[go: up one dir, main page]

CN111797900B - A method and device for classifying arteries and veins in OCT-A images - Google Patents

A method and device for classifying arteries and veins in OCT-A images Download PDF

Info

Publication number
CN111797900B
CN111797900B CN202010518961.7A CN202010518961A CN111797900B CN 111797900 B CN111797900 B CN 111797900B CN 202010518961 A CN202010518961 A CN 202010518961A CN 111797900 B CN111797900 B CN 111797900B
Authority
CN
China
Prior art keywords
oct
image
images
classification
information
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.)
Active
Application number
CN202010518961.7A
Other languages
Chinese (zh)
Other versions
CN111797900A (en
Inventor
赵一天
谢建洋
苏攀
蒋珊珊
毛浩宇
杨建龙
刘江
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Ningbo Institute of Material Technology and Engineering of CAS
Cixi Institute of Biomedical Engineering CIBE of CAS
Original Assignee
Ningbo Institute of Material Technology and Engineering of CAS
Cixi Institute of Biomedical Engineering CIBE of CAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Ningbo Institute of Material Technology and Engineering of CAS, Cixi Institute of Biomedical Engineering CIBE of CAS filed Critical Ningbo Institute of Material Technology and Engineering of CAS
Priority to CN202010518961.7A priority Critical patent/CN111797900B/en
Publication of CN111797900A publication Critical patent/CN111797900A/en
Application granted granted Critical
Publication of CN111797900B publication Critical patent/CN111797900B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/23—Clustering techniques
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/30—Determination of transform parameters for the alignment of images, i.e. image registration
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
    • G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
    • G06V10/757—Matching configurations of points or features
    • G—PHYSICS
    • G06—COMPUTING OR CALCULATING; 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 OR CALCULATING; 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)
  • Theoretical Computer Science (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • General Engineering & Computer Science (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Medical Informatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • General Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Quality & Reliability (AREA)
  • Multimedia (AREA)
  • Software Systems (AREA)
  • Databases & Information Systems (AREA)
  • Computing Systems (AREA)
  • Radiology & Medical Imaging (AREA)
  • Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
  • Eye Examination Apparatus (AREA)

Abstract

本发明公开了一种OCT‑A图像的动静脉分类方法。包括获取眼底彩照和OCT‑A图像，对眼底彩照和OCT‑A图像的分割结果分别进行拓扑估计，获得拓扑树集合；在对眼底彩照和OCT‑A图像粗匹配后，进行关键点匹配，使两种模态图像之间的拓扑树匹配，获得拓扑信息；根据优势集理论对眼底彩照的近视盘区域血管进行动脉和静脉分类；根据拓扑信息采用标签传播算法对全局进行动静脉分类，获得OCT‑A血管的动静脉分类结果。本发明还公开了一种OCT‑A图像的动静脉分类装置。本发明通过拓扑信息传递的方式实现OCT‑A的血管动静脉分类，解决了OTC‑A图像中因为缺少颜色对比信息而无法进行血管分类的问题。

The present invention discloses an arteriovenous classification method for OCT-A images. The method includes obtaining a color fundus photograph and an OCT-A image, performing topological estimation on the segmentation results of the color fundus photograph and the OCT-A image respectively, and obtaining a topological tree set; after roughly matching the color fundus photograph and the OCT-A image, performing key point matching to match the topological trees between the two modal images and obtain topological information; classifying the blood vessels in the near-optic disc area of the color fundus photograph as arteries and veins according to the dominant set theory; and using a label propagation algorithm to classify the arteries and veins globally according to the topological information to obtain the arteriovenous classification results of the OCT-A blood vessels. The present invention also discloses an arteriovenous classification device for OCT-A images. The present invention realizes the arteriovenous classification of blood vessels of OCT-A by means of topological information transmission, and solves the problem that blood vessels cannot be classified in OCT-A images due to the lack of color contrast information.

