CN106815853B

CN106815853B - Method and device for segmenting retinal blood vessels in fundus image

Info

Publication number: CN106815853B
Application number: CN201611158462.1A
Authority: CN
Inventors: 崔彤哲; 王小珂; 周永新; 谌记文; 陈国桢; 孙毅
Original assignee: Hinacom Software And Technology Ltd
Current assignee: Hinacom Software And Technology Ltd
Priority date: 2016-12-14
Filing date: 2016-12-14
Publication date: 2020-06-05
Anticipated expiration: 2036-12-14
Also published as: CN106815853A

Abstract

The invention discloses a method and a device for segmenting retinal blood vessels in an fundus image. Wherein, the method comprises the following steps: acquiring a fundus image; processing the fundus image based on the Hatheri matrix to obtain a first retinal vessel map; performing binarization processing on the first retinal vessel map to obtain a second retinal vessel map; and reconstructing interrupted blood vessels in the second retinal blood vessel map to obtain a third retinal blood vessel map. The invention solves the technical problems of high complexity and inaccurate segmentation result in the prior art when retinal blood vessels in the fundus image are segmented.

Description

Method and device for segmenting retinal blood vessels in fundus image

Technical Field

The invention relates to the field of computer internet, in particular to a method and a device for segmenting retinal blood vessels in a fundus image.

Background

Among various kinds of information that humans receive from the outside, about 80% or more of the information comes from the vision. The eye is the most important sensory organ of the human body. The optic nerve of the fundus is the part of the brain which extends outwards and is an important visual organ, and fundus diseases often cause hypopsia or permanent loss. In addition, ocular fundus changes due to ocular fundus circulatory disorder diseases and systemic diseases may be manifested to varying degrees in the retina and choroid. Therefore, the fundus picture can be processed to provide an early diagnosis or prognosis judgment function for certain general diseases.

The traditional fundus examination method depends on fundus pictures and observation and analysis of doctors for examination, and is difficult to accurately, objectively and comprehensively analyze fundus images due to the influence of various factors such as shooting, flushing and expanding, the resolving power of human eyes and the like. The fundus retinal blood vessel is used as an important characteristic of a fundus image and is a precondition and a basis for various analysis works, so that the fundus retinal blood vessel image extraction method has extremely important significance for researching blood vessel segmentation, and whether the outline of the fundus retinal blood vessel can be completely extracted is a key factor for the quality of subsequent fundus image processing and analysis.

The segmentation of the retinal blood vessel can position the position of the blood vessel to reduce misdiagnosis of lesion, and meanwhile, the blood vessel tree can be detected through the segmentation of the blood vessel and the geometric relationship of the blood vessel tree is established to assist the examination of the optic disc and the recess, and various parameters of the blood vessel can be obtained through the segmented blood vessel to examine the abnormality of the blood vessel. One of the most important signs that hypertension and diabetes exhibit on the retina is abnormal malformation of blood vessels, such as: arteriovenous imprints, copper wire or silver wire arteries, and changes in tortuosity and branch angle, among others. However, in the face of large-scale crowd analysis and observation, limitation, lack of accuracy, consistency and repeatability occur, so that retinal blood vessels need to be divided and displayed to a viewer as a retinal image with blood vessel morphological measurement data (such as tortuosity, branch angle and the like) and abnormality detection marks (such as arteriovenous indentation, copper wire-shaped or silver wire-shaped artery, abnormal protruding area and the like) to help the viewer to make final judgment. The viewer can observe the form of the retina and the detection result of the abnormality only through the user interface, and then identify the pathological change signs of the retina. Therefore, the screening efficiency and accuracy can be greatly improved. The first condition for analyzing retinal blood vessels is that blood vessels can be accurately segmented from fundus images.

