CN109544525A - A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching - Google Patents
A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching Download PDFInfo
- Publication number
- CN109544525A CN109544525A CN201811357420.XA CN201811357420A CN109544525A CN 109544525 A CN109544525 A CN 109544525A CN 201811357420 A CN201811357420 A CN 201811357420A CN 109544525 A CN109544525 A CN 109544525A
- Authority
- CN
- China
- Prior art keywords
- gaussian
- blood vessel
- self
- scale
- gaussian kernel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 210000004204 blood vessel Anatomy 0.000 title claims abstract description 52
- 238000000034 method Methods 0.000 title claims abstract description 46
- 238000000605 extraction Methods 0.000 claims abstract description 14
- 239000000284 extract Substances 0.000 claims abstract description 9
- 239000011159 matrix material Substances 0.000 claims abstract description 6
- 230000000877 morphologic effect Effects 0.000 claims description 5
- 230000003044 adaptive effect Effects 0.000 abstract description 9
- 230000008859 change Effects 0.000 abstract description 6
- 230000000694 effects Effects 0.000 abstract description 6
- 230000002792 vascular Effects 0.000 abstract description 4
- 230000008569 process Effects 0.000 description 6
- 230000011218 segmentation Effects 0.000 description 6
- 238000012549 training Methods 0.000 description 4
- 238000012545 processing Methods 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 206010020772 Hypertension Diseases 0.000 description 2
- 210000001367 artery Anatomy 0.000 description 2
- 230000008901 benefit Effects 0.000 description 2
- 206010012601 diabetes mellitus Diseases 0.000 description 2
- 201000010099 disease Diseases 0.000 description 2
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 2
- 238000001914 filtration Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 238000002577 ophthalmoscopy Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 208000024172 Cardiovascular disease Diseases 0.000 description 1
- 208000002177 Cataract Diseases 0.000 description 1
- 208000034189 Sclerosis Diseases 0.000 description 1
- 230000033115 angiogenesis Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 208000001969 capillary hemangioma Diseases 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002059 diagnostic imaging Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 210000000416 exudates and transudate Anatomy 0.000 description 1
- 230000001631 hypertensive effect Effects 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 210000003734 kidney Anatomy 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 201000001119 neuropathy Diseases 0.000 description 1
- 230000007823 neuropathy Effects 0.000 description 1
- 208000033808 peripheral neuropathy Diseases 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 230000004256 retinal image Effects 0.000 description 1
- 230000004233 retinal vasculature Effects 0.000 description 1
- 210000001210 retinal vessel Anatomy 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 210000003462 vein Anatomy 0.000 description 1
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
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
- G06T5/92—Dynamic range modification of images or parts thereof based on global image properties
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/136—Segmentation; Edge detection involving thresholding
-
- 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
- G06T2207/20036—Morphological image processing
-
- 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)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Eye Examination Apparatus (AREA)
- Image Analysis (AREA)
Abstract
The present invention discloses a kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching, comprising: 1) extracts the green channel of colored eyeground picture, the pretreatment that smooth, gray value reverses by mean filter；2) self-adapting window model adapts to the variation in blood vessel fucoid direction by 12 directions, the variation of blood vessel fucoid width is adapted to by the window width of dynamic change, Gaussian scale-space is constructed using Gaussian function standard deviation as scale factor on 12 directions, then each Gaussian kernel in Gaussian scale-space and pre-processed results matching, corresponding position pixel value in finally all matching results, which is compared, to be adaptive selected maximum value and carrys out constitutive characteristic matrix, realizes feature extraction；3) bianry image is converted by self-adaption thresholding to reflect vascular distribution.The present invention combines to identify that blood vessel, recognition effect are more preferable with self-adaption thresholding using self-adapting window model.
Description
Technical field
The invention belongs to computer fields, propose a kind of eyeground picture blood vessel identification based on self-adapting window Model Matching
Method can extract eyeground picture blood vessel network well.
Background technique
When we see a doctor to hospital, especially when we be with kidney trouble, diabetes, neuropathy, artery sclerosis,
When the patient of the disease such as cataract, hypertension, oculist often helps us to check eyeground, and funduscopy is for understanding the state of an illness
Very important, many diseases can be by checking that funduscopy comes out.We will mainly observe view in fact at observing eye bottom
Retinal vasculature, because retinal vessel is uniquely can be in order to avoid sore hurts the blood vessel observed, to its observation to me in human body
Check that cardiovascular disease has great importance.Such as we observe the eyeground of hypertensive patient and are observed that retina is dynamic
Arteries and veins hardening, the eyeground for observing diabetic can see capillary hemangioma, some exudates, blutpunkte, can therefrom learn
Other physical feeling blood vessel situations, doctor can understand the state of an illness of patient according to these indexs.
