Eye age determination method and device, equipment, readable storage medium and program product

Abstract

The invention relates to an eye age determination method and device, computer equipment, a computer readable storage medium and a computer program product. The method comprises the steps of obtaining an original fundus image of a user, and performing image processing on the original fundus image to obtain a target fundus image; inputting the target fundus image into a pre-trained eye age recognition model to obtain eye age information of the user output by the eye age recognition model; wherein the eye age identification model is obtained by training based on a deformation field algorithm and a retina image age causal feature extraction algorithm based on a double-task learning framework. By adopting the method, the accuracy of determining the eye age can be improved.

Description

Technical Field

[0001] This application relates to the field of data processing technology, and in particular to a method, apparatus, computer device, computer-readable storage medium, and computer program product for determining eye age. Background Technology

[0002] Eye age refers to the physiological age of the eyes determined by assessing the state, structure, and function of eye tissues. Eye age may differ from actual age. Eye age can reflect not only a user's eye health but also their overall health.

[0003] Therefore, there is an urgent need for a method that can accurately determine a user's eye age. Summary of the Invention

[0004] Therefore, it is necessary to provide a method, device, computer equipment, computer-readable storage medium, and computer program product that can accurately determine eye age in order to address the above-mentioned technical problems.

[0005] In a first aspect, this application provides a method for determining visual age, including:

[0006] The system acquires the user's original fundus image and performs image processing on the original fundus image to obtain the target fundus image.

[0007] The target fundus image is input into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model;

[0008] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0009] In one embodiment, the eye age recognition model includes a first model and a second model. A target fundus image is input into the pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model. This includes: inputting the target fundus image into the first model of the eye age recognition model, so that the first model extracts features from the target fundus image to obtain target features, where the target features are features of the natural aging type; and inputting the target features into the second model of the eye age recognition model to obtain the user's eye age information output by the second model.

[0010] In one embodiment, image processing is performed on the original fundus image to obtain a target fundus image, including: performing a first image processing on the original fundus image based on a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image; and performing a second image processing on the first fundus image based on a color normalization algorithm to obtain the target fundus image.

[0011] In one embodiment, the training process of the eye age recognition model includes: acquiring multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group, wherein each training fundus image group includes a first training fundus image and a second training fundus image, and the acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition time of the second training fundus image; acquiring an initial eye age recognition model, and training the initial eye age recognition model based on the multiple training fundus image groups, the acquisition information, and the eye age information to obtain an eye age recognition model.

[0012] In one embodiment, an initial age recognition model is trained based on multiple training fundus image groups, acquisition information, and age information to obtain an age recognition model. This includes: performing image processing on a first training fundus image and a second training fundus image in the multiple training fundus image groups to obtain multiple target training fundus image groups, each target training fundus image group including a target training fundus image and a second target training fundus image; and training the initial age recognition model using each target training fundus image group, acquisition information, and age information based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to obtain an age recognition model.

[0013] In one embodiment, the initial age recognition model includes an initial first model and an initial second model. Based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial age recognition model is trained using fundus image groups, acquired information and age information for each target to obtain the age recognition model. This includes: training the initial first model and the initial second model using fundus image groups, acquired information and age information for each target to obtain the first model and the second model, and determining the age recognition model based on the first model and the second model.

[0014] Secondly, this application also provides a device for determining eye age, comprising:

[0015] The acquisition module is used to acquire the user's original fundus image and perform image processing on the original fundus image to obtain the target fundus image;

[0016] The execution module is used to input the target fundus image into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model;

[0017] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0018] Thirdly, this application also provides a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method described in any of the embodiments of the first aspect above.

[0019] Fourthly, this application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any of the embodiments of the first aspect above.

[0020] Fifthly, this application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the embodiments of the first aspect above.

[0021] The aforementioned methods, apparatus, computer devices, computer-readable storage media, and computer program products for determining eye age first acquire the user's original fundus image and perform image processing on the original fundus image to obtain a target fundus image; then, the target fundus image is input into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model; wherein, the eye age recognition model is trained based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework. The eye age determination method provided in this application uses an eye age recognition model trained based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to determine the user's eye age, which improves the accuracy of eye age determination compared to the prior art that relies on human experience to determine eye age. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the drawings used in the description of the embodiments of this application or related technologies will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating a method for determining eye age in one embodiment;

[0024] Figure 2 This is a flowchart illustrating a method for obtaining user eye age information output by an eye age recognition model in one embodiment.