Description

Artery and vein classification method and device for OCT-A image
Technical Field
The invention relates to the technical field of image processing, in particular to an artery and vein classification method and device of OCT-A images.
Background
Vascular diseases are one of the most common public health safety in the world, the most common related diseases comprise diabetes, arteriosclerosis, cardiovascular diseases, hypertension and the like, the early clinical manifestations of the diseases are not obvious, the diseases are not easy to be found and are important, and the later sudden manifestations cause serious threat to the life safety of patients. Early screening and diagnosis is therefore of great importance to the public's life health safety. However, due to the complexity and invasiveness of current vascular disease diagnostic approaches, large-scale screening cannot be achieved. Retinal blood vessels, which are the only vascular tissue available in humans under atraumatic conditions, have been shown in clinical studies to have a close relationship in terms of morphological changes with vascular-related systemic diseases (diabetes, hypertension and cardiovascular diseases) and are therefore considered to have great potential in the screening of vascular diseases.
Clinical studies have shown that lesions of diabetes, hypertension and other cardiovascular diseases can cause changes in the caliber of retinal blood vessels, including various degrees of arterial and venous distension, and therefore the arterial to venous caliber ratio (AVR) is commonly used in medicine as a clinical diagnostic basis for related diseases. However, due to the huge number of potential patients, the screening of the fundus is mainly realized through manual film reading in clinic at present, however, the method requires the ophthalmologist to have abundant clinical experience, and is difficult to perform large-scale screening by relying on ophthalmologist manual work. Moreover, because the manual detection experience standards are not uniform, partial missed diagnosis or misdiagnosis may be caused. Therefore, the method has very important significance for automatic detection, diagnosis and interventional therapy of early vascular diseases through computer-aided retinal vascular specificity analysis. Among them, arterial and venous classification of retinal blood vessels is of great research importance as a deep understanding of vascular structure changes.
The research discovers that the abnormal retinal microvasculature is caused by vascular diseases, the current common retinal vascular classification method mostly acts on fundus color photographs, and arterial and venous classification is realized by means of blood vessel color differences in images, however, the retinal color photographs cannot observe capillaries in the macular area, so that the early automatic screening of related diseases is very challenging. With the wide use of OCT-a, it is possible to observe retinal microvasculature, but due to the lack of color information, arteriovenous classification of blood vessels cannot be directly achieved in OCT-a images.
Disclosure of Invention
The invention aims to provide an artery and vein classification method and device for OCT-A images, which can solve the problem that the existing retinal vessel artery and vein classification algorithm can not realize the classification of micro-vessels.
According to a first aspect of the present invention, there is provided an arteriovenous classification method of an OCT-a image, comprising:
acquiring fundus color photographs and OCT-A images, respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images, and obtaining a topology tree set;
after the eye bottom color photograph and the OCT-A image are subjected to rough matching, key point matching is carried out, so that topology trees between the two modal images are matched, and topology information is obtained;
artery and vein classification is carried out on blood vessels in a near vision disc area of fundus color photograph according to an dominance set theory;
and (3) performing global artery and vein classification by adopting a label propagation algorithm according to the topology information to obtain an artery and vein classification result of the OCT-A blood vessel.
Further, "obtaining fundus color illumination and OCT-A image, respectively performing topology estimation on segmentation results of fundus color illumination and OCT-A image, and obtaining a topology tree set" specifically includes:
dividing fundus color illumination and OCT-A images by a morphological refinement algorithm to obtain a vascular structure with unit pixel width;
detecting key nodes of the vascular structure, and disconnecting the segmented image at bifurcation points and intersection points in the key nodes to obtain a plurality of independent vascular segments;
performing central line fitting on the blood vessel segments by adopting cubic spline fitting to obtain central line information of each segment of blood vessel;
establishing an undirected graph of a retinal blood vessel network according to the central line information, extracting characteristic vectors of vessel segments, and establishing a weighted undirected graph of retinal blood vessels;
extracting a video disc area of fundus color illumination according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the video disc area, and determining a starting point of an OCT-A image;
after virtual points are introduced to connect the starting points of all the vessel trees, an independent vessel tree is obtained through a minimum spanning tree algorithm, and a topology tree set is obtained.
Further, "after coarse matching of the fundus color photograph and the OCT-a image, performing key point matching to match a topology tree between two modality images, the obtaining of topology information" specifically includes:
performing rough matching on fundus color illumination and OCT-A images according to a registration algorithm;
intercepting the coincident part of the OCT-A image in the fundus color illumination segmentation result after rough matching and extracting key nodes, and intercepting the coincident part of the fundus color illumination in the OCT-A image segmentation result after rough matching and extracting the key nodes;
matching key nodes under two modes through a Gaussian regression process;
and completing topology tree matching according to consistency of key nodes contained in the topology tree, and obtaining topology tree information.
Further, "classifying arteries and veins of a near vision disc region blood vessel of fundus color photograph according to the dominance set theory" specifically includes:
selecting points in a blood vessel section within a preset multiplying power video disc range as clustering objects, extracting feature vectors from each point, and obtaining a point set;
clustering the selected point sets according to the advantage sets and classifying all the points into two types;
defining arteries and veins according to the average value of the brightness information of the midpoints of each class;
and counting the label distribution of the middle points of each blood vessel segment and performing active vein marking on the blood vessel segments.
Further, "global artery and vein classification is performed by adopting a label propagation algorithm according to topology information, and an artery and vein classification result of the OCT-A blood vessel is obtained" specifically comprises:
the label of the initial vessel section is transmitted downwards through the topology information, so that the retinal vessel is completely classified;
the arteriovenous classification of OCT-a vessels was obtained from a complete classification of retinal vessels.
According to a second aspect of the present invention, there is provided an arteriovenous classification device of OCT-a image, comprising:
the acquisition module is used for: acquiring fundus color photographs and OCT-A images, respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images, and obtaining a topology tree set;
a first processing module: after the eye bottom color photograph and the OCT-A image are subjected to rough matching, key point matching is carried out, so that topology trees between the two modal images are matched, and topology information is obtained;
and a classification module: artery and vein classification is carried out on blood vessels in a near vision disc area of fundus color photograph according to an dominance set theory;
