CN101373479A - Method and system for searching computer picture of mammary gland x-ray radiography - Google Patents

Method and system for searching computer picture of mammary gland x-ray radiography Download PDF

Info

Publication number
CN101373479A
CN101373479A CNA2008101970936A CN200810197093A CN101373479A CN 101373479 A CN101373479 A CN 101373479A CN A2008101970936 A CNA2008101970936 A CN A2008101970936A CN 200810197093 A CN200810197093 A CN 200810197093A CN 101373479 A CN101373479 A CN 101373479A
Authority
CN
China
Prior art keywords
interest
area
image
picture
feature
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.)
Pending
Application number
CNA2008101970936A
Other languages
Chinese (zh)
Inventor
宋恩民
姜娈
金人超
许向阳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Huazhong University of Science and Technology
Original Assignee
Huazhong University of Science and Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Huazhong University of Science and Technology filed Critical Huazhong University of Science and Technology
Priority to CNA2008101970936A priority Critical patent/CN101373479A/en
Publication of CN101373479A publication Critical patent/CN101373479A/en
Pending legal-status Critical Current

Links

Images

Landscapes

  • Apparatus For Radiation Diagnosis (AREA)

Abstract

The invention discloses a method for retrieving computerized mammary gland X-ray radiography images and a system thereof. Firstly, a computerized mammary gland X-ray radiography image is input to a system; finally, single view and/or multiple view features which a user pays attention to and selects and weights relating to the importance of input features are transmitted to a search module based on feature similarity measure rules by processing a module for abstracting interesting zones to be inquired, an interesting organization partition module and an interesting organization feature abstracting module, thus getting a plurality of interesting zones with known diagnostic results which are retrieved from the established reference data of the computerized mammary gland X-ray interesting zone with known diagnostic results and needed by users as well as are similar to the interesting zones inquired; finally, the retrieval results are displayed to users in the descending order of similarity.