Medical image is a part of medical big data, and Medical Image Processing also just becomes the important of medical big data research
Direction, carrying out the identification of eyeground picture blood vessel using computerized algorithm has very big clinical meaning, but carries on the back in retinal image
The contrast of scape and blood vessel structure is low, and blood vessel fucoid direction is indefinite, these eyeground picture blood vessel structures of the variation of blood vessel diameter
Inherent characteristics carry out very big challenge to this work belt.
There are many methods to identify for eyeground picture medium vessels at present, these methods can be broadly divided into supervised learning side
Method and unsupervised learning method.This two major classes method is very different in terms of mechanism of classifying.Specifically in supervised learning side
Image segmentation algorithm needs to learn the various rules to pixel classifications in picture by training dataset in method, so in the method
The criticality of middle training data is self-evident, it might even be possible to say that the order of accuarcy of training data mark directly affects supervised segmentation
The segmentation effect of method；And algorithm can extract the natural mode in the picture of eyeground in non-supervisory dividing method, pass through these moulds
Formula may determine that whether certain pixel belongs to blood vessel or non-vascular.Non-supervisory method does not need to instruct in advance compared with supervised segmentation method
Practice, so not needing training set, it will usually have higher execution speed and robustness.
1989 by S Chaudhuri, the IEEE of S Chatterjee, N Katz and M Nelson invention
TRANSACTIONS ON MEDICAL IMAGING paper " Detection of Blood Vessels in Retinal
It proposes in Images Using Two-Dimensional Matched Filters. " a kind of based on two-dimentional matched filtering
Eyeground picture blood vessel segmentation algorithm is extracted the green channel images of eyeground picture in this method first, then uses the mean value of 5*5
Filter noise reduction, the pretreatment link to this this method terminate.With a kind of method based on two-dimentional matched filtering after pretreatment
Blood vessel segmentation is carried out, this method considers three useful features of eyeground picture, is respectively: 1) vessel boundary is substantially in two
Parallel lines can regard two parallel lines as in the case where short enough, so vessel boundary is in piecewise linearity structure.2) with
Grey scale pixel value curve approximation on the vertical direction in blood vessel track on blood vessel track is in normal distribution curve.3) in eyeground figure
In piece, the direction of fucoid is variation, and the width of blood vessel is in 2~10pixels.Considered based on three above, is proposed a kind of solid
The Gaussian kernel of window is determined to match the angiogenesis eigenmatrix in the picture of eyeground, and threshold is then carried out using fixed global threshold
Value is handled to obtain two-value picture, which can show the blood vessel structure in the picture of eyeground.
However, the algorithm that S Chaudhuri et al. is proposed extraordinary can not extract blood vessel structure, their method
An original hypothesis be the blood vessel in very short range diameter may be considered it is constant, thus by the diameter of all blood vessels
It is assumed to be 2 σ (σ is fixed empirical value), however blood vessel diameter is but coarse and fine, this just largely affects of model
With effect.The blood vessel structure in the picture of eyeground can be more accurately matched using self-adapting window model, improve the identification of algorithm
Effect.
Summary of the invention
Other than three useful features in Chaudhuri et al. method it is contemplated that in the picture of eyeground, although
It is considered that the diameter of blood vessel and direction are constant in the case where short enough, but it is big in the diameter of entire eyeground picture medium vessels
Small and distribution arrangement is not fixed and invariable.Based on considerations above, the present invention proposes a kind of based on self-adapting window model
The eyeground picture blood vessel recognition methods matched.
To achieve the goals above, present invention employs following technical solutions:
A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching the following steps are included:
S1, read in the colour eyeground RGB picture, len, min, max, step, wherein len be may be considered it is linear most
The minimum value of long blood vessel segment length, min σjInitial value, i.e. minimum value, max σjStop value, i.e. maximum value, step be
Step-length, wherein the subscript j of σ marks different scale factors；
S2, RGB image turn gray scale；
S3, morphological operation are completed to this pretreatment, obtain feature extraction algorithm input picture；
S4, feature extraction generate eigenmatrix；
S5, self-adaption thresholding realize that blood vessel structure extracts.
