KR20230123080A - Self supervised learning based apparatus and method for retinal image quality assessment - Google Patents

Abstract

Disclosed is an artificial intelligence-based eyeball image quality information management apparatus and method, and a method for determining quality information of an eyeball image according to an embodiment of the present invention includes dividing a blood vessel region from a plurality of eyeball images, corresponding to the blood vessel region Calculating the number of lines to generate, generating a plurality of artificial intelligence learning labels based on the number of lines and the additional information of each of the plurality of eyeball images, generating an eyeball image quality information model based on the artificial intelligence learning labels and determining quality information corresponding to the input of the subject's eyeball image based on the eyeball image quality information model.

Description

Apparatus and method for managing eye image quality information based on self-supervised learning {SELF SUPERVISED LEARNING BASED APPARATUS AND METHOD FOR RETINAL IMAGE QUALITY ASSESSMENT}

The present application relates to an apparatus and method for managing eye image quality information based on self-supervised learning.

Currently, in order to diagnose eye health, the center of the retina must be photographed with an eye camera. An eye camera is actually a microscope connected to a digital camera to examine the inside of the eye, taking high-resolution pictures for monitoring and diagnosis of degenerative eye disease or trauma. The conventional eye camera described above uses optical systems having different magnifications and angles of view when observing the cornea and when observing the retina. For example, as a corneal observation optical system, a low-magnification optical system is used because the distance and center must be aligned while viewing all of the sclera, iris, and cornea, while a high-magnification optical system is used to magnify and observe the retina. do. In such an eye camera, when the retinal focus is adjusted using the retina observation optical system, it is impossible to perform or readjust the distance or center alignment with the cornea. Therefore, if the distance between the eye camera and the eye to be examined or the alignment state of the eye to be examined is different, a correct eye picture cannot be obtained.

In addition, there is a problem in that when an eyeball image of poor quality is used in an automatic examination using artificial intelligence, an erroneous result is derived and the probability of misdiagnosis increases.

In addition, existing eyeball images have inconvenience in managing quality by directly designing and extracting features such as sharpness, focus value, average value, and representative value.

In addition, in order to build a video quality management system, an expert must directly inspect the quality and build a database, and this work is inefficient in terms of time, manpower, and cost.

The background technology of the present application is disclosed in Korean Patent Registration No. 10-2250694.

The present invention is to solve the above-mentioned problems of the prior art, and receives eye image data obtained by photographing a subject's eye area, divides a blood vessel region in the eye image data, calculates a line corresponding to the blood vessel region, and calculates a line based on the line. Provides a device that generates a virtual label, uses the generated virtual label as an input, creates a model through artificial intelligence-based learning, and inputs eye image data into the generated model to determine the quality information of the subject's eye image aims to do

However, the technical problem to be achieved by the embodiments of the present application is not limited to the technical problems described above, and other technical problems may exist.

As a means for achieving the above technical problem, an eyeball image quality management apparatus according to an embodiment of the present application includes a division unit dividing a blood vessel region from a plurality of eyeball images and a calculation unit calculating the number of lines corresponding to the blood vessel region. a label generator for generating a plurality of artificial intelligence learning labels based on the number of lines and the additional information of each of the plurality of eyeball images; a model generator for generating an eyeball image quality information model based on the artificial intelligence learning labels; and a quality determining unit configured to determine quality information of an eyeball image of the subject corresponding to an input of the subject's eyeball image based on the eyeball image quality information model.

Also, the segmentation unit may segment the blood vessel region from the eyeball image data based on a blood vessel region segmentation model pre-constructed through artificial intelligence-based learning to identify the blood vessel region.

Also, the label generating unit may generate a virtual label using the number of lines and the value of the boundary line.

In addition, the virtual label is generated based on the following [Equation 1],

_Pi may be the virtual label, h _i may be the number of lines, and Tau ⁺ and Tau ^- may be boundary line values.

In addition, the model generation unit may generate a model through artificial intelligence-based learning using the generated virtual label as an input.

Meanwhile, a method for managing eye image quality information according to an embodiment of the present disclosure includes dividing blood vessel regions from a plurality of eye image images, calculating lines corresponding to the blood vessel regions, the number of lines, and each of the plurality of eye image images. Generating a plurality of artificial intelligence learning labels based on additional information of, generating an eyeball image quality information model based on the artificial intelligence learning labels, and inputting an eyeball image of a subject based on the eyeball image quality information model It may include determining quality information corresponding to .