Description

A kind of searching computer picture of mammary gland x-ray radiography method and system
Technical field
The invention belongs to the Computer Analysis and the application technology of medical image, be specifically related to a kind of searching computer picture of mammary gland x-ray radiography method and system thereof.
Background technology
In clinical practice, radiologist must read the sheet experience and improves the diagnostic level of self by constantly saving bit by bit.Each case diagnosis process can think that the radiologist is at first according to " knowledge " (the existing sheet experience of readding) that is stored in the brain, obtain waiting to diagnose the tentative diagnosis of case, and then, obtain waiting the diagnostic result of diagnosing case final according to the pathological biopsy inspection in the suspicious lesion tissue of tentative diagnosis zone.Yet pathologic finding still all gives sufferer very big burden on health mentally, so the quantity of the accuracy of raising radiologist tentative diagnosis and then minimizing pathologic finding is very important.Because it is less that young radiologist reads the sheet experience, " knowledge " amount of storing in the brain is limited, if can be in the process of its training and learning and self study, provide some similar to it and made a definite diagnosis case images for some difficult case of distinguishing diagnosis, can help young radiologist to increase " knowledge " amount, make them make tentative diagnosis more accurately in subsequently clinical, this radiologist to youth is a very significant thing to patient still.This process of similar image that provides belongs to the searching computer field.
Searching computer can be divided into two kinds of text-based image retrieval and CBIRs.Traditional text-based image retrieval method is based on the retrieval of key word, for example image name, pictograph description etc., but use text description because color, texture, shape, spatial relationship and the semantic information of image entities are very difficult, so text-based image retrieval is just showed a lot of weak points.CBIR is to extract for example features such as gray scale, texture, shape from image itself, directly replaces text message to remove to retrieve similar image with these features.Provide similar and known diagnosis as a result the process of case image can use the CBIR method to realize, but because the own characteristic of object medical image to be retrieved, as: gray level resolution height, image similarity contain much information greatly, contained, color type is few etc., the difficulty that makes method realize is bigger.Therefore, content-based medical image search method not only need carry out conventional image retrieval under the guiding of the medical knowledge base of having set up, promptly use general iconology feature to reach visual similar, but also the relevant characteristics of image of the pathology that needs some and medical science to combine and reach similar on the pathology.
Studies show that, picture of mammary gland x-ray radiography (breast molybdenum target x ray technology) is a kind of inspection method of effectively carrying out early stage clinical asymptomatic breast cancer, by FDA (Food and Drug Adminstration) (Food and DrugAdministrator, FDA) Ren Ke conventional breast disease screening method.But limited by image-forming principle, normal galactophore tissue at the picture of mammary gland x-ray radiography middle-high density also can present the high brightness close with abnormal structure, young Asia women its focus fine and close for breast tissue then is difficult for being found more, cause easily and fail to pinpoint a disease in diagnosis and mistaken diagnosis, and the early diagnosis of breast cancer depends on radiologist's experience, professional ability, degree of fatigue to a great extent, the subjective factor influence is bigger, especially in the generaI investigation activity of mammary gland disease, be difficult to guarantee efficient and the accuracy diagnosed.In this case, training and learning and the self study of effectively strengthening young radiologist is very important.
Therefore, how to design one and from known diagnosis result's computer picture of mammary gland x-ray radiography image reference database, accurately and effectively retrieve and inquire about the method for case similar image and the focus that system will become our research thereof.
Summary of the invention
The object of the present invention is to provide a kind of searching computer picture of mammary gland x-ray radiography method, some computer picture of mammary gland x-ray radiography images of computer picture of mammary gland x-ray radiography image similarity can be returned and inquire about to this method for the user, and have higher recall ratio and precision ratio; The present invention further provides the searching computer picture of mammary gland x-ray radiography system that realizes this method.
The search method of computer picture of mammary gland x-ray radiography image provided by the invention, its step comprises:
Step (1) input one width of cloth computer picture of mammary gland x-ray radiography image;
Step (2) user-interactive ground encloses the zone of its concern of drawing on image, obtain area-of-interest to be checked;
Step (3) at first computing machine is cut apart area-of-interest to be checked, determines the border of tissue of interest;
Step (4) is based on the segmentation result to tissue of interest, and computing machine extracts the relevant single-view of several regions and the iconology feature of many views, and this iconology feature comprises geometric properties, morphological feature, gray feature and textural characteristics; The user selects single-view and/or many view feature of its concern, and the relevant weight of input feature vector importance;
Step (5) is utilized the characteristic similarity measurement criterion of weighting, retrieves the some area-of-interests similar to area-of-interest to be checked from the picture of mammary gland x-ray radiography area-of-interest computer picture reference database of setting up the known diagnosis result; Comprise in the reference database: region of interest area image, corresponding pathologic finding of area-of-interest and diagnostic result message file, and the relevant iconology representative record file of area-of-interest of computing machine extraction;
Step (6) is shown to the user with area-of-interest and regional corresponding pathologic finding and the diagnostic result information that retrieval obtains.
The searching system of computer picture of mammary gland x-ray radiography image provided by the invention is characterized in that: it comprise load module, area-of-interest extraction module to be checked, tissue of interest cut apart module, tissue of interest characteristic extracting module, based on the retrieval module and the output module of characteristic similarity measurement criterion;
Load module is used to receive the computer picture of mammary gland x-ray radiography image of input, and sends area-of-interest extraction module to be checked to;
Area-of-interest extraction module to be checked is used for carrying out alternately with the user, the area-of-interest that circle draws on image, and, extract the area-of-interest to be checked in the picture of mammary gland x-ray radiography of importing according to the described process of above-mentioned steps (2), send tissue of interest to and cut apart module;
Tissue of interest is cut apart module according to the described process of above-mentioned steps (3), tissue of interest in the area-of-interest of area-of-interest extraction module extraction to be checked is split, obtain the boundary information of tissue of interest, send the tissue of interest characteristic extracting module to;
The tissue of interest characteristic extracting module is according to the boundary information of the tissue of interest that receives, according to the described process of above-mentioned steps (4), calculate the eigenwert that comprises geometric properties, morphological feature, gray feature and textural characteristics of single-view and many views; And carry out alternately with the user, select its single-view of paying close attention to and/or many view feature and the relevant weight of input feature vector importance for the user, and send information to retrieval module based on the characteristic similarity measurement criterion;
Based on the retrieval module of characteristic similarity measurement criterion according to the described process of above-mentioned steps (5), utilize the characteristic similarity measurement criterion of weighting, in the known diagnosis result's who has set up picture of mammary gland x-ray radiography area-of-interest computer picture reference database, retrieve the some known diagnosis results' similar area-of-interest to inquiring about area-of-interest, and the area-of-interest of k similarity value correspondence sends output module to as result for retrieval before selecting according to user's needs, k is the number of the similar image of user's concern, by output module area-of-interest and regional corresponding pathologic finding and the diagnostic result information that retrieval obtains is shown to the user.
The present invention is by input one width of cloth computer picture of mammary gland x-ray radiography image, and the zone of being enclosed its concern of drawing by user-interactive ground on image obtains area-of-interest to be checked; Computing machine is cut apart area-of-interest to be checked and is extracted the relevant iconology feature of several regions then, method provides and imports the iconology feature of the equal angular picture of mammary gland x-ray radiography image of the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue of the corresponding sufferer of computer picture of mammary gland x-ray radiography image and heteropleural breast tissue simultaneously, and the user can select single-view and/or the many view feature and the relevant weight of input feature vector importance of its concern in view of the above; Last method is utilized the characteristic similarity measurement criterion of weighting, and (proper vector is similar more, then image is similar more), (comprise the reference database: the region of interest area image from the picture of mammary gland x-ray radiography area-of-interest computer picture reference database of setting up the known diagnosis result, the corresponding sufferer essential information of area-of-interest file, corresponding pathologic finding of area-of-interest and diagnostic result message file, the iconology representative record file that the area-of-interest that computing machine extracts is relevant) retrieves the some area-of-interests similar in, and area-of-interest and regional corresponding pathologic finding and the diagnostic result information that retrieval obtains is shown to the user to area-of-interest to be checked.In a word, the inventive method accurately and is effectively retrieved the region of interest area image that obtains enclosing to the user the similar some known diagnosis results of the region of interest to be checked of picture from the picture of mammary gland x-ray radiography area-of-interest computer picture reference database of setting up the known diagnosis result by a searching computer picture of mammary gland x-ray radiography method and system thereof.
