CN103778600A - Image processing system - Google Patents
Image processing system Download PDFInfo
- Publication number
- CN103778600A CN103778600A CN201210411592.7A CN201210411592A CN103778600A CN 103778600 A CN103778600 A CN 103778600A CN 201210411592 A CN201210411592 A CN 201210411592A CN 103778600 A CN103778600 A CN 103778600A
- Authority
- CN
- China
- Prior art keywords
- tumor
- profiles
- candidate
- feature
- image
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10116—X-ray image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Multimedia (AREA)
- Medical Informatics (AREA)
- Quality & Reliability (AREA)
- Radiology & Medical Imaging (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Ultra Sonic Daignosis Equipment (AREA)
Abstract
An image processing system for breast imaging is provided. The image processing system comprise the following components of: an image collector which is used for obtaining two-dimensional breast ultrasonic images; a tumor detector which is used for detecting M interest areas containing breast tumors from the obtained two-dimensional breast ultrasonic images and scoring each interest area to obtain a detection score of the corresponding interest area, wherein M is larger than 0; a multi-parameter segmenting device which is used for segmenting K candidate tumor contours from each of the M interest areas through using K tumor segmenting algorithms, and recording detection scores for each candidate tumor contour, wherein K is larger than 0; a feature scoring device which is used for evaluating and scoring each candidate tumor contour according to at least one predetermined feature; a fusion device which is used for selecting finally-segmented tumor contours from M*K candidate tumor contours according to detection scores and feature scores of the M*K candidate tumor contours.
Description
Technical field
The application relates to a kind of image processing system for breast ultrasound ripple image, relate in particular to and a kind ofly detect the region-of-interest of multiple tumors of breast and give a mark from breast ultrasound ripple image, use multiple lesion segmentation approach of multiparameter that each region-of-interest is carried out dividing processing and given a mark, and from multiple segmentation results, select the image processing techniques of the segmentation result that aggregative index is higher.
Background technology
Breast cancer is women's second largest killer, and early detection is to reduce the key of mortality ratio (40% or more than).Ultrasound wave is used to breast imaging as the supplemental diagnostics test of Mammogram (X ray) more and more, in the time that Mammogram may Reduced susceptibility occur or in the time that Mammogram exists unacceptable radiation risk, itself is also used as First Line imaging technique.Therefore, computer-aided diagnosis (CAD) system can help the doctor who lacks experience to avoid mistaken diagnosis, reduces the quantity of benign lesion biopsy under the prerequisite of not mistaken diagnosis cancer, and reduces the variation of various detections.
In computer assisted breast ultrasound ripple diagnostic system, core technology comprises lesion detection and cuts apart, and lesion segmentation is to determine that tumour is optimum or pernicious key.Existing lesion detection is processed the single testing result of output, and lesion segmentation processing is carried out lesion segmentation processing for described single testing result.In this process, the segmentation result that detects mistake or incorrect testing result and use unsuitable parameter all can lead to errors.Such area of computer aided breast ultrasound ripple diagnosis is powerful not, and the accuracy of segmentation result is also unstable.
Summary of the invention
The object of the present invention is to provide a kind of image processing system for breast ultrasound ripple image, detect the region-of-interest of multiple tumors of breast and give a mark from breast ultrasound ripple image, use multiple lesion segmentation approach of multiparameter to carry out dividing processing to each region-of-interest, and each tumor's profiles being partitioned into is carried out to the feature marking of multiple features, after this from multiple segmentation results, select the image processing techniques of the segmentation result that aggregative index is higher, make segmentation result not be subject to the impact of the incorrect testing result of part and/or partitioning parameters improper use, thereby improve, tumor of breast detects and the stability of dividing processing result.
Another object of the present invention is to provide a kind of image processing system for breast ultrasound ripple image, use multiple lesion segmentation approach of multiparameter to carry out dividing processing to each region-of-interest from marking the breast ultrasound ripple image of multiple region-of-interests, and each tumor's profiles being partitioned into is carried out to the feature marking of multiple features, select the image processing techniques of the segmentation result that characteristic exponent is higher, make segmentation result not be subject to the impact of part partitioning parameters improper use, thereby improve the stability of tumor of breast dividing processing result.
According to an aspect of the present invention, provide a kind of image processing system for breast image, comprising: image collection device, for obtaining two-dimentional breast ultrasound ripple image; Lesion detection device, detects M region-of-interest (ROI) that comprises tumor of breast for the two-dimentional breast ultrasound ripple image obtaining from image collection device, and is that each ROI marking detects mark as it, wherein, and M > 0; Multiparameter dispenser, for using respectively K lesion segmentation algorithm of parameter based on different or parameter combinations from each K of being partitioned into candidate's tumor's profiles of a described M ROI, and be the detection mark that each candidate's tumor's profiles record is partitioned into the ROI of described candidate's tumor's profiles, wherein, K > 0; Feature scoring device, for evaluating marking as its feature scores according at least one predetermined feature to the each of the M*K being partitioned into candidate's tumor's profiles; Fusion device for according to detection mark and the feature scores of described M*K candidate's tumor's profiles, selects candidate's tumor's profiles as the tumor's profiles being finally partitioned in the middle of described M*K candidate's tumor's profiles.
