Artificial intelligence-based myopia risk prediction method, apparatus and device, and medium

Abstract

The invention provides a myopia risk prediction method and device based on artificial intelligence, equipment and a medium, and belongs to the technical field of artificial intelligence. The method comprises the following steps: acquiring fundus image data of a to-be-predicted individual; segmenting the fundus image data; extracting biomechanical characteristics of the sclera at the rear part of the eye of the individual to be predicted from the segmented fundus image data; and inputting the biomechanical characteristics into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result of the to-be-predicted individual. By adopting the technical scheme of the invention, the biomechanical characteristics of the posterior eye sclera of the individual can be non-invasively evaluated, so that the myopia risk of the individual can be accurately predicted based on the biomechanical characteristics of the posterior eye sclera.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to an artificial intelligence-based myopia risk prediction method, device, equipment and medium. Background Art

[0002] Pathological myopia not only affects the quality of vision, but also significantly increases the risk of fundus lesions (such as retinal detachment, choroidal neovascularization, macular degeneration and posterior scleral staphyloma), and even leads to blindness. The earlier the age of onset of myopia, the greater the possibility of developing pathological myopia.

[0003] Therefore, early prevention and intervention of myopia is crucial to reduce the incidence of pathological myopia. However, the relevant technical intervention methods and early warning mechanisms for myopia mainly rely on routine ophthalmic examinations, such as refractive power measurement and axial length measurement, etc. These examinations can only indirectly assess the progression of myopia and cannot accurately predict the risk of myopia in individuals. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to propose an artificial intelligence-based myopia risk prediction method, device, equipment and medium, which non-invasively evaluates the biomechanical properties of the sclera at the back of the eye of an individual, so as to accurately predict the myopia risk of an individual based on the biomechanical properties of the sclera at the back of the eye.

[0005] To achieve the above-mentioned purpose, the first aspect of the embodiment of the present application proposes a myopia risk prediction method based on artificial intelligence, the method comprising:

[0006] Obtain fundus image data of the individual to be predicted;

[0007] Segmenting the fundus image data;

[0008] Extracting the biomechanical characteristics of the posterior sclera of the individual to be predicted from the segmented fundus image data;

[0009] The biomechanical characteristics are input into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result of the individual to be predicted.

[0010] In some embodiments, before inputting the biomechanical features into a pre-trained myopia risk prediction model for training, the method further comprises:

[0011] Acquiring sample image data of the back of the eye;

[0012] Constructing an algorithm training data set based on the sample image data;

[0013] A machine learning classification algorithm is trained based on the algorithm training data set and preset compliance labels to obtain a myopia risk prediction model.

[0014] In some embodiments, the algorithm training data set includes high compliance training samples and low compliance training samples;

[0015] The step of constructing an algorithm training data set based on the sample image data includes:

[0016] Registering the sample image data based on a preset three-dimensional registration algorithm;

[0017] Performing strain extraction processing on the registered sample image data to obtain the strain magnitude of the registered sample image data;

[0018] When the strain magnitude is greater than or equal to a preset strain threshold, dividing the registered sample image data into the high-compliance training sample;

[0019] When the strain magnitude is smaller than a preset strain threshold, the registered sample image data is divided into the low-compliance training sample.

[0020] In some embodiments, the sample image data includes: a baseline image and a mechanical loading image;

[0021] The registering the sample image data based on a preset three-dimensional registration algorithm includes:

[0022] Performing an affine transformation on the baseline image and the mechanical loading image based on a preset three-dimensional registration algorithm to obtain an affine-transformed baseline image and an affine-transformed mechanical loading image;

[0023] Based on the three-dimensional registration algorithm, the baseline image after the affine transformation and the mechanical loading image after the affine transformation are elastically transformed.

[0024] In some embodiments, the step of performing strain extraction processing on the registered sample image data to obtain the strain magnitude of the registered sample image data includes:

[0025] Calculate the displacement field of each pixel in the registered sample image data;

[0026] Calculating the partial derivative of the displacement field to obtain the three-dimensional strain tensor of the sclera at the back of the eye in the sample image data;

[0027] The norm of the three-dimensional strain tensor is calculated to obtain the strain magnitude of the registered sample image data.

[0028] In some embodiments, the compliance tags include: a high compliance tag and a low compliance tag;

[0029] The method of training a machine learning classification algorithm based on the algorithm training data set and a preset compliance label to obtain a myopia risk prediction model includes:

[0030] Extracting a first biomechanical feature of the posterior sclera of the eye from the high-compliance training sample by using a preset feature extraction model;

[0031] Extracting a second biomechanical feature of the posterior sclera of the eye from the low-compliance training sample by using a preset feature extraction model;

[0032] The first biomechanical feature and the high compliance label are input into a machine learning classification algorithm to train the machine learning classification algorithm, and the second biomechanical feature and the low compliance label are input into the machine learning classification algorithm to train the machine learning classification algorithm to obtain a myopia risk prediction model.

[0033] To achieve the above-mentioned purpose, a second aspect of the embodiment of the present application proposes a myopia risk prediction device based on artificial intelligence, the device comprising:

[0034] An acquisition module, used for acquiring fundus image data of the individual to be predicted;

[0035] A feature extraction module, used for segmenting the fundus image data, and extracting the biomechanical features of the posterior sclera of the individual to be predicted from the segmented fundus image data;

[0036] The risk prediction module is used to input the biomechanical characteristics into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result for the individual to be predicted.

[0037] To achieve the above-mentioned purpose, the third aspect of an embodiment of the present application proposes a computer device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the method described in the first aspect when executing the computer program.

[0038] To achieve the above-mentioned purpose, the fourth aspect of an embodiment of the present application proposes a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method described in the first aspect is implemented.

[0039] To achieve the above-mentioned purpose, the fifth aspect of the embodiments of the present application proposes a computer program product, which stores a computer program. When the computer program is executed by a processor, it implements the method described in the first aspect.

[0040] The artificial intelligence-based myopia risk prediction method and device, computer equipment and storage medium proposed in the present application first obtain the fundus image data of the individual to be predicted, then segment the fundus image data, and extract the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted from the segmented fundus image data, and finally input the biomechanical characteristics into a pre-trained myopia risk prediction model for training, thereby obtaining the myopia risk prediction result of the individual to be predicted.