Description of drawings
Fig. 1 is the process flow diagram of searching computer picture of mammary gland x-ray radiography method of the present invention;
Fig. 2 is the structural representation of searching computer picture of mammary gland x-ray radiography of the present invention system;
Fig. 3 is the process synoptic diagram that the area-of-interest background trend is removed in the embodiment of the invention;
Fig. 4 suppresses result schematic diagram for embodiment of the invention adjacent tissue;
Fig. 5 is an embodiment of the invention image segmentation process synoptic diagram;
Fig. 6 is the feature list that the embodiment of the invention extracts tissue of interest;
Fig. 7 is for searching many view area of mating on the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue synoptic diagram to G-G ' in the embodiment of the invention;
Fig. 8 is for searching many view area of mating on the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue synoptic diagram to R-R ' in the embodiment of the invention;
Fig. 9 is the application legend of searching computer picture of mammary gland x-ray radiography.
Embodiment
The present invention is further detailed explanation below in conjunction with accompanying drawing and example.
As shown in Figure 1, the inventive method may further comprise the steps:
(1) input one width of cloth computer picture of mammary gland x-ray radiography image.
(2) obtain area-of-interest to be checked by the draw zone of its concern of user-interactive ground circle on image.
(2.1) on image, enclose the zone of its concern of drawing by user-interactive ground, and calculate the center in circle picture zone;
(2.2) center with circle picture zone is the center of circle, do a circle with circle picture zone homalographic, calculate the intensity-weighted center of gravity of this circle, the center of circle is moved to this center of gravity place, recomputate and enclose the intensity-weighted center of gravity of picture zone homalographic circle, after repeating these process several times (by user's input), the intensity-weighted center of gravity calculated to the end.Why do like this rather than directly adopt the user draw a circle to approve the zone the center, be because there is deviation in the user to twice delineation of priority in same zone, the process of such iteration can reduce this deviation to a certain extent, increases the stability and the consistance of system;
(2.3) center of gravity so that iteration stops in (2.2) step is the center, on the computer picture of mammary gland x-ray radiography image of input, get one with database in the big or small identical square window image (wide) of area-of-interest as 125 * 125 pixels as area-of-interest to be checked.
(3) computing machine is cut apart area-of-interest to be checked, determines the border of tissue of interest (as suspicious lesion tissue).The inventive method adopts a kind of interactively tissue of interest dividing method, method has made full use of the size that the user encloses the picture zone, be the information of related organization's size, making the expectation that segmentation result more is close to the users has increased the stability of system and user's degree of belief.Be implemented as follows:
(3.1) background trend is removed;
Tissue of interest generally can be surrounded by fine and close gland tissue, and some also can be positioned near skin line or the wall of the chest line, and these are all given to cut apart and bring certain difficulty.In order to reduce the influence that background normal structure is on every side cut apart tissue of interest, strengthen the contrast of tissue of interest, the inventive method at first adopts the least square fitting method to simulate big background planes (i.e. " background trend " plane) such as and former area-of-interest, try to achieve the error image of original area-of-interest and background plane, and error image is demarcated (difference of original area-of-interest and background plane is changed to 0 less than 0 pixel with its gray-scale value; Otherwise, its gray-scale value is changed to difference multiply by 255) and obtain the area-of-interest that background trend is removed.
Fig. 3 (a) is that a width of cloth is positioned near the original area-of-interest of skin line, and Fig. 3 (b) is the background plane that Fig. 3 (a) comes out through least square fitting, and Fig. 3 (c) is the area-of-interest that Fig. 3 (a) removes through background trend.
(3.2) adjacent tissue suppresses
In order further to reduce the influence that contiguous normal gland tissue is cut apart tissue of interest, the present invention uses " adjacent tissue's inhibition " on the region of interest area image that background trend is removed.
Enclose the area in picture zone according to the user, roughly estimate the radius of this tissue of interest, the pixel that is in the radius keeps its original gray-scale value, and being in the outer grey scale pixel value of radius will suppress, far away more from region of interest centers, the inhibition amplitude is big more.Concrete computing formula is shown in (1):
L(x，y)＝R(x，y，σ，r)·I(x，y) (1)
The image that obtains after representing to suppress of L wherein through adjacent tissue, the region of interest area image of I for removing through background trend, R is an inhibition function, its computing formula is shown in (2):
R ( x , y , σ , r ) = e ( r - x 2 + y 2 ) 2 2 σ 2 ( x 2 + y 2 ) > r 1 else - - - ( 2 )
Wherein r represents to enclose with the user radius of the border circular areas of picture zone homalographic.σ is a constant, is used to regulate the inhibition degree, if excessively suppress, may make tissue of interest " less divided " (the tissue of interest mistake is divided into normal structure); If the inhibition degree is not enough, then do not reach the effect of inhibition, make interest groups tissue region " over-segmentation " (the normal structure mistake is divided into the interest groups tissue region).Thereby need suitable adjustment inhibiting factor, make that the inhibition amplitude is suitable.The result that Fig. 3 (c) obtains after suppressing through adjacent tissue as shown in Figure 4.
" adjacent tissue's inhibition " method has made full use of the information of the relevant tissue of interest size that the user provides.This is embodied in two aspects: at first, " adjacent tissue's inhibition " do not do any inhibition enclosing in the border circular areas of picture zone homalographic with the user, thereby avoided " less divided " effectively; Secondly, inhibiting factor σ is directly proportional with radius r, and tissue of interest is more little, and is big more to the inhibition degree of adjacent tissue, thereby helps avoiding " over-segmentation ".
(3.3) based on landform dividing method segmented sense region-of-interest;
The inventive method adopts multilayer landform dividing method to determine the border of tissue of interest, the basic thought of multilayer landform dividing method is to use a plurality of different threshold values to carry out region growing and obtain from inside to outside multilayer cut zone, with outermost cut zone as final segmentation result, wherein method is specifically to determine number of plies t according to the number of selected threshold value, and t is a positive integer.
The multilayer landform is cut apart the partitioning algorithm that is based on region growing.Why adopt the algorithm based on region growing to mainly contain following reason: at first, tissue of interest to be split is positioned at picture centre, thus can be simply with region of interest centers as seed points; Secondly, the area-of-interest picture structure is fairly simple, is fit to use the region growing algorithm.At last, very fast based on the algorithm speed of region growing, can shorten system response time.
To be specifically described this dividing method with three layers of landform dividing method in the present embodiment.The implementation step of three layers of landform dividing method is as follows:
Step0: initially the central pixel point with area-of-interest adds S set (S is initially sky) as seed points, and sets initial current threshold value, and this threshold value is relevant with the seed points gray-scale value, is one of percentage of seed points gray-scale value as value, and putting i is 0;
Step1: with certain some adjacency among all and the S and with the absolute value of the gray value differences of seed points S exterior pixel point adding S less than current threshold value;
Step2: repetitive process step1, when not having new pixel to add S, i adds 1;
Step3: if generated three layers of landform (being that i equals 3), S set is final segmentation result, finishes; Otherwise, according to cut zone and on every side the contrast of background area calculate a new threshold value as current threshold value, jump toward step1;
The concrete computing method of three threshold values as shown in Equation (3) in the present embodiment
T i = I seed 100 i = 1 T i - 1 + C i - 1 × 255 3 i = 2,3 - - - ( 3 )
Wherein, I SeedBe the gray-scale value of seed points, C I-1It is i-1 layer segmentation result and the contrast of background area on every side.C I-1Formula is as (4),
C i - 1 = I R i - 1 ‾ - I B i - 1 ‾ I R i - 1 ‾ + I B i - 1 ‾ - - - ( 4 )
Wherein,
Figure A200810197093D00133
With Be respectively i-1 layer segmentation area and reach the average gray value of background area on every side.In the inventive method " background area " is defined as: enclose remaining area after removing current segmentation result on the relevant border circular areas of the tissue of interest size information of picture with the user at one.Fig. 5 (a)-Fig. 5 (c) is be added to result on Fig. 4 of three layers of segmentation result, and Fig. 5 (d) is be added to result on the former figure of the profile with segmentation result.
