CN111797900A - Arteriovenous classification method and device of OCT-A image - Google Patents
Arteriovenous classification method and device of OCT-A image Download PDFInfo
- Publication number
- CN111797900A CN111797900A CN202010518961.7A CN202010518961A CN111797900A CN 111797900 A CN111797900 A CN 111797900A CN 202010518961 A CN202010518961 A CN 202010518961A CN 111797900 A CN111797900 A CN 111797900A
- Authority
- CN
- China
- Prior art keywords
- oct
- blood vessel
- image
- classification
- arteriovenous
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 34
- 210000004204 blood vessel Anatomy 0.000 claims abstract description 137
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 44
- 210000003462 vein Anatomy 0.000 claims abstract description 34
- 230000011218 segmentation Effects 0.000 claims abstract description 32
- 210000001367 artery Anatomy 0.000 claims abstract description 24
- 230000008901 benefit Effects 0.000 claims abstract description 17
- 208000001491 myopia Diseases 0.000 claims abstract description 12
- 210000001210 retinal vessel Anatomy 0.000 claims description 37
- 238000012545 processing Methods 0.000 claims description 25
- 230000008569 process Effects 0.000 claims description 11
- 238000000605 extraction Methods 0.000 claims description 9
- 230000000877 morphologic effect Effects 0.000 claims description 7
- 238000003709 image segmentation Methods 0.000 claims description 6
- 238000001514 detection method Methods 0.000 claims description 5
- 210000004220 fundus oculi Anatomy 0.000 claims description 3
- 238000005286 illumination Methods 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 abstract description 3
- 201000010099 disease Diseases 0.000 description 7
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 7
- 238000012216 screening Methods 0.000 description 5
- 208000019553 vascular disease Diseases 0.000 description 5
- 238000003745 diagnosis Methods 0.000 description 4
- 208000024172 Cardiovascular disease Diseases 0.000 description 3
- 206010020772 Hypertension Diseases 0.000 description 3
- 206010012601 diabetes mellitus Diseases 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000004232 retinal microvasculature Effects 0.000 description 3
- 230000002792 vascular Effects 0.000 description 3
- 238000011835 investigation Methods 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- 238000012546 transfer Methods 0.000 description 2
- 230000000007 visual effect Effects 0.000 description 2
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 1
- 206010003210 Arteriosclerosis Diseases 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 208000011775 arteriosclerosis disease Diseases 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000007635 classification algorithm Methods 0.000 description 1
- 238000003759 clinical diagnosis Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000010339 dilation Effects 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 239000012467 final product Substances 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 210000004088 microvessel Anatomy 0.000 description 1
- 230000004660 morphological change Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 231100000915 pathological change Toxicity 0.000 description 1
- 230000036285 pathological change Effects 0.000 description 1
- 230000005180 public health Effects 0.000 description 1
- 210000001525 retina Anatomy 0.000 description 1
- 230000002207 retinal effect Effects 0.000 description 1
- 238000011524 similarity measure Methods 0.000 description 1
- 230000006641 stabilisation Effects 0.000 description 1
- 238000011105 stabilization Methods 0.000 description 1
- 230000009885 systemic effect Effects 0.000 description 1
- 238000002560 therapeutic procedure Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Images
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
- 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; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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; CALCULATING OR 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; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30041—Eye; Retina; Ophthalmic
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30101—Blood vessel; Artery; Vein; Vascular
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Eye Examination Apparatus (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Computing Systems (AREA)
Abstract
The invention discloses an arteriovenous classification method of an OCT-A image. Acquiring fundus color photographs and OCT-A images, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set; after the eye fundus color photograph and the OCT-A image are roughly matched, key point matching is carried out, so that the topological trees between the two modal images are matched, and topological information is obtained; classifying the arteries and veins of the near vision disc region of the fundus color photograph according to an advantage set theory; and (4) carrying out vein classification on the whole situation by adopting a label propagation algorithm according to the topological information to obtain the arteriovenous classification result of the OCT-A blood vessel. The invention also discloses an arteriovenous classification device of the OCT-A image. The invention realizes the classification of the blood vessel artery and vein of OCT-A by a topological information transmission mode, and solves the problem that the blood vessel classification can not be carried out due to the lack of color contrast information in the OTC-A image.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a method and a device for classifying arteriovenous of an OCT-A image.
Background
Vascular diseases are one of the most common public health and 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 discovered and pay attention, and the late sudden manifestations pose serious threats to the life safety of patients. Therefore, early screening and diagnosis are of great significance to the public's life health and safety. However, due to the complexity and invasiveness of the current diagnosis means for vascular diseases, large-scale general investigation cannot be realized. Retinal blood vessels, which are the only blood vessel tissues available in the human body under the condition of non-wound, are proved to have close relationship with the morphological changes of blood vessel related systemic diseases (diabetes, hypertension and cardiovascular diseases) in clinical research, so that the retinal blood vessels are considered to have great potential in the screening of blood vessel diseases.