[0041] Thus, compared to the method of indirectly evaluating myopia progression only by measuring diopter and axial length, the present application trains a machine learning model based on pre-evaluating the biomechanical characteristics of the sclera at the back of the eye of an individual, and obtains a myopia risk prediction model. Therefore, for the individual to be predicted, it is only necessary to non-invasively obtain the fundus image data of the individual to be predicted, and extract the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted based on the fundus image data. The myopia risk prediction model can be used to train the biomechanical characteristics to obtain the myopia risk prediction result of the individual to be predicted. In other words, the present application realizes the non-invasive evaluation of the biomechanical characteristics of the sclera at the back of the eye of an individual, thereby accurately predicting the myopia risk of an individual based on the biomechanical characteristics of the sclera at the back of the eye. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 A schematic diagram of the steps of the myopia risk prediction method based on artificial intelligence provided in some embodiments of the present application;

[0043] Figure 2A A baseline image involved in the artificial intelligence-based myopia risk prediction method provided in the embodiment of the present application;

[0044] Figure 2B A mechanical loading image involved in the artificial intelligence-based myopia risk prediction method provided in an embodiment of the present application;

[0045] Figure 3 A schematic diagram of image segmentation results of fundus image data involved in the artificial intelligence-based myopia risk prediction method provided in an embodiment of the present application;

[0046] Figure 4 A schematic diagram of the steps of the artificial intelligence-based myopia risk prediction method in other embodiments provided in the embodiments of the present application;

[0047] Figure 5 for Figure 4 Schematic diagram of detailed process flow of step S402;

[0048] Figure 6 A schematic diagram of a displacement field visualization result involved in the artificial intelligence-based myopia risk prediction method provided in an embodiment of the present application;

[0049] Figure 7 A schematic diagram of the system framework of a terminal device involved in the artificial intelligence-based myopia risk prediction method provided in an embodiment of the present application;

[0050] Figure 8 A schematic diagram of the structure of a myopia risk prediction device based on artificial intelligence provided in an embodiment of the present application;

[0051] Fig. 9 A schematic diagram of the hardware structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0052] In order to make the purpose, technical solution and advantages of the present application more clearly understood, the present application is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not used to limit the present application.

[0053] It should be noted that, although the functional modules are divided in the device schematic diagram and the logical order is shown in the flowchart, in some cases, the steps shown or described may be performed in a different order than the module division in the device or the order in the flowchart. The terms "first", "second", etc. in the specification, claims and the above drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0054] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art to which this application belongs. The terms used herein are only for the purpose of describing the embodiments of this application and are not intended to limit this application.

[0055] First, some nouns involved in this application are analyzed:

[0056] Artificial intelligence (AI) is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new type of intelligent machine that can respond in a similar way to human intelligence. Research in this field includes robotics, speech recognition, image recognition, natural language processing and expert systems. AI can simulate the information process of human consciousness and thinking. AI is also a theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0057] Optical Coherence Tomography (OCT): It is a new type of tomographic imaging technology that has developed rapidly in recent years and has the most promising application prospects, especially in the detection and imaging of biological tissues in vivo. It has been tried in the clinical diagnosis of ophthalmology, dentistry and dermatology. OCT is a non-contact, high-resolution tomography and biological microscope imaging technology. It can be used for in vivo viewing, axial sectioning and measurement of the posterior segment of the eye (including the retina, retinal nerve fiber layer, macula and optic disc). It is a diagnostic method that is particularly used to help detect and manage eye diseases (including but not limited to macular holes, cystoid macular edema, diabetic retinopathy, age-related macular degeneration and glaucoma).

[0058] Next, the overall concept of the embodiments of the present application is described.

[0059] Myopia, especially high myopia and pathological myopia, has become a prominent problem in global public health. According to the World Health Organization, there are about 2.8 billion people suffering from myopia in the world, of which more than 100 million are high myopia. With the increasing incidence of myopia, it is estimated that by 2050, the number of myopic patients in the world will reach 5 billion, and the number of patients with pathological myopia will reach 900 million.

[0060] Pathological myopia not only affects the quality of vision, but also significantly increases the risk of fundus lesions, such as retinal detachment, choroidal neovascularization, macular degeneration and posterior scleral staphyloma, and may even lead to blindness.

[0061] The earlier the age of onset of myopia, the greater the possibility of developing high myopia or pathological myopia. Therefore, early prevention and intervention are crucial to reducing the incidence of pathological myopia. However, existing intervention methods and early warning mechanisms still have limitations, because existing interventions or warnings mainly rely on routine ophthalmic examinations (such as refractive power measurement and axial length measurement), but these routine examinations can only be used to indirectly assess the progression of myopia, and cannot accurately predict an individual's myopia risk, especially it is impossible to fully understand the biomechanical properties of the sclera at the back of the eye.

[0062] In recent years, studies have shown that the biomechanical properties of the sclera at the back of the eye, especially the strain response of the sclera under intraocular pressure, are closely related to the development of myopia. As the outer membrane of the eyeball, the sclera is elastic and tough and can resist the effects of intraocular pressure. When the biomechanical properties of the sclera are abnormal, such as excessive compliance (i.e., the sclera is overstretched under the action of intraocular pressure), the axial length may grow abnormally, thereby increasing the individual's risk of high myopia or pathological myopia. Therefore, evaluating the biomechanical properties of the sclera has become an important direction for predicting the progression of myopia. However, there is currently a lack of non-invasive means to directly evaluate the biomechanical properties of the sclera in clinical practice. Therefore, the development of technology that can non-invasively and conveniently evaluate the biomechanical properties of the sclera is of great significance for achieving early intervention and accurate prediction of myopia.

[0063] Based on this, the embodiments of the present application propose a myopia risk prediction method, device, computer equipment, computer-readable storage medium and computer program product based on artificial intelligence, by first acquiring the fundus image data of the individual to be predicted, then segmenting the fundus image data, and extracting the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted from the segmented fundus image data, and finally inputting the biomechanical characteristics into a pre-trained myopia risk prediction model for training, thereby obtaining the myopia risk prediction result of the individual to be predicted.

[0064] Thus, compared to the method of indirectly evaluating myopia progression only by measuring diopter and axial length, the present application trains a machine learning model based on pre-evaluating the biomechanical characteristics of the sclera at the back of the eye of an individual, and obtains a myopia risk prediction model. Therefore, for the individual to be predicted, it is only necessary to non-invasively obtain the fundus image data of the individual to be predicted, and extract the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted based on the fundus image data. The myopia risk prediction model can be used to train the biomechanical characteristics to obtain the myopia risk prediction result of the individual to be predicted. In other words, the present application realizes the non-invasive evaluation of the biomechanical characteristics of the sclera at the back of the eye of an individual, thereby accurately predicting the myopia risk of an individual based on the biomechanical characteristics of the sclera at the back of the eye.