With the rapid development of computers, researchers have developed methods for automatically segmenting fundus blood vessels, which are largely classified into supervised methods and unsupervised methods. Supervised segmentation methods rely on a classification training set for training and generally have better segmentation results than unsupervised. Unsupervised segmentation methods include morphology-based segmentation methods, tracking-based segmentation methods, and snake model-based algorithms. Although the supervised method can obtain a better blood vessel segmentation result, the method needs a better training set for training, for example, when the blood vessel segmentation is performed by using a neural network method, a training set with various changing pictures is needed for training, and if the training degree is not enough, the result is biased, which causes limitation in use. In the unsupervised segmentation method, the algorithm based on morphology, such as region growing, morphology reconstruction and the like, has high time calculation complexity which can reach as much as 13 minutes; although the width of the blood vessel can be accurately obtained when the blood vessel is segmented, the tracking method is easy to fall into impasse in the intersection points and branches of the blood vessel, so that inaccurate blood vessel segmentation is obtained; the contour model-based method has high computational complexity and time complexity, and the general segmentation time is ten minutes and several minutes. Therefore, the above-described fundus blood vessel segmentation methods have both unsuitability and limitation in practical applications.

Aiming at the problems of high complexity and inaccurate segmentation result in retinal vessel segmentation in an fundus image in the prior art, no effective solution is provided at present.

Disclosure of Invention

The embodiment of the invention provides a method and a device for segmenting retinal blood vessels in an fundus image, which at least solve the technical problems of high complexity and inaccurate segmentation result when segmenting the retinal blood vessels in the fundus image in the prior art.

According to an aspect of the embodiments of the present invention, there is provided a method for segmenting retinal blood vessels in a fundus image, including: acquiring a fundus image; processing the fundus image based on the Hatheri matrix to obtain a first retinal vessel map; performing binarization processing on the first retinal vessel map to obtain a second retinal vessel map; and reconstructing interrupted blood vessels in the second retinal blood vessel map to obtain a third retinal blood vessel map.

According to another aspect of the embodiments of the present invention, there is also provided a segmentation apparatus for retinal blood vessels in a fundus image, including: an acquisition module for acquiring a fundus image; the first processing module is used for processing the fundus image based on the Hatheri matrix to obtain a first retinal vessel map; the second processing module is used for carrying out binarization processing on the first retinal vessel map to obtain a second retinal vessel map; and the reconstruction module is used for reconstructing the interrupted blood vessel in the second retinal blood vessel map to obtain a third retinal blood vessel map.

In the embodiment of the invention, by acquiring the fundus image, the fundus image is firstly processed based on the Hatheri matrix to obtain a first retinal vessel map, then the first retinal vessel map is subjected to binarization processing to obtain a second retinal vessel map, finally, interrupted vessels in the second retinal vessel map are reconstructed to obtain a third retinal vessel map, so that the purpose of segmenting retinal vessels from the fundus image is achieved, the retinal vessels can be accurately segmented from the fundus image by using a Hathera matrix, the reconstruction of the interrupted blood vessel after the second retinal vessel map is obtained can lead the retinal vessel to be more accurate, is beneficial to the subsequent morphological measurement of the blood vessel and the subsequent analysis and research of diseases, and further solves the technical problems of high complexity and inaccurate segmentation result when segmenting retinal blood vessels in the fundus image in the prior art.

Drawings

The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the invention and together with the description serve to explain the invention without limiting the invention. In the drawings:

fig. 1 is a flowchart of a segmentation method for retinal blood vessels in a fundus image according to embodiment 1 of the present invention;

FIG. 2 is a vessel pixel diagram according to embodiment 1 of the present invention;

FIG. 3 is a weighted directed node graph according to embodiment 1 of the present invention;

fig. 4 is a structural diagram of a segmentation apparatus for retinal blood vessels in a fundus image according to embodiment 2 of the present invention;

fig. 5 is a block diagram of an alternative segmentation apparatus for retinal blood vessels in a fundus image according to embodiment 2 of the present invention;

fig. 6 is a block diagram of an alternative segmentation apparatus for retinal blood vessels in a fundus image according to embodiment 2 of the present invention;

fig. 7 is a block diagram of an alternative segmentation apparatus for retinal blood vessels in a fundus image according to embodiment 2 of the present invention; and

fig. 8 is a structural diagram of an alternative segmentation apparatus for retinal blood vessels in a fundus image according to embodiment 2 of the present invention.

Detailed Description

In order to make the technical solutions of the present invention better understood, 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 only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.

