Method and apparatus for segmenting retinal blood vessels in fundus images

Abstract

The application belongs to the technical field of computer vision and medical image processing, and particularly relates to a retinal blood vessel segmentation method and device in fundus images, aiming to solve the problems of poor adaptability to different image qualities, weak anti-interference ability to composite noise and low blood vessel reconstruction accuracy in the prior art. The method comprises the following steps: acquiring a fundus image to be processed; analyzing the fundus image to determine prior analysis indexes and noise classification information, and then determining processing parameters; processing the fundus image by using the processing parameters and performing noise removal processing to generate an initial blood vessel segmentation map; identifying breakpoints and candidate destination points in the initial blood vessel segmentation map; performing multi-dimensional checking between the breakpoints and each candidate destination point to determine the optimal connection and generate a reconstructed retinal blood vessel segmentation map. The application significantly improves the accuracy, robustness and adaptability to complex lesion images through adaptive parameter adjustment and multi-dimensional checking reconstruction.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of computer vision and medical image processing, and particularly relates to a retinal blood vessel segmentation method and device in fundus images. BACKGROUND

[0002] The fundus retinal blood vessels are the only blood vessel system that can be directly and non-invasively observed in the human body. Changes in the morphological structure, such as the tortuosity, diameter, branching angle, arteriovenous diameter ratio (AVR) of the blood vessels, are closely related to many systemic cardiovascular and cerebrovascular diseases such as hypertension, diabetes, and arteriosclerosis. Therefore, the segmentation of retinal blood vessels in fundus images is one of the key technologies for computer-aided diagnosis, and its accuracy directly affects the early screening and condition assessment of various cardiovascular and cerebrovascular diseases.

[0003] Existing retinal blood vessel segmentation methods, such as the Hessian matrix method based on multi-scale analysis, can effectively enhance the linear structure of blood vessels, but usually use fixed calculation parameters. This "one-size-fits-all" approach has poor adaptability when facing fundus images of different sources and qualities. For example, for low-contrast or high-noise images, fixed parameters can easily lead to missed detection of a large number of fine blood vessels or false detection of noise; and for complex images with dense blood vessels or lesions, the segmentation accuracy is also difficult to guarantee.

[0004] In addition, after preliminary segmentation, due to image artifacts, lesion occlusion, or algorithm limitations, the blood vessel network often appears to be broken and discontinuous. Traditional blood vessel reconstruction methods mostly rely on simple shortest path connection or single morphological rules, which can easily produce physiologically unreasonable false connections when dealing with complex topological structures such as blood vessel intersections and branches, and can mistakenly connect blood vessel segments that do not belong to the same branch, which seriously affects the reliability of subsequent blood vessel morphological quantitative analysis. At the same time, the complex noise formed by the high-brightness reflection of the optic disc area, hemorrhagic spots, and exudates, etc. also brings great challenges to accurate segmentation, and the existing technology lacks an effective mechanism for hierarchical and targeted suppression of these different types of noise.

[0005] Therefore, how to design a retinal blood vessel segmentation method that can adaptively adjust parameters to adapt to different image qualities, intelligently verify and accurately reconstruct interrupted blood vessels, and effectively suppress the interference of complex noise, has become a technical problem to be solved in the field. SUMMARY

[0006] In order to solve the above problems in the prior art, i.e. to solve the problems of poor adaptability of the segmentation algorithm to different image qualities, weak anti-interference ability to complex noise, and low accuracy of blood vessel reconstruction in the prior art, on the one hand, the present application provides a retinal blood vessel segmentation method in a fundus image, comprising the following steps:

[0007] obtaining an eye fundus image to be processed;

[0008] analyzing the eye fundus image to determine at least one priori analysis index representing distribution characteristics of the image and to identify noise regions in the image to generate noise classification information;

[0009] dynamically determining processing parameters for extracting blood vessel segmentation features based on the priori analysis index and the noise classification information;

[0010] processing the eye fundus image using the dynamically determined processing parameters and performing noise removal processing based on the noise classification information to generate an initial blood vessel segmentation map containing interrupted blood vessels;

[0011] identifying interruption points of blood vessels in the initial blood vessel segmentation map and candidate destination points on a blood vessel skeleton;

[0012] performing multi-dimensional verification between the interruption points and each of the candidate destination points to determine a destination point corresponding to the interruption point and to connect the destination point to the interruption point to generate a reconstructed retinal blood vessel segmentation map.

[0013] In some preferred embodiments, the priori analysis index includes at least one of the following:

[0014] a blood vessel density obtained by statistical analysis of a sliding window on the eye fundus image;

[0015] a noise intensity calculated by a ratio of a global gray variance and an average edge gradient of the eye fundus image;

[0016] a gray uniformity calculated by a skewness of a gray distribution of the eye fundus image.

[0017] In some preferred embodiments, the processing parameters include at least a window size for Hessian matrix calculation and a feature screening threshold for screening blood vessel features; and the dynamically determining the processing parameters for extracting blood vessel segmentation includes:

[0018] matching an initial window size according to the blood vessel density and the noise intensity;

[0019] matching an initial feature screening threshold according to the noise intensity and the gray uniformity;

[0020] respectively adaptively adjusting the initial window size and the initial feature screening threshold for different types of noise regions according to the noise classification information.

[0021] In some preferred embodiments, the generating the initial blood vessel segmentation map containing interrupted blood vessels includes:

[0022] perform adaptive Hessian matrix feature extraction on the fundus image based on the dynamically determined processing parameter, to generate a blood vessel feature map;

[0023] perform binarization on the blood vessel feature map, to generate a binary blood vessel map;

[0024] perform type-specific elimination on false blood vessel targets corresponding to noise regions in the binary blood vessel map based on the noise classification information, to generate the initial blood vessel segmentation map.

[0025] In some preferred embodiments, the multi-dimension verification includes:

[0026] topology verification based on blood vessel orientation, diameter, or branching angle;

[0027] texture feature verification based on pixel gray scale, edge, or gradient features on the connection path;

[0028] special node-specific rule adaptation verification for blood vessel intersection points, branching points, and terminal capillary.