[0065] Based on the overall concept of the above-mentioned embodiments of the present application, specific embodiments of the myopia risk prediction method based on artificial intelligence, the device, the computer equipment, the computer readable storage medium and the computer program product provided in the embodiments of the present application are proposed. First, each specific embodiment of the myopia risk prediction method based on artificial intelligence in the embodiments of the present application is described in detail.

[0066] The embodiments of the present application can acquire and process relevant data based on artificial intelligence technology. Among them, artificial intelligence (AI) is the theory, method, technology and application system that uses digital computers or machines controlled by digital computers to simulate, extend and expand human intelligence, perceive the environment, acquire knowledge and use knowledge to obtain the best results.

[0067] AI basic technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing technology, operation / interaction systems, mechatronics, etc. AI software technologies mainly include computer vision technology, robotics technology, biometrics technology, speech processing technology, natural language processing technology, and machine learning / deep learning.

[0068] It should be noted that in each specific implementation of the present application, when it comes to the need to perform relevant processing based on data related to user identity or characteristics such as user information, user behavior data, user historical data, and user location information, the user's permission or consent will be obtained first, and the collection, use, and processing of these data will comply with relevant laws, regulations, and standards. In addition, when the embodiment of the present application needs to obtain the user's sensitive personal information, the user's separate permission or consent will be obtained through a pop-up window or by jumping to a confirmation page. After clearly obtaining the user's separate permission or consent, the necessary user-related data for the normal operation of the embodiment of the present application will be obtained.

[0069] In addition, the myopia risk prediction method based on artificial intelligence provided in the embodiment of the present application relates to the field of artificial intelligence technology. The myopia risk prediction method based on artificial intelligence provided in the embodiment of the present application can be applied to the terminal, can also be applied to the server side, and can also be software running in the terminal or the server side. In some embodiments, the terminal can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc.; the server side can be configured as an independent physical server, or it can be configured as a server cluster or a distributed system composed of multiple physical servers, and can also be configured to provide cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms and other basic cloud computing services. Cloud servers; software can be applications that implement myopia risk prediction methods based on artificial intelligence, etc., but are not limited to the above forms.

[0070] Alternatively, the present application embodiment can also be used in numerous general or special computer system environments or configurations. For example: personal computers, server computers, handheld devices or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer computer devices, network PCs, minicomputers, mainframe computers, distributed computing environments including any of the above systems or devices, etc. The present application can be described in the general context of computer executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. The present application can also be practiced in distributed computing environments, in which tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules can be located in local and remote computer storage media including storage devices.

[0071] For ease of understanding and explanation, the following text will take the application of the artificial intelligence-based myopia risk prediction method provided by the embodiment of the present application to a terminal device as an example for detailed description. The implementation of the artificial intelligence-based myopia risk prediction method provided by the embodiment of the present application in any of the above-mentioned forms of the subject application can refer to the process of the artificial intelligence-based myopia risk prediction method applied by the terminal device described later.

[0072] Please refer to Figure 1 , Figure 1 A schematic diagram of the steps of the myopia risk prediction method based on artificial intelligence in some embodiments of the present application. It should be understood that although Figure 1 The execution order of some method steps is shown in the figure, but based on different design requirements of actual applications, the myopia risk prediction method based on artificial intelligence provided in the embodiment of the present application can certainly adopt an execution order different from the method steps shown in the figure. That is, Figure 1 The order of the steps of the method shown does not constitute a limitation on the execution logic order of the method for predicting myopia risk based on artificial intelligence provided in the embodiment of the present application. Figure 1 Reasonable changes in the order of the method steps shown should be included in the scope of protection of the artificial intelligence-based myopia risk prediction method provided in the embodiments of the present application.

[0073] like Figure 1 As shown, in some embodiments, the artificial intelligence-based myopia risk prediction method provided in the embodiments of the present application may include but is not limited to steps S101 to S104.

[0074] Step S101, obtaining fundus image data of the individual to be predicted.

[0075] When the terminal device predicts the myopia risk for an individual to be predicted, it first obtains the fundus image data of the individual to be predicted through the OCT device.

[0076] In some embodiments, the terminal device can obtain the fundus image data of the individual to be predicted by exchanging data with the OCT device. For example, the terminal device sends a data request instruction to the OCT device, so that the OCT device responds to the data request instruction and starts to collect images of the posterior ocular tissue of the individual to be predicted, and directly reports the collected fundus image data of the individual to be predicted to the terminal device, so that the terminal device can obtain the fundus image data of the individual to be predicted.

[0077] In some embodiments, the fundus image data may include: Figure 2A Baseline images without intraocular pressure as shown and Figure 2B The mechanical loading image of the intraocular pressure is shown. The terminal device can generate a baseline image and a mechanical loading image in the process of acquiring the image data of the posterior ocular tissue of the individual to be predicted through the OCT device.

[0078] In the embodiment of the present application, the fundus image data of the individual to be predicted is obtained by exchanging data between the terminal device and the OCT device, thereby achieving the purpose of non-invasively collecting fundus image data for myopia risk prediction. In addition, since the process of collecting individual fundus image data is non-invasive, high-quality three-dimensional images of the posterior tissue of the eyeball can be obtained without causing discomfort to the patient.

[0079] In some embodiments, when the terminal device collects image data of the posterior eye tissue of the individual to be predicted through the OCT device, the terminal device can further clean and standardize the collected OCT image (fundus image data) to remove noise, and convert the OCT image data into the Digital Imaging and Communications in Medicine (DICOM) format to facilitate subsequent analysis of the OCT image as input to the algorithm.

[0080] Step S102, segmenting the fundus image data.

[0081] After acquiring the fundus image data of the individual to be predicted, the terminal device can segment each frame of the fundus image data by calling a pre-trained feature extraction model.

[0082] In some embodiments, the terminal device can obtain a feature extraction model by using a preset deep learning segmentation algorithm for model training. For example, the terminal device collects a large number of posterior ocular tissue images as samples through an OTC device, and generates a segmented image based on a portion of the large number of posterior ocular tissue images, and uses the segmented image as a result label, thereby training a preset deep learning segmentation algorithm based on the sample and label to obtain a feature extraction model.

[0083] It should be noted that the preset deep learning segmentation algorithms include convolutional neural network (CNN), recurrent neural network (RNN) and deep graph cut, etc. It should be understood that based on different design requirements of actual applications, in different feasible implementations, the deep learning segmentation algorithm used by the terminal device to train the feature extraction model can of course also be other algorithms not listed here, and the myopia risk prediction method based on artificial intelligence provided in the embodiment of the present application is not limited to the specific type of the deep learning segmentation algorithm.

