A system, method and medium for ophthalmic slit image detection

Abstract

The present invention relates to the technical field of ophthalmic disease diagnosis, and discloses a system for ophthalmic slit image detection, including: a video acquisition module that acquires an examination video of a patient's eye by integrating a high-definition camera into a slit lamp microscope; a key frame extraction module that extracts key frames from the examination video of the patient's eye; a semantic segmentation module that divides each key frame image into different sub-regions through a semantic segmentation model; a global graph data construction module that constructs global graph data according to the sub-regions of the key frame images; a diagnosis result output module that inputs the global graph data into a trained diagnosis model, and the output value represents the diagnosis result of the patient's eye; the present invention constructs the examination video of the patient's eye into a global graph data representation by extracting key frames and image semantic segmentation, and performs aggregation analysis on different sub-regions of each key frame through the diagnosis model, so as to improve the semantic understanding of multiple ophthalmic slit images and improve the accuracy of the diagnosis result.

Description

Technical Field

[0001] The present invention relates to the technical field of ophthalmic disease diagnosis, and more specifically, it relates to a system, method and medium for detecting ophthalmic slit images. Background Art

[0002] The slit lamp microscope is an indispensable tool in the diagnosis of eye diseases. It uses a high-intensity light source to form a parallel light beam through a group of lenses, and then forms a slit light band through an adjustable diaphragm and projects it onto the patient's eye. The length, width, direction and brightness of the slit light band can all be adjusted. Therefore, doctors can observe the anterior segment of the patient's eye through the slit lamp microscope, including: the cornea, iris and lens, etc., and rely on the doctor's professional experience to discover the pathological signs of the patient's eye. However, this diagnosis method highly depends on the doctor's professional experience, and there may be subjective differences in the diagnosis results among different doctors.

[0003] With the development of artificial intelligence, more and more intelligent models are applied to the field of medical diagnosis. Currently, feature extraction is performed from a single ophthalmic slit image through a CNN (Convolutional Neural Network Model) to identify and classify ophthalmic diseases. In theory, it can provide consistent diagnosis results. However, analyzing only a single ophthalmic slit image has certain limitations, that is, it has high requirements for the imaging quality of the image. A single ophthalmic slit image may not fully display the complete ophthalmic pathological signs, and there is a lack of semantic understanding of multiple ophthalmic slit images, resulting in low accuracy of the diagnosis results. Summary of the Invention

[0004] The present invention provides a system, method and medium for detecting ophthalmic slit images to solve the technical problems in the above background art.

[0005] The present invention provides a system for detecting ophthalmic slit images, including:

[0006] A video acquisition module, which is used to collect the examination video of the patient's eye by integrating a high-definition camera with the slit lamp microscope;

[0007] A key frame extraction module, which is used to extract key frames from the examination video of the patient's eye to obtain an image dataset with the number of key frame images being M;

[0008] Where M is a custom parameter and is a positive integer greater than 1;

[0009] A semantic segmentation module, which is used to traverse the M key frame images of the image dataset and divide each key frame image into different sub-regions through a semantic segmentation model;

[0010] The sub-regions include: eyelids, conjunctiva, sclera, cornea, anterior chamber, iris, pupil, lens, lens capsule and anterior vitreous;

[0011] A global graph data construction module, which is used to construct global graph data according to sub-regions of M key-frame images;

[0012] The global graph data includes: nodes and edges between nodes;

[0013] The nodes of the global graph data establish data connections with the sub-regions of each key-frame image;

[0014] Edges are formed between nodes corresponding to adjacent sub-regions of the same key-frame image;

[0015] Edges are formed between nodes corresponding to the same sub-region of adjacent key-frame images;

[0016] A diagnostic model training module, which is used to construct and train a diagnostic model;

[0017] A diagnostic result output module, which is used to input the global graph data into the trained diagnostic model, and the output value represents the diagnostic result of the patient's eye;

[0018] The diagnostic results include: refractive error, conjunctival disease, corneal disease, anterior chamber disease, iris disease and eye trauma.