(4) based in the step (3) to the segmentation result of tissue of interest, computing machine extracts the relevant single-view iconology feature of several regions.According to corresponding several views of input computer picture of mammary gland x-ray radiography image (being the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue of sufferer and the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue), the inventive method provides the right iconology feature of many view area with area-of-interest coupling to be retrieved simultaneously.The user can select single-view and/or the many view feature and the relevant weight of input feature vector importance of its concern in view of the above.
The purpose that the image correlated characteristic extracts is certain expression of extracting picture materials by various image analysis technologies, makes this expression of image can be as the foundation of image retrieval.Reasonably select characteristics of image, can improve the speed and the effect of retrieval effectively.When being retrieved, the picture of mammary gland x-ray radiography image should merely not consider the similarity on the vision meaning, what also should pay attention to is the similarity of pathological characteristics on the medical significance, if the feature of extracting makes that the picture number identical with the actual pathology kind of query image is many more in the result for retrieval image, then feature extraction this time is effective more.
The iconology feature of selecting for use generally can be divided into geometric properties, morphological feature, gray feature and textural characteristics etc.The eigenwert of selecting should be followed following characteristics:
1. identifiability: the eigenwert of inhomogeneity object has notable difference;
2. reliability: the eigenwert of homogeneous object applications similar;
3. independence: strong correlation does not have between the eigenwert;
(4.1) according to above rule, the inventive method at first to area-of-interest to be checked extract 32 single-view associated pictures learn features as the Partial Feature vector with pending image retrieval, shown in tabulation among Fig. 6.
Bin Zheng etc. is wherein arranged (specifically referring to Bin Zheng, Amy Lu, Lara A.Hardesty, et al.A method to improve visual similarity of breast masses for an interactivecomputer-aided diagnosis environment.Medical Physics, 2006,33 (1): 111-117) 14 features of Ti Chuing (feature 1-feature 14 in the table), Nicholas Petrick etc. are (specifically referring to Nicholas Petrick, Heang-Ping Chan, Datong Wei, et al.Automated detection ofbreast masses on mammograms using adaptive contrast enhancement andtexture classification.Med.Phys, 1996,23 (10): 1685-1695) 10 features of Ti Chuing (feature 15-feature 24 in the table).In addition, the inventive method has proposed 8 new features, feature 25-feature 32 in the table, and its concrete computing method are as follows:
Feature 25: cut apart the center skew and the maximum radius length ratio F of cutting apart tissue of interest of tissue of interest 25, formula is as (5),
F 25 = d S Max i ( r i ) - - - ( 5 )
D wherein SBe the distance between the pixel that has maximum gradation value in regional center and the zone.Radius length r iBe defined as the distance of pixel i on regional center and the profile, Max i(r i) be radius length r iMaximal value.Because the normal tissue regions of the grey scale change in true focus zone is more violent, so under the identical situation of maximum radius length, the center skew in true lesion tissue zone is more likely bigger than normal.So this eigenwert in true lesion tissue zone should be bigger than normal.
Feature 26: the gradient F of having cut apart gray-scale value in the tissue of interest 26, formula is as (6),
F 26 = R ( R - 1 ) ( R - 2 ) Σ i = 1 R ( I i - I R ‾ σ ) 3 - - - ( 6 )
Wherein, R is that regional interior pixel is counted, and σ is the standard deviation of gray-scale value in the zone, I iRepresent the gray-scale value of the pixel i in the zone,
Figure A200810197093D00143
Represent the average gray value of regional interior pixel.Relative normal tissue regions, the gradient in true focus zone is bigger than normal.
Feature 27: the average sharpness value F of all pixels on the segmented sense interest group driving wheel exterior feature 27, acutance is utilizes the sobel operator to calculate the Grad of gained to original area-of-interest, and this feature is used to measure the contrast of tissue of interest at boundary.Often there is border more clearly in true focus zone, so the average acutance of its profile is bigger than normal.
Feature 28: the average gradient F of all pixels on the segmented sense interest group driving wheel exterior feature 28, gradient is on current pixel and the directions of rays along regional center to current pixel apart from the gray value differences of the pixel at 10 pixel distance places of current pixel, and this feature also is used to describe the contrast of boundary.
Feature 29: the mean radius length F of all pixels on the segmented sense interest group driving wheel exterior feature 29, formula is as (7),
F 29 = 1 n B Σ i = 1 n B r i - - - ( 7 )
Wherein, radius length r iBe defined as the distance of pixel i on regional center and the profile, n BBe the number of pixels on the profile.Statistical result showed, the mean radius length in true focus zone is bigger than normal.
Feature 30: the mean radius length ratio F of pixel on the standardization radius length standard deviation of pixel and the segmented sense interest group driving wheel exterior feature on the segmented sense interest group driving wheel exterior feature 30, the ratio of feature 21 and feature 29.Formula is as (8),
F 30 = F 21 F 29 = 1 n B Σ i = 1 n B ( r i ′ - r i ′ ‾ ) 2 1 n B Σ i = 1 n B r i - - - ( 8 )
Wherein, standardization radius length
Figure A200810197093D00153
Be defined as radius length r iWith maximum radius length M ax i(r i) the ratio,
Figure A200810197093D00154
Be the standardization radius length
Figure A200810197093D00155
Mean value.True focus zone is round or ellipse mostly, so its standardization radius length standard deviation is often less than normal, so this eigenwert of true focus zone is less than normal.
Feature 31: the ratio F of having cut apart all number of pixels in local maximum pixel number and the zone in the tissue of interest 31If some gray values of pixel points are to be the center with this pixel, the maximal value in the window of 5 * 5 sizes in all pixel gray-scale values, then this pixel is a local maximum pixel.True focus zone is brighter, the darker even form on every side in center often, thus in the window of one 5 * 5 size central pixel point to have the possibility of maximum gradation value lower, so regional this eigenwert of true focus is less than normal.
Feature 32: cut apart local maximum pixel average height F in the tissue of interest 32It is the center with local maximum pixel therewith that the height of a local maximum pixel is local maximum pixel gray-scale value, the minimum gradation value in the window of 5 * 5 sizes poor.This eigenwert of true focus zone is less than normal.
After 32 feature extractions are finished, again feature is carried out normalized one by one, normalized method as shown in Equation (9), the span after each feature normalization is within [0,1].
f ij ′ = f ij - min i ( f ij ) max i ( f ij ) - min i ( f ij ) - - - ( 9 )
(4.2) picture of mammary gland x-ray radiography has its limitation as a kind of effective generaI investigation means, because being the breast tissue with a three-dimensional, its image-forming principle is mapped as a two-dimensional plane image, so overlapping normal structure is easy to a large amount of small cancer focuses is covered, thereby causes failing to pinpoint a disease in diagnosis and mistaken diagnosis of a lot of early-stage cancers.Sufferer will be taken several picture of mammary gland x-ray radiography images (being the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue and the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue) usually in clinical, and the information that the radiologist understands in the comprehensive multiple image improves the accuracy of tentative diagnosis.Therefore, the inventive method provides with the right iconology feature of many view area of area-of-interest to be retrieved coupling as the Partial Feature vector simultaneously with pending image retrieval.
With many view area of area-of-interest to be retrieved coupling to mainly comprising: many view area of mating on the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue to the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue on many view area of mating right.As shown in Figure 7, many view area of mating on the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue are in separately same position on the image to G-G ', as: regional G and the regional G ' distance that projects to nipple on the axis of image (nipple is to the vertical line of wall of the chest line) separately equates, is d.As shown in Figure 8, many view area of mating on the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue are to R-R ', because the difference of shooting angle, region R and region R ' generally be not in separately same position on the image.The inventive method according to the region R center on axis (nipple is to the vertical line of wall of the chest line) project to nipple apart from d, finding apart from teat spacing on the axis on the other width of cloth picture of mammary gland x-ray radiography image is the position of d, with this position is that a matching area band vertical with the axis is set at the center, and in the matching area band search brightness the highest zone, set this zone and be matching area R ', many view area of determining thus to mate on the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue are to R-R '.
Many view area according to the coupling that finds are right, geometric properties, morphological feature, gray feature and the textural characteristics etc. of many view area that the inventive method is mated according to the Rule Extraction of feature extraction to being correlated with, as:
