Macular localization system based on retinal structure

Through the macular area positioning system based on retinal structure, the macular area positioning problem in the existing technology is solved by using the optic disc and vascular segmentation module combined with the main vascular arch detection, automatic positioning is achieved, accuracy and reliability are improved, and manual labeling is saved time.

CN112581439BActive Publication Date: 2024-11-12ZHONGKE TAIMING (NANJING) TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202011447213.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-12-11
Publication Date
2024-11-12
Estimated Expiration
2040-12-11

AI Technical Summary

Technical Problem

The prior art is difficult to provide accurate data labels for training neural networks, affecting the localization of macular areas, especially because the lack of clear boundaries of macular areas makes it difficult for doctors to provide accurate labels.

Method used

Through the macular area positioning system based on the retinal structure, the optic disc vascular segmentation module and the macular area of ​​interest determination module are used, and the macular area location is calculated by combining the main vascular arch detection and linear combination to achieve automatic macular area positioning. The system includes modules for visual disk parameter calculation, macular area of ​​interest division, main vascular arch determination and macular area position calculation, and uses a convolutional neural network structured by U-Net for training and segmentation.

Benefits of technology

Automatic macular area positioning is realized, saving time for manual labeling, improving the accuracy and reliability of positioning, and meeting doctors' needs for retinal macular area.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN112581439B_ABST
    Figure CN112581439B_ABST
Patent Text Reader

Abstract

The present invention provides a macular area positioning system based on retinal structure, and relates to the technical field of macular positioning. The present invention calculates the parameters of the optic disc and the parameters of the blood vessels based on the retinal blood vessels and optic disc in the fundus image; determines the main vascular arch structure of the retinal result based on the parameters of the blood vessels and optic disc, and obtains the position of the macular area according to its structure. Automatic macular area positioning based on retinal structure is realized, saving the time loss of manual annotation for each image, and automatic macular area positioning based on retinal structure is realized, saving the time loss of manual annotation for each image. Tests show that the accuracy, reliability and robustness of the present invention meet the needs of human doctors for positioning the retinal macular area.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of macular positioning, and in particular to a macular area positioning system based on retinal structure. Background Art

[0002] The macula is one of the most sensitive and dangerous areas of the retina. Damage to the macula often has a direct impact on vision. The positioning of the macula is particularly important in retinal images.

[0003] The existing method for locating the macula area of ​​the fundus requires manual annotation of some feature points in each image, and fitting the possible macula area based on the feature points.

[0004] However, the lack of clear boundaries of the macula makes it difficult for doctors to provide accurate data labels for training neural networks, which affects the positioning of the macula. Summary of the invention

[0005] 1. Technical issues to be resolved

[0006] In view of the deficiencies of the prior art, the present invention provides a macular area positioning system based on the retinal structure, which solves the problem that the existing methods are difficult to provide accurate data labels for training neural networks, thus affecting the positioning of the macula.

[0007] (II) Technical solution

[0008] To achieve the above objectives, the present invention is implemented through the following technical solutions:

[0009] A macular area positioning system based on retinal structure, the system comprising:

[0010] An optic disc blood vessel segmentation module is used to obtain a binary optic disc image and a blood vessel image by segmenting the retinal fundus image;

[0011] A macular region of interest determination module, used to determine the macular region of interest based on the optic disc image;

[0012] A main vascular arch determination module is used to perform connected domain detection and average vascular width calculation in the macular region of interest based on the vascular image; determine the preferred main vascular arch based on the number of pixels in the connected domain, and obtain the candidate main vascular arch through a linear combination of the number of pixels in the connected domain and the average vascular width;

[0013] The macular area position calculation module is used to locate the macular area based on the two main vascular arches, the long axis and the radius.