and a second processing module: and (3) performing global artery and vein classification by adopting a label propagation algorithm according to the topology information to obtain an artery and vein classification result of the OCT-A blood vessel.
Further, the obtaining module specifically includes:
a first dividing unit: dividing fundus color illumination and OCT-A images by a morphological refinement algorithm to obtain a vascular structure with unit pixel width;
a first detection unit: detecting key nodes of the vascular structure, and disconnecting the segmented image at bifurcation points and intersection points in the key nodes to obtain a plurality of independent vascular segments;
a first processing unit: performing central line fitting on the blood vessel segments by adopting cubic spline fitting to obtain central line information of each segment of blood vessel;
a first extraction unit: establishing an undirected graph of a retinal blood vessel network according to the central line information, extracting characteristic vectors of vessel segments, and establishing a weighted undirected graph of retinal blood vessels;
a second processing unit: extracting a video disc area of fundus color illumination according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the video disc area, and determining a starting point of an OCT-A image;
a third processing unit: after virtual points are introduced to connect the starting points of all the vessel trees, an independent vessel tree is obtained through a minimum spanning tree algorithm, and a topology tree set is obtained.
Further, the first processing module specifically includes:
a first matching unit: performing rough matching on fundus color illumination and OCT-A images according to a registration algorithm;
a second extraction unit: intercepting the coincident part of the OCT-A image in the fundus color illumination segmentation result after rough matching and extracting key nodes, and intercepting the coincident part of the fundus color illumination in the OCT-A image segmentation result after rough matching and extracting the key nodes;
a second matching unit: matching key nodes under two modes through a Gaussian regression process;
a first acquisition unit: and completing topology tree matching according to consistency of key nodes contained in the topology tree, and obtaining topology tree information.
Further, the classification module specifically includes:
a third extraction unit: selecting points in a blood vessel section within a preset multiplying power video disc range as clustering objects, extracting feature vectors from each point, and obtaining a point set;
a first classification unit: clustering the selected point sets according to the advantage sets and classifying all the points into two types;
a first defining unit: defining arteries and veins according to the average value of the brightness information of the midpoints of each class;
a first statistical unit: and counting the label distribution of the middle points of each blood vessel segment and performing active vein marking on the blood vessel segments.
Further, the second processing module specifically includes:
a fourth processing unit: the label of the initial vessel section is transmitted downwards through the topology information, so that the retinal vessel is completely classified;
a second acquisition unit: the arteriovenous classification of OCT-a vessels was obtained from a complete classification of retinal vessels.
The beneficial effects of the invention are as follows: 1. the artery and vein classification method of the blood vessel in the OCT-A image is provided for the first time, an effective tool is provided for evaluating the structural change of the retina microvasculature, and the early screening of vascular diseases is also possible. 2. The OCT-A image is guided by fundus illumination for the first time, and the blood vessel artery and vein classification of the OCT-A is realized by se:Sub>A topological information transmission mode, so that the problem that the blood vessel classification cannot be carried out due to the lack of color contrast information in the OTC-A image is solved. 3. The topological information transfer method is put forward for the first time, and key points are matched by using a Gaussian regression process, so that the matching between topological subgraphs is carried out on the basis of a topological structure, and the connection between fundus color photograph blood vessels and OCT-A blood vessel trees is realized. 4. The problem that fundus color photograph blood vessels and OCT-A blood vessels cannot be completely overlapped due to registration accuracy errors is solved through matching among topology trees, accuracy of pixel-level label propagation is compensated through information among the topology trees, and meanwhile robustness of OCT-A blood vessel classification is improved. 5. Preliminary experiments are performed in a batch of paired fundus color photographs and OCT-A images, and the results show that the algorithm used in the invention can accurately classify OCT-A blood vessels.
Drawings
FIG. 1 is a flow chart of an arteriovenous classification method of an OCT-A image in accordance with an embodiment of the present invention;
fig. 2 is a block diagram of an arteriovenous classification device for OCT-a images according to an embodiment of the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Fig. 1 shows a flow of an arteriovenous classification method of an OCT-a image according to an embodiment of the present invention, including:
s11, acquiring fundus color illumination and OCT-A images, and respectively carrying out topology estimation on segmentation results of the fundus color illumination and the OCT-A images to obtain a topology tree set.
The execution subject of the method may be a server.
And S12, performing key point matching after performing rough matching on the fundus color photograph and the OCT-A image, so that topology trees between the two modal images are matched, and topology information is obtained.
S13, classifying arteries and veins of the blood vessels of the near vision disc area of the fundus color photograph according to the theory of the dominance set.
S14, performing global artery and vein classification by adopting a label propagation algorithm according to the topology information to obtain an artery and vein classification result of the OCT-A blood vessel.
In the embodiment of the present specification, the technical solution for achieving the above object of the present invention is as follows: 1) Performing topology estimation on segmentation results of the eye fundus color illumination and the OCT-A image by using a graph theory-based method respectively, and establishing a complete topology tree set; 2) Performing coarse registration on the eye fundus color photograph and the OCT-A image by adopting a registration algorithm, and then performing key point matching by using a Gaussian regression process on the basis of the coarse registration to realize matching of topological trees between two modal images, thereby realizing topology information transmission; 3) Artery and vein classification is carried out on blood vessels in the region close to the optic disc of fundus color photograph by adopting an advantage set theory; 4) Global arteriovenous classification is realized by adopting a label propagation algorithm according to topology information, so that arteriovenous classification of OCT-A blood vessels is realized.
As a preferred embodiment, the "obtaining fundus color photograph and OCT-a image, performing topology estimation on the segmentation results of fundus color photograph and OCT-a image, respectively, to obtain a topology tree set" specifically includes: dividing fundus color illumination and OCT-A images by a morphological refinement algorithm to obtain a vascular structure with unit pixel width; detecting key nodes of the vascular structure, and disconnecting the segmented image at bifurcation points and intersection points in the key nodes to obtain a plurality of independent vascular segments; performing central line fitting on the blood vessel segments by adopting cubic spline fitting to obtain central line information of each segment of blood vessel; establishing an undirected graph of a retinal blood vessel network according to the central line information, extracting characteristic vectors of vessel segments, and establishing a weighted undirected graph of retinal blood vessels; extracting a video disc area of fundus color illumination according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the video disc area, and determining a starting point of an OCT-A image; after virtual points are introduced to connect the starting points of all the vessel trees, an independent vessel tree is obtained through a minimum spanning tree algorithm, and a topology tree set is obtained.