Example 1

In accordance with an embodiment of the present invention, there is provided a method embodiment of a method for segmenting retinal blood vessels in a fundus image, it should be noted that the steps illustrated in the flowchart of the figure may be performed in a computer system such as a set of computer executable instructions and that, although a logical order is illustrated in the flowchart, in some cases, the steps illustrated or described may be performed in an order different than here.

Fig. 1 is a method for segmenting retinal blood vessels in a fundus image according to an embodiment of the present invention, as shown in fig. 1, the method comprising the steps of:

in step S102, a fundus image is acquired.

Specifically, the fundus image may be acquired by taking a photograph.

Step S104, the fundus image is processed based on the Hatheri matrix, and a first retinal vessel map is obtained.

In particular, the Hessian matrix, which is a Hessian matrix, is a function of a real-valued multivariate function f (x)₁,x₂,...,x_n) If the function f (x)₁,x₂,...,x_n) If the second partial derivatives of (a) are present, the Hessian matrix is defined as:

wherein D is_iRepresenting the differential operator on the ith variable,

i.e. f (x)₁,x₂,...,x_n) The Hessian matrix of (a) is:

the Hessian matrix is actually the second derivative in the case of multivariate and describes the change in gray scale gradient in each direction. In the n eigenvalues of the Hessian matrix, the eigenvector corresponding to the eigenvalue with the largest amplitude represents the direction of the largest curvature of the P point, and the eigenvector corresponding to the eigenvalue with the smallest amplitude represents the direction of the smallest curvature.

By processing the fundus image using the hese matrix, a first retinal blood vessel map showing the presence of blood vessels can be obtained.

And step S106, performing binarization processing on the first retinal vessel map to obtain a second retinal vessel map.

Specifically, the first retinal blood vessel map in color can be converted into a second retinal blood vessel map in black and white by performing binarization processing on the first retinal blood vessel map.

And step S108, reconstructing interrupted blood vessels in the second retinal blood vessel map to obtain a third retinal blood vessel map.

In particular, the morphology of the retinal blood vessels contains valuable information, but there may be a case where the blood vessels are discontinuous in the second retinal blood vessel map obtained as described above, and if a continuous blood vessel is divided into many segments, the morphological measurement of the blood vessels is inaccurate, so that it is necessary to find the interrupted blood vessels and reconnect the interrupted portions.

In the above-described embodiment of the present invention, by acquiring the fundus image, the fundus image is first processed based on the haise matrix to obtain the first retinal vessel map, then the first retinal vessel map is subjected to binarization processing to obtain a second retinal vessel map, finally, interrupted vessels in the second retinal vessel map are reconstructed to obtain a third retinal vessel map, so that the purpose of segmenting retinal vessels from the fundus image is achieved, the retinal vessels can be accurately segmented from the fundus image by using a Hathera matrix, the reconstruction of the interrupted blood vessel after the second retinal vessel map is obtained can lead the retinal vessel to be more accurate, is beneficial to the subsequent morphological measurement of the blood vessel and the subsequent analysis and research of diseases, and further solves the technical problems of high complexity and inaccurate segmentation result when segmenting retinal blood vessels in the fundus image in the prior art.

In an alternative embodiment, before step S104, step S202 is included: preprocessing the fundus image; wherein the pretreatment comprises: and acquiring a green channel image of the fundus image, and performing filtering and contrast enhancement processing on the green channel image.

In an alternative embodiment, the processing of filtering and enhancing contrast for the green channel image in step S202 includes: filtering the green channel image by using an image processing mean filtering algorithm; and carrying out contrast enhancement processing on the green channel image subjected to filtering processing by using an adaptive histogram equalization algorithm.

Specifically, an image processing mean filtering algorithm can be adopted to filter and remove noise from the green channel image, wherein the filtering operator adopts three times of the radius of the video disc. The green channel image after filtering may be subjected to contrast processing by Adaptive Histogram Equalization (AHE), which is a computer image processing technique for improving the contrast of an image, to enhance the contrast. The conventional histogram equalization algorithm uses the same histogram transformation for the pixels of the entire image, and is good for an image in which the distribution of pixel values is relatively equalized, however, for an image in which the distribution of pixel values is unbalanced, that is, an image including a portion significantly darker or brighter than other regions, such as a fundus image, the contrast of the portion significantly darker or brighter than other regions will not be effectively enhanced, unlike the conventional histogram equalization algorithm, the AHE algorithm changes the image contrast by calculating a local histogram of the image and then redistributing the brightness. Therefore, the adaptive histogram equalization algorithm is more suitable for the present invention to improve the local contrast of the fundus image and to obtain more image details.