[0014] Furthermore, the macular region of interest determination module includes:

[0015] An optic disc parameter calculation unit is used to perform ellipse fitting on the optic disc in the optic disc image to obtain the center and radius of the optic disc, and the major axis and minor axis of the ellipse;

[0016] A macular region of interest calculation unit, used to determine the macular region of interest on the blood vessel image based on the optic disc radius, and to dilate the optic disc in the optic disc image;

[0017] The quadrant division unit is used to divide the macular area of ​​interest into four quadrants based on the major axis and the minor axis of the ellipse.

[0018] Furthermore, the expansion multiple of the optic disc is 1.2 times, and a square centered on the optic disc is made with a side length of 16 times the radius of the optic disc as the region of interest.

[0019] Furthermore, the main vascular arch determination module includes:

[0020] The preferred main vascular arch calculation unit is used to make a Venn difference between the vascular image and the expanded optic disc image, and then perform 8-connected domain calculation, retaining the area with a connected domain pixel number of not less than 30, and obtain a connected domain calculation graph, from which the connected domain with the largest number of pixels is found as the preferred main vascular arch;

[0021] The blood vessel priority ranking unit is used to take the blood vessel widths at the optic disc radius positions of multiple preset multiples in the connected domain calculation map with the optic disc center as the center, and perform linear combination with the corresponding first preset proportional coefficient to obtain the average width of each blood vessel in the connected domain calculation map; and perform linear combination of the number of pixels in the connected domain and the average blood vessel width based on the second preset proportional coefficient, and rank the blood vessels according to the result.

[0022] The alternative main vascular arch calculation unit is used to determine the quadrant where the preferred main vascular arch is located; it is also used to determine whether each blood vessel and the preferred main vascular arch are in the same quadrant on the left and right but in different quadrants on the upper and lower sides based on the primary and secondary order of the blood vessels. If so, the alternative main vascular arch is obtained; otherwise, the next blood vessel is judged.

[0023] Furthermore, the blood vessel width is the distance from the outer edge of the blood vessel to the centerline of the blood vessel; the preset multiples are 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9 and 2.0, and the corresponding first preset proportional coefficients are 0.3, 0.3, 0.1, 0.06, 0.06, 0.06, 0.06, 0.06; the second preset proportional coefficients corresponding to the number of pixels in the connected domain and the average width of the blood vessel are 0.3 and 0.7.

[0024] Furthermore, the macular area position calculation module includes:

[0025] The macular positioning unit is used to make a straight line where the long axis is located based on the vascular image after the two main vascular arches are located, and to make a line parallel to the straight line where the long axis is located at a position 5 times the optic disc radius on one side of the two main vascular arches, and form a closed area with the two main vascular arches; then a circle is fitted to the closed area, and the center of the circle is used as the center of the macular area, and the circle fitting result is used as the positioning of the macular area.

[0026] Furthermore, the macular area position calculation module also includes:

[0027] The closed area generating unit is used to draw a perpendicular line from the nearest point of the main vascular arch to the straight line where the long axis is located to the straight line where the long axis is located when the straight line where the long axis is located, the parallel line to the straight line where the long axis is located and the two main vascular arches cannot form a closed area; and then form a closed area by the straight line where the long axis is located, the parallel line to the straight line where the long axis is located, the perpendicular line and the two main vascular arches.

[0028] Furthermore, the optic disc blood vessel segmentation module includes:

[0029] An optic disc segmentation unit, used for taking the retinal fundus image as an input of the trained optic disc segmentation network model to obtain an optic disc image;

[0030] The blood vessel segmentation unit is used to use the retinal fundus image as the input of the trained blood vessel segmentation network model to obtain a blood vessel image.

[0031] Furthermore, the optic disc segmentation network model and the blood vessel segmentation network model are convolutional neural networks with a U-Net structure.