Preferably, the characteristic extraction procedure based on self-adapting window Model Matching the following steps are included:
(1), Gaussian kernel k is constructedi,j(x,y,σj)=exp (- μ2/2*σj 2) be a M*N matrix, the wherein subscript i of k,
J marks different Gaussian kernels in the Gaussian scale-space and Gaussian scale-space of different directions respectively, and x, y are the position rope of Gaussian kernel
Draw, σjFor the standard deviation criteria of Gaussian kernel, μ is Gaussian kernel independent variable, is overlapped on the corresponding straight line vertical with fucoid with fucoid
Position where pixel on a neighborhood (line segment) for line segment, M=len, N=6* σj, we are in | μ |=3* σjIt clips at place
The unlimited long-tail of Gaussian curve makes | μ | < 3* σjThe width of Gaussian kernel is thus determined, at this time Degree (ki,j())=0 °, wherein
Degree () takes the direction (degree) of Gaussian kernel.
(2)、σjIt is initialized as min, with σjGaussian scale-space is constructed for scale factor, wherein σj=σj-1+ step, until
σjUntil=max；
(3), each Gaussian kernel does matching operation in pre-processed results and Gaussian scale-space；
(4), Gaussian kernel presses Degree (ki,j())=Degree (ki-1,j())+15 ° of rotations；
(5), step 2)~4 are repeated), 12 different directions are shared until Degree ()=180 °；
(6), corresponding in the matching result of each Gaussian kernel in the Gaussian scale-space in 12 directions and pre-processed results
The pixel of position is compared, and selects the maximum value of pixel value as eigenmatrix element, this process is adaptive, because calculating
Method automatically selects element of the best match of fucoid and model as eigenmatrix in all matching results.
Preferably, the thresholding in step S6, which handles us, selects otsu thresholding, using the side of maximum variance between class
Method threshold value.
A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching of the invention, comprising: 1) carry out
Extract the green channel of colored eyeground picture, the pretreatment that smooth, gray value reverses by mean filter；2) self-adapting window
Model adapts to the variation in blood vessel fucoid direction by 12 directions, and blood vessel fucoid is adapted to by the window width of dynamic change
The variation of width constructs Gaussian scale-space by scale factor of Gaussian function standard deviation on 12 directions, then Gauss
Each Gaussian kernel in scale space and pre-processed results matching, corresponding position pixel value in finally all matching results into
Row is relatively adaptive selected maximum value and carrys out constitutive characteristic matrix, realizes feature extraction；3) it is converted by self-adaption thresholding
Bianry image reflects vascular distribution.The present invention is combined with self-adaption thresholding using self-adapting window model to identify blood
Pipe, recognition effect are more preferable.
Compared with prior art, the invention has the following advantages that
Self-adapting window Model Matching algorithm proposed by the present invention not only allows for eyeground picture medium vessels fucoid direction
Uncertainty, and self-adapting window model can match the blood vessel pipe of different in width with the model of dynamics change window size
Mark, with fucoid formed it is better couple, solve in existing method due to blood vessel fucoid width is different and bring identification fucoid
Cross problem thick and that identification is not complete.Present invention incorporates self-adaption thresholding processing, can effectively avoid be in existing method
Selection suitable global threshold and the multiple trial carried out.The present invention appoints in the identification of eyeground picture blood vessel compared with the conventional method
There is better performance in business.
Detailed description of the invention
For a better understanding of the present invention, it is described in detail in conjunction with attached drawing, in which:
Fig. 1 is flow chart of the invention.
In attached drawing: 1, reading in colour RGB eye fundus image and parameter；2, RGB turns gray scale；3, morphological operation；4, building is high
This core；5, Gaussian scale-space is constructed；6, pre-processed results match in entire Gaussian scale-space；7, in 0 °~180 ° ranges
Change Gaussian kernel direction and carries out loop control；8, adaptive features select；9, self-adaption thresholding, final obtain extract result.
Specific embodiment
Each aspect of the present invention feature and illustrative examples will be described from now on.This can be realized with MATLAB
Corresponding software is invented, MATLAB has very big advantage to matrix data processing.
The present invention is a kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching, can be specifically divided into
Three pretreatment, feature extraction, classification modules.Wherein, the purpose of preprocessing process is to improve eyeground picture medium vessels and background
Contrast, be characterized extraction and prepare.The present invention uses self-adapting window Model Matching blood vessel in characteristic extraction procedure, can be with
Direction and the width for better adapting to the variation of eyeground picture medium vessels fucoid are exactly further widening blood vessel on this process nature
Facilitate subsequent carry out Threshold segmentation with the contrast of background.Classified using self-adaption thresholding, by eigenmatrix be divided into blood vessel and
Background two parts.