[0019] Furthermore, key-frame extraction is performed on the examination video of the patient's eye to obtain an image dataset with the number of key-frame images being M, including the following steps:

[0020] Step S201, dividing the examination video of the patient's eye at equal time intervals to obtain N sub-frame images, where N is a custom parameter greater than M;

[0021] Step S202, calculating the similarity value between two adjacent sub-frame images among the N sub-frame images through the structural similarity index, and if it is determined that the similarity value is less than or equal to the similarity threshold, then the latter sub-frame image is used as a key frame, where the similarity threshold is a custom parameter between 0 and 1;

[0022] Step S203, when the number of key-frame images is greater than M, increase the similarity threshold, and when the number of key-frame images is less than M, decrease the similarity threshold until M key-frame images are obtained.

[0023] Furthermore, the semantic segmentation model is TransUNet, and the sample labels of the training samples for training the semantic segmentation model are obtained through the annotation method of ophthalmology experts.

[0024] Furthermore, the diagnostic model includes: a first hidden layer, a second hidden layer and a first classifier;

[0025] The first hidden layer is used to extract the image features of the sub-regions to which each node of the global graph data is data-connected, and expand the image features into vector representations;

[0026] The second hidden layer inputs the global graph data and outputs comprehensive features;

[0027] The comprehensive features are input into the first classifier, and the classification space of the first classifier represents the diagnosis result of the patient's eye.

[0028] Furthermore, the calculation formula of the first hidden layer includes:

[0029] ;

[0030] ;

[0031] ;

[0032] where 1 ≤ c ≤ C, C represents the number of channels of the key-frame image, assigned as 3, represents the image features output by the first hidden layer, , and represent the original graph, channel score, and global average pooling score of the original graph of the c-th channel of the sub-region associated with the data of each node, and represent the first weight parameter and the first bias parameter respectively, , , and represent the first weight parameter, second weight parameter, first bias parameter, and second bias parameter of the c-th channel of the sub-region associated with the data of each node respectively, represents the pixel value at the i-th row and j-th column of the original graph of the c-th channel of the sub-region associated with the data of each node, and H and W represent the number of rows and columns of the original graph of the sub-region respectively, represents the concatenation operation of the original graphs of C channels, represents element-wise multiplication, sigmoid represents the sigmoid activation function, and ReLU represents the ReLU activation function.

[0033] Furthermore, the calculation formula of the second hidden layer includes:

[0034] ;

[0035] ;

[0036] where 1 ≤ m ≤ M, 1 ≤ u ≤ U, U represents the number of nodes of each key-frame image, Feature represents the comprehensive features output by the second hidden layer, represents the set of nodes that have an edge connection with the u-th node in the m-th key-frame image, and respectively represent the vector representations of the image features of the u-th and v-th nodes of the m-th key frame image, represents the updated feature of the u-th node of the m-th key frame image, 、 and respectively represent the second weight parameter, the third weight parameter, and the fourth weight parameter, represents stacking the updated features of U nodes of M key frame images.

[0037] Furthermore, pre-train the first hidden layer before training the diagnostic model. During the pre-training of the first hidden layer, connect the second classifier, and the classification space of the second classifier represents whether there is bleeding in the anterior chamber.

[0038] Furthermore, obtain the sample labels of the training samples for training the diagnostic model by means of ophthalmologist annotation.

[0039] A method for ophthalmic slit image detection according to the present invention includes the following steps:

[0040] Step S301, collect the examination video of the patient's eye with a slit lamp microscope integrated with a high-definition camera;

[0041] Step S302, extract key frames from the examination video of the patient's eye to obtain an image data set with the number of key frame images being M;

[0042] Step S303, traverse the M key frame images of the image data set, and divide each key frame image into different sub-regions through a semantic segmentation model;

[0043] Step S304, construct global graph data according to the sub-regions of the M key frame images;

[0044] Step S305, construct and train a diagnostic model;

[0045] Step S306, input the global graph data into the trained diagnostic model, and the output value represents the diagnostic result of the patient's eye.