[0029] In some preferred embodiments, the determination of the destination point corresponding to the interruption point and the connection include:

[0030] for each candidate connection path connecting the interruption point and each candidate destination point, calculate a comprehensive weighted score thereof, which is obtained by weighted summation of path length of the candidate connection path and matching degree obtained in the topology verification and the texture feature verification;

[0031] determine the candidate connection path with the optimal comprehensive weighted score as an effective blood vessel connection path, and determine the endpoint of the effective blood vessel connection path as the destination point corresponding to the interruption point;

[0032] perform smooth fitting connection along the effective blood vessel connection path using a preset curve fitting algorithm.

[0033] In another aspect, the embodiments of the present application provide a device for segmenting retinal blood vessels in a fundus image, the device comprising:

[0034] an analysis module configured to analyze the acquired fundus image to determine at least one prior analysis indicator representing distribution characteristics of the image, and to identify noise regions in the image to generate noise classification information;

[0035] an image processing module configured to dynamically determine processing parameters based on the prior analysis indicator and the noise classification information, and further configured to process the fundus image using the processing parameters and perform noise elimination processing based on the noise classification information, to generate an initial blood vessel segmentation map containing interrupted blood vessels;

[0036] The intelligent reconstruction module is configured to identify an interruption point of the blood vessels in the initial blood vessel segmentation map and a candidate destination point on a blood vessel skeleton, and perform multi-dimensional verification between the interruption point and each candidate destination point to determine a destination point corresponding to the interruption point and connect the destination point to the interruption point to generate a reconstructed retinal blood vessel segmentation map.

[0037] Advantages of the present application:

[0038] 1) The present application dynamically matches the optimal processing parameters through image prior analysis and noise classification, so that the segmentation algorithm can adapt to different quality and different lesion types of fundus images, can effectively deal with fundus images of different quality, and improves the accuracy and robustness of segmentation;

[0039] 2) A layered anti-interference system is constructed, complex noise such as optic disc and hemorrhage points is identified by K-means clustering, and targeted inhibition and removal is performed at different stages, which significantly reduces the interference of noise on blood vessel segmentation and improves the purity and accuracy of the segmentation result;

[0040] 3) The multi-dimensional verification reconstruction algorithm replaces the simple shortest path connection, accurately judges the connection relationship of the interrupted blood vessels, effectively avoids the wrong connection in the complex areas such as blood vessel intersection and branch, ensures the physiological rationality and topological structure integrity of the reconstructed blood vessel network, and thus guarantees the reliability of the final segmentation result. BRIEF DESCRIPTION OF DRAWINGS

[0041] Other features, objects and advantages of the present application will become more apparent from the following detailed description of non-limiting embodiments made with reference to the accompanying drawings:

[0042] Figure 1 is a flowchart of a method for segmenting retinal blood vessels in a fundus image according to an embodiment of the present application;

[0043] Figure 2 is a logic diagram of adaptive parameter determination according to an embodiment of the present application;

[0044] Figure 3 is a schematic diagram of a blood vessel reconstruction process according to an embodiment of the present application, which clearly shows that the candidate destination point is a data stream, and gradually passes through the multi-dimensional verification process of topological structure verification, texture feature verification and special node rule adaptation, and the interruption point is a fixed reference information participating in the progressive screening logic of each level verification;

[0045] Figure 4 is a structural block diagram of a device for segmenting retinal blood vessels in a fundus image according to an embodiment of the present application;

[0046] Figure 5Fig. 1 is a structural schematic diagram of a computer system of a server for implementing the server of the method, system and electronic device embodiments of the present application. DETAILED DESCRIPTION

[0047] The present application will be further described below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are intended to explain the present application and are not intended to limit the present application. In addition, it should be noted that only the parts related to the present application are shown in the accompanying drawings for the convenience of description.

[0048] It should be noted that the embodiments and features in the embodiments of the present application can be combined with each other without conflict. The present application will be described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.

[0049] The present application proposes a retinal blood vessel segmentation method and device in fundus images, aiming to solve the problems of poor adaptability of existing segmentation algorithms to different image qualities, weak anti-interference ability to composite noise, and low blood vessel reconstruction accuracy. The present application extracts three core indexes of blood vessel density, noise intensity, and gray uniformity through image prior analysis, dynamically matches the calculation window size (W) of the Heaviside matrix and the feature screening threshold (T), solves the adaptability defects of fixed parameters to different images, and improves the reconstruction accuracy and anti-interference ability.

[0050] In order to more clearly describe the retinal blood vessel segmentation method in fundus images of the present application, the following will be described in conjunction with Figure 1 The method in the embodiments of the present application will be described in detail.

[0051] The retinal blood vessel segmentation method in fundus images of the first embodiment of the present application, as shown in Figure 1 The steps S100-S600 are described in detail as follows:

[0052] S100, acquiring a fundus image to be processed and performing preprocessing.

[0053] The original fundus image to be subjected to blood vessel segmentation is acquired. Since the contrast of the retinal blood vessels in the green channel is the highest in the fundus color image, the green channel image needs to be extracted first.

[0054] Preferably, in order to suppress noise and enhance the contrast between blood vessels and background, the green channel image is preprocessed. The preprocessing includes:

[0055] The mean filtering algorithm is used to filter the green channel image to reduce Gaussian noise and salt and pepper noise in the image;

[0056] The adaptive histogram equalization algorithm is used to enhance the contrast of the filtered green channel image, and a pre-processed fundus image is obtained; the local contrast of the image can be effectively enhanced, the edge of the fine blood vessels is clearer, and the problem of over-amplifying noise is avoided.

[0057] After pre-processing, an optimized fundus image I with a size of MxN (M is the number of rows and N is the number of columns) is obtained as the input basis for all subsequent steps.

[0058] S200, analyzing the fundus image to determine at least one priori analysis index representing the distribution characteristics of the image, and identifying a noise region in the image to generate noise classification information.

[0059] Preferably, in order to quantify the overall quality and content distribution of the image, the priori analysis index includes at least one of the following:

[0060] The blood vessel density (VD) obtained by statistical analysis based on a sliding window for the fundus image;

[0061] The noise intensity (NI) calculated by the ratio of the global gray variance and the average edge gradient of the fundus image;

[0062] The gray uniformity (GU) calculated by the skewness of the gray distribution of the fundus image.