[0025] Figure 3 This is a flowchart illustrating a method for obtaining a target fundus image in one embodiment;

[0026] Figure 4 This is a flowchart illustrating the training process of an age recognition model in one embodiment.

[0027] Figure 5 This is a flowchart illustrating a method for obtaining an eye age recognition model in one embodiment;

[0028] Figure 6 This is a flowchart illustrating the eye age determination method in another embodiment;

[0029] Figure 7 This is a structural block diagram of an eye age determination device in one embodiment;

[0030] Figure 8 This is an internal structural diagram of a computer device in one embodiment;

[0031] Figure 9 This is a diagram of the internal structure of a computer device in another embodiment. Detailed Implementation

[0032] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0033] It should be noted that the terms "first," "second," etc., used in this application can be used to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from the second element. The terms "comprising" and "having," and any variations thereof, used in this application, are intended to cover non-exclusive inclusion. The term "multiple" used in this application refers to two or more. The term "and / or" used in this application refers to one of the embodiments, or any combination of multiple embodiments.

[0034] Eye age refers to the physiological age of the eyes determined by assessing the state, structure, and function of eye tissues. Eye age may differ from actual age. Eye age can reflect not only a user's eye health but also their overall health.

[0035] Therefore, there is an urgent need for a method that can accurately determine a user's eye age.

[0036] In view of this, this application provides a method for determining eye age by first acquiring the user's original fundus image and then processing the original fundus image to obtain a target fundus image; then, inputting the target fundus image into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model; wherein, the eye age recognition model is trained based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework. The eye age determination method provided in this application uses an eye age recognition model trained based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to determine the user's eye age, which improves the accuracy of eye age determination compared to the prior art that relies on human experience to determine eye age.

[0037] The eye age determination method provided in this application can be implemented by a computer device, which can be a terminal or a server.

[0038] In one exemplary embodiment, such as Figure 1 As shown, a method for determining visual age is provided, which includes the following steps:

[0039] Step 101: Obtain the user's original fundus image and perform image processing on the original fundus image to obtain the target fundus image.

[0040] Optionally, the fundus image may include retinal vessels, optic discs, macula, and retinal background.

[0041] In some exemplary embodiments, the computer device may first acquire the user's raw fundus image.

[0042] Specifically, the user's original fundus image can be acquired using non-contact imaging technology. During the acquisition process, automatic focusing and automatic exposure are used to ensure that the computer device can obtain a clear and properly exposed original fundus image.

[0043] In one alternative approach, if more than 25% of the peripheral retinal area cannot be observed in the acquired raw fundus image, or if there are artifacts in the central retinal area that significantly affect the analysis, the user's raw fundus image is reacquired.

[0044] Furthermore, after obtaining the user's original fundus image, the computer device can perform image processing on the original fundus image to obtain the target fundus image.

[0045] For example, the image processing can be image preprocessing, such as noise reduction, contrast enhancement, geometric correction, and artifact removal.

[0046] This image processing can also include feature extraction, such as blood vessel segmentation, macular and optic disc localization, and lesion detection.

[0047] Image processing can also be a standardization process, such as size and resolution standardization, grayscale standardization, and quantization parameter extraction.

[0048] Step 102: Input the target fundus image into the pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model.

[0049] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0050] Optionally, the age recognition model can be a traditional machine learning model, such as a random forest model or a gradient boosting tree model. It can also be a deep learning model, such as a convolutional neural network model, a Transformer model, or a regression model. Furthermore, it can be a hybrid model, that is, a model composed of at least two different models.

[0051] The deformation field algorithm refers to an algorithm used to describe and calculate the shape changes of an image in space, such as stretching, twisting, bending, and translation. The deformation field algorithm quantifies the spatial mapping relationship of an image from its initial state to its target state by constructing a continuous spatial transformation field, i.e., the deformation field.