In an alternative embodiment, step S106 includes step S302 of performing binarization processing on the first retinal vessel map using a local entropy threshold segmentation algorithm.

Specifically, according to the superiority of the entropy threshold segmentation algorithm, the local entropy threshold segmentation algorithm is adopted to carry out binarization processing on the first retinal vessel map, so that the background and the target can be well separated.

Introduction of the principle of the local entropy thresholding algorithm: let f (x, y) be the gray value at the point (x, y) in the image, f (x, y)>0, for an image of size M N, define H_fFor the entropy of the image, i.e.:

in the formula

Is a gray scale distribution. If M N is a local window of the image, then it is called H_fIs the local entropy of the image.

The local entropy can reflect the discrete degree of the image gray scale, the image gray scale is relatively uniform at the place with the maximum local entropy, the image gray scale has larger discreteness at the place with the small local entropy, and the position with large pixel value difference in the window is the edge of the target and the background in general, so the target with the relatively uniform gray scale can be segmented according to the local entropy. Since the local entropy is a contribution common to multiple pixel points in the window and is insensitive to noise of a single point, the local entropy also has a filtering effect. The logarithm operand in the definition of the local entropy formula is larger, the running speed is slow, and the definition is 0<p_ij<<1, the approximation formula can be obtained by rounding off the high power (or equivalently infinitesimal) with a taylor expansion:

calculating the local entropy value of the image according to the formula, determining the position of the maximum local entropy of the image, calculating the similarity between the local entropy of the image and the segmented local entropy from the position of the maximum local entropy, and determining the segmented region according to the similarity. When the segmentation is finished, a binary target image of the image can be obtained.

In an alternative embodiment, after step S106, a step S402 of post-processing the second retinal vessel map is included; wherein the post-processing comprises: and determining a lesion region of the second retinal blood vessel map, and removing noise caused by the lesion region.

In particular, since the fundus image may have internal variations including lesions, it is very difficult to segment perfect blood vessels, and therefore post-processing of the second retinal blood vessel map is required.

In an alternative embodiment, the determining the lesion region of the second retinal vessel map in step S402 includes:

step S502, a connected region is marked on the second retina blood vessel icon.

Step S504, an isolated region is determined, wherein the isolated region is a connected region with the area smaller than a preset threshold value.

In step S506, the lesion area is determined according to the eccentricity, circularity and/or linearity of the isolated area.

Specifically, when determining a lesion region of the second retinal vessel map, firstly, a connected region needs to be injected into the second retinal vessel map, the connected region with the area smaller than the preset threshold is determined according to the preset threshold, the connected region with the area smaller than the preset threshold can be defined as an isolated region, and then some geometric characteristics of the isolated region, including but not limited to the area, the circumference, the euler number, the eccentricity, the circularity, the linearity, and the like of the isolated region, are calculated.

Wherein, the Euler number is the number of the communicating areas minus the number of the holes in the communicating areas; eccentricity is the eccentricity of an ellipse consisting of the major axis and the minor axis of an isolated region, and the calculation formula is as follows:

a denotes a major axis of the isolated region, b denotes a minor axis of the isolated region; the circularity is a measure of the degree of similarity between an isolated region and a circle, and is obtained from a calculation formula of the circumference of the circle and the area of the circle, and the calculation formula of the circularity is:

A_sdenotes the area of the isolated region, L_sRepresenting the perimeter of the isolated region; the linearity can be obtained according to the ratio of the perimeter to the area of the isolated region, and the specific calculation formula is as follows:

the noise caused by the pathological changes is generally in a circular or elliptical shape, while the blood vessel shape is generally in a linear shape, so after calculating the eccentricity, circularity and linearity of the isolated region, the eccentricity, circularity and linearity can be compared with a preset eccentricity threshold, circularity threshold and linearity threshold, it should be noted that one or more factors of the eccentricity, circularity and linearity can be selected for comparison, the pathological change region is determined through comparison, and then the noise caused by the pathological change region is removed, so that the second retinal blood vessel map is more accurate.