[0063] Further preferably, in this embodiment, the VD value reflects the density of blood vessels, and the calculation method is specifically as follows:

[0064] By setting a plurality of non-overlapping sliding windows (for example, the window size is 50x50 pixels) on the image, the proportion of suspected blood vessel pixels in each window determined by preliminary threshold segmentation (such as Otsu method) is counted, and then the proportion of all windows is averaged to obtain the blood vessel density.

[0065] Specifically, in this embodiment, the blood vessel density (VD) is:

[0066] ;

[0067] Wherein, is the total number of non-overlapping windows divided by the image, is the number of suspected blood vessel pixels in the i-th window (determined by Otsu threshold segmentation, the pixel with a gray value lower than the background average is determined as a suspected blood vessel); is the total number of pixels in a single window; and The larger the value is, the more dense the blood vessels are; if VD=0, output "no blood vessel image" prompt.

[0068] Further preferably, in the present embodiment, the NI value synthetically evaluates the noise interference level of the image; the NI value is obtained by calculating the ratio of the mean of the edge gradient calculated by the operator and the global gray variance of the image. The NI value is obtained by calculating the ratio of the mean of the edge gradient calculated by the operator and the global gray variance of the image.

[0069] Specifically, in the present embodiment, the noise intensity (NI) is:

[0070] ;

[0071] wherein, is the global gray variance of the image, is the mean of the edge gradient calculated by the operator; (Avoiding the denominator being 0); the greater the NI value, the stronger the noise, and generally .

[0072] Further preferably, in the present embodiment, the GU value reflects the balance of the gray distribution of the image. The GU value is obtained by calculating the skewness of the gray distribution of the image and performing normalization processing.

[0073] Specifically, in the present embodiment, the gray uniformity (GU) is:

[0074] ;

[0075] ;

[0076] wherein, is the skewness of the gray distribution, describing the degree of asymmetry of the gray distribution; the present embodiment preferably (the maximum skewness based on the statistics of 5000+ clinical samples, covering 20 hospitals and 10 common types of fundus lesions); M is the number of rows, N is the number of columns, is the pixel gray value, is the global gray mean, is the global gray variance of the image; the smaller the absolute value of the GU value skewness, the closer to 1, indicating that the gray distribution is more symmetrical, more uniform, and the image quality is generally better. The more uniform it is; when , , it is determined that the gray distribution is extremely unbalanced.

[0077] In a preferred embodiment, in order to distinguish different types of noise and perform targeted processing, the present embodiment adopts a K-means clustering algorithm to identify the noise regions in the image to generate noise classification information, specifically: based on the gray value of the pixel, the circularity of the connected region where the pixel is located, and the area characteristics, the pixel is classified into one of the background / blood vessels, the optic disc highlight region, the hemorrhage point region, or the exudate region by using the clustering algorithm.​

[0078] For example, optic disc highlight region usually has the characteristics of high gray level and high circularity; bleeding point has the characteristics of medium-high gray level and higher circularity; and exudate shows medium gray level but irregular shape.

[0079] Further preferably, in a specific embodiment, identifying noise regions in the image to generate noise classification information comprises:

[0080] A feature vector is extracted for each pixel in the image, such as X = (gray level, circularity, area of the connected region it belongs to);

[0081] The number of clusters K = 3 (corresponding to three main noise types of optic disc highlight, bleeding point and exudate) is set, and an initial cluster center based on clinical experience is used to cluster all pixels;

[0082] After clustering, a noise classification map N is generated, in which each pixel is labeled as: 1 (optic disc highlight), 2 (bleeding point) or 3 (exudate); non-noise pixels (background / blood vessels) are uniformly labeled as 0.

[0083] S300, based on the prior analysis index and noise classification information, dynamically determining the processing parameters for extracting blood vessel segmentation features.

[0084] Preferably, the processing parameters at least include the window size for Hessian matrix calculation and the feature screening threshold for screening blood vessel features.

[0085] Further preferably, the dynamic determination of the processing parameters for extracting blood vessel segmentation refers to as shown in Figure 2 , specifically:

[0086] Based on the prior analysis index and noise classification information, the core blood vessel feature extraction algorithm, Hessian matrix analysis, dynamically matches the optimal processing parameters, including:

[0087] According to the blood vessel density and the noise intensity, an initial window size (W) is matched;

[0088] According to the noise intensity and the gray level uniformity, an initial feature screening threshold (T) is matched;

[0089] According to the noise classification information, the initial window size and the initial feature screening threshold of different types of noise regions are adaptively adjusted respectively to obtain the adjusted window size and the feature screening threshold.

[0090] Further preferably, in a specific embodiment, according to the hierarchical combination of blood vessel density (VD) and noise intensity (NI), the optimal initial window size W is queried and matched from a preset rule table.

[0091] For example, images with dense vessels (high VD value) and low noise (low NI value) match small windows (e.g. 3x3) to capture details, images with sparse vessels (low VD value) or high noise (high NI value) match large windows (e.g. 7x7) to enhance noise resistance. The window size is selected from 3x3, 5x5, 7x7, based on the retinal vessel width statistics (1-20 pixels), 3x3 window fits thin vessels, 7x7 window enhances noise resistance, larger window will cause vessel details loss; the feature screening threshold is selected from [0.3, 0.6], lower than 0.3 will introduce a lot of noise, higher than 0.6 will miss thin vessels, this range is verified by experiments to balance the detection rate and false detection rate.

[0092] Specifically, in the present embodiment, the preset rule table refers to Table 1 below.

[0093] Table 1

[0094] Further preferably, in a specific embodiment, the optimal initial feature screening threshold T is matched according to the classification combination of noise intensity (NI) and gray uniformity (GU). For example, images with low noise (low NI value) and good gray uniformity (high GU value) match a lower T to detect more vessels, and images with high noise and poor gray uniformity match a higher T to suppress noise.

[0095] Specifically, in the present embodiment, the preset rule table refers to Table 2 below.