[0052] The retinal image age causal feature extraction algorithm based on a dual-task learning framework encodes baseline and follow-up images using the same model, predicts the current age using baseline features, and simultaneously predicts the age change interval by combining baseline and follow-up features. Its core principle is that if the model relies solely on age-related non-causal confounding features (such as device artifacts or cohort-specific bias) for prediction, while it can estimate the baseline age well, it is difficult to accurately infer the amount of age change within an individual. Therefore, by introducing an age interval prediction task and a time-series consistency constraint, the shared encoder is forced to extract features from the images that are stable and regularly changing within the individual over time. This results in purer, more biologically meaningful aging-related causal features, enhancing the model's generalization ability across centers and devices, and providing reliable image-based biomarkers for aging mechanism research and disease progression monitoring.

[0053] For example, in addition to the user's eye age data, the eye age information may also include the user's eye features and structures, such as fundus structure, lens and cornea, and ocular surface tissue; it may also include the user's eye function indicators, such as visual function, intraocular pressure and circulation; and it may also include age deviation, that is, the difference between eye age and actual age, the magnitude of which can reflect the rate of eye aging.

[0054] In some exemplary embodiments, after obtaining a target fundus image, the computer device can input the target fundus image into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model.

[0055] The aforementioned method for determining eye age first acquires the user's original fundus image and processes it to obtain a target fundus image. Then, the target fundus image is input into a pre-trained eye age recognition model to obtain the user's eye age information output by the model. This eye age recognition model is trained using a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework. The eye age determination method provided in this application uses an eye age recognition model trained on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to determine the user's eye age. Compared to existing technologies that rely on human experience to determine eye age, this method improves the accuracy of eye age determination.

[0056] In one exemplary embodiment, such as Figure 2 As shown, the eye age recognition model includes a first model and a second model. The target fundus image is input into the pre-trained eye age recognition model to obtain the user's eye age information output by the model. This includes the following steps:

[0057] Step 201: Input the target fundus image into the first model in the eye age recognition model so that the first model can extract features from the target fundus image to obtain target features.

[0058] Optionally, the first model can be a spatiotemporal feature encoder model. A spatiotemporal feature encoder model refers to a deep learning model that can simultaneously process dynamic changes in the time dimension and structural information in the spatial dimension.

[0059] The target feature is a characteristic of natural aging. For example, in addition to the target feature, the target fundus image may also contain features related to pathological changes, abnormal states, or non-aging factors. While these features do not fall under the category of natural aging, they directly affect eye health and may even interfere with the accuracy of eye age assessment. Therefore, it is necessary to obtain the target feature from the target fundus image to determine the user's eye age information based on the target feature, thereby improving the accuracy of eye age information.

[0060] In some exemplary embodiments, after obtaining a target fundus image, the computer device can input the target fundus image into a first model in the eye age recognition model, so that the first model can extract features from the target fundus image to obtain target features.

[0061] Specifically, after the computer device inputs the target fundus image into the first model of the eye age recognition model, the first model can analyze the fundus features at different levels in the target fundus image and integrate the analysis results to obtain the target features, avoiding the omission of local details.

[0062] In an optional embodiment of this application, the first model extracts features from the target fundus image to obtain target features. The target features may include structural feature information, such as information related to optic disc area and vascular branch angles; and may also include texture feature information, such as information related to vascular wall transparency and macular pigment deposition.

[0063] Step 202: Input the target features into the second model in the eye age recognition model to obtain the user's eye age information output by the second model.

[0064] Optionally, the second model can be a multilayer perceptron model, a multilayer neural network model, etc. This second model is used to predict the user's eye age information based on target features.

[0065] In some exemplary embodiments, after the computer device inputs a target fundus image into a first model in an eye age recognition model to extract features from the target fundus image and obtain target features, the target features can be input into a second model in the eye age recognition model to obtain the user's eye age information output by the second model.

[0066] Specifically, the computer device can input the target feature into the second model in the eye age recognition model, so that the second model can predict the eye age information based on the target feature, and obtain the user's eye age information output by the second model.

[0067] In one exemplary embodiment, such as Figure 3 As shown, image processing is performed on the original fundus image to obtain the target fundus image, including the following steps:

[0068] Step 301: Perform first image processing on the original fundus image based on the contrast-limited adaptive histogram equalization algorithm to obtain the first fundus image.

[0069] Optionally, the first image processing can be image enhancement processing. This image enhancement processing can be used to adjust the visual effects or feature distribution of the image to highlight key information, suppress noise interference, and improve image quality.