Clinical studies have shown that changes in retinal vascular caliber, including varying degrees of arterial and venous dilation, are caused by the pathological changes of diabetes, hypertension, and other cardiovascular diseases, and thus the Ratio of arterial-to-venous diameter Ratio (AVR) is commonly used in medicine as a clinical diagnosis basis for related diseases. However, due to the huge number of potential patients, the fundus screening is mainly achieved through manual film reading in clinic at present, however, the ophthalmologist is required to have abundant clinical experience, and large-scale general investigation is difficult to perform manually only by means of an ophthalmologist. Moreover, the manual detection experience standard is not uniform, so that partial missed diagnosis or misdiagnosis can be caused. Therefore, the method has very important significance for automatic detection, diagnosis and interventional therapy of early vascular diseases through computer-aided retinal vessel specific analysis. The arterial and venous classification of retinal vessels is of great research interest as a deep understanding of vascular structural changes.
Research finds that vascular diseases firstly cause retinal microvasculature abnormity, most of the current common retinal blood vessel classification methods are applied to fundus color photography, artery and vein classification is realized by means of blood vessel color difference in images, and however, the fundus color photography cannot observe capillary vessels in the macular region, which presents a great challenge to early automatic screening of related diseases. With the wide use of OCT-A, the possibility is provided for observing retinal microvasculature, but due to the lack of color information, arteriovenous classification of blood vessels cannot be directly realized in an OCT-A image.
Disclosure of Invention
The invention aims to provide an arteriovenous classification method and device of an OCT-A image, which can solve the problem that the mutexisting retinal vessel arteriovenous classification algorithm cannot realize 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, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set;
after the eye fundus color photograph and the OCT-A image are roughly matched, key point matching is carried out, so that the topological trees between the two modal images are matched, and topological information is obtained;
classifying the arteries and veins of the near vision disc region of the fundus color photograph according to an advantage set theory;
and (4) carrying out vein classification on the whole situation by adopting a label propagation algorithm according to the topological information to obtain the arteriovenous classification result of the OCT-A blood vessel.
Further, "acquiring a fundus color photograph and an OCT-a image, and performing topology estimation on segmentation results of the fundus color photograph and the OCT-a image, respectively, to obtain a topology tree set" specifically includes:
segmenting the fundus color photograph and the OCT-A image by a morphological thinning algorithm to obtain a blood vessel structure with unit pixel width;
detecting key nodes of a blood vessel structure, and disconnecting the segmentation images at bifurcation points and intersection points in the key nodes to obtain a plurality of independent blood vessel sections;
performing center line fitting on the blood vessel sections by adopting cubic spline fitting to obtain center line information of each section of blood vessel;
establishing an undirected graph of a retinal vessel network according to the center line information, extracting a characteristic vector of a vessel section, and establishing a weighted undirected graph of a retinal vessel;
mutextracting a optic disc region of the fundus color photograph according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the optic disc region, and determining a starting point of an OCT-A image;
and introducing virtual points to connect the starting points of the blood vessel trees, and obtaining independent blood vessel trees by a minimum spanning tree algorithm to obtain a topological tree set.
Further, "after rough matching of the fundus color photograph and the OCT-a image, matching of key points is performed to match a topology tree between two modality images, and obtaining topology information" specifically includes:
roughly matching the fundus color photograph with the OCT-A image according to a registration algorithm;
intercepting the overlapped part of the OCT-A images in the eye ground color photograph segmentation result after the rough matching and mut mutextracting key nodes, and intercepting the overlapped part of the eye ground color photograph in the OCT-A image segmentation result after the rough matching and mut mutextracting key nodes;
matching key nodes in the two modes through a Gaussian regression process;
and completing topology tree matching according to the consistency of key nodes contained in the topology tree to obtain topology tree information.
Further, "moving arteries and veins of the near vision disc region blood vessels of the fundus oculi color photograph according to the dominance set theory" specifically includes:
selecting points in a blood vessel section within a preset multiplying power optic disc range as clustering objects, and extracting a characteristic vector for each point to obtain a point set;
clustering the selected point sets according to the advantage sets and dividing all points into two types;
defining an artery and a vein according to the mean value of the brightness information of the midpoints in each class;
and counting the label distribution of the midpoint of each blood vessel section and carrying out arteriovenous marking on the blood vessel section.
Further, "classifying the global moving veins by using a label propagation algorithm according to the topological information to obtain an arteriovenous classification result of the OCT-a blood vessel" specifically includes:
transmitting the label of the initial blood vessel section downwards through topological information to realize the complete classification of retinal blood vessels;
the arteriovenous classification of the OCT-a vessels is obtained from the complete classification of retinal vessels.
According to a second aspect of the present invention, there is provided an arteriovenous classification device of an OCT-a image, comprising:
an acquisition module: acquiring fundus color photographs and OCT-A images, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set;
a first processing module: after the eye fundus color photograph and the OCT-A image are roughly matched, key point matching is carried out, so that the topological trees between the two modal images are matched, and topological information is obtained;
a classification module: classifying the arteries and veins of the near vision disc region of the fundus color photograph according to an advantage set theory;
a second processing module: and (4) carrying out vein classification on the whole situation by adopting a label propagation algorithm according to the topological information to obtain the arteriovenous classification result of the OCT-A blood vessel.