[0096] Table 2

[0097] Further preferably, in a specific embodiment, on the basis of the initial matching parameters, local fine-tuning is performed according to the noise classification map N generated in step S200. For the region identified as the optic disc (N=1), W is further increased and T is increased; for the bleeding or exudation region (N=2 or 3), W and T are moderately increased, and the adjustment amplitude is less than that of the optic disc region. This is intended to further suppress noise response in specific noise regions, which has a higher priority than the global initial matching result.

[0098] Specifically, in an embodiment, based on statistical analysis of clinical samples, it is ensured that the parameter correction can quickly converge the vessel continuity and noise false detection rate to a preset range, and the local fine-tuning preferably includes the following contents:

[0099] Optic disc region (N=1): +2 based on the initial matching window size W (e.g. original 3x3 is adjusted to 5x5), +0.1 based on the feature threshold T;

[0100] Hemorrhage / Exudation Area (N=2 or 3): W base +1, T base +0.05; Normal Area (N=0): Keep the original matching parameters unchanged;

[0101] Constraint: The adjusted W should not exceed the range of {3, 5, 7}, and T should not exceed the range of [0.3, 0.6].

[0102] S400, using the dynamically determined processing parameters to process the fundus image, and performing noise removal processing based on the noise classification information to generate an initial blood vessel segmentation map containing interrupted blood vessels.

[0103] Preferably, the generating of the initial blood vessel segmentation map containing interrupted blood vessels comprises:

[0104] Based on the dynamically determined processing parameters, adaptive Hessian matrix feature extraction is performed on the fundus image to generate a blood vessel feature map;

[0105] The blood vessel feature map is binarized to generate a binary blood vessel map; wherein the local entropy calculation window size is consistent with the window size of the Hessian matrix calculation;

[0106] Based on the noise classification information, type-specific removal is performed on the pseudo-blood vessel targets in the binary blood vessel map corresponding to the noise area to generate the initial blood vessel segmentation map.

[0107] In a preferred embodiment, before the blood vessel feature Figure Two is binarized, feedback calibration is further included, specifically comprising:

[0108] The blood vessel continuity representing the proportion of blood vessel pixels in the effective connected region in the blood vessel feature map is calculated, and the noise false detection rate representing the proportion of isolated suspected blood vessel pixels is calculated;

[0109] When the blood vessel continuity is lower than the preset continuity lower limit, the feature screening threshold is reduced and adaptive Hessian matrix feature extraction is re-executed; when the noise false detection rate is higher than the preset false detection rate upper limit, the feature screening threshold is increased and adaptive Hessian matrix feature extraction is re-executed; until the preset quality requirement is met.

[0110] Wherein, the continuity lower limit and the false detection rate upper limit are determined based on statistical determination of clinical fundus images, covering 20 hospitals and 10 common fundus lesion types.

[0111] Further preferably, in a specific embodiment, reference is made to the fundus image as shown in Figure 2As shown, the fundus image is processed using dynamically determined processing parameters, and noise removal processing is performed based on the noise classification information to generate an initial blood vessel segmentation map containing interrupted blood vessels, specifically including the following:

[0112] S401, adaptive Hessian matrix feature extraction: based on the adjusted window size W and feature screening threshold T adapted to the pixel or region obtained in step S300, the preprocessed fundus image is calculated by Hessian matrix.

[0113] Specifically, in this embodiment, adaptive Hessian matrix feature extraction specifically includes:

[0114] 1) For each pixel, construct its Hessian matrix:

[0115] ;

[0116] Based on the statistical analysis of the blood vessel width of the fundus image, this embodiment preferably uses a Gaussian difference operator with a standard deviation of 1.0 to calculate the second-order partial derivative of the pixel in the window to construct the Hessian matrix ; wherein: , are the second-order partial derivatives in the x and y directions, is the mixed second-order partial derivative (approximated by a Gaussian difference operator).

[0117] 2) Solve the eigenvalues and ( ), and calculate the blood vessel feature response value ):

[0118] ;

[0119] Wherein, the response value , the greater the value, the value is closer to 1, indicating that the pixel ) is more likely to be a blood vessel.

[0120] 3) Use the dynamically determined threshold T to screen the response value, and retain the pixels with a response value greater than the dynamically determined feature screening threshold T, and set the rest to zero to generate a gray blood vessel feature map:

[0121] ;

[0122] Wherein, is the screened feature map (gray value , not a binary map, retaining feature response intensity information).

[0123] S402, feedback calibration:

[0124] To avoid the poor quality of feature maps caused by improper parameter matching, a feedback calibration mechanism is introduced. The feature map is calculated

[0125] VC=total number of vessel pixels in the effective connected area / total number of suspected vessel pixels

[0126] FNR=number of isolated suspected vessel pixels / total number of suspected vessel pixels

[0127] Wherein, the effective connected area refers to the connected area of pixels with an area of 1 pixel, which is used to exclude isolated noise. The difference between the noise area determination threshold is derived from the different area characteristics of blood vessels and noise.

[0128] If the VC is lower than the preset lower limit (based on clinical sample statistics, the embodiment preferably 0.6), it means that more blood vessels may be missed, so the T is appropriately reduced (such as ) and the feature extraction is performed again; if the FNR is higher than the preset upper limit (based on clinical sample statistics, the embodiment preferably 0.2), it means that too much noise is introduced, so the T is appropriately increased (such as ) and the extraction is performed again; this cycle continues until both indicators meet the preset requirements to ensure the quality of the output (hereinafter referred to as ).

[0129] S403, binarization is performed on the calibrated :

[0130] The calibrated blood vessel feature map is binarized to generate a binary blood vessel map . The embodiment preferably adopts a local entropy threshold segmentation algorithm, and the window size for calculating the local entropy is consistent with the window size W of the Heather matrix to ensure the coherence of the adaptive logic.

[0131] S404, noise removal: based on the part of the noise classification map N generated in step S200 which is identified as different types of noise area, the corresponding morphological removal rule is adopted to perform type-specific noise removal on the binary blood vessel map .