In the dividing, the blood vessel region may be segmented from the eyeball image data based on a blood vessel region segmentation model pre-constructed through artificial intelligence-based learning to identify the blood vessel region.

Also, in the generating of the learning label, a virtual label may be generated using the number of lines and the value of the boundary line.

In addition, the virtual label is generated based on the following [Equation 1],

_Pi may be the virtual label, h _i may be the number of lines, and Tau ⁺ and Tau ^- may be boundary line values.

In addition, the generating of the eyeball image quality information model may generate the model through artificial intelligence-based learning using the generated virtual label as an input.

The above-described problem solving means are merely exemplary and should not be construed as intended to limit the present disclosure. In addition to the exemplary embodiments described above, additional embodiments may exist in the drawings and detailed description of the invention.

According to the above-mentioned problem solving means of the present application, it is possible to provide a self-supervised learning-based eyeball image quality information management device and method.

According to the above-mentioned problem solving means of the present application, it is possible to obtain a high-quality label without an ophthalmic imaging specialist.

According to the above-described problem solving means of the present application, even if there is no quality label in the eyeball image, a virtual quality label can be obtained and applied to a deep learning methodology that requires big data.

According to the above-described problem solving means of the present application, it is possible to avoid performance degradation and errors of the artificial intelligence system due to low quality.

According to the above-described problem solving means of the present application, an image of good quality may be registered in a process of registering an image of a patient through determination of eye image quality information.

1 is a schematic configuration diagram of an eyeball image quality information management system including a processing device according to an embodiment of the present invention.
2 is a schematic configuration diagram of an apparatus for managing eyeball image quality information according to an embodiment of the present disclosure.
3 is a diagram illustratively illustrating results of segmenting a blood vessel region from eye image data according to an exemplary embodiment of the present disclosure.
4 is a diagram illustratively illustrating a result of determining quality information by inputting an eyeball image to an eyeball image quality information model according to an embodiment of the present disclosure.
5 is an operational flowchart of a method for managing eyeball image quality information according to an embodiment of the present invention.

Hereinafter, embodiments of the present application will be described in detail so that those skilled in the art can easily practice with reference to the accompanying drawings. However, the present disclosure may be implemented in many different forms and is not limited to the embodiments described herein. And in order to clearly describe the present application in the drawings, parts irrelevant to the description are omitted, and similar reference numerals are attached to similar parts throughout the specification.

Throughout this specification, when a part is said to be "connected" to another part, this includes not only the case of being "directly connected" but also the case of being "electrically connected" with another element in between. do.

Throughout the present specification, when a member is referred to as being “on,” “above,” “on top of,” “below,” “below,” or “below” another member, this means that a member is located in relation to another member. This includes not only the case of contact but also the case of another member between the two members.

Throughout the present specification, when a part "includes" a certain component, it means that it may further include other components without excluding other components unless otherwise stated.

The present application relates to an apparatus 40 and method for managing eye image quality information based on self-supervised learning.

1 is a schematic configuration diagram of an eyeball image quality information management system 10 including a processing device according to an embodiment of the present invention.

Referring to FIG. 1 , an eyeball image quality information management system 10 according to an embodiment of the present invention may include an eyeball imaging device 30 and an eyeball image quality information management device 40 according to an embodiment of the present invention. can Exemplarily, the eyeball imaging device 30 is a terminal capable of transmitting to the eyeball image quality information management device 40, and the specific type is not limited. For example, the eyeball imaging device 30 may be at least one of a desktop computer, a laptop computer, and a hospital information system server installed in a hospital, and may be a smartphone, tablet, or handheld used by a doctor or nurse ( handheld) may be a portable computing device such as a PC, PDA, or PMP. Also, the eyeball imaging device 30 may be a computing device used by a patient or a third party. In addition, the eyeball imaging device 30 may obtain an eyeball image using a Musandong eyeball camera commonly used in ophthalmology.

The eyeball imaging device 30 , the eyeball image quality information management device 40 , and a user terminal (not shown) may communicate with each other through the network 20 . The network 20 refers to a connection structure capable of exchanging information between nodes such as terminals and servers, and examples of such a network 20 include a 3rd Generation Partnership Project (3GPP) network and a Long LTE (LTE) network. Term Evolution (Term Evolution) network, 5G network, WIMAX (World Interoperability for Microwave Access) network, Internet, LAN (Local Area Network), Wireless LAN (Wireless Local Area Network), WAN (Wide Area Network), PAN (Personal Area Network) Network), wifi network, Bluetooth network, satellite broadcasting network, analog broadcasting network, DMB (Digital Multimedia Broadcasting) network, etc. are included, but are not limited thereto.

User terminals (not shown) include, for example, a smart phone, a smart pad, a tablet PC, and the like and PCS (Personal Communication System), GSM (Global System for Mobile communication), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), Wibro (Wireless Broadband Internet) It may be any kind of wireless communication device such as a terminal.

According to an embodiment of the present application, the eyeball image quality information management apparatus 40 may receive an eyeball image obtained by capturing an eyeball region of a subject. For example, the subject's ocular region may be any one of the retina, optic disc, macula, vitreous body, and choroid, but is not limited thereto, and various parts constituting the subject's eyeball in consideration of the subject's disease, symptoms, etc. Of course, it can be determined.

2 is a schematic configuration diagram of an apparatus for managing eyeball image quality information according to an embodiment of the present disclosure.

An eyeball image quality information management apparatus 40 according to an embodiment of the present application includes a division unit 110 for dividing a blood vessel region from a plurality of eyeball images, a calculation unit 120 for calculating the number of lines corresponding to the blood vessel region, A label generation unit 130 generating a plurality of artificial intelligence learning labels based on the number of lines and the additional information of each of the plurality of eyeball images, and a model generating an eyeball image quality information model based on the artificial intelligence learning labels It may include a unit 140 and a quality determination unit 150 that determines quality information of the subject's eyeball image corresponding to the input of the subject's eyeball image based on the eyeball image quality information model.

According to an embodiment of the present application, the segmentation unit 110 may segment a blood vessel region from a plurality of eyeball images.

Specifically, the segmentation unit 110 divides an eye image composed of pixels through a deep learning-based algorithm model, groups pixels having similar properties, and assigns a label to distinguish and divide regions around blood vessels and blood vessel regions in the eye image. can do.

In this regard, FIG. 3 is a diagram exemplarily illustrating a result of segmenting a blood vessel region from eye image data according to an exemplary embodiment of the present disclosure.

For example, the segmentation unit 110 may perform blood vessel segmentation by constructing LeNet, AlexNet, GoogleNet, VGGNet, and ResNet Unet, which are algorithm models based on deep learning, when eye image data is received. It is not limited, and any methodology required for blood vessel segmentation may be applied. Illustratively, the type of learning model disclosed herein may be Unet.

Blood vessel segmentation does not process the entire eye image data, but divides the eye image data into several parts to utilize important parts for processing the eye image data. The eye image data is composed of pixels and has similar properties using segmentation. Pixels with n are grouped together to obtain more granular information about the eye image data.

In addition, each grouped portion is assigned a label, and the label can be used to designate a boundary separating the vessel region from the rest of the eye image data.

In addition, the segmentation unit 110 may segment a blood vessel region from the subject's eye image based on a blood vessel region segmentation model pre-constructed through artificial intelligence-based learning to identify the blood vessel region.

Here, the vascular region segmentation model built through artificial intelligence-based learning may be implemented as a deep learning model suitable for image segmentation, such as Unet. Unet is a convolutional network for image segmentation in the biomedical field and has the advantage of high speed with an end-to-end structure.

In addition, the Unet algorithm can generate a deep convolutional neural network using the Convolution-Batch Normalization-ReLU layer block. The U-Net model architecture can output images with more granularity than traditional encoder-decoder models that take an image as input and downsample for the layers until the layers in the image are sampled. In other words, Unet passes through the network through skip connections between layers of the same size between downsampling and upsampling. By returning, you can convey high-level abstract information.

Here, artificial intelligence refers to the field of researching artificial intelligence or methodology to create it, and machine learning (Machine Learning) is the field of defining various problems dealt with in the field of artificial intelligence and studying methodologies to solve them. means Machine learning is also defined as an algorithm that improves the performance of a certain task through constant experience.

As another example, the vascular segmentation model built through artificial intelligence-based learning is semantic region segmentation that is learned to determine the classification category for each pixel of an input image (medical image data) in a predefined class (object type) ( Semantic Segmentation) based model. In this regard, a segmentation model based on semantic region segmentation according to an embodiment of the present disclosure may be a model that considers each tissue type described above as a class for region segmentation.

On the other hand, an Artificial Neural Network (ANN) is a model used in machine learning, and is composed of artificial neurons (nodes) that form a network 20 by synaptic coupling. It means an overall model with problem solving ability can do. An artificial neural network may be defined by a connection pattern between neurons in different layers, a learning process for updating model parameters, and an activation function for generating output values.