[0032] Furthermore, the system also includes:

[0033] The segmentation network training module is used to train the optic disc segmentation network model and the blood vessel segmentation network model, and the training method of the optic disc segmentation network model and the blood vessel segmentation network model includes:

[0034] Use CrossEntropy Loss (SoftMax Loss) as the loss function;

[0035] Use mini-batch gradient descent as the learning algorithm for convolutional neural networks;

[0036] Each convolutional neural network is trained for 250 epochs. During the training, the learning rate is reduced according to the set decay strategy. The learning rate gradually decreases with the deepening of training. The learning rate decays once every epoch. The initial learning rate is set to 0.005, the decay rate is set to 0.986, and the final learning rate is fixed to 0.0001.

[0037] L2 WeightDecay regularization is used for each parameter in each convolutional neural network.

[0038] (III) Beneficial effects

[0039] The present invention provides a macular area positioning system based on retinal structure. Compared with the prior art, it has the following beneficial effects:

[0040] The present invention calculates the parameters of the optic disc and the parameters of the blood vessels based on the retinal blood vessels and optic disc in the fundus image; based on the parameters of the blood vessels and optic disc, the main vascular arch structure of the retinal result is determined, and the position of the macular area is obtained according to its structure. Automatic macular area positioning based on the retinal structure is realized, saving the time loss of manual annotation for each image, and the automatic macular area positioning based on the retinal structure is realized, saving the time loss of manual annotation for each image. The test shows that the accuracy, reliability and robustness of the present invention meet the needs of human doctors for positioning the retinal macular area. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0042] Figure 1 is a system block diagram of an embodiment of the present invention;

[0043] Figure 2 This is a flow chart of the blood vessel and optic disc segmentation network training according to an embodiment of the present invention;

[0044] Figure 3 It is a schematic diagram of locating the macular area by the macular area position calculation module according to an embodiment of the present invention. DETAILED DESCRIPTION

[0045] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention are clearly and completely described. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0046] The embodiment of the present application provides a macular area positioning system based on retinal structure, which solves the problem that existing methods are difficult to provide accurate data labels for training neural networks, affecting the positioning of the macula, and realizes automatic retinal structure macular area positioning.

[0047] The technical solution in the embodiment of the present application is to solve the above technical problems. The overall idea is as follows: through deep learning, the blood vessels and optic disc of the image are first segmented, and then the geometric parameter features of the segmentation results are extracted, the segmentation results are regionally located, and the parameters of the network are continuously adjusted during the training process. Tests have shown that the accuracy, reliability, and robustness of the present invention meet the doctor's needs for retinal macular area positioning. The retinal macular area positioning method based on deep learning and retinal structure described in the present invention can be used to diagnose macular lesions for people of all ages, and can be used in hospital clinics, doctors' rapid diagnosis and other fields.

[0048] In order to better understand the above technical solution, the above technical solution will be described in detail below in conjunction with the accompanying drawings and specific implementation methods.

[0049] Embodiment 1:

[0050] like Figure 1 As shown, the present invention provides a macular area positioning system based on retinal structure, the system comprising:

[0051] An optic disc blood vessel segmentation module is used to obtain a binary optic disc image and a blood vessel image by segmenting the retinal fundus image;

[0052] A macular region of interest determination module, used to determine the macular region of interest based on the optic disc image;

[0053] A main vascular arch determination module is used to perform connected domain detection and average vascular width calculation in the macular region of interest based on the vascular image; determine the preferred main vascular arch based on the number of pixels in the connected domain, and obtain the candidate main vascular arch through a linear combination of the number of pixels in the connected domain and the average vascular width;

[0054] The macular area position calculation module is used to locate the macular area based on the two main vascular arches, the long axis and the radius.

[0055] Furthermore, the macular region of interest determination module includes:

[0056] An optic disc parameter calculation unit is used to perform ellipse fitting on the optic disc in the optic disc image to obtain the center and radius of the optic disc, and the major axis and minor axis of the ellipse;

[0057] A macular region of interest calculation unit, used to determine the macular region of interest on the blood vessel image based on the optic disc radius, and to dilate the optic disc in the optic disc image;

[0058] The quadrant division unit is used to divide the macular area of ​​interest into four quadrants based on the major axis and the minor axis of the ellipse.