Further, the obtaining module specifically includes:
a first dividing unit: segmenting the fundus color photograph and the OCT-A image by a morphological thinning algorithm to obtain a blood vessel structure with unit pixel width;
a first detection unit: detecting key nodes of a blood vessel structure, and disconnecting the segmentation images at bifurcation points and intersection points in the key nodes to obtain a plurality of independent blood vessel sections;
a first processing unit: performing center line fitting on the blood vessel sections by adopting cubic spline fitting to obtain center line information of each section of blood vessel;
a first extraction unit: establishing an undirected graph of a retinal vessel network according to the center line information, extracting a characteristic vector of a vessel section, and establishing a weighted undirected graph of a retinal vessel;
a second processing unit: mutextracting a optic disc region of the fundus color photograph according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the optic disc region, and determining a starting point of an OCT-A image;
a third processing unit: and introducing virtual points to connect the starting points of the blood vessel trees, and obtaining independent blood vessel trees by a minimum spanning tree algorithm to obtain a topological tree set.
Further, the first processing module specifically includes:
a first matching unit: roughly matching the fundus color photograph with the OCT-A image according to a registration algorithm;
a second extraction unit: intercepting the overlapped part of the OCT-A images in the eye ground color photograph segmentation result after the rough matching and mut mutextracting key nodes, and intercepting the overlapped part of the eye ground color photograph in the OCT-A image segmentation result after the rough matching and mut mutextracting key nodes;
a second matching unit: matching key nodes in the two modes through a Gaussian regression process;
a first acquisition unit: and completing topology tree matching according to the consistency of key nodes contained in the topology tree to obtain 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 optic disc range as clustering objects, and extracting a characteristic vector for each point to obtain a point set;
a first classification unit: clustering the selected point sets according to the advantage sets and dividing all points into two types;
a first defining unit: defining an artery and a vein according to the mean value of the brightness information of the midpoints in each class;
a first statistical unit: and counting the label distribution of the midpoint of each blood vessel section and carrying out arteriovenous marking on the blood vessel section.
Further, the second processing module specifically includes:
a fourth processing unit: transmitting the label of the initial blood vessel section downwards through topological information to realize the complete classification of retinal blood vessels;
a second acquisition unit: the arteriovenous classification of the OCT-a vessels is obtained from the complete classification of retinal vessels.
The invention has the beneficial effects that: 1. the method for classifying the arteriovenous vessels in the OCT-A image is provided for the first time, an effective tool is provided for evaluating the structural change of the retinal microvasculature, and the possibility is provided for early screening of vascular diseases. 2. The OCT-A image is guided by fundus color photography for the first time, the classification of the blood vessels and the artery and the vein of the OCT-A is realized in a topological information transmission mode, and the problem that the blood vessels cannot be classified due to the lack of color contrast information in the OTC-A image is solved. 3. The topological information transfer method is proposed for the first time, and the matching of key points is carried out 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 the fundus color-illuminated blood vessel and the OCT-A blood vessel tree is realized. 4. By matching the topological trees, the problem that the fundus color-illuminated blood vessel and the OCT-A blood vessel cannot be completely overlapped due to registration accuracy errors is solved, the accuracy of pixel level label propagation is compensated by using the information between the topological trees, and meanwhile the robustness of OCT-A blood vessel classification is improved. 5. The results of pre-mut mutexperiments in a batch of paired fundus color photographs and OCT-A images show that the algorithm used in the invention can accurately classify the OCT-A blood vessels.
Drawings
FIG. 1 is a flow chart of a method of arteriovenous classification of an OCT-A image according to an embodiment of the invention;
fig. 2 is a structural diagram of an arteriovenous classification device of an OCT-a image according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be obtained by a person skilled in the art without any inventive step based on the embodiments of the present invention, are within the scope of the present 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:
and S11, acquiring fundus color photographs and OCT-A images, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set.
The execution subject of the method may be a server.
And S12, after the eye fundus color photograph and the OCT-A image are roughly matched, key point matching is carried out, so that the topological trees between the two mode images are matched, and topological information is obtained.
And S13, classifying the arteries and veins of the near vision disc region blood vessels of the fundus color photograph according to the dominance set theory.
S14, globally classifying the arteriovenous by adopting a label propagation algorithm according to the topological information to obtain an arteriovenous 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) respectively carrying out topology estimation on segmentation results of the fundus color photograph and the OCT-A image by using a method based on graph theory, and establishing a complete topology tree set; 2) coarse registration is carried out on the fundus color photograph and the OCT-A image by adopting a registration algorithm, then key point matching is carried out by using a Gaussian regression process on the basis of the coarse registration, and the matching of a topological tree between two modal images is realized, so that the transmission of topological information is realized; 3) classifying veins and veins of the blood vessels in the fundus color-illumination near optic disc region by adopting an advantage set theory; 4) and global arteriovenous classification is realized by adopting a label propagation algorithm according to the topological information, so that the arteriovenous classification of the OCT-A blood vessel is realized.
