Eye disease recognition method, device, equipment and storage medium based on anterior segment images

Abstract

The present invention discloses an eye disease recognition method, device, equipment and storage medium based on anterior segment images, belonging to the technical field of image processing. The eye disease recognition method of the present invention includes obtaining anterior segment images; performing standardization processing on the anterior segment images to obtain standard anterior segment images; performing segmentation processing on the standard anterior segment images to obtain a plurality of local anterior segment images; extracting pathological features in the local anterior segment images; performing feature importance analysis on each pathological feature to determine target pathological features; and recognizing blinding eye diseases according to the target pathological features and the trained AI analysis model. This eye disease recognition method can realize the simultaneous recognition of multiple anterior segment diseases, and has a relatively high recognition efficiency.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular, to a method, device, equipment, and storage medium for identifying eye diseases based on anterior segment images. Background Art

[0002] Currently, when identifying anterior segment diseases, it is often based on a single anterior segment image to identify one anterior segment disease, and it is often impossible to identify multiple anterior segment diseases simultaneously, resulting in a problem of low identification efficiency. Therefore, how to provide a method for identifying eye diseases based on anterior segment images to achieve the simultaneous identification of multiple anterior segment diseases and improve the identification efficiency has become an urgent problem to be solved. Summary of the Invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. For this purpose, the present invention provides a method for identifying eye diseases based on anterior segment images, which can achieve the simultaneous identification of multiple anterior segment diseases and has a relatively high identification efficiency.

[0004] The present invention also provides an apparatus for identifying eye diseases based on anterior segment images.

[0005] The present invention also provides a device for identifying eye diseases based on anterior segment images.

[0006] The present invention also provides a computer-readable storage medium.

[0007] According to a first aspect embodiment of the present invention, a method for identifying eye diseases based on anterior segment images includes:

[0008] Obtain an anterior segment image;

[0009] Perform normalization processing on the anterior segment image to obtain a standard anterior segment image;

[0010] Perform segmentation processing on the standard anterior segment image to obtain a plurality of local anterior segment images;

[0011] Extract pathological features from the local anterior segment images;

[0012] Perform feature importance analysis on each pathological feature to determine target pathological features;

[0013] Identify blinding eye diseases according to the target pathological features and a trained AI analysis model.

[0014] The eye disease recognition method based on anterior segment images according to an embodiment of the present invention has at least the following beneficial effects: By acquiring anterior segment images, standardizing the anterior segment images to obtain standard anterior segment images. Acquiring anterior segment images can conveniently obtain the entire anterior segment structure information. Segmenting the standard anterior segment images to obtain multiple local anterior segment images and extracting pathological features in the local anterior segment images can conveniently obtain pathological features of various anterior segment diseases. Furthermore, analyzing the importance of each pathological feature to determine the target pathological feature, and recognizing blinding eye diseases based on the target pathological feature and the trained AI analysis model can achieve the simultaneous recognition of multiple anterior segment diseases with relatively high recognition efficiency.

[0015] According to some embodiments of the present invention, the acquiring of the anterior segment images includes:

[0016] Acquiring the anterior segment images and the epidemiological parameters corresponding to the anterior segment images.

[0017] According to some embodiments of the present invention, the standardizing of the anterior segment images to obtain standard anterior segment images includes:

[0018] Screening the anterior segment images to obtain the anterior segment images after screening;

[0019] Performing annotation processing on the anterior segment images after screening according to the cross-validation method to obtain standard anterior segment images.

[0020] According to some embodiments of the present invention, the segmenting of the standard anterior segment images to obtain multiple local anterior segment images includes:

[0021] Obtaining the three-dimensional structure parameters of the standard anterior segment images;

[0022] Performing segmentation processing on the standard anterior segment images according to the three-dimensional structure parameters and the trained multi-object segmentation model to obtain multiple local anterior segment images.

[0023] According to some embodiments of the present invention, the analyzing the importance of each pathological feature to determine the target pathological feature includes:

[0024] Analyzing the importance of each pathological feature according to at least one of the Pearson correlation coefficient method, the bilateral multivariate method, and the feature iterative deletion method to determine the target pathological feature.