In the embodiments of the present description, the present invention converts the visual topology estimation into a graph optimization problem, i.e. a complex vascular network is represented by a graph G containing points and edges, and then the topology estimation is implemented in the graph G using a minimum spanning tree algorithm. The specific operation flow is as follows:
1) Establishing a diagram: adopting a morphological refinement algorithm on the basis of the segmented image, and iteratively removing pixels outside the blood vessel to obtain a blood vessel structure with a width of only one pixel; the key nodes are then detected using a sliding window of 3*3, where the detected points are divided into: the method comprises the steps of terminating nodes (only one adjacent pixel exists in a neighborhood), connecting points (two adjacent pixels exist in the neighborhood), bifurcation points (three adjacent pixels exist in the neighborhood), crossing points (the number of the adjacent pixels in the neighborhood is greater than three), disconnecting a segmented image at the bifurcation points and the crossing points to obtain a plurality of independent blood vessel segments, and then performing center line fitting by adopting three spline fitting to obtain center line information of each blood vessel segment. And taking the starting point and the end point of each segment of blood vessel as a point V in the graph theory, connecting adjacent points according to the neighborhood information to form a side E in the graph theory, and thus establishing an undirected graph G (V, E) of the retinal blood vessel network. Then, feature vectors of the vessel segment are extracted, including direction, gray scale, contrast, and diameter information (only direction and diameter information is extracted as its feature vector for OCT-a because OCT-a lacks color information), and feature vector clustering is performed at neighboring point set Vi (i=1, 2, …, N) using the dominant set to reduce feature vector dimension, and this is taken as similarity measure W between two points. A weighted undirected graph G (V, E, W) of retinal blood vessels is created.
2) Determining a starting point, extracting a video disc region by adopting an automatic segmentation algorithm for fundus illumination, taking a set of intersection points of the edge of the video disc region and blood vessels as a starting point S of a blood vessel tree, deleting points and edges falling in the video disc range, wherein the number of the intersection points is the number of retinal blood vessel trees in principle; for OCT-a images, points that fall within the pixels of the image edge are taken as starting points.
3) Topology estimation, introducing virtual points, connecting with the starting points of each vessel tree, converting the vessel topology estimation problem into a minimum graph problem with the virtual points as the starting points, calculating a minimum graph by adopting a minimum spanning tree algorithm, and finally removing the virtual points to obtainAnd (3) an independent vessel tree, so as to complete the topological estimation of the fundus color photograph vessel and the OCT-A vessel.
As a preferred embodiment, "after coarse matching of an fundus color photograph and an OCT-a image, performing key point matching to match a topology tree between two modality images, obtaining topology information" specifically includes: performing rough matching on fundus color illumination and OCT-A images according to a registration algorithm; intercepting the coincident part of the OCT-A image in the fundus color illumination segmentation result after rough matching and extracting key nodes, and intercepting the coincident part of the fundus color illumination in the OCT-A image segmentation result after rough matching and extracting the key nodes; matching key nodes under two modes through a Gaussian regression process; and completing topology tree matching according to consistency of key nodes contained in the topology tree, and obtaining topology tree information.
In the embodiment of the specification, the registration algorithm is used for realizing the rough registration of fundus color illumination and OCT-A images, but due to the limitation of registration accuracy, the complete matching of retinal blood vessels and OCT-A blood vessels can not be realized, and the pixel-level guidance can not be realized, so the invention provides the topology information transfer algorithm to realize the matching of vessel topology trees under two image modes. The operation method is as follows: firstly, intercepting the coincident part of OCT-A images in fundus color illumination segmentation results according to registration results, extracting the key nodes mentioned above, and extracting the key nodes in the OCT-A segmentation results; then matching key nodes in the two modes by adopting a Gaussian regression process; and finally, matching the topology tree under two modes according to the consistency of the key nodes contained in the topology tree.
Specifically, to find a set of corresponding points in the fundus illumination and OCT-a images, we first define points where the coincident feature points between the two modalities are considered to be a set of matches within six pixel errors. Thus selecting a set of corresponding points according to our definition in the result of the coarse registration. Wherein->Represents the key point in fundus color photograph, +.>The key points corresponding to fundus color illumination in OCT-A are represented, and l represents the number of point pairs. Then the set +.>Using a gaussian regression process as a set of training sets, estimation is performed by fitting the position coordinates between the corresponding pointsCorresponding mean value in Gaussian function>Sum of variances->The calculation method is as follows:
wherein K represents a vectorK is a kernel function defining a mapping consisting of affine transformation and non-linear mapping,/->Representing noise variance->Representing a symmetric matrix of size L×L, wherein the element calculation formula is +.>，Representation->Is a matrix of size L x D where D is the size of the feature vector, and since only the coordinate information of the key point is used in the present invention, d=2.
Then, the coordinates of the projections of all key points in the eye bottom map in the OCT-A image are calculated according to the trained model, and then a new model is calculated according to the definition of the matching pointsMatching point pairs to obtain more complete corresponding setWill be->Fitting by using Gaussian regression process as training set to obtain +.>Iterating in this way, finally when the set of matching points +.>Exit the cycle at steady state, resulting in +.>I.e. the final set of matching points, in principle m is equal to the number of key points in the corresponding retinal image portion of the OCT-a image.
After obtaining the vessel topology information and the set of corresponding points in the two modes, matching between the topology trees is realized by judging the coincidence degree of key points contained between the two groups of vessel trees. Specifically, for the topology tree in any fundus color photograph, the collection of key points contained in the topology tree is countedWherein F represents fundus illumination, i represents corresponding topology tree, and then +.f. by comparing the set of key points contained in the vessel tree contained in OCT-A image>Wherein O represents the OCT-A image. The two vessel trees with the highest overlap ratio are considered to be matched vessel trees. So far, the matching and the fusion of the vessel tree in the two modal images are realized.
As a preferred embodiment, "classifying arteries and veins of a near-vision disc region blood vessel of fundus color photograph according to the dominance set theory" specifically includes: selecting points in a blood vessel section within a preset multiplying power video disc range as clustering objects, extracting feature vectors from each point, and obtaining a point set; clustering the selected point sets according to the advantage sets and classifying all the points into two types; defining arteries and veins according to the average value of the brightness information of the midpoints of each class; and counting the label distribution of the middle points of each blood vessel segment and performing active vein marking on the blood vessel segments.