As a preferred embodiment, "acquiring a fundus color photograph and an OCT-a image, and performing topology estimation on segmentation results of the fundus color photograph and the OCT-a image, respectively, to obtain a topology tree set" specifically includes: segmenting the fundus color photograph and the OCT-A image by a morphological thinning algorithm to obtain a blood vessel structure with unit pixel width; detecting key nodes of a blood vessel structure, and disconnecting the segmentation images at bifurcation points and intersection points in the key nodes to obtain a plurality of independent blood vessel sections; performing center line fitting on the blood vessel sections by adopting cubic spline fitting to obtain center line information of each section of blood vessel; establishing an undirected graph of a retinal vessel network according to the center line information, extracting a characteristic vector of a vessel section, and establishing a weighted undirected graph of a retinal vessel; mutextracting a optic disc region of the fundus color photograph according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the optic disc region, and determining a starting point of an OCT-A image; and introducing virtual points to connect the starting points of the blood vessel trees, and obtaining independent blood vessel trees by a minimum spanning tree algorithm to obtain a topological tree set.
In the embodiment of the specification, the invention converts the visual topology estimation into a graph optimization problem, namely, a complex blood vessel network is represented by a graph G containing points and edges, and then the topology estimation is realized by using a minimum spanning tree algorithm in the graph G. The specific operation flow is as follows:
1) and (3) establishing a graph: iteratively removing external pixels of the blood vessels by adopting a morphological thinning algorithm on the basis of the segmented image to obtain a blood vessel structure with the width of only one pixel; and then detecting key nodes by adopting a 3-by-3 sliding window, wherein the detected points are divided into: the method comprises the steps of dividing an image into a plurality of independent blood vessel sections, performing center line fitting by adopting cubic spline fitting to obtain center line information of each blood vessel section, wherein a terminal node (only one adjacent pixel in a neighborhood), a connecting point (two adjacent pixels in the neighborhood), a bifurcation point (three adjacent pixels in the neighborhood), and a cross point (the number of adjacent pixels in the neighborhood is more than three). And taking the starting point and the end point of each segment of blood vessel as points V in the graph theory, and connecting adjacent points according to neighborhood information to form an edge E in the graph theory so as to establish an undirected graph G (V, E) of the retinal blood vessel network. Then, a feature vector of the blood vessel segment is mut mutextracted, wherein the feature vector comprises direction, gray scale, contrast and diameter information (since OCT-a lacks color information, only the direction and diameter information of OCT-a is mut mutextracted as its feature vector), and the feature vector dimension is reduced by clustering the feature vector at a neighboring point set Vi (i =1,2, …, N) using the advantage set, and is taken as a similarity measure W between two points. A weighted undirected graph G (V, E, W) of retinal blood vessels is established.
2) Determining a starting point, extracting a optic disc region by adopting an automatic segmentation algorithm for fundus color photographs, collecting intersection points of the edge of the optic disc region and blood vessels as a starting point S of a blood vessel tree, and deleting points and edges falling in the optic disc range, wherein the number of the intersection points is the number of the retinal blood vessel trees in principle; for the OCT-a image, a point falling within a pixel of the edge of the image is taken as a starting point.
3) Topology estimation, introducing virtual points, connecting with the starting points of each blood vessel tree, converting the blood vessel topology estimation problem into a minimum graph problem with the virtual points as the starting points, calculating a minimum subgraph by adopting a minimum spanning tree algorithm, and finally removing the virtual points to obtain the final productAnd (4) completing topology estimation of fundus color-lighting blood vessels and OCT-A blood vessels by using independent blood vessel trees.
As a preferred embodiment, "after rough matching of the fundus oculi color photograph and the OCT-a image, performing key point matching to match the topology tree between the two modality images, and obtaining topology information" specifically includes: roughly matching the fundus color photograph with the OCT-A image according to a registration algorithm; intercepting the overlapped part of the OCT-A images in the eye ground color photograph segmentation result after the rough matching and mut mutextracting key nodes, and intercepting the overlapped part of the eye ground color photograph in the OCT-A image segmentation result after the rough matching and mut mutextracting key nodes; matching key nodes in the two modes through a Gaussian regression process; and completing topology tree matching according to the consistency of key nodes contained in the topology tree to obtain topology tree information.
In the embodiment of the specification, the coarse registration of the fundus color photograph and the OCT-A image is realized by using a registration algorithm, but due to the limitation of registration accuracy, the complete matching of the retinal blood vessel and the OCT-A blood vessel cannot be realized, and the pixel-level guidance cannot be realized, so that the topological information transfer algorithm is provided by the invention, and the matching of the blood vessel topological trees in two image modes is realized. The operation method comprises the following steps: firstly, intercepting the overlapped part of an OCT-A image in the fundus color photograph segmentation result according to the registration result, mut mutextracting the key nodes mentioned above, and mut mutextracting the key nodes in the OCT-A segmentation result; matching the key nodes in the two modes by adopting a Gaussian regression process; and finally, completing the matching of the topological trees under the two modes according to the consistency of the key nodes contained in the topological trees.