[0059] Furthermore, the expansion multiple of the optic disc is 1.2 times, and a square centered on the optic disc is made with a side length of 16 times the radius of the optic disc as the region of interest.

[0060] Furthermore, the main vascular arch determination module includes:

[0061] The preferred main vascular arch calculation unit is used to make a Venn difference between the vascular image and the expanded optic disc image, and then perform an 8-connected domain calculation, retaining the area with a connected domain pixel number of not less than 30, to obtain a connected domain calculation graph, from which the connected domain with the largest number of connected domain pixels is found as the preferred main vascular arch; the connected domain calculation graph is an area that only contains blood vessels but not the optic disc.

[0062] The blood vessel priority ranking unit is used to take the blood vessel widths at the optic disc radius positions of multiple preset multiples in the connected domain calculation map with the optic disc center as the center, and perform linear combination with the corresponding first preset proportional coefficient to obtain the average width of each blood vessel in the connected domain calculation map; and perform linear combination of the number of pixels in the connected domain and the average blood vessel width based on the second preset proportional coefficient, and rank the blood vessels according to the result.

[0063] The alternative main vascular arch calculation unit is used to determine the quadrant where the preferred main vascular arch is located; it is also used to determine whether each blood vessel and the preferred main vascular arch are in the same quadrant on the left and right but in different quadrants on the upper and lower sides based on the primary and secondary order of the blood vessels. If so, the alternative main vascular arch is obtained; otherwise, the next blood vessel is judged.

[0064] Furthermore, the blood vessel width is the distance from the outer edge of the blood vessel to the centerline of the blood vessel; the preset multiples are 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9 and 2.0, and the corresponding first preset proportional coefficients are 0.3, 0.3, 0.1, 0.06, 0.06, 0.06, 0.06, 0.06; the second preset proportional coefficients corresponding to the number of pixels in the connected domain and the average width of the blood vessel are 0.3 and 0.7.

[0065] Furthermore, the macular area position calculation module includes:

[0066] The macular positioning unit is used to make a straight line where the long axis is located based on the vascular image after the two main vascular arches are located, and to make a line parallel to the straight line where the long axis is located at a position 5 times the optic disc radius on one side of the two main vascular arches, and form a closed area with the two main vascular arches; then a circle is fitted to the closed area, and the center of the circle is used as the center of the macular area, and the circle fitting result is used as the positioning of the macular area.

[0067] Furthermore, the macular area position calculation module also includes:

[0068] The closed area generating unit is used to draw a perpendicular line from the nearest point of the main vascular arch to the straight line where the long axis is located to the straight line where the long axis is located when the straight line where the long axis is located, the parallel line to the straight line where the long axis is located and the two main vascular arches cannot form a closed area; and then form a closed area by the straight line where the long axis is located, the parallel line to the straight line where the long axis is located, the perpendicular line and the two main vascular arches.

[0069] Furthermore, the optic disc blood vessel segmentation module includes:

[0070] An optic disc segmentation unit, used for taking the retinal fundus image as an input of the trained optic disc segmentation network model to obtain an optic disc image;

[0071] The blood vessel segmentation unit is used to use the retinal fundus image as the input of the trained blood vessel segmentation network model to obtain a blood vessel image.

[0072] Furthermore, the optic disc segmentation network model and the blood vessel segmentation network model are convolutional neural networks with a U-Net structure.