[0132] For example, in the area marked as optic disc, morphological opening operation is used to remove false vessels; in the area marked as hemorrhage point, isolated targets with high circularity and small area are removed; in the area marked as exudate, irregular areas are screened and removed. After this step, an initial blood vessel segmentation map (binary map, 1 for blood vessel pixels and 0 for background) is obtained which is relatively clean but may still have blood vessel fractures.

[0133] Specifically, in the embodiment, the type-specific elimination includes: for the optic disc highlight area, adopting a gray threshold and a morphological opening operation for elimination; and / or, for the bleeding point or exudate area, elimination based on circularity and area threshold.

[0134] The noise processing adopted in the application is a composite noise hierarchical suppression processing, including a preprocessing stage of adopting a clustering algorithm to analyze the gray value, circularity, and connected region area characteristics of pixels, distinguishing different types of noise such as optic disc highlight, bleeding point, and exudate, and outputting a noise classification map; a Heaviside matrix stage of adjusting Heaviside matrix parameters for different noise regions based on the noise classification map to suppress noise characteristics; and a post-processing stage of performing a type-specific noise elimination operation on the binary blood vessel map based on the geometric characteristics and morphological characteristics of different noises. Thus, targeted suppression and elimination are realized at different stages to ensure the accuracy of segmentation.

[0135] S500, identifying an interruption point of the blood vessel in the initial blood vessel segmentation map and a candidate destination point on the blood vessel skeleton.

[0136] Preferably, in order to repair the fracture in the initial segmentation map , it is necessary to locate the problem point, specifically including the following steps:

[0137] 1) extracting the blood vessel skeleton; performing connected region labeling on to obtain a plurality of connected regions , taking the largest connected region as the main blood vessel network (based on clinical data, the main blood vessel network of the retina is usually the largest connected region), performing thinning processing to extract the center line thereof to obtain the skeleton of the main blood vessel, and constructing a topological atlas G recording node (end point, intersection point), direction, diameter, and the like.

[0138] In the embodiment, the thinning processing adopts the Zhang-Suen algorithm, and the iteration termination condition is no pixel change for two consecutive iterations.

[0139] 2) identifying the interruption point P; for other small blood vessel segments other than the main stem , performing neighborhood traversal, if there is only one blood vessel pixel marked as an end point in the neighborhood of the pixel , calculating the distance from the end point to the center of the optic disc, and selecting the end point closest to the center of the optic disc as the interruption point P of the blood vessel segment. Specifically, in the embodiment, the optic disc is located through the gray peak value and the circularity, the gray peak value region is found through the gray histogram, the optic disc region is determined by combining the circularity screening (such as circularity

[0140] ), and the geometric center of the region is taken as the center of the optic disc. ​

[0141] 3) Selecting candidate destination points Q; in the vicinity of each breakpoint P (e.g. 50x50 pixels), all nodes in the vessel skeleton G are selected as candidate destination points , and sorted by distance to P, and the top n are selected as candidate destination points . In this embodiment, n is preferably 5, and if there are less than 5, all are kept.

[0142] Preferably, if there are no candidate destination points, the broken vessel is marked as unreconstructable, and a prompt is given in the final report.

[0143] S600, performing multi-dimensional verification between the breakpoint and each candidate destination point to determine the destination point corresponding to the breakpoint and connect them to generate a reconstructed retinal vessel segmentation map.

[0144] In a preferred embodiment, as shown in Figure 3 , the multi-dimensional verification, for each breakpoint and its corresponding set of candidate destination points , performs a three-level progressive verification:

[0145] First-level verification: topological structure verification based on vessel orientation, diameter or branching angle;

[0146] Second-level verification: texture feature verification based on pixel gray scale, edge or gradient features along the connection path;

[0147] Third-level verification: special node-specific rule adaptation verification for vessel intersection points, branching points and terminal capillary.

[0148] Further preferably, in a specific embodiment, the topological structure verification judges the homology in macroscopic structure between the vessel segment where P is located and the vessel segment where Q is located, specifically:

[0149] Orientation consistency: calculate the orientation vector of the vessel segment where P is located and the orientation vector of the vessel segment where Q is located , judge whether the included angle between them is within a reasonable range;

[0150] ;

[0151] In this embodiment, the threshold of orientation consistency is preferably , and all angles are vector included angles, and the calculation range is ;

[0152] Diameter matching degree: calculate the vessel diameter at the breakpoint P ​​a ratio of a diameter of a blood vessel at the candidate destination point to a diameter of a blood vessel at the candidate destination point

[0153]

[0154] wherein, in the embodiment, the threshold value of the diameter matching degree is preferably 0.8

[0155] branch angle rationality: if the candidate destination point is a branch point, the branch angle between the interrupted blood vessel and the main blood vessel is determined, and whether the branch angle formed after connection conforms to the physiological range is determined

[0156]

[0157] wherein, in the embodiment, the threshold value of the branch angle rationality is preferably 0.8

[0158] Further preferably, only when all the topological verification items pass, i.e., only when , , all meet the threshold value, the candidate point can enter the next level of verification. The above threshold values are determined based on clinical sample statistics and conform to the physiological structure characteristics of retinal blood vessels.

[0159] Further preferably, in a specific embodiment, the texture feature verification verifies whether the pixels on the potential connection path from P to conform to the micro-texture features of blood vessels, specifically:

[0160] a texture feature template is established based on the blood vessel segment where the P point is located, including gray scale distribution, edge smoothness, gradient direction consistency, etc.

[0161] a preliminary path from P to is searched, and the termination condition is to reach the Q point, and the path can pass through non-blood vessel areas; and the matching degree of the pixels on the path with the texture template is calculated, and only when the matching degree is higher than a preset threshold value, the candidate point passes this level of verification;

[0162] wherein, the matching degree includes the fitting degree of the gray value distribution of the pixels on the candidate path with the gray scale distribution model of the preset blood vessel template; the similarity of the edge smoothness of the candidate path region with the edge smoothness of the preset blood vessel template; and the consistency of the gradient direction of the pixels on the candidate path with the preset blood vessel direction vector.