[0084] Step S103, extracting the biomechanical characteristics of the posterior sclera of the individual to be predicted from the segmented fundus image data.

[0085] It should be noted that the terminal device obtains the segmentation result (segmented fundus image data) by calling the feature extraction model to segment each frame of the fundus image data. Figure 3 As shown. In addition, the biomechanical characteristics of the sclera at the back of the eye include: sclera thickness, curvature, etc. It should be understood that, based on different design requirements of actual applications, in different feasible implementations, the terminal device extracts the biomechanical characteristics of the sclera at the back of the eye from the segmented fundus image data through the feature extraction module, which may of course also include other features not listed here. The artificial intelligence-based myopia risk prediction method provided in the embodiment of the present application does not limit the types of other features of the sclera at the back of the eye.

[0086] The terminal device further calls the feature extraction model to extract key biomechanical parameters from the segmented fundus image data, thereby extracting biomechanical characteristics such as the sclera thickness and curvature of the sclera at the back of the eye of the individual to be predicted from the segmented fundus image data.

[0087] Step S104, inputting the biomechanical characteristics into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result of the individual to be predicted.

[0088] After the terminal device extracts the biomechanical characteristics of the sclera at the back of the eye from the segmented fundus image data by calling the feature extraction model, the terminal device further inputs the biomechanical characteristics into a pre-trained myopia risk prediction model to train the biomechanical characteristics through the myopia risk prediction model, thereby predicting the myopia risk of the individual to be predicted based on the biomechanical characteristics, and outputting the myopia risk prediction result of the individual to be predicted.

[0089] It should be noted that the myopia risk prediction model can be obtained based on model training of the machine learning classification algorithm XGBoost. For example, the terminal device extracts biomechanical features such as scleral thickness and curvature from sample image data of the back of the eye with confirmed myopia risk results, and then inputs the extracted biomechanical features and myopia risk results into the machine learning classification algorithm XGBoost for model training, thereby obtaining a myopia risk prediction model.

[0090] In an embodiment of the present application, when predicting the myopia risk for an individual to be predicted through a terminal device, the fundus image data of the individual to be predicted is first obtained through an OCT device, and then a pre-trained feature extraction model is called to segment each frame of the fundus image data, and the feature extraction model is called to extract key biomechanical parameters for the segmented fundus image data, thereby extracting the biomechanical characteristics of the sclera thickness, curvature and other biomechanical characteristics of the sclera at the posterior part of the eye of the individual to be predicted from the segmented fundus image data, and finally, the extracted biomechanical characteristics are input into a pre-trained myopia risk prediction model, so as to train the biomechanical characteristics through the myopia risk prediction model, thereby predicting the myopia risk of the individual to be predicted based on the biomechanical characteristics, and outputting the myopia risk prediction result of the individual to be predicted.

[0091] Thus, compared to the method of indirectly evaluating myopia progression only by measuring diopter and axial length, the present application trains a machine learning model based on pre-evaluating the biomechanical characteristics of the sclera at the back of the eye of an individual, and obtains a myopia risk prediction model. Therefore, for the individual to be predicted, it is only necessary to non-invasively obtain the fundus image data of the individual to be predicted, and extract the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted based on the fundus image data. The myopia risk prediction model can be used to train the biomechanical characteristics to obtain the myopia risk prediction result of the individual to be predicted. In other words, the present application realizes the non-invasive evaluation of the biomechanical characteristics of the sclera at the back of the eye of an individual, thereby accurately predicting the myopia risk of an individual based on the biomechanical characteristics of the sclera at the back of the eye.

[0092] In addition, the embodiments of the present application use conventional OCT equipment to collect images of individual posterior ocular tissue for non-invasive testing, which greatly reduces the complexity of clinical operations, thereby providing an efficient and convenient technical solution for large-scale myopia risk screening for adolescents.

[0093] Before calling the myopia risk prediction model to predict the myopia risk of an individual according to the biomechanical characteristics of the sclera at the back of the eye, the terminal device can first obtain the myopia risk prediction model based on model training of the machine learning classification algorithm XGBoost.

[0094] Please refer to Figure 4 , Figure 4 A schematic flowchart of the steps of the artificial intelligence-based myopia risk prediction method in other embodiments provided in the embodiments of the present application.

[0095] like Figure 4 As shown, in some embodiments, in the aforementioned step S104, the biomechanical characteristics are input into a pre-trained myopia risk prediction model for training, and before the myopia risk prediction result of the individual to be predicted is obtained, the artificial intelligence-based myopia risk prediction method provided in the embodiment of the present application may also include steps S401 to S403 as shown below.

[0096] Step S401, obtaining sample image data of the back of the eye.

[0097] During the process of training the myopia risk prediction model, the terminal device first collects sample image data of the back of the eye through the OCT device.

[0098] In some embodiments, the terminal device may use an OCT device to collect images of posterior ocular tissue for multiple individuals with known myopia risk results, thereby using the collected OCT images as sample image data.

[0099] In other embodiments, the terminal device may also use an OCT device to collect images of posterior ocular tissue for multiple individuals with unknown myopia risk results, thereby using the collected OCT images as sample image data.

[0100] Step S402: construct an algorithm training data set based on the sample image data.

[0101] After the terminal device acquires the image of the posterior eye tissue through the OCT device to obtain sample image data of the posterior eye, it further classifies and adds labels to the sample image data, thereby constructing an algorithm training data set for model training of the machine learning classification algorithm XGBoost.

[0102] In some embodiments, when the terminal device collects images of posterior ocular tissue for individuals with known myopia risk results through an OCT device to obtain sample image data of the posterior eye, the terminal device can directly classify and add corresponding labels to the sample image data corresponding to each individual based on the myopia risk result of each individual, thereby constructing an algorithm training data set.

[0103] In other embodiments, when the OCT device collects posterior ocular tissue images for individuals with unknown myopia risk results to obtain sample image data of the posterior eye, the terminal device can perform image registration and strain extraction processing on each piece of sample image data, thereby classifying and adding corresponding labels to each piece of sample image data based on the strain of each piece of sample image data, thereby constructing an algorithm training data set.

[0104] Step S403, training a machine learning classification algorithm based on the algorithm training data set and preset compliance labels to obtain a myopia risk prediction model.

[0105] It should be noted that the preset compliance label can be a label added to the classified sample image data when the terminal device constructs an algorithm training data set based on the sample image data of the back of the eye. The compliance label is used to guide XGBoost to accurately classify different sample image data in the algorithm training data set when training the model for the machine learning classification algorithm XGBoost.

[0106] After the terminal device constructs the algorithm training data set, it can input the algorithm training data set together with the compliance labels corresponding to the different sample image data in the algorithm training data set into the machine learning classification algorithm XGBoost, so that XGBoost can perform model training based on the biomechanical characteristics of the posterior sclera of the eye in each sample image data in the algorithm training data set, so as to finally accurately classify the different sample image data in the algorithm training data set according to the compliance labels. In this way, the terminal device can store the trained XGBoost that can accurately classify the sample image data as a myopia risk prediction model. Afterwards, when the terminal device predicts the myopia risk for the individual to be predicted, by inputting the biomechanical characteristics of the individual into the myopia risk prediction model, the myopia risk of the individual can be predicted according to the biomechanical characteristics through the myopia risk prediction model.

[0107] In some embodiments, the algorithm training data set includes: high compliance training samples and low compliance training samples. Figure 5 , Figure 5 for Figure 4 Schematic diagram of the detailed step flow of step S402.

[0108] like Figure 5 As shown, the above step S402, constructing an algorithm training data set based on the sample image data, may include steps S501 to S504 as shown below.

[0109] Step S501 : registering the sample image data based on a preset three-dimensional registration algorithm.

[0110] When it is necessary to explain, the preset three-dimensional registration algorithm can be the classic three-dimensional registration algorithm Ants. It should be understood that based on different design requirements of actual applications, in different feasible implementations, the terminal device can of course also adopt other registration algorithms different from Ants. The artificial intelligence-based myopia risk prediction method provided in the embodiment of the present application is not limited to the specific type of three-dimensional registration algorithm.

[0111] The terminal device performs data cleaning and standardization on the acquired sample image data of the back of the eye to remove the noise, and then further converts the sample image data into DICOM format, and then calls the classic 3D registration algorithm Ants to perform registration processing on each sample image data.

[0112] In some embodiments, each sample image data obtained by the terminal device through OCT image acquisition of the individual by the OCT device may also include a baseline image and a mechanical loading image. In this case, the above step S501, registering the sample image data based on a preset three-dimensional registration algorithm, may include the steps shown as follows:

[0113] Performing an affine transformation on the baseline image and the mechanical loading image based on a preset three-dimensional registration algorithm to obtain an affine-transformed baseline image and an affine-transformed mechanical loading image;

[0114] Based on the three-dimensional registration algorithm, the baseline image after the affine transformation and the mechanical loading image after the affine transformation are elastically transformed.

[0115] When the terminal device calls the classic three-dimensional registration algorithm Ants to perform registration processing on each sample image data, it first performs affine transformation on the baseline image and the mechanical loading image based on the three-dimensional registration algorithm Ants to obtain the baseline image after affine transformation and the mechanical loading image after affine transformation. After that, the terminal device further performs elastic transformation on the baseline image after affine transformation and the mechanical loading image after affine transformation based on the three-dimensional registration algorithm Ants.

[0116] It should be noted that the elastic transformation of the baseline image after affine transformation and the mechanical loading image after affine transformation performed by the terminal device based on the three-dimensional registration algorithm Ants includes: mapping the baseline image m(x) after affine transformation to the mechanical loading image f(x) through a displacement field u(x).

[0117] Step S502 : performing strain extraction processing on the registered sample image data to obtain the strain magnitude of the registered sample image data.

[0118] The terminal device further performs strain extraction processing on the registered sample image data, thereby obtaining the strain size of each registered sample image data.

[0119] In some embodiments, the terminal device may calculate a displacement field for the registered sample image data, and further extract a three-dimensional strain tensor, thereby calculating an overstrain value based on the three-dimensional strain tensor.

[0120] Based on this, the above step S502 may include the following steps:

[0121] Calculate the displacement field of each pixel in the registered sample image data;

[0122] Calculating the partial derivative of the displacement field to obtain the three-dimensional strain tensor of the sclera at the back of the eye in the sample image data;

[0123] The norm of the three-dimensional strain tensor is calculated to obtain the strain magnitude of the registered sample image data.

[0124] After the terminal device calls the classic three-dimensional registration algorithm Ants to perform registration processing on each sample image data, it first calculates the displacement field of each pixel in the registered sample image data, and then further calculates the partial derivative of the displacement field to obtain the three-dimensional strain tensor of the sclera at the back of the eye in the sample image data. Finally, by calculating the norm of the three-dimensional strain tensor, the strain size of the registered sample image data can be obtained.

[0125] Please note that, please refer to Figure 6 The displacement field visualization result is shown. Since the terminal device performs elastic transformation on the affine transformed baseline image and the affine transformed mechanical loading image based on the three-dimensional registration algorithm Ants, the affine transformed baseline image m(x) is mapped to the mechanical loading image f(x) through a displacement field u(x). The deformation is defined in the physical space of the image and determines the displacement difference between the corresponding features in the two images. Therefore, if the feature at position y in f(x) matches the feature at position x in m(x), then the displacement field should give u(x)=yx at x. Such a displacement field can deform the image f(x) into the image m(x) through the combined mapping fdeformed(x)=f(x+u(x)). In the perfect case, m(x)=fdeformed(x). The gradient descent optimization of the elastic mapping can be roughly summarized as:

[0126] Calculate the similarity gradient:

[0127] Update the displacement field:

[0128] Regularized displacement field: u(x)←Gσ*u(x)

[0129] Among them, Π is the similarity, δ is the gradient step size, and Gσ is the Gaussian smoother.

[0130] In addition, in biomechanics, strain is a measure that describes the degree of deformation of a material or tissue under the action of force. It is an important parameter of the mechanical properties of a material and can be used to evaluate the elasticity, plasticity and other properties of the material. In the biomechanical measurement of posterior ocular tissue, the measurement of strain is crucial to understanding the biomechanical behavior of ocular tissue. In the three-dimensional case, the strain tensor can be used to represent the three-dimensional strain field. For example, the strain tensor that can be used has 6 independent components: ∈xx, ∈yy, ∈zz, ∈xy, ∈xz, ∈yz. In addition, the components of the strain tensor can be expressed as partial derivatives of the displacement field. Therefore, assuming the displacement field is u=(ux,uy,uz), the components of the strain tensor can be calculated as follows:

[0131]

[0132] At the same time, when the strain in the image is large, the terminal device can understand that the sclera has high compliance, and when the strain in the image is small, it can be understood that the sclera has low compliance. Therefore, the terminal device can express the strain magnitude by calculating the norm of the strain tensor, and the formula is as follows:

[0133]

[0134] Here, ∈ is the norm of the strain tensor, which is also the size of the strain in the image.

[0135] Step S503 : when the strain magnitude is greater than or equal to a preset strain threshold, classifying the registered sample image data as the high-compliance training sample.

[0136] It should be noted that the preset strain threshold may be a strain value of the sclera at the back of the eye preset by the terminal device, and the preset strain threshold may be used to assess whether an individual has a risk of myopia. For example, when the strain value of the sclera at the back of the eye of an individual is greater than or equal to the preset strain threshold, the terminal device predicts that the individual has a risk of myopia, and when the strain value of the sclera at the back of the eye of an individual is less than the preset strain threshold, the terminal device may predict that the individual does not have a risk of myopia.

[0137] After obtaining the strain size of each registered sample image data, the terminal device compares the strain size with the preset strain threshold. When the strain size of the registered sample image data is greater than or equal to the preset strain threshold, the terminal device can classify the registered sample image data as a high-compliance training sample in the algorithm training data set.

[0138] Step S504 : when the strain magnitude is less than a preset strain threshold, classify the registered sample image data as the low-compliance training sample.

[0139] After the terminal device compares the strain size of the registered sample image data with the preset strain threshold, if the strain size of the registered sample image data is less than the preset strain threshold, the terminal device will classify the registered sample image data as a low-compliance training sample in the algorithm training data set.

[0140] In some embodiments, the compliance label includes: a high compliance label and a low compliance label. Based on this, the above step S403, based on the algorithm training data set and the preset compliance label to train the machine learning classification algorithm to obtain the myopia risk prediction model, may include the following steps:

[0141] Extracting a first biomechanical feature of the posterior sclera of the eye from the high-compliance training sample by using a preset feature extraction model;

[0142] Extracting a second biomechanical feature of the posterior sclera of the eye from the low-compliance training sample by using a preset feature extraction model;

[0143] The first biomechanical feature and the high compliance label are input into a machine learning classification algorithm to train the machine learning classification algorithm, and the second biomechanical feature and the low compliance label are input into the machine learning classification algorithm to train the machine learning classification algorithm to obtain a myopia risk prediction model.

[0144] When the terminal device trains the machine learning classification algorithm based on the constructed algorithm training data set and the preset compliance label, for the high-compliance training samples of the algorithm training data set, the terminal device extracts the sclera thickness, curvature, etc. of the sclera at the back of the eye from the high-compliance training samples by calling the pre-trained feature extraction model. For the low-compliance training samples of the algorithm training data set, the terminal device also extracts the sclera thickness, curvature, etc. of the sclera at the back of the eye from the low-compliance training samples by calling the pre-trained feature extraction model.

[0145] Afterwards, the terminal device further inputs the first biomechanical feature extracted from the high-compliance training sample together with the high-compliance label into the machine learning classification algorithm to train the machine learning classification algorithm. At the same time, the terminal device also inputs the second biomechanical feature extracted from the low-compliance training sample together with the low-compliance label into the machine learning classification algorithm to train the machine learning classification algorithm. In this way, after the machine learning classification algorithm converges in model training based on the first biomechanical feature and the high-compliance label, and converges in model training based on the second biomechanical feature and the low-compliance label, the terminal device can store the machine learning classification algorithm as a myopia risk prediction model.

[0146] For example, Figure 7 In the system framework shown, when the terminal device constructs the algorithm training sample data to train the machine learning classification algorithm to obtain the myopia risk prediction model, it first collects data through the data acquisition module to build an algorithm training data set. That is, the image data of the back of the eye is obtained through the OCT device to generate a baseline image without intraocular pressure and a mechanical loading image with intraocular pressure. This process is non-invasive and can obtain high-quality three-dimensional images of the posterior tissue of the eye without causing discomfort to the individual. Afterwards, the terminal device cleans and standardizes the collected OCT images through the data preprocessing module, removes noise, and converts the image data into DICOM format to facilitate the input and analysis of subsequent algorithms.

[0147] After that, the terminal device uses the classic three-dimensional registration algorithm Ants for image registration based on the image registration and strain extraction module. The registered image data is used to calculate the displacement field, and further extract the three-dimensional strain tensor to evaluate the deformation of the sclera at the back of the eye under the action of intraocular pressure. In addition, the terminal device uses a pre-trained deep learning model to segment each frame of the fundus OCT image through the biomechanical feature extraction module, and extracts key biomechanical parameters from the segmented OCT image, including scleral thickness, curvature, etc. Therefore, through the deep learning model, these features are quantified and their correlation with compliance is analyzed, and then the features with higher correlation and compliance labels are used to train the machine learning classification algorithm XGBoost. Therefore, the terminal device inputs the extracted biomechanical features into the pre-trained machine learning classification algorithm XGBoost through the compliance prediction module, so that XGBoost can predict the individual's myopia risk based on the biomechanical features.

[0148] In this embodiment, the terminal device uses a three-dimensional registration algorithm (such as the Ants registration tool) to accurately calculate the displacement field of the posterior ocular tissue under different intraocular pressure conditions, thereby optimizing the application effect of the registration algorithm in processing complex tissue deformation by applying the three-dimensional registration technology to the OCT image for the image registration of the posterior sclera. In addition, the terminal device combines the technology of affine transformation and elastic transformation to further ensure the global and local alignment of the OCT image, thereby improving the accuracy of extracting the biomechanical characteristics of the posterior sclera and using it for subsequent assessment of myopia risk.

[0149] In addition, the strain data is used for compliance classification when constructing the algorithm training data set through the terminal device, and the high compliance and low compliance groups in the algorithm training sample data set are divided based on the displacement field and strain tensor obtained by three-dimensional registration. Thus, combined with the relationship between scleral compliance and myopia risk, the training-based machine learning classification algorithm can provide more accurate results in the early prediction of myopia progression. That is, compared with the prediction method that only relies on axial length or refractive power, this embodiment can significantly improve the prediction accuracy of myopia risk by combining compliance classification with artificial intelligence prediction through the terminal device.

[0150] Next, a complete embodiment of the myopia risk prediction method based on artificial intelligence proposed in the embodiments of the present application is presented.

[0151] First, the terminal device performs the steps of data acquisition and preprocessing. That is, the terminal device collects the patient's fundus images through the OCT device, including baseline images and loading images, to ensure the integrity of the image data. The collected image data is formatted into DICOM format, and necessary noise elimination and standardization are performed.

[0152] Afterwards, the terminal device performs three-dimensional registration and displacement field calculation on the image. That is, the terminal device uses the registration algorithm to perform affine and elastic transformations on the baseline image and the loaded image, and calculates the displacement field of each pixel in the image. In this process, the terminal device can enhance the registration accuracy through the CNN network, thereby ensuring that the displacement field after image alignment can accurately reflect the deformation of the posterior ocular tissue.

[0153] Afterwards, the terminal device can further extract the strain field and classify the data for the image. That is, the terminal device calculates the partial derivative of the displacement field to obtain the three-dimensional strain tensor of the posterior eye tissue. The image data is divided into high compliance and low compliance according to the strain value, marking patients with potential myopia risk.

[0154] Finally, the terminal device performs the steps of feature extraction and prediction model training. That is, the terminal device uses the segmentation algorithm to extract biomechanical features such as scleral thickness and curvature from the OCT image, and inputs these features into the XGBoost model for training, and optimizes the model parameters through cross-validation, and finally generates the patient's myopia risk prediction results.

[0155] See also Figure 8 The embodiment of the present application also provides a myopia risk prediction device based on artificial intelligence, which can implement the above-mentioned myopia risk prediction method based on artificial intelligence, and the device includes:

[0156] An acquisition module, used for acquiring fundus image data of the individual to be predicted;

[0157] A feature extraction module, used for segmenting the fundus image data, and extracting the biomechanical features of the posterior sclera of the individual to be predicted from the segmented fundus image data;

[0158] The risk prediction module is used to input the biomechanical characteristics into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result for the individual to be predicted.

[0159] In some embodiments, the myopia risk prediction device based on artificial intelligence provided by the embodiments of the present application further includes:

[0160] The model training module is used to obtain sample image data of the back of the eye; construct an algorithm training data set based on the sample image data; and train a machine learning classification algorithm based on the algorithm training data set and preset compliance labels to obtain a myopia risk prediction model.

[0161] In some embodiments, the algorithm training data set includes high compliance training samples and low compliance training samples;

[0162] The model training module is also used to align the sample image data based on a preset three-dimensional alignment algorithm; perform strain extraction processing on the aligned sample image data to obtain the strain size of the aligned sample image data; and, when the strain size is greater than or equal to a preset strain threshold, divide the aligned sample image data into the high-compliance training sample; and, when the strain size is less than the preset strain threshold, divide the aligned sample image data into the low-compliance training sample.

[0163] In some embodiments, the sample image data includes: a baseline image and a mechanical loading image;

[0164] The model training module is also used to perform affine transformation on the baseline image and the mechanical loading image based on a preset three-dimensional registration algorithm to obtain the baseline image after affine transformation and the mechanical loading image after affine transformation; and to perform elastic transformation on the baseline image after affine transformation and the mechanical loading image after affine transformation based on the three-dimensional registration algorithm.

[0165] In some embodiments, the model training module is also used to calculate the displacement field of each pixel point in the registered sample image data; calculate the partial derivative of the displacement field to obtain the three-dimensional strain tensor of the sclera at the back of the eye in the sample image data; and calculate the norm of the three-dimensional strain tensor to obtain the strain magnitude of the registered sample image data.

[0166] In some embodiments, the compliance tags include: a high compliance tag and a low compliance tag;

[0167] The model training module is also used to extract the first biomechanical feature of the sclera at the back of the eye from the high-compliance training samples through a preset feature extraction model; to extract the second biomechanical feature of the sclera at the back of the eye from the low-compliance training samples through a preset feature extraction model; and to input the first biomechanical feature and the high-compliance label into the machine learning classification algorithm to train the machine learning classification algorithm, and to input the second biomechanical feature and the low-compliance label into the machine learning classification algorithm to train the machine learning classification algorithm to obtain a myopia risk prediction model.

[0168] It should be noted that the specific implementation of the artificial intelligence-based myopia risk prediction device provided in the embodiment of the present application is basically the same as the specific implementation of the above-mentioned artificial intelligence-based myopia risk prediction method, and will not be repeated here.

[0169] The embodiment of the present application also provides a computer device, the computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the above-mentioned artificial intelligence-based myopia risk prediction method when executing the computer program. The computer device can be any intelligent terminal including a tablet computer, a car computer, etc.

[0170] See also Fig. 9 , Fig. 9 The hardware structure of a computer device according to another embodiment is shown, and the computer device includes:

[0171] The processor 901 may be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an application-specific integrated circuit (Application Specific Integrated Circuit, ASIC), or one or more integrated circuits, and is used to execute relevant programs to implement the technical solutions provided in the embodiments of the present application;

[0172] The memory 902 can be implemented in the form of a read-only memory (ROM), a static storage device, a dynamic storage device, or a random access memory (RAM). The memory 902 can store an operating system and other applications. When the technical solution provided in the embodiments of this specification is implemented by software or firmware, the relevant program code is stored in the memory 902, and the processor 901 calls and executes the myopia risk prediction method based on artificial intelligence in the embodiments of this application;

[0173] Input / output interface 903, used to implement information input and output;

[0174] Communication interface 904, used to realize communication interaction between the device and other devices, which can be realized by wired mode (such as USB, network cable, etc.) or wireless mode (such as mobile network, WIFI, Bluetooth, etc.);

[0175] A bus 905 that transmits information between various components of the device (e.g., the processor 901, the memory 902, the input / output interface 903, and the communication interface 904);

[0176] The processor 901 , the memory 902 , the input / output interface 903 and the communication interface 904 are connected to each other in communication within the device via a bus 905 .

[0177] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the above-mentioned artificial intelligence-based myopia risk prediction method is implemented.

[0178] The memory, as a non-transient computer-readable storage medium, can be used to store non-transient software programs and non-transient computer executable programs. In addition, the memory may include a high-speed random access memory, and may also include a non-transient memory, such as at least one disk storage device, a flash memory device, or other non-transient solid-state storage device. In some embodiments, the memory may optionally include a memory remotely disposed relative to the processor, and these remote memories may be connected to the processor via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0179] An embodiment of the present application also provides a computer program product, which stores a computer program, and when the computer program is executed by a processor, the above-mentioned artificial intelligence-based myopia risk prediction method is implemented.

[0180] The embodiments of the present application provide an artificial intelligence-based myopia risk prediction method, an artificial intelligence-based myopia risk prediction device, a computer device, a computer-readable storage medium, and a computer program product, which first obtain fundus image data of the individual to be predicted, then segment the fundus image data, and extract the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted from the segmented fundus image data, and finally input the biomechanical characteristics into a pre-trained myopia risk prediction model for training, thereby obtaining the myopia risk prediction result of the individual to be predicted.

[0181] Thus, compared to the method of indirectly evaluating myopia progression only by measuring diopter and axial length, the present application trains a machine learning model based on pre-evaluating the biomechanical characteristics of the sclera at the back of the eye of an individual, and obtains a myopia risk prediction model. Therefore, for the individual to be predicted, it is only necessary to non-invasively obtain the fundus image data of the individual to be predicted, and extract the biomechanical characteristics of the sclera at the back of the eye of the individual to be predicted based on the fundus image data. The myopia risk prediction model can be used to train the biomechanical characteristics to obtain the myopia risk prediction result of the individual to be predicted. In other words, the present application realizes the non-invasive evaluation of the biomechanical characteristics of the sclera at the back of the eye of an individual, thereby accurately predicting the myopia risk of an individual based on the biomechanical characteristics of the sclera at the back of the eye.

[0182] In addition, the embodiments of the present application use conventional OCT equipment to collect images of individual posterior ocular tissue for non-invasive testing, which greatly reduces the complexity of clinical operations, thereby providing an efficient and convenient technical solution for large-scale myopia risk screening for adolescents.

[0183] The embodiments described in the embodiments of the present application are intended to more clearly illustrate the technical solutions of the embodiments of the present application and do not constitute a limitation on the technical solutions provided in the embodiments of the present application. Those skilled in the art will appreciate that with the evolution of technology and the emergence of new application scenarios, the technical solutions provided in the embodiments of the present application are also applicable to similar technical problems.

[0184] Those skilled in the art will appreciate that the technical solutions shown in the figures do not constitute a limitation on the embodiments of the present application, and may include more or fewer steps than shown in the figures, or a combination of certain steps, or different steps.

[0185] The device embodiments described above are merely illustrative, and the units described as separate components may or may not be physically separated, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0186] Those skilled in the art will appreciate that all or some of the steps in the methods disclosed above, and the functional modules / units in the systems and devices may be implemented as software, firmware, hardware, or a suitable combination thereof.

[0187] The terms "first", "second", "third", "fourth", etc. (if any) in the specification of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments of the present application described herein can be implemented in an order other than those illustrated or described herein. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device comprising a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0188] It should be understood that in the present application, "at least one (item)" means one or more, and "plurality" means two or more. "And / or" is used to describe the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" can mean: only A exists, only B exists, and A and B exist at the same time, where A and B can be singular or plural. The character " / " generally indicates that the objects associated before and after are in an "or" relationship. "At least one of the following" or similar expressions refers to any combination of these items, including any combination of single or plural items. For example, at least one of a, b or c can mean: a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, c can be single or multiple.

[0189] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the above units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0190] The units described above as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0191] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0192] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including multiple instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods of various embodiments of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (Read-Only Memory, referred to as ROM), random access memory (Random Access Memory, referred to as RAM), disk or optical disk and other media that can store programs.

[0193] The preferred embodiments of the present invention are described above with reference to the accompanying drawings, but the scope of the rights of the present invention is not limited thereto. Any modification, equivalent substitution and improvement made by a person skilled in the art without departing from the scope and essence of the present invention should be within the scope of the rights of the present invention.

Claims

1. A myopia risk prediction method based on artificial intelligence, characterized in that: The method comprises: Obtain fundus image data of the individual to be predicted; Segmenting the fundus image data; Extracting the biomechanical characteristics of the posterior sclera of the individual to be predicted from the segmented fundus image data; The biomechanical characteristics are input into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result of the individual to be predicted.

2. The method according to claim 1, characterized in that Before inputting the biomechanical features into a pre-trained myopia risk prediction model for training, the method further includes: Acquiring sample image data of the back of the eye; Constructing an algorithm training data set based on the sample image data; A machine learning classification algorithm is trained based on the algorithm training data set and preset compliance labels to obtain a myopia risk prediction model.

3. The method according to claim 2, characterized in that The algorithm training data set includes high-compliance training samples and low-compliance training samples; The step of constructing an algorithm training data set based on the sample image data includes: Registering the sample image data based on a preset three-dimensional registration algorithm; Performing strain extraction processing on the registered sample image data to obtain the strain magnitude of the registered sample image data; When the strain magnitude is greater than or equal to a preset strain threshold, dividing the registered sample image data into the high-compliance training sample; When the strain magnitude is smaller than a preset strain threshold, the registered sample image data is divided into the low-compliance training sample.

4. The method according to claim 3, characterized in that The sample image data includes: a baseline image and a mechanical loading image; The registering the sample image data based on a preset three-dimensional registration algorithm includes: Performing an affine transformation on the baseline image and the mechanical loading image based on a preset three-dimensional registration algorithm to obtain an affine-transformed baseline image and an affine-transformed mechanical loading image; Based on the three-dimensional registration algorithm, the baseline image after the affine transformation and the mechanical loading image after the affine transformation are elastically transformed.

5. The method according to claim 3, characterized in that: The step of performing strain extraction processing on the registered sample image data to obtain the strain magnitude of the registered sample image data includes: Calculate the displacement field of each pixel in the registered sample image data; Calculating the partial derivative of the displacement field to obtain the three-dimensional strain tensor of the sclera at the back of the eye in the sample image data; The norm of the three-dimensional strain tensor is calculated to obtain the strain magnitude of the registered sample image data.

6. The method according to any one of claims 3 to 5, characterized in that The compliance labels include: high compliance labels and low compliance labels; The method of training a machine learning classification algorithm based on the algorithm training data set and a preset compliance label to obtain a myopia risk prediction model includes: Extracting a first biomechanical feature of the posterior sclera of the eye from the high-compliance training sample by using a preset feature extraction model; Extracting a second biomechanical feature of the posterior sclera of the eye from the low-compliance training sample by using a preset feature extraction model; The first biomechanical feature and the high compliance label are input into a machine learning classification algorithm to train the machine learning classification algorithm, and the second biomechanical feature and the low compliance label are input into the machine learning classification algorithm to train the machine learning classification algorithm to obtain a myopia risk prediction model.

7. A myopia risk prediction device based on artificial intelligence, characterized in that: The device comprises: An acquisition module, used for acquiring fundus image data of the individual to be predicted; A feature extraction module, used for segmenting the fundus image data, and extracting the biomechanical features of the posterior sclera of the individual to be predicted from the segmented fundus image data; The risk prediction module is used to input the biomechanical characteristics into a pre-trained myopia risk prediction model for training to obtain a myopia risk prediction result for the individual to be predicted.

8. A computer device, characterized in that: The computer device includes a memory and a processor, the memory stores a computer program, and the processor implements the artificial intelligence-based myopia risk prediction method according to any one of claims 1 to 6 when executing the computer program.

9. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the artificial intelligence-based myopia risk prediction method according to any one of claims 1 to 6 is implemented.

10. A computer program product, wherein the computer program product stores a computer program, characterized in that: When the computer program is executed by a processor, the artificial intelligence-based myopia risk prediction method according to any one of claims 1 to 6 is implemented.

Images