Pretreated detailed process are as follows: the green channel images for extracting image first, experiments have shown that the image is relative to it
There is higher contrast in his channel；Then the influence of noise is reduced using the mean filter noise reduction of 5*5；Finally carry out gray scale
Value reverses.
Above-mentioned preprocessing process includes following feature:
Here joined in prior art basis gray value reverse operation, make perpendicular on blood vessel fucoid direction with blood vessel
The grey scale pixel value curve that fucoid is overlapped can be similar to normal distribution curve.
The detailed process of feature extraction are as follows: 1) construct Gaussian kernel；2) poor with Gaussian function parameter and standard on 12 directions
Gaussian scale-space is constructed for scale factor；3) each Gaussian kernel in pre-processed results and the Gaussian scale-space in 12 directions
Matching；4) adaptive method construct eigenmatrix is used.
Features described above extraction algorithm includes following feature:
Building and the variation that fucoid direction is not only only accounted for when the matched model of blood vessel fucoid in the picture of eyeground, are also examined
The variation in entire eyeground picture medium vessels fucoid width is considered, has proposed self-adapting window model on this basis and come preferably
Match direction and all uncertain fucoid of width in the picture of eyeground.
More fully, the eyeground picture feature extraction process based on self-adapting window Model Matching is completed by following steps:
(1), Gaussian kernel is constructed and when 0 ° of direction is in pronucleus direction；
(2), with the poor (σ of Gaussian kernel parameter and standardj) it is that scale factor constructs Gaussian scale-space, σjIt is initialized as min, is pressed
σj=σj-1+ step increases, until σjUntil=max；
(3), all Gaussian kernels in pre-processed results image and current scale space are matched；
(4), Gaussian kernel presses Degree (ki,j())=Degree (ki-1,j())+15 ° of rotations；
(5) if, Degree (ki,j() < 180 ° repetition step 2)~4), it otherwise carries out in next step；
(6), select maximum value as the final position pixel value, structure in the corresponding position pixel value of all result pictures
At eigenmatrix；
Classification realizes binaryzation, including following feature using self-adaption thresholding method:
Global threshold is generated using adaptive otsu thresholding, and then carry out this result of binaryzation can clearly indicate
Vascular distribution.
A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching, as shown in Fig. 1.It include: to read
Enter colored RGB eye fundus image and parameter (1)；RGB turns gray scale (2)；Morphological operation (3)；It constructs Gaussian kernel (4)；Construct Gauss
Scale space (5)；Pre-processed results match (6) in entire Gaussian scale-space；Change Gaussian kernel within the scope of 0 °~180 °
Direction carries out loop control (7)；Adaptive features select (8)；Self-adaption thresholding (9)；Final obtain extracts result.
Read in the colour eyeground RGB picture and parameter (1): reading in picture is the colored jpg format chart shot by fundus camera
Piece, this example need parameter and its value: min=1, step=1, max=10, len=19.
RGB turns gray scale (2): selecting green channel image as gray level image.
Morphological operation (3): with the mean filter of 5*5 to picture noise reduction, carry out gray value reverse.
It constructs Gaussian kernel (4): building Gaussian kernel ki,j(x,y,σj)=exp (- μ2/2*σj 2) be a M*N matrix,
Middle M=len, N=6* σj, we are in | μ |=3* σjPlace, which clips the unlimited long-tail of Gaussian curve, to be made | μ | < 3* σjHeight has thus been determined
The width of this core, at this time Degree (ki,j())=0 °, wherein Degree () takes the direction (degree) of Gaussian kernel.
It constructs Gaussian scale-space (5): σjIt is initialized as min, with σjGaussian scale-space is constructed for scale factor, wherein
σj=σj-1+ step, until σjUntil=max；
Pre-processed results match (6) in entire Gaussian scale-space: each in pre-processed results and Gaussian scale-space
Gaussian kernel carries out convolution.
Change Gaussian kernel direction within the scope of 0 °~180 ° and carry out loop control (7): Gaussian kernel presses Degree (ki,j())=
Degree(ki-1,j())+15 ° of rotations.
If Degree (ki,j()) < 180 °, repeat 5)~7), otherwise continue to the next step.
Adaptive features select (8): of each Gaussian kernel and pre-processed results in the Gaussian scale-space in 12 directions
With in result, the pixel of corresponding position is compared, and the maximum value of pixel value is adaptive selected as eigenmatrix element.