[0025] According to some embodiments of the present invention, the recognizing of blinding eye diseases based on the target pathological feature and the trained AI analysis model includes:

[0026] Construct a multi-view convolution module according to the graph convolution and the target pathological features;

[0027] Obtain an AI analysis model according to the multi-view convolution module and the multi-region selection attention mechanism module;

[0028] Identify blinding eye diseases according to the trained AI analysis model.

[0029] According to some embodiments of the present invention, the eye disease identification method further includes:

[0030] Identify the anterior segment image according to the artificial intelligence algorithm to obtain identification data;

[0031] Configure a corresponding reference file according to the identification data;

[0032] Optimize the reference file according to the obtained feedback data to obtain the final reference file.

[0033] An eye disease identification device based on an anterior segment image according to an embodiment of the second aspect of the present invention includes:

[0034] An anterior segment image acquisition module for acquiring an anterior segment image;

[0035] A normalization processing module for normalizing the anterior segment image to obtain a standard anterior segment image;

[0036] An image segmentation module for segmenting the standard anterior segment image to obtain a plurality of local anterior segment images;

[0037] A feature extraction module for extracting pathological features from the local anterior segment images;

[0038] A feature analysis module for analyzing the importance of each pathological feature to determine the target pathological feature;

[0039] An identification module for identifying blinding eye diseases according to the target pathological features and the trained AI analysis model.

[0040] The eye disease recognition device based on anterior segment images according to an embodiment of the present invention has at least the following beneficial effects: Such an eye disease recognition device obtains anterior segment images through an anterior segment image acquisition module, and a standardization processing module performs standardization processing on the anterior segment images to obtain standard anterior segment images. By obtaining anterior segment images, it is possible to conveniently obtain the entire anterior segment structure information. An image segmentation module performs segmentation processing on the standard anterior segment images to obtain multiple local anterior segment images, and a feature extraction module extracts pathological features in the local anterior segment images, making it possible to conveniently obtain the pathological features of various anterior segment diseases. Furthermore, a feature analysis module performs feature importance analysis on each pathological feature to determine target pathological features, and an identification module identifies blinding eye diseases based on the target pathological features and a trained AI analysis model, capable of simultaneously identifying multiple anterior segment diseases with relatively high identification efficiency.

[0041] An eye disease recognition device based on anterior segment images according to a third aspect embodiment of the present invention includes:

[0042] At least one processor, and,

[0043] A memory communicatively connected to the at least one processor; wherein,

[0044] The memory stores instructions, and when the instructions are executed by the at least one processor, the at least one processor implements the eye disease recognition method as described in the first aspect embodiment when executing the instructions.

[0045] The eye disease recognition device based on anterior segment images according to an embodiment of the present invention has at least the following beneficial effects: Such an electronic device adopts the above-mentioned eye disease recognition method based on anterior segment images. By obtaining anterior segment images, performing standardization processing on the anterior segment images to obtain standard anterior segment images. By obtaining anterior segment images, it is possible to conveniently obtain the entire anterior segment structure information. Performing segmentation processing on the standard anterior segment images to obtain multiple local anterior segment images, and extracting pathological features in the local anterior segment images, making it possible to conveniently obtain the pathological features of various anterior segment diseases. Furthermore, performing feature importance analysis on each pathological feature to determine target pathological features, and identifying blinding eye diseases based on the target pathological features and a trained AI analysis model, capable of simultaneously identifying multiple anterior segment diseases with relatively high identification efficiency.

[0046] A computer-readable storage medium according to a fourth aspect embodiment of the present invention stores computer-executable instructions, and the computer-executable instructions are used to cause a computer to execute the eye disease recognition method based on anterior segment images as described in the first aspect embodiment.