Described image processing system can also comprise: combiner, for the M*K being partitioned into candidate's tumor's profiles calculated respectively to similarity each other, and not the highest candidate's tumor's profiles from every group of similarity higher than removing detection mark in the middle of multiple candidate's tumor's profiles of predetermined value, thereby obtain N candidate's tumor's profiles, wherein, N < < M*K; Wherein, feature scoring device is given a mark as its feature scores to each evaluation the in described N candidate's tumor's profiles according at least one predetermined feature, fusion device, according to detection mark and the feature scores of described N candidate's tumor's profiles, is selected candidate's tumor's profiles.
Described image processing system can also comprise: pretreater, and for the two-dimentional breast ultrasound ripple image obtaining is carried out to pre-service, described pre-service comprises two-dimentional breast ultrasound ripple image execution Denoising disposal and/or the image enhancement processing to obtaining; Wherein, multiparameter dispenser is to carrying out described lesion segmentation processing through pretreated two-dimentional breast ultrasound ripple image.
Lesion detection device can use deformations bit model method, template matching method or Adaboost method to carry out detection and the marking of described M ROI that comprises tumor of breast.
Described different parameter can be different iterationses, different ratio or diverse ways, and described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
Feature scoring device can use support vector regression method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.
Described contour feature can be grey-scale contrast, intensity contrast or Gestar feature.
Described tumour feature can be rear acoustical signature or echo mode feature.
Fusion device can be normalized detection mark and the feature scores of each candidate's tumor's profiles respectively, calculates the comprehensive mark of each candidate's tumor's profiles according to following formula, and selects candidate's tumor's profiles that comprehensive mark is the highest:
Score
combined＝w
ds×NDS+w
rs×NRS
Wherein, Score
combinedthe comprehensive mark of candidate's tumor's profiles, w
dsand w
rsbe respectively to give the detection mark of candidate's tumor's profiles and the weights of feature scores, NDS and NRS are respectively detection mark and the feature scores of candidate's tumor's profiles.
Fusion device can use s-score normalization algorithm, min-max normalization algorithm, Tanh estimator or double sigmoid algorithm to carry out described normalized.
Feature scoring device can be used support vector regression method, the Jaccard index that calculates each candidate's tumor's profiles according to the detection mark of at least one feature in following characteristics and candidate's tumor's profiles is as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature, wherein, fusion device is from selecting in the middle of described M*K candidate's tumor's profiles candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
Described different parameter can be different iterations, different ratio, different step, different Denoising disposal or different image enchancing methods, and described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
According to a further aspect in the invention, provide a kind of image processing system for breast image, comprising: image collection device, mark for obtaining the two-dimentional breast ultrasound ripple image that has M tumour ROI, wherein, M > 0; Multiparameter dispenser, for using respectively K lesion segmentation algorithm of parameter based on different or parameter combinations from each K of being partitioned into candidate's tumor's profiles of a described M ROI, wherein, K > 0; Feature scoring device, for according at least one predetermined feature, each evaluation of the M*K being partitioned into candidate's tumor's profiles being given a mark as its feature scores, and from described M*K candidate's tumor's profiles, select candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
Described different parameter can be different iterations, different ratio or different steps, and described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
Feature scoring device can use support vector regression method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.
Described contour feature can be grey-scale contrast, intensity contrast or Gestar feature.Described tumour feature can be rear acoustical signature or echo mode feature.
According to a further aspect in the invention, provide a kind of image processing method for breast image, comprising: a) obtain two-dimentional breast ultrasound ripple image; B) detect M ROI that comprises tumor of breast from the two-dimentional breast ultrasound ripple image obtaining, and be that each ROI marking detects mark as it, wherein, M > 0; C) each be partitioned into K the candidate tumor's profiles of K the lesion segmentation algorithm that uses respectively parameter based on different or parameter combinations from a described M ROI, and be the detection mark that each candidate's tumor's profiles record is partitioned into the ROI of described candidate's tumor's profiles, wherein, K > 0; D) according at least one predetermined feature, each evaluation the in the M*K being partitioned into candidate's tumor's profiles given a mark as its feature scores; E), according to detection mark and the feature scores of described M*K candidate's tumor's profiles, in the middle of described M*K candidate's tumor's profiles, select candidate's tumor's profiles as the tumor's profiles being finally partitioned into.
Described method can also comprise: g) M*K being partitioned into candidate's tumor's profiles calculated respectively to similarity each other, and not the highest candidate's tumor's profiles from every group of similarity higher than removing detection mark in the middle of multiple candidate's tumor's profiles of predetermined value, thereby obtain N candidate's tumor's profiles, wherein, N < < M*K; Wherein, in steps d) in, according at least one predetermined feature, each evaluation the in described N candidate's tumor's profiles given a mark as its feature scores, and step e) in, according to detection mark and the feature scores of described N candidate's tumor's profiles, select candidate's tumor's profiles.