In the embodiment of the present specification, since the arteries and veins of the fundus color photograph blood vessel have differences in color and caliber and are obvious in the near vision disc region, an accurate clustering result can be obtained relatively easily, and meanwhile, since the arteries of the retinal blood vessel have higher oxygen content than veins, a very strong visual difference is shown in the image, and the visual difference can be used as a key index for distinguishing the arteries and veins. Therefore, the invention provides an artery and vein classification method for realizing the blood vessels in the near-optic disc region by using the advantage set, wherein the blood vessels in the near-optic disc region are firstly gathered into two types by virtue of advantages, and then the brightness information in the two types of blood vessels is extracted as an index for distinguishing the blood vessel types, so that the artery and vein classification of the blood vessels is realized. The specific operation steps are as follows:
firstly, selecting points in a blood vessel section within a range of 1.5 to 2 times of a video disc as clustering objects, and extracting feature vectors containing information such as color, contrast, blood vessel width and the like from each point; then clustering the selected point set by using the advantage set, and dividing all the points into two types; taking the average value of the brightness information of the points in each class as the basis of the classification labels, wherein a group of points with large brightness are defined as arteries, and a group of points with small brightness are defined as veins; and finally, counting the label distribution of the middle points of each section of the blood vessel, defining the classification labels of the blood vessel sections by adopting a voting mechanism, and if the number of the points of the blood vessel sections, which are labeled as arteries, is more than the number of the points of the blood vessel sections, which are labeled as veins, then considering the blood vessel sections as arteries, otherwise, marking the blood vessel sections as veins.
As a preferred embodiment, "global artery and vein classification by using a label propagation algorithm according to topology information, obtaining an arteriovenous classification result of an OCT-a blood vessel" specifically includes: the label of the initial vessel section is transmitted downwards through the topology information, so that the retinal vessel is completely classified; the arteriovenous classification of OCT-a vessels was obtained from a complete classification of retinal vessels.
In the embodiment of the present disclosure, in obtaining the topology of the complete retinal blood vessel and the arteriovenous classification label of the starting blood vessel segment in the fundus color photograph, the present invention uses a label propagation algorithm to implement global arteriovenous classification, specifically, we consider that all the blood vessel segments in the same blood vessel tree are consistent with the label of the starting blood vessel segment of the blood vessel tree, so that the complete classification of the retinal blood vessel can be implemented by transmitting the topology information obtained by the label of the starting blood vessel segment obtained by the above steps downward. OCT-a vessel arteriovenous classification is also included.
Fig. 2 shows a structure of an arteriovenous classification device of an OCT-a image according to an embodiment of the present invention, including:
the acquisition module 21: acquiring fundus color photographs and OCT-A images, respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images, and obtaining a topology tree set;
the first processing module 22: after the eye bottom color photograph and the OCT-A image are subjected to rough matching, key point matching is carried out, so that topology trees between the two modal images are matched, and topology information is obtained;
classification module 23: artery and vein classification is carried out on blood vessels in a near vision disc area of fundus color photograph according to an dominance set theory;
the second processing module 24: and (3) performing global artery and vein classification by adopting a label propagation algorithm according to the topology information to obtain an artery and vein classification result of the OCT-A blood vessel.
As a preferred embodiment, the acquisition module specifically includes: a first dividing unit: dividing fundus color illumination and OCT-A images by a morphological refinement algorithm to obtain a vascular structure with unit pixel width; a first detection unit: detecting key nodes of the vascular structure, and disconnecting the segmented image at bifurcation points and intersection points in the key nodes to obtain a plurality of independent vascular segments; a first processing unit: performing central line fitting on the blood vessel segments by adopting cubic spline fitting to obtain central line information of each segment of blood vessel; a first extraction unit: establishing an undirected graph of a retinal blood vessel network according to the central line information, extracting characteristic vectors of vessel segments, and establishing a weighted undirected graph of retinal blood vessels; a second processing unit: extracting a video disc area of fundus color illumination according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the video disc area, and determining a starting point of an OCT-A image; a third processing unit: after virtual points are introduced to connect the starting points of all the vessel trees, an independent vessel tree is obtained through a minimum spanning tree algorithm, and a topology tree set is obtained.
As a preferred embodiment, the first processing module specifically includes: a first matching unit: performing rough matching on fundus color illumination and OCT-A images according to a registration algorithm; a second extraction unit: intercepting the coincident part of the OCT-A image in the fundus color illumination segmentation result after rough matching and extracting key nodes, and intercepting the coincident part of the fundus color illumination in the OCT-A image segmentation result after rough matching and extracting the key nodes; a second matching unit: matching key nodes under two modes through a Gaussian regression process; a first acquisition unit: and completing topology tree matching according to consistency of key nodes contained in the topology tree, and obtaining topology tree information.
As a preferred embodiment, the classification module specifically includes: a third extraction unit: selecting points in a blood vessel section within a preset multiplying power video disc range as clustering objects, extracting feature vectors from each point, and obtaining a point set; a first classification unit: clustering the selected point sets according to the advantage sets and classifying all the points into two types; a first defining unit: defining arteries and veins according to the average value of the brightness information of the midpoints of each class; a first statistical unit: and counting the label distribution of the middle points of each blood vessel segment and performing active vein marking on the blood vessel segments.
As a preferred embodiment, the second processing module specifically includes: a fourth processing unit: the label of the initial vessel section is transmitted downwards through the topology information, so that the retinal vessel is completely classified; a second acquisition unit: the arteriovenous classification of OCT-a vessels was obtained from a complete classification of retinal vessels.
Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features but not others included in other embodiments, combinations of features of different embodiments are meant to be within the scope of the invention and form different embodiments. For example, in the claims, any of the claimed embodiments may be used in any combination.
Those of ordinary skill in the art will appreciate that: the above embodiments are only for illustrating the technical solution of the present invention, not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: it is still possible to modify the technical solutions described in the foregoing embodiments or to make equivalent substitutions for some or all of the technical features thereof, without departing from the essence of the corresponding technical solutions from the scope of the present invention defined by the claims.