Specifically, to find the set of corresponding points in the fundus color photograph and OCT-a image, we first define that the coincident feature points between the two modalities are considered as a set of matched points within a six pixel error range. Thus, a set of corresponding points is selected in the result of the coarse registration according to our definition. WhereinRepresenting key points in the fundus color photograph,key points corresponding to fundus color photographs in the OCT-A are shown, and l represents the number of the point pairs. Then assemble the setUsing a Gaussian regression process method as a set of training sets, the corresponding mean in the Gaussian function is estimated by fitting the position coordinates between the corresponding pointsSum varianceThe calculation method is as follows:
wherein K represents a vectorK is a kernel function, a mapping composed of affine transformation and nonlinear mapping is defined,which represents the variance of the noise, is,representing a symmetric matrix of size LxL, in which the elements are calculated by the formula,To representIs a matrix of size L × D, where D is the size of the eigenvector, and in the present invention, since only the coordinate information of the keypoints is used, D = 2.
Then, calculating coordinates of all key points projected in the fundus image in the OCT-A image according to the trained model, and then calculating new matching point pairs according to the definition of the matching points to obtain a more complete corresponding setThen will beFitting a Gaussian regression process by using the training set to obtainIterate in this way, eventually when matching the set of pointsOut of circulation on stabilization, obtainedI.e. the final set of matching points, m is in principle equal to the number of key points in the corresponding retina image part of the OCT-a image.
After the blood vessel topological information and the corresponding point set in the two modes are obtained, matching between the topological trees is realized by judging the contact ratio of key points contained between the two groups of blood vessel trees. Specifically, for any topological tree in fundus color photography, the set of key points contained in the topological tree is countedWhere F represents the fundus color photograph and i represents the corresponding topological tree, and then by comparing the set of keypoints contained in the vascular tree contained in the OCT-A imageWhere O represents an OCT-A image. The two vessel trees with the highest degree of coincidence are considered as matching vessel trees. So far, the matching and fusion of the blood vessel trees in the two modal images are realized.
As a preferred embodiment, "moving arteries and veins of the near vision disk region blood vessels of the fundus color photograph according to the dominance set theory" specifically includes: selecting points in a blood vessel section within a preset multiplying power optic disc range as clustering objects, and extracting a characteristic vector for each point to obtain a point set; clustering the selected point sets according to the advantage sets and dividing all points into two types; defining an artery and a vein according to the mean value of the brightness information of the midpoints in each class; and counting the label distribution of the midpoint of each blood vessel section and carrying out arteriovenous marking on the blood vessel section.
In the embodiment of the specification, the arteries and veins of the fundus color-illumination blood vessels have difference in color and caliber and are obviously represented in the near vision disc area, so that an accurate clustering result can be easily obtained, and meanwhile, the arteries of the retinal blood vessels have higher oxygen content than the veins, so that very strong visual difference is represented in an image and can be used as a key index for distinguishing the arteries and the veins. Therefore, the invention provides the arteriovenous classification method for realizing the blood vessels in the near vision disc region by using the advantage set. The specific operation steps are as follows:
firstly, selecting points in a blood vessel section in a 1.5-to 2-fold optic disc range as clustering objects, and extracting a characteristic vector containing information such as color, contrast, blood vessel width and the like for each point; then, clustering the selected point set by using the advantage set, and dividing all points into two types; taking the mean value of the brightness information of the points in each class as the basis of a classification label, wherein a group of points with high brightness is defined as an artery, and a group with low brightness is defined as a vein; and finally, counting the label distribution of the middle point of each blood vessel section, defining the classification labels of the blood vessel sections by adopting a voting mechanism, if the number of the points with the labels as the artery in the blood vessel sections is more than that of the points with the labels as the vein, considering the blood vessel section as the artery, and otherwise, marking the blood vessel section as the vein.
As a preferred embodiment, "classifying the global moving veins by using a label propagation algorithm according to the topology information to obtain the arteriovenous classification result of the OCT-a blood vessel" specifically includes: transmitting the label of the initial blood vessel section downwards through topological information to realize the complete classification of retinal blood vessels; the arteriovenous classification of the OCT-a vessels is obtained from the complete classification of retinal vessels.
In the embodiment of the present specification, in obtaining a topology structure of a complete retinal blood vessel and an arteriovenous classification label of an initial blood vessel section in fundus color photography, the present invention uses a label propagation algorithm to implement global arteriovenous classification, specifically, it is considered that all blood vessel sections in the same blood vessel tree are consistent with a label of the initial blood vessel section of the blood vessel tree, and therefore, the label of the initial blood vessel section obtained in the above steps is transmitted downwards through the topology information obtained in the above steps to implement complete classification of the retinal blood vessel. The OCT-A blood 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, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set;
the first processing module 22: after the eye fundus color photograph and the OCT-A image are roughly matched, key point matching is carried out, so that the topological trees between the two modal images are matched, and topological information is obtained;
the classification module 23: classifying the arteries and veins of the near vision disc region of the fundus color photograph according to an advantage set theory;
the second processing module 24: and (4) carrying out vein classification on the whole situation by adopting a label propagation algorithm according to the topological information to obtain the arteriovenous classification result of the OCT-A blood vessel.