Described method can also comprise: b) front in execution step, the two-dimentional breast ultrasound ripple image obtaining is carried out to pre-service, and described pre-service comprises two-dimentional breast ultrasound ripple image execution Denoising disposal and/or the image enhancement processing to obtaining; Wherein, in step c), to carrying out described lesion segmentation processing through pretreated two-dimentional breast ultrasound ripple image.
Step b) in, can use deformations bit model method, template matching method or Adaboost method to carry out detection and the marking of the ROI that described M comprises tumor of breast.
Step c) in, described different parameter can be different iterations, different ratio or different steps, and described lesion segmentation approach is Level Set Method, figure cutting method, robs segmentation method or watershed algorithm.
In steps d) in, can use support vector regression method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.
Described contour feature can be grey-scale contrast, intensity contrast or Gestar feature.Described tumour feature can be rear acoustical signature or echo mode feature.
Step e) in, detection mark and the feature scores of each candidate's tumor's profiles can be normalized respectively, calculate the comprehensive mark of each candidate's tumor's profiles according to following formula, and select candidate's tumor's profiles that comprehensive mark is the highest:
Score
combined＝w
ds×NDS+w
rs×NRS
Wherein, Score
combinedthe comprehensive mark of candidate's tumor's profiles, w
dsand w
rsbe respectively to give the detection mark of candidate's tumor's profiles and the weights of feature scores, NDS and NRS are respectively detection mark and the feature scores of candidate's tumor's profiles.
Step e) in, can use s-score normalization algorithm, min-max normalization algorithm, Tanh estimator or double sigmoid algorithm to carry out described normalized.
In steps d) in, can use support vector regression method, the Jaccard index that calculates each candidate's tumor's profiles according to the detection mark of at least one feature in following characteristics and candidate's tumor's profiles is as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.Wherein, step e) in, can be from selecting in the middle of described M*K candidate's tumor's profiles candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
Described different parameter can be different iterations, different ratio, different step, different Denoising disposal or different image enchancing methods, and described lesion segmentation approach is Level Set Method, figure cutting method, robs segmentation method or watershed algorithm.
According to a further aspect in the invention, provide a kind of image processing method for breast image, comprising: a) obtaining mark has the two-dimentional breast ultrasound ripple image of M tumour ROI, wherein, M > 0; B) each be partitioned into K the candidate tumor's profiles of K the lesion segmentation algorithm that uses respectively parameter based on different or parameter combinations from a described M ROI, wherein, K > 0; C) according at least one predetermined feature, each evaluation the in the M*K being partitioned into candidate's tumor's profiles given a mark as its feature scores; D) from described M*K candidate's tumor's profiles, select candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
Step b) in, described different parameter can be different iterations, different ratio or different steps, and described lesion segmentation approach is Level Set Method, figure cutting method, robs segmentation method or watershed algorithm.
Step c) in, can use support vector regression method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.
Described contour feature can be grey-scale contrast, intensity contrast or Gestar feature.Described tumour feature can be rear acoustical signature or echo mode feature.
Accompanying drawing explanation
By the description of carrying out below in conjunction with accompanying drawing, above and other object of the present invention and feature will become apparent, wherein:
Fig. 1 be illustrate according to the logic diagram of the image processing system of exemplary embodiment of the present invention with and the schematic diagram of carries out image processing;
Fig. 2 is the process flow diagram illustrating according to the image processing method of exemplary embodiment of the present invention;
Fig. 3 illustrates the candidate's tumor's profiles being partitioned into according to the image processing method of exemplary embodiment of the present invention;
Fig. 4 illustrates according to the process flow diagram of the image processing method of another exemplary embodiment of the present invention.
Embodiment
Below, describe with reference to the accompanying drawings embodiments of the invention in detail.
Fig. 1 be illustrate according to the logic diagram of the image processing system of exemplary embodiment of the present invention with and the schematic diagram of carries out image processing.
With reference to Fig. 1, image processing system comprises image collection device 100, lesion detection device 110, multiparameter dispenser 120, feature scoring device 140 and fusion device 150.
(or through pretreated) two-dimentional breast ultrasound ripple image that lesion detection device 110 obtains from image collection device 100 detects M region-of-interest (ROI) that comprises tumor of breast, and be that each region-of-interest marking detects mark as it, wherein, M > 0.Each ROI is the rectangular region that comprises actual tumour, and these regions can be overlapped.In Fig. 1, lesion detection device 110 downsides illustrate the ROI of multiple light collimation mark notes.Lesion detection device 110 can use deformations bit model (DPM) method, template matching method or Adaboost method to carry out detection and the marking of described M ROI that comprises tumor of breast, but lesion detection approach of the present invention is not limited to said method.
For this M ROI, each be partitioned into K the candidate tumor's profiles of K the lesion segmentation algorithm that multiparameter dispenser 120 uses respectively parameter based on different or parameter combinations from a described M ROI, and be the detection mark that each candidate's tumor's profiles record is partitioned into the ROI of described candidate's tumor's profiles, wherein, K > 0.Described different parameter is (but being not limited to) different iterations, different ratio or different steps, and described lesion segmentation approach can be (but being not limited to) Level Set Method, figure cutting method, region growing method or watershed algorithm.In the embodiment that comprises pretreater, described different parameter also comprises different Denoising disposals or different image enchancing methods.By this step, image processing system obtains M*K the candidate's tumor's profiles being partitioned into.In Fig. 1, multiparameter dispenser 120 downsides illustrate the part candidate's tumor's profiles obtaining through the processing of described step.Wherein, two candidate's tumor's profiles of the top are closely similar.
