Method, device and equipment for recognizing retinal nerve fiber layer defect images

Abstract

The present invention discloses a method, device and equipment for identifying retinal nerve fiber layer defects, wherein the method includes: obtaining an eye image to be identified, including: retinal nerve fiber layer information; performing image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, the first identification area including at least two first feature points, each first feature point being used to characterize a retinal nerve fiber defect starting point; selecting any first feature point for parameter configuration to obtain an identification feature; based on the identification feature, performing sub-feature identification on the at least two first feature points respectively to obtain at least two sub-feature curves, each sub-feature curve being used to characterize the area of ​​retinal nerve fiber defects corresponding to the first feature point; obtaining a target identification area based on the at least two sub-feature curves, the target identification area being the retinal nerve fiber layer defect area. In the above manner, the present invention improves labeling efficiency and quality.

Description

Technical Field

[0001] The present invention relates to the technical field of fundus images, and in particular to a method, device and equipment for identifying retinal nerve fiber layer defects. Background Art

[0002] Detecting RNFLD (retinal nerve fiber layer defect) requires automated annotation of the shape of the RNFLD boundary. Prior to automated annotation, deep learning models for RNFLD detection relied on high-quality RNFLD recognition. Currently, RNFLD identification relies primarily on manual work and subjective experience by physicians. In practice, RNFLD identification is not only difficult to accurately describe the curved boundary of the RNFLD, but also involves a high workload and significant variability, introducing noise into model training. By automatically adapting the identification and description of different nerve fiber orientations to different fundus images, RNFLD shape recognition can be standardized, further improving detection quality.

[0003] In terms of modeling the direction of optic nerve fibers on fundus images, existing technologies focus on forming a unified mathematical description rather than automatically generating a matching function for any given fundus image without labeled nerve fiber directions.

[0004] Based on this, how to provide a method for automatically modeling the direction of nerve fibers on any given fundus image, and use the generated modeling to better assist in the identification of RNFLD, and how to improve the performance of RNFLD detection by performing polar coordinate expansion on the images obtained by the generated modeling are problems that urgently need to be solved by people in this field. Summary of the Invention

[0005] To solve the above problems, a method, device and apparatus for identifying retinal nerve fiber layer defects according to embodiments of the present invention are proposed.

[0006] According to one aspect of an embodiment of the present invention, a method for identifying retinal nerve fiber layer defects is provided, comprising:

[0007] Acquire an eye image to be identified, wherein the eye image to be labeled includes: retinal nerve fiber layer information;

[0008] Performing image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, where the first identification area includes at least two first feature points, each first feature point being used to represent a starting point of a retinal nerve fiber defect;

[0009] Select any first feature point to configure parameters and obtain identification features;

[0010] performing sub-feature recognition on the at least two first feature points according to the recognition feature to obtain at least two sub-feature curves, each sub-feature curve being used to characterize an area of ​​retinal nerve fiber loss corresponding to the first feature point;

[0011] A target identification area is obtained according to the at least two sub-feature annotation curves, and the target identification area is the retinal nerve fiber layer defect annotation area.

[0012] Optionally, selecting any first feature point to configure parameters to obtain an identification feature includes:

[0013] Select any first feature point to obtain at least one second feature point corresponding to the first feature point, each second feature point being used to represent a retinal nerve fiber defect termination point;

[0014] Select the second characteristic point of the target according to the configuration requirements;

[0015] The characteristic relationship between the first characteristic point and the target second characteristic point is calculated, and the parameter configuration is completed based on the characteristic relationship to obtain the identification feature. In this way, the corresponding curve parameter configuration can be obtained based on the starting point and the ending point of the retinal nerve fiber defect, which is used for drawing the subsequent annotation curve.

[0016] Optionally, the eye image to be identified and the first identification area are obtained by a preset image acquisition device.

[0017] Optionally, performing sub-feature recognition on the at least two first feature points based on the recognition feature to obtain at least two sub-feature curves includes:

[0018] Respectively obtaining coordinate information corresponding to the at least two first feature points;

[0019] respectively acquiring simulation information corresponding to the at least two first characteristic coordinate information, wherein the simulation information is obtained from a preset simulation model;

[0020] Calculating displacement information of each of the first feature points based on the coordinate information and the simulation information;

[0021] The at least two sub-characteristic curves are obtained according to the displacement information corresponding to each of the at least two first characteristic points.