Self-adaption thresholding (9): obtaining bianry image using ostu thresholding, and wherein blood vessel network is white, and background is in
Black.
The present invention can identify the blood vessel of eyeground picture more accurately, even if picture quality is poor, picture comparison
Spending in lower eyeground picture still has relatively good recognition effect.
Claims (3)
1. a kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching, it is characterised in that: including following step
It is rapid:
S1, the colour eyeground RGB picture, len, min, max, step are read in, wherein len is linear longest blood vessel segment length
Minimum value, min σjInitial value, i.e. minimum value, max σjStop value, i.e. maximum value, step is step-length, wherein σ
Subscript j marks different scale factors；σjFor the standard deviation criteria of Gaussian kernel；
S2, RGB image turn gray scale；
S3, morphological operation are completed to this pretreatment, obtain feature extraction algorithm input picture；
S4, feature extraction generate eigenmatrix；
S5, self-adaption thresholding realize that blood vessel structure extracts.
2. the method according to claim 1, wherein characteristic extraction procedure the following steps are included:
(1), Gaussian kernel k is constructedi,j(x,y,σj)=exp (- μ2/2*σj 2) be a M*N matrix, the wherein subscript i of k, j difference
Different Gaussian kernels in the Gaussian scale-space and Gaussian scale-space of different directions are marked, x, y are the location index of Gaussian kernel, σj
For the standard deviation criteria of Gaussian kernel, μ is Gaussian kernel independent variable, the pixel institute being overlapped on the corresponding straight line vertical with fucoid with fucoid
Position on an a neighborhood i.e. line segment for line segment, M=len, N=6* σj, | μ |=3* σjPlace clip Gaussian curve without
Limit for length's tail makes | μ | < 3* σjThe width of Gaussian kernel is thus determined, at this time Degree (ki,j())=0 °, wherein Degree ()
Take the direction of Gaussian kernel；
(2)、σjIt is initialized as min, with σjGaussian scale-space is constructed for scale factor, wherein σj=σj-1+ step, until σj=
Until max；
(3), each Gaussian kernel does matching operation in pre-processed results and Gaussian scale-space；
(4), Gaussian kernel presses Degree (ki,j())=Degree (ki-1,j())+15 ° of rotations；
(5), step (2)~(4) are repeated, 12 different directions are shared until Degree ()=180 °；
(6), in the matching result of each Gaussian kernel in the Gaussian scale-space in 12 directions and pre-processed results, corresponding position
Pixel be compared, select the maximum value of pixel value as eigenmatrix element.
3. the method according to claim 1, wherein thresholding selects otsu thresholding, using most generous between class
The method threshold value of difference.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811357420.XA CN109544525A (en) | 2018-11-15 | 2018-11-15 | A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811357420.XA CN109544525A (en) | 2018-11-15 | 2018-11-15 | A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109544525A true CN109544525A (en) | 2019-03-29 |
Family
ID=65847656
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811357420.XA Pending CN109544525A (en) | 2018-11-15 | 2018-11-15 | A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109544525A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112716446A (en) * | 2020-12-28 | 2021-04-30 | 深圳硅基智能科技有限公司 | Method and system for measuring pathological change characteristics of hypertensive retinopathy |
| CN115409762A (en) * | 2021-05-28 | 2022-11-29 | 南京博视医疗科技有限公司 | Method and device for automatically detecting overlap of planning point and blood vessel |
| CN116071355A (en) * | 2023-03-06 | 2023-05-05 | 山东第一医科大学第二附属医院 | Auxiliary segmentation system and method for peripheral blood vessel image |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104732499A (en) * | 2015-04-01 | 2015-06-24 | 武汉工程大学 | Retina image enhancement algorithm based on multiple scales and multiple directions |
| CN104899876A (en) * | 2015-05-18 | 2015-09-09 | 天津工业大学 | Eyeground image blood vessel segmentation method based on self-adaption difference of Gaussians |
| US20160163041A1 (en) * | 2014-12-05 | 2016-06-09 | Powel Talwar | Alpha-matting based retinal vessel extraction |
| CN106651846A (en) * | 2016-12-20 | 2017-05-10 | 中南大学湘雅医院 | Method for segmenting vasa sanguinea retinae image |
-
2018