As a preferred embodiment, the obtaining module specifically includes: a first dividing unit: segmenting the fundus color photograph and the OCT-A image by a morphological thinning algorithm to obtain a blood vessel structure with unit pixel width; a first detection unit: detecting key nodes of a blood vessel structure, and disconnecting the segmentation images at bifurcation points and intersection points in the key nodes to obtain a plurality of independent blood vessel sections; a first processing unit: performing center line fitting on the blood vessel sections by adopting cubic spline fitting to obtain center line information of each section of blood vessel; a first extraction unit: establishing an undirected graph of a retinal vessel network according to the center line information, extracting a characteristic vector of a vessel section, and establishing a weighted undirected graph of a retinal vessel; a second processing unit: mutextracting a optic disc region of the fundus color photograph according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the optic disc region, and determining a starting point of an OCT-A image; a third processing unit: and introducing virtual points to connect the starting points of the blood vessel trees, and obtaining independent blood vessel trees by a minimum spanning tree algorithm to obtain a topological tree set.
As a preferred embodiment, the first processing module specifically includes: a first matching unit: roughly matching the fundus color photograph with the OCT-A image according to a registration algorithm; a second extraction unit: intercepting the overlapped part of the OCT-A images in the eye ground color photograph segmentation result after the rough matching and mut mutextracting key nodes, and intercepting the overlapped part of the eye ground color photograph in the OCT-A image segmentation result after the rough matching and mut mutextracting key nodes; a second matching unit: matching key nodes in the two modes through a Gaussian regression process; a first acquisition unit: and completing topology tree matching according to the consistency of key nodes contained in the topology tree to obtain 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 optic disc range as clustering objects, and extracting a characteristic vector for each point to obtain a point set; a first classification unit: clustering the selected point sets according to the advantage sets and dividing all points into two types; a first defining unit: defining an artery and a vein according to the mean value of the brightness information of the midpoints in each class; a first statistical unit: and counting the label distribution of the midpoint of each blood vessel section and carrying out arteriovenous marking on the blood vessel section.
As a preferred embodiment, the second processing module specifically includes: a fourth processing unit: transmitting the label of the initial blood vessel section downwards through topological information to realize the complete classification of retinal blood vessels; a second acquisition unit: the arteriovenous classification of the OCT-a vessels is obtained from the complete classification of retinal vessels.
Furthermore, those skilled in the art will appreciate that while some embodiments herein include some features included in other embodiments, rather than other features, 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 understand that: the above embodiments are only used for illustrating the technical solutions of the present invention, and not for limiting the same, although the present invention is described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it is possible to modify the solutions described in the above embodiments or to substitute some or all of the technical features of the embodiments, without departing from the scope of the present invention as defined in the claims.
Claims (10)
1. An arteriovenous classification method of an OCT-A image is characterized by comprising the following steps:
acquiring fundus color photographs and OCT-A images, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set;
after roughly matching the fundus color photograph and the OCT-A image, matching key points to match a topology tree between the two modal images to obtain topology information;
classifying the arteries and veins of the near vision disc region of the fundus color photograph according to an advantage set theory;
and carrying out overall motion vein classification by adopting a label propagation algorithm according to the topological information to obtain an arteriovenous classification result of the OCT-A blood vessel.
2. The method for classifying arteriovenous of an OCT-a image according to claim 1, wherein the "acquiring a fundus color photograph and an OCT-a image, performing topology estimation on the segmentation results of the fundus color photograph and the OCT-a image, respectively, and obtaining a topology tree set" specifically comprises:
segmenting the fundus color photograph and the OCT-A image by a morphological thinning algorithm to obtain a blood vessel structure with unit pixel width;
detecting key nodes of the blood vessel structure, and disconnecting the segmentation images at bifurcation points and intersection points in the key nodes to obtain a plurality of independent blood vessel sections;
performing center line fitting on the blood vessel sections by adopting cubic spline fitting to obtain center line information of each section of blood vessel;
establishing an undirected graph of a retinal vessel network according to the center line information, extracting a characteristic vector of the vessel section, and establishing a weighted undirected graph of a retinal vessel;
mutextracting a optic disc region of the fundus color photograph according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the optic disc region, and determining a starting point of an OCT-A image;
and introducing virtual points to connect the starting points of the blood vessel trees, and obtaining independent blood vessel trees by a minimum spanning tree algorithm to obtain a topological tree set.