Claims (2)

1.一种OCT-A图像的动静脉分类方法，其特征为，包括：1. A method for classifying arteries and veins in OCT-A images, characterized by comprising: 获取眼底彩照和OCT-A图像，对所述眼底彩照和OCT-A图像的分割结果分别进行拓扑估计，获得拓扑树集合；Obtain fundus color images and OCT-A images, and perform topological estimation on the segmentation results of the fundus color images and OCT-A images respectively to obtain a set of topological trees; 在对所述眼底彩照和OCT-A图像粗匹配后，进行关键点匹配，使两种模态图像之间的拓扑树匹配，获得拓扑信息；After coarse matching of the fundus color image and the OCT-A image, key point matching is performed to match the topology tree between the two modal images and obtain topological information. 根据优势集理论对所述眼底彩照的近视盘区域血管进行动脉和静脉分类；Based on the dominance set theory, the blood vessels in the myopic disc region of the fundus photographs were classified into arteries and veins. 根据所述拓扑信息采用标签传播算法对全局进行动静脉分类，获得OCT-A血管的动静脉分类结果；Based on the topology information, a label propagation algorithm is used to classify arteries and veins globally, and the artery and vein classification results of OCT-A vessels are obtained. 其中，“获取眼底彩照和OCT-A图像，对所述眼底彩照和OCT-A图像的分割结果分别进行拓扑估计，获得拓扑树集合”具体包括：Specifically, "acquiring fundus color images and OCT-A images, performing topological estimation on the segmentation results of the fundus color images and OCT-A images respectively, and obtaining a set of topological trees" includes: 通过形态学细化算法对眼底彩照和OCT-A图像进行分割，获得单位像素宽的血管结构；The fundus color photograph and OCT-A image were segmented using a morphological thinning algorithm to obtain the vascular structure with a unit pixel width; 检测所述血管结构的关键节点，并在所述关键节点中的分叉点和交叉点处将分割图像断开，获得若干独立的血管段；The key nodes of the vascular structure are detected, and the segmented image is broken at the bifurcation and intersection points of the key nodes to obtain several independent vascular segments; 采用三次样条拟合对所述血管段进行中心线拟合，获得每一段血管的中心线信息；The centerline of the blood vessel segment was fitted using cubic spline fitting to obtain the centerline information of each segment. 根据所述中心线信息建立视网膜血管网络的无向图并提取所述血管段的特征向量，建立视网膜血管的带权无向图；Based on the centerline information, an undirected graph of the retinal vascular network is constructed and the feature vectors of the vascular segments are extracted to construct a weighted undirected graph of the retinal vessels. 根据自动分割算法提取所述眼底彩照的视盘区域，根据所述视盘区域确定血管树的起点，并确定OCT-A图像的起始点；The optic disc region of the fundus image is extracted using an automatic segmentation algorithm. The starting point of the vascular tree is determined based on the optic disc region, and the starting point of the OCT-A image is also determined. 引入虚拟的点将各血管树的起点相连后，通过最小生成树算法得到独立的血管树，获得拓扑树集合；By introducing virtual points to connect the starting points of each vascular tree, the minimum spanning tree algorithm is used to obtain independent vascular trees, thus obtaining a set of topological trees. “在对所述眼底彩照和OCT-A图像粗匹配后，进行关键点匹配，使两种模态图像之间的拓扑树匹配，获得拓扑信息”具体包括："After coarse matching of the fundus color image and the OCT-A image, key point matching is performed to match the topology trees between the two modalities and obtain topological information." This specifically includes: 根据配准算法对所述眼底彩照和OCT-A图像进行粗匹配；The fundus color photograph and the OCT-A image are coarsely matched according to the registration algorithm; 截取粗匹配后所述眼底彩照分割结果中OCT-A图像的重合部分并提取关键节点，截取粗匹配后所述OCT-A图像分割结果中眼底彩照的重合部分并提取关键节点；Extract the overlapping portion of the OCT-A image in the fundus color image segmentation result after coarse matching and extract key nodes. 通过高斯回归过程对两种模态下的关键节点进行匹配；Key nodes in the two modes were matched using a Gaussian regression process; 根据拓扑树中包含关键节点的一致性完成拓扑树匹配，获得拓扑树信息；The topology tree is matched based on the consistency of key nodes in the topology tree to obtain topology tree information; “根据优势集理论对所述眼底彩照的近视盘区域血管进行动脉和静脉分类”具体包括："Classifying the blood vessels in the myopic disc region of the aforementioned fundus color photograph into arteries and veins based on the dominance set theory" specifically includes: 选取预设倍率视盘范围内的血管段中的点作为聚类对象，对每一个点提取特征向量，获得点集；Points in the blood vessel segments within the preset magnification optic disc range are selected as clustering objects, and feature vectors are extracted from each point to obtain a point set; 根据优势集对选择的所述点集进行聚类并将所有的点分成两类；Cluster the selected point set based on the advantage set and divide all points into two classes; 根据每类中点的亮度信息的均值定义动脉与静脉；Arteries and veins are defined based on the mean brightness information of the midpoints in each class; 统计每一段血管段中点的标签分布并对所述血管段进行动静脉标记；The distribution of labels at the midpoints of each blood vessel segment is statistically analyzed, and the blood vessel segments are marked with arteries and veins. “根据所述拓扑信息采用标签传播算法对全局进行动静脉分类，获得OCT-A血管的动静脉分类结果”具体包括："Based on the aforementioned topological information, a label propagation algorithm is used to perform global arterial and venous classification to obtain the OCT-A vessel arterial and venous classification results," specifically including: 通过所述拓扑信息对起始血管段的标签向下传递，实现视网膜血管的完全分类；By passing the labels of the starting vessel segments downwards using the aforementioned topological information, complete classification of retinal vessels can be achieved. 从所述视网膜血管的完全分类中获取OCT-A血管的动静脉分类。The arterial and venous classification of OCT-A vessels is obtained from the complete classification of retinal vessels. 2.