02 | 华南理工大学 | Computer-assisted emphysema analysis system based on chest CT (Computerized Tomography) image |
| CN103345753A (en) * | 2013-07-10 | 2013-10-09 | 深圳先进技术研究院 | Method system for brain image processing |
| CN103425986A (en) * | 2013-08-31 | 2013-12-04 | 西安电子科技大学 | Breast lump image feature extraction method based on edge neighborhood weighing |
| CN103425986B (en) * | 2013-08-31 | 2016-08-10 | 西安电子科技大学 | Mammary gland tumor image characteristic extracting method based on edge neighborhood weighting |
| CN105354405A (en) * | 2014-08-20 | 2016-02-24 | 中国科学院上海高等研究院 | Machine learning based immunohistochemical image automatic interpretation system |
| CN104794426A (en) * | 2015-01-13 | 2015-07-22 | 宁夏医科大学 | Method for improving prostate tumor MRI (Magnetic Resonance Imaging) image identification rate based on CAD (Computer-Aided Diagnosis) system |
| CN104732213B (en) * | 2015-03-23 | 2018-04-20 | 中山大学 | A kind of area of computer aided Mass detection method based on mammary gland magnetic resonance image |
| CN104732213A (en) * | 2015-03-23 | 2015-06-24 | 中山大学 | Computer-assisted lump detecting method based on mammary gland magnetic resonance image |
| CN104771228A (en) * | 2015-03-23 | 2015-07-15 | 中山大学 | Method and device for judging whether breast mass is benign or malignant |
| CN104771228B (en) * | 2015-03-23 | 2017-04-05 | 中山大学 | A kind of determination methods and device of the good evil of mammary gland tumor |
| CN104915961A (en) * | 2015-06-08 | 2015-09-16 | 北京交通大学 | Lump image region display method and system based on mammary X-ray image |
| US10297024B2 (en) | 2015-09-30 | 2019-05-21 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
| US10636143B2 (en) | 2015-09-30 | 2020-04-28 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
| US11250567B2 (en) | 2015-09-30 | 2022-02-15 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for determining a breast region in a medical image |
| CN105374025A (en) * | 2015-09-30 | 2016-03-02 | 上海联影医疗科技有限公司 | Breast image obtaining method and device and mammography system |
| CN105374025B (en) * | 2015-09-30 | 2018-05-04 | 上海联影医疗科技有限公司 | Breast image acquisition methods and device, mammography system |
| 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 |
| WO2017092615A1 (en) * | 2015-11-30 | 2017-06-08 | 上海联影医疗科技有限公司 | Computer aided diagnosis system and method |
| US10467757B2 (en) | 2015-11-30 | 2019-11-05 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for computer aided diagnosis |
| US10825180B2 (en) | 2015-11-30 | 2020-11-03 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for computer aided diagnosis |
| CN107277528A (en) * | 2017-05-12 | 2017-10-20 | 慧影医疗科技（北京）有限公司 | Big image transmission optimization display methods |
| CN110603567B (en) * | 2017-05-30 | 2023-07-28 | 富士胶片富山化学株式会社 | Dispensing inspection assisting device and dispensing inspection assisting method |
| CN110603567A (en) * | 2017-05-30 | 2019-12-20 | 富士胶片富山化学株式会社 | Dispensing monitoring support device and dispensing monitoring support method |
| CN107492099A (en) * | 2017-08-28 | 2017-12-19 | 京东方科技集团股份有限公司 | Medical image analysis method, medical image analysis system and storage medium |
| CN107492099B (en) * | 2017-08-28 | 2021-08-20 | 京东方科技集团股份有限公司 | Medical image analysis method, medical image analysis system, and storage medium |
| CN111936989A (en) * | 2018-03-29 | 2020-11-13 | 谷歌有限责任公司 | Similar Medical Image Search |
| CN109146864A (en) * | 2018-08-10 | 2019-01-04 | 华侨大学 | The method that galactophore image is split based on the differential evolution algorithm of fuzzy entropy |
| CN111281430B (en) * | 2018-12-06 | 2024-02-23 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method, device and readable storage medium |
| CN111281430A (en) * | 2018-12-06 | 2020-06-16 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method, device and readable storage medium |
| WO2020133510A1 (en) * | 2018-12-29 | 2020-07-02 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and device |
| CN112672691A (en) * | 2018-12-29 | 2021-04-16 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and equipment |
| CN112672691B (en) * | 2018-12-29 | 2024-03-29 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic imaging method and equipment |
| CN110458883A (en) * | 2019-03-07 | 2019-11-15 | 腾讯科技（深圳）有限公司 | A kind of processing system of medical imaging, method, apparatus and equipment |