1. effective size of focus: effective size in focus zone equals the square root of maximum radial distance and minimum radial distance product between its borderline point, and two focuses zones that the right effective size of focus of matching area equals to mate are the absolute value of the difference of size effectively.In addition, the maximum radial distance between its borderline point of focus zone also will be employed;
2. distance feature: the focus regional center on axis (nipple is to the vertical line of wall of the chest line) projection and the distance of nipple, the absolute value of the difference of two focus zone distance feature that the right distance feature of matching area equals to mate, it is right that eigen is primarily aimed at many view area of mating on the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue.
3. comprehensive density feature: the comprehensive density feature in focus zone is as a kind of quantization method of focus, discover that this feature and projection visual angle and metaplasia are irrelevant, has high " view unchangeability " (Y.H.Chang, W.F.Good, J.K.Leader, X.H.Wang, B.Zheng, L.A.Hardesty, C.M.Hakim, and D.Gur, " Integrated density of a lesion:Aquantitative, mammographically derived, invariable measure; " Med.Phys.30 (7), 1805-1811 (2003) .).Comprehensive density feature SID specifically is defined as:
SID = ( AVE R - AVE R 1 ) * AR 3
Wherein R is the interest groups tissue region that is partitioned into, and R1 is the background tissues zone, AVE RBe the average gray of the interest groups tissue region that is partitioned into, AVE R1Be the average gray in background tissues zone, AR is the area of the interest groups tissue region that is partitioned into.
4. based on information-theoretical similarity measurement feature: the inventive method will extract some based on information-theoretical similarity measurement feature as, entropy, combination entropy, conditional entropy and (standardization) mutual information etc., they can measure effectively the focus zone to similarity degree.
(4.3) single-view and the many views correlated characteristic that proposes according to step (4.1) and (4.2), single-view and/or many view feature and the relevant weight of input feature vector importance that the inventive method provides the user to select its concern by User Interface.
(5) according to the feature of user's selection and the feature weight of input, (proper vector is similar more to utilize the characteristic similarity measurement criterion of weighting, then image is similar more), (comprise the reference database: the region of interest area image from the picture of mammary gland x-ray radiography area-of-interest computer picture reference database of setting up the known diagnosis result, the corresponding sufferer essential information of area-of-interest file, the iconology representative record file that the area-of-interest that corresponding pathologic finding of area-of-interest and diagnostic result message file, computing machine extract is relevant) retrieves the some known diagnosis results' similar area-of-interest in to area-of-interest to be checked.
(5.1) for the realization of search method and system, the present invention has set up a known diagnosis result's picture of mammary gland x-ray radiography area-of-interest computer picture reference database.Comprise in the reference database: region of interest area image, the corresponding sufferer essential information of area-of-interest file, corresponding pathologic finding of area-of-interest and diagnostic result message file, the iconology representative record file that the area-of-interest that computing machine extracts is relevant.
Wherein, the extracting method of area-of-interest is consistent with the extracting method in the step (2), gets a square window image (wide as 125 * 125 pixels) on the sufferer picture of mammary gland x-ray radiography image diagnosing; The corresponding sufferer essential information of area-of-interest, corresponding pathologic finding of area-of-interest and diagnostic result information are provided by professional radiologist; To the area-of-interest in the reference database carry out with step (3) in consistent tissue of interest cut apart and and the middle consistent single-view iconology feature extraction relevant of step (4) with many views can obtain the iconology representative record that the area-of-interest in the reference database is correlated with.
(5.2) according to the feature of user's selection and the feature weight of input, (proper vector is similar more to utilize the characteristic similarity measurement criterion of weighting, then image is similar more), in the known diagnosis result's who has set up picture of mammary gland x-ray radiography area-of-interest computer picture reference database, retrieve the some known diagnosis results' similar area-of-interest to inquiring about area-of-interest.The present invention is defined as follows the similarity criterion:
Area-of-interest Y in known diagnosis result's the picture of mammary gland x-ray radiography area-of-interest computer picture reference database Q, its proper vector is designated as V (Y Q) (from the representative record file, obtaining), area-of-interest X to be checked, its proper vector is designated as V (X), and the weight vector of proper vector (user's setting) is designated as w, then define the similarity criterion and be the inverse of two Euclidean distances between the weighted feature vector square, promptly
Sim ( Y Q , X ) = 1 | | w ( V ( Y Q ) - V ( X ) ) | | 2 - - - ( 10 )
According to the similarity criterion, each area-of-interest in area-of-interest to be checked and known diagnosis result's the picture of mammary gland x-ray radiography area-of-interest computer picture reference database all can calculate a similarity value, all similarity values that obtain are carried out descending ordering, and select the area-of-interest of preceding k (the user-interactive interface is selected to finish) similarity value correspondence as result for retrieval according to user's needs.
(6) the corresponding pathologic finding and the diagnostic result information in area-of-interest that retrieval is obtained and zone are shown to the user.
As shown in Figure 2, assistant diagnosis system of the present invention comprise load module 100, area-of-interest extraction module 200 to be checked, tissue of interest cut apart module 300, tissue of interest characteristic extracting module 400, based on the retrieval module 500 and the output module 600 of characteristic similarity measurement criterion.
Load module 100 is used to receive the computer picture of mammary gland x-ray radiography image of input, and sends area-of-interest extraction module 200 to be checked to.
Area-of-interest extraction module 200 to be checked, sends tissue of interest to and cuts apart module 300 according to the area-of-interest to be checked in the picture of mammary gland x-ray radiography of the extracted region input of user-interactive ground its concern that circle draws on image according to the described step of above-mentioned steps (2).
Tissue of interest is cut apart module 300 according to the described process of above-mentioned steps (3), tissue of interest in the area-of-interest of area-of-interest extraction module 200 extractions to be checked is split, obtain the boundary information of tissue of interest, send tissue of interest characteristic extracting module 400 to.
Tissue of interest characteristic extracting module 400 is according to the boundary information of the tissue of interest that receives, according to the described process of above-mentioned steps (4), calculate the relevant single-view in a series of zones and many view feature, as geometric properties, morphological feature, gray feature and textural characteristics etc., the user selects the relevant weight of the single-view of its concern and/or many view feature and input feature vector importance and sends retrieval module 500 based on the characteristic similarity measurement criterion to.
Based on the retrieval module 500 of characteristic similarity measurement criterion according to the described process of above-mentioned steps (5), (proper vector is similar more to utilize the characteristic similarity measurement criterion of weighting, then image is similar more), in the known diagnosis result's who has set up picture of mammary gland x-ray radiography area-of-interest computer picture reference database, retrieve the some known diagnosis results' similar area-of-interest to inquiring about area-of-interest, and the area-of-interest of k similarity value correspondence sends output module 600 to as result for retrieval before selecting according to user's needs, by output module 600 area-of-interest and regional corresponding pathologic finding and the diagnostic result information that retrieval obtains is shown to the user.K is imported by interactive interface by the user for the number of the similar image of user's concern.
Example:
Relate to several parameters in a kind of searching computer picture of mammary gland x-ray radiography method that the present invention proposes and the system thereof, these parameters will be carried out the comprehensive adjustment setting at the data characteristics of concrete processing to reach the superperformance of total system, list the parameter of setting at deal with data set of the present invention herein:
Inhibiting factor σ was for enclosing 1.5 times of radius r of the border circular areas of picture zone homalographic with the user during adjacent tissue suppressed to handle in the step (3.2);
Step is defined as " background area " in (3.3): the remaining area after removing current segmentation result on the border circular areas.The center of circle of this border circular areas is seed points, radius for drawing a circle to approve 1.6 times of border circular areas radius of regional homalographic with the user.
The inventive method is by searching computer picture of mammary gland x-ray radiography method and system thereof, accurately and effectively retrieves some known diagnosis results' similar on the area-of-interest vision to be checked that obtains enclosing picture to the user and the pathology area-of-interest from the picture of mammary gland x-ray radiography area-of-interest computer picture reference database of setting up the known diagnosis result.Realization of the present invention is not limited to the disclosed scope of above-mentioned example, and persons skilled in the art can adopt the mode that is different from above-mentioned example to realize technical scheme of the present invention according to above-mentioned disclosed content.