一种OCT-A图像的动静脉分类装置，其特征为，包括：2. A device for classifying arteries and veins in OCT-A images, characterized in that it comprises: 获取模块：获取眼底彩照和OCT-A图像，对所述眼底彩照和OCT-A图像的分割结果分别进行拓扑估计，获得拓扑树集合；Acquisition module: Acquires fundus color images and OCT-A images, performs topological estimation on the segmentation results of the fundus color images and OCT-A images respectively, and obtains a set of topological trees; 第一处理模块：在对所述眼底彩照和OCT-A图像粗匹配后，进行关键点匹配，使两种模态图像之间的拓扑树匹配，获得拓扑信息；First processing module: After coarse matching of the fundus color image and the OCT-A image, key point matching is performed to match the topology tree between the two modal images and obtain topological information; 分类模块：根据优势集理论对所述眼底彩照的近视盘区域血管进行动脉和静脉分类；Classification module: Based on the dominance set theory, the blood vessels in the myopic disc region of the fundus photograph are classified into arteries and veins; 第二处理模块：根据所述拓扑信息采用标签传播算法对全局进行动静脉分类，获得OCT-A血管的动静脉分类结果；The second processing module: Based on the topology information, the label propagation algorithm is used to perform global arterial and venous classification to obtain the arterial and venous classification results of OCT-A vessels; 所述获取模块具体包括：The acquisition module specifically includes: 第一分割单元：通过形态学细化算法对眼底彩照和OCT-A图像进行分割，获得单位像素宽的血管结构；First segmentation unit: The fundus color photograph and OCT-A image are segmented using a morphological thinning algorithm to obtain the vascular structure with a unit pixel width; 第一检测单元：检测所述血管结构的关键节点，并在所述关键节点中的分叉点和交叉点处将分割图像断开，获得若干独立的血管段；First detection unit: detects key nodes of the vascular structure, and breaks the segmented image at the bifurcation and intersection points of the key nodes to obtain several independent vascular segments; 第一处理单元：采用三次样条拟合对所述血管段进行中心线拟合，获得每一段血管的中心线信息；First processing unit: Uses cubic spline fitting to fit the centerline of the blood vessel segment to obtain the centerline information of each blood vessel segment; 第一提取单元：根据所述中心线信息建立视网膜血管网络的无向图并提取所述血管段的特征向量，建立视网膜血管的带权无向图；First extraction unit: Based on the centerline information, establish an undirected graph of the retinal vascular network and extract the feature vectors of the vascular segments to establish a weighted undirected graph of the retinal vessels; 第二处理单元：根据自动分割算法提取所述眼底彩照的视盘区域，根据所述视盘区域确定血管树的起点，并确定OCT-A图像的起始点；The second processing unit extracts the optic disc region from the fundus color image using an automatic segmentation algorithm, determines the starting point of the vascular tree based on the optic disc region, and determines the starting point of the OCT-A image. 第三处理单元：引入虚拟的点将各血管树的起点相连后，通过最小生成树算法得到独立的血管树，获得拓扑树集合；The third processing unit: After introducing virtual points to connect the starting points of each vascular tree, the independent vascular trees are obtained through the minimum spanning tree algorithm, and a set of topological trees is obtained. 所述第一处理模块具体包括：The first processing module specifically includes: 第一匹配单元：根据配准算法对所述眼底彩照和OCT-A图像进行粗匹配；First matching unit: Performs coarse matching between the fundus color photograph and the OCT-A image according to the registration algorithm; 第二提取单元：截取粗匹配后所述眼底彩照分割结果中OCT-A图像的重合部分并提取关键节点，截取粗匹配后所述OCT-A图像分割结果中眼底彩照的重合部分并提取关键节点；Second extraction unit: Extract the overlapping part of the OCT-A image in the fundus color image segmentation result after coarse matching and extract key nodes; 第二匹配单元：通过高斯回归过程对两种模态下的关键节点进行匹配；The second matching unit: matches key nodes in two modes through a Gaussian regression process; 第一获取单元：根据拓扑树中包含关键节点的一致性完成拓扑树匹配，获得拓扑树信息；First acquisition unit: Complete topology tree matching based on the consistency of key nodes contained in the topology tree to obtain topology tree information; 所述分类模块具体包括：The classification module specifically includes: 第三提取单元：选取预设倍率视盘范围内的血管段中的点作为聚类对象，对每一个点提取特征向量，获得点集；The third extraction unit: Selects points in the blood vessel segments within the preset magnification visual disc range as clustering objects, extracts feature vectors for each point, and obtains a point set; 第一分类单元：根据优势集对选择的所述点集进行聚类并将所有的点分成两类；First classification unit: Cluster the selected point set according to the advantage set and divide all points into two classes; 第一定义单元：根据每类中点的亮度信息的均值定义动脉与静脉；First definition unit: Define arteries and veins based on the mean value of the brightness information of the midpoints in each class; 第一统计单元：统计每一段血管段中点的标签分布并对所述血管段进行动静脉标记；First statistical unit: Statistically analyze the label distribution of the midpoints of each blood vessel segment and mark the arteries and veins of the blood vessel segment; 所述第二处理模块具体包括：The second processing module specifically includes: 第四处理单元：通过所述拓扑信息对起始血管段的标签向下传递，实现视网膜血管的完全分类；Fourth processing unit: The labels of the starting vessel segments are passed down through the topological information to achieve complete classification of retinal vessels; 第二获取单元：从所述视网膜血管的完全分类中获取OCT-A血管的动静脉分类。Second acquisition unit: Acquires the arteriovenous classification of OCT-A vessels from the complete classification of retinal vessels.
CN202010518961.7A 2020-06-09 2020-06-09 A method and device for classifying arteries and veins in OCT-A images Active CN111797900B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010518961.7A CN111797900B (en) 2020-06-09 2020-06-09 A method and device for classifying arteries and veins in OCT-A images