[0046] The present invention provides a storage medium that stores non-temporary computer-readable instructions. When the non-temporary computer-readable instructions are executed by a computer, the steps in the above-mentioned method for ophthalmic slit image detection are executed.

[0047] The beneficial effects of the present invention are as follows: By extracting key frames and image semantic segmentation, the present invention constructs the examination video of the patient's eye into a global graph data representation, and performs aggregation analysis on different sub-regions of each key frame through a diagnostic model, thereby improving the semantic understanding of multiple ophthalmic slit images and the accuracy of the diagnostic result. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 is a schematic diagram of a system for ophthalmic slit image detection according to the present invention;

[0049] Figure 2 is a flowchart for extracting key frames from the examination video of the patient's eye according to the present invention;

[0050] Figure 3 is a flowchart of a method for ophthalmic slit image detection according to the present invention.

[0051] In the figure: video acquisition module 101, key frame extraction module 102, semantic segmentation module 103, global graph data construction module 104, diagnostic model training module 105, diagnostic result output module 106. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0052] Now, the subject matter described herein will be discussed with reference to example embodiments. It should be understood that discussing these embodiments is only to enable those skilled in the art to better understand and thus implement the subject matter described herein. Without departing from the scope of protection of the content of this specification, changes can be made to the functions and arrangements of the elements discussed. Each example can omit, substitute, or add various processes or components as needed. Additionally, the features described in relation to some examples can also be combined in other examples.

[0053] It should be noted that unless otherwise defined, the technical terms or scientific terms used in one or more embodiments of the present invention should have the ordinary meaning understood by those of ordinary skill in the art to which the present invention pertains. The "first", "second", and similar terms used in one or more embodiments of the present invention do not denote any order, quantity, or importance, but are only used to distinguish different components. The terms such as "including" or "comprising" mean that the elements or items appearing before this word cover the elements or items listed after this word and their equivalents, without excluding other elements or items. The terms such as "connected" or "coupled" are not limited to physical or mechanical connections, but can include electrical connections, whether direct or indirect. The terms such as "upper", "lower", "left", "right", etc. are only used to represent relative positional relationships, and when the absolute position of the object being described changes, the relative positional relationship may also change accordingly.

[0054] As Figures 1 to 3 shown, a system for ophthalmic slit image detection includes:

[0055] A video acquisition module 101, which is used to collect the examination video of the patient's eye by integrating a high-definition camera with a slit lamp microscope;

[0056] The key frame extraction module 102 is used to extract key frames from the examination video of the patient's eye to obtain an image dataset with the number of key frame images being M;

[0057] where M is a custom parameter and a positive integer greater than 1;

[0058] The semantic segmentation module 103 is used to traverse the M key frame images of the image dataset and divide each key frame image into different sub-regions through a semantic segmentation model;

[0059] The sub-regions include: eyelids, conjunctiva, sclera, cornea, anterior chamber, iris, pupil, lens, lens capsule, and anterior vitreous;

[0060] The global graph data construction module 104 is used to construct global graph data based on the sub-regions of the M key frame images;

[0061] The global graph data includes: nodes and edges between nodes;

[0062] The nodes of the global graph data establish a data connection with the sub-regions of each key frame image;

[0063] Edges are formed between the nodes corresponding to adjacent sub-regions of the same key frame image;

[0064] Edges are formed between the nodes corresponding to the same sub-region of adjacent key frame images;

[0065] The diagnostic model training module 105 is used to construct and train a diagnostic model;

[0066] The diagnostic result output module 106 is used to input the global graph data into the trained diagnostic model, and the output value represents the diagnostic result of the patient's eye;

[0067] The diagnostic results include: refractive error, conjunctival disease, corneal disease, anterior chamber disease, iris disease, and eye trauma.