[0047] The computer-readable storage medium according to an embodiment of the present invention has at least the following beneficial effects: By acquiring an anterior segment image and performing normalization processing on the anterior segment image, a standard anterior segment image is obtained when this computer-readable storage medium executes the above-mentioned eye disease recognition method. The entire anterior segment structure information can be conveniently obtained by acquiring the anterior segment image. By performing segmentation processing on the standard anterior segment image to obtain multiple local anterior segment images and extracting pathological features in the local anterior segment images, the pathological features of various anterior segment diseases can be conveniently obtained. Furthermore, by performing feature importance analysis on each pathological feature to determine the target pathological feature, and based on the target pathological feature and the trained AI analysis model, the blinding eye diseases are recognized, and the simultaneous recognition of multiple anterior segment diseases can be achieved, with relatively high recognition efficiency.

[0048] Additional aspects and advantages of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] The following further describes the present invention with reference to the drawings and embodiments, where:

[0050] Figure 1 is a flowchart of the eye disease recognition method based on anterior segment images according to an embodiment of the present invention;

[0051] Figure 2 is Figure 1 a flowchart of step S200 in

[0052] Figure 3 is Figure 1 a flowchart of step S300 in

[0053] Figure 4 is Figure 1 a flowchart of step S600 in

[0054] Figure 5 is Figure 1 another flowchart of the eye disease recognition method based on anterior segment images of

[0055] Figure 6 is a schematic structural diagram of the eye disease recognition device based on anterior segment images according to an embodiment of the present invention.

[0056] Reference numerals: 610, anterior segment image acquisition module; 620, normalization processing module; 630, image segmentation module; 640, feature extraction module; 650, feature analysis module; 660, recognition module. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0057] Embodiments of the present invention will be described in detail below. Examples of the embodiments are shown in the accompanying drawings, where like or similar reference numerals denote like or similar elements or elements having like or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary only for explaining the present invention and should not be construed as limiting the present invention.

[0058] In the description of the present invention, it should be understood that with respect to the orientation description, such as the orientation or positional relationship indicated by up, down, front, back, left, right, etc., is based on the orientation or positional relationship shown in the accompanying drawings. It is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and thus should not be construed as limiting the present invention.

[0059] In the description of the present invention, the meaning of several is more than one, and the meaning of multiple is more than two. Understandings such as greater than, less than, exceeding, etc. do not include the present number, and understandings such as above, below, within, etc. include the present number. If there is a description of first and second, it is only for the purpose of distinguishing technical features and should not be construed as indicating or implying relative importance or implicitly indicating the quantity of the indicated technical features or implicitly indicating the sequence relationship of the indicated technical features.

[0060] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installation, connection, etc. should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific content of the technical solution.

[0061] In the description of the present invention, the description referring to terms such as "one embodiment", "some embodiments", "schematic embodiments", "examples", "specific examples", or "some examples", etc. means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0062] In a first aspect, referring to Figure 1 , the method for identifying eye diseases based on anterior segment images according to an embodiment of the present invention includes:

[0063] S100, obtaining an anterior segment image;

[0064] S200, performing normalization processing on the anterior segment image to obtain a standard anterior segment image;

[0065] S300, performing segmentation processing on the standard anterior segment image to obtain a plurality of local anterior segment images;

[0066] S400, extract the pathological features in the local anterior segment images;

[0067] S500, perform feature importance analysis on each pathological feature to determine the target pathological feature;

[0068] S600, identify blinding eye diseases based on the target pathological feature and the trained AI analysis model.

[0069] In the process of identifying anterior segment diseases, first obtain anterior segment images. It should be noted that these anterior segment images are a type of AS-OCT three-dimensional image. This three-dimensional image has the characteristics of non-contact three-dimensional imaging, high detection sensitivity, high resolution, rapid examination, objective quantitative measurement, etc. The corresponding entire anterior segment structure information can be obtained through the acquired anterior segment images. Such anterior segment images can be collected by an AS-OCT device and obtained from multiple angles through a scanning mode, or it can be in other ways, not limited to this. Furthermore, perform standardization processing on the anterior segment images to obtain standard anterior segment images. Specifically, the acquired anterior segment images can be screened to eliminate abnormal images, and then the screened anterior segment images can be labeled through methods such as cross-validation to obtain standard anterior segment images. Since multiple anterior segment diseases need to be identified by analyzing different parts of the anterior segment structure respectively, the standard anterior segment images are segmented to obtain multiple local anterior segment images. Analyze the anterior segment diseases based on these local anterior segment images, extract the pathological features in the local anterior segment images, perform feature importance analysis and feature correlation analysis on different types of pathological features, and determine the target pathological feature. Identify blinding eye diseases based on the target pathological feature and the trained AI analysis model. Among them, the trained AI analysis model is based on a hybrid model of two-dimensional convolutional neural network and three-dimensional deep neural network, and also includes a multi-view graph convolutional module and a multi-region selective attention mechanism module, which can improve the comprehensiveness and accuracy of ophthalmic disease identification. This identification method can achieve the simultaneous identification of multiple anterior segment diseases with relatively high identification efficiency.