According to exemplary embodiment of the present invention, in order to save computational resource and operation time, image processing system also comprises combiner 130.The M*K that combiner 130 is partitioned into multiparameter dispenser 120 candidate's tumor's profiles calculates respectively similarity each other, and not the highest candidate's tumor's profiles from every group of similarity higher than removing detection mark in the middle of multiple candidate's tumor's profiles of predetermined value, thereby obtain N candidate's tumor's profiles, wherein, N < < M*K.If image processing system is not considered the factor of operation time and calculation process amount, can not comprise combiner 130.
Score
combined＝w
ds×NDS+w
rs×NRS
Wherein, Score
combinedthe comprehensive mark of candidate's tumor's profiles, w
dsand w
rsbe respectively to give the detection mark of candidate's tumor's profiles and the weights of feature scores, NDS and NRS are respectively detection mark and the feature scores of candidate's tumor's profiles.
According to optional exemplary embodiment of the present invention, feature scoring device 140 is used SVR method, and the Jaccard index that calculates each candidate's tumor's profiles according to the detection mark of at least one feature in textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature and candidate's tumor's profiles is as its feature scores; In this case, fusion device 150 is from selecting in the middle of described M*K candidate's tumor's profiles candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
Image processing system according to the present invention detects multiple ROI from two-dimentional breast ultrasound ripple image, and described multiple ROI are used as the basis that mammary gland is cut apart; The present invention also adopts the multiple lesion segmentation approach based on multiparameter to carry out lesion segmentation processing to described multiple ROI, obtains candidate's tumor's profiles of greater number, and each candidate's tumor's profiles is carried out to the evaluation marking of multiple features; Thus, the marking of image processing system in can comprehensive detection processing and evaluate the feature marking of processing, thereby according to the highest candidate's tumor's profiles of the comprehensive marking of certain Standard Selection as its final lesion segmentation result.What this processing mode was more traditional only detects a tumor region, the mode of only carrying out a kind of lesion segmentation processing for this tumor region is compared, more can guarantee the Stability and veracity of Output rusults, and not be subject to the impact of the indivedual mistakes that occur in arbitrary link.
In some cases, before by computer aided system execution processing, the prior breast ultrasound ripple image taking of doctor's possibility is in the region that marks one or more tumours existence.Now,, as input, wherein, process the region that one or more tumours of mark exist in advance respectively in the region that image processing system can be using two-dimentional breast ultrasound ripple image and one or more tumours of mark exist in advance as ROI.According to another exemplary embodiment of the present invention, image processing system of the present invention comprises image collection device 200, multiparameter dispenser 120 and feature scoring device 140.Image collection device 100 can obtain the two-dimentional breast ultrasound ripple image that has marked M tumour ROI, wherein, and M > 0.Image collection device 100 can as required, carry out suitable processing (as ROI being adjusted into rectangle etc.) to the tumour ROI of mark.Multiparameter dispenser 120 is for using respectively K lesion segmentation algorithm of parameter based on different or parameter combinations from each K of being partitioned into candidate's tumor's profiles of a described M ROI, wherein, K > 0, thus be partitioned into altogether M*K candidate's tumor's profiles.Feature scoring device 140 as previously mentioned, is evaluated marking as its feature scores at least one feature that basis is predetermined to the each of the M*K being partitioned into candidate's tumor's profiles; In addition, from M*K candidate's tumor's profiles described in feature scoring device 140, select candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
Describe in detail according to the image processing method of exemplary embodiment of the present invention hereinafter with reference to Fig. 2-Fig. 4.
Fig. 2 is the process flow diagram illustrating according to the image processing method of exemplary embodiment of the present invention.
With reference to Fig. 2, at step S100, image processing system obtains two-dimentional breast ultrasound ripple image.Image processing system can obtain described two-dimentional breast ultrasound ripple image from connected supersonic imaging device, also can read described two-dimentional breast ultrasound ripple image from information storage medium.
The ultrasonograph that supersonic imaging device is taken often contains " noise " of such as spot etc.According to a preferred embodiment of the invention, in order to obtain preferably image processing effect, at step S105, image processing system is carried out pre-service to the two-dimentional breast ultrasound ripple image obtaining, and described pre-service comprises two-dimentional breast ultrasound ripple image execution Denoising disposal and/or the image enhancement processing to obtaining.But step S105 is optional step, rather than the step that must carry out.
After this, at step S110, (or through pretreated) the two-dimentional breast ultrasound ripple image obtaining from image processing system detects M ROI that comprises tumor of breast, and is that each ROI marking detects mark as it, wherein, M > 0.Can use deformations bit model (DPM) method, template matching method or AdaBoost method to carry out detection and the marking of described M ROI that comprises tumor of breast.