[0163] ​​​​​​​​Specifically, in the embodiment, the texture feature template comprises:

[0164] Gray distribution curve: for K pixels in the blood vessel segment where P is located, the gray value is counted , and a normal distribution curve is fitted ( is the mean value, and is the standard deviation);

[0165] Edge smoothness: the standard deviation of the edge gradient of the blood vessel segment is calculated ;

[0166] Gradient direction consistency: the proportion of pixels whose gradient direction is to the angle between the blood vessel segment and the blood vessel direction vector is counted .

[0167] Specifically, the matching degree of the pixels on the path and the texture template is calculated , specifically:

[0168] Search for the initial path L of P to Q (the search termination condition is to reach the point Q, and the path can pass through non-blood vessel areas), containing M pixels;

[0169] Gray matching: the fitting degree of the gray value of each pixel in L and is calculated , and the number of pixels is counted ;

[0170] Edge matching: the standard deviation of the edge gradient of the L region is calculated , and if , it is marked as edge matching;

[0171] Gradient direction matching: the proportion of pixels whose gradient direction in L is to the angle between the blood vessel segment and the blood vessel direction vector is counted , and if , it is marked as gradient direction matching;

[0172] The matching degree is determined as:

[0173] ;

[0174] Wherein, when passes the verification (the threshold is determined by experiment optimization, which can balance the matching accuracy and calculation efficiency), otherwise Q is removed.

[0175] Further preferably, in a specific embodiment, special node-specific rule adaptation verification is performed for vascular intersection, branch point, and terminal capillary; that is, different connection judgment rules and / or verification thresholds are used for different types of special nodes. The judgment of all special nodes is based on 5x5 pixel neighborhood analysis.

[0176] Specifically, in the present embodiment, the special nodes include:

[0177] There are different directions of the vessel segments in the neighborhood, the angle calculation and the topology verification are consistent, and are both based on the vessel direction vector, and the angle of each segment is marked as an intersection point;

[0178] There is 1 main vessel (diameter ) and 1-2 capillaries (diameter ) in the neighborhood, which is marked as a branch point;

[0179] The diameter of the vessel at the breakpoint P is pixels, which is marked as a terminal capillary.

[0180] Specifically, in the present embodiment, the special rule adaptation includes:

[0181] Intersection point rule: only select nodes in the radiation direction consistent with the direction of P as Q, and prohibit connection across vessels with different directions;

[0182] Branch point rule: only allow "capillary to main vessel" connection, that is, "capillary to main vessel" connection, and prohibit reverse or same-level connection;

[0183] Terminal capillary rule: appropriately relax the texture verification threshold, and narrow the search range, for example, the texture verification threshold is reduced to 0.7, and the search range is reduced to 30x30 pixels.

[0184] In a preferred embodiment, the determination of the destination point corresponding to the breakpoint and the connection thereof comprises:

[0185] For each candidate connection path connecting the breakpoint and each candidate destination point, a comprehensive weighted score is calculated, which is obtained by weighted summation according to the path length of the candidate connection path and the matching degree obtained in the topology structure verification and the texture feature verification;

[0186] The candidate connection path with the optimal comprehensive weighted score is determined as the effective vessel connection path, and the endpoint of the effective vessel connection path is determined as the destination point corresponding to the breakpoint;

[0187] adopting a preset curve fitting algorithm to perform smooth fitting connection along the effective blood vessel connection path.

[0188] Further preferably, in a specific embodiment, determining the destination point corresponding to the breakpoint and connecting, specifically includes:

[0189] 1) For each candidate connection path connecting the breakpoint and each candidate destination point, determining a path score item, including:

[0190] ;

[0191] ;

[0192] Wherein, is the path length score, is the path length, is the maximum possible path length in the search range, preferably 50 in the embodiment, is the minimum possible path length, preferably 10 in the embodiment; is the topological matching degree, based on the homologous blood vessel direction angle in clinical statistics, preferably the optimal direction angle is .

[0193] 2) Calculate the comprehensive weighted score of each path, and determine the candidate connection path with the optimal comprehensive weighted score as the effective blood vessel connection path:

[0194] ;

[0195] ;

[0196] Wherein, is the comprehensive weighted score of the candidate connection path, , , is the corresponding item weight; through orthogonal experiment optimization, preferably =0.3, , ; take the candidate connection path with the maximum value as the effective blood vessel connection path.

[0197] After three levels of verification, there may still be multiple candidate points Q qualified, and the most suitable candidate point is determined as the destination point through the comprehensive weighted score.

[0198] 3) Adopting a preset curve fitting algorithm to perform smooth fitting connection along the effective blood vessel connection path; in the present embodiment, a second order Bezier curve fitting is adopted to ensure morphological continuity:

[0199] ;

[0200] wherein, the control point M is the intersection of the tangent line extension of the vessel segment where the break point P locates and the tangent line extension of the vessel segment where the destination point Q locates; if the two tangent lines have no intersection, the midpoint of the line segment PQ is taken and is offset by 5 pixels along the weighted average direction of the two vessel direction vectors (the weight is proportional to the vessel diameter), ensuring that the curvature difference of the curvature of the connected segment and the original vessel curvature is less than 0.1 .

[0201] Preferably, after the connection is completed, verification is performed, specifically:

[0202] if the connected segment is in normal communication with the original vessel and has no distortion (the curvature change ), the connected segment is retained; otherwise, the search is restarted; after the connection of all break points is completed, the final complete and continuous reconstructed retinal vessel segmentation map is output .

[0203] In a preferred embodiment, the following steps are further included:

[0204] On the basis of the reconstructed retinal vessel segmentation map, the vessel segments are distinguished as arteries or veins, and an arterial-venous marker is generated;

[0205] at least one vascular morphological parameter is quantified, the vascular morphological parameter including a vessel tortuosity, a branch angle, or an arteriovenous diameter ratio, wherein the arteriovenous diameter ratio is calculated according to a ratio of an average diameter of an artery to an average diameter of a vein measured in a preset region;

[0206] a quantitative analysis result including the arterial-venous marker and the morphological parameter is generated.

[0207] Specifically, in the embodiment, the vessel segments are distinguished as arteries or veins (A / V), specifically: the reconstructed vessel tree is divided into independent vessel segments, the gray mean value, the width, the branch density and other features of each segment are extracted, and a pre-trained support vector machine (SVM) classifier is input to automatically label the arteries and veins; the branch density is the number of branch points per unit area (unit: ).