| CN110033456B (en) * | 2019-03-07 | 2021-07-09 | 腾讯科技（深圳）有限公司 | Medical image processing method, device, equipment and system |
| CN110033456A (en) * | 2019-03-07 | 2019-07-19 | 腾讯科技（深圳）有限公司 | A kind of processing method of medical imaging, device, equipment and system |
| CN110349662B (en) * | 2019-05-23 | 2023-01-13 | 复旦大学 | Cross-image set outlier sample discovery method and system for filtering lung mass misdetection results |
| CN110349662A (en) * | 2019-05-23 | 2019-10-18 | 复旦大学 | The outliers across image collection that result is accidentally surveyed for filtering pulmonary masses find method and system |
| CN111508590B (en) * | 2020-04-03 | 2023-03-31 | 上海理工大学 | Efficient identification detection method for ribs and vertebrae in liver CT perfusion image |
| CN111508590A (en) * | 2020-04-03 | 2020-08-07 | 上海理工大学 | Efficient identification detection method for ribs and vertebrae in liver CT perfusion image |
Also Published As
| Publication number | Publication date |
|---|---|
| CN100484474C (en) | 2009-05-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN100484474C (en) | Galactophore cancer computer auxiliary diagnosis system based on galactophore X-ray radiography | |
| Sechopoulos et al. | Artificial intelligence for breast cancer detection in mammography and digital breast tomosynthesis: State of the art | |
| CN109583440B (en) | Medical image auxiliary diagnosis method and system combining image recognition and report editing | |
| EP0757544B1 (en) | Computerized detection of masses and parenchymal distortions | |
| CN108898595B (en) | Construction method and application of positioning model of focus region in chest image | |
| Giger | Computer-aided diagnosis of breast lesions in medical images | |
| US7308126B2 (en) | Use of computer-aided detection system outputs in clinical practice | |
| CN108416360B (en) | Cancer diagnosis system and method based on breast molybdenum target calcification features | |
| US7903861B2 (en) | Method for classifying breast tissue density using computed image features | |
| US6760468B1 (en) | Method and system for the detection of lung nodule in radiological images using digital image processing and artificial neural network | |
| US20050207630A1 (en) | Lung nodule detection and classification | |
| Lee et al. | A computer-aided design mammography screening system for detection and classification of microcalcifications | |
| Elizabeth et al. | Computer-aided diagnosis of lung cancer based on analysis of the significant slice of chest computed tomography image | |
| US20080002870A1 (en) | Automatic detection and monitoring of nodules and shaped targets in image data | |
| Wu et al. | Coarse-to-fine lung nodule segmentation in CT images with image enhancement and dual-branch network | |
| US8285013B2 (en) | Method and apparatus for detecting abnormal patterns within diagnosis target image utilizing the past positions of abnormal patterns | |
| US20120257804A1 (en) | Communicative cad system for assisting breast imaging diagnosis | |
| CN113269225A (en) | Non-invasive liver epithelium sample vascular smooth muscle lipoma image classification device based on image omics | |
| Wang et al. | A method of ultrasonic image recognition for thyroid papillary carcinoma based on deep convolution neural network. | |
| Thakur et al. | A systematic review of machine and deep learning techniques for the identification and classification of breast cancer through medical image modalities | |
| JP2024507290A (en) | Method and system for breast ultrasonic image diagnosis using weakly-supervised deep learning artificial intelligence ial intelligence} | |
| Fan et al. | Research on abnormal target detection method in chest radiograph based on YOLO v5 algorithm | |
| Saranya et al. | A dense kernel point convolutional neural network for chronic liver disease classification with hybrid chaotic slime mould and giant trevally optimizer | |
| Saltanat et al. | An efficient pixel value based mapping scheme to delineate pectoral muscle from mammograms | |
| CN115375632A (en) | Lung nodule intelligent detection system and method based on CenterNet model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| C17 | Cessation of patent right | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20090506 Termination date: 20120713 |