An artificial neural network may include an input layer, an output layer, and optionally one or more hidden layers. Each layer may include one or more neurons, and the artificial neural network may include neurons and synapses connecting the neurons. In an artificial neural network, each neuron may output a function value of an activation function for input signals, weights, and biases input through a synapse.

In this case, the number of nodes in the input layer may be a dimension of an input eyeball image, and the number of nodes in an output layer may be a vector dimension of eyeball image quality information. That is, the number of input nodes for the learning model constructed by the apparatus 40 for managing eye image quality information disclosed herein corresponds to the number of pixels of an eye image, and the number of output nodes is a scalar value such as good or bad quality. so it can be one.

Model parameters refer to parameters determined through learning, and include weights of synaptic connections and biases of neurons. In addition, hyperparameters mean parameters that must be set before learning in a machine learning algorithm, and include a learning rate, number of iterations, mini-batch size, initialization function, and the like.

In addition, the purpose of learning an artificial neural network can be seen as determining model parameters that minimize a loss function. The loss function may be used as an index for determining optimal model parameters in the learning process of an artificial neural network.

Machine learning can be classified into supervised learning, unsupervised learning, and reinforcement learning according to learning methods.

Supervised learning refers to a method of training an artificial neural network given a label for training data, and a label is the correct answer (or result value) that the artificial neural network must infer when learning data is input to the artificial neural network. can mean Unsupervised learning may refer to a method of training an artificial neural network in a state in which a label for training data is not given. Reinforcement learning may refer to a learning method in which an agent defined in an environment learns to select an action or action sequence that maximizes a cumulative reward in each state.

According to an embodiment of the present application, the calculation unit 120 may calculate the number of lines corresponding to the blood vessel area.

Specifically, the calculation unit 120 may detect lines corresponding to blood vessel regions classified through a blood vessel region segmentation model in the eyeball image data and count the number of lines.

Here, the line is a parameter in the coordinate space of the feature pixels of the image by using the feature of having an intersection in the parameter space in the case of pixels in the coordinate space appearing in the form of a curve in the image and pixels existing on the same straight line in the coordinate space. After mapping into space, it means a line expressed by finding an intersection through a voting process and extracting a straight line component.

According to an embodiment of the present application, the label generator 130 may generate a plurality of artificial intelligence learning labels based on the number of lines and additional information of each of a plurality of eyeball images.

Here, the additional information may include, but is not limited to, axial length and ocular age, choroidal blood vessels, optic nerve head shape, blood vessel conditions, macular degeneration, diabetic retinopathy, glaucoma, and the like.

Specifically, the label generation unit 130 adds the number of lines corresponding to the blood vessel area, the axial length, the eye age, the choroidal blood vessel, the shape of the optic disc, the state of the blood vessel, macular degeneration, diabetic retinopathy, glaucoma, and the like. By generating a plurality of artificial intelligence learning labels based on the information, it is possible to create artificial intelligence learning labels related to various eye diseases by linking the blood vessel area and additional information of the eye image.

According to an embodiment of the present application, the label generator 130 may generate a virtual label using the number of lines and the boundary value.

Specifically, a virtual label _pi may be generated by using a value h _i obtained by counting lines as a boundary line value. A virtual label according to the boundary value can be calculated by [Equation 1] below.

[Equation 1]

In this regard, the variable h _i in [Equation 1] is a value obtained by calculating the number of lines, and Tau ⁺ and Tau ^- are boundary values, and a virtual label p _i is generated using the calculation result and the two boundary values. can do.

In addition, h _i information in which a value obtained by calculating a line is between two boundary line values may not be used to remove ambiguity of the generated virtual label.

According to an embodiment of the present application, the model generation unit 140 may generate an eyeball image quality information model based on the artificial intelligence learning label.

Specifically, the model generation unit 140 may generate an eyeball image quality information model for classifying the quality of an eyeball image based on the learning label. To this end, the model generation unit 140 creates a feature vector by setting elements related to classifying quality from learning labels as characteristics, maps a label indicating a specific class to an object variable, and maps the feature vector and the object variable to an object variable. It is possible to create an eye image quality information model in which the correlation with is learned.

In addition, the model generation unit 140 may generate an eye image quality information model through artificial intelligence-based learning using the generated virtual label as an input.

Specifically, the model generating unit 140 classifies eye images to be labeled based on the number of lines of blood vessel regions in the unlabeled eye image, learns virtual labels, and generates an eye image quality information model, thereby generating an eye image without a label. Quality information may be determined for an image.

For example, an eye image quality information model that determines the quality information of an eye image through artificial intelligence-based learning using a virtual label generated for an eye image having a boundary value of 10 and counting lines of 15 as an input. It is possible to determine the quality information of the eye image by generating.

In addition, the process of inputting an eye image without a label to the eye image quality information model and inputting the eye image to the eye image quality information model again when an eye image of a certain grade or less is output until a certain grade is satisfied is learned, thereby providing eye image quality information. model can be created.

On the other hand, the labeled data is fed back to the eye image quality information model, and by learning, the accuracy of machine learning can be maintained continuously.

In addition, the aforementioned artificial intelligence-based learning may be learned in an unsupervised form using an unlabeled data set. Unsupervised learning may refer to a method of training an artificial neural network in a state in which a label for training data is not given.

Unsupervised learning is used to discover hidden features or structures of data. Illustratively, unsupervised learning may include K-Means divided into clustering, Hierarchical Cluster Analysis (HCA), and Expectation Maximization. In addition, Principal Component Analysis (PCA), Kernel PCA, Locally-Linear Embedding (LLE, Locally-Linear Embedding), t-SNE ( t-distributed StochasticNeighbor Embedding). In addition, Apriori and Eclat, which are classified as association rule learning, may be included. Hierarchical clustering algorithms allow each group to be subdivided into smaller groups. Visualization algorithms take large, unlabeled, high-dimensional data and create a 2D or 3D representation that can be plotted. Dimensionality reduction is used to simplify data without losing too much information. For example, since a car's mileage is highly correlated with its age, a dimensionality reduction algorithm could combine the two features into a single feature representing the wear and tear of the car.

In addition, the K-means clustering algorithm is a traditional classification technique that repeatedly subdivides a target group into K clusters based on the average value of distance (similarity).

According to an embodiment of the present application, the quality determination unit 150 may determine quality information corresponding to the input of the subject's eyeball image based on the eyeball image quality information model.

Specifically, the quality determination unit 150 may classify and determine quality information corresponding to an eyeball image input of a subject according to a preset grade based on an eyeball image quality information model learned through a virtual label.

In this regard, FIG. 4 is a diagram illustratively illustrating a result of determining quality information by inputting an eyeball image to an eyeball image quality information model according to an embodiment of the present disclosure.

Referring to FIG. 4 , the quality determination unit 150 may classify the quality of an eyeball image through an eyeball image quality information model learned with a generated virtual label.

Figure 4(a) is the result of classifying the image as a good level, and Figure 4(b) shows the result of classifying the image as a bad level. A model is created through learning based on high-quality images in which virtual labels are generated, and even if the number of lines calculated for the input eye image data is less than the boundary value, the quality information of the eye image is obtained through the generated eye image quality information model. can decide

For example, the quality of the eyeball image may be evaluated by dividing the quality into good and bad grades, but it is not limited to the two classification methods.

According to an embodiment of the present application, an output unit (not shown) may output an image obtained by inputting eyeball image data to a generated model.

Specifically, the output unit (not shown) may input an eyeball image of the subject to the blood vessel region segmentation model input, transmit the image of the divided blood vessel region to an external device, and output the divided image.

Also, the output unit (not shown) may display the blood vessel area on the screen and update the output image data according to the user's preference such as contrast and brightness.

With the configuration as described above, not only can the user arbitrarily set the output of the image preferred by the user, but also select and set the output of the image more accurately suited to the user's preference by referring to the blood vessel area displayed on the screen as a criterion. can

Hereinafter, based on the details described above, the operation flow of the present application will be briefly reviewed.

5 is an operational flowchart of a method for managing eyeball image quality information according to an embodiment of the present invention.

The eyeball image quality information management method shown in FIG. 5 may be performed by the previously described eyeball image quality information management apparatus. Therefore, even if omitted below, the description of the apparatus for managing eye image quality information can be equally applied to the description of the method for managing eye image quality information.

Referring to FIG. 5 , in step S501, the segmentation unit 110 may segment a blood vessel region from a plurality of eyeball images.

Next, in step S502, the calculation unit 120 may calculate the number of lines corresponding to the blood vessel area.

Next, in step S503, the label generation unit 130 may generate a plurality of artificial intelligence learning labels based on the number of lines and the additional information of each of the plurality of eyeball images.

Next, in step S504, the model generation unit 140 may generate an eyeball image quality information model based on the artificial intelligence learning label.

Next, in step S505, the quality determination unit 150 may determine quality information corresponding to the input of the subject's eyeball image based on the eyeball image quality information model.

In the foregoing description, steps S501 to S505 may be further divided into additional steps or combined into fewer steps, depending on the implementation of the present application. Also, some steps may be omitted if necessary, and the order of steps may be changed.

The eyeball image quality information management method according to an embodiment of the present application may be implemented in the form of program instructions that can be executed through various computer means and recorded in a computer readable medium. The computer readable medium may include program instructions, data files, data structures, etc. alone or in combination. Program instructions recorded on the medium may be those specially designed and configured for the present invention or those known and usable to those skilled in computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks and magnetic tapes, optical media such as CD-ROMs and DVDs, and magnetic media such as floptical disks. - includes hardware devices specially configured to store and execute program instructions, such as magneto-optical media, and ROM, RAM, flash memory, and the like. Examples of program instructions include high-level language codes that can be executed by a computer using an interpreter, as well as machine language codes such as those produced by a compiler. The hardware devices described above may be configured to act as one or more software modules to perform the operations of the present invention, and vice versa.

In addition, the above-described eyeball image quality management and control method may be implemented in the form of a computer program or application stored in a recording medium and executed by a computer. Those skilled in the art will understand that it can be easily modified into other specific forms without changing the technical spirit or essential features of the present application. Therefore, the embodiments described above should be understood as illustrative in all respects and not limiting. For example, each component described as a single type may be implemented in a distributed manner, and similarly, components described as distributed may be implemented in a combined form.

The scope of the present application is indicated by the following claims rather than the detailed description above, and all changes or modifications derived from the meaning and scope of the claims and equivalent concepts thereof should be construed as being included in the scope of the present application.

10: eye image quality information management system
20: Network
30: eye imaging device
40: ocular image quality information management device
110: division part
120: calculation unit
130: label generation unit
140: model generating unit
150: quality determining unit

Claims (11)

An apparatus for determining quality information of an eye image of a subject,
a division unit dividing a blood vessel region from a plurality of eye images;
a calculator configured to calculate the number of lines corresponding to the blood vessel area;
a label generation unit generating a plurality of artificial intelligence learning labels based on the number of lines and the additional information of each of the plurality of eyeball images;
a model generating unit generating an eyeball image quality information model based on the artificial intelligence learning label; and
a quality determining unit determining quality information of the eyeball image of the subject corresponding to an input of the subject's eyeball image based on the eyeball image quality information model;
To include, the ocular image quality information management device. According to claim 1,
The division part,
and dividing the blood vessel region in the subject's eye image based on a blood vessel region segmentation model pre-constructed through artificial intelligence-based learning to identify the blood vessel region. According to claim 1,
The label generator,
An apparatus for managing eye image quality information, wherein a virtual label is generated using the number of lines and the value of the boundary line. According to claim 3,
The virtual label,
It is generated based on [Equation 1] below,

P _i is the virtual label, h _i is the number of lines, and Tau ⁺ and Tau ^- are boundary line values. According to claim 4,
The model generator,
and generating the eye image quality information model through artificial intelligence-based learning using the generated virtual label as an input. A method for determining quality information of an eye image of a subject,
segmenting blood vessel regions from the plurality of eye images;
calculating the number of lines corresponding to the blood vessel area;
generating a plurality of artificial intelligence learning labels based on the number of lines and additional information of each of the plurality of eyeball images;
generating an eye image quality information model based on the artificial intelligence learning label; and
determining quality information corresponding to an input of an eyeball image of a subject based on the eyeball image quality information model;
A method for managing eye image quality information, comprising: According to claim 1,
The dividing step is
and dividing the blood vessel region in the subject's eye image based on a blood vessel region segmentation model pre-constructed through artificial intelligence-based learning to identify the blood vessel region. According to claim 1,
The step of generating the learning label,
The method of managing eye image quality information, wherein a virtual label is generated using the number of lines and the boundary line value. According to claim 3,
The virtual label,
It is generated based on [Equation 1] below,

P _i is the virtual label, h _i is the number of lines, and Tau ⁺ and Tau ^- are boundary line values. According to claim 4,
The step of generating the eyeball image quality information model,
and generating the eye image quality information model through artificial intelligence-based learning using the generated virtual label as an input. A computer-readable recording medium recording a program for executing the method of any one of claims 1 to 5 in a computer.

Images