[0073] Furthermore, the system also includes:

[0074] The segmentation network training module is used to train the optic disc segmentation network model and the blood vessel segmentation network model, and the training method of the optic disc segmentation network model and the blood vessel segmentation network model includes:

[0075] Use CrossEntropy Loss (SoftMax Loss) as the loss function;

[0076] Use mini-batch gradient descent as the learning algorithm for convolutional neural networks;

[0077] Each convolutional neural network is trained for 250 epochs. During the training, the learning rate is reduced according to the set decay strategy. The learning rate gradually decreases with the deepening of training. The learning rate decays once every epoch. The initial learning rate is set to 0.005, the decay rate is set to 0.986, and the final learning rate is fixed to 0.0001.

[0078] L2 WeightDecay regularization is used for each parameter in each convolutional neural network.

[0079] The beneficial effects of this embodiment are:

[0080] The present invention automatically segments retinal blood vessels and optic discs in fundus images by using a convolutional neural network; calculates optic disc parameters and blood vessel parameters based on the segmentation results; determines the main vascular arch structure of the retinal result based on the blood vessel and optic disc parameters, and obtains the position of the macular area according to the structure.

[0081] The implementation process of the embodiment of the present invention is described in detail below:

[0082] This embodiment uses a 2-neural network architecture and a macular localization algorithm based on the retinal main vascular arch and optic disc, wherein the two neural networks used are convolutional neural networks.

[0083] First, the first convolutional neural network used in the present invention is a blood vessel segmentation network. After training, after inputting a retinal fundus image, a new image can be output, and the pixel intensity range on the image is between 0 and 1. Each pixel value represents the probability that it belongs to a retinal vessel, and the blood vessel segmentation process can effectively eliminate the changes in pigmentation, illumination and non-vascular lesions.

[0084] The second convolutional neural network is the optic disc segmentation network. After training, when a fundus image is input, it can output an optic disc segmentation probability similar to the first network. These two convolutional neural networks use the U-Net structure.

[0085] In the training and use of fundus images in the blood vessel and optic disc segmentation datasets for preprocessing, the data is mapped to the range of -1 to 1 by subtracting the mean and dividing by the variance, which can accelerate network training and convergence.

[0086] Then the dataset is divided into training set and test set with a ratio of 4:1.

[0087] For deep learning, a large amount of training data is required. However, in the case of limited training data, in order to improve the generalization ability and accuracy of the convolutional neural network, the segmentation network training module first performs data enhancement on the data. The data enhancement methods include random left and right mirror flip (50%), random top and bottom mirror flip (50%), random rotation (0-90 degrees), random cropping (480×640), random saturation jitter, random contrast jitter, random brightness jitter, etc.

[0088] After each pooling operation, the size of the feature map is reduced to half of its original size. As the size of the feature map continues to decrease, the convolutional neural network can fully learn the semantic information in the image, but it loses local detail information.

[0089] In order to better utilize the local detail information of the low layer, the present invention connects the upsampled high-level feature map with the bottom-level feature map to simultaneously obtain high-level semantic information and detail information, thereby improving the accuracy of blood vessel and optic disc segmentation. Figure 2 As shown, the segmentation network training module trains the blood vessel and optic disc segmentation network on the training set separated from the blood vessel and optic disc segmentation data set, and tests it on the test set. The training method of the optic disc segmentation network model and the blood vessel segmentation network model includes:

[0090] Use CrossEntropy Loss (SoftMax Loss) as the loss function;

[0091] Use mini-batch gradient descent as the learning algorithm for convolutional neural networks;

[0092] Each convolutional neural network is trained for 250 epochs. During the training, the learning rate is reduced according to the set decay strategy. The learning rate gradually decreases with the deepening of training. The learning rate decays once every epoch. The initial learning rate is set to 0.005, the decay rate is set to 0.986, and the final learning rate is fixed to 0.0001.

[0093] L2 WeightDecay regularization is used for each parameter in each convolutional neural network.

[0094] The model with the highest dice index on the test set was selected as the final blood vessel and optic disc segmentation network model.

[0095] The input of the blood vessel and optic disc segmentation network is an RGB three-channel fundus image. The network structure of each layer is as follows:

[0096]

[0097]

[0098] After training the blood vessel and optic disc segmentation network, the blood vessel and optic disc segmentation network can be used to locate the macular area.