[0022] Optionally, a method for acquiring the at least two first feature points includes:

[0023] The retinal nerve fiber layer information is input into a preset recognition model to obtain the at least two first feature points.

[0024] Optionally, the recognition model is obtained by:

[0025] Acquire an initial recognition model, where the initial recognition model is used as a model to be trained for the recognition model;

[0026] Inputting a training sample into the initial recognition model to obtain the recognition model, the training sample includes: a polar coordinate transformation image modeled by optic nerve fibers.

[0027] Specifically, the initial recognition model is obtained by:

[0028] And X0=(x0+δs·cosα,y0+δs·sinα) determines the initial recognition model of the fundus nerve fiber direction. The above model is obtained by simulating the electric field environment constructed with the center of the optic disc as positive charge and the center of the macula as negative charge. Among them, the direction of the optic nerve fiber at each position is determined by parameters q0, x0, x1, α, and δ. s Determine that the parameter q0 represents the amount of positive charge, the parameter x0 represents the position of positive charge, the parameter x1 represents the position of negative charge, the parameter α represents the angle of electric field line, and the parameter δ s are fixed parameters, E(x) refers to the electric field lines between the positive charge q0 at (x0, y0) and the negative charge q1 at (x1, y1), k refers to the Coulomb constant, and

[0029] Specifically, the method for obtaining the recognition model includes: adjusting the corresponding coefficient of the parameter x0 to obtain the recognition model of the nerve fiber direction of the fundus image.

[0030] Specifically, for the parameter δ s include: Wherein, dist represents the distance from the center of the optic disc to the center of the macula in the fundus image, and den represents the point for constructing each sub-feature annotation curve, which is a preset parameter.

[0031] Optionally, the polar coordinate transformation image modeled by the optic nerve fibers includes constructing the polar coordinate transformation image using the center of the optic disc as the polar coordinate center. In this way, calculations performed under the polar coordinate transformation image can convert angular offset features into linear offset features, thereby reducing the corresponding computational complexity.

[0032] According to another aspect of an embodiment of the present invention, a device for marking retinal nerve fiber layer defects is provided, the device comprising:

[0033] An acquisition module is used to acquire an eye image to be identified, wherein the eye image to be identified includes: retinal nerve fiber layer information;

[0034] an identification module, configured to perform image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, wherein the first identification area includes at least two first feature points, each first feature point being used to represent a starting point of a retinal nerve fiber defect;

[0035] An execution module is used to select any first feature point to configure parameters and obtain an identification feature;

[0036] a processing module, configured to perform sub-feature recognition on the at least two first feature points based on the recognition feature to obtain at least two sub-feature curves, each sub-feature curve being used to characterize an area of ​​retinal nerve fiber coloboma corresponding to the first feature point;

[0037] The labeling module is used to obtain a target identification area based on the at least two sub-characteristic curves, and the target identification area is the retinal nerve fiber layer defect area.

[0038] According to another aspect of an embodiment of the present invention, there is provided a computing device, comprising: a processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus;

[0039] The memory is used to store at least one executable instruction, and the executable instruction enables the processor to execute operations corresponding to the above-mentioned retinal nerve fiber layer defect marking method.

[0040] According to another aspect of the present invention, a computer storage medium is provided, wherein the storage medium stores at least one executable instruction, and the executable instruction enables a processor to perform operations corresponding to the above-mentioned retinal nerve fiber layer defect marking method.

[0041] According to the solution provided by the above-mentioned embodiment of the present invention, by obtaining an eye image to be identified, the eye image to be identified includes: retinal nerve fiber layer information; based on the retinal nerve fiber layer information, image recognition is performed on the eye image to be identified to obtain a first identification area, and the first identification area includes at least two first feature points, each first feature point is used to characterize a starting point of a retinal nerve fiber defect; any first feature point is selected for parameter configuration to obtain an identification feature; based on the identification feature, sub-feature identification is performed on the at least two first feature points respectively to obtain at least two sub-feature curves, each sub-feature curve is used to characterize the area of ​​retinal nerve fiber defect corresponding to the first feature point; based on the at least two sub-feature curves, a target identification area is obtained, and the target identification area is the retinal nerve fiber layer defect area, thereby improving the efficiency and quality of labeling.