In an alternative embodiment, step S108 includes:

in step S602, a break point of a broken blood vessel in the second retinal blood vessel map is determined.

And S604, determining a destination point corresponding to the interruption point by using a Dijkstra algorithm, wherein the destination point is a node on a vascular skeleton, and the vascular skeleton is a connected region with the largest area in the second retinal vascular map.

And step S606, connecting the interruption point and the destination point corresponding to the interruption point.

Specifically, when determining the interruption point of the blood vessel interrupted in the second retinal blood vessel map in step S602, a connected region mark may be performed on the second retinal blood vessel map, the largest connected region is selected as a blood vessel skeleton, a 3 × 3 domain depth search traversal is performed on each pixel of other connected regions, if a pixel has only one connection point with a pixel value of 1 in the 3 × 3 domain, the connection point is marked as an end point of the connected region, and a point closest to the visual disk in the end point of each connected region is taken as an interruption point, as shown in fig. 2, a non-shaded square in fig. 2 represents a blood vessel pixel, and the pixel value is 1, and it can be seen that the blood vessel pixel located at the center has only one connection point with a pixel value of 1 in the 3 × 3 domain.

Specifically, when determining the destination point corresponding to the interruption point by using Dijkstra algorithm, that is, Dijkstra algorithm in step S604, it is assumed that a center line pixel of a blood vessel in the first retinal blood vessel graph is higher than a pixel at an edge of the blood vessel, a pixel value of the highest pixel value in the 7 × 7 field of the interruption point in the connected region is selected as a source node of the graph, other pixels form other nodes of the graph, and the boundary cost between the node and the node is composed of pixel cost and direction cost, where the pixel cost is a pixel intensity difference between two nodes, and the direction cost refers to a change in direction when the source node tracks the node. And searching the shortest path of the source node by utilizing a Dijkstra algorithm until the searched node is the point on the blood vessel skeleton, and finishing the search, wherein the searched node on the blood vessel skeleton is the target point.

Among them, Dijkstra's algorithm is a typical shortest path search algorithm for calculating the shortest path from one node to all other nodes. The main characteristic is that the expansion is towards the outer layer by taking the starting point as the center until the expansion reaches the end point. The algorithm idea is that G (V, E) is a directed graph with weight, vertex set V in the graph is divided into two groups, the first group is a vertex set S (only a source node is in the S initially, and every time a shortest path is obtained, the corresponding vertex is added into the set S until all the vertices are added into the S); the second group is a set of vertices U for which no shortest path is determined, and the shortest path length from source node V to each vertex in S is always kept no greater than the shortest path length from source node V to any vertex in U during the joining process. The algorithm comprises the following steps:

a. initially, S only contains the source point, i.e., S ═ v, where v is 0 in distance. U includes vertices other than v, i.e., U ═ rest vertices, where v has an edge with vertex U in U, then < U, v > is normally weighted, and where U is not an edge-out adjacency point for v, then the < U, v > is weighted to ∞.

b. And selecting a vertex k with the minimum distance v from the U, and adding k into S (the selected distance is the length of the shortest path from v to k).

c. Modifying the distance of each vertex in the U by taking k as a newly considered middle point; if the distance from the source point v to the vertex u (passing through the vertex k) is shorter than the original distance (not passing through the vertex k), the distance value of the vertex u is modified, and the weight of the distance of the vertex k of the modified distance value is added to the upper side.

d. Repeating steps b and c until all vertices are contained in S.

As shown in fig. 3, fig. 3 is a directed graph with weights, initially, S ═ a }, node B with the shortest path is added to obtain S ═ a, B }, then shortest path node C is added to obtain S ═ a, B, C }, and finally node D is added to obtain S ═ a, B, C, D }, that is, shortest path a- > B- > C- > D from a to D is obtained.

Specifically, in step S602, after the interruption point and the destination point are obtained, the termination point and the destination point are connected, and the reconnection of the blood vessel is completed.

Example 2

According to an embodiment of the present invention, there is provided a product embodiment of a device for segmenting retinal blood vessels in a fundus image, and fig. 4 is a device for segmenting retinal blood vessels in a fundus image according to an embodiment of the present invention, and as shown in fig. 4, the device includes an obtaining module 101, a first processing module 103, a second processing module 105, and a reconstructing module 107.

