CN104657984A - Automatic extraction method of three-dimensional breast full-volume image regions of interest - Google Patents

Automatic extraction method of three-dimensional breast full-volume image regions of interest Download PDF

Info

Publication number
CN104657984A
CN104657984A CN201510044837.0A CN201510044837A CN104657984A CN 104657984 A CN104657984 A CN 104657984A CN 201510044837 A CN201510044837 A CN 201510044837A CN 104657984 A CN104657984 A CN 104657984A
Authority
CN
China
Prior art keywords
image
region
interest
area
tumor
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
Application number
CN201510044837.0A
Other languages
Chinese (zh)
Other versions
CN104657984B (en
Inventor
汪源源
王欣
郭翌
余锦华
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.)
Fudan University
Original Assignee
Fudan University
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 Fudan University filed Critical Fudan University
Priority to CN201510044837.0A priority Critical patent/CN104657984B/en
Publication of CN104657984A publication Critical patent/CN104657984A/en
Application granted granted Critical
Publication of CN104657984B publication Critical patent/CN104657984B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

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
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/08—Detecting organic movements or changes, e.g. tumours, cysts, swellings
    • A61B8/0825—Detecting organic movements or changes, e.g. tumours, cysts, swellings for diagnosis of the breast, e.g. mammography
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/08—Detecting organic movements or changes, e.g. tumours, cysts, swellings
    • A61B8/0833—Detecting organic movements or changes, e.g. tumours, cysts, swellings involving detecting or locating foreign bodies or organic structures
    • A61B8/085—Detecting organic movements or changes, e.g. tumours, cysts, swellings involving detecting or locating foreign bodies or organic structures for locating body or organic structures, e.g. tumours, calculi, blood vessels, nodules
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B8/00—Diagnosis using ultrasonic, sonic or infrasonic waves
    • A61B8/52—Devices using data or image processing specially adapted for diagnosis using ultrasonic, sonic or infrasonic waves
    • 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
    • G06T2207/10132—Ultrasound image
    • G06T2207/10136—3D ultrasound image
    • 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
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30004—Biomedical image processing
    • G06T2207/30068—Mammography; Breast

Landscapes

  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Engineering & Computer Science (AREA)
  • Medical Informatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Physics & Mathematics (AREA)
  • Radiology & Medical Imaging (AREA)
  • Surgery (AREA)
  • Public Health (AREA)
  • Biomedical Technology (AREA)
  • Molecular Biology (AREA)
  • Pathology (AREA)
  • Animal Behavior & Ethology (AREA)
  • Biophysics (AREA)
  • Heart & Thoracic Surgery (AREA)
  • Veterinary Medicine (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Quality & Reliability (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Vascular Medicine (AREA)
  • Ultra Sonic Daignosis Equipment (AREA)
  • Image Processing (AREA)

Abstract

The invention belongs to the field of image processing, and particularly relates to an automatic extraction method of regions of interest in three-dimensional breast full-volume images (ABVS). The method comprises the following steps: processing the continuous cross section two-dimensional images in three-dimensional ABVS images by using a maximum direction-based phase information method to obtain the candidate regions of interest on each cross section image; removing the unrelated regions according to the prior knowledge such as the continuity and position characteristic of breast tumor on the two-dimensional cross section images; obtaining the shape and texture features of the residual suspected tumor regions, inputting the shape and texture shapes to a two-valued logistic regression classifier to obtain the probability of each region becoming tumor and selecting the region with the maximum probability as the tumor region; obtaining the minimum ellipsoid comprising the region of interest according to the selected region to serve as the region of interest. The automatic extraction method provided by the invention can be used for realizing the automatic extraction of tumor regions of interest in the three-dimensional ABVS images, obtaining the correct positions of tumor, decreasing the workload of the manual operation and providing important reference to further tumor detection.

Description

The extraction method of three-D ultrasonic mammary gland total volume interesting image regions
Technical field
The invention belongs to technical field of image processing, be specially the extraction method of three-D ultrasonic mammary gland total volume interesting image regions.
Background technology
The advantages such as ultrasonic imaging is created because having nothing, real-time, repeatability is strong, low cost, have important application clinically.Compared with the two-dimensional ultrasonic imaging of traditional hand-held, ABVS has brand-new imaging pattern, can standardization autoscan mammary gland, carries out digitized processing, avoid the individual difference of user to image; ABVS can carry out full milk scanning, and comparatively conventional Ultrasound adds the coronal section of reconstruction, thus can provide information more more than two dimensional image, has good repeatability.
Because the volume of tumour is relative to less whole ABVS image, the segmentation accuracy rate of directly carrying out tumour is low.Therefore, user is usually needed in hundreds of width cross-sectional image, manually to mark the center of tumour or interested region, further to analyze.The method of this artificial demarcation is very consuming time, and depends on the experience of user.
For this problem, the present invention proposes a kind of method of full-automatic extraction ABVS area-of-interest.The method does not need user's advance demand flag tumour, can the interested region of Automatic-searching, and finally obtains the minimum ellipsoid comprising area-of-interest.Method of the present invention is applied in the automatic analysis system of ABVS image, the accuracy of whole system can be improved.
Summary of the invention
The object of the invention is the method for the area-of-interest proposed in a kind of automatic extraction three-D ultrasonic mammary gland total volume image.
The present invention proposes a kind of extraction method of three-D ultrasonic mammary gland total volume interesting image regions, and its concrete steps are:
1. first the ABVS image of DICOM form is carried out image reconstruction according to the distance between the pixel on three-dimensional, make it corresponding with actual galactophore image size; Image sheet, 750 sagittal image sheets of 820 width transversal section are obtained after rebuilding; According to the difference of scan depths, obtain the image sheet of 98 ~ 294 width coronal-planes; On each tangent plane, the distance between adjacent two width images is 0.2 mm, and the distance between the neighbor pixel on every width tangent plane picture is also 0.2 mm;
2. in coronal image, be generally oval feature according to mammary gland, calculate the minimum value map image of the front ten width coronal image sheets of ABVS image after rebuilding, after its threshold process, obtain the template of mammary gland on coronal-plane with Hough elliptic transformation, and be applied to all coronal image sheets; Utilize the relation between ABVS image coronal-plane and transversal section three-dimensional coordinate, mammary gland position on coronal-plane is projected to transversal section, determine mammary gland position in transversal section, tentatively reduce the hunting zone of transversal section area-of-interest, remove the impact of the interference such as mammary gland external noise, artifact;
3. 820 width images of pair transversal section, merge into the new image of a width every ten width images by minimum value mapping, obtain 82 width images after merging; Wherein every width image is adopted based on maximum direction phase method, in conjunction with knub position characteristic and texture feature, extracts candidate region interested, obtain the bianry image of respective regions;
4., by step 3, obtain the bianry image of candidate region interested, a series of continuous transversal section; According to the continuity of tumour, if the connected region of certain a slice and front and back adjacent sheet zero lap, be then called irrelevant sheet, first removed; To remaining sheet, divide into groups according to its continuity and connected region center, often group is expressed as a suspected tumor;
5. each group of cross-sectional view photo pair processed through step 4 is classified, extract the shape of its bianry image and corresponding grey scale image, Texture eigenvalue, input to logistic regression sorter, obtaining often group may be the probability of tumour, the cross-sectional view photo of maximum probability person corresponding to real tumor region;
6., in the bianry image of the tumor region serialgram determined in step 5, find maximum connected region, determine the minimum ellipse comprising this connected region; According to the coordinate corresponding relation between sagittal plane and cross-sectional image, find the sagittal view photo that elliptical center horizontal ordinate is corresponding, adopt the method extracting candidate region interested in step 3 to obtain the candidate region interested of sagittal view photo, and then determine minimum ellipse sagittal plane comprising candidate region interested; Utilize two ellipses of sagittal plane and transversal section, determine that ABVS image comprises three main shafts of the minimum ellipsoid of area-of-interest long, thus obtain the area-of-interest of tumour.
The relevant technical details related to regard to each step of the inventive method is below further described specifically.
Be scan along mammary gland cross-sectional direction the sequential chart photo that obtains by the automatic total volume Image-forming instrument of mammary gland to rebuild and obtain about step 1. ABVS image, picture format is generally DICOM form, has the plane that three orthogonal: transversal section, sagittal plane and coronal-plane; The two-dimensional ultrasonic image that ultrasonoscopy in transversal section and sagittal plane and traditional ultrasonic hand-held instrument obtain is more close, and the ultrasonoscopy on coronal-plane can observe the general profile of mammary gland; Comprise the full detail of image in DICOM file, comprise the gray-scale value of image slices vegetarian refreshments and the pixel interval of image.Therefore, can by the reading of DICOM fileinfo, the ABVS image of reconstruction of three-dimensional.The scan depths of ABVS image is generally at 20 ~ 60 mm, and according to the difference of scan depths, the size of the ABVS image after reconstruction is (98-294) × 750 × 820.As shown in Figure 1, after rebuilding, the image of transversal section, sagittal plane and coronal-plane as shown in Figure 2 for three tangent plane schematic diagram of original ABVS image.
According to the relation of the distance between the ABVS image slices vegetarian refreshments read from DICOM file with actual galactophore image, rebuild ABVS image, its neighbor dot spacing is 0.2 mm.
About the template that this step of step 2. is to obtain mammary gland, get rid of the interference such as artifact, noise because of reason generations such as loose contacts around mammary gland.Its implementation method is: first get the front 10 width image sheets of coronal-plane representing skin of mammary gland top layer, calculate the minimum value map image of front 10 width coronal image sheets, then the principle that this ten width image maps by minimum value merged.Image after being combined adopts the threshold process of OSTU [1]method obtains bianry image; Adopt morphologic dilation erosion method to obtain the general profile of mammary gland, then detect the position of the ellipse finding mammary gland corresponding by the method for Hough elliptic transformation, generate a width two-value template image, oval inner value is 1, and oval outside value is 0; For ABVS image, the horizontal ordinate of the corresponding cross-sectional image of ordinate of its coronal image, therefore, the Prototype drawing that coronal image can be utilized to generate, determines mammary region on cross-sectional image, removes the impact of the noise outside mammary gland, artifact interference.
In order to reduce the operand of Hough elliptic transformation, according to priori, the major and minor axis of ellipse, direction and ratio of semi-minor axis length are limited.Here limit oval main axis length between 400 ~ 800, ratio of semi-minor axis length is greater than 0.75, transverse with xthe angle of axle positive dirction is between (0, π).Oval equation is:
(1)
Wherein arepresent transverse, brepresent oval minor axis, θrepresent transverse with xthe angle of axle positive dirction, ( x 0, y 0) be oval center.
After ellipse coronal-plane comprising mammary gland being detected, utilize the relation between ABVS image coronal-plane and transversal section three-dimensional coordinate, mammary gland position on coronal-plane is projected to transversal section, determine mammary gland position in transversal section, tentatively reduce the hunting zone of candidate region interested, transversal section, remove the impact of the interference such as mammary gland external noise, artifact.The result that this segment template extracts as shown in Figure 3.Utilize the template action of coronal-plane in view picture ABVS image, the image of three tangent planes obtained as shown in Figure 4.
Object about this step of step 3. extracts the area-of-interest of cross-sectional view photo.First, every for the consecutive image of transversal section ten width are carried out minimum value and map merging, reduce operand.Then, the every width cross-sectional image after being combined carries out the extraction of candidate region interested.
The leaching process of two dimension breast ultrasound tumor image candidate region interested is:
1) Image semantic classification.
A) speckle noise of ultrasonoscopy is relatively more serious, so first carry out anisotropy elimination of spot noise (SRAD).For the image shown in Fig. 5 (a), filtered result is as shown in Fig. 5 (b).
B) scope of gray-scale value is reduced [2], utilize linear normalization formula (2) to adjust the gray-scale value of pixel in image, result is as shown in Fig. 5 (c).
(2)
Wherein l n for number of greyscale levels, lboundwith uboundbe taken as respectively q(0.5) and q(0.95), qit is the Quantile Function of histogram cumulative distribution.
C) low echo area territory is strengthened [3], the self-adaptation Z-type equation adopting formula (3) to represent carries out greyscale transformation:
(3)
Wherein z a , z b , z c determine the shape of Z-type function. z a with z c determine the scope of curve nonlinear transformation, z a be traditionally arranged to be 20, z c for the average of image, and z b then determine the inclined degree of curve, the gradient according to intensity profile obtains:
(4)
(5)
If the image obtained after pre-service is , as shown in Fig. 5 (d).
2) ask for maximum direction phase diagram pMO.According to the method mentioned in document [4], along six direction (0 °, 30 °, 60 °, 90 °, 120 °, 150 °), filtering is carried out to image at the frequency domain of image, finally extract the phase information with ceiling capacity direction.This six direction covers whole frequency spectrum (0 ~ 360 °).
To image i, in order to calculate PMO matrix, first need two dimension log-gaborwave filter is to image filtering, two-dimentional here log-gaborthe transport function of wave filter is the Gaussian function under logarithmic scale, is decomposed into two parts: radial wave filter and angular filter.For each direction θ 0, the wave filter of structure is:
(6)
Wherein in braces, front portion represents radial wave filter, and rear portion is angular filter. ωfor radial coordinate; θfor angle coordinate; ω 0for filter centre frequency, arrange the centre frequency of four different scales, its value is respectively 1/3,1/ (3 × 1.7), 1/ (3 × 1.7 2), 1/ (3 × 1.7 - 3); σ θ determine with θcentered by angular bandwidth, be set to 30 °; determine radial bandwidth, be usually set to 0.55, while ensureing filter effect, avoid aliasing.
Image through different center frequency, 24 altogether of different angles log-gaborafter filter filtering, 24 image arrays are changed in the inversion through Fast Fourier Transform (FFT) (FFT) ( xfor direction, srepresent centre frequency yardstick), then calculate the phasing matrix in each angle lPA:
(7)
Wherein represent in angle θtime phasing matrix; nrepresent the quantity of different center frequency wave filter; for yardstick sunder wild phase bit matrix, its calculating formula is:
(8)
(9)
Wherein imag( x, s) be imaginary part, real( x, s) be real part.
After six angle filtering, obtain six lPAeigenmatrix.Then, these six matrixes are merged into an eigenmatrix.Here, extract certain a bit on there is ceiling capacity angle phase characteristic value as the final phase characteristic value of this point, thus obtain a new phase characteristic matrix.Because the energy of local features structural information, so the angle with maximum local energy is closest to boundary direction.So, phase characteristic pMOeach pixel is defined as:
(10)
Wherein represent in angle θon energy matrix, its calculating formula is:
(11)
Here imag( x, s) and real( x, s) represent that image passes through respectively log-gaborimaginary part after wave filter and real part. ρthe angle with maximum local energy obtained by formula (5), represent ρthe superposition of the phasing matrix of four yardstick centre frequencies in angle.
Through process above, image can be obtained imaximum direction phase diagram pMO.To the pretreated image such as shown in Fig. 6 (a), its maximum direction phase diagram is as shown in Fig. 6 (b).Then, in order to outstanding tumor region, remove ground unrest, will pMOwith 256- be multiplied, adopt the median filter of 5 × 5 to carry out filtering to the image after being multiplied, result is as shown in Fig. 6 (c).But, after the phase extraction of maximum direction, pMOmost of grey scale pixel value scope be [0,0.5], therefore entire image seems fuzzyyer gloomy.Two formula are below adopted to carry out gray correction:
(12)
(13)
To sum up, the maximum direction phase diagram of gray level image is obtained.In order to the region representation tumor region making wherein gray-scale value large, to the phase diagram negate of maximum direction, last gained image is as shown in Fig. 6 (d).
3) because tumor type is varied, the texture of its ultrasonoscopy and shape difference are also larger, and some inside tumor gray scale is uneven, but there is continuous print phase place at the edge of tumour; Some inside tumor gray scale is comparatively even, but the phase place at edge is not obvious or with the phase place aliasing in background.Therefore, in the choosing of candidate region interested, the information in conjunction with gray level image and maximum direction phase diagram is needed.With OSTU threshold method to pretreated image i, normalized pMOwith pMO+Ithree width images carry out Threshold segmentation process respectively, obtain the bianry image containing a lot of connected regions.Remove the noise region that in image, area is less than 300 pixels and is greater than 50% with the connected ratio at edge.Calculate the energy function of remaining all connected regions e:
(14)
Wherein compactrepresent the degree of compacting of connected region, its value is the area of connected region and the area ratio of its Least Chimb shape; areafor the area of connected region; centdisfor the distance of connected region center and picture centre; eccentricityfor the eccentricity of the minimum external ellipse of connected region; for region and coincident rate, its value is region accounts for the total quantity of image edge pixels point ratio with the quantity of the pixel of coincident. w 1 ~ w 4represent that the degree of compacting in region, area, centre distance and eccentricity affect size to region energy functional value respectively, generally getting it is 1, also can carry out certain adjustment according to concrete supersonic tumor image data base, and after adjustment, scope is 1 ± 0.5. wfor the weight of coincident rate, its span is generally (0,0.5).
Through said process, obtain pretreated image respectively i, normalized pMOwith pMO+Ithe candidate region interested of three width images.Calculate the energy function of three width interesting image candidate regions e, select emaximum region is the final candidate region interested of image.Image i, normalized pMOwith pMO+Ithree width images and Threshold segmentation thereof and regioselective result are as shown in Figure 7.
About step 4. through previous step, obtain the bianry image of the candidate region interested of a series of continuous transversal section.Tumour is continuous print at the image sheet of transversal section, and in the bianry image of its correspondence, connected region has overlap, as shown in Figure 8.In fig. 8, the candidate region interested of adjacent two width image sheets overlaps each other, and the area of candidate region interested first changes from small to big, more from large to small.Bianry image connected region and the non-overlapping unrelated images sheet of front and back abutment flange can be removed according to the continuity of tumour.Judge a certain image sheet be whether unrelated images sheet according to being:
(15)
Wherein bW k represent the kthe bianry image of width cross-sectional image, mwith nbe respectively the pixel number that image is axial and horizontal.When r( k) <50 time, think kthe candidate region interested of width image and adjacent sheet zero lap, remove the in image collection kwidth image.As an example, Fig. 9 gives the continuous three width cross-sectional images of ABVS image and the bianry image of corresponding candidate region interested thereof, wherein the candidate region interested of the second width image and front and back two width image all zero laps, therefore remove it from the image set of this ABVS.
To remaining image sheet, first according to the continuity of tumor region in cross-sectional view photo, continuous print image sheet is divided into one group; Then often organizing in continuous print image sheet, the distance between the center, candidate region interested of calculating adjacent sheet, is classified as one group by the image sheet of distance within 50 pixels.Like this, the final continuous cross-section slices dividing the often group expression suspected tumor obtained.In these groups except real tumor region, further comprises nipple rear shade and low echo area territory that some cause because of reasons such as probe contacts are bad.
Object about this step of step 5. extracts the shape and textural characteristics of often organizing bianry image corresponding to continuous two dimensional image and gray level image, and utilize these features to classify.
By calculating shape and the textural characteristics that can obtain every width image sheet, but because often comprise the image sheet that quantity do not wait in group, each feature that every picture group photo obtains has the multiple values corresponding with image sheet quantity, therefore need to conclude it, obtain one and can be used for describing the eigenwert often organizing image global feature.
1) shape facility:
The aspect ratio of the candidate region interested in group in cross-sectional view photo: in general the aspect ratio of tumour is less than 1, and the aspect ratio of such as shadow region, mammary gland rear, some non-tumor areas is greater than 1, therefore can using the aspect ratio of area-of-interest as the feature distinguishing tumour and non-tumour.This parameter uses the axially most major diameter of tumour usually l short with horizontal most major diameter l long ratio represent:
(16)
The aspect ratio of every width interesting image regions in calculating group m abr , get the aspect ratio of its intermediate value as this group.
The Duplication of area-of-interest between sequential chart photo: can be seen by Fig. 8, the area of the area-of-interest of the continuous cross-sectional view photo of tumour changes from small to big, again from large to small, in general, the image sheet being positioned at centre position has the maximum area-of-interest of area.
The area-of-interest Duplication defining adjacent two width images is:
(17)
Wherein bW k in expression group kthe bianry image of width image.Duplication is defined as by formula (17) kwidth image and kthe overlapping area of+1 width interesting image candidate region and kthe area ratio of+1 width interesting image candidate region.Like this, one is contained lthe suspected tumor image sets of width image, we can obtain l-1 Duplication.In order to this l-1 feature is concluded, and asks its average, standard deviation, sum of products gradient mean value respectively, as the feature of classification.
2) textural characteristics:
Textural characteristics comprises gray level co-occurrence matrixes [5]characteristic sum gray feature.
Gray level co-occurrence matrixes [5]feature comprises entropy, contrast, the degree of correlation, energy and homogeney.Easy in order to what calculate, first ask the minimum enclosed rectangle of candidate region interested in every width bianry image, then by minimum enclosed rectangle respectively longitudinally with laterally to external expansion 20 pixels, obtain the rectangle comprising candidate region interested.Calculate the textural characteristics being positioned at the gray level image of this rectangle, the average of the gray level co-occurrence matrixes feature of all image sheets in then calculating group, as the feature for classifying.
Gray feature comprises: a) gray average of candidate region interested and standard deviation ; B) ratio of candidate region interested and the gray average of background area in the rectangle comprising candidate region interested with the ratio of standard deviation , in general, the gray average of area-of-interest inside is lower than background area gray average, and gray standard deviation is also lower than surrounding background area gray standard deviation; C) area-of-interest forward-and-rearward sound shadow feature meAP, in order to distinguish the low echo area territories such as nipple rear shade, area-of-interest forward-and-rearward sound shadow feature meAPcalculating formula be:
(18)
Wherein have:
(19)
(20)
(21)
Here mwith nbe respectively the pixel number that image is axial and horizontal, i h with bW h represent in 82 width ABVS cross-sectional view photos respectively hthe bianry image of width image and its candidate region interested, mBWrepresent the candidate region interested that area is maximum, meAfor mBWin suspected tumor front gray average corresponding to candidate region interested, mePfor mBWin suspected tumor rear gray average corresponding to candidate region interested, mefor the gray average of all candidate regions interested of suspected tumor.
To sum up, the final feature extracted has five shape facilities, ten textural characteristics.These features are inputed to logistic regression sorter classify, often being organized suspected tumor region may be the probability of tumour, and select probability the maximum is tumour.In order to the validity of verification method, adopt the method for ten cross validations.
Object about this step of step 6. is the bianry image of serialgram and the gray level image of tumor region by obtaining, determines that three main shafts of the ellipsoid comprising area-of-interest are long, thus obtains the area-of-interest of spheroid-like.
First, find the candidate region interested that the area that is positioned at tumor region centre position is maximum, with morphologic plavini by candidate region interested to external expansion 10 pixels, calculate the minimum external ellipse in the region after expansion, oval major axis length is a, minor axis length is b.As shown in Figure 1, the sequence number of the corresponding sagittal view photo of transverse axis coordinate on transversal section, if the center point coordinate of the external ellipse of area-of-interest be ( x 0, y 0), then get x 0sheet sagittal view photo, at this sagittal view picture yaxial coordinate ( y 0- b/ 2, y 0+ b/ 2) the above-mentioned candidate region interested extracting method based on maximum direction phase place is utilized to obtain sagittal plane candidate region interested in region.Calculate this candidate region interested to the external ellipse of external expansion 10 pixel rear regions and major axis long c.Like this, three main shafts just obtaining the ellipsoid comprising area-of-interest are long a, bwith c.
The equation of ellipsoid is:
(22)
Wherein ( x 0, y 0, z 0) be the coordinate of ellipsoid central point, a, b, cbe respectively ellipsoid axial length in three orthogonal directions.As shown in Figure 10, the area-of-interest that actual ABVS image finally obtains as shown in figure 11 for ABVS image and elliposoidal area-of-interest schematic diagram thereof.
Accompanying drawing explanation
Fig. 1. ABVS image transversal section, sagittal plane and coronal-plane schematic diagram, wherein Horizontal plane represents schematic cross-sectional view, and Sagittal plane represents sagittal plane schematic diagram, and Coronal plane represents coronal-plane schematic diagram.
Fig. 2. after rebuilding ABVS image in transversal section, the view of sagittal plane and coronal-plane.Tumor of breast is comprised in square wire frame.
Fig. 3. coronal-plane template obtains, the template for extracting in oval wire frame.
Fig. 4. applying oval template restriction area-of-interest hunting zone: three tangent planes of (a) original image, is the figure of each tangent plane after the oval template of employing in wire frame; B () applies three tangent planes after oval template.
Fig. 5. cross-sectional image pre-service.Wherein, (a) original image; (b) SRAD filtered image; C image that () tonal range reduces; D () low echo area territory strengthens image.
Fig. 6. cross-sectional image maximum direction phase diagram.Wherein, (a) original image; (b) pMOimage; (c) pMO× (256- i) image after medium filtering; Image after the adjustment of (d) gray scale.
Fig. 7. the extraction of area-of-interest.Wherein, (a), (b), (c) are respectively i, pMO, i+PMO; D (), (e), (f) are respectively the OSTU bianry image of three width images; G (), (h), (i) are respectively the candidate region interested that three width images obtain; Be final candidate region interested in rectangular wire frame.
Fig. 8. the candidate region interested of tumor region consecutive image and correspondence thereof.Left side is continuous print original graph photo, and right side is the bianry image of its correspondence, and white portion part is candidate region interested.
Fig. 9. the candidate region interested of non-tumor region consecutive image.Left side is continuous print original graph photo, and right side is the bianry image of its correspondence, and white portion is candidate region interested.
Figure 10. ABVS image and elliposoidal area-of-interest schematic diagram thereof.Wherein, (a) transversal section area-of-interest figure; (b) sagittal plane area-of-interest figure; (c) coronal-plane area-of-interest figure.
Figure 11. tumour region of interesting extraction result: the image of the maximum area-of-interest of (a) tumour three tangent planes; The image of b area-of-interest three tangent planes that () extracts.
Embodiment
In three-D ultrasonic mammary gland total volume imaging (ABVS) propose the present invention, the extraction method of area-of-interest is tested.ABVS image takes from the ACUSON S2000 of Siemens Company tMultrasonic instrument.This system is equipped with wide-band linearity probe (14L5BV), can obtain the mammary gland volume images of 15.4 cm × 16.8 cm × (2 ~ 6) cm.Acquire 15 routine ABVS images in this experiment altogether, each example has 98 ~ 294 width coronal-planes, 820 width cross-sectional images, 750 width sagittal view pictures.
First, rebuild original ABVS image, after rebuilding, three tangent planes (transversal section, sagittal plane, coronal-plane) of image as shown in Figure 2, wherein can see the general profile of breast in coronal image.As shown in Figure 2, breast contours, generally close to ellipse, therefore finds with Hough transformation the ellipse representing breast, as shown in Figure 3 on the image of coronal-plane.In ensuing process, only need to consider the region in process ellipse, can operand be reduced, get rid of the interference of breast external noise, artifact etc.The image of each tangent plane after Figure 4 shows that original each tangent plane original image and applying oval template.
Shown in Fig. 5 is carry out pretreated process to cross-sectional image.Fig. 5 (a) is depicted as original image.The feature that contrast for ultrasonoscopy is low, speckle noise is serious, first adopt anisotropy elimination of spot noise, its result is as shown in Fig. 5 (b).Then, reduce tonal range, the gray-scale value be actually being positioned in the middle of 0 ~ 255 carries out certain stretching, and its result is as shown in Fig. 5 (c).Finally, strengthen low echo area territory gray-scale value, its result is as shown in Fig. 5 (d).
Figure 6 shows that the maximum direction phase diagram of cross-sectional image.Because the interior intensity value of some tumour differs greatly, if only utilize gray-scale map to carry out Threshold segmentation, complete tumor region cannot be obtained, therefore, increase phase information.Figure 7 shows that i, pMOwith i+PMOthe extraction of three width interesting image regions and the last candidate region interested chosen from three.Here, the candidate region interested finally chosen is by image i+PMOcarry out that Threshold segmentation and regional choice obtain.
Fig. 8 and Fig. 9 is the candidate region interested of non-tumor region and tumor region transversal section consecutive image and correspondence thereof respectively.This shows, the area-of-interest of tumor region image is continuous print, overlaps each other; But not the area-of-interest of tumor region image does not overlap each other.The unrelated images sheet of a part can be got rid of according to this feature.
Figure 11 is tumour region of interesting extraction result.The image removing irrelevant sheet is divided into groups, then calculates texture and the shape facility of every picture group photo, and utilize logistic regression sorter to classify, adopt its validity of method validation of ten cross validations.
In an experiment, process 15 routine ABVS images altogether, wherein 14 examples can obtain the group containing tumor region, and rate of accuracy reached is to 93.3%.Utilize the gray level image containing the image sheet in the group of tumor region and bianry image, in conjunction with sagittal image, calculate the main shaft comprising three orthogonal directionss of the minimum ellipsoid of tumor region long, obtain the elliposoidal area-of-interest of tumour, its schematic diagram as shown in Figure 10.Three tangent planes that Figure 11 (a) is ABVS image, three tangent planes of the actual area-of-interest finally obtained are as shown in Figure 11 (b).Adopt ellipsoid represent interested region, the ellipsoid obtained than cubical area closer to real tumor region.
To sum up, the present invention is suitable for the extraction of tumour area-of-interest in three-dimensional mammary gland total volume image, and the process of whole extraction is completely automatically, accurately, can realize the location of tumor region, no longer rely on the mark of user, more objective and efficient.
list of references
[1] N. Otsu, “A threshold selection method from gray level histograms,” IEEE Trans. Syst., Man Cybern., vol. 9, pp. 62–66, 1979.
[2] B. Liu, H.D. Cheng, J.H. Huang, J.W. Tian, X.L. Tang, J.F. Liu, “Fully automatic and segmentation-robust classification of breast tumors based on local texture analysis of ultrasound images,” Pattern Recognit., vol.43, pp. 280–298, 2010.
[3] M. Xian, Y.T. Zhang, H.D. Cheng, “Fully automatic segmentation of breast ultrasound images based on breast characteristics in space and frequency domains,” Pattern Recognit., vol.48, pp. 485–497, 2015.
[4] J. Shan, H.D. Cheng, Y. Wang, “Completely automated segmentation approach for breast ultrasound images using multiple-domain feature,” Ultrasound in Med. & Biol., vol.38, pp. 262–275, 2012.
[5] R. M. Haralick, K. Shanmugam, I. H. Dinstein, “Textural features for image classification,” IEEE Trans. Syst., Man Cybern., vol.3, pp. 10–621, 1973。

Claims (6)

1. an extraction method for the middle area-of-interest of three-D ultrasonic mammary gland total volume imaging (ABVS), is characterized in that concrete step is:
(1) first the ABVS image of DICOM form is carried out image reconstruction according to the distance between the pixel on three-dimensional, make it corresponding with actual galactophore image size; Image sheet, 750 sagittal image sheets of 820 width transversal section are obtained after rebuilding; According to the difference of scan depths, obtain the image sheet of 98 ~ 294 width coronal-planes;
(2) in coronal image, oval feature is generally according to mammary gland, calculate the minimum value map image of the front ten width coronal image sheets of ABVS image after rebuilding, after its threshold process, obtain the template of mammary gland on coronal-plane with Hough elliptic transformation, and be applied to all coronal image sheets; Utilize the relation between ABVS image coronal-plane and transversal section three-dimensional coordinate, mammary gland position on coronal-plane is projected to transversal section, determine mammary gland position in transversal section, tentatively reduce the hunting zone of transversal section area-of-interest, remove the impact of mammary gland external noise, artifact interference;
(3) to 820 width images of transversal section, every ten width are merged into the new image of a width by minimum value mapping, after merging, obtains 82 width images; Wherein every width image is adopted based on maximum direction phase method, in conjunction with knub position characteristic and texture feature, extracts candidate region interested, obtain the bianry image of respective regions;
(4) by step (3), the bianry image of candidate region interested, a series of continuous transversal section is obtained; According to the continuity of tumour, if the connected region of certain a slice and front and back adjacent sheet zero lap, be then called irrelevant sheet, first removed; To remaining sheet, divide into groups according to its continuity and connected region center, often group is expressed as a suspected tumor;
(5) group cross-sectional view photo each in step (4) is classified; Extract the shape of its bianry image and corresponding grey scale image, textural characteristics, inputing to logistic regression sorter, to obtain every group may be the probability of tumour, the cross-sectional view photo of maximum probability person corresponding to real tumor region;
(6), in the bianry image of the tumor region serialgram determined in step (5), find maximum connected region, determine the minimum ellipse comprising this connected region; Find the sagittal view photo that elliptical center horizontal ordinate is corresponding, adopt the algorithm extracting candidate region interested in step (3) to obtain the candidate region interested of sagittal view photo, and then determine the minimum ellipse comprising this region; Utilize two ellipses of sagittal plane and transversal section, determine that ABVS image comprises three main shafts of the minimum ellipsoid of area-of-interest long, thus obtain the area-of-interest of tumour.
2. the extraction method of three-D ultrasonic mammary gland total volume interesting image regions according to claim 1, it is characterized in that in step (1) and step (2), ABVS image scans along mammary gland cross-sectional direction the sequential chart photo reconstruction obtained by instrument and obtains, its form is DICOM, has the plane that three orthogonal: transversal section, sagittal plane and coronal-plane; The two-dimensional ultrasonic image that ultrasonoscopy in transversal section and sagittal plane and traditional ultrasonic hand-held instrument obtain is more close, and the ultrasonoscopy on coronal-plane can observe the general profile of mammary gland.
3. the extraction method of three-D ultrasonic mammary gland total volume interesting image regions according to claim 1, it is characterized in that adopting in step (3) area-of-interest exacting method based on maximum direction phase place and tumour priori cross-sectional view photo to be carried out to the extraction of area-of-interest, detailed process is:
(1) ultrasonoscopy is carried out to the pre-service of image noise reduction and contrast strengthen: first adopt anisotropy Method of Noise to reduce the speckle noise of ultrasonoscopy, then improve picture contrast by gray scale stretching conversion;
(2) image obtained above-mentioned pre-service asks for its maximum direction phase diagram (PMO); Due in ultrasonic galactophore image, the direction of tumor boundaries is constantly change, at the frequency domain of image along six direction: 0 °, and 30 °, 60 °, 90 °, 120 °, 150 °, carry out filtering to image, final extraction has the phase information in ceiling capacity direction; This six direction covers whole frequency spectrum 0 ~ 360 °;
To image i, calculate PMO matrix, use two dimension here log-gaborwave filter to image filtering, this two dimension log-gaborthe transport function of wave filter is the Gaussian function under logarithmic scale, is decomposed into two parts: radial wave filter and angular filter; For each direction θ 0, the wave filter of structure is:
(1)
Wherein, in braces, front portion represents radial wave filter, and rear portion is angular filter, ωfor radial coordinate, θfor angle coordinate, ω 0for filter centre frequency, arrange the centre frequency of four different scales, its value is respectively 1/3,1/ (3 × 1.7), 1/ (3 × 1.7 2), 1/ (3 × 1.7 - 3); σ θ determine with θcentered by angular bandwidth, determine radial bandwidth, while ensureing filter effect, avoid aliasing;
Image through different center frequency, 24 altogether of different angles log-gaborafter filter filtering, 24 image arrays are changed in the inversion through Fast Fourier Transform (FFT) (FFT) , xfor direction, srepresent centre frequency yardstick, then calculate the phasing matrix in each angle lPA:
(2)
Wherein, represent in angle θtime phasing matrix, nrepresent the quantity of different center frequency wave filter, for yardstick sunder wild phase bit matrix, its calculating formula is:
(3)
(4)
Wherein imag( x, s) be imaginary part, real( x, s) be real part;
After six angle filtering, obtain six lPAeigenmatrix; Then, these six matrixes are merged into an eigenmatrix; Here, extract certain a bit on there is ceiling capacity angle phase characteristic value as the final phase characteristic value of this point, thus obtain a new phase characteristic matrix; So, phase characteristic pMOeach pixel is defined as:
(5)
Wherein represent in angle θon energy matrix, its calculating formula is:
(6)
Here imag( x, s) and real( x, s) represent that image passes through respectively log-gaborimaginary part after wave filter and real part; ρthe angle with maximum local energy obtained by formula (5), represent ρthe superposition of the phasing matrix of four yardstick centre frequencies in angle, obtains the PMO figure of ultrasonoscopy;
(3) in the choosing of candidate region interested, in conjunction with the information of gray level image and maximum direction phase diagram, with OSTU threshold method to pretreated image i, normalized pMOwith pMO+Ithree width images carry out Threshold segmentation process respectively, obtain the bianry image containing a lot of connected regions; Remove the noise region that in image, area is less than 300 pixels and is greater than 50% with the connected ratio at edge, calculate the energy function of remaining all connected regions e:
(7)
Wherein compactrepresent the degree of compacting of connected region, its value is the area of connected region and the area ratio of its Least Chimb shape; areafor the area of connected region; centdisfor the distance of connected region center and picture centre; eccentricityfor the eccentricity of the minimum external ellipse of connected region; for region and coincident rate, its value is region accounts for the total quantity of image edge pixels point ratio with the quantity of the pixel of coincident; w 1 ~ w 4represent that the degree of compacting in region, area, centre distance and eccentricity affect size to region energy functional value respectively, wfor the weight of coincident rate;
Through said process, obtain pretreated image respectively i, normalized pMOwith pMO+Ithe candidate region interested of three width images; Calculate the energy function of three width interesting image regions e, select emaximum region is the final candidate region interested of image.
4. the extraction method of three-D ultrasonic mammary gland total volume interesting image regions according to claim 1, is characterized in that in step (4), judge a certain image sheet be whether unrelated images sheet according to being:
(15)
Wherein bW k represent the kthe bianry image of width cross-sectional image, mwith nbe respectively the pixel number that image is axial and horizontal; When r( k) <50 time, think kthe candidate region interested of width image and adjacent sheet zero lap, remove the in image collection kwidth image;
To remaining image sheet, first according to the continuity of tumor region in cross-sectional view photo, continuous print image sheet is divided into one group; Then often organizing in continuous print image sheet, the distance between the center, candidate region interested of calculating adjacent sheet, is classified as one group by the image sheet of distance within 50 pixels; Like this, the final continuous cross-section slices dividing the often group expression suspected tumor obtained.
5. the extraction method of three-D ultrasonic mammary gland total volume interesting image regions according to claim 1, it is characterized in that in step (5), extract texture and the shape facility of every picture group photo, input to logistic regression sorter, obtaining this group may be the probability of tumour; The group of maximum probability represents tumor region;
In the feature extracted, textural characteristics comprises gray level co-occurrence matrixes characteristic sum gray feature; Shape facility comprises coincidence factor characteristic sum aspect ratio features between group internal sheet.
6. the extraction method of three-D ultrasonic mammary gland total volume interesting image regions according to claim 1, it is characterized in that in step (6), by bianry image and the gray level image of the serialgram of the tumor region of acquisition, determine that three main shafts of the ellipsoid comprising area-of-interest are long, thus obtain the area-of-interest of spheroid-like, detailed process is:
First, find the candidate region interested that the area that is positioned at tumor region centre position is maximum, with morphologic plavini by candidate region interested to external expansion 10 pixels, calculate the minimum external ellipse in the region after expansion, oval major axis length is a, minor axis length is b;the sequence number of the corresponding sagittal view photo of transverse axis coordinate on transversal section, if the center point coordinate of the external ellipse of area-of-interest be ( x 0, y 0), then get x 0sheet sagittal view photo, at this sagittal view picture yaxial coordinate ( y 0- b/ 2, y 0+ b/ 2) the above-mentioned candidate region interested extracting method based on maximum direction phase place is utilized to obtain sagittal plane candidate region interested in region; Calculate this candidate region interested to the external ellipse of external expansion 10 pixel rear regions and major axis long c,like this, three main shafts just obtaining the ellipsoid comprising area-of-interest are long a, bwith c.
CN201510044837.0A 2015-01-28 2015-01-28 The extraction method of three-D ultrasonic mammary gland total volume interesting image regions Active CN104657984B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201510044837.0A CN104657984B (en) 2015-01-28 2015-01-28 The extraction method of three-D ultrasonic mammary gland total volume interesting image regions

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201510044837.0A CN104657984B (en) 2015-01-28 2015-01-28 The extraction method of three-D ultrasonic mammary gland total volume interesting image regions

Publications (2)

Publication Number Publication Date
CN104657984A true CN104657984A (en) 2015-05-27
CN104657984B CN104657984B (en) 2018-10-16

Family

ID=53249060

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201510044837.0A Active CN104657984B (en) 2015-01-28 2015-01-28 The extraction method of three-D ultrasonic mammary gland total volume interesting image regions

Country Status (1)

Country Link
CN (1) CN104657984B (en)

Cited By (36)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104899841A (en) * 2015-06-15 2015-09-09 惠仁望都医疗设备科技有限公司 Generation calculation method for nuclear magnetic resonance image
CN105513057A (en) * 2015-11-30 2016-04-20 首都医科大学 Tumor auxiliary diagnosis method
CN106023188A (en) * 2016-05-17 2016-10-12 天津大学 Breast tumor feature selection method based on Relief algorithm
CN106023293A (en) * 2016-05-26 2016-10-12 北京爱科声科技有限公司 Three-dimensional reconstruction method based on C-scan ultrasonic image
CN106469445A (en) * 2015-08-18 2017-03-01 青岛海信医疗设备股份有限公司 A kind of calibration steps of 3-D view, device and system
CN106846252A (en) * 2017-02-09 2017-06-13 深圳市医诺智能科技发展有限公司 The anisotropy Zoom method and its system of a kind of image target area
CN106875409A (en) * 2017-03-24 2017-06-20 云南大学 A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method
CN106897682A (en) * 2017-02-15 2017-06-27 电子科技大学 Leucocyte automatic identifying method in a kind of leukorrhea based on convolutional neural networks
CN107610105A (en) * 2017-08-31 2018-01-19 沈阳东软医疗系统有限公司 Position ROI method, apparatus, equipment and machinable medium
CN107766874A (en) * 2017-09-07 2018-03-06 沈燕红 A kind of measuring method and measuring system of ultrasound volume biological parameter
CN107928707A (en) * 2017-12-07 2018-04-20 苏州掌声医疗科技有限公司 A kind of area method for fast measuring and system suitable for portable ultrasonic device
CN108269261A (en) * 2016-12-30 2018-07-10 亿阳信通股份有限公司 A kind of Bones and joints CT image partition methods and system
CN108369268A (en) * 2015-12-14 2018-08-03 皇家飞利浦有限公司 System and method for medical supply tracking
WO2018141607A1 (en) * 2017-02-02 2018-08-09 Elekta Ab (Publ) System and method for detecting brain metastases
CN108629378A (en) * 2018-05-10 2018-10-09 上海鹰瞳医疗科技有限公司 Image-recognizing method and equipment
CN108830852A (en) * 2018-07-13 2018-11-16 上海深博医疗器械有限公司 Three-D ultrasonic tumour auxiliary measurement system and method
CN109215023A (en) * 2018-09-17 2019-01-15 青岛海信医疗设备股份有限公司 A kind of method and apparatus of determining organ and tumor contact area
CN109357636A (en) * 2018-12-10 2019-02-19 电子科技大学 A kind of phase amplification calculation method based on black part structure light scan
CN109377481A (en) * 2018-09-27 2019-02-22 上海联影医疗科技有限公司 Image quality evaluating method, device, computer equipment and storage medium
CN110246125A (en) * 2019-05-31 2019-09-17 天津大学 Teat placement automatic testing method based on ABUS coronal image
CN110524886A (en) * 2018-12-04 2019-12-03 北京星驰恒动科技发展有限公司 Threedimensional model forming method and 3D printing method in 3D printing method
CN111275617A (en) * 2020-01-09 2020-06-12 云南大学 Automatic splicing method and system for ABUS breast ultrasound panorama and storage medium
WO2020133510A1 (en) * 2018-12-29 2020-07-02 深圳迈瑞生物医疗电子股份有限公司 Ultrasonic imaging method and device
CN111402277A (en) * 2020-02-17 2020-07-10 艾瑞迈迪医疗科技（北京）有限公司 Object outer contour segmentation method and device for medical image
CN111436972A (en) * 2020-04-13 2020-07-24 王时灿 Three-dimensional ultrasonic gynecological disease diagnosis device
CN111832563A (en) * 2020-07-17 2020-10-27 江苏大学附属医院 Intelligent breast tumor identification method based on ultrasonic image
CN112348082A (en) * 2020-11-06 2021-02-09 上海依智医疗技术有限公司 Deep learning model construction method, image processing method and readable storage medium
CN112568932A (en) * 2021-02-26 2021-03-30 深圳中科乐普医疗技术有限公司 Puncture needle development enhancement method and system and ultrasonic imaging equipment
US20210338194A1 (en) * 2016-06-27 2021-11-04 Taihao Medical Inc. Analysis method for breast image and electronic apparatus using the same
CN113689402A (en) * 2021-08-24 2021-11-23 北京长木谷医疗科技有限公司 Deep learning-based femoral medullary cavity form identification method, device and storage medium
CN114332547A (en) * 2022-03-17 2022-04-12 浙江太美医疗科技股份有限公司 Medical object classification method and apparatus, electronic device, and storage medium
CN114305502A (en) * 2020-09-29 2022-04-12 深圳迈瑞生物医疗电子股份有限公司 Mammary gland ultrasonic scanning method, device and storage medium
CN114581468A (en) * 2022-03-04 2022-06-03 平顶山学院 Activated sludge strain segmentation method based on anisotropic phase stretch transformation
CN116168276A (en) * 2023-02-27 2023-05-26 脉得智能科技(无锡)有限公司 Multi-modal feature fusion-based breast nodule classification method, device and storage medium
US11915347B2 (en) 2018-06-11 2024-02-27 Shanghai United Imaging Healthcare Co., Ltd. Systems and methods for reconstructing cardiac images
GB2605473B (en) * 2020-10-30 2024-03-27 Ibm Logistic model to determine 3D z-wise lesion connectivity

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050027188A1 (en) * 2002-12-13 2005-02-03 Metaxas Dimitris N. Method and apparatus for automatically detecting breast lesions and tumors in images
CN101085364A (en) * 2006-06-07 2007-12-12 沈阳东软医疗系统有限公司 Method for detecting mammary cancer armpit lymph gland transferring focus
CN102289657A (en) * 2011-05-12 2011-12-21 西安电子科技大学 Breast X ray image lump detecting system based on visual attention mechanism
CN103337074A (en) * 2013-06-18 2013-10-02 大连理工大学 Active contour model based method for segmenting mammary gland DCE-MRI focus
CN103914697A (en) * 2012-12-29 2014-07-09 上海联影医疗科技有限公司 Extraction method for region of interest of breast three-dimension image

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050027188A1 (en) * 2002-12-13 2005-02-03 Metaxas Dimitris N. Method and apparatus for automatically detecting breast lesions and tumors in images
CN101085364A (en) * 2006-06-07 2007-12-12 沈阳东软医疗系统有限公司 Method for detecting mammary cancer armpit lymph gland transferring focus
CN102289657A (en) * 2011-05-12 2011-12-21 西安电子科技大学 Breast X ray image lump detecting system based on visual attention mechanism
CN103914697A (en) * 2012-12-29 2014-07-09 上海联影医疗科技有限公司 Extraction method for region of interest of breast three-dimension image
CN103337074A (en) * 2013-06-18 2013-10-02 大连理工大学 Active contour model based method for segmenting mammary gland DCE-MRI focus

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
SHAN JUAN: "A fully automatic segmentation method for breast ultrasound images", 《UTAH STATE UNIVERSITY》 *
粟华等: "基于相位的C-V模型乳腺超声图像分割方法", 《东南大学学报(自然科学版)》 *