At step S120, each be partitioned into K the candidate tumor's profiles of K the lesion segmentation algorithm that image processing system uses respectively parameter based on different or parameter combinations from a described M ROI, and be the detection mark that each candidate's tumor's profiles record is partitioned into the ROI of described candidate's tumor's profiles, wherein, K > 0.Through the processing of step S120, will obtain M*K candidate's tumor's profiles.Fig. 3 illustrates that a ROI who detects from the two-dimentional breast ultrasound ripple image in left side carries out 3 lesion segmentation algorithms and cuts apart the 3 candidate's tumor's profiles (right side) that obtain.
Here, described K lesion segmentation algorithm is to use different parameters or the kinds of tumors partitioning algorithm of parameter combinations.Described different parameter is different iterations, different ratio or different steps.In the pretreated situation of execution step S105, described different parameter can also comprise different Denoising disposals or different image enchancing methods.Described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
After this, at step S130, image processing system calculates respectively similarity each other to the M*K being partitioned into candidate's tumor's profiles, and not the highest candidate's tumor's profiles from every group of similarity higher than removing detection mark in the middle of multiple candidate's tumor's profiles of predetermined value, thereby obtain N candidate's tumor's profiles, wherein, N < < M*K.
If image processing system is not considered the long and excessive impact of calculation process amount operation time, can not carry out S130.
At step S140, image processing system is given a mark as its feature scores to each evaluation the in described N candidate's tumor's profiles according at least one predetermined feature.In the situation that not performing step S130, will each evaluation the in the M*K being partitioned into candidate's tumor's profiles be given a mark as its feature scores.
According to exemplary embodiment of the present invention, image processing system use SVR method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.Wherein, described contour feature is grey-scale contrast, intensity contrast or Gestar feature, and described tumour feature is rear acoustical signature or echo mode feature.But, the invention is not restricted to use support vector regression method, also can use other characteristics of image method for evaluating similarity, and the feature that the present invention is also not limited to listing at this is evaluated marking.
After this,, at step S150, image processing system, according to detection mark and the feature scores of described N candidate's tumor's profiles, selects candidate's tumor's profiles as the final tumor's profiles being partitioned into from described N candidate's tumor's profiles.
With multiple candidate's tumor's profiles using and detect mark and feature scores as input, according to preference and stressing, can carry out by multiple fusion method the comprehensive evaluation of candidate's tumor's profiles, thereby select the highest candidate's tumor's profiles of comprehensive evaluation as the tumor's profiles being finally partitioned into.
According to exemplary embodiment of the present invention, image processing system is normalized detection mark and the feature scores of each candidate's tumor's profiles respectively, calculate the comprehensive mark of each candidate's tumor's profiles according to following formula, and select candidate's tumor's profiles that comprehensive mark is the highest:
Score
combined＝w
ds×NDS+w
rs×NRS
Wherein, Score
combinedthe comprehensive mark of candidate's tumor's profiles, w
dsand w
rsbe respectively to give the detection mark of candidate's tumor's profiles and the weights of feature scores, NDS and NRS are respectively detection mark and the feature scores of candidate's tumor's profiles.
Here can use s-score normalization algorithm, min-max normalization algorithm, Tanh estimator or double sigmoid algorithm to carry out the normalized of described mark.
According to an alternative embodiment of the invention, at step S140, image processing system uses support vector regression method, and the Jaccard index that calculates each candidate's tumor's profiles according to the detection mark of at least one feature in following characteristics and candidate's tumor's profiles is as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.At step S150, image processing system is from selecting in the middle of described M*K candidate's tumor's profiles candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
According to another optional embodiment of the present invention, the tumor's profiles being partitioned into that image processing system output is final.
Fig. 4 illustrates according to the process flow diagram of the image processing method of another exemplary embodiment of the present invention.
With reference to Fig. 4, at step S200, image processing system obtains one or more regions of two-dimentional breast ultrasound ripple image and mark.Image processing system is using each region of mark as ROI.Here suppose to have M ROI.
At step S220, each be partitioned into K the candidate tumor's profiles of K the lesion segmentation algorithm that image processing system uses respectively parameter based on different or parameter combinations from a described M ROI, wherein, K > 0.Can carry out dividing processing described here as step S120 in Fig. 2, different, in the situation of input tab area, there is not detection mark.
At step S240, image processing system is given a mark as its feature scores to each evaluation the in the M*K being partitioned into candidate's tumor's profiles according at least one predetermined feature, and from described M*K candidate's tumor's profiles, selects candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
According to finding out the description of exemplary embodiment of the present invention, thereby image processing system of the present invention and method can be cut apart and obtain multiple candidate's tumor's profiles each ROI execution multiparameter detecting, and described multiple candidate's tumor's profiles are carried out to feature marking, thereby the mark of comprehensive detection and feature scores are chosen best candidate's tumor's profiles as final segmentation result, what this processing mode was more traditional only detects a tumor region, the mode of only carrying out a kind of lesion segmentation processing for this tumor region is compared, more can guarantee the Stability and veracity of Output rusults, and be not subject to the impact of the indivedual mistakes that occur in arbitrary link.