The device comprises an acquisition module 101, a storage module and a processing module, wherein the acquisition module is used for acquiring fundus images; the first processing module 103 is used for processing the fundus image based on the Hatheri matrix to obtain a first retinal vessel map; the second processing module 105 is configured to perform binarization processing on the first retinal vessel map to obtain a second retinal vessel map; and a reconstructing module 107, configured to reconstruct the interrupted blood vessel in the second retinal blood vessel map to obtain a third retinal blood vessel map.

In the above embodiment, the fundus image is acquired by the acquisition module 101, the fundus image is processed by the first processing module 103 based on the haise matrix to obtain the first retinal vessel map, then the first retinal vessel map is binarized by the second processing module 105 to obtain the second retinal vessel map, and finally the blood vessels interrupted in the second retinal vessel map are reconstructed by the reconstruction module 107 to obtain the third retinal vessel map, so that the retinal vessel is segmented from the fundus image.

It should be noted here that the above-mentioned obtaining module 101, the first processing module 103, the second processing module 105 and the reconstructing module 107 correspond to steps S102 to S108 in embodiment 1, and the above-mentioned modules are the same as the examples and application scenarios realized by the corresponding steps, but are not limited to what is disclosed in embodiment 1. It should be noted that the modules described above as part of an apparatus may be implemented in a computer system such as a set of computer-executable instructions.

In an alternative embodiment, as shown in fig. 5, the apparatus further comprises: a preprocessing module 201, configured to preprocess the fundus image before the first processing module 103 processes the fundus image based on the haise matrix; wherein the pretreatment comprises: and acquiring a green channel image of the fundus image, and performing filtering and contrast enhancement processing on the green channel image.

In an alternative embodiment, the pre-processing module 201 performs filtering and contrast enhancement on the green channel image, including: filtering the green channel image by using an image processing mean filtering algorithm; and carrying out contrast enhancement processing on the green channel image subjected to filtering processing by using an adaptive histogram equalization algorithm.

In an alternative embodiment, the second processing module 105 performs binarization processing on the first retinal vessel map, including: and carrying out binarization processing on the first retinal vessel map by using a local entropy threshold segmentation algorithm.

In an alternative embodiment, as shown in fig. 6, the apparatus further comprises: a post-processing module 301, configured to perform post-processing on the second retinal blood vessel map after the second retinal blood vessel map is obtained by the second processing module; wherein the post-processing comprises: and determining a lesion region of the second retinal blood vessel map, and removing noise caused by the lesion region.

In an alternative embodiment, as shown in fig. 7, the post-processing module 301 comprises an annotation module 401, a first determination module 403, and a second determination module 405. The labeling module 401 is configured to label a second retinal blood vessel icon with a connected region; the first determining module 403 is configured to determine an isolated region, where the isolated region is a connected region with an area smaller than a preset threshold; the second determination module 405 is used to determine the lesion area based on the eccentricity, circularity, and/or linearity of the isolated area.

In an alternative embodiment, as shown in fig. 8, the reconstruction module 107 includes a third determination module 501, a fourth determination module 503, and a connection module 505. Wherein the third determination module 501 is configured to determine a break point of a broken blood vessel in the second retinal blood vessel map; the fourth determining module 503 is configured to determine a destination point corresponding to the interruption point by using Dijkstra algorithm, where the destination point is a node on a blood vessel skeleton, and the blood vessel skeleton is a connected region with a largest area in the second retinal blood vessel map; the connection module 505 is used for connecting the interruption point and the destination point corresponding to the interruption point.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

In the above embodiments of the present invention, the descriptions of the respective embodiments have respective emphasis, and for parts that are not described in detail in a certain embodiment, reference may be made to related descriptions of other embodiments.

In the embodiments provided in the present application, it should be understood that the disclosed technology can be implemented in other ways. The above-described embodiments of the apparatus are merely illustrative, and for example, the division of the units may be a logical division, and in actual implementation, there may be another division, for example, multiple units or components may be combined or integrated into another system, or some features may be omitted, or not executed. In addition, the shown or discussed mutual coupling or direct coupling or communication connection may be an indirect coupling or communication connection through some interfaces, units or modules, and may be in an electrical or other form.