[0068] It should be noted that the eyelids include the skin, eyelashes, meibomian glands, etc. of the upper and lower eyelids; the conjunctiva is a transparent mucous membrane covering the inner side of the eyelids and the surface of the eyeball; the sclera is the tough white outer layer of the eyeball; the cornea is the transparent refractive structure at the front of the eyeball, responsible for most of the eye's focusing ability; the anterior chamber is the space filled with aqueous humor between the cornea and the iris; the iris contains the colored circular muscle tissue of the pupil, used to control the size of the pupil; the pupil is the opening in the center of the iris, through which light enters the eye; the lens is located behind the iris and is responsible for further focusing light onto the retina; the lens capsule is the transparent capsule that wraps the lens; the anterior vitreous is the front 1 / 3 part of the transparent gel-like substance inside the eyeball.

[0069] It should be noted that refractive errors can be identified by observing the corneal curvature and lens shape. Refractive errors include myopia, hyperopia, astigmatism, presbyopia, etc.; conjunctival diseases can be conjunctivitis, manifested as symptoms such as conjunctival redness and swelling, and increased secretions; corneal diseases can be diagnosed by observing the transparency, thickness and surface condition of the cornea. Corneal diseases include keratitis, corneal ulcer, corneal scar, corneal degeneration, etc.; anterior chamber diseases can be identified by observing the clarity and cell components of the anterior chamber. Anterior chamber diseases include hypopyon and hyphema, etc.; iris diseases can be diagnosed by observing the color, texture and morphological changes of the iris. Iris diseases include iritis, iris atrophy, etc.; the slit lamp microscope can clearly show the location and degree of eye trauma, such as corneal abrasion and foreign body entry, etc.

[0070] In one embodiment of the present invention, the diagnosis results of the patient's eye can also include lens diseases, eyelid diseases, dry eye, glaucoma, etc.

[0071] It should be noted that lens diseases can be cataracts; eyelid diseases can include meibomian gland dysfunction, blepharitis, trichiasis, ptosis, etc.; dry eye can be diagnosed by observing the wetness of the cornea; although the slit lamp microscope is mainly used to observe the anterior segment structure of the eye, indirect signs indicating glaucoma such as corneal edema and iris abnormalities may also be noticed during the examination process.

[0072] In one embodiment of the present invention, by installing a high-definition camera at the eyepiece of the slit lamp microscope or on the optical path adapter, the examination video of the patient's eye is collected, and the examination video is transmitted to the key frame extraction module for processing through wireless transmission or wired transmission, and can also be transmitted to the display for the doctor to observe. In addition, video file storage can also be performed for the doctor to review or for reexamination, without the patient having to repeat the examination multiple times, improving the patient's satisfaction with the visit.

[0073] In one embodiment of the present invention, as Figure 2 shown, to obtain an image dataset with the number of key frame images being M from the examination video of the patient's eye, the following steps are included:

[0074] Step S201, dividing the examination video of the patient's eye at equal time intervals to obtain N sub-frame images, where N is a custom parameter greater than M;

[0075] Step S202, calculating the similarity value between adjacent two sub-frame images among the N sub-frame images through SSIM (structural similarity index), and if it is determined that the similarity value is less than or equal to the similarity threshold, then the latter sub-frame image is used as the key frame, where the similarity threshold is a custom parameter between 0 and 1. Preferably, the similarity threshold is set to 0.5;

[0076] Step S203: When the number of key-frame images is greater than M, increase the similarity threshold; when the number of key-frame images is less than M, decrease the similarity threshold until M key-frame images are obtained.

[0077] It should be noted that N is set according to the duration of the eye examination video of the patient. Generally, the duration of diagnosing eye diseases through a slit lamp microscope is about 3 to 5 minutes. If the duration is 4 minutes, then 120 sub-frame images can be obtained by frame division at an interval of 2 seconds. Preferably, M is set to 20 and N is set to 100.

[0078] In an embodiment of the present invention, the similarity value can also be calculated as the sum of the absolute values of the differences of each pixel point between two adjacent sub-frame images.