[0070] In some embodiments, step S100 includes:

[0071] Obtain the anterior segment images and the corresponding epidemiological investigation parameters of the anterior segment images.

[0072] In the process of identifying anterior segment diseases, anterior segment images need to be acquired. These anterior segment images are a type of AS-OCT three-dimensional image. The corresponding entire anterior segment structure information can be obtained from the acquired anterior segment images. Such anterior segment images can be collected by an AS-OCT device and acquired from multiple angles through a scanning mode. To improve the recognition accuracy, it is also necessary to obtain the epidemiological parameters corresponding to each anterior segment image. Among them, the epidemiological parameters include age, gender, uncorrected visual acuity, etc. These epidemiological parameters can help analyze and identify different ophthalmic diseases and improve the analysis accuracy.

[0073] Referring to Figure 2 , in some embodiments, step S200 includes:

[0074] S210, screening the anterior segment images to obtain the screened anterior segment images;

[0075] S220, performing annotation processing on the screened anterior segment images according to the cross-validation method to obtain standard anterior segment images.

[0076] In the process of identifying anterior segment diseases, anterior segment images are first acquired. After the anterior segment images are acquired, to avoid the interference of abnormal data, it is necessary to screen the anterior segment images and eliminate the abnormal anterior segment images. To improve the image annotation quality, it is also necessary to perform annotation processing on the screened anterior segment images according to the cross-validation method to obtain standard anterior segment images. It should be noted that the cross-validation method here refers to performing multiple annotation processes on each region of the screened anterior segment images, and then forming the final annotation for each region by synthesizing multiple annotations within the region. Through this annotation method, the annotation quality of the images can be improved and the data accuracy can be enhanced.

[0077] Referring to Figure 3 , in some embodiments, step S300 includes:

[0078] S310, obtaining the three-dimensional structure parameters of the standard anterior segment images;

[0079] S320, performing segmentation processing on the standard anterior segment images according to the three-dimensional structure parameters and the trained multi-object segmentation model to obtain multiple local anterior segment images.

[0080] After screening and annotating the acquired anterior segment images to obtain standard anterior segment images, to improve the accuracy of anterior segment structure segmentation, it is necessary to obtain the three-dimensional structure parameters of the standard anterior segment images, and perform segmentation and detection on the standard anterior segment images according to the three-dimensional structure parameters and the trained multi-object segmentation model to obtain multiple local anterior segment images. Specifically, the multi-object segmentation model is mainly an end-to-end multi-object segmentation model based on multi-information interaction. The convolutional layers used in the segmentation model can be expressed as:

[0081]

[0082] Among them, H cij represents the value at position c in the feature map i in the hidden layer generated by the input sample j, and W k,m,i represents the weight value at position k of the convolution kernel from the m channels of the input sample to the feature map i in the hidden layer. represents the pixel value corresponding to the data of the m channels of the input sample j at the position d°c + k. d represents the stride vector of the downsampling convolution, determining how many pixels to sample at intervals, and ° represents the element-wise product.

[0083] The cross-entropy loss function used by the multi-object segmentation model for segmenting the standard anterior segment image is

[0084]

[0085] Among them, n is the number of categories, y represents the true annotation, represents the segmentation prediction result.

[0086] The MES loss function used by the multi-object segmentation model for detecting the standard anterior segment image is

[0087]

[0088] Among them, y i and represent the label value and prediction value of the model, and n represents the number of regression coordinate points.