[0070] In some exemplary embodiments, after obtaining the original fundus image, the computer device may perform a first image processing on the original fundus image based on a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image.

[0071] Specifically, computer equipment can convert the original fundus image from the RGB color space to the LAB color space and divide the converted original fundus image into small blocks of fixed size.

[0072] Furthermore, the contrast-limited adaptive histogram equalization algorithm is applied to the brightness channel of each small block. Finally, the processed original fundus image is converted back to the RGB color space to obtain the first fundus image.

[0073] In an optional embodiment of this application, the computer device may also use a linear transformation algorithm and a histogram equalization algorithm to perform a first image processing on the original fundus image to obtain a first fundus image.

[0074] Step 302: Perform second image processing on the first fundus image based on the color normalization algorithm to obtain the target fundus image.

[0075] Optionally, the second image processing can be image normalization. Image normalization can reduce color variations between fundus images under different conditions, such as different fundus camera models, photography settings, and exposure changes.

[0076] In some exemplary embodiments, after obtaining a first fundus image, the computer device may perform a second image processing on the first fundus image based on a color normalization algorithm to obtain a target fundus image.

[0077] Specifically, the second image processing, based on a color normalization algorithm, is applied to the first fundus image to obtain the target fundus image, which can be represented as follows: Where P refers to the target fundus image, This refers to the first fundus image. This refers to applying a Gaussian filter with a standard deviation of s to the first fundus image. , and This refers to adjustable parameters. Among them, if... It is 4. -4, If the value is 128, the resolution of the target fundus image can be adjusted to 512×512.

[0078] In one exemplary embodiment, such as Figure 4 As shown, the training process of the eye age recognition model includes the following steps:

[0079] Step 401: Obtain multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group.

[0080] Each training fundus image group includes a first training fundus image and a second training fundus image. The acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition times of the second training fundus image.

[0081] In some exemplary embodiments, the computer device can acquire multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group.

[0082] Specifically, the first and second training fundus images included in each training fundus image group can be acquired using non-contact imaging technology. During the acquisition process, automatic focusing and automatic exposure are used to ensure the image quality of the first and second training fundus images.

[0083] In one alternative approach, at least 75% of the peripheral retinal area can be observed in both the first and second training fundus images, and there are no artifacts in the central retinal area that would significantly affect the analysis.

[0084] Step 402: Obtain the initial eye age recognition model, and train the initial eye age recognition model based on multiple training fundus image groups, collected information and eye age information to obtain the eye age recognition model.

[0085] In some exemplary embodiments, after acquiring multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group, the computer device can acquire an initial eye age recognition model, and train the initial eye age recognition model based on multiple training fundus image groups, acquisition information, and eye age information to obtain an eye age recognition model.

[0086] In one exemplary embodiment, such as Figure 5 As shown, an initial eye age recognition model is trained based on multiple sets of training fundus images, acquired information, and eye age information to obtain an eye age recognition model, including the following steps:

[0087] Step 501: Perform image processing on the first and second training fundus images in multiple training fundus image groups to obtain multiple target training fundus image groups.

[0088] Each target training fundus image group includes a target training fundus image and a second target training fundus image.

[0089] In some exemplary embodiments, after obtaining multiple training fundus image groups, the computer device can perform a first image processing on the first training fundus image and the second training fundus image in the multiple training fundus image groups based on a contrast-limited adaptive histogram equalization algorithm to obtain a first training fundus image group.

[0090] Specifically, the computer device can convert the first training fundus image and the second training fundus image from the RGB color space to the LAB color space, and divide the converted first training fundus image and the second training fundus image into small blocks of fixed size.

[0091] Furthermore, the contrast-limited adaptive histogram equalization algorithm is applied to the brightness channel of each small block. Finally, the processed first and second training fundus images are converted back to the RGB color space to obtain the first training fundus image group.

[0092] In an optional embodiment of this application, the computer device may further perform first image processing on the first training fundus image and the second training fundus image in the plurality of training fundus image groups using a linear transformation algorithm and a histogram equalization algorithm to obtain the first training fundus image group.