[0079] In an embodiment of the present invention, the semantic segmentation model is TransUNet, and it can also be Mask R-CNN or PSPNet, etc. The sample labels for training the semantic segmentation model are obtained by the annotation of ophthalmology experts.

[0080] In an embodiment of the present invention, the diagnostic model includes: a first hidden layer, a second hidden layer, and a first classifier;

[0081] The first hidden layer is used to extract the image features of the sub-regions associated with the data of each node of the global graph data and expand the image features into vector representations;

[0082] The second hidden layer inputs the global graph data and outputs comprehensive features;

[0083] The comprehensive features are input into the first classifier, and the classification space of the first classifier represents the diagnostic result of the patient's eye.

[0084] In an embodiment of the present invention, the calculation formula of the first hidden layer includes:

[0085] ;

[0086] ;

[0087] ;

[0088] where 1 ≤ c ≤ C, C represents the number of channels of the key-frame image, and is assigned as 3, represents the image features output by the first hidden layer, , and represent the original graph, channel score, and original graph global average pooling score of the c-th channel of the sub-region associated with the data of each node, and respectively represent the first weight parameter and the first bias parameter, , , and respectively represent the first weight parameter, the second weight parameter, the first bias parameter, and the second bias parameter of the c-th channel of the sub-region where each node has data connection, represents the pixel value at the i-th row and j-th column of the c-th channel of the original graph of the sub-region where each node has data connection, and H and W respectively represent the number of rows and columns of the original graph of the sub-region, represents the splicing operation on the original graphs of C channels, represents element-wise multiplication, sigmoid represents the sigmoid activation function, and ReLU represents the ReLU activation function.

[0089] It should be noted that the number of channels of the key frame image is 3, corresponding to the RGB three channels respectively. If the key frame image is grayscale processed, the number of channels of the key frame image is 1.

[0090] In an embodiment of the present invention, the first hidden layer can be constructed based on a convolutional neural network model.

[0091] In an embodiment of the present invention, the calculation formula of the second hidden layer includes:

[0092] ;

[0093] ;

[0094] where 1≤m≤M, 1≤u≤U, U represents the number of nodes of each key frame image, Feature represents the comprehensive feature output by the second hidden layer, represents the set of nodes that have edge connections with the u-th node in the m-th key frame image, and respectively represent the vector representations of the image features of the u-th and v-th nodes of the m-th key frame image, represents the updated feature of the u-th node of the m-th key frame image, , and respectively represent the second weight parameter, the third weight parameter, and the fourth weight parameter, represents the stacking operation on the updated features of U nodes of M key frame images.

[0095] In an embodiment of the present invention, the first hidden layer is pre-trained before the diagnostic model is trained, and a second classifier is connected during the pre-training of the first hidden layer. The classification space of the second classifier represents whether there is bleeding in the anterior chamber.

[0096] In one embodiment of the present invention, the first hidden layer is pre-trained before the diagnostic model is trained, and a third classifier is connected during the pre-training of the first hidden layer. The classification space of the third classifier represents whether the lens is cloudy.

[0097] It should be noted that pre-training the first hidden layer can capture the important features of the key-frame images, which helps to accelerate the training speed of the diagnostic model.

[0098] In one embodiment of the present invention, the sample labels of the training samples for training the diagnostic model are obtained by the annotation of ophthalmic experts.

[0099] It should be noted that the maximum number of iterations during the training of the diagnostic model is set, the difference between the value output by the diagnostic model in each iteration and the sample label of the training sample is specified as the loss function, and the weight parameters and bias parameters of the diagnostic model are updated through backpropagation and gradient descent algorithms until the maximum number of iterations is reached or within a continuous number of iterations, the loss value of the diagnostic model is within the preset range, which indicates that the diagnostic model has reached the convergence state, and then the training is terminated.

[0100] In one embodiment of the present invention, as Figure 3 shown, the present invention provides a method for ophthalmic slit image detection, including the following steps:

[0101] Step S301, collecting an examination video of the patient's eye by integrating a high-definition camera with a slit lamp microscope;

[0102] Step S302, extracting key frames from the examination video of the patient's eye to obtain an image data set with the number of key-frame images being M;

[0103] Step S303, traversing the M key-frame images of the image data set, and dividing each key-frame image into different sub-regions through a semantic segmentation model;

[0104] Step S304, constructing global graph data according to the sub-regions of the M key-frame images;

[0105] Step S305, constructing and training a diagnostic model;

[0106] Step S306, inputting the global graph data into the trained diagnostic model, and the output value represents the diagnostic result of the patient's eye.

[0107] In one embodiment of the present invention, the present invention provides a storage medium that stores non-temporary computer-readable instructions, which can execute a method for ophthalmic slit image detection as described above when executed by a computer.

[0108] The above has described the embodiments of this example, but this example is not limited to the above specific implementation manners. The above specific implementation manners are merely illustrative rather than restrictive. Under the inspiration of this example, those of ordinary skill in the art can also make many forms, all of which fall within the protection scope of this example.

Claims

1. A system for ophthalmic fissure image detection, characterized in that: include: A video acquisition module, which is used to integrate a high-definition camera into a slit lamp microscope to acquire a patient's eye examination video; A key frame extraction module is used to extract key frames from the patient's eye examination video to obtain an image data set with M key frame images; Where M is a custom parameter and is a positive integer greater than 1; A semantic segmentation module is used to traverse the M key frame images of the image dataset and divide each key frame image into different sub-regions through a semantic segmentation model; Subregions include: eyelid, conjunctiva, sclera, cornea, anterior chamber, iris, pupil, lens, lens capsule, and anterior vitreous; A global graph data construction module, which is used to construct global graph data according to the sub-regions of the M key frame images; Global graph data includes: nodes and edges between nodes; The nodes of the global graph data establish data connections with the sub-regions of each key frame image; The nodes corresponding to adjacent sub-regions of the same key frame image form edges; The nodes corresponding to the same sub-region of adjacent key frame images form edges; A diagnostic model training module, which is used to build and train a diagnostic model; A diagnosis result output module, which is used to input the global image data into the trained diagnosis model, and the output value represents the diagnosis result of the patient's eye; Diagnoses included: refractive error, conjunctival disease, corneal disease, anterior chamber disease, iris disease, and ocular trauma; The diagnostic model includes: a first hidden layer, a second hidden layer and a first classifier; The first hidden layer is used to extract the image features of the sub-regions connected to each node of the global graph data, and expand the image features into vector representation; The second hidden layer inputs global graph data and outputs comprehensive features; The comprehensive features are input into a first classifier, and the classification space of the first classifier represents the diagnosis result of the patient's eye; The calculation formula for the first hidden layer includes: ； ； ； Where 1≤c≤C, C represents the number of channels of the key frame image, which is assigned a value of 3. represents the image features output by the first hidden layer, , and Represents the original image, channel score and global average pooling score of the cth channel of the sub-region connected by each node. and represent the first weight parameter and the first bias parameter respectively, , , and They respectively represent the first weight parameter, the second weight parameter, the first bias parameter and the second bias parameter of the cth channel of the sub-region to which each node is connected. represents the pixel value of the i-th row and j-th column of the original image of the c-th channel of the sub-region to which each node is connected. H and W represent the number of rows and columns of the original image of the sub-region, respectively. Indicates that the original images of C channels are spliced ​​together. ReLU represents the ReLU activation function.

2. A system for ophthalmic fissure image detection according to claim 1, characterized in that: Extracting key frames from a patient's eye examination video to obtain an image dataset with M key frame images includes the following steps: Step S201, dividing the patient's eye examination video into N frame images at equal time intervals, where N is a custom parameter greater than M; Step S202, calculating the similarity value of two adjacent sub-frame images among the N sub-frame images by using the structural similarity index, and determining that the similarity value is less than or equal to a similarity threshold, then taking the next sub-frame image as a key frame, wherein the similarity threshold is a custom parameter between 0 and 1; Step S203: when the number of key frame images is greater than M, the similarity threshold is increased; when the number of key frame images is less than M, the similarity threshold is decreased until M key frame images are obtained.

3. A system for ophthalmic fissure image detection according to claim 1, characterized in that: The semantic segmentation model is TransUNet, and the sample labels of the training samples used to train the semantic segmentation model are obtained by annotation by ophthalmologists.

4. A system for ophthalmic fissure image detection according to claim 1, characterized in that: The calculation formula for the second hidden layer includes: ； ； Where 1≤m≤M, 1≤u≤U, U represents the number of nodes in each key frame image, Feature represents the comprehensive features of the output of the second hidden layer, represents the set of nodes that have edges connected to the u-th node in the m-th key frame image, and Respectively represent the vector representation of the image features of the u-th and v-th nodes of the m-th key frame image, represents the updated features of the u-th node of the m-th key frame image, , and represent the second weight parameter, the third weight parameter and the fourth weight parameter respectively, It means stacking the updated features of U nodes of M key frame images.

5. The system for ophthalmic fissure image detection according to claim 1, characterized in that: The first hidden layer is pre-trained before the diagnosis model is trained. During the pre-training of the first hidden layer, the second classifier is connected. The classification space of the second classifier indicates whether there is anterior chamber bleeding.

6. A system for ophthalmic fissure image detection according to claim 1, characterized in that: The sample labels of the training samples used to train the diagnosis model are obtained by annotation by ophthalmologists.

7. A storage medium, characterized in that: The device stores non-transitory computer-readable instructions. When the non-transitory computer-readable instructions are executed by a computer, a method for ophthalmic fissure image detection is performed, comprising the following steps: Step S301, collecting a patient's eye examination video using a high-definition camera integrated into a slit lamp microscope; Step S302, extracting key frames from the patient's eye examination video to obtain an image data set with M key frame images; Where M is a custom parameter and is a positive integer greater than 1; Step S303, traversing the M key frame images of the image data set, and dividing each key frame image into different sub-regions through a semantic segmentation model; Subregions include: eyelid, conjunctiva, sclera, cornea, anterior chamber, iris, pupil, lens, lens capsule, and anterior vitreous; Step S304, constructing global graph data according to the sub-regions of the M key frame images; Global graph data includes: nodes and edges between nodes; The nodes of the global graph data establish data connections with the sub-regions of each key frame image; The nodes corresponding to adjacent sub-regions of the same key frame image form edges; The nodes corresponding to the same sub-region of adjacent key frame images form edges; Step S305, constructing and training a diagnostic model; Step S306, inputting the global image data into the trained diagnostic model, and the output value represents the diagnosis result of the patient's eye; Diagnoses included: refractive error, conjunctival disease, corneal disease, anterior chamber disease, iris disease, and ocular trauma; The diagnostic model includes: a first hidden layer, a second hidden layer and a first classifier; The first hidden layer is used to extract the image features of the sub-regions connected to each node of the global graph data, and expand the image features into vector representation; The second hidden layer inputs global graph data and outputs comprehensive features; The comprehensive features are input into a first classifier, and the classification space of the first classifier represents the diagnosis result of the patient's eye; The calculation formula for the first hidden layer includes: ； ； ； Where 1≤c≤C, C represents the number of channels of the key frame image, which is assigned a value of 3. represents the image features output by the first hidden layer, , and Represents the original image, channel score and global average pooling score of the cth channel of the sub-region connected by each node. and represent the first weight parameter and the first bias parameter respectively, , , and They respectively represent the first weight parameter, the second weight parameter, the first bias parameter and the second bias parameter of the cth channel of the sub-region to which each node is connected. represents the pixel value of the i-th row and j-th column of the original image of the c-th channel of the sub-region to which each node is connected. H and W represent the number of rows and columns of the original image of the sub-region, respectively. Indicates that the original images of C channels are spliced ​​together. ReLU represents the ReLU activation function.

Images