In addition, can also mark to doctor the breast image execution multiparameter lesion segmentation of ROI, and choose best candidate's tumor's profiles as final segmentation result according to the feature marking of each candidate's tumor's profiles being partitioned into, improve the stability of the result of lesion segmentation, made segmentation result not be subject to one or partial parameters improper use or indivedual wrong impact.
Although represent with reference to preferred embodiment and described the present invention, it should be appreciated by those skilled in the art that in the case of not departing from the spirit and scope of the present invention that are defined by the claims, can carry out various modifications and conversion to these embodiment.
Claims (13)
1. for an image processing system for breast image, comprising:
Image collection device, for obtaining two-dimentional breast ultrasound ripple image;
Lesion detection device, detects M region-of-interest that comprises tumor of breast for the two-dimentional breast ultrasound ripple image obtaining from image collection device, and is that each region-of-interest marking detects mark as it, wherein, and M > 0;
Multiparameter dispenser, for using respectively K lesion segmentation algorithm of parameter based on different or parameter combinations from each K of being partitioned into candidate's tumor's profiles of a described M region-of-interest, and be the detection mark that each candidate's tumor's profiles record is partitioned into the region-of-interest of described candidate's tumor's profiles, wherein, K > 0;
Feature scoring device, for evaluating marking as its feature scores according at least one predetermined feature to the each of the M*K being partitioned into candidate's tumor's profiles;
Fusion device for according to detection mark and the feature scores of described M*K candidate's tumor's profiles, selects candidate's tumor's profiles as the tumor's profiles being finally partitioned in the middle of described M*K candidate's tumor's profiles.
2. image processing system as claimed in claim 1, also comprises:
Combiner, for the M*K being partitioned into candidate's tumor's profiles calculated respectively to similarity each other, and not the highest candidate's tumor's profiles from every group of similarity higher than removing detection mark in the middle of multiple candidate's tumor's profiles of predetermined value, thereby obtain N candidate's tumor's profiles, wherein, N < < M*K;
Wherein, feature scoring device is given a mark as its feature scores to each evaluation the in described N candidate's tumor's profiles according at least one predetermined feature, fusion device, according to detection mark and the feature scores of described N candidate's tumor's profiles, is selected candidate's tumor's profiles.
3. image processing system as claimed in claim 1, also comprises:
Pretreater, for the two-dimentional breast ultrasound ripple image obtaining is carried out to pre-service, described pre-service comprises two-dimentional breast ultrasound ripple image execution Denoising disposal and/or the image enhancement processing to obtaining;
Wherein, multiparameter dispenser is to carrying out described lesion segmentation processing through pretreated two-dimentional breast ultrasound ripple image.
4. image processing system as claimed in claim 1, wherein, lesion detection device uses deformations bit model method, template matching method or Adaboost method to carry out detection and the marking of described M region-of-interest that comprises tumor of breast.
5. image processing system as claimed in claim 1, wherein, described different parameter is different iterationses, different ratio or diverse ways, and described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
6. image processing system as claimed in claim 4, wherein, feature scoring device use support vector regression method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.
7. image processing system as claimed in claim 1, wherein, fusion device is normalized detection mark and the feature scores of each candidate's tumor's profiles respectively, calculates the comprehensive mark of each candidate's tumor's profiles according to following formula, and selects candidate's tumor's profiles that comprehensive mark is the highest:
Score
combined＝w
ds×NDS+w
rs×NRS
Wherein, Score
combinedthe comprehensive mark of candidate's tumor's profiles, w
dsand w
rsbe respectively to give the detection mark of candidate's tumor's profiles and the weights of feature scores, NDS and NRS are respectively detection mark and the feature scores of candidate's tumor's profiles.
8. image processing system as claimed in claim 7, wherein, fusion device uses s-score normalization algorithm, min-max normalization algorithm, Tanh estimator or double sigmoid algorithm to carry out described normalized.
9. image processing system as claimed in claim 4,
Wherein, feature scoring device is used support vector regression method, the Jaccard index that calculates each candidate's tumor's profiles according to the detection mark of at least one feature in following characteristics and candidate's tumor's profiles is as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature
Wherein, fusion device is from selecting in the middle of described M*K candidate's tumor's profiles candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
10. image processing system as claimed in claim 3, wherein, described different parameter is different iterations, different ratio, different step, different Denoising disposal or different image enchancing methods, and described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
11. 1 kinds of image processing systems for breast image, comprising:
Image collection device, marks for obtaining the two-dimentional breast ultrasound ripple image that has M tumour region-of-interest, wherein, and M > 0;
Multiparameter dispenser, for using respectively K lesion segmentation algorithm of parameter based on different or parameter combinations from each K of being partitioned into candidate's tumor's profiles of a described M region-of-interest, wherein, K > 0;
Feature scoring device, for according at least one predetermined feature, each evaluation of the M*K being partitioned into candidate's tumor's profiles being given a mark as its feature scores, and from described M*K candidate's tumor's profiles, select candidate's tumor's profiles that feature scores is the highest as the tumor's profiles being finally partitioned into.
12. image processing systems as claimed in claim 11, wherein, described different parameter is different iterations, different ratio or different steps, and described lesion segmentation approach is Level Set Method, figure cutting method, region growing method or watershed algorithm.
13. image processing systems as claimed in claim 11, wherein, feature scoring device use support vector regression method according to the Jaccard index of the each candidate's tumor's profiles of at least one feature calculation in following characteristics as its feature scores: textural characteristics, space characteristics, strength characteristic, contour feature or tumour feature.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210411592.7A CN103778600B (en) | 2012-10-25 | 2012-10-25 | Image processing system |
| KR1020130021930A KR102128325B1 (en) | 2012-10-25 | 2013-02-28 | Image Processing System |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201210411592.7A CN103778600B (en) | 2012-10-25 | 2012-10-25 | Image processing system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103778600A true CN103778600A (en) | 2014-05-07 |
| CN103778600B CN103778600B (en) | 2019-02-19 |
Family
ID=50570799
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201210411592.7A Active CN103778600B (en) | 2012-10-25 | 2012-10-25 | Image processing system |
Country Status (2)
| Country | Link |
|---|---|
| KR (1) | KR102128325B1 (en) |
| CN (1) | CN103778600B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105488800A (en) * | 2015-11-30 | 2016-04-13 | 上海联影医疗科技有限公司 | Feature extracting method and computer-aided diagnosis method and system |
| CN106683137A (en) * | 2017-01-11 | 2017-05-17 | 中国矿业大学 | Monocular multi-target identification and positioning method based on artificial mark |
| CN107636725A (en) * | 2015-05-18 | 2018-01-26 | 皇家飞利浦有限公司 | The image partition method and system known |
| US10467757B2 (en) | 2015-11-30 | 2019-11-05 | Shanghai United Imaging Healthcare Co., Ltd. | System and method for computer aided diagnosis |
| CN110569837A (en) * | 2018-08-31 | 2019-12-13 | 阿里巴巴集团控股有限公司 | Method and device for optimizing damage detection result |
| CN111554383A (en) * | 2020-04-24 | 2020-08-18 | 浙江杜比医疗科技有限公司 | Neural network for breast tumor detection and detection system thereof |
| WO2021087687A1 (en) * | 2019-11-04 | 2021-05-14 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic image analyzing method, ultrasonic imaging system and computer storage medium |
Families Citing this family (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| KR102315574B1 (en) * | 2014-12-03 | 2021-10-20 | 삼성전자주식회사 | Apparatus and method for classification of data, apparatus and method for segmentation of region of interest |
| KR102333479B1 (en) | 2015-03-31 | 2021-12-03 | 한국전자통신연구원 | Method and apparatus for breast lesion diagnosis |
| KR102667231B1 (en) * | 2020-12-21 | 2024-05-20 | 주식회사 인피니트헬스케어 | System and method for contouring a set of medical images based on deep learning algorighm and anatomical characteristics |
| CN115089292A (en) * | 2022-06-09 | 2022-09-23 | 北京三春晖医疗器械有限公司 | Needle distribution control system and method and tumor therapeutic apparatus |
| CN115641430B (en) * | 2022-10-12 | 2024-01-30 | 阿里巴巴（中国）有限公司 | Method, device, medium and computer equipment for determining interest surface |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070255095A1 (en) * | 2006-03-31 | 2007-11-01 | Gilreath Mark G | System and method for assessing a patient condition |
| CN101887514A (en) * | 2009-05-13 | 2010-11-17 | 联咏科技股份有限公司 | Human face detecting device and human face detecting method thereof |
| CN102113335A (en) * | 2009-06-09 | 2011-06-29 | 索尼公司 | Image processing apparatus, method and program |
| CN102467741A (en) * | 2010-11-18 | 2012-05-23 | 索尼公司 | Method and device for detecting spot area in image |
| CN102479327A (en) * | 2010-11-23 | 2012-05-30 | 现代摩比斯株式会社 | Object recognition system, sysetm for recognizing obstacle for a vehicle and method thereof |
Family Cites Families (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1998035501A2 (en) * | 1997-02-06 | 1998-08-13 | Koninklijke Philips Electronics N.V. | Image segmentation and object tracking method and corresponding system |
| US20050129297A1 (en) | 2003-12-15 | 2005-06-16 | Kamath Vidya P. | Classification of breast lesion method and system |
| KR100863745B1 (en) * | 2005-08-18 | 2008-10-16 | 주식회사 메디슨 | Apparatus and method for processing an ultrasound image in an ultrasound image diagnostic system |
| US8218850B2 (en) | 2006-02-10 | 2012-07-10 | Synarc Inc. | Breast tissue density measure |
| GB0602739D0 (en) | 2006-02-10 | 2006-03-22 | Ccbr As | Breast tissue density measure |
| US20100158332A1 (en) | 2008-12-22 | 2010-06-24 | Dan Rico | Method and system of automated detection of lesions in medical images |
| KR101439412B1 (en) * | 2010-11-22 | 2014-09-11 | 한국전자통신연구원 | Method of segmenting lesions in images |
-
2012
- 2012-10-25 CN CN201210411592.7A patent/CN103778600B/en active Active