[0208] The pre-trained support vector machine (SVM) classifier described in the embodiment, the training data of the SVM classifier is 1000 fundus images labeled with arteries and veins, the proportion of artery and vein samples is 1:1, the division ratio of the training set and the test set is 7:3; the classification features are determined by mutual information entropy screening to ensure that the correlation of the features and the arterial-venous difference is optimal.

[0209] Further preferably, the quantitative analysis result further contains an abnormal index prompt (such as AVR<0.7).

[0210] The device for segmenting retinal blood vessels in fundus images of the second embodiment of the present application is used to execute the method described in the first embodiment, and refers to FIG. 1, and comprises: Figure 4

[0211] An analysis module is configured to analyze the acquired fundus images to determine at least one prior analysis index representing the distribution characteristics of the images and to identify noise regions in the images to generate noise classification information;

[0212] An image processing module is configured to dynamically determine processing parameters based on the prior analysis index and the noise classification information, and is further configured to process the fundus images using the processing parameters and to perform noise removal processing based on the noise classification information to generate an initial blood vessel segmentation map containing interrupted blood vessels;

[0213] An intelligent reconstruction module is configured to identify the interruption points of the blood vessels in the initial blood vessel segmentation map and candidate destination points on the blood vessel skeleton, and to perform multi-dimensional verification between the interruption points and each of the candidate destination points to determine the destination points corresponding to the interruption points and to connect them to generate a reconstructed retinal blood vessel segmentation map.

[0214] It should be noted that the device for segmenting retinal blood vessels in fundus images provided in the above embodiments is only exemplified by the division of the above functional modules, and in actual application, the above functions can be completed by different functional modules according to needs, i.e., the modules or steps in the embodiments of the present application are further decomposed or combined, for example, the modules of the above embodiments can be combined into one module, or can be further split into multiple sub-modules to complete all or part of the functions described above. The names of the modules and steps involved in the embodiments of the present application are only for distinguishing the modules and steps, and should not be considered as an improper limitation of the present application.

[0215] The device of the third embodiment of the present application comprises:

[0216] At least one processor;

[0217] and a memory in communication connection with the at least one processor;

[0218] Wherein the memory stores instructions executable by the processor, and the instructions are used to be executed by the processor to realize the segmentation method of the retinal blood vessels in the fundus images described above.

[0219] The computer readable storage medium of the fourth embodiment of the present application stores computer instructions, and the computer instructions are used to be executed by the computer to realize the segmentation method of the retinal blood vessels in the fundus images described above.

[0220] ​Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the storage device and the processing device described above and the related descriptions can refer to the corresponding processes in the foregoing method embodiments, and will not be described here again.

[0221] Reference will be made to the following Figure 5 , which shows a structural schematic diagram of a computer system of a server for implementing the system, method and electronic device embodiments of the present application. Figure 5 The server shown is merely an example and should not bring any limitation to the functions and use range of the embodiments of the present application.

[0222] As Figure 5 shown, the computer system includes a central processing unit (CPU) 501 which can perform various appropriate actions and processes according to programs stored in a read only memory (ROM) 502 or loaded from a storage portion 508 into a random access memory (RAM) 503. In the RAM 503, various programs and data required for system operation are also stored. The CPU 501, the ROM 502 and the RAM 503 are connected to each other through a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0223] The following components are connected to the I / O interface 505: an input portion 506 including a keyboard, a mouse, etc.; an output portion 507 including a display such as a cathode ray tube (CRT), a liquid crystal display (LCD), etc., and a speaker, etc.; a storage portion 508 including a hard disk, etc.; and a communication portion 509 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication portion 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as necessary. A removable media 511 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc. is mounted on the drive 510 as necessary, so that a computer program read therefrom is installed into the storage portion 508 as necessary.

[0224] In particular, the processes described above with reference to the flowcharts can be implemented as a computer software program according to embodiments of the present disclosure. For example, embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer readable medium, the computer program comprising program code for performing the methods illustrated by the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via the communication section 509, and / or installed from the removable medium 511. When the computer program is executed by the central processing unit (CPU) 501, the above-described functions defined in the methods of the present application are performed. It should be noted that the computer readable medium of the present application can be a computer readable signal medium or a computer readable storage medium or any combination of the two. The computer readable storage medium may, for example, be, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus or device, or any suitable combination of the above. More specific examples of the computer readable storage medium can include, but are not limited to, an electrical connection having one or more wires, a portable computer diskette, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or Flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, the computer readable storage medium can be any tangible medium that contains or stores a program that can be used by or in connection with an instruction execution system, apparatus or device. In the present application, the computer readable signal medium can include a data signal carried in a baseband or as part of a carrier wave, in which the computer readable program code is carried. Such a propagated data signal can take any of a variety of forms, including but not limited to electro-magnetic, optical, or any suitable combination thereof. The computer readable signal medium can also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate or transport a program for use by or in connection with an instruction execution system, apparatus or device. Program code contained on the computer readable medium can be transmitted by any suitable medium, including but not limited to wireless, wire line, optical fiber cable, RF, etc., or any suitable combination of the above.

[0225] Computer program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, Smalltalk, C++ or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computer (for example, through the Internet using an Internet Service Provider).

[0226] The computer program instructions can also be loaded onto a computer or other programmable information processing apparatus to cause a series of operations to be performed on the computer or other programmable information processing apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable information processing apparatus implement the functions / acts specified in the flowchart and / or block diagram block or blocks.

[0227] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. For example, singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0228] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the disclosure. For example, singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise. It will be further understood that the terms "comprises" and / or "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and / or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0229] The technical scheme of the present application has been described in combination with the preferred embodiments shown in the drawings, but it is easy for those skilled in the art to understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical schemes after the changes or replacements will all fall within the protection scope of the present application.

Claims

1. A method of segmenting retinal blood vessels in a fundus image, characterized by, The method comprises the following steps: obtaining an eye fundus image to be processed; analyzing the eye fundus image to determine at least one priori analysis index representing the distribution characteristics of the image and identify noise regions in the image to generate noise classification information; dynamically determining processing parameters for extracting blood vessel segmentation features based on the priori analysis index and the noise classification information; wherein the priori analysis index comprises at least one of the following: blood vessel density obtained by sliding window statistics based on the eye fundus image; noise intensity calculated by the ratio of the global gray variance and the average edge gradient of the eye fundus image; gray uniformity calculated by the skewness of the gray distribution of the eye fundus image; the processing parameters at least include the window size for Hessian matrix calculation and the feature screening threshold for screening blood vessel features; the dynamic determination of the processing parameters for extracting blood vessel segmentation is specifically: matching the initial window size according to the blood vessel density and the noise intensity; matching the initial feature screening threshold according to the noise intensity and the gray uniformity; according to the noise classification information, the initial window size and the initial feature screening threshold of different types of noise regions are respectively adaptively adjusted; processing the eye fundus image by using the dynamically determined processing parameters, and performing noise removal processing based on the noise classification information to generate an initial blood vessel segmentation image containing interrupted blood vessels; identifying the breakpoints of the blood vessels in the initial blood vessel segmentation image and the candidate destination points on the blood vessel skeleton; multi-dimensional verification is performed between the breakpoints and each of the candidate destination points to determine the destination point corresponding to the breakpoint and connect them to generate a reconstructed retinal blood vessel segmentation image.

2. The method of claim 1, wherein, It also includes preprocessing of the eye fundus image to be processed, specifically including: obtaining a green channel image of the eye fundus image; using a mean filter algorithm to filter the green channel image; using an adaptive histogram equalization algorithm to enhance the contrast of the filtered green channel image to obtain a preprocessed eye fundus image.

3. The method of claim 1, wherein the step of segmenting the retinal blood vessels in the fundus image is performed by using a convolutional neural network. The method for generating an initial blood vessel segmentation image containing interrupted blood vessels comprises: performing adaptive Hessian matrix feature extraction on the eye fundus image based on the dynamically determined processing parameters to generate a blood vessel feature map; performing binaryzation on the blood vessel feature map to generate a binary blood vessel map; based on the noise classification information, performing type-specific removal on the pseudo-blood vessel targets corresponding to the noise regions in the binary blood vessel map to generate the initial blood vessel segmentation image.

4. The method of claim 3, wherein the step of segmenting the retinal blood vessels in the fundus image is performed by using a neural network. Before binaryzation of the blood vessel feature map, feedback calibration is also included, specifically including: calculating the blood vessel continuity representing the proportion of blood vessel pixels in the effective connected region in the blood vessel feature map, and the noise false detection rate representing the proportion of isolated suspected blood vessel pixels; when the blood vessel continuity is lower than the preset lower limit of continuity, the feature screening threshold is reduced and adaptive Hessian matrix feature extraction is re-executed; when the noise false detection rate is higher than the preset upper limit of false detection rate, the feature screening threshold is increased and adaptive Hessian matrix feature extraction is re-executed; until the preset quality requirement is met.

5. The method of claim 1, wherein, The multi-dimensional verification includes: topological structure verification based on the direction, diameter or branch angle of the blood vessels; The texture feature verification is based on pixel gray scale, edge or gradient features on the connection path; Special node-specific rule adaptation verification is performed on the special nodes such as vascular intersection points, branch points and terminal capillary.

6. The method of claim 5, wherein the step of segmenting the retinal blood vessels in the fundus image is performed by using a convolutional neural network. The determination of the destination point corresponding to the breakpoint and the connection thereof comprise: For each candidate connection path connecting the breakpoint and each candidate destination point, a comprehensive weighted score is calculated, which is obtained by weighted summation according to the path length of the candidate connection path and the matching degree obtained in the topology verification and the texture feature verification; The candidate connection path with the optimal comprehensive weighted score is determined as the effective vascular connection path, and the endpoint of the effective vascular connection path is determined as the destination point corresponding to the breakpoint; A preset curve fitting algorithm is used to perform smooth fitting connection along the effective vascular connection path.

7. The method of claim 1, wherein, The method further comprises: Based on the reconstructed retinal vascular segmentation map, arteries or veins are distinguished, and an arterial-venous marker is generated; At least one vascular morphological parameter is quantified, including vascular tortuosity, branch angle or arteriovenous diameter ratio, wherein the arteriovenous diameter ratio is calculated according to the ratio of the average diameter of the artery to the average diameter of the vein measured in a preset area; An analysis result containing the arterial-venous marker and the quantified morphological parameter is generated.

8. An apparatus for segmenting retinal blood vessels in a fundus image, characterized by comprising: It comprises: An analysis module configured to analyze the acquired fundus image to determine at least one priori analysis indicator representing the distribution characteristics of the image and to identify noise regions in the image to generate noise classification information; An image processing module configured to dynamically determine processing parameters based on the priori analysis indicator and the noise classification information; and further configured to process the fundus image using the processing parameters and to perform noise removal processing based on the noise classification information to generate an initial vascular segmentation map containing interrupted blood vessels; An intelligent reconstruction module configured to identify the breakpoint of the blood vessels in the initial vascular segmentation map and the candidate destination points on the vascular skeleton; and to perform multi-dimensional verification between the breakpoint and each candidate destination point to determine the destination point corresponding to the breakpoint and to connect the same, thereby generating a reconstructed retinal vascular segmentation map. The priori analysis indicator comprises at least one of the following: The vascular density obtained by sliding window statistics based on the fundus image; the noise intensity calculated by the ratio of the global gray scale variance to the average edge gradient of the fundus image; the gray scale uniformity calculated by the skewness of the gray scale distribution of the fundus image; The processing parameters at least include the window size for Heaviside matrix calculation and the feature screening threshold for screening blood vessel features; the dynamically determined processing parameters for extracting the vascular segmentation are as follows: The initial window size is matched according to the vascular density and the noise intensity; The initial feature screening threshold is matched according to the noise intensity and the gray scale uniformity; According to the noise classification information, the initial window size and the initial feature screening threshold of different types of noise regions are adaptively adjusted respectively.

Images