3. The method for classifying arteriovenous of an OCT-a image according to claim 1, wherein the "performing key point matching after rough matching of the fundus color photograph and the OCT-a image to match a topology tree between two modality images to obtain topology information" specifically comprises:
roughly matching the fundus color photograph with the OCT-A image according to a registration algorithm;
intercepting the overlapped part of the OCT-A images in the eye ground color photograph segmentation result after coarse matching and mut mutextracting key nodes, and intercepting the overlapped part of the eye ground color photograph in the OCT-A image segmentation result after coarse matching and mut mutextracting key nodes;
matching key nodes in the two modes through a Gaussian regression process;
and completing topology tree matching according to the consistency of key nodes contained in the topology tree to obtain topology tree information.
4. The method for classifying arteriovenous vessels of an OCT-a image according to claim 1, wherein "classifying the vessels of the near-vision disc region of the fundus oculi with color illumination according to the dominance set theory" specifically includes:
selecting points in a blood vessel section within a preset multiplying power optic disc range as clustering objects, and extracting a characteristic vector for each point to obtain a point set;
clustering the selected point set according to the advantage set and dividing all points into two types;
defining an artery and a vein according to the mean value of the brightness information of the midpoints in each class;
and counting the label distribution of the midpoint of each blood vessel section and carrying out arteriovenous marking on the blood vessel section.
5. The method of claim 1, wherein the step of obtaining an arteriovenous classification result of an OCT-a vessel by globally performing arteriovenous classification using a label propagation algorithm according to the topology information specifically comprises:
transmitting the label of the initial blood vessel section downwards through the topological information to realize the complete classification of the retinal blood vessels;
obtaining arteriovenous classification of OCT-A blood vessel from the complete classification of retinal blood vessel.
6. An arteriovenous classification device of an OCT-A image, characterized by comprising:
an acquisition module: acquiring fundus color photographs and OCT-A images, and respectively carrying out topology estimation on segmentation results of the fundus color photographs and the OCT-A images to obtain a topology tree set;
a first processing module: after roughly matching the fundus color photograph and the OCT-A image, matching key points to match a topology tree between the two modal images to obtain topology information;
a classification module: classifying the arteries and veins of the near vision disc region of the fundus color photograph according to an advantage set theory;
a second processing module: and carrying out overall motion vein classification by adopting a label propagation algorithm according to the topological information to obtain an arteriovenous classification result of the OCT-A blood vessel.
7. The apparatus for classifying arteriovenous of OCT-A images of claim 6, wherein said acquisition module comprises:
a first dividing unit: segmenting the fundus color photograph and the OCT-A image by a morphological thinning algorithm to obtain a blood vessel structure with unit pixel width;
a first detection unit: detecting key nodes of the blood vessel structure, and disconnecting the segmentation images at bifurcation points and intersection points in the key nodes to obtain a plurality of independent blood vessel sections;
a first processing unit: performing center line fitting on the blood vessel sections by adopting cubic spline fitting to obtain center line information of each section of blood vessel;
a first extraction unit: establishing an undirected graph of a retinal vessel network according to the center line information, extracting a characteristic vector of the vessel section, and establishing a weighted undirected graph of a retinal vessel;
a second processing unit: mutextracting a optic disc region of the fundus color photograph according to an automatic segmentation algorithm, determining a starting point of a blood vessel tree according to the optic disc region, and determining a starting point of an OCT-A image;
a third processing unit: and introducing virtual points to connect the starting points of the blood vessel trees, and obtaining independent blood vessel trees by a minimum spanning tree algorithm to obtain a topological tree set.
8. The apparatus for classifying arteriovenous of OCT-A images of claim 6, wherein said first processing module comprises:
a first matching unit: roughly matching the fundus color photograph with the OCT-A image according to a registration algorithm;
a second extraction unit: intercepting the overlapped part of the OCT-A images in the eye ground color photograph segmentation result after coarse matching and mut mutextracting key nodes, and intercepting the overlapped part of the eye ground color photograph in the OCT-A image segmentation result after coarse matching and mut mutextracting key nodes;
a second matching unit: matching key nodes in the two modes through a Gaussian regression process;
a first acquisition unit: and completing topology tree matching according to the consistency of key nodes contained in the topology tree to obtain topology tree information.
9. The apparatus of claim 6, wherein the classification module comprises:
a third extraction unit: selecting points in a blood vessel section within a preset multiplying power optic disc range as clustering objects, and extracting a characteristic vector for each point to obtain a point set;
a first classification unit: clustering the selected point set according to the advantage set and dividing all points into two types;
a first defining unit: defining an artery and a vein according to the mean value of the brightness information of the midpoints in each class;
a first statistical unit: and counting the label distribution of the midpoint of each blood vessel section and carrying out arteriovenous marking on the blood vessel section.
10. The apparatus for classifying arteriovenous of OCT-A images of claim 6, wherein said second processing module comprises:
a fourth processing unit: transmitting the label of the initial blood vessel section downwards through the topological information to realize the complete classification of the retinal blood vessels;
a second acquisition unit: obtaining arteriovenous classification of OCT-A blood vessel from the complete classification of retinal blood vessel.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010518961.7A CN111797900B (en) | 2020-06-09 | 2020-06-09 | Artery and vein classification method and device for OCT-A image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010518961.7A CN111797900B (en) | 2020-06-09 | 2020-06-09 | Artery and vein classification method and device for OCT-A image |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111797900A true CN111797900A (en) | 2020-10-20 |
CN111797900B 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 | Artery and vein classification method and device for OCT-A image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111797900B (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113450256A (en) * | 2021-06-30 | 2021-09-28 | 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 | Method for detecting retinal vessel bifurcation and intersection points in OCTA image |
WO2023193404A1 (en) * | 2022-04-09 | 2023-10-12 | 中山大学中山眼科中心 | Method for labeling capillaries in fundus color photography on basis of conditional generative adversarial network |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050191627A1 (en) * | 2001-09-28 | 2005-09-01 | Incyte Corporation | Enzymes |
CN101340929A (en) * | 2005-10-28 | 2009-01-07 | 勒帕斯公司 | Compositions and methods for the treatment and prevention of fibrotic, inflammatory, and neovascularization conditions of the eye |
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 |
-
2020
- 2020-06-09 CN CN202010518961.7A patent/CN111797900B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20050191627A1 (en) * | 2001-09-28 | 2005-09-01 | Incyte Corporation | Enzymes |
CN101340929A (en) * | 2005-10-28 | 2009-01-07 | 勒帕斯公司 | Compositions and methods for the treatment and prevention of fibrotic, inflammatory, and neovascularization conditions of the eye |
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)
Title |
---|
DAO-YI YU等: "Retinal capillary perfusion: spatial and temporal heterogeneity", 《PROGRESS IN RETINAL AND EYE RESEARCH》, vol. 70, pages 23 - 54, XP085706938, DOI: 10.1016/j.preteyeres.2019.01.001 * |
YITIAN ZHAO等: "Retinal vascular network topology reconstruction and artery/vein classification via dominant set clustering", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》, vol. 39, no. 2, pages 341 - 356, XP011769474, DOI: 10.1109/TMI.2019.2926492 * |
陈艳花: "血管内OCT图像序列中血管三维重建方法的研究", 《中国优秀硕士学位论文全文数据库 医药卫生科技辑》, no. 03, pages 062 - 27 * |
黄晓莉: "糖尿病视网膜微血管病变的基础与临床研究", 《中国博士学位论文全文数据库 医药卫生科技辑》, no. 09, pages 065 - 10 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113450256A (en) * | 2021-06-30 | 2021-09-28 | 重庆理工大学 | Automatic aorta segmentation method and system based on CT image of level set |
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 | Method for detecting retinal vessel bifurcation and intersection points in OCTA image |
WO2023193404A1 (en) * | 2022-04-09 | 2023-10-12 | 中山大学中山眼科中心 | Method for labeling capillaries in fundus color photography on basis of conditional generative adversarial network |
Also Published As
Publication number | Publication date |
---|---|
CN111797900B (en) | 2024-04-09 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
WO2021169128A1 (en) | Method and apparatus for recognizing and quantifying fundus retina vessel, and device and storage medium | |
Cao et al. | Hierarchical method for cataract grading based on retinal images using improved Haar wavelet | |
CN108764286B (en) | Classification and identification method of feature points in blood vessel image based on transfer learning | |
Niemeijer et al. | Segmentation of the optic disc, macula and vascular arch in fundus photographs | |
Goldbaum et al. | Automated diagnosis and image understanding with object extraction, object classification, and inferencing in retinal images | |
Mendonça et al. | Automatic localization of the optic disc by combining vascular and intensity information | |
Ding et al. | A novel deep learning pipeline for retinal vessel detection in fluorescein angiography | |
CN111797900B (en) | Artery and vein classification method and device for OCT-A image | |
CN112465772B (en) | Fundus colour photographic image blood vessel evaluation method, device, computer equipment and medium | |
Chetoui et al. | Explainable end-to-end deep learning for diabetic retinopathy detection across multiple datasets | |
CN111797901A (en) | Retinal artery and vein classification method and device based on topological structure estimation | |
CN113420826B (en) | Liver focus image processing system and image processing method | |
US11783488B2 (en) | Method and device of extracting label in medical image | |
CN112164043A (en) | Method and system for splicing multiple fundus images | |
Fu et al. | Optic disc segmentation by U-net and probability bubble in abnormal fundus images | |
Oloumi et al. | Computer-aided diagnosis of plus disease in retinal fundus images of preterm infants via measurement of vessel tortuosity | |
Lin et al. | Retinal vascular tree reconstruction with anatomical realism | |
Shaik et al. | Glaucoma identification based on segmentation and fusion techniques | |
Hu et al. | Multi-scale interactive network with artery/vein discriminator for retinal vessel classification | |
CN116309235A (en) | Fundus image processing method and system for diabetes prediction | |
Mangrulkar | Retinal image classification technique for diabetes identification | |
CN116664592A (en) | Image-based arteriovenous blood vessel separation method and device, electronic equipment and medium | |
Wang et al. | Study on correlation between subjective and objective metrics for multimodal retinal image registration | |
Kulkarni et al. | Eye gaze–based optic disc detection system | |
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 |