Cited By (54)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN104899841B (en) * 2015-06-15 2018-10-09 惠仁望都医疗设备科技有限公司 A kind of generation computational methods of nuclear magnetic resonance image
CN104899841A (en) * 2015-06-15 2015-09-09 惠仁望都医疗设备科技有限公司 Generation calculation method for nuclear magnetic resonance image
CN106469445A (en) * 2015-08-18 2017-03-01 青岛海信医疗设备股份有限公司 A kind of calibration steps of 3-D view, device and system
CN105513057A (en) * 2015-11-30 2016-04-20 首都医科大学 Tumor auxiliary diagnosis method
CN108369268B (en) * 2015-12-14 2022-10-18 皇家飞利浦有限公司 System and method for medical device tracking
CN108369268A (en) * 2015-12-14 2018-08-03 皇家飞利浦有限公司 System and method for medical supply tracking
CN106023188A (en) * 2016-05-17 2016-10-12 天津大学 Breast tumor feature selection method based on Relief algorithm
CN106023293A (en) * 2016-05-26 2016-10-12 北京爱科声科技有限公司 Three-dimensional reconstruction method based on C-scan ultrasonic image
US11944486B2 (en) * 2016-06-27 2024-04-02 Taihao Medical Inc. Analysis method for breast image and electronic apparatus using the same
US20210338194A1 (en) * 2016-06-27 2021-11-04 Taihao Medical Inc. Analysis method for breast image and electronic apparatus using the same
CN108269261A (en) * 2016-12-30 2018-07-10 亿阳信通股份有限公司 A kind of Bones and joints CT image partition methods and system
WO2018141607A1 (en) * 2017-02-02 2018-08-09 Elekta Ab (Publ) System and method for detecting brain metastases
CN106846252B (en) * 2017-02-09 2019-11-15 深圳市医诺智能科技发展有限公司 A kind of the anisotropy Zoom method and its system of image target area
CN106846252A (en) * 2017-02-09 2017-06-13 深圳市医诺智能科技发展有限公司 The anisotropy Zoom method and its system of a kind of image target area
CN106897682A (en) * 2017-02-15 2017-06-27 电子科技大学 Leucocyte automatic identifying method in a kind of leukorrhea based on convolutional neural networks
CN106875409A (en) * 2017-03-24 2017-06-20 云南大学 A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method
CN106875409B (en) * 2017-03-24 2019-06-21 云南大学 A kind of light-type incisional hernia sticking patch three-dimensional ultrasound pattern feature extracting method
CN107610105B (en) * 2017-08-31 2020-12-11 东软医疗系统股份有限公司 Method, device and equipment for positioning ROI and machine-readable storage medium
CN107610105A (en) * 2017-08-31 2018-01-19 沈阳东软医疗系统有限公司 Position ROI method, apparatus, equipment and machinable medium
CN107766874A (en) * 2017-09-07 2018-03-06 沈燕红 A kind of measuring method and measuring system of ultrasound volume biological parameter
CN107928707A (en) * 2017-12-07 2018-04-20 苏州掌声医疗科技有限公司 A kind of area method for fast measuring and system suitable for portable ultrasonic device
CN107928707B (en) * 2017-12-07 2023-08-15 成都优途科技有限公司 Area rapid measurement method and system suitable for portable ultrasonic equipment
CN108629378A (en) * 2018-05-10 2018-10-09 上海鹰瞳医疗科技有限公司 Image-recognizing method and equipment
US11915347B2 (en) 2018-06-11 2024-02-27 Shanghai United Imaging Healthcare Co., Ltd. Systems and methods for reconstructing cardiac images
CN108830852A (en) * 2018-07-13 2018-11-16 上海深博医疗器械有限公司 Three-D ultrasonic tumour auxiliary measurement system and method
CN109215023A (en) * 2018-09-17 2019-01-15 青岛海信医疗设备股份有限公司 A kind of method and apparatus of determining organ and tumor contact area
CN109215023B (en) * 2018-09-17 2021-11-05 青岛海信医疗设备股份有限公司 Method and device for determining contact area between organ and tumor
CN109377481A (en) * 2018-09-27 2019-02-22 上海联影医疗科技有限公司 Image quality evaluating method, device, computer equipment and storage medium
CN110524886A (en) * 2018-12-04 2019-12-03 北京星驰恒动科技发展有限公司 Threedimensional model forming method and 3D printing method in 3D printing method
CN109357636B (en) * 2018-12-10 2019-12-17 电子科技大学 phase amplification calculation method based on black piece structured light scanning
CN109357636A (en) * 2018-12-10 2019-02-19 电子科技大学 A kind of phase amplification calculation method based on black part structure light scan
WO2020133510A1 (en) * 2018-12-29 2020-07-02 深圳迈瑞生物医疗电子股份有限公司 Ultrasonic imaging method and device
CN112672691B (en) * 2018-12-29 2024-03-29 深圳迈瑞生物医疗电子股份有限公司 Ultrasonic imaging method and equipment
CN112672691A (en) * 2018-12-29 2021-04-16 深圳迈瑞生物医疗电子股份有限公司 Ultrasonic imaging method and equipment
CN110246125A (en) * 2019-05-31 2019-09-17 天津大学 Teat placement automatic testing method based on ABUS coronal image
CN111275617A (en) * 2020-01-09 2020-06-12 云南大学 Automatic splicing method and system for ABUS breast ultrasound panorama and storage medium
CN111275617B (en) * 2020-01-09 2023-04-07 云南大学 Automatic splicing method and system for ABUS breast ultrasound panorama and storage medium
CN111402277A (en) * 2020-02-17 2020-07-10 艾瑞迈迪医疗科技（北京）有限公司 Object outer contour segmentation method and device for medical image
CN111402277B (en) * 2020-02-17 2023-11-14 艾瑞迈迪医疗科技（北京）有限公司 Object outline segmentation method and device for medical image
CN111436972A (en) * 2020-04-13 2020-07-24 王时灿 Three-dimensional ultrasonic gynecological disease diagnosis device
CN111832563A (en) * 2020-07-17 2020-10-27 江苏大学附属医院 Intelligent breast tumor identification method based on ultrasonic image
CN114305502A (en) * 2020-09-29 2022-04-12 深圳迈瑞生物医疗电子股份有限公司 Mammary gland ultrasonic scanning method, device and storage medium
GB2605473B (en) * 2020-10-30 2024-03-27 Ibm Logistic model to determine 3D z-wise lesion connectivity
CN112348082A (en) * 2020-11-06 2021-02-09 上海依智医疗技术有限公司 Deep learning model construction method, image processing method and readable storage medium
CN112348082B (en) * 2020-11-06 2021-11-09 上海依智医疗技术有限公司 Deep learning model construction method, image processing method and readable storage medium
CN112568932A (en) * 2021-02-26 2021-03-30 深圳中科乐普医疗技术有限公司 Puncture needle development enhancement method and system and ultrasonic imaging equipment
CN113689402B (en) * 2021-08-24 2022-04-12 北京长木谷医疗科技有限公司 Deep learning-based femoral medullary cavity form identification method, device and storage medium
CN113689402A (en) * 2021-08-24 2021-11-23 北京长木谷医疗科技有限公司 Deep learning-based femoral medullary cavity form identification method, device and storage medium
CN114581468A (en) * 2022-03-04 2022-06-03 平顶山学院 Activated sludge strain segmentation method based on anisotropic phase stretch transformation
CN114581468B (en) * 2022-03-04 2023-04-28 平顶山学院 Activated sludge strain segmentation method based on anisotropic phase stretching transformation
CN114332547B (en) * 2022-03-17 2022-07-08 浙江太美医疗科技股份有限公司 Medical object classification method and apparatus, electronic device, and storage medium
CN114332547A (en) * 2022-03-17 2022-04-12 浙江太美医疗科技股份有限公司 Medical object classification method and apparatus, electronic device, and storage medium
CN116168276A (en) * 2023-02-27 2023-05-26 脉得智能科技(无锡)有限公司 Multi-modal feature fusion-based breast nodule classification method, device and storage medium
CN116168276B (en) * 2023-02-27 2023-10-31 脉得智能科技(无锡)有限公司 Multi-modal feature fusion-based breast nodule classification method, device and storage medium

Also Published As

Publication number Publication date
CN104657984B (en) 2018-10-16

Similar Documents

Publication Publication Date Title
CN104657984A (en) Automatic extraction method of three-dimensional breast full-volume image regions of interest
CN102722890B (en) Non-rigid heart image grading and registering method based on optical flow field model
Zhao et al. Segmentation of ultrasound images of thyroid nodule for assisting fine needle aspiration cytology
Baselice Ultrasound image despeckling based on statistical similarity
CN104252708B (en) A kind of x-ray chest radiograph image processing method and system
Chang et al. Graph-based learning for segmentation of 3D ultrasound images
CN109919929A (en) A kind of fissuring of tongue feature extracting method based on wavelet transformation
CN108122221A (en) The dividing method and device of diffusion-weighted imaging image midbrain ischemic area
CN115843373A (en) Multi-scale local level set ultrasonic image segmentation method fusing Gabor wavelets
CN108830856B (en) GA automatic segmentation method based on time series SD-OCT retina image
CN104838422A (en) Image processing device and method
CN116580068B (en) Multi-mode medical registration method based on point cloud registration
Luo et al. Automatic liver parenchyma segmentation from abdominal CT images using support vector machines
CN111951215A (en) Image detection method and device and computer readable storage medium
CN106780718A (en) A kind of three-dimensional rebuilding method of paleontological fossil
Hacihaliloglu et al. Statistical shape model to 3D ultrasound registration for spine interventions using enhanced local phase features
CN106778793A (en) The repeatable measuring method and device of a kind of image feature
Tsantis et al. Multiresolution edge detection using enhanced fuzzy c‐means clustering for ultrasound image speckle reduction
Karthikeyan et al. Lungs segmentation using multi-level thresholding in CT images
Gupta et al. Brain tumor detection using curvelet transform and support vector machine
CN103530844A (en) Splicing method based on mycobacterium tuberculosis acid-fast staining image
CN113205553A (en) Light stripe center extraction method based on three-channel feature fusion
Haji et al. A novel neutrosophic method for automatic seed point selection in thyroid nodule images
CN102201038A (en) Method for detecting P53 protein expression in brain tumor
CN102184529A (en) Empirical-mode-decomposition-based edge detecting method

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