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010518961.7A CN111797900B (en) 2020-06-09 2020-06-09 A method and device for classifying arteries and veins in OCT-A images

Publications (2)

Publication Number Publication Date
CN111797900A CN111797900A (en) 2020-10-20
CN111797900B true CN111797900B (en) 2024-04-09

Family

ID=72804134

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010518961.7A Active CN111797900B (en) 2020-06-09 2020-06-09 A method and device for classifying arteries and veins in OCT-A images

Country Status (1)

Country Link
CN (1) CN111797900B (en)

Families Citing this family (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113450256B (en) * 2021-06-30 2023-03-03 重庆理工大学 Automatic aorta segmentation method and system based on CT image of level set
CN113516643A (en) * 2021-07-13 2021-10-19 重庆大学 Detection of retinal blood vessel bifurcations and intersections in OCTA images
CN114782339A (en) * 2022-04-09 2022-07-22 中山大学中山眼科中心 A Conditional Generative Adversarial Network-based Method for Capillary Labeling in Color Fundus Photographs
CN119479050B (en) * 2024-10-31 2025-07-15 中国医学科学院阜外医院 Retina image processing method and device and computer readable storage medium
CN120563941B (en) * 2025-07-30 2025-12-05 瀚依科技(杭州)有限公司 Blood vessel classification method, apparatus, electronic device, medium and computer program product

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101340929A (en) * 2005-10-28 2009-01-07 勒帕斯公司 Compositions and methods for treating and preventing fibrotic, inflammatory and neovascular disorders
CN101687031A (en) * 2006-10-27 2010-03-31 勒帕斯公司 Compositions and methods for treating ocular diseases and conditions
CN102048550A (en) * 2009-11-02 2011-05-11 上海交通大学医学院附属仁济医院 Method for automatically generating liver 3D (three-dimensional) image and accurately positioning liver vascular domination region
CN107845098A (en) * 2017-11-14 2018-03-27 南京理工大学 Liver cancer image full-automatic partition method based on random forest and fuzzy clustering
CN110349175A (en) * 2019-06-25 2019-10-18 深圳先进技术研究院 A kind of arteriovenous malformation dividing method, system and electronic equipment
WO2019237148A1 (en) * 2018-06-13 2019-12-19 Commonwealth Scientific And Industrial Research Organisation Retinal image analysis

Family Cites Families (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20050191627A1 (en) * 2001-09-28 2005-09-01 Incyte Corporation Enzymes

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101340929A (en) * 2005-10-28 2009-01-07 勒帕斯公司 Compositions and methods for treating and preventing fibrotic, inflammatory and neovascular disorders
CN101687031A (en) * 2006-10-27 2010-03-31 勒帕斯公司 Compositions and methods for treating ocular diseases and conditions
CN102048550A (en) * 2009-11-02 2011-05-11 上海交通大学医学院附属仁济医院 Method for automatically generating liver 3D (three-dimensional) image and accurately positioning liver vascular domination region
CN107845098A (en) * 2017-11-14 2018-03-27 南京理工大学 Liver cancer image full-automatic partition method based on random forest and fuzzy clustering
WO2019237148A1 (en) * 2018-06-13 2019-12-19 Commonwealth Scientific And Industrial Research Organisation Retinal image analysis
CN110349175A (en) * 2019-06-25 2019-10-18 深圳先进技术研究院 A kind of arteriovenous malformation dividing method, system and electronic equipment

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
Retinal capillary perfusion: spatial and temporal heterogeneity;Dao-Yi Yu等;《Progress in Retinal and Eye Research》;第70卷;23-54 *
Retinal vascular network topology reconstruction and artery/vein classification via dominant set clustering;Yitian Zhao等;《IEEE Transactions on Medical Imaging》;第39卷(第2期);341-356 *
糖尿病视网膜微血管病变的基础与临床研究;黄晓莉;《中国博士学位论文全文数据库 医药卫生科技辑》(第09期);E065-10 *
血管内OCT图像序列中血管三维重建方法的研究;陈艳花;《中国优秀硕士学位论文全文数据库 医药卫生科技辑》(第03期);E062-27 *

Also Published As

Publication number Publication date
CN111797900A (en) 2020-10-20

Similar Documents

Publication Publication Date Title
CN111797900B (en) A method and device for classifying arteries and veins in OCT-A images
Fu et al. Optic disc segmentation by U-net and probability bubble in abnormal fundus images
Trucco et al. Validating retinal fundus image analysis algorithms: issues and a proposal
CN109472781B (en) Diabetic retinopathy detection system based on serial structure segmentation
Giancardo et al. Exudate-based diabetic macular edema detection in fundus images using publicly available datasets
CN111797901A (en) A Retinal Artery and Vein Classification Method and Device Based on Topological Structure Estimation
US20220383661A1 (en) Method and device for retinal image recognition, electronic equipment, and storage medium
Kang et al. AVNet: A retinal artery/vein classification network with category-attention weighted fusion
CN106204555B (en) A Optic Disc Localization Method Combining Gbvs Model and Phase Consistency
Jordan et al. A review of feature-based retinal image analysis
CN104573712A (en) Arteriovenous retinal blood vessel classification method based on eye fundus image
Khitran et al. Automated system for the detection of hypertensive retinopathy
CN111161287A (en) Retinal vessel segmentation method based on symmetric bidirectional cascade network deep learning
SG172159A1 (en) A method and system for determining the position of an optic cup boundary
Hu et al. Multi-scale interactive network with artery/vein discriminator for retinal vessel classification
Mao et al. Deep learning with skip connection attention for choroid layer segmentation in oct images
Oloumi et al. Computer-aided diagnosis of plus disease in retinal fundus images of preterm infants via measurement of vessel tortuosity
CN108921836A (en) A kind of method and device for extracting eye fundus image mark
Salih et al. Fast optic disc segmentation using FFT-based template-matching and region-growing techniques
CN115511883B (en) Method, apparatus and storage medium for determining curvature of retinal fundus blood vessel
CN114359284B (en) Method for analyzing retinal fundus images and related products
CN116630237A (en) Image quality detection method and related device, electronic equipment and storage medium
Zou et al. Supervised vessels classification based on feature selection
Zhou et al. Computer aided diagnosis for diabetic retinopathy based on fundus image
Rodtook et al. Optic disc localization using graph traversal algorithm along blood vessel in polar retinal image

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