[0099] Optic disc blood vessel segmentation module, including:

[0100] An optic disc segmentation unit, used for taking the retinal fundus image as an input of the trained optic disc segmentation network model to obtain an optic disc image;

[0101] The blood vessel segmentation unit is used to use the retinal fundus image as the input of the trained blood vessel segmentation network model to obtain a blood vessel image.

[0102] It is possible to obtain a binary optic disc image and a blood vessel image by segmenting the retinal fundus image.

[0103] The module for determining the region of interest in the macular area includes:

[0104] An optic disc parameter calculation unit is used to perform ellipse fitting on the optic disc in the optic disc image to obtain the center and radius of the optic disc, and the major axis and minor axis of the ellipse;

[0105] a macular region of interest calculation unit, used to dilate the optic disc in the optic disc image and determine the macular region of interest on the blood vessel image based on the optic disc radius;

[0106] The quadrant division unit is used to divide the macular area of ​​interest into four quadrants based on the long axis and short axis of the ellipse, wherein the long axis is the nasal side and the temporal side on the left and the short axis is the two main vascular arches on the left.

[0107] Furthermore, the expansion multiple of the optic disc is 1.2 times, and a square centered on the optic disc is made with a side length of 16 times the radius of the optic disc as the region of interest.

[0108] It can be implemented to determine the macular region of interest based on the optic disc image.

[0109] Due to blood vessels, optic disc, RGB color Figure 3 The positions of the two are consistent, so the determined regional positions can be converted to each other on each image. For example, the macular region of interest obtained from the optic disc image can be located on the vascular image. The same is true for the fitted ellipse, major axis, minor axis, blood vessels, etc.

[0110] The main vascular arch determination module includes:

[0111] The preferred main vascular arch calculation unit is used to make a Venn difference between the vascular image and the expanded optic disc image, and then perform an 8-connected domain calculation, retaining the area with a connected domain pixel number of not less than 30, and obtain a connected domain calculation graph, from which the connected domain with the largest number of connected domain pixels is found as the preferred main vascular arch; the preferred main vascular arch occupies one quadrant.

[0112] The blood vessel priority ranking unit is used to take the blood vessel widths at the optic disc radius positions of multiple preset multiples in the connected domain calculation map with the optic disc center as the center, and perform linear combination with the corresponding first preset proportional coefficient to obtain the average width of each blood vessel in the connected domain calculation map; and perform linear combination of the number of pixels in the connected domain and the average blood vessel width based on the second preset proportional coefficient, and rank the blood vessels according to the result.

[0113] The alternative main vascular arch calculation unit is used to determine the quadrant where the preferred main vascular arch is located; it is also used to determine whether each blood vessel and the preferred main vascular arch are in the same quadrant on the left and right but in different quadrants on the upper and lower sides based on the primary and secondary order of the blood vessels. If so, the alternative main vascular arch is obtained; otherwise, the next blood vessel is judged.

[0114] Among them, the blood vessel width is the distance from the outer edge of the blood vessel to the centerline of the blood vessel; the preset multiples are 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9 and 2.0, and the corresponding first preset proportional coefficients are 0.3, 0.3, 0.1, 0.06, 0.06, 0.06, 0.06, 0.06; the second preset proportional coefficients corresponding to the number of pixels in the connected domain and the average width of the blood vessel are 0.3 and 0.7.

[0115] The preferred main vascular arch and the candidate main vascular arch, namely the superior and inferior temporal vascular arch, can be obtained.

[0116] Macular area position calculation module, including:

[0117] like Figure 3 As shown, the macular positioning unit is used to make a straight line where the long axis is located based on the vascular image after the two main vascular arches are located, and to make a parallel line to the straight line where the long axis is located at a position 5 times the optic disc radius on one side of the two main vascular arches, and form a closed area with the two main vascular arches; then a circle is fitted to the closed area, and the center of the circle is used as the center of the macular area, and the circle fitting result is used as the positioning of the macular area.

[0118] The closed area generating unit is used to draw a perpendicular line from the nearest point of the main vascular arch to the straight line where the long axis is located to the straight line where the long axis is located when the straight line where the long axis is located, the parallel line to the straight line where the long axis is located and the two main vascular arches cannot form a closed area; and then form a closed area by the straight line where the long axis is located, the parallel line to the straight line where the long axis is located, the perpendicular line and the two main vascular arches.

[0119] Ultimately, the position of the corresponding blood vessel image is determined based on the position of the two main vascular arches in the macular area of ​​interest, and the macular area is located in combination with the long axis and radius.

[0120] In summary, compared with the prior art, the present invention has the following beneficial effects:

[0121] 1. The present invention uses a convolutional neural network to automatically segment the retinal blood vessels and optic disc in the fundus image; calculates the parameters of the optic disc and the blood vessels based on the segmentation results; determines the main vascular arch structure of the retinal result based on the blood vessel and optic disc parameters, and obtains the position of the macular area based on its structure. Automatic macular area positioning based on the retinal structure is achieved, saving the time loss of manual annotation for each image.

[0122] 2. When training the vascular optic disc segmentation network, data enhancement methods such as random rotation, random mirror flipping, random cropping, random brightness jitter, random contrast jitter, and random saturation jitter are used. CrossEntropy Loss (SoftMax Loss) is used as the loss function for training the vascular optic disc segmentation network, and the stochastic gradient descent algorithm is used as the training method. When training the vascular optic disc segmentation network, the learning rate of each training is less than or equal to the learning rate of the previous training. The information of the retinal structure is captured more accurately, and the final accuracy meets the level required by doctors.

[0123] It should be noted that, through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the above technical solution is essentially or the part that contributes to the prior art can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including several instructions to enable a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment. In this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the term "include", "comprise" or any other variant thereof is intended to cover non-exclusive inclusion, so that the process, method, article or device including a series of elements includes not only those elements, but also includes other elements that are not explicitly listed, or also includes elements inherent to such process, method, article or device. Without more constraints, an element defined by the phrase "comprising a..." does not exclude the existence of other identical elements in the process, method, article or apparatus comprising the element.

[0124] The above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A macular area positioning system based on retinal structure, characterized in that: The system includes: An optic disc blood vessel segmentation module is used to obtain a binary optic disc image and a blood vessel image by segmenting the retinal fundus image; A macular region of interest determination module, used to determine the macular region of interest based on the optic disc image; A main vascular arch determination module is used to perform connected domain detection and average vascular width calculation in the macular region of interest based on the vascular image; determine the preferred main vascular arch based on the number of pixels in the connected domain, and obtain the candidate main vascular arch through a linear combination of the number of pixels in the connected domain and the average vascular width; A macular area position calculation module, used for locating the macular area based on two main vascular arches, a long axis and a radius; The macular region of interest determination module comprises: An optic disc parameter calculation unit is used to perform ellipse fitting on the optic disc in the optic disc image to obtain the center and radius of the optic disc, and the major axis and minor axis of the ellipse; A macular region of interest calculation unit, used to determine the macular region of interest on the blood vessel image based on the optic disc radius, and to dilate the optic disc in the optic disc image; A quadrant division unit, used to divide the macular area of ​​interest into four quadrants based on the major axis and the minor axis of the ellipse; The main vascular arch determination module comprises: The preferred main vascular arch calculation unit is used to make a Venn difference between the vascular image and the expanded optic disc image, and then perform 8-connected domain calculation, retaining the area with a connected domain pixel number of not less than 30, and obtain a connected domain calculation graph, from which the connected domain with the largest number of pixels is found as the preferred main vascular arch; a blood vessel primary and secondary sorting unit, for taking the optic disc center as the center, taking the blood vessel widths at the optic disc radius positions of multiple preset multiples in the connected domain calculation map, and performing linear combination with the corresponding first preset proportional coefficient to obtain the average width of each blood vessel in the connected domain calculation map; and performing linear combination of the number of pixels in the connected domain and the average blood vessel width based on the second preset proportional coefficient, and sorting the blood vessels according to the result; The candidate main vascular arch calculation unit is used to determine the quadrant where the preferred main vascular arch is located; it is also used to determine, based on the order of the primary and secondary vessels, whether each vessel and the preferred main vascular arch are in the same quadrant on the left and right but in different quadrants on the upper and lower sides. If so, the candidate main vascular arch is obtained; otherwise, the next vessel is determined; The macular area position calculation module comprises: The macular positioning unit is used to make a straight line where the long axis is located based on the vascular image after the two main vascular arches are located, and to make a line parallel to the straight line where the long axis is located at a position 5 times the optic disc radius on one side of the two main vascular arches, and form a closed area with the two main vascular arches; then a circle is fitted to the closed area, and the center of the circle is used as the center of the macular area, and the circle fitting result is used as the positioning of the macular area.

2. The macular area positioning system based on retinal structure according to claim 1, characterized in that: The expansion multiple of the optic disc is 1.2 times, and a square centered on the optic disc is made with a side length of 16 times the radius of the optic disc as the region of interest.

3. The macular area positioning system based on retinal structure according to claim 1, characterized in that: The blood vessel width is the distance from the outer edge of the blood vessel to the centerline of the blood vessel; the preset multiples are 1.3, 1.4, 1.5, 1.6, 1.7, 1.8, 1.9 and 2.0, and the corresponding first preset proportional coefficients are 0.3, 0.3, 0.1, 0.06, 0.06, 0.06, 0.06 and 0.06; the second preset proportional coefficients corresponding to the number of pixels in the connected domain and the average width of the blood vessel are 0.3 and 0.

7.

4. The macular area positioning system based on retinal structure according to claim 1, characterized in that: The macular area position calculation module also includes: The closed area generating unit is used to draw a perpendicular line from the nearest point of the main vascular arch to the straight line where the long axis is located to the straight line where the long axis is located when the straight line where the long axis is located, the parallel line to the straight line where the long axis is located and the two main vascular arches cannot form a closed area; and then form a closed area by the straight line where the long axis is located, the parallel line to the straight line where the long axis is located, the perpendicular line and the two main vascular arches.

5. The macular area positioning system based on retinal structure as claimed in claim 1, characterized in that: The optic disc blood vessel segmentation module comprises: An optic disc segmentation unit, used for taking the retinal fundus image as an input of the trained optic disc segmentation network model to obtain an optic disc image; The blood vessel segmentation unit is used to use the retinal fundus image as the input of the trained blood vessel segmentation network model to obtain a blood vessel image.

6. The macular area positioning system based on retinal structure as claimed in claim 4, characterized in that: The optic disc segmentation network model and the blood vessel segmentation network model are convolutional neural networks with a U-Net structure.

7. The macular area positioning system based on retinal structure according to claim 6, characterized in that: The system further comprises: The segmentation network training module is used to train the optic disc segmentation network model and the blood vessel segmentation network model, and the training method of the optic disc segmentation network model and the blood vessel segmentation network model includes: Use CrossEntropy Loss as the loss function; Use mini-batch gradient descent as the learning algorithm for convolutional neural networks; Each convolutional neural network is trained for 250 epochs. During the training, the learning rate is reduced according to the set decay strategy. The learning rate gradually decreases with the deepening of training. The learning rate decays once every epoch. The initial learning rate is set to 0.005, the decay rate is set to 0.986, and the final learning rate is fixed to 0.0001. L2 Weight Decay regularization is used for each parameter in each convolutional neural network.