-
2013
- 2013-02-28 KR KR1020130021930A patent/KR102128325B1/en active IP Right Grant
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070255095A1 (en) * | 2006-03-31 | 2007-11-01 | Gilreath Mark G | System and method for assessing a patient condition |
| CN101887514A (en) * | 2009-05-13 | 2010-11-17 | 联咏科技股份有限公司 | Human face detecting device and human face detecting method thereof |
| CN102113335A (en) * | 2009-06-09 | 2011-06-29 | 索尼公司 | Image processing apparatus, method and program |
| CN102467741A (en) * | 2010-11-18 | 2012-05-23 | 索尼公司 | Method and device for detecting spot area in image |
| CN102479327A (en) * | 2010-11-23 | 2012-05-30 | 现代摩比斯株式会社 | Object recognition system, sysetm for recognizing obstacle for a vehicle and method thereof |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107636725A (en) * | 2015-05-18 | 2018-01-26 | 皇家飞利浦有限公司 | The image partition method and system known |
| CN105488800A (en) * | 2015-11-30 | 2016-04-13 | 上海联影医疗科技有限公司 | Feature extracting method and computer-aided diagnosis method and system |
| 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 |
| CN106683137A (en) * | 2017-01-11 | 2017-05-17 | 中国矿业大学 | Monocular multi-target identification and positioning method based on artificial mark |
| CN106683137B (en) * | 2017-01-11 | 2019-12-31 | 中国矿业大学 | Artificial mark based monocular and multiobjective identification and positioning method |
| CN110569837A (en) * | 2018-08-31 | 2019-12-13 | 阿里巴巴集团控股有限公司 | Method and device for optimizing damage detection result |
| WO2021087687A1 (en) * | 2019-11-04 | 2021-05-14 | 深圳迈瑞生物医疗电子股份有限公司 | Ultrasonic image analyzing method, ultrasonic imaging system and computer storage medium |
| CN111554383A (en) * | 2020-04-24 | 2020-08-18 | 浙江杜比医疗科技有限公司 | Neural network for breast tumor detection and detection system thereof |
| CN111554383B (en) * | 2020-04-24 | 2023-09-05 | 浙江杜比医疗科技有限公司 | Neural network for breast tumor detection and detection system thereof |
Also Published As
| Publication number | Publication date |
|---|---|
| KR20140057130A (en) | 2014-05-12 |
| KR102128325B1 (en) | 2020-07-01 |
| CN103778600B (en) | 2019-02-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103778600A (en) | Image processing system | |
| US11651489B2 (en) | System and method for diagnostic vector classification support | |
| CN107480677B (en) | Method and device for identifying interest region in three-dimensional CT image | |
| JP6059261B2 (en) | Intelligent landmark selection to improve registration accuracy in multimodal image integration | |
| JP4999163B2 (en) | Image processing method, apparatus, and program | |
| US9943286B2 (en) | Ultrasonographic images processing | |
| Kim et al. | Machine-learning-based automatic identification of fetal abdominal circumference from ultrasound images | |
| TWI552013B (en) | Method, apparatus and coumputer program product of classifying suspicious region of interest in ultrasonic attenuation image mapping tissue of patient | |
| US9042611B2 (en) | Automated vascular region separation in medical imaging | |
| US9324155B2 (en) | Systems and methods for determining parameters for image analysis | |
| RU2636262C2 (en) | Method and system for processing ultrasound imaging data | |
| Chlebus et al. | Deep learning based automatic liver tumor segmentation in CT with shape-based post-processing | |
| US20070036406A1 (en) | System and method for toboggan-based object detection in cutting planes | |
| Lal et al. | Automatic segmentation of tumors in B-Mode breast ultrasound images using information gain based neutrosophic clustering | |
| CN112568933B (en) | Ultrasonic imaging method, apparatus and storage medium | |
| CN104021391B (en) | Handle the method and apparatus and breast cancer diagnosis apparatus of ultrasound image | |
| US7853062B2 (en) | System and method for polyp detection in tagged or non-tagged stool images | |
| Zhou et al. | Automatic organ localizations on 3D CT images by using majority-voting of multiple 2D detections based on local binary patterns and Haar-like features | |
| Zhou et al. | Variabilities in reference standard by radiologists and performance assessment in detection of pulmonary embolism in CT pulmonary angiography | |
| CN111383323B (en) | Ultrasonic imaging method and system and ultrasonic image processing method and system | |
| JP2024507290A (en) | Method and system for breast ultrasonic image diagnosis using weakly-supervised deep learning artificial intelligence ial intelligence} | |
| Pereira et al. | A 3D Lung Nodule Candidate Detection by Grouping DCNN 2D Candidates. | |
| CN111986165B (en) | Calcification detection method and device in breast image | |
| Wijata et al. | Machine Learning Detects a Biopsy Needle in Ultrasound Images | |
| Belhaouair et al. | On a new approach to the automated detection of thoracic organs diseases using the spot feature in the analysis of digital X-ray images |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |