CN106846301A - Retinal images sorting technique and device - Google Patents
Retinal images sorting technique and device Download PDFInfo
- Publication number
- CN106846301A CN106846301A CN201611248700.8A CN201611248700A CN106846301A CN 106846301 A CN106846301 A CN 106846301A CN 201611248700 A CN201611248700 A CN 201611248700A CN 106846301 A CN106846301 A CN 106846301A
- Authority
- CN
- China
- Prior art keywords
- image
- region
- interest
- area
- optic disk
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- 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
- G06T2207/30041—Eye; Retina; Ophthalmic
-
- 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
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Investigating Or Analysing Biological Materials (AREA)
- Image Processing (AREA)
Abstract
The present invention relates to a kind of retinal images sorting technique and device, methods described includes：Obtain sample retinal images；The area-of-interest of the sample retinal images is extracted, and extracts optic disk region and angiosomes in the area-of-interest；The characteristics of image in the optic disk region in the area-of-interest and the region outside the angiosomes is extracted, with according to described image features training Image Classifier；Obtain targeted retinal image to be sorted；The target region of interest of the targeted retinal image is extracted, and extracts optic disk region and angiosomes in the target region of interest；The characteristics of image in the optic disk region in the target region of interest and the region outside the angiosomes is extracted, the targeted retinal image is classified with according to described image grader.Retinal images sorting technique of the invention and device, can improve efficiency and the degree of accuracy of retinal images classification, simple and easy to apply, applied widely.
Description
Technical field
The present invention relates to technical field of image processing, more particularly to a kind of retinal images sorting technique and device.
Background technology
The development of medical science is healthy closely related with the mankind, therefore digital image processing techniques cause life from the beginning
The great interest of thing medical field.Just there is Document system to point out a quite varied application of image procossing early in late nineteen seventies
Occasion is Medical Image Processing.No matter being all the extremely many necks of image procossing species medically in basic subject or clinical practice
Domain.But because the treatment technology difficulty of medical image is big so that many treatment are extremely difficult to clinical practice degree.In recent years
Come, with the reduction of digital image processing apparatus Cheng Mu, improve all kinds of medical image qualities with digital image processing techniques and reached
To the practical stage.
PVR is the one of the main reasons of blinding.Additionally, hypertension, cerebrovascular sclerosis, coronary sclerosis etc.
Cardiovascular and cerebrovascular disease is the main cause that current China the elderly is dead and disables, and the tissue level of such disease injury is first
In microcirculation and the change of capilary level.Eye ground capilary be human body uniquely can with atraumatic directly observe compared with
The course of disease of the diseases such as the capilary of deep layer, its change degree and hypertension, the order of severity and more after situation it is closely related.Pass through
Inspection to retinal vasculature can be found that the diseases such as hypertension, diabetes, artery sclerosis.
Entered with the situation to PVR by checking and assessing retina color digital photo mostly in the prior art
Row judges, wastes time and energy very much.Therefore, one of existing technical problem urgently to be resolved hurrily is：How a kind of high efficiency, height are provided
The retinal images sorting technique of the degree of accuracy.
The content of the invention
In order to solve the above-mentioned technical problem, on the one hand the application provides a kind of retinal images sorting technique, including：
Obtain sample retinal images；
Extract the area-of-interest of the sample retinal images, and extract the optic disk region in the area-of-interest and
Angiosomes；
The image for extracting the optic disk region in the area-of-interest and the region outside the angiosomes is special
Levy, with according to described image features training Image Classifier；
Obtain targeted retinal image to be sorted；
The target region of interest of the targeted retinal image is extracted, and extracts regarding in the target region of interest
Disk area and angiosomes；
Extract the figure in the optic disk region in the target region of interest and the region outside the angiosomes
As feature, the targeted retinal image is classified with according to described image grader.
Alternatively, the area-of-interest for extracting the sample retinal images, including：
The corresponding binary image of the sample retinal images is obtained, and to binary image polarity expansion and corruption
Erosion is processed, to obtain the area-of-interest.
Alternatively, it is described to extract the optic disk region in the area-of-interest and the area outside the angiosomes
The characteristics of image in domain, with according to described image features training Image Classifier, including：
It is random to extract from the optic disk region in the area-of-interest and the region outside the angiosomes
Multiple images fritter；
The characteristics of image of the multiple image fritter is extracted according to neural network algorithm, and is determined according to described image feature
Final image feature, with according to the final image features training Image Classifier.
Alternatively, the optic disk region extracted in the area-of-interest, including：
Make gaussian pyramid decomposition to the area-of-interest, and in the afterbody image for obtaining is decomposed, with gray scale
Maximum point is optic disk center, determines the round region of target by radius of preset length, using the border circular areas as described
Optic disk region.
Alternatively, the sample retinal images are RGB triple channel images；
Correspondingly, the angiosomes extracted in the area-of-interest, including：
The green channel images I in the triple channel image is chosen firstg, to the green channel images IgCarry out out fortune
ObtainImage；
To describedImage carries out difference operation and medium filtering obtains IgoImage；
To the IgoImage carry out opening operation and with the IgoImage carries out difference operation and obtainsImage；
To describedImage is split using Ostu split plot designs, obtains the angiosomes.
Alternatively, described image grader includes softmax graders.
On the other hand, present invention also offers a kind of retinal images sorter, including：
Sample image acquiring unit, for obtaining sample retinal images；
Image-region extraction unit, for extracting the area-of-interest of the sample retinal images, and extracts the sense
Optic disk region and angiosomes in interest region；
Classifier training unit, for extracting the optic disk region and the angiosomes in the area-of-interest
Outside region characteristics of image, with according to described image features training Image Classifier；
Target image acquiring unit, for obtaining targeted retinal image to be sorted；
Target area acquiring unit, for extracting the target region of interest of the targeted retinal image, and extracts institute
State optic disk region and the angiosomes in target region of interest；
Target image taxon, for extracting the optic disk region and the blood in the target region of interest
The characteristics of image in the region outside the domain of area under control, classifies with according to described image grader to the targeted retinal image.
Alternatively, described image area extracting unit is specifically for obtaining the corresponding binaryzation of the sample retinal images
Image, and to binary image polarity expansion and corrosion treatment, to obtain the area-of-interest.
Alternatively, the classifier training unit specifically for：
It is random to extract from the optic disk region in the area-of-interest and the region outside the angiosomes
Multiple images fritter；
The characteristics of image of the multiple image fritter is extracted according to neural network algorithm, and is determined according to described image feature
Final image feature, with according to the final image features training Image Classifier.
Alternatively, described image area extracting unit is decomposed specifically for making gaussian pyramid to the area-of-interest,
And in the afterbody image for obtaining is decomposed, mesh is determined as optic disk center, by radius of preset length with the point that gray scale is maximum
Round region is marked, using the border circular areas as the optic disk region.
Retinal images sorting technique of the invention and device, by extracting the area-of-interest of sample retinal images,
And optic disk region and angiosomes in the area-of-interest are extracted, then extract the optic disk area in the area-of-interest
The characteristics of image in the region outside domain and the angiosomes, with according to described image features training Image Classifier, and then
The target region of interest of targeted retinal image is extracted, and extracts optic disk region and blood vessel in the target region of interest
Region, and extract the image in the optic disk region in the target region of interest and the region outside the angiosomes
Feature, classifies with according to described image grader to the targeted retinal image, can improve retinal images classification
Efficiency and the degree of accuracy, it is simple and easy to apply, it is applied widely.
Brief description of the drawings
In order to illustrate more clearly about the embodiment of the present invention or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are the present invention
Some embodiments, for those of ordinary skill in the art, on the premise of not paying creative work, can also basis
These accompanying drawings obtain other accompanying drawings.
Fig. 1 is the schematic flow sheet of the retinal images sorting technique of one embodiment of the invention；
Fig. 2 is the sample retinal images of one embodiment of the invention；
Fig. 3 is the area-of-interest schematic diagram of one embodiment of the invention；
Optic disk regions and angiosomes schematic diagram of the Fig. 4 for one embodiment of the invention；
Fig. 5 is the structural representation of the retinal images sorter of one embodiment of the invention.
Specific embodiment
To make the purpose, technical scheme and advantage of the embodiment of the present invention clearer, below in conjunction with the embodiment of the present invention
In accompanying drawing, the technical scheme in the embodiment of the present invention is explicitly described, it is clear that described embodiment be the present invention
A part of embodiment, rather than whole embodiments.Based on the embodiment in the present invention, those of ordinary skill in the art are not having
The every other embodiment obtained under the premise of creative work is made, the scope of protection of the invention is belonged to.
Fig. 1 is the schematic flow sheet of the retinal images sorting technique of one embodiment of the invention；As shown in figure 1, the party
Method includes：
S1：Obtain sample retinal images；
For example, the image of known classification of certain amount (such as N width) is chosen first as training image { (x(1),c(1)),...,(x(N),c(N)), wherein c=1,2,3,4,5 represent classification, roughly equal per class picture number.
S2：The area-of-interest of the sample retinal images is extracted, and extracts the optic disk area in the area-of-interest
Domain and angiosomes；
As the preferred of the present embodiment, the area-of-interest of the sample retinal images is extracted described in this step, can
To include：
The corresponding binary image of the sample retinal images is obtained, and to binary image polarity expansion and corruption
Erosion is processed, to obtain the area-of-interest.
Further, the optic disk area in the area-of-interest is extracted as the preferred of the present embodiment, described in this step
Domain, including：
Make gaussian pyramid decomposition to the area-of-interest, and in the afterbody image for obtaining is decomposed, with gray scale
Maximum point is optic disk center, determines the round region of target by radius of preset length, using the border circular areas as described
Optic disk region.
For example, to each image zooming-out region of interest.Color digital image is converted into gray level image, gray value G
Use with formula (1) computing：
G=0.2989*R+0.5870*G+0.1140*B (1)
On this basis, the minimum value of gray value is mapped to 0, maximum is mapped to 255, other value linear stretches, so
Statistic histogram afterwards；Obtain the corresponding number of pixels of the image each gray value；Between finding pixel c1~c2 again, number of pixels is most
Few corresponding gray value, the pixel value that will be greater than the gray value is set to 1,0 is set to less than the value of the pixel, then to the binary picture
As being expanded respectively and erosion operation, obtain area-of-interest (referring to Fig. 3)；Wherein, c1 is preferably the value between 5 to 10,
C2 is preferably the value between 40 to 80；
Optic disk extraction is carried out to described image；Specifically, gaussian pyramid decomposition is carried out to described image, and is being decomposed
To afterbody image in take the maximum point of gray scale for optic disk center (i.e. as the center of circle) makees the circle that radius is r, the circle
Region is optic disk region (referring to Fig. 4).
S3：Extract the image in the optic disk region in the area-of-interest and the region outside the angiosomes
Feature, with according to described image features training Image Classifier (such as softmax graders)；
Further, as the preferred of the present embodiment, the sample retinal images are RGB triple channel images；
Correspondingly, the angiosomes in the area-of-interest is extracted described in this step, can be included：
The green channel images I in the triple channel image is chosen firstg, to the green channel images IgCarry out out fortune
ObtainImage；
To describedImage carries out difference operation and medium filtering obtains IgoImage；
To the IgoImage carry out opening operation and with the IgoImage carries out difference operation and obtainsImage；
To describedImage is split using Ostu split plot designs, obtains the angiosomes.
Further, as the preferred of the present embodiment, that extracts in the area-of-interest described in step S3 described regards
The characteristics of image in the region outside disk area and the angiosomes, with according to described image features training Image Classifier,
Can also include：
It is random to extract from the optic disk region in the area-of-interest and the region outside the angiosomes
Multiple images fritter；
The characteristics of image of the multiple image fritter is extracted according to neural network algorithm, and is determined according to described image feature
Final image feature, with according to the final image features training Image Classifier.
S4：Obtain targeted retinal image to be sorted；
S5：The target region of interest of the targeted retinal image is extracted, and extracts the target region of interest
Optic disk region and angiosomes；
S6：Extract the optic disk region in the target region of interest and the region outside the angiosomes
Characteristics of image, classifies with according to described image grader to the targeted retinal image.
The present invention is illustrated with a specific embodiment below, but does not limit protection scope of the present invention.
The retinal images sorting technique of the present embodiment is comprised the following steps：
A1：Sample retinal images are obtained, wherein, described image is color digital image (referring to Fig. 2), with R, G, B tri-
Individual passage storage, is always divided into five classes.The image of known classification of certain amount (N) is chosen first as training image { (x(1),
c(1)),...,(x(N),c(N)), c=1,2,3,4,5 represent classification, and roughly equal per class picture number；
A2：To each image zooming-out region of interest.Color digital image is converted into gray level image, under gray value G is used
Face formula (1) computing：
G=0.2989*R+0.5870*G+0.1140*B (1)
The minimum value of gray value is mapped to 0, maximum is mapped to 255, then other value linear stretches count Nogata
Figure.Obtain the corresponding number of pixels of the image each gray value.Between finding pixel c1~c2, the minimum corresponding gray scale of number of pixels
Value, the pixel value that will be greater than the gray value is set to 1,0 is set to less than the value of the pixel, then the binary map is expanded respectively and rotten
Erosion computing, obtains area-of-interest (referring to Fig. 3), and c1 can be taken as the value between 5~10, and c2 can be taken as between 40~80
Value.
A3：Optic disk is extracted to each image；
Specifically, gaussian pyramid decomposition is carried out to described image, gray scale is taken in the afterbody image for obtaining is decomposed
Maximum point is optic disk center (as the center of circle), makees the circle that radius is r, and the border circular areas are optic disk region (referring in Fig. 4
Border circular areas).
A4：Vessel extraction is carried out to each image；
Specifically, green channel images I is chosen firstg, opening operation is carried out to it and is obtainedDifference operation is done to obtain simultaneously
Do medium filtering and obtain Igo；To IgoIt is opening operation and and IgoDifference operation is done to obtainOstu split plot designs are used to split to it
Obtain segmentation result (referring to Fig. 4).
A5：Clipping image background area, and scale it into less image sx×sy× 3, such as 128x128x3.
A6：Obtain image fritter；Specifically, for the image that each reduces, optic disk and blood vessel are avoided in region of interest
Region, extracts n image fritter at random, and its size is px×py× 3, n are more than 100.
A7：Sample image to all of training grader carries out the treatment of step A2~A6, obtains substantial amounts of small image
Block；
A8：The small image block learning feature obtained from step A7；
Specifically, using 3 layers of neutral net, ground floor and the pixel summation that third layer number of unit is each image fritter
(Np=px×py× 3), number of unit is Y in the second layer, and less than 1000, therefore the data of ground floor unit are image fritter
Pixel value xi, then by weightsAnd side-play amountAs the input of the second layer, exported accordinglyWherein, f is sigmoid functions；Using the value as the input of third layer, weighted sum is also passed through
Calculations of offset is exportedTherefore total training parameter is 2Np×Y+Y+Np, the seasonal output valve of training
Being approximately equal to input value, i.e. cost function is：It is optimized and obtains corresponding parameter value；
A9：Feature extraction；Specifically, the W that step A8 is obtained1, this feature size is px×py×3×Y.For step A5
The image for obtaining, to each of which fritter px×pyConvolution is done with features described above, image size is obtained for (sx-px)×(sy-py)+1,
Convolution feature sizes are (sx-px+1)×(sy-py+1)×Y；It is the nonoverlapping of m × n that this feature is divided into several sizes
Region, takes the average in the region for final feature, and its size is Θ=floor [(sx-px+1)/m]×floor[(sy-py+1)/
n]×Y；
A10：The final feature obtained using step A9, trains softmax graders, and corresponding cost function is：
Training parameter number c × Θ, λ are a constant；{ value is false expression for wherein 1 { value is genuine expression formula }=1,1
Formula }=0；Optimize the cost function and reach minimum value, obtain the parameters of grader
A11：Need to judge the image I of classification for one, processed by step A2 to A8, obtain corresponding feature y,
Length is Θ, calculates θoptY, the corresponding classification of its maximum is classification results.
Fig. 5 is the structural representation of the retinal images sorter of one embodiment of the invention；As shown in figure 5, the dress
Put including：
Sample image acquiring unit 10, for obtaining sample retinal images；
Image-region extraction unit 20, the area-of-interest for extracting the sample retinal images, and extract described
Optic disk region and angiosomes in area-of-interest；
Classifier training unit 30, for extracting the optic disk region and the area vasculosa in the area-of-interest
The characteristics of image in the region outside domain, with according to described image features training Image Classifier；
Target image acquiring unit 40, for obtaining targeted retinal image to be sorted；
Target area acquiring unit 50, for extracting the target region of interest of the targeted retinal image, and extracts
Optic disk region and angiosomes in the target region of interest；
Target image taxon 60, for extracting the optic disk region in the target region of interest and described
The characteristics of image in the region outside angiosomes, is divided the targeted retinal image with according to described image grader
Class.
Device described in the present embodiment can be used for performing above method embodiment, and its principle is similar with technique effect, this
Place repeats no more.
Used as the preferred of the present embodiment, described image area extracting unit 20 can be specifically for obtaining the sample view
The corresponding binary image of film image, and to binary image polarity expansion and corrosion treatment, it is described interested to obtain
Region.
Used as the preferred of the present embodiment, the classifier training unit 30 can be specifically for：
It is random to extract from the optic disk region in the area-of-interest and the region outside the angiosomes
Multiple images fritter；
The characteristics of image of the multiple image fritter is extracted according to neural network algorithm, and is determined according to described image feature
Final image feature, with according to the final image features training Image Classifier.
Used as the preferred of the present embodiment, described image area extracting unit 20 can also be specifically for the region of interest
Gaussian pyramid decomposition is made in domain, and in the afterbody image for obtaining is decomposed, with the maximum point of gray scale as optic disk center, with pre-
If length determines the round region of target for radius, using the border circular areas as the optic disk region.
Device described in the present embodiment can be used for performing above method embodiment, and its principle is similar with technique effect, this
Place repeats no more.
It should be noted that for device embodiment, because it is substantially similar to embodiment of the method, so description
Fairly simple, the relevent part can refer to the partial explaination of embodiments of method.
Above example is merely to illustrate technical scheme, rather than its limitations；Although with reference to the foregoing embodiments
The present invention has been described in detail, it will be understood by those within the art that：It still can be to foregoing each implementation
Technical scheme described in example is modified, or carries out equivalent to which part technical characteristic；And these are changed or replace
Change, do not make the spirit and scope of the essence disengaging various embodiments of the present invention technical scheme of appropriate technical solution.
Claims (10)
1. a kind of retinal images sorting technique, it is characterised in that including：
Obtain sample retinal images；
The area-of-interest of the sample retinal images is extracted, and extracts optic disk region and blood vessel in the area-of-interest
Region；
The characteristics of image in the optic disk region in the area-of-interest and the region outside the angiosomes is extracted, with
According to described image features training Image Classifier；
Obtain targeted retinal image to be sorted；
The target region of interest of the targeted retinal image is extracted, and extracts the optic disk area in the target region of interest
Domain and angiosomes；
The image for extracting the optic disk region in the target region of interest and the region outside the angiosomes is special
Levy, the targeted retinal image is classified with according to described image grader.
2. method according to claim 1, it is characterised in that the region of interest of the extraction sample retinal images
Domain, including：
Obtain the corresponding binary image of the sample retinal images, and to binary image polarity expansion and corrosion at
Reason, to obtain the area-of-interest.
3. method according to claim 1, it is characterised in that the optic disk area in the extraction area-of-interest
The characteristics of image in the region outside domain and the angiosomes, with according to described image features training Image Classifier, including：
It is random to extract multiple from the optic disk region in the area-of-interest and the region outside the angiosomes
Image fritter；
The characteristics of image of the multiple image fritter is extracted according to neural network algorithm, and is determined according to described image feature final
Characteristics of image, with according to the final image features training Image Classifier.
4. method according to claim 1, it is characterised in that the optic disk region in the extraction area-of-interest,
Including：
Make gaussian pyramid decomposition to the area-of-interest, and in the afterbody image for obtaining is decomposed, it is maximum with gray scale
Point for optic disk center, the round region of target is determined by radius of preset length, using the border circular areas as the optic disk
Region.
5. method according to claim 1, it is characterised in that the sample retinal images are RGB triple channel images；
Correspondingly, the angiosomes extracted in the area-of-interest, including：
The green channel images I in the triple channel image is chosen firstg, to the green channel images IgOpening operation is carried out to obtain
ArriveImage；
To describedImage carries out difference operation and medium filtering obtains IgoImage；
To the IgoImage carry out opening operation and with the IgoImage carries out difference operation and obtainsImage；
To describedImage is split using Ostu split plot designs, obtains the angiosomes.
6. method according to claim 1, it is characterised in that described image grader includes softmax graders.
7. a kind of retinal images sorter, it is characterised in that including：
Sample image acquiring unit, for obtaining sample retinal images；
Image-region extraction unit, for extracting the area-of-interest of the sample retinal images, and extracts described interested
Optic disk region and angiosomes in region；
Classifier training unit, for extracting the optic disk region and the angiosomes in the area-of-interest outside
Region characteristics of image, with according to described image features training Image Classifier；
Target image acquiring unit, for obtaining targeted retinal image to be sorted；
Target area acquiring unit, for extracting the target region of interest of the targeted retinal image, and extracts the mesh
Optic disk region and angiosomes in mark area-of-interest；
Target image taxon, for extracting the optic disk region and the area vasculosa in the target region of interest
The characteristics of image in the region outside domain, classifies with according to described image grader to the targeted retinal image.
8. device according to claim 7, it is characterised in that described image area extracting unit is described specifically for obtaining
The corresponding binary image of sample retinal images, and to binary image polarity expansion and corrosion treatment, to obtain
State area-of-interest.
9. device according to claim 7, it is characterised in that the classifier training unit specifically for：
It is random to extract multiple from the optic disk region in the area-of-interest and the region outside the angiosomes
Image fritter；
The characteristics of image of the multiple image fritter is extracted according to neural network algorithm, and is determined according to described image feature final
Characteristics of image, with according to the final image features training Image Classifier.
10. device according to claim 7, it is characterised in that described image area extracting unit is specifically for described
Area-of-interest makees gaussian pyramid decomposition, and in the afterbody image for obtaining is decomposed, with the maximum point of gray scale as optic disk
Center, the round region of target is determined by radius of preset length, using the border circular areas as the optic disk region.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611248700.8A CN106846301B (en) | 2016-12-29 | 2016-12-29 | Retina image classification method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201611248700.8A CN106846301B (en) | 2016-12-29 | 2016-12-29 | Retina image classification method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106846301A true CN106846301A (en) | 2017-06-13 |
| CN106846301B CN106846301B (en) | 2020-06-23 |
Family
ID=59113738
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201611248700.8A Active CN106846301B (en) | 2016-12-29 | 2016-12-29 | Retina image classification method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106846301B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109685770A (en) * | 2018-12-05 | 2019-04-26 | 合肥奥比斯科技有限公司 | Retinal vessel curvature determines method |
| CN110914835A (en) * | 2017-07-28 | 2020-03-24 | 新加坡国立大学 | Method for modifying retinal fundus images for a deep learning model |
| CN111000563A (en) * | 2019-11-22 | 2020-04-14 | 北京理工大学 | Automatic measuring method and device for retinal artery and vein diameter ratio |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070109499A1 (en) * | 2005-10-12 | 2007-05-17 | Siemens Corporate Research Inc | System and Method For Robust Optic Disk Detection In Retinal Images Using Vessel Structure And Radon Transform |
| US20110170751A1 (en) * | 2008-01-16 | 2011-07-14 | Rami Mangoubi | Systems and methods for detecting retinal abnormalities |
| CN102843957A (en) * | 2009-08-24 | 2012-12-26 | 新加坡保健服务集团有限公司 | Method and system for detecting disc haemorrhages |
| CN103870838A (en) * | 2014-03-05 | 2014-06-18 | 南京航空航天大学 | Eye fundus image characteristics extraction method for diabetic retinopathy |
| CN104573712A (en) * | 2014-12-31 | 2015-04-29 | 浙江大学 | Arteriovenous retinal blood vessel classification method based on eye fundus image |
| CN105513077A (en) * | 2015-12-11 | 2016-04-20 | 北京大恒图像视觉有限公司 | System for screening diabetic retinopathy |
| CN105913025A (en) * | 2016-04-12 | 2016-08-31 | 湖北工业大学 | Deep learning face identification method based on multiple-characteristic fusion |
| CN106056595A (en) * | 2015-11-30 | 2016-10-26 | 浙江德尚韵兴图像科技有限公司 | Method for automatically identifying whether thyroid nodule is benign or malignant based on deep convolutional neural network |
-
2016
- 2016-12-29 CN CN201611248700.8A patent/CN106846301B/en active Active
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070109499A1 (en) * | 2005-10-12 | 2007-05-17 | Siemens Corporate Research Inc | System and Method For Robust Optic Disk Detection In Retinal Images Using Vessel Structure And Radon Transform |
| US20110170751A1 (en) * | 2008-01-16 | 2011-07-14 | Rami Mangoubi | Systems and methods for detecting retinal abnormalities |
| CN102843957A (en) * | 2009-08-24 | 2012-12-26 | 新加坡保健服务集团有限公司 | Method and system for detecting disc haemorrhages |
| CN103870838A (en) * | 2014-03-05 | 2014-06-18 | 南京航空航天大学 | Eye fundus image characteristics extraction method for diabetic retinopathy |
| CN104573712A (en) * | 2014-12-31 | 2015-04-29 | 浙江大学 | Arteriovenous retinal blood vessel classification method based on eye fundus image |
| CN106056595A (en) * | 2015-11-30 | 2016-10-26 | 浙江德尚韵兴图像科技有限公司 | Method for automatically identifying whether thyroid nodule is benign or malignant based on deep convolutional neural network |
| CN105513077A (en) * | 2015-12-11 | 2016-04-20 | 北京大恒图像视觉有限公司 | System for screening diabetic retinopathy |
| CN105913025A (en) * | 2016-04-12 | 2016-08-31 | 湖北工业大学 | Deep learning face identification method based on multiple-characteristic fusion |
Non-Patent Citations (5)
| Title |
|---|
| MARC LALONDE 等: "Fast and Robust Optic Disc Detection Using Pyramidal Decomposition and Hausdorff-Based Template Matching", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| MOHD HANAFI AHMAD HIJAZI 等: "Retinal Image Classification using a Histogram Based Approach", 《THE 2010 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS》 * |
| SAHINAZ SAFARI SANJANI 等: "Blood Vessel Segmentation in Retinal Fundus Images", 《HTTPS://WEB.STANFORD.EDU/~JBBOIN/DOC/VESSEL_SEGMENTATION_REPORT.PDF》 * |
| SWATI GUPTA 等: "Diagnosis of Diabetic Retinopathy using Machine Learning", 《JOURNAL OF RESEARCH AND DEVELOPMENT》 * |
| 高玮玮 等: "基于多模板匹配的局部自适应区域生长法在视网膜内出血自动检测中的应用", 《光谱学与光谱分析》 * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110914835A (en) * | 2017-07-28 | 2020-03-24 | 新加坡国立大学 | Method for modifying retinal fundus images for a deep learning model |
| CN113284101A (en) * | 2017-07-28 | 2021-08-20 | 新加坡国立大学 | Method for modifying retinal fundus images for a deep learning model |
| CN110914835B (en) * | 2017-07-28 | 2024-04-19 | 新加坡国立大学 | Method for modifying retinal fundus image for deep learning model |
| CN109685770A (en) * | 2018-12-05 | 2019-04-26 | 合肥奥比斯科技有限公司 | Retinal vessel curvature determines method |
| CN109685770B (en) * | 2018-12-05 | 2020-10-09 | 合肥奥比斯科技有限公司 | Method for determining retinal vascular tortuosity |
| CN111000563A (en) * | 2019-11-22 | 2020-04-14 | 北京理工大学 | Automatic measuring method and device for retinal artery and vein diameter ratio |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106846301B (en) | 2020-06-23 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108021916B (en) | Deep learning diabetic retinopathy sorting technique based on attention mechanism | |
| Rezaoana et al. | Detection and classification of skin cancer by using a parallel CNN model | |
| Soomro et al. | Boosting sensitivity of a retinal vessel segmentation algorithm with convolutional neural network | |
| CN108095683A (en) | The method and apparatus of processing eye fundus image based on deep learning | |
| Karegowda et al. | Exudates detection in retinal images using back propagation neural network | |
| Shahin et al. | Automated detection of diabetic retinopathy in blurred digital fundus images | |
| Lau et al. | Automatically early detection of skin cancer: Study based on nueral netwok classification | |
| Beaufort et al. | Automatic recognition of coccoliths by dynamical neural networks | |
| CN109087302A (en) | A kind of eye fundus image blood vessel segmentation method and apparatus | |
| CN107871318A (en) | A kind of coronary calcification plaque detection method based on model migration | |
| CN110276356A (en) | Eye fundus image aneurysms recognition methods based on R-CNN | |
| Min et al. | Fully automatic computer-aided mass detection and segmentation via pseudo-color mammograms and mask r-cnn | |
| CN108185984A (en) | The method that eyeground color picture carries out eyeground lesion identification | |
| CN111815562B (en) | Retina blood vessel segmentation method combining U-Net and self-adaptive PCNN | |
| CN107330876A (en) | A kind of image automatic diagnosis method based on convolutional neural networks | |
| CN107704886A (en) | A kind of medical image hierarchy system and method based on depth convolutional neural networks | |
| Aravind et al. | Automatic detection of microaneurysms and classification of diabetic retinopathy images using SVM technique | |
| Abiyev et al. | Fuzzy neural networks for identification of breast cancer using images' shape and texture features | |
| CN106846301A (en) | Retinal images sorting technique and device | |
| Yang et al. | Classification of diabetic retinopathy severity based on GCA attention mechanism | |
| Junjun et al. | Diabetic retinopathy detection based on deep convolutional neural networks for localization of discriminative regions | |
| Sreng et al. | Automatic exudate extraction for early detection of Diabetic Retinopathy | |
| Wankhede et al. | Retinal blood vessel segmentation using graph cut analysis | |
| Acharya et al. | Swarm intelligence based adaptive gamma corrected (SIAGC) retinal image enhancement technique for early detection of diabetic retinopathy | |
| Zhang et al. | Hierarchical detection of red lesions in retinal images by multiscale correlation filtering |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |