CN105956198A - Nidus position and content-based mammary image retrieval system and method - Google Patents
Nidus position and content-based mammary image retrieval system and method Download PDFInfo
- Publication number
- CN105956198A CN105956198A CN201610445569.8A CN201610445569A CN105956198A CN 105956198 A CN105956198 A CN 105956198A CN 201610445569 A CN201610445569 A CN 201610445569A CN 105956198 A CN105956198 A CN 105956198A
- Authority
- CN
- China
- Prior art keywords
- image
- collection
- similarity
- retrieved
- circle
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/50—Information retrieval; Database structures therefor; File system structures therefor of still image data
- G06F16/58—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually
- G06F16/583—Retrieval characterised by using metadata, e.g. metadata not derived from the content or metadata generated manually using metadata automatically derived from the content
-
- G06F19/321—
Landscapes
- Engineering & Computer Science (AREA)
- Library & Information Science (AREA)
- Theoretical Computer Science (AREA)
- Databases & Information Systems (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Apparatus For Radiation Diagnosis (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The invention provides a nidus position and content-based mammary image retrieval system and method. The system comprises an image preprocessing unit, an image nidus position similarity measurement unit, an image content similarity measurement unit and an image comprehensive similarity measurer. The method comprises the steps of obtaining a to-be-retrieved image of an x-ray image of a mammary molybdenum target and a historical image set; selecting a standard image; preprocessing the to-be-retrieved image and the historical image set; performing image nidus position similarity measurement on the preprocessed to-be-retrieved image and the preprocessed historical image set; and performing image content similarity measurement on the preprocessed to-be-retrieved image of the x-ray image of the mammary molybdenum target and the preprocessed historical image set to obtain an image comprehensive similarity image sequence number so as to obtain a retrieval result of the to-be-retrieved image. According to the system and method, a nidus position-based similarity measurement method is added, thereby effectively improving the retrieval performance of the x-ray image of the mammary molybdenum target and further assisting a doctor in diagnosing mammary diseases.
Description
Technical field
The invention belongs to medical image post-procession technique field, be specifically related to a kind of mammary gland based on lesions position Yu content
Image indexing system and method.
Background technology
At present, breast carcinoma examination is to realize breast carcinoma early to examine the important means early controlled, and can reduce the mortality rate of 30%.Breast
Gland mammography image is breast carcinoma detection in early days, the important evidence of diagnosis, and the different manifestations of focus in galactophore image becomes early
The sole criterion of phase Diagnosis of Breast cancer, but its diagnosis has bigger difficulty, can be effectively by the retrieval of galactophore image
Auxiliary diagnosis.
CBIR technology starts from the initial stage nineties, develops into nowadays its purposes medically and more comes
The most extensive, wherein the meaning in terms of galactophore image retrieval is the most great, is that the feature according to galactophore image carries out retrieving,
Retrieval result is considered as the useful value of medical diagnosis.Being developed so far, retrieval technique has had more ripe development, the most still
So at retrieval aspect of performance existing defects, be mainly manifested in retrieval performance relatively low on, it main reason is that the letter that image contains
Breath amount and similarity measurement mode, the content of extracted image useful information and the method for similarity retrieval.Therefore, as
What improves the performance retrieved, it is still necessary to discussed further and research.
Summary of the invention
For the deficiencies in the prior art, the present invention proposes a kind of galactophore image searching system based on lesions position Yu content
And method.
The technical scheme is that a kind of galactophore image searching system based on lesions position Yu content, including image
Pretreatment unit, image focus position similarity measurement unit, picture material similarity measurement unit and image synthesis similarity
Tolerance device；
Described image pre-processing unit, for obtaining the image F to be retrieved of breast molybdenum target radioscopic image0, history image collection
(F1,F2,…,Fn), selection standard image FC, to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out pretreatment,
Obtain pretreated image set (I0,I1,I2,…,In), wherein, including pretreated image I to be retrieved0After pretreatment
History image collection (I1,I2,…,In)；Described standard picture FCBreast molybdenum target X ray picture normal for form, that be of moderate size
Picture；
Described image focus position similarity measurement unit, pretreated to be checked for mammograms
Rope image I0With pretreated history image collection (I1,I2,…,In) carry out image focus position similarity measurement, obtain image
Lesions position similarity set (S1,S2,…,Sn)；
Described picture material similarity measurement unit, for the figure pretreated to be retrieved to mammograms
As I0With pretreated history image collection (I1,I2,…,In) carry out picture material similarity measurement, obtain picture material similar
Property set (E1,E2,…,En)；
Described image synthesis similarity measurements measuring device, for by image focus position similarity set (S1,S2,…,Sn) according to
Image focus position similarity sorts from big to small, and marking serial numbers, by the image focus position similarity set after sequence
The weight of sequence number distribution A%, by picture material similarity set (E1,E2,…,En) according to picture material similarity from small to large
Sequence, and marking serial numbers, by the weight of sequence number distribution (100-A) % of the picture material similarity set after sequence, complex chart
As lesions position similarity image sequence number and picture material similarity image sequence number draw image synthesis similarity image sequence number,
To image searching result to be retrieved: i.e. image synthesis similarity image sequence number is the least, represent that this image gets over phase with image to be retrieved
Seemingly.
Preferably, described image pre-processing unit includes: image denoising device and image intensifier；
Described image denoising device, for respectively to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out noise reduction
Process, obtain the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,P2,…,Pn)；
Described image intensifier, for respectively to the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,
P2,…,Pn) carry out image enhancement processing, obtain pretreated image I to be retrieved0With pretreated history image collection (I1,
I2,…,In)；
Preferably, described image focus position similarity measurement unit includes: lesions position central point and radius determiner,
Image to be retrieved and image library and standard picture aligner, image focus position similarity determiner；
Described lesions position central point and radius determiner, be used for determining pretreated image set (I0,I1,I2,…,
In) lesions position center point coordinate collection { (x0,y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,
R1,R2,…,Rn)；
Preferably, described pretreated image set (I is determined0,I1,I2,…,In) lesions position center point coordinate collection
{(x0,y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,R2,…,Rn) method particularly includes:
Use classical Da-Jin algorithm Threshold Segmentation Algorithm to pretreated image set (I0,I1,I2,…,In) carry out two-value
Change processes, and retains highlight regions in the image after binary conversion treatment, using the half of highlight regions X-direction maximum as focus
Place-centric point abscissa, using the half of highlight regions Y direction maximum as lesions position central point vertical coordinate, obtains disease
Stove place-centric point coordinates collection { (x0,y0),(x1,y1),(x2,y2)…,(xn,yn), travel through highlight regions institute a little, from first
Individual point starts, and utilizes right angled triangle Pythagorean theorem, obtains a little to the distance of central point, calculates in highlight regions institute successively a little
With the distance of central point, using its maximum as lesions position radius, obtain lesions position radius collection (R0,R1,R2,…,Rn)。
Described image to be retrieved and image library and standard picture aligner, for utilizing based on CPD method for registering, will locate in advance
Image set (I after reason0,I1,I2,…,In) and standard picture FCMate, the lesion center point coordinates collection after being changed
{(X0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn), and by lesion center point coordinates the collection { (X after conversion0,Y0),(X1,
Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) show at standard picture FCOn, utilize and turn
Lesion center point coordinates collection { (X after changing0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,
R1,R2,…,Rn) determine focal area (Circle0,Circle1,Circle2,…,Circlen)；
Described image focus position similarity determiner, for calculating foci region (Circle respectively0,Circle1,
Circle2,…,CirclenThe focal area Circle of image to be retrieved in)0Focal area (Circle with history image1,
Circle2,…,Circlen) common factor and union, order
Obtain image focus position similarity set (S1,S2,S3,…,Sn)；
Preferably, described picture material similarity measurement unit includes: characteristics of image rectangular histogram extractor and picture material
Similarity determiner；
Described characteristics of image rectangular histogram extractor, is used for extracting pretreated image set (I0,I1,I2,…,In) ash
Degree feature, shape facility and textural characteristics, build its image grey level histogram, straight based on edge orientation histogram, direction gradient
Side's figure and local binary patterns rectangular histogram, obtain gray feature vector (α0,α1,α2,…,αn), shape eigenvectors (β0,β1,
β2,…,βn) and texture feature vector (γ0,γ1,γ2,…,γn), merge gray feature vector (α0,α1,α2,…,αn), shape
Shape characteristic vector (β0,β1,β2,…,βn) and texture feature vector (γ0,γ1,γ2,…,γn), obtain multi-scale HoGC feature
Vector (ω0,ω1,ω2,…,ωn)；
Described picture material similarity determiner, for using EMD method by the multi-scale HoGC feature of image to be retrieved
Vector ω0Multi-scale HoGC characteristic vector (ω with history image1,ω2,…,ωn) carry out similarity measurement, obtain in image
Hold similarity set (E1,E2,…,En)。
Use the method that galactophore image searching system based on lesions position Yu content carries out image retrieval, including following step
Rapid:
Step 1: obtain the image F to be retrieved of breast molybdenum target radioscopic image0, history image collection (F1,F2,…,Fn), choose
Standard picture FC；
Step 2: to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out pretreatment, obtain pretreated
Image set (I0,I1,I2,…,In)；
Step 2.1: respectively to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out noise reduction process, dropped
Image P to be retrieved after making an uproar0With the history image collection (P after noise reduction1,P2,…,Pn)；
Step 2.2: respectively to the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,P2,…,Pn) enter
Row image enhancement processing, obtains pretreated image I to be retrieved0With pretreated history image collection (I1,I2,…,In)；
Step 3: the image I pretreated to be retrieved to mammograms0With pretreated history image collection
(I1,I2,…,In) carry out image focus position similarity measurement；
Step 3.1: determine pretreated image set (I0,I1,I2,…,In) lesions position center point coordinate collection { (x0,
y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,R2,…,Rn)；
Step 3.2: utilize based on CPD method for registering, by pretreated image set (I0,I1,I2,…,In) and standard drawing
As FCMate, lesion center point coordinates the collection { (X after being changed0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn), and
By lesion center point coordinates the collection { (X after conversion0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection
(R0,R1,R2,…,Rn) show at standard picture FCOn, utilize lesion center point coordinates the collection { (X after conversion0,Y0),(X1,Y1),
(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) determine focal area (Circle0,Circle1,
Circle2,…,Circlen)；
Step 3.3: calculating foci region (Circle respectively0,Circle1,Circle2,…,CirclenFigure to be retrieved in)
The focal area Circle of picture0Focal area (Circle with history image1,Circle2,…,Circlen) common factor and also
Collection, order Obtain image focus position similarity
Set (S1,S2,S3,…,Sn)；
Step 4: the image I pretreated to be retrieved to mammograms0With pretreated history image collection
(I1,I2,…,In) carry out picture material similarity measurement；
Step 4.1: extract pretreated image set (I0,I1,I2,…,In) gray feature, shape facility and texture
Feature, builds its image grey level histogram, based on edge orientation histogram, histograms of oriented gradients and local binary patterns Nogata
Figure, obtains gray feature vector (α0,α1,α2,…,αn), shape eigenvectors (β0,β1,β2,…,βn) and texture feature vector
(γ0,γ1,γ2,…,γn)；
Step 4.2: merge gray feature vector (α0,α1,α2,…,αn), shape eigenvectors (β0,β1,β2,…,βn) and
Texture feature vector (γ0,γ1,γ2,…,γn), obtain multi-scale HoGC characteristic vector (ω0,ω1,ω2,…,ωn)；
Step 4.3: use EMD method by multi-scale HoGC characteristic vector ω of image to be retrieved0With combining of history image
Close histogram feature vector (ω1,ω2,…,ωn) carry out similarity measurement, obtain picture material similarity set (E1,E2,…,
En)；
Step 5: by image focus position similarity set (S1,S2,…,Sn) according to image focus position similarity from greatly
To little sequence, and marking serial numbers, by the weight of the sequence number distribution A% of the image focus position similarity set after sequence, will figure
As content similarities set (E1,E2,…,En) sort from small to large according to picture material similarity, and marking serial numbers, will sequence
After picture material similarity set sequence number distribution (100-A) % weight, synthetic image lesions position similarity image sequence
Number and picture material similarity image sequence number draw image synthesis similarity image sequence number, obtain image searching result to be retrieved:
I.e. image synthesis similarity image sequence number is the least, represents that this image is the most similar to image to be retrieved.
Preferably, described step 3.2 comprises the following steps:
Step 3.2.1: extract pretreated image set (I0,I1,I2,…,In) breast contours in image and standard drawing
As FCBreast contours；
Step 3.2.2: utilize affine transformation based on CPD, registrates pretreated image set (I0,I1,I2,…,In) figure
Breast contours in Xiang and standard picture FCBreast contours, obtain registration transformation matrix (T0,T1,T2,…,Tn)；
Step 3.2.3: by registration transformation matrix (T0,T1,T2,…,Tn) to lesions position center point coordinate collection { (x0,
y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,R2,…,Rn) change, after being changed
Lesions position center point coordinate collection { (X0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn)}；
Step 3.2.4: by lesions position center point coordinate the collection { (X after conversion0,Y0),(X1,Y1),(X2,Y2),…,
(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) show at standard picture FCOn, by the lesions position after conversion
Heart point coordinates collection { (X0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) really
Fixed circle (Circle0,Circle1,Circle2,…,Circlen) as focal area.
Beneficial effects of the present invention:
The present invention proposes a kind of galactophore image searching system based on lesions position and content and method, traditional based on
On the basis of the image search method of content, add based on lesions position method for measuring similarity, it is possible to effectively improve breast
The retrieval performance of gland molybdenum target X-ray image such that it is able to further auxiliary doctor's diagnosis to mastopathy.
Accompanying drawing explanation
Fig. 1 is structural frames based on lesions position Yu the galactophore image searching system of content in the specific embodiment of the invention
Figure；
Fig. 2 is flow process based on lesions position Yu the galactophore image search method of content in the specific embodiment of the invention
Figure；
Fig. 3 is to the image pretreated to be retrieved of mammograms and pre-in the specific embodiment of the invention
History image collection after process carries out the flow chart of image focus position similarity measurement；
Fig. 4 is the flow chart utilizing in the specific embodiment of the invention and determining focal area based on CPD method for registering；
Fig. 5 is to the image pretreated to be retrieved of mammograms and pre-in the specific embodiment of the invention
History image collection after process carries out the flow chart of picture material similarity measurement；
Fig. 6 is to obtain at the image of image searching result to be retrieved according to history image collection in the specific embodiment of the invention
Reason process flow diagram flow chart.
Detailed description of the invention
Below in conjunction with the accompanying drawings the specific embodiment of the invention is described in detail.
A kind of galactophore image searching system based on lesions position Yu content, as it is shown in figure 1, include Image semantic classification list
Unit, image focus position similarity measurement unit, picture material similarity measurement unit and image synthesis similarity measurements measuring device.
Image pre-processing unit, for obtaining the image F to be retrieved of breast molybdenum target radioscopic image0, history image collection (F1,
F2,…,Fn), selection standard image FC, to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out pretreatment, obtain
Pretreated image set (I0,I1,I2,…,In), wherein, including pretreated image I to be retrieved0Go through with pretreated
History image set (I1,I2,…,In)。
Standard picture FCBreast molybdenum target radioscopic image normal for form, that be of moderate size.
Image pre-processing unit includes: image denoising device and image intensifier.
Image denoising device, for respectively to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out at noise reduction
Reason, obtains the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,P2,…,Pn)。
Image intensifier, for respectively to the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,
P2,…,Pn) carry out image enhancement processing, obtain pretreated image I to be retrieved0With pretreated history image collection (I1,
I2,…,In)。
Image focus position similarity measurement unit, for the figure pretreated to be retrieved to mammograms
As I0With pretreated history image collection (I1,I2,…,In) carry out image focus position similarity measurement, obtain image focus
Position similarity set (S1,S2,…,Sn)。
Image focus position similarity measurement unit includes: lesions position central point and radius determiner, image to be retrieved
With image library and standard picture aligner, image focus position similarity determiner.
Lesions position central point and radius determiner, be used for determining pretreated image set (I0,I1,I2,…,In)
Lesions position center point coordinate collection { (x0,y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,
R2,…,Rn)。
In present embodiment, determine pretreated image set (I0,I1,I2,…,In) lesions position center point coordinate
Collection { (x0,y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,R2,…,Rn) concrete grammar
For:
Use classical Da-Jin algorithm Threshold Segmentation Algorithm to pretreated image set (I0,I1,I2,…,In) carry out two-value
Change processes, and retains highlight regions in the image after binary conversion treatment, using the half of highlight regions X-direction maximum as focus
Place-centric point abscissa, using the half of highlight regions Y direction maximum as lesions position central point vertical coordinate, obtains disease
Stove place-centric point coordinates collection { (x0,y0),(x1,y1),(x2,y2)…,(xn,yn), travel through highlight regions institute a little, from first
Individual point starts, and utilizes right angled triangle Pythagorean theorem, obtains a little to the distance of central point, calculates in highlight regions institute successively a little
With the distance of central point, using its maximum as lesions position radius, obtain lesions position radius collection (R0,R1,R2,…,Rn)。
Image to be retrieved and image library and standard picture aligner, for utilizing based on CPD method for registering, after pretreatment
Image set (I0,I1,I2,…,In) and standard picture FCCarry out district to join, lesion center point coordinates the collection { (X after being changed0,
Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn), and by lesion center point coordinates the collection { (X after conversion0,Y0),(X1,Y1),(X2,
Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) show at standard picture FCOn, utilize the disease after conversion
Stove center point coordinate collection { (X0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,
Rn) determine focal area (Circle0,Circle1,Circle2,…,Circlen)。
Image focus position similarity determiner, for calculating foci region (Circle respectively0,Circle1,Circle2,…,
CirclenThe focal area Circle of image to be retrieved in)0Focal area (Circle with history image1,Circle2,…,
Circlen) common factor and union, orderObtain
Image focus position similarity set (S1,S2,S3,…,Sn)。
Picture material similarity measurement unit, for the image I pretreated to be retrieved to mammograms0
With pretreated history image collection (I1,I2,…,In) carry out picture material similarity measurement, obtain picture material similarity collection
Close (E1,E2,…,En)。
Picture material similarity measurement unit includes: characteristics of image rectangular histogram extractor and picture material similarity determine
Device.
Characteristics of image rectangular histogram extractor, is used for extracting pretreated image set (I0,I1,I2,…,In) gray scale special
Levy, shape facility and textural characteristics, build its image grey level histogram, based on edge orientation histogram, histograms of oriented gradients
With local binary patterns rectangular histogram, obtain gray feature vector (α0,α1,α2,…,αn), shape eigenvectors (β0,β1,β2,…,
βn) and texture feature vector (γ0,γ1,γ2,…,γn), merge gray feature vector (α0,α1,α2,…,αn), shape facility
Vector (β0,β1,β2,…,βn) and texture feature vector (γ0,γ1,γ2,…,γn), obtain multi-scale HoGC characteristic vector
(ω0,ω1,ω2,…,ωn)。
Picture material similarity determiner, for using EMD method by the multi-scale HoGC characteristic vector of image to be retrieved
ω0Multi-scale HoGC characteristic vector (ω with history image1,ω2,…,ωn) carry out similarity measurement, obtain picture material phase
Like property set (E1,E2,…,En)。
Image synthesis similarity measurements measuring device, for by image focus position similarity set (S1,S2,…,Sn) according to image
Lesions position similarity sorts from big to small, and marking serial numbers, by the sequence number of the image focus position similarity set after sequence
The weight of distribution A%, by picture material similarity set (E1,E2,…,En) sort from small to large according to picture material similarity,
And marking serial numbers, by the weight of sequence number distribution (100-A) % of the picture material similarity set after sequence, synthetic image focus
Position similarity image sequence number and picture material similarity image sequence number draw image synthesis similarity image sequence number, obtain to be checked
Rope image searching result: i.e. image synthesis similarity image sequence number is the least, represents that this image is the most similar to image to be retrieved.
Use the method that galactophore image searching system based on lesions position Yu content carries out image retrieval, as in figure 2 it is shown,
Comprise the following steps:
Step 1: obtain the image F to be retrieved of breast molybdenum target radioscopic image0, history image collection (F1,F2,…,Fn), choose
Standard picture FC, standard picture FCBreast molybdenum target radioscopic image normal for form, that be of moderate size.
Step 2: to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out pretreatment, obtain pretreated
Image set (I0,I1,I2,…,In), wherein, including pretreated image I to be retrieved0With pretreated history image collection
(I1,I2,…,In)。
Step 2.1: respectively to image F to be retrieved0With history image collection (F1,F2,…,Fn) carry out noise reduction process, dropped
Image P to be retrieved after making an uproar0With the history image collection (P after noise reduction1,P2,…,Pn)。
In present embodiment, the alternative approach of spatial domain is used to select median filter to be filtered, it is achieved to be retrieved
Image F0With history image collection (F1,F2,…,Fn) noise reduction process, reduce to image F to be retrieved0With history image collection (F1,
F2,…,FnNoise in).
Step 2.2: respectively to the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,P2,…,Pn) enter
Row image enhancement processing, obtains pretreated image I to be retrieved0With pretreated history image collection (I1,I2,…,In)。
In present embodiment, use the method that contrast strengthens to the image P to be retrieved after noise reduction0With the history after noise reduction
Image set (P1,P2,…,Pn) carry out image enhancement processing.Emphasize breast molybdenum target radioscopic image entirety or local characteristics, expanded view
Difference between different objects feature in Xiang, suppresses uninterested feature, increases the contrast of suspected abnormality and surrounding tissue.
Step 3: the image I pretreated to be retrieved to mammograms0With pretreated history image collection
(I1,I2,…,In) carry out image focus position similarity measurement, as shown in Figure 3.
Step 3.1: determine pretreated image set (I0,I1,I2,…,In) lesions position center point coordinate collection { (x0,
y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,R2,…,Rn)。
In present embodiment, use classical Da-Jin algorithm Threshold Segmentation Algorithm to pretreated image set (I0,I1,
I2,…,In) carry out binary conversion treatment, retain highlight regions in the image after binary conversion treatment, highlight regions X-direction is maximum
The half of value is as lesions position central point abscissa, using the half of highlight regions Y direction maximum as in lesions position
Heart point vertical coordinate, obtains lesions position center point coordinate collection { (x0,y0),(x1,y1),(x2,y2)…,(xn,yn), travel through highlighted
Region institute a little, from the beginning of first point, utilizes right angled triangle Pythagorean theorem, obtains a little to the distance of central point, counts successively
Calculate in highlight regions a little with the distance of central point, using its maximum as lesions position radius, obtain lesions position radius
Collection (R0,R1,R2,…,Rn)。
Step 3.2: utilize based on CPD method for registering, by pretreated image set (I0,I1,I2,…,In) and standard drawing
As FCCarry out district to join, lesion center point coordinates the collection { (X after being changed0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn), and
By lesion center point coordinates the collection { (X after conversion0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection
(R0,R1,R2,…,Rn) show at standard picture FCOn, utilize lesion center point coordinates the collection { (X after conversion0,Y0),(X1,Y1),
(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) determine focal area (Circle0,Circle1,
Circle2,…,Circlen), as shown in Figure 4.
In present embodiment, utilize based on CPD (Coherence Point Drift) method for registering, can be by different size
Mammary gland profile be standardized, reduce lesions position Similarity measures error.
Step 3.2.1: extract pretreated image set (I0,I1,I2,…,In) breast contours in image and standard drawing
As FCBreast contours.
Step 3.2.2: utilize affine transformation based on CPD (Coherence Point Drift), registrate pretreated
Image set (I0,I1,I2,…,In) breast contours in image and standard picture FCBreast contours, obtain registration transformation matrix
(T0,T1,T2,…,Tn)。
Step 3.2.3: by registration transformation matrix (T0,T1,T2,…,Tn) to lesions position center point coordinate collection { (x0,
y0),(x1,y1),(x2,y2)…,(xn,yn) and lesions position radius collection (R0,R1,R2,…,Rn) change, after being changed
Lesions position center point coordinate collection { (X0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn)}。
Step 3.2.4: by lesions position center point coordinate the collection { (X after conversion0,Y0),(X1,Y1),(X2,Y2),…,
(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) show at standard picture FCOn, by the lesions position after conversion
Heart point coordinates collection { (X0,Y0),(X1,Y1),(X2,Y2),…,(Xn,Yn) and lesions position radius collection (R0,R1,R2,…,Rn) really
Fixed circle (Circle0,Circle1,Circle2,…,Circlen) as focal area.
Step 3.3: calculating foci region (Circle respectively0,Circle1,Circle2,…,CirclenFigure to be retrieved in)
The focal area Circle of picture0Focal area (Circle with history image1,Circle2,…,Circlen) common factor and also
Collection, order Obtain image focus position phase
Like property set (S1,S2,S3,…,Sn)。
In present embodiment, Lesions position similarity ratio is the biggest more phase
Seemingly.
Step 4: the image I pretreated to be retrieved to mammograms0With pretreated history image collection
(I1,I2,…,In) carry out picture material similarity measurement, as shown in Figure 5.
Step 4.1: extract pretreated image set (I0,I1,I2,…,In) gray feature, shape facility and texture
Feature, builds image grey level histogram, based on edge orientation histogram (Edge Direction Histogram, EDH), direction
Histogram of gradients (Histogram of oriented gradients, HOG) and local binary patterns rectangular histogram (Local
Binary Pattern, LBP), obtain gray feature vector (α0,α1,α2,…,αn), shape eigenvectors (β0,β1,β2,…,
βn) and texture feature vector (γ0,γ1,γ2,…,γn)。
Step 4.2: merge gray feature vector (α0,α1,α2,…,αn), shape eigenvectors (β0,β1,β2,…,βn) and
Texture feature vector (γ0,γ1,γ2,…,γn) obtain multi-scale HoGC characteristic vector (ω0,ω1,ω2,…,ωn)。
Step 4.3: use EMD (Earth Mover's Distance) method that the multi-scale HoGC of image to be retrieved is special
Levy vector ω0Multi-scale HoGC characteristic vector (ω with history image1,ω2,…,ωn) carry out similarity measurement, obtain image
Content similarities set (E1,E2,…,En)。
In present embodiment, (E1,E2,…,En)=(0.0223,0.0668,0.2641,0.3103,0.3179 ...,
0.6624).Numerical value is the least, and feature histogram is the most similar, i.e. draws mammograms content similarities.
Step 5: by image focus position similarity set (S1,S2,…,Sn) according to image focus position similarity from greatly
To little sequence, and marking serial numbers, by the weight of the sequence number distribution 40% of the image focus position similarity set after sequence, will figure
As content similarities set (E1,E2,…,En) sort from small to large according to picture material similarity, and marking serial numbers, will sequence
After picture material similarity set sequence number distribution 60% weight, synthetic image lesions position similarity image sequence number and figure
As content similarities picture numbers draws image synthesis similarity image sequence number, obtain image searching result to be retrieved: i.e. image
Comprehensive similarity image sequence number is the least, represents that this image is the most similar to image to be retrieved.
In present embodiment, on the basis of traditional CBIR method, add based on focus position
Put method for measuring similarity, it is possible to effectively improve the retrieval performance of mammograms such that it is able to further assist
Doctor's diagnosis to mastopathy, obtains image processing process such as Fig. 6 of image searching result to be retrieved according to history image collection
Shown in.
Claims (7)
1. a galactophore image searching system based on lesions position Yu content, it is characterised in that include image pre-processing unit,
Image focus position similarity measurement unit, picture material similarity measurement unit and image synthesis similarity measurements measuring device；
Described image pre-processing unit, for obtaining the image F to be retrieved of breast molybdenum target radioscopic image0, history image collection (F1,
F2..., Fn), selection standard image FC, to image F to be retrieved0With history image collection (F1, F2..., Fn) carry out pretreatment, obtain
Pretreated image set (I0, I1, I2..., In), wherein, including pretreated image I to be retrieved0Go through with pretreated
History image set (I1, I2..., In)；Described standard picture FCBreast molybdenum target radioscopic image normal for form, that be of moderate size；
Described image focus position similarity measurement unit, for the figure pretreated to be retrieved to mammograms
As I0With pretreated history image collection (I1, I2..., In) carry out image focus position similarity measurement, obtain image focus
Position similarity set (S1, S2..., Sn)；
Described picture material similarity measurement unit, for the image I pretreated to be retrieved to mammograms0With
Pretreated history image collection (I1, I2..., In) carry out picture material similarity measurement, obtain picture material similarity set
(E1, E2..., En)；
Described image synthesis similarity measurements measuring device, for by image focus position similarity set (S1, S2..., Sn) according to image
Lesions position similarity sorts from big to small, and marking serial numbers, by the sequence number of the image focus position similarity set after sequence
The weight of distribution A%, by picture material similarity set (E1, E2..., En) sort from small to large according to picture material similarity,
And marking serial numbers, by the weight of sequence number distribution (100-A) % of the picture material similarity set after sequence, synthetic image focus
Position similarity image sequence number and picture material similarity image sequence number draw image synthesis similarity image sequence number, obtain to be checked
Rope image searching result: i.e. image synthesis similarity image sequence number is the least, represents that this image is the most similar to image to be retrieved.
Galactophore image searching system based on lesions position Yu content the most according to claim 1, it is characterised in that described
Image pre-processing unit includes: image denoising device and image intensifier；
Described image denoising device, for respectively to image F to be retrieved0With history image collection (F1, F2..., Fn) carry out noise reduction process,
Obtain the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1, P2..., Pn)；
Described image intensifier, for respectively to the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1,
P2..., Pn) carry out image enhancement processing, obtain pretreated image I to be retrieved0With pretreated history image collection (I1,
I2..., In)。
Galactophore image searching system based on lesions position Yu content the most according to claim 1, it is characterised in that described
Image focus position similarity measurement unit includes: lesions position central point and radius determiner, image to be retrieved and image library
With standard picture aligner, image focus position similarity determiner；
Described lesions position central point and radius determiner, be used for determining pretreated image set (I0, I1, I2..., In) disease
Stove place-centric point coordinates collection { (x0, y0), (x1, y1), (x2, y2) ..., (xn, yn) and lesions position radius collection (R0, R1,
R2..., Rn)；
Described image to be retrieved and image library and standard picture aligner, for utilizing based on CPD method for registering, after pretreatment
Image set (I0, I1, I2..., In) and standard picture FCMate, lesion center point coordinates the collection { (X after being changed0,
Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn), and by lesion center point coordinates the collection { (X after conversion0, Y0), (X1, Y1), (X2,
Y2) ..., (Xn, Yn) and lesions position radius collection (R0, R1, R2..., Rn) show at standard picture FCOn, utilize the disease after conversion
Stove center point coordinate collection { (X0, Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn) and lesions position radius collection (R0, R1, R2...,
Rn) determine focal area (Circle0, Circle1, Circle2..., Circlen)；
Described image focus position similarity determiner, for calculating foci region (Circle respectively0, Circle1, Circle2...,
CirclenThe focal area Circle of image to be retrieved in)0Focal area (Circle with history image1, Circle2...,
Circlen) common factor and union, orderObtain
Image focus position similarity set (S1, S2, S3..., Sn)。
Galactophore image searching system based on lesions position Yu content the most according to claim 1, it is characterised in that described
Picture material similarity measurement unit includes: characteristics of image rectangular histogram extractor and picture material similarity determiner；
Described characteristics of image rectangular histogram extractor, is used for extracting pretreated image set (I0, I1, I2..., In) gray scale special
Levy, shape facility and textural characteristics, build its image grey level histogram, based on edge orientation histogram, histograms of oriented gradients
With local binary patterns rectangular histogram, obtain gray feature vector (α0, α1, α2..., αn), shape eigenvectors (β0, β1, β2...,
βn) and texture feature vector (γ0, γ1, γ2..., γn), merge gray feature vector (α0, α1, α2..., αn), shape facility
Vector (β0, β1, β2..., βn) and texture feature vector (γ0, γ1, γ2..., γn), obtain multi-scale HoGC characteristic vector
(ω0, ω1, ω2..., ωn)；
Described picture material similarity determiner, for using EMD method by the multi-scale HoGC characteristic vector of image to be retrieved
ω0Multi-scale HoGC characteristic vector (ω with history image1, ω2..., ωn) carry out similarity measurement, obtain picture material phase
Like property set (E1, E2..., En)。
Galactophore image searching system based on lesions position Yu content the most according to claim 3, it is characterised in that described
Determine pretreated image set (I0, I1, I2..., In) lesions position center point coordinate collection { (x0, y0), (x1, y1), (x2,
y2) ..., (xn, yn) and lesions position radius collection (R0, R1, R2..., Rn) method particularly includes:
Use classical Da-Jin algorithm Threshold Segmentation Algorithm to pretreated image set (I0, I1, I2..., In) carry out at binaryzation
Reason, retains highlight regions in the image after binary conversion treatment, using the half of highlight regions X-direction maximum as lesions position
Central point abscissa, using the half of highlight regions Y direction maximum as lesions position central point vertical coordinate, obtains focus position
Put center point coordinate collection { (x0, y0), (x1, y1), (x2, y2) ..., (xn, yn), travel through highlight regions institute a little, from first point
Start, utilize right angled triangle Pythagorean theorem, obtain a little to the distance of central point, calculate successively in highlight regions a little with in
The distance of heart point, using its maximum as lesions position radius, obtains lesions position radius collection (R0, R1, R2..., Rn)。
6. use the galactophore image searching system based on lesions position Yu content described in claim 1 to carry out the side of image retrieval
Method, it is characterised in that comprise the following steps:
Step 1: obtain the image F to be retrieved of breast molybdenum target radioscopic image0, history image collection (F1, F2..., Fn), selection standard
Image FC；
Step 2: to image F to be retrieved0With history image collection (F1, F2..., Fn) carry out pretreatment, obtain pretreated image
Collection (I0, I1, I2..., In)；
Step 2.1: respectively to image F to be retrieved0With history image collection (F1, F2..., Fn) carry out noise reduction process, after obtaining noise reduction
Image P to be retrieved0With the history image collection (P after noise reduction1, P2..., Pn)；
Step 2.2: respectively to the image P to be retrieved after noise reduction0With the history image collection (P after noise reduction1, P2..., Pn) carry out image
Enhancement process, obtains pretreated image I to be retrieved0With pretreated history image collection (I1, I2..., In)；
Step 3: the image I pretreated to be retrieved to mammograms0With pretreated history image collection (I1,
I2..., In) carry out image focus position similarity measurement；
Step 3.1: determine pretreated image set (I0, I1, I2..., In) lesions position center point coordinate collection { (x0, y0),
(x1, y1), (x2, y2) ..., (xn, yn) and lesions position radius collection (R0, R1, R2..., Rn)；
Step 3.2: utilize based on CPD method for registering, by pretreated image set (I0, I1, I2..., In) and standard picture FC
Mate, lesion center point coordinates the collection { (X after being changed0, Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn), and will turn
Lesion center point coordinates collection { (X after changing0, Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn) and lesions position radius collection (R0,
R1, R2..., Rn) show at standard picture FCOn, utilize lesion center point coordinates the collection { (X after conversion0, Y0), (X1, Y1), (X2,
Y2) ..., (Xn, Yn) and lesions position radius collection (R0, R1, R2..., Rn) determine focal area (Circle0, Circle1,
Circle2..., Circlen)；
Step 3.3: calculating foci region (Circle respectively0, Circle1, Circle2..., CirclenImage to be retrieved in)
Focal area Circle0Focal area (Circle with history image1, Circle2..., Circlen) common factor and union, order Obtain the similarity set of image focus position
(S1, S2, S3..., Sn)；
Step 4: the image I pretreated to be retrieved to mammograms0With pretreated history image collection (I1,
I2..., In) carry out picture material similarity measurement；
Step 4.1: extract pretreated image set (I0, I1, I2..., In) gray feature, shape facility and textural characteristics,
Build its image grey level histogram, based on edge orientation histogram, histograms of oriented gradients and local binary patterns rectangular histogram,
To gray feature vector (α0, α1, α2..., αn), shape eigenvectors (β0, β1, β2..., βn) and texture feature vector (γ0,
γ1, γ2..., γn)；
Step 4.2: merge gray feature vector (α0, α1, α2..., αn), shape eigenvectors (β0, β1, β2..., βn) and texture
Characteristic vector (γ0, γ1, γ2..., γn), obtain multi-scale HoGC characteristic vector (ω0, ω1, ω2..., ωn)；
Step 4.3: use EMD method by multi-scale HoGC characteristic vector ω of image to be retrieved0Comprehensive Nogata with history image
Figure characteristic vector (ω1, ω2..., ωn) carry out similarity measurement, obtain picture material similarity set (E1, E2..., En)；
Step 5: by image focus position similarity set (S1, S2..., Sn) according to image focus position similarity from big to small
Sequence, and marking serial numbers, by the weight of the sequence number distribution A% of the image focus position similarity set after sequence, by image
Hold similarity set (E1, E2..., En) sort from small to large according to picture material similarity, and marking serial numbers, after sequence
Picture material similarity set sequence number distribution (100-A) % weight, synthetic image lesions position similarity image sequence number and
Picture material similarity image sequence number draws image synthesis similarity image sequence number, obtains image searching result to be retrieved: i.e. scheme
As comprehensive similarity image sequence number is the least, represent that this image is the most similar to image to be retrieved.
Galactophore image search method based on lesions position Yu content the most according to claim 5, it is characterised in that described
Step 3.2 comprises the following steps:
Step 3.2.1: extract pretreated image set (I0, I1, I2..., In) breast contours in image and standard picture FC
Breast contours；
Step 3.2.2: utilize affine transformation based on CPD, registrates pretreated image set (I0, I1, I2..., In) in image
Breast contours and standard picture FCBreast contours, obtain registration transformation matrix (T0, T1, T2..., Tn)；
Step 3.2.3: by registration transformation matrix (T0, T1, T2..., Tn) to lesions position center point coordinate collection { (x0, y0),
(x1, y1), (x2, y2) ..., (xn, yn) and lesions position radius collection (R0, R1, R2..., Rn) change, after being changed
Lesions position center point coordinate collection { (X0, Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn)}；
Step 3.2.4: by lesions position center point coordinate the collection { (X after conversion0, Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn)}
With lesions position radius collection (R0, R1, R2..., Rn) show at standard picture FCOn, by the lesions position center point coordinate after conversion
Collection { (X0, Y0), (X1, Y1), (X2, Y2) ..., (Xn, Yn) and lesions position radius collection (R0, R1, R2..., Rn) circle that determines
(Circle0, Circle1, Circle2..., Circlen) as focal area.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610445569.8A CN105956198B (en) | 2016-06-20 | 2016-06-20 | A kind of galactophore image searching system and method based on lesions position and content |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610445569.8A CN105956198B (en) | 2016-06-20 | 2016-06-20 | A kind of galactophore image searching system and method based on lesions position and content |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105956198A true CN105956198A (en) | 2016-09-21 |
| CN105956198B CN105956198B (en) | 2019-04-26 |
Family
ID=56907041
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610445569.8A Active CN105956198B (en) | 2016-06-20 | 2016-06-20 | A kind of galactophore image searching system and method based on lesions position and content |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105956198B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846317A (en) * | 2017-02-27 | 2017-06-13 | 北京连心医疗科技有限公司 | A kind of feature based extracts the method for retrieving medicine image with Similarity matching |
| CN107341265A (en) * | 2017-07-20 | 2017-11-10 | 东北大学 | A kind of galactophore image searching system and method for merging depth characteristic |
| CN107392204A (en) * | 2017-07-20 | 2017-11-24 | 东北大学 | A kind of galactophore image microcalcifications automatic checkout system and method |
| CN107833631A (en) * | 2017-11-20 | 2018-03-23 | 新乡医学院 | A kind of medical image computer-aided analysis method |
| CN108537893A (en) * | 2017-03-02 | 2018-09-14 | 南京同仁医院有限公司 | A kind of three-dimensional visualization model generation method of thyroid gland space occupying lesion |
| CN110837572A (en) * | 2019-11-15 | 2020-02-25 | 北京推想科技有限公司 | Image retrieval method and device, readable storage medium and electronic equipment |
| CN111091906A (en) * | 2019-10-31 | 2020-05-01 | 中电药明数据科技(成都)有限公司 | Auxiliary medical diagnosis method and system based on real world data |
| CN111583320A (en) * | 2020-03-17 | 2020-08-25 | 哈尔滨医科大学 | Breast cancer ultrasonic image typing method and system fusing deep convolutional network and image omics characteristics and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101147159A (en) * | 2005-02-21 | 2008-03-19 | 三菱电机株式会社 | Fast method of object detection by statistical template matching |
| CN101564323A (en) * | 2009-04-20 | 2009-10-28 | 华中科技大学 | Auxiliary equipment for diagnosing galactophore nidus based on galactophore X-ray photograph |
| CN102306173A (en) * | 2011-08-25 | 2012-01-04 | 重庆理工大学 | Image similarity comparison method |
| US20120283574A1 (en) * | 2011-05-06 | 2012-11-08 | Park Sun Young | Diagnosis Support System Providing Guidance to a User by Automated Retrieval of Similar Cancer Images with User Feedback |
| CN104915961A (en) * | 2015-06-08 | 2015-09-16 | 北京交通大学 | Lump image region display method and system based on mammary X-ray image |
-
2016
- 2016-06-20 CN CN201610445569.8A patent/CN105956198B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101147159A (en) * | 2005-02-21 | 2008-03-19 | 三菱电机株式会社 | Fast method of object detection by statistical template matching |
| CN101564323A (en) * | 2009-04-20 | 2009-10-28 | 华中科技大学 | Auxiliary equipment for diagnosing galactophore nidus based on galactophore X-ray photograph |
| US20120283574A1 (en) * | 2011-05-06 | 2012-11-08 | Park Sun Young | Diagnosis Support System Providing Guidance to a User by Automated Retrieval of Similar Cancer Images with User Feedback |
| CN102306173A (en) * | 2011-08-25 | 2012-01-04 | 重庆理工大学 | Image similarity comparison method |
| CN104915961A (en) * | 2015-06-08 | 2015-09-16 | 北京交通大学 | Lump image region display method and system based on mammary X-ray image |
Non-Patent Citations (1)
| Title |
|---|
| 邢春: ""基于多特征融合图像检索系统设计与实现"", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106846317A (en) * | 2017-02-27 | 2017-06-13 | 北京连心医疗科技有限公司 | A kind of feature based extracts the method for retrieving medicine image with Similarity matching |
| CN106846317B (en) * | 2017-02-27 | 2021-09-17 | 北京连心医疗科技有限公司 | Medical image retrieval method based on feature extraction and similarity matching |
| CN108537893A (en) * | 2017-03-02 | 2018-09-14 | 南京同仁医院有限公司 | A kind of three-dimensional visualization model generation method of thyroid gland space occupying lesion |
| CN107341265A (en) * | 2017-07-20 | 2017-11-10 | 东北大学 | A kind of galactophore image searching system and method for merging depth characteristic |
| CN107392204A (en) * | 2017-07-20 | 2017-11-24 | 东北大学 | A kind of galactophore image microcalcifications automatic checkout system and method |
| CN107341265B (en) * | 2017-07-20 | 2020-08-14 | 东北大学 | Mammary gland image retrieval system and method fusing depth features |
| CN107833631A (en) * | 2017-11-20 | 2018-03-23 | 新乡医学院 | A kind of medical image computer-aided analysis method |
| CN111091906A (en) * | 2019-10-31 | 2020-05-01 | 中电药明数据科技(成都)有限公司 | Auxiliary medical diagnosis method and system based on real world data |
| CN111091906B (en) * | 2019-10-31 | 2023-06-20 | 中电药明数据科技(成都)有限公司 | Auxiliary medical diagnosis method and system based on real world data |
| CN110837572A (en) * | 2019-11-15 | 2020-02-25 | 北京推想科技有限公司 | Image retrieval method and device, readable storage medium and electronic equipment |
| CN111583320A (en) * | 2020-03-17 | 2020-08-25 | 哈尔滨医科大学 | Breast cancer ultrasonic image typing method and system fusing deep convolutional network and image omics characteristics and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105956198B (en) | 2019-04-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105956198A (en) | Nidus position and content-based mammary image retrieval system and method | |
| US10614573B2 (en) | Method for automatically recognizing liver tumor types in ultrasound images | |
| CN108537773B (en) | Method for intelligently assisting in identifying pancreatic cancer and pancreatic inflammatory diseases | |
| Yang et al. | Lung lesion localization of COVID-19 from chest CT image: A novel weakly supervised learning method | |
| ur Rehman et al. | An appraisal of nodules detection techniques for lung cancer in CT images | |
| Nguyen et al. | Pulmonary nodule detection based on faster R-CNN with adaptive anchor box | |
| CN101103924A (en) | Galactophore cancer computer auxiliary diagnosis method based on galactophore X-ray radiography and system thereof | |
| Jony et al. | Detection of lung cancer from CT scan images using GLCM and SVM | |
| CN111899850A (en) | Medical image information processing method, display method and readable storage medium | |
| Feuerstein et al. | Automatic mediastinal lymph node detection in chest CT | |
| Kriti et al. | A review of segmentation algorithms applied to B-mode breast ultrasound images: A characterization approach | |
| Pawełczyk et al. | Towards detecting high-uptake lesions from lung CT scans using deep learning | |
| Wu et al. | Automatic lung segmentation in CT images using dilated convolution based weighted fully convolutional network | |
| Teuwen et al. | Soft tissue lesion detection in mammography using deep neural networks for object detection | |
| Dhalia Sweetlin et al. | Patient-Specific Model Based Segmentation of Lung Computed Tomographic Images. | |
| Duan et al. | Segmentation of pulmonary vascular tree by incorporating vessel enhancement filter and variational region-growing | |
| Liu et al. | Segmentation of mass in mammograms by a novel integrated active contour method | |
| Shi | Lung nodule detection using convolutional neural networks | |
| Chen et al. | Scrfd: spatial coherence based rib fracture detection | |
| Xiong et al. | Deep Ensemble Learning Network for Kidney Lesion Detection | |
| Li et al. | Intelligent system for predicting breast tumors using machine learning | |
| Smaoui et al. | Designing a new approach for the segmentation of the cancerous breast mass | |
| Wang et al. | Staging of hepatocellular carcinoma using deep feature in contrast-enhanced MR images | |
| Cao et al. | An adaptive pulmonary nodule detection algorithm | |
| Liu et al. | Detection and station mapping of mediastinal lymph nodes on thoracic computed tomography using spatial prior from multi-atlas label fusion |
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 |