[0042] The above description is only an overview of the technical solutions of the embodiments of the present invention. In order to more clearly understand the technical means of the embodiments of the present invention, they can be implemented in accordance with the contents of the specification. In order to make the above and other purposes, features and advantages of the embodiments of the present invention more obvious and easy to understand, the specific implementation methods of the embodiments of the present invention are listed below. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Various other advantages and benefits will become apparent to those skilled in the art upon reading the detailed description of the preferred embodiment below. The accompanying drawings are for illustration purposes only and are not to be considered as limiting the embodiments of the present invention. The same reference numerals are used throughout the accompanying drawings to denote the same components. In the accompanying drawings:

[0044] Figure 1 A flow chart showing a method for identifying retinal nerve fiber layer defects provided by an embodiment of the present invention is shown;

[0045] Figure 2 shows a schematic diagram of the distribution of the second characteristic points provided by an embodiment of the present invention;

[0046] Figure 3 A schematic diagram of at least two sub-feature annotation curves provided by an embodiment of the present invention is shown;

[0047] Figure 4 A schematic structural diagram of a retinal nerve fiber layer defect identification device provided by an embodiment of the present invention is shown;

[0048] Figure 5 A schematic structural diagram of a computing device provided by an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0049] Exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided to enable a more thorough understanding of the present invention and to fully convey the scope of the present invention to those skilled in the art.

[0050] Detecting RNFLD (retinal nerve fiber layer defect) requires automated annotation of the shape of the RNFLD boundary. Prior to automated annotation, deep learning models for RNFLD detection relied on high-quality RNFLD recognition. Currently, RNFLD identification relies primarily on manual work and subjective experience by physicians. In practice, RNFLD identification is not only difficult to accurately describe the curved boundary of the RNFLD, but also involves a high workload and significant variability, introducing noise into model training. By automatically adapting the identification and description of different nerve fiber orientations to different fundus images, RNFLD shape recognition can be standardized, further improving detection quality.

[0051] In terms of modeling the direction of optic nerve fibers on fundus images, existing technologies focus on forming a unified mathematical description rather than automatically generating a matching function for any given fundus image without labeled nerve fiber directions.

[0052] Based on this, how to provide a method for automatically modeling the direction of nerve fibers on any given fundus image, and use the generated modeling to better assist in the identification of RNFLD, and how to improve the performance of RNFLD detection by performing polar coordinate expansion on the images obtained by the generated modeling are problems that urgently need to be solved by people in this field.

[0053] In order to solve the above problems, Figure 1 FIG. 1 is a flow chart showing a method for identifying retinal nerve fiber layer defects according to an embodiment of the present invention. Figure 1 As shown, the method includes the following steps:

[0054] Step 11, obtaining an eye image to be identified, wherein the eye image to be identified includes: retinal nerve fiber layer information;

[0055] Step 12: performing image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, wherein the first identification area includes at least two first feature points, each first feature point being used to represent a starting point of a retinal nerve fiber defect;

[0056] Step 13: Select any first feature point to configure parameters and obtain identification features;

[0057] Step 14: Based on the identification feature, perform sub-feature identification on the at least two first feature points to obtain at least two sub-feature curves, as shown in FIG. Figure 3 As shown, each sub-characteristic curve corresponds to a first characteristic point, that is, the first characteristic point 1-1 corresponds to the sub-characteristic curve 1-1, and the first characteristic point 1-2 corresponds to the sub-characteristic curve 1-2. Each sub-characteristic curve is used to characterize the area of ​​retinal nerve fiber defect corresponding to the first characteristic point;

[0058] Step 15: obtaining a target identification area based on the at least two sub-characteristic curves, wherein the target identification area is the retinal nerve fiber layer defect area.

[0059] It can be understood that each of the sub-feature identification curves is actually used to describe the fiber orientation of the retinal nerve fibers with defects.