The units described as separate parts may or may not be physically separate, and parts displayed as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of the embodiment.

In addition, functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may exist alone physically, or two or more units are integrated into one unit. The integrated unit can be realized in a form of hardware, and can also be realized in a form of a software functional unit.

The integrated unit, if implemented in the form of a software functional unit and sold or used as a stand-alone product, may be stored in a computer readable storage medium. Based on such understanding, the technical solution of the present invention may be embodied in the form of a software product, which is stored in a storage medium and includes instructions for causing a computer device (which may be a personal computer, a server, or a network device) to execute all or part of the steps of the method according to the embodiments of the present invention. And the aforementioned storage medium includes: a U-disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a removable hard disk, a magnetic or optical disk, and other various media capable of storing program codes.

The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims

1. A method for segmenting retinal blood vessels in a fundus image, comprising:

acquiring a fundus image;

processing the fundus image based on a Hatheri matrix to obtain a first retinal vessel map;

performing binarization processing on the first retinal vessel map to obtain a second retinal vessel map;

reconstructing interrupted blood vessels in the second retinal vessel map to obtain a third retinal vessel map,

wherein reconstructing the disrupted blood vessels in the second retinal vessel map comprises: determining a break point of a broken blood vessel in the second retinal vessel map; determining a destination point corresponding to the break point by using a Dijkstra algorithm, wherein the destination point is a node on a vascular skeleton, and the vascular skeleton is a connected region with the largest area in the second retinal vascular map; and connecting the interruption point and the destination point corresponding to the interruption point.

2. The method of claim 1, wherein prior to processing the fundus image based on the Hathers matrix, comprising: preprocessing the fundus image;

wherein the pre-processing comprises: and acquiring a green channel image of the fundus image, and performing filtering and contrast enhancement processing on the green channel image.

3. The method of claim 2, wherein filtering and contrast enhancing the green channel image comprises:

filtering the green channel image by using an image processing mean filtering algorithm;

and performing contrast enhancement processing on the green channel image subjected to the filtering processing by using an adaptive histogram equalization algorithm.

4. The method according to claim 1, wherein the binarizing processing of the first retinal blood vessel map includes: and carrying out binarization processing on the first retinal vessel map by using a local entropy threshold segmentation algorithm.

5. The method of claim 1, wherein after obtaining the second retinal vessel map, comprising:

post-processing the second retinal vessel map;

wherein the post-processing comprises: determining a lesion region of the second retinal vessel map, and removing noise caused by the lesion region.

6. The method of claim 5, wherein determining the lesion area of the second retinal vessel map comprises:

labeling a connected region on the second retinal vessel map;

determining an isolated region, wherein the isolated region is the connected region with the area smaller than a preset threshold value;

determining the lesion area according to the eccentricity, circularity and/or linearity of the isolated area.

7. A device for segmenting retinal blood vessels in a fundus image, comprising:

an acquisition module for acquiring a fundus image;

the first processing module is used for processing the fundus image based on the Hatheri matrix to obtain a first retinal vessel map;

the second processing module is used for carrying out binarization processing on the first retinal vessel map to obtain a second retinal vessel map;

a reconstruction module for reconstructing the interrupted blood vessel in the second retinal blood vessel map to obtain a third retinal blood vessel map,

wherein the reconstruction module comprises a third determination module, a fourth determination module and a connection module, wherein the third determination module is used for determining the break point of the broken blood vessel in the second retinal blood vessel map; the fourth determining module is configured to determine a destination point corresponding to the interruption point by using a Dijks tra algorithm, where the destination point is a node on a blood vessel skeleton, and the blood vessel skeleton is a connected region with a largest area in the second retinal blood vessel map; the connection module is used for connecting the interruption point and the destination point corresponding to the interruption point.

8. The apparatus of claim 7, further comprising: a preprocessing module for preprocessing the fundus image before the first processing module processes the fundus image based on a Hatheri matrix;

9. The apparatus of claim 7, further comprising: a post-processing module, configured to perform post-processing on the second retinal vessel map after the second retinal vessel map is obtained by the second processing module;