[0089] In some embodiments, step S500 includes:

[0090] Performing feature importance analysis on each pathological feature according to at least one of the Pearson correlation coefficient method, the bilateral multivariate method, and the feature iterative deletion method to determine the target pathological feature.

[0091] In order to make full use of the correlations between different regions and different features related to multiple diseases in the anterior segment image, after segmenting the standard anterior segment image to obtain multiple local anterior segment images, different types of pathological features can be extracted using methods such as radiomics. Performing feature importance analysis on each pathological feature according to at least one of the Pearson correlation coefficient method, the bilateral multivariate method, and the feature iterative deletion method, sorting the feature importance, and analyzing each pathological feature to determine the target pathological feature. It should be noted that when performing feature importance analysis on each pathological feature using the Pearson correlation coefficient method, the Pearson correlation coefficient method can be expressed as

[0092]

[0093] Among them, E[(X - μ X )(Y - μ Y )] represents the covariance of X and Y, and σ X , σ Y are the standard deviations of X and Y respectively.

[0094] Since the Pearson correlation coefficient method has a fast calculation speed, is suitable for analyzing a large amount of data, and can improve the recognition efficiency.

[0095] In addition, since the Pearson coefficient is only sensitive to linear relationships, for non-linear models, a feature iterative deletion method can be used for feature analysis. Through the model training samples, then score each pathological feature and sort them, remove the pathological feature with the smallest feature score, and then use the remaining pathological features to train the model again for the next iteration. Finally, select the required pathological features and use these pathological features as the target pathological features.

[0096] Referring to Figure 4 , in some embodiments, step S600 includes:

[0097] S610, construct a multi-view convolutional module according to graph convolution and target pathological features;

[0098] S620, obtain an AI analysis model according to the multi-view convolutional module and the multi-region selective attention mechanism module;

[0099] S630, identify blinding eye diseases according to the trained AI analysis model.

[0100] In order to make full use of the various disease pathological features and the spatial structure information of the anterior segment image, a hybrid model of two-dimensional convolutional neural network and three-dimensional deep neural network can be used as the basic network framework of the AI analysis model. Use graph convolution to process the target pathological features to construct the correlation between anterior segment images at different angles, forming a multi-view convolutional module. At the same time, in order to characterize the differences between different regions in the anterior segment image, a multi-region selective attention mechanism module can also be constructed. By integrating the multi-view convolutional module and the multi-region selective attention mechanism module into the hybrid model, an AI analysis model is obtained. Through this multi-view multi-disease AI analysis model, blinding eye diseases can be conveniently identified, improving the comprehensiveness of ophthalmic disease recognition and the accuracy of analysis.

[0101] Referring to Figure 5 , in some embodiments, the eye disease recognition method further includes:

[0102] S700, identify the anterior segment image according to the artificial intelligence algorithm to obtain identification data;

[0103] S800. Configure the corresponding reference file according to the recognition data;

[0104] S900. Optimize the reference file according to the obtained feedback data to obtain the final reference file.

[0105] In the process of identifying anterior segment diseases, the anterior segment image can be identified according to the artificial intelligence algorithm to obtain the recognition data. Specifically, different disease analysis algorithms and calculation service functions can be called according to different received commands, which can improve the recognition efficiency. By obtaining the recognition data, the corresponding reference file can be queried and configured in the resource library according to the recognition data. For example, the disease analysis materials and treatment plans stored in the resource library, etc. In order to provide a more matching solution for different anterior segment diseases, the feedback data can also be obtained through the interactive feedback system, and the reference file can be optimized according to the feedback data to generate the final reference file. It should be noted that the interactive feedback system can control the output of the feedback data according to principles such as rule-driven and data-driven. This method can improve the accuracy of the feedback data, so that the generated reference file can more accurately assist in the analysis of the corresponding anterior segment diseases and improve the accuracy of disease recognition and analysis.