[0060] In an optional embodiment of the present invention, in step 13, selecting any first feature point to perform parameter configuration to obtain a marked feature includes:

[0061] Step 131: Select any first feature point and obtain at least one second feature point corresponding to the first feature point. Figure 2 As shown, Figure 2 A second characteristic point distribution diagram provided in an embodiment of the present invention is shown in FIG. Figure 2 The image includes: an eyeball to be identified, ocular blood vessels 1, ocular blood vessels 2, an optic disc, an optic cup and other physiological structures of the eyeball, and also includes a first identification area and a second identification area indicated by dotted lines, wherein a point in each dotted line represents a first feature point and a second feature point, respectively, and each second feature point is used to represent a retinal nerve fiber defect termination point;

[0062] Step 132, selecting a second characteristic point of the target according to the configuration requirements;

[0063] Step 133 : Calculate the characteristic relationship between the first characteristic point and the target second characteristic point, complete the parameter configuration according to the characteristic relationship, and obtain the recognition feature.

[0064] It is understandable that, since the distribution of retinal nerve fibers in the eye in real life is limited to a certain area, usually, the location of this fixed area is determined by the distance between the center of the optic disc and the center of the macula. Therefore, when the starting point of the retinal nerve fiber defect (hereinafter referred to as the defect) is determined, the defect end point corresponding to the determined defect starting point is also in a relatively fixed area. At the same time, in this fixed area, the marking of the actual retinal nerve fiber distribution curvature needs to be as close as possible. In order to facilitate user observation, when there is a clear demand for a specific curvature or marking angle (ie, configuration requirements), it is necessary to provide the user with at least one optional damage end point (ie, the second feature point) for selecting a more appropriate marking curve (ie, sub-feature marking curve).

[0065] Typically, the user will select the point that is closest or closest to the actual retinal nerve fiber damage end point as the second feature point.

[0066] In an optional embodiment of the present invention, in step 12, the eye image to be identified and the first identification area are obtained by a preset image acquisition device.

[0067] It is understandable that after obtaining the eye image to be labeled, the pre-set feature recognition model is used to extract features of the eye image to be identified and find the damaged area of ​​the retinal nerve fibers in the eye image to be identified (i.e., the first recognition area).

[0068] In an optional embodiment of the present invention, the at least two first feature points mentioned in step 12 are obtained by:

[0069] Step 121 : inputting the retinal nerve fiber layer information into a preset recognition model to obtain the at least two first feature points.

[0070] The identification model mentioned in step 121 is obtained by:

[0071] Step 1211, obtaining an initial recognition model, wherein the initial recognition model is used as a model to be trained for the recognition model;

[0072] Step 1212: Input a training sample into the initial recognition model to obtain the recognition model. The training sample includes: a polar coordinate transformation image modeled by optic nerve fibers.

[0073] It can be understood that the recognition model here can be regarded as a mother model or basic model before training. After training samples with corresponding eye features or retinal nerve fiber damage points (starting points and / or ending points), the recognition model that can be used to identify the first recognition area shown is obtained.

[0074] In an optional embodiment of the present invention, in step 14, performing sub-feature annotation on the at least two first feature points to obtain at least two sub-feature annotation curves includes:

[0075] Step 141, respectively obtaining coordinate information corresponding to the at least two first feature points;

[0076] Step 142: construct a model and obtain corresponding simulation information, wherein the simulation information is obtained from a preset simulation model;

[0077] Step 143: Calculate the displacement information of each of the first feature points based on the coordinate information and the simulation information;

[0078] Step 144 : Obtain the at least two sub-feature annotation curves based on the displacement information corresponding to the at least two first feature points.

[0079] Exemplarily, here we take the electric field model to simulate the retinal nerve fiber layer of the human eye as an example, use the recognition model to obtain at least two defect starting points (i.e., the first feature point), and obtain the position information or coordinate information of the at least two defect starting points. For optic nerve fiber modeling, the electric field environment is referenced for modeling, and the electric field environment is constructed with the center of the optic disc as a positive charge and the center of the macula as a negative charge, and the relevant electric field information is obtained. In the optic nerve fiber model, the direction of the optic nerve fiber at each position is determined by the parameters q0, x0, x1, α, and δ. s These parameters are the corresponding electric field information, specifically including: the parameter q0 represents the amount of positive charge, the parameter x0 represents the position of positive charge, the parameter x1 represents the position of negative charge, the parameter α represents the angle of electric field line, the parameter δ s are fixed parameters, E(x) refers to the electric field lines between the positive charge q0 at (x0, y0) and the negative charge q1 at (x1, y1), and k refers to the Coulomb constant.