- 2018-11-15 CN CN201811357420.XA patent/CN109544525A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20160163041A1 (en) * | 2014-12-05 | 2016-06-09 | Powel Talwar | Alpha-matting based retinal vessel extraction |
| CN104732499A (en) * | 2015-04-01 | 2015-06-24 | 武汉工程大学 | Retina image enhancement algorithm based on multiple scales and multiple directions |
| CN104899876A (en) * | 2015-05-18 | 2015-09-09 | 天津工业大学 | Eyeground image blood vessel segmentation method based on self-adaption difference of Gaussians |
| CN106651846A (en) * | 2016-12-20 | 2017-05-10 | 中南大学湘雅医院 | Method for segmenting vasa sanguinea retinae image |
Non-Patent Citations (2)
| Title |
|---|
| SUBHASIS CHAUDHURI ET AL.: ""Detection of Blood Vessels in Retinal Images Using Two-Dimensional Matched Filters"", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
| 卢鑫: ""视网膜血管识别技术研究与算法实现"", 《中国优秀博硕士学位论文全文数据库(硕士) 医药卫生科技辑》 * |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112716446A (en) * | 2020-12-28 | 2021-04-30 | 深圳硅基智能科技有限公司 | Method and system for measuring pathological change characteristics of hypertensive retinopathy |
| CN112716446B (en) * | 2020-12-28 | 2023-03-24 | 深圳硅基智能科技有限公司 | Method and system for measuring pathological change characteristics of hypertensive retinopathy |
| CN115409762A (en) * | 2021-05-28 | 2022-11-29 | 南京博视医疗科技有限公司 | Method and device for automatically detecting overlap of planning point and blood vessel |
| CN115409762B (en) * | 2021-05-28 | 2023-11-24 | 南京博视医疗科技有限公司 | Method and device for automatically detecting overlapping of planning point and blood vessel |
| CN116071355A (en) * | 2023-03-06 | 2023-05-05 | 山东第一医科大学第二附属医院 | Auxiliary segmentation system and method for peripheral blood vessel image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Xiao et al. | Weighted res-unet for high-quality retina vessel segmentation | |
| Li et al. | Computer‐assisted diagnosis for diabetic retinopathy based on fundus images using deep convolutional neural network | |
| Imran et al. | Comparative analysis of vessel segmentation techniques in retinal images | |
| Hasan et al. | DRNet: Segmentation and localization of optic disc and Fovea from diabetic retinopathy image | |
| Almotiri et al. | Retinal vessels segmentation techniques and algorithms: a survey | |
| Wang et al. | Hierarchical retinal blood vessel segmentation based on feature and ensemble learning | |
| US11213197B2 (en) | Artificial neural network and system for identifying lesion in retinal fundus image | |
| Shen et al. | Domain-invariant interpretable fundus image quality assessment | |
| Biswas et al. | Dilated deep neural network for segmentation of retinal blood vessels in fundus images | |
| Chalakkal et al. | Quality and content analysis of fundus images using deep learning | |
| Hassan et al. | Joint segmentation and quantification of chorioretinal biomarkers in optical coherence tomography scans: A deep learning approach | |
| Dash et al. | An unsupervised approach for extraction of blood vessels from fundus images | |
| CN110751636B (en) | Fundus image retinal arteriosclerosis detection method based on improved coding and decoding network | |
| Zhao et al. | High‐quality retinal vessel segmentation using generative adversarial network with a large receptive field | |
| Cheng et al. | Retinal blood vessel segmentation based on Densely Connected U-Net | |
| Shi et al. | MD-Net: A multi-scale dense network for retinal vessel segmentation | |
| CN109544525A (en) | A kind of eyeground picture blood vessel recognition methods based on self-adapting window Model Matching | |
| CN111080643A (en) | Method and device for classifying diabetes and related diseases based on fundus images | |
| Fu et al. | MSCNN-AM: A multi-scale convolutional neural network with attention mechanisms for retinal vessel segmentation | |
| Mohammedhasan et al. | A New Early Stage Diabetic Retinopathy Diagnosis Model Using Deep Convolutional Neural Networks and Principal Component Analysis. | |
| Zhang et al. | Two-step registration on multi-modal retinal images via deep neural networks | |
| Kassim et al. | Deep U-Net regression and hand-crafted feature fusion for accurate blood vessel segmentation | |
| Zhang et al. | Microaneurysm (MA) detection via sparse representation classifier with MA and Non-MA dictionary learning | |
| Zhao et al. | Attention residual convolution neural network based on U-net (AttentionResU-Net) for retina vessel segmentation | |
| LU500798B1 (en) | Full-Automatic Segmentation Method for Coronary Artery Calcium Lesions Based on Non-Contrast Chest CT |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190329 |