[0106] In a second aspect, referring to Figure 6 , the eye disease recognition device based on the anterior segment image according to the embodiment of the present invention includes:

[0107] An anterior segment image acquisition module 610, configured to acquire an anterior segment image;

[0108] A normalization processing module 620, configured to perform normalization processing on the anterior segment image to obtain a standard anterior segment image;

[0109] An image segmentation module 630, configured to perform segmentation processing on the standard anterior segment image to obtain a plurality of local anterior segment images;

[0110] A feature extraction module 640, configured to extract pathological features in the local anterior segment image;

[0111] A feature analysis module 650, configured to perform feature importance analysis on each pathological feature to determine the target pathological feature;

[0112] A recognition module 660, configured to recognize blinding eye diseases according to the target pathological features and the trained AI analysis model.

[0113] In the process of identifying anterior segment diseases, the anterior segment image acquisition module 610 first acquires anterior segment images. The corresponding entire anterior segment structure information can be obtained from the acquired anterior segment images. Furthermore, the normalization processing module 620 performs normalization processing on the anterior segment images to obtain standard anterior segment images. Specifically, the acquired anterior segment images can be screened to eliminate abnormal images, and then the screened anterior segment images can be labeled through cross-checking and other methods to obtain standard anterior segment images. Since different parts of the anterior segment structure need to be analyzed separately for the identification of multiple anterior segment diseases, the image segmentation module 630 performs segmentation processing on the standard anterior segment images to obtain multiple local anterior segment images. Based on these local anterior segment images, the anterior segment diseases are analyzed. The feature extraction module 640 extracts the pathological features in the local anterior segment images, and the feature analysis module 650 performs feature importance analysis and feature correlation analysis on different types of pathological features to determine the target pathological features. The identification module 660 identifies blinding eye diseases based on the target pathological features and the trained AI analysis model. The trained AI analysis model is based on a hybrid model of two-dimensional convolutional neural network and three-dimensional deep neural network, and also includes a multi-view graph convolutional module and a multi-region selective attention mechanism module, which can improve the comprehensiveness and accuracy of ophthalmic disease identification. This identification method can achieve the simultaneous identification of multiple anterior segment diseases with relatively high identification efficiency.

[0114] In a third aspect, the ophthalmic disease identification device according to an embodiment of the present invention includes at least one processor and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions, and the instructions are executed by the at least one processor so that when the at least one processor executes the instructions, the ophthalmic disease identification method as in the embodiment of the first aspect is implemented.

[0115] The ophthalmic disease identification device according to an embodiment of the present invention has at least the following beneficial effects: This electronic device adopts the above-mentioned ophthalmic disease identification method. By acquiring anterior segment images and performing normalization processing on the anterior segment images, standard anterior segment images are obtained. The entire anterior segment structure information can be conveniently obtained by acquiring anterior segment images. By performing segmentation processing on the standard anterior segment images to obtain multiple local anterior segment images and extracting the pathological features in the local anterior segment images, the pathological features of multiple anterior segment diseases can be conveniently obtained. Furthermore, by performing feature importance analysis on each pathological feature to determine the target pathological features and identifying blinding eye diseases based on the target pathological features and the trained AI analysis model, the simultaneous identification of multiple anterior segment diseases can be achieved with relatively high identification efficiency.

[0116] Fourthly, the present invention also provides a computer-readable storage medium. The computer-readable storage medium stores computer-executable instructions for causing a computer to execute the eye disease recognition method based on anterior segment images according to the embodiments of the first aspect.

[0117] The computer-readable storage medium according to the embodiments of the present invention has at least the following beneficial effects: By executing the above-mentioned eye disease recognition method, this computer-readable storage medium can obtain anterior segment images, perform normalization processing on the anterior segment images to obtain standard anterior segment images. Obtaining anterior segment images can facilitate the acquisition of the entire anterior segment structure information. By performing segmentation processing on the standard anterior segment images to obtain multiple local anterior segment images and extracting pathological features in the local anterior segment images, the pathological features of various anterior segment diseases can be conveniently obtained. Furthermore, by performing feature importance analysis on each pathological feature to determine the target pathological feature, and according to the target pathological feature and the trained AI analysis model, blinding eye diseases can be recognized, enabling the simultaneous recognition of multiple anterior segment diseases with relatively high recognition efficiency.