[0080] In the optic nerve fiber model, for a fixed parameter δ s include: Wherein, dist represents the distance from the center of the optic disc to the center of the macula in the fundus image, and den represents the point for constructing each sub-feature annotation curve, which is a preset parameter.

[0081] For example, it is assumed that each sub-characteristic curve is required to consist of 240 points with equal distances, and the value of den is 240. The specific value range of den needs to be determined in combination with actual needs.

[0082] In the optic nerve fiber model, the starting coordinate information of the defect starting point (i.e., the first feature point) can usually be expressed as:

[0083] X0=(x0+δs·cosα, y0+δs·sinα)

[0084] Since the optic nerve fiber model is constructed based on the electric field model, the direction of the optic nerve fiber should satisfy the direction of the electric field line at the same starting point under the electric field model. That is, the distance that the nerve fiber moves from the defect starting point along the derivative direction of the current electric field line based on the previous point can be expressed as:

[0085]

[0086] In an optional embodiment of the present invention, for step 1212, the polar coordinate transformation image modeled by the optic nerve fibers includes: constructing the polar coordinate transformation image with the center of the optic disc as the polar coordinate center.

[0087] In an optional embodiment of the present invention, in step 13, selecting any first feature point for parameter configuration further includes:

[0088] Step 1311, obtaining all image features in the training sample in step 1212;

[0089] Step 1312: performing data integration on the image features to obtain at least one first feature sequence;

[0090] Step 1313, obtaining the features of the eye image to be labeled;

[0091] Step 1314: performing data integration on the eye image features to be labeled to obtain a second feature sequence;

[0092] Step 1315 : Use the second feature sequence to perform a similarity comparison with each first feature sequence. If the comparison result is less than a preset threshold, the feature corresponding to the first feature sequence corresponding to the comparison result is used as the parameter configuration result.

[0093] For example, when labeling retinal nerve fiber coloboma images, after the initial model is trained as a recognition model using the training samples, a corresponding comparison learning network can be trained to predict the labeling results in subsequent eye images to be identified. In this way, combining machine learning to predict the features of the image to be identified can reduce the labeling workload.

[0094] In the above-mentioned embodiment of the present invention, the input of the detection network is a polar coordinate transformation image sampled and generated according to the modeling of the direction of the optic nerve fibers. The RNFLD detection results are subdivided into multiple optic disc peripheral sectors, which can automatically model the direction of the nerve fibers on any given fundus image. In this way, the annotation tool developed according to this solution is easy to use, can greatly reduce the workload of doctors in annotation, and can improve the annotation quality.

[0095] Figure 4 FIG. 1 shows a schematic diagram of the structure of a retinal nerve fiber layer defect identification device provided by an embodiment of the present invention. Figure 4 As shown, the device includes:

[0096] An acquisition module is used to acquire an eye image to be identified, wherein the eye image to be identified includes: retinal nerve fiber layer information;

[0097] an identification module, configured to perform image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, wherein the first identification area includes at least two first feature points, each first feature point being used to represent a starting point of a retinal nerve fiber defect;

[0098] An execution module is used to select any first feature point to configure parameters and obtain an identification feature;

[0099] a processing module, configured to perform sub-feature recognition on the at least two first feature points based on the recognition feature to obtain at least two sub-feature curves, each sub-feature curve being used to characterize an area of ​​retinal nerve fiber coloboma corresponding to the first feature point;

[0100] The labeling module is used to obtain a target identification area based on the at least two sub-characteristic curves, and the target identification area is the retinal nerve fiber layer defect area.

[0101] It should be understood that the above Figures 1 to 3 The description of the exemplary method embodiments is merely an illustrative example of the technical solution of the present invention and does not limit the retinal nerve fiber layer defect labeling method of the present invention. In other embodiments, the execution steps and order of the retinal nerve fiber layer defect labeling method of the present invention may differ from the above-described embodiment, and the present invention does not limit this.

[0102] It should be noted that this embodiment is an apparatus embodiment corresponding to the above method embodiment, and all implementation methods in the above method embodiment are applicable to the embodiment of this apparatus and can achieve the same technical effects.

[0103] An embodiment of the present invention provides a non-volatile computer storage medium storing at least one executable instruction. The computer executable instruction can execute the retinal nerve fiber layer defect marking method in any of the above method embodiments.

[0104] Figure 5 The schematic diagram of the structure of the computing device provided by the embodiment of the present invention is shown. The specific embodiment of the present invention does not limit the specific implementation of the computing device.

[0105] like Figure 5 As shown, the computing device may include: a processor, a communication interface, a memory, and a communication bus.

[0106] The processor, communication interface, and memory communicate with each other via a communication bus. The communication interface is used to communicate with other devices, such as client devices or other server network elements. The processor is used to execute a program, specifically the steps described in the embodiment of the retinal nerve fiber layer defect labeling method for a computing device.

[0107] Specifically, the program may include program codes including computer operation instructions.

[0108] The processor may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present invention. The one or more processors included in the computing device may be processors of the same type, such as one or more CPUs, or processors of different types, such as one or more CPUs and one or more ASICs.

[0109] Memory is used to store programs. The memory may include high-speed RAM memory, and may also include non-volatile memory (non-volatile memory), such as at least one disk storage.

[0110] The program can be specifically configured to cause a processor to execute the retinal nerve fiber layer defect labeling method described in any of the above-described method embodiments. The specific implementation of each step in the program can be found in the corresponding descriptions of the corresponding steps and units in the above-described retinal nerve fiber layer defect labeling method embodiments, and will not be repeated here. Those skilled in the art will clearly understand that, for ease and brevity of description, the specific operating processes of the devices and modules described above can refer to the corresponding process descriptions in the above-described method embodiments, and will not be repeated here.

[0111] The algorithm or display provided herein is not inherently related to any particular computer, virtual system or other device. Various general-purpose systems can also be used together with the teachings based on this. According to the above description, it is obvious that the structure required for constructing such systems. In addition, the embodiment of the present invention is not directed to any specific programming language. It should be understood that various programming languages ​​can be utilized to implement the content of the embodiment of the present invention described herein, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of the embodiment of the present invention.

[0112] In the description provided herein, numerous specific details are described. However, it is understood that embodiments of the present invention may be practiced without these specific details. In some instances, well-known methods, structures, and techniques are not shown in detail so as not to obscure the understanding of this description.

[0113] Similarly, it should be understood that in order to streamline the embodiments of the present invention and aid in understanding one or more of the various aspects of the invention, in the above description of exemplary embodiments of the present invention, the various features of the embodiments of the present invention are sometimes grouped together into a single embodiment, figure, or description thereof.

[0114] Those skilled in the art will appreciate that the modules in the devices of the embodiments can be adaptively changed and installed in one or more devices different from the embodiments. The modules, units, or components in the embodiments can be combined into one module, unit, or component, and furthermore, they can be divided into multiple submodules, subunits, or subcomponents.

[0115] Furthermore, those skilled in the art will appreciate that although some embodiments herein include certain features and not other features included in other embodiments, the combination of features from different embodiments is intended to be within the scope of the present invention and to form different embodiments.

[0116] The various component embodiments of the present invention can be implemented in hardware, or in software modules running on one or more processors, or in a combination thereof. It should be understood by those skilled in the art that a microprocessor or digital signal processor (DSP) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of the present invention. The embodiments of the present invention can also be implemented as a device or apparatus program (e.g., a computer program and a computer program product) for executing part or all of the methods described herein. Such a program implementing an embodiment of the present invention can be stored on a computer-readable medium, or can have the form of one or more signals. Such a signal can be downloaded from an Internet website, or provided on a carrier signal, or provided in any other form.

[0117] It should be noted that the above embodiments illustrate rather than limit the present invention, and the word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. The present invention can be implemented using hardware comprising a number of different elements and using a suitably programmed computer. The use of the words first, second, and third, etc., does not denote any order. These words may be interpreted as names. The steps in the above embodiments, unless otherwise specified, should not be construed as limiting the order of execution.

Claims

1. A method for identifying retinal nerve fiber layer defects, characterized in that: The method comprises: Acquiring an eye image to be identified, wherein the eye image to be identified includes: retinal nerve fiber layer information; Performing image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, where the first identification area includes at least two first feature points, each first feature point being used to represent a starting point of a retinal nerve fiber defect; Select any first feature point to configure parameters and obtain identification features; performing sub-feature recognition on the at least two first feature points according to the recognition feature to obtain at least two sub-feature curves, each sub-feature curve being used to characterize an area of ​​retinal nerve fiber loss corresponding to the first feature point; Obtaining a target identification area based on the at least two sub-characteristic curves, wherein the target identification area is a retinal nerve fiber layer defect area; The selecting any first feature point to configure parameters to obtain the identification feature includes: Select any first feature point to obtain at least one second feature point corresponding to the first feature point, each second feature point being used to represent a retinal nerve fiber defect termination point; Select the second characteristic point of the target according to the configuration requirements; A feature relationship between the first feature point and the target second feature point is calculated, and the parameter configuration is completed according to the feature relationship to obtain the recognition feature.

2. The method according to claim 1, wherein The eye image to be identified and the first identification area are obtained by a preset image acquisition device.

3. The method according to claim 1, wherein The step of performing sub-feature recognition on the at least two first feature points based on the recognition feature to obtain at least two sub-feature curves includes: Respectively obtaining coordinate information corresponding to the at least two first feature points; respectively acquiring simulation information corresponding to the at least two first characteristic coordinate information, wherein the simulation information is obtained from a preset simulation model; Calculating displacement information of each of the first feature points based on the coordinate information and the simulation information; The at least two sub-characteristic curves are obtained according to the displacement information corresponding to each of the at least two first characteristic points.

4. The method according to claim 1, wherein The method of obtaining the at least two first feature points includes: The retinal nerve fiber layer information is input into a preset recognition model to obtain the at least two first feature points.

5. The method according to claim 4, wherein The identification model is obtained in the following manner: Acquire an initial recognition model, where the initial recognition model is used as a model to be trained for the recognition model; Inputting a training sample into the initial recognition model to obtain the recognition model, the training sample includes: a polar coordinate transformation image modeled by optic nerve fibers.

6. The method according to claim 5, wherein The polar coordinate conversion image modeled by optic nerve fibers includes: constructing the polar coordinate conversion image with the center of the optic disc as the polar coordinate center.

7. A retinal nerve fiber layer defect identification device, characterized in that: The device comprises: An acquisition module is used to acquire an eye image to be identified, wherein the eye image to be identified includes: retinal nerve fiber layer information; an identification module, configured to perform image recognition on the eye image to be identified based on the retinal nerve fiber layer information to obtain a first identification area, wherein the first identification area includes at least two first feature points, each first feature point being used to represent a starting point of a retinal nerve fiber defect; An execution module is used to select any first feature point to configure parameters and obtain an identification feature; a processing module, configured to perform sub-feature recognition on the at least two first feature points based on the recognition feature to obtain at least two sub-feature curves, each sub-feature curve being used to characterize an area of ​​retinal nerve fiber coloboma corresponding to the first feature point; a labeling module, configured to obtain a target identification area based on the at least two sub-characteristic curves, wherein the target identification area is a retinal nerve fiber layer defect area; The selecting any first feature point to configure parameters to obtain the identification feature includes: Select any first feature point to obtain at least one second feature point corresponding to the first feature point, each second feature point being used to represent a retinal nerve fiber defect termination point; Select the second characteristic point of the target according to the configuration requirements; A feature relationship between the first feature point and the target second feature point is calculated, and the parameter configuration is completed according to the feature relationship to obtain the recognition feature.

8. A computing device comprising: A processor, a memory, a communication interface, and a communication bus, wherein the processor, the memory, and the communication interface communicate with each other via the communication bus; The memory is used to store at least one executable instruction, and when the at least one executable instruction is executed, the processor executes the retinal nerve fiber layer defect marking method according to any one of claims 1 to 6.

9. A computer storage medium, wherein at least one executable instruction is stored in the storage medium, and when the executable instruction is executed, a computing device executes the retinal nerve fiber layer defect marking method according to any one of claims 1 to 6.

Images