Claims (7)

1. the search method of a computer picture of mammary gland x-ray radiography image, its step comprises:
Step (1) input one width of cloth computer picture of mammary gland x-ray radiography image;
Step (2) user-interactive ground encloses the zone of its concern of drawing on image, obtain area-of-interest to be checked;
Step (3) computing machine is cut apart area-of-interest to be checked, determines the border of tissue of interest;
Step (4) is based on the segmentation result to tissue of interest, and computing machine extracts the relevant single-view of several regions and the iconology feature of many views, and this iconology feature comprises geometric properties, morphological feature, gray feature and textural characteristics; The user selects single-view and/or many view feature of its concern, and the relevant weight of input feature vector importance;
Step (5) is utilized the characteristic similarity measurement criterion of weighting, retrieves the some area-of-interests similar to area-of-interest to be checked from the picture of mammary gland x-ray radiography area-of-interest computer picture reference database of setting up the known diagnosis result; Comprise in the reference database: region of interest area image, corresponding pathologic finding of area-of-interest and diagnostic result message file, and the relevant iconology representative record file of area-of-interest of computing machine extraction;
Step (6) is shown to the user with area-of-interest and regional corresponding pathologic finding and the diagnostic result information that retrieval obtains.
2. method according to claim 1 is characterized in that: step (2) comprises following process:
Step (2.1) is enclosed the zone of its concern of drawing by user-interactive ground on image, and calculates the center that the user encloses the picture zone;
Step (2.2) is the center of circle with the center that the user encloses the picture zone, do one and enclose the circle of picture zone homalographic with the user, calculate the intensity-weighted center of gravity of this circle, the center of circle is moved to this center of gravity place, recomputate the intensity-weighted center of gravity of enclosing picture zone homalographic circle with the user, repeat this process, multiplicity is imported by the user, gets the intensity-weighted center of gravity of calculating to the end;
Center of gravity when step (2.3) stops with iteration in the step (2.2) is the center, gets a square window image as area-of-interest to be checked on the picture of mammary gland x-ray radiography image of input.
3. method according to claim 1 and 2 is characterized in that: step (3) is cut apart tissue of interest according to following step in area-of-interest:
Step (3.1) adopts the least square fitting method to simulate big background planes such as and former area-of-interest, try to achieve the error image of original area-of-interest and background plane, error image obtains the region of interest area image that background trend is removed through after demarcating;
Step (3.2) is on the region of interest area image that background trend is removed, enclose the area in picture zone according to the user, roughly estimate the radius of this tissue of interest, the method of using " adjacent tissue's inhibition " keeps its original gray-scale value to the pixel that is in the radius, and the gray-scale value that is in the outer pixel of radius suppresses:
L(x，y)＝R(x，y，σ，r)·I(x，y)
R ( x , y , σ , r ) = e ( r - x 2 + y 2 ) 2 2 σ 2 ( x 2 + y 2 ) > r 1 else
The image that obtains after representing to suppress of L wherein through adjacent tissue, the region of interest area image of I for removing through background trend, R is an inhibition function, r represents to enclose with the user radius of the border circular areas of picture zone homalographic; σ is a constant, is used to regulate the inhibition degree;
Tissue of interest in the region of interest area image that step (3.3) obtains step (3.2) according to following process is cut apart:
(3.3.1) initial central pixel point with area-of-interest adds S set as seed points, and S is initially sky, and sets number of plies t and initial current threshold values, and t is a positive integer, and putting i is 0;
(3.3.2) with certain some adjacency among all and the S and with the absolute value of the gray value differences of seed points S exterior pixel point adding S less than current threshold value;
(3.3.3) repetitive process step (3.3.2), when not having new pixel to add S, i adds 1;
(3.3.4) when i=t, S set is final segmentation result, finishes; Otherwise, according to cut zone and on every side the contrast of background area calculate a new threshold value as current threshold value, jump toward step (3.3.2).
4. according to claim 1,2 or 3 described methods is characterized in that: the iconology feature that single-view is relevant in the step (4) specifically comprises the mean value of full figure gray-scale value, the mean value of full figure gray scale fluctuation, the standard deviation of full figure gray scale fluctuation, region significance, normalized mean radius length, the standard deviation of radius length, the measure of skewness of radius length, form factor, the standard deviation of gray-scale value in the zone, the standard deviation of profile gradient, the measure of skewness of profile gradient, the background area gray standard deviation, the fluctuation of background area average gray, the skew of standardization center, profile length, region area, contrast, circularity, girth area ratio, the average of standardization radius length, the standard deviation of standardization radius length, the entropy of standardization radius length, the area ratio of standardization radius length, the mistake remainder of standardization radius length.
5. method according to claim 4, it is characterized in that: the iconology feature that single-view is relevant in the step (4) also comprises following characteristics: center skew and maximum radius length ratio, the gradient of gray-scale value in the zone, the average acutance of profile, the profile average gradient, mean radius length, standardization radius length standard deviation and mean radius length ratio, the a little bigger ratio of local pole, and a little bigger average height of local pole.
6. according to claim 1,2 or 3 described methods, it is characterized in that: with many view area of area-of-interest to be retrieved coupling to mainly comprising: many view area of mating on the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue to the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue on many view area of mating right; Many view area of mating on the equal angular picture of mammary gland x-ray radiography image of heteropleural breast tissue are to being in separately same position on the image; Many view area of mating on the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue are not to being in separately same position on the image; When step (4) extract to as if many views associated picture when learning feature, right with adopting following method to search many view area of coupling, and extract the right iconology feature of many view area of coupling:
According to the distance that project to nipple of regional center on the axis, on the axis on the other width of cloth picture of mammary gland x-ray radiography image, find the position that equates apart from teat spacing, with this position is that a matching area band is set at the center, and in the matching area band, search for the highest zone of brightness, come thus to determine that many view area of mating on the different angles picture of mammary gland x-ray radiography image of homonymy breast tissue are right;
Many view area according to the coupling that finds are right, extract many view area to relevant geometric properties, morphological feature, gray feature and textural characteristics iconology feature, specifically comprise: the right correlated characteristic of many view area of coupling: effective size of focus, distance feature, comprehensive density feature and based on information-theoretical similarity measurement feature.
7. the searching system of a computer picture of mammary gland x-ray radiography image is characterized in that: it comprise load module (100), area-of-interest extraction module to be checked (200), tissue of interest cut apart module (300), tissue of interest characteristic extracting module (400), based on the retrieval module (500) and the output module (600) of characteristic similarity measurement criterion;
Load module (100) is used to receive the computer picture of mammary gland x-ray radiography image of input, and sends area-of-interest extraction module to be checked (200) to;
Area-of-interest extraction module to be checked (200) is used for carrying out alternately with the user, the area-of-interest that circle draws on image, and, extract the area-of-interest to be checked in the picture of mammary gland x-ray radiography of importing according to the described process of above-mentioned steps (2), send tissue of interest to and cut apart module (300);
Tissue of interest is cut apart module (300) according to the described process of above-mentioned steps (3), tissue of interest in the area-of-interest of area-of-interest extraction module to be checked (200) extraction is split, obtain the boundary information of tissue of interest, send tissue of interest characteristic extracting module (400) to;
Tissue of interest characteristic extracting module (400) is according to the boundary information of the tissue of interest that receives, according to the described process of above-mentioned steps (4), calculate the eigenwert that comprises geometric properties, morphological feature, gray feature and textural characteristics of single-view and many views; And carry out alternately with the user, select its single-view of paying close attention to and/or many view feature and the relevant weight of input feature vector importance for the user, and send information to retrieval module (500) based on the characteristic similarity measurement criterion;
Based on the retrieval module (500) of characteristic similarity measurement criterion according to the described process of above-mentioned steps (5), utilize the characteristic similarity measurement criterion of weighting, in the known diagnosis result's who has set up picture of mammary gland x-ray radiography area-of-interest computer picture reference database, retrieve the some known diagnosis results' similar area-of-interest to inquiring about area-of-interest, and the area-of-interest of k similarity value correspondence sends output module (600) to as result for retrieval before selecting according to user's needs, k is the number of the similar image of user's concern, by output module (600) area-of-interest and regional corresponding pathologic finding and the diagnostic result information that retrieval obtains is shown to the user.
CNA2008101970936A 2008-09-27 2008-09-27 Method and system for searching computer picture of mammary gland x-ray radiography Pending CN101373479A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CNA2008101970936A CN101373479A (en) 2008-09-27 2008-09-27 Method and system for searching computer picture of mammary gland x-ray radiography

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CNA2008101970936A CN101373479A (en) 2008-09-27 2008-09-27 Method and system for searching computer picture of mammary gland x-ray radiography

Publications (1)

Publication Number Publication Date
CN101373479A true CN101373479A (en) 2009-02-25

Family

ID=40447645

Family Applications (1)

Application Number Title Priority Date Filing Date
CNA2008101970936A Pending CN101373479A (en) 2008-09-27 2008-09-27 Method and system for searching computer picture of mammary gland x-ray radiography

Country Status (1)

Country Link
CN (1) CN101373479A (en)

Cited By (33)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
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
CN103477353A (en) * 2011-03-16 2013-12-25 皇家飞利浦有限公司 Method and system for intelligent linking of medical data
CN103578099A (en) * 2012-08-08 2014-02-12 深圳市慧康精密仪器有限公司 Method for extracting tumor elasticity characteristics based on ultrasonic elastography
CN103720474A (en) * 2012-10-12 2014-04-16 西门子公司 Inspection parameter calculating
WO2014114103A1 (en) * 2013-01-25 2014-07-31 中兴通讯股份有限公司 Method and apparatus for extracting test output information of automatic test platform
CN103988205A (en) * 2011-12-13 2014-08-13 国际商业机器公司 Method, device, and computer program for generating medical case-specific representative images and radiographic image interpretation information
CN104205166A (en) * 2012-02-17 2014-12-10 爱克发医疗保健公司 Method for defining a region of interest in a radiation image of a breast
CN104376121A (en) * 2014-12-04 2015-02-25 犹杰 Picture self-adaptation matching and combined displaying system, method and user terminal
CN104732213A (en) * 2015-03-23 2015-06-24 中山大学 Computer-assisted lump detecting method based on mammary gland magnetic resonance image
CN104809717A (en) * 2014-01-28 2015-07-29 上海西门子医疗器械有限公司 Non-rigid medical image interest point marking and positioning method and device and medical device
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
CN104978753A (en) * 2014-04-02 2015-10-14 深圳市蓝韵实业有限公司 Dividing algorithm for nipple area in mammary gland X-ray image
CN106061379A (en) * 2014-01-10 2016-10-26 佳能株式会社 Processing device, processing method, and program
CN106133724A (en) * 2014-03-27 2016-11-16 富士胶片株式会社 Similar cases retrieval device, similar cases search method and similar cases search program
CN106164903A (en) * 2014-03-27 2016-11-23 富士胶片株式会社 Similar cases retrieval device, similar cases search method and similar cases search program
CN106446004A (en) * 2016-07-30 2017-02-22 北京航空航天大学 Digital pathological full-slice image retrieval method
CN106560827A (en) * 2015-09-30 2017-04-12 松下知识产权经营株式会社 Control Method
CN104182755B (en) * 2014-08-30 2017-04-12 西安电子科技大学 Mammary gland molybdenum target X-ray image block feature extraction method based on tower-shaped principal component analysis (PCA)
CN106776710A (en) * 2016-11-18 2017-05-31 广东技术师范学院 A kind of picture and text construction of knowledge base method based on vertical search engine
CN106777400A (en) * 2017-03-03 2017-05-31 长春理工大学 A kind of computer control system for being applied to medical image retrieval
CN106846317A (en) * 2017-02-27 2017-06-13 北京连心医疗科技有限公司 A kind of feature based extracts the method for retrieving medicine image with Similarity matching
CN107092774A (en) * 2017-03-10 2017-08-25 深圳华大智造科技有限公司 A kind of method and device that reference information is provided
WO2017185727A1 (en) * 2016-04-28 2017-11-02 深圳市前海安测信息技术有限公司 Big-data-based breast screening system and method
CN107862688A (en) * 2017-11-07 2018-03-30 山东浪潮云服务信息科技有限公司 A kind of method and device for medical image aid in diagosis
CN109363699A (en) * 2018-10-16 2019-02-22 杭州依图医疗技术有限公司 A kind of method and device of breast image lesion identification
CN110475503A (en) * 2017-03-30 2019-11-19 富士胶片株式会社 The working method of medical image processing device and endoscopic system and medical image processing device
CN110706794A (en) * 2019-09-26 2020-01-17 中国科学院深圳先进技术研究院 Medical image processing system and medical image processing method
CN110727819A (en) * 2019-10-10 2020-01-24 北京航空航天大学 Method for retrieving scale-adaptive pathological full-section image database
CN114018202A (en) * 2021-11-08 2022-02-08 绍兴职业技术学院 Novel algorithm for rapidly evaluating roundness
US20220218294A1 (en) * 2021-01-13 2022-07-14 Hologic, Inc. Systems and methods for applying force to breasts for imaging
CN115578389A (en) * 2022-12-08 2023-01-06 青岛澳芯瑞能半导体科技有限公司 Defect detection method of groove MOS device
CN116965843A (en) * 2023-09-19 2023-10-31 南方医科大学南方医院 Mammary gland stereotactic system
CN110996772B (en) * 2017-08-15 2024-04-02 玛雷迪夫美国公司 Breast cancer detection