[0118] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited to the above embodiments. Various changes can be made without departing from the spirit of the present invention within the scope of knowledge possessed by those of ordinary skill in the art. In addition, the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

Claims

1. An eye disease recognition method based on anterior segment images, characterized in that, Including: Obtain anterior segment images; Perform normalization processing on the anterior segment images to obtain standard anterior segment images; Obtain three-dimensional structure parameters of the standard anterior segment images; Perform segmentation processing on the standard anterior segment images according to the three-dimensional structure parameters and a trained multi-object segmentation model to obtain multiple local anterior segment images, where the multiple local anterior segment images correspond to different parts of the anterior segment structure; Extract pathological features from the local anterior segment images; Perform feature importance analysis on each of the pathological features to determine target pathological features; Use graph convolution to process the target pathological features to construct the correlation between anterior segment images at different angles, forming a multi-view convolution module; Integrate the multi-view convolution module and a multi-region selective attention mechanism module into a hybrid model using two-dimensional convolutional neural network and three-dimensional deep neural network to obtain an AI analysis model, where the multi-region selective attention mechanism module is used to represent the differences between different regions in the anterior segment images; Identify blinding eye diseases according to the trained AI analysis model.

2. The eye disease recognition method according to claim 1, characterized in that The obtaining of the anterior segment images includes: Obtain anterior segment images and epidemiological parameters corresponding to the anterior segment images.

3. The eye disease recognition method according to claim 1, wherein The performing of normalization processing on the anterior segment images to obtain standard anterior segment images includes: Screen the anterior segment images to obtain the anterior segment images after screening; Perform annotation processing on the anterior segment images after screening according to the cross-validation method to obtain standard anterior segment images; Wherein, the cross-validation method refers to performing multiple annotation processes on each region of the anterior segment images after screening, and then synthesizing the multiple annotations within the region for each region to form a final annotation.

4. The eye disease recognition method according to claim 1, wherein The performing of feature importance analysis on each of the pathological features to determine target pathological features includes: Perform feature importance analysis on each of the pathological features according to at least one of the Pearson correlation coefficient method, the bilateral multivariate method, and the feature iterative deletion method to determine target pathological features.

5. The eye disease recognition method according to any one of claims 1 to 4, characterized in that, The eye disease identification method further includes: Identify the anterior segment images according to an artificial intelligence algorithm to obtain identification data; Configure corresponding reference files according to the identification data; Optimize the reference files according to the obtained feedback data to obtain final reference files.

6. An eye disease recognition device based on anterior segment images, characterized in that Including: An anterior segment image acquisition module for obtaining anterior segment images; A normalization processing module for performing normalization processing on the anterior segment images to obtain standard anterior segment images; An image segmentation module for obtaining three-dimensional structure parameters of the standard anterior segment images; performing segmentation processing on the standard anterior segment images according to the three-dimensional structure parameters and a trained multi-object segmentation model to obtain multiple local anterior segment images, where the multiple local anterior segment images correspond to different parts of the anterior segment structure; A feature extraction module for extracting pathological features from the local anterior segment images; A feature analysis module for performing feature importance analysis on each of the pathological features to determine target pathological features; An identification module, configured to process the target pathological features using graph convolution to construct the correlation between anterior segment images from different angles, forming a multi-view convolution module; integrating the multi-view convolution module and the multi-region selective attention mechanism module into a hybrid model adopting a two-dimensional convolutional neural network and a three-dimensional deep neural network to obtain an AI analysis model, where the multi-region selective attention mechanism module is used to characterize the differences between different regions in the anterior segment image; identifying blinding eye diseases according to the trained AI analysis model.

7. An ophthalmopathy recognition device based on anterior segment images, characterized in that, Comprising: at least one processor, and, a memory communicatively connected to the at least one processor; wherein, the memory stores instructions, and the instructions are executed by the at least one processor so that when the at least one processor executes the instructions, the method for identifying eye diseases based on anterior segment images according to any one of claims 1 to 5 is implemented.

8. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer-executable instructions, and the computer-executable instructions are used to cause a computer to execute the method for identifying eye diseases based on anterior segment images according to any one of claims 1 to 5.

Images