Cited By (49)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN103477353A (en) * 2011-03-16 2013-12-25 皇家飞利浦有限公司 Method and system for intelligent linking of medical data
CN103988205A (en) * 2011-12-13 2014-08-13 国际商业机器公司 Method, device, and computer program for generating medical case-specific representative images and radiographic image interpretation information
CN104205166A (en) * 2012-02-17 2014-12-10 爱克发医疗保健公司 Method for defining a region of interest in a radiation image of a breast
CN104205166B (en) * 2012-02-17 2017-10-13 爱克发医疗保健公司 Method for limiting area-of-interest in the radiation image of mammary gland
CN103578099B (en) * 2012-08-08 2016-12-21 深圳市慧康精密仪器有限公司 The extracting method of tumor elastic characteristic based on ultrasonic elastograph imaging
CN103578099A (en) * 2012-08-08 2014-02-12 深圳市慧康精密仪器有限公司 Method for extracting tumor elasticity characteristics based on ultrasonic elastography
CN103720474A (en) * 2012-10-12 2014-04-16 西门子公司 Inspection parameter calculating
CN103720474B (en) * 2012-10-12 2017-03-01 西门子公司 Calculate inspection parameter
WO2014114103A1 (en) * 2013-01-25 2014-07-31 中兴通讯股份有限公司 Method and apparatus for extracting test output information of automatic test platform
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
CN106061379A (en) * 2014-01-10 2016-10-26 佳能株式会社 Processing device, processing method, and program
US10102622B2 (en) 2014-01-10 2018-10-16 Canon Kabushiki Kaisha Processing apparatus, processing method, and non-transitory computer-readable storage medium
CN104809717B (en) * 2014-01-28 2018-03-23 上海西门子医疗器械有限公司 Non-rigid medical image interest point identification, localization method and device, Medical Devices
CN104809717A (en) * 2014-01-28 2015-07-29 上海西门子医疗器械有限公司 Non-rigid medical image interest point marking and positioning method and device and medical device
CN106133724A (en) * 2014-03-27 2016-11-16 富士胶片株式会社 Similar cases retrieval device, similar cases search method and similar cases search program
CN106164903A (en) * 2014-03-27 2016-11-23 富士胶片株式会社 Similar cases retrieval device, similar cases search method and similar cases search program
CN106133724B (en) * 2014-03-27 2019-09-03 富士胶片株式会社 Similar cases retrieve device and similar cases search method
CN106164903B (en) * 2014-03-27 2019-06-18 富士胶片株式会社 Similar cases retrieve device and similar cases search method
CN104978753A (en) * 2014-04-02 2015-10-14 深圳市蓝韵实业有限公司 Dividing algorithm for nipple area in mammary gland X-ray image
CN104182755B (en) * 2014-08-30 2017-04-12 西安电子科技大学 Mammary gland molybdenum target X-ray image block feature extraction method based on tower-shaped principal component analysis (PCA)
CN104376121B (en) * 2014-12-04 2018-03-27 深圳大数点科技有限公司 System, method and the user terminal that a kind of picture Adaptive matching combination is presented
CN104376121A (en) * 2014-12-04 2015-02-25 犹杰 Picture self-adaptation matching and combined displaying system, method and user terminal
CN104732213A (en) * 2015-03-23 2015-06-24 中山大学 Computer-assisted lump detecting method based on mammary gland magnetic resonance image
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
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
CN106560827B (en) * 2015-09-30 2021-11-26 松下知识产权经营株式会社 Control method
CN106560827A (en) * 2015-09-30 2017-04-12 松下知识产权经营株式会社 Control Method
WO2017185727A1 (en) * 2016-04-28 2017-11-02 深圳市前海安测信息技术有限公司 Big-data-based breast screening system and method
CN106446004B (en) * 2016-07-30 2020-01-17 北京航空航天大学 Digital pathological full-section image retrieval method
CN106446004A (en) * 2016-07-30 2017-02-22 北京航空航天大学 Digital pathological full-slice image retrieval method
CN106776710A (en) * 2016-11-18 2017-05-31 广东技术师范学院 A kind of picture and text construction of knowledge base method based on vertical search engine
CN106846317B (en) * 2017-02-27 2021-09-17 北京连心医疗科技有限公司 Medical image retrieval method based on feature extraction and similarity matching
CN106846317A (en) * 2017-02-27 2017-06-13 北京连心医疗科技有限公司 A kind of feature based extracts the method for retrieving medicine image with Similarity matching
CN106777400A (en) * 2017-03-03 2017-05-31 长春理工大学 A kind of computer control system for being applied to medical image retrieval
CN107092774A (en) * 2017-03-10 2017-08-25 深圳华大智造科技有限公司 A kind of method and device that reference information is provided
US11412917B2 (en) 2017-03-30 2022-08-16 Fujifilm Corporation Medical image processor, endoscope system, and method of operating medical image processor
CN110475503A (en) * 2017-03-30 2019-11-19 富士胶片株式会社 The working method of medical image processing device and endoscopic system and medical image processing device
CN110996772B (en) * 2017-08-15 2024-04-02 玛雷迪夫美国公司 Breast cancer detection
CN107862688A (en) * 2017-11-07 2018-03-30 山东浪潮云服务信息科技有限公司 A kind of method and device for medical image aid in diagosis
CN109363699A (en) * 2018-10-16 2019-02-22 杭州依图医疗技术有限公司 A kind of method and device of breast image lesion identification
CN110706794A (en) * 2019-09-26 2020-01-17 中国科学院深圳先进技术研究院 Medical image processing system and medical image processing method
CN110727819A (en) * 2019-10-10 2020-01-24 北京航空航天大学 Method for retrieving scale-adaptive pathological full-section image database
US20220218294A1 (en) * 2021-01-13 2022-07-14 Hologic, Inc. Systems and methods for applying force to breasts for imaging
CN114018202B (en) * 2021-11-08 2024-02-02 绍兴职业技术学院 Novel algorithm for rapidly evaluating roundness
CN114018202A (en) * 2021-11-08 2022-02-08 绍兴职业技术学院 Novel algorithm for rapidly evaluating roundness
CN115578389A (en) * 2022-12-08 2023-01-06 青岛澳芯瑞能半导体科技有限公司 Defect detection method of groove MOS device
CN116965843A (en) * 2023-09-19 2023-10-31 南方医科大学南方医院 Mammary gland stereotactic system
CN116965843B (en) * 2023-09-19 2023-12-01 南方医科大学南方医院 Mammary gland stereotactic system

Similar Documents

Publication Publication Date Title
CN101373479A (en) Method and system for searching computer picture of mammary gland x-ray radiography
Gu et al. Automatic lung nodule detection using a 3D deep convolutional neural network combined with a multi-scale prediction strategy in chest CTs
US20220319008A1 (en) Automated tumor identification and segmentation with medical images
Valente et al. Automatic 3D pulmonary nodule detection in CT images: a survey
Way et al. Computer‐aided diagnosis of pulmonary nodules on CT scans: improvement of classification performance with nodule surface features
Qin et al. Pulmonary nodule segmentation with CT sample synthesis using adversarial networks
US8335359B2 (en) Systems, apparatus and processes for automated medical image segmentation
CN109583440A (en) It is identified in conjunction with image and reports the medical image aided diagnosis method edited and system
US8355553B2 (en) Systems, apparatus and processes for automated medical image segmentation using a statistical model
US11348243B2 (en) Systems and methods for medical image style transfer using deep neural networks
Ge et al. Computer‐aided detection of lung nodules: false positive reduction using a 3D gradient field method and 3D ellipsoid fitting
Narayanan et al. Performance analysis of a computer-aided detection system for lung nodules in CT at different slice thicknesses
Han et al. A texture feature analysis for diagnosis of pulmonary nodules using LIDC-IDRI database
CN106682435A (en) System and method for automatically detecting lesions in medical image through multi-model fusion
CN1979509A (en) Method and apparatus for creating a multi-resolution framework for improving medical imaging workflow
Zhu et al. Increasing computer‐aided detection specificity by projection features for CT colonography
Xu et al. Quantifying the margin sharpness of lesions on radiological images for content‐based image retrieval
Maitra et al. Automated digital mammogram segmentation for detection of abnormal masses using binary homogeneity enhancement algorithm
US20230368515A1 (en) Similarity determination apparatus, similarity determination method, and similarity determination program
Shi et al. Automatic nodule segmentation method for CT images using aggregation-U-Net generative adversarial networks
Gundreddy et al. Assessment of performance and reproducibility of applying a content‐based image retrieval scheme for classification of breast lesions
Zhao et al. A deep-learning based automatic pulmonary nodule detection system
CN101052993A (en) Multi-scale filter synthesis for medical image registration
CN101799806A (en) Chest digital X-ray image retrieval system based on multi-information fusion
Junior et al. Evaluating margin sharpness analysis on similar pulmonary nodule retrieval

Legal Events

Date Code Title Description
C06 Publication
PB01 Publication
C10 Entry into substantive examination
SE01 Entry into force of request for substantive examination
C02 Deemed withdrawal of patent application after publication (patent law 2001)
WD01 Invention patent application deemed withdrawn after publication

Open date: 20090225