[0093] In some exemplary embodiments, after obtaining the first training fundus image group, the computer device can also perform a second image processing on the first training fundus image group based on a color normalization algorithm to obtain multiple target training fundus image groups.

[0094] Specifically, the first training fundus image group is processed using a second image processing method based on a color normalization algorithm to obtain multiple target training fundus image groups, which can be represented as follows: Where P refers to the target training fundus image set. This refers to the first set of training fundus images. This refers to applying a Gaussian filter with a standard deviation of s to the first set of training fundus images. , and This refers to adjustable parameters.

[0095] Step 502: Based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial eye age recognition model is trained using the fundus image group trained by each target, the collected information and eye age information, so as to obtain the eye age recognition model.

[0096] For example, the deformation field algorithm refers to using the vascular topology extracted by the vascular segmentation network as a stable spatial framework to guide deformation, thus resisting the influence of sudden changes in local appearance. Vascular structures are very stable in the short term and not easily changed. The overall topological structure of the blood vessels (major branches, bifurcation points) also remains largely unchanged in a short period. Therefore, aligning with blood vessels as the "skeleton" is itself a strong constraint that is insensitive to local lesions. Lesions (such as hemorrhage and exudation) are usually not identified as blood vessels in the vascular segmentation step. Therefore, the deformation field is entirely determined by the healthy, stable vascular morphology.

[0097] In some exemplary embodiments, after obtaining the target training fundus image set, the computer device can train the initial eye age recognition model using the target training fundus image set, the collected information, and the eye age information based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, in order to obtain the eye age recognition model.

[0098] Specifically, an existing, reliable vascular segmentation AI model (e.g., a U-Net-based model) is used to process baseline and follow-up color fundus images, respectively. This yields two binarized vascular mask images—one representing the baseline vascular skeleton. One is a follow-up vascular skeleton. Calculate a smooth deformation field, i.e., an elastic mesh, when this deformation field is applied to follow-up vascular maps. At that time, it can be compared with the baseline vascular map. To the greatest extent possible, they overlap.

[0099] This can be achieved by minimizing an energy function, which can be expressed as: Where D is the similarity term, which can measure... and after deformation The measure of the difference between them, such as the binary cross-entropy or the sum of squared differences, allows the optimizer to continuously adjust the deformation field T to minimize this difference; R(T) is a regularization term that ensures that the deformation field T is smooth and physically reasonable, preventing unrealistic distortions such as tearing or excessive folding. It is a parameter that controls the smoothness.

[0100] Furthermore, the optimal deformation field T can be determined through the above calculations. This field is optimized for the angiogram but represents a spatial transformation model across the entire retinal plane. Applying this same deformation field T to the original, color follow-up images and resampling them generates the final aligned color image. At this point, the anatomical location corresponds precisely to the baseline image.

[0101] A retinal image age causal feature extraction algorithm based on a dual-task learning framework can be used to extract image features from retinal images that are causally related to changes in physiological age. Specifically, through a dual-task learning framework, a shared encoder is constrained to learn features that change stably over time within an individual, thereby filtering out confounding features that are age-related but not causal.

[0102] The model can adopt an architecture of shared encoder + dual prediction heads, as follows:

[0103] Let the baseline image be Follow-up images are All images are retinal images of the same patient. The corresponding label is baseline age. Follow-up age And the age interval Delta a = - .

[0104] The encoder employs a deep convolutional network, such as ResNet-50, or a visual Transformer, such as ViT-Base. It takes a single retinal image as input and outputs a global feature vector. The same encoder processes both baseline and follow-up images to ensure feature space consistency.

[0105] The baseline age prediction header structure consists of two fully connected layers with ReLU activation in between, and outputs the predicted baseline age.

[0106] The age interval prediction head H_d_age has the same structure as H_age. The input is the concatenation (or difference) of baseline features and follow-up features, and the output is the predicted age interval.

[0107] The baseline age prediction loss uses smoothed L1 loss (Huber loss) to reduce the impact of outliers:

[0108] ;

[0109] ;

[0110] The age interval prediction loss also uses smoothed L1 loss:

[0111] ;

[0112] The overall loss function simultaneously optimizes age prediction accuracy and feature causality:

[0113] ;

[0114] By employing a dual-task learning framework and temporal consistency constraints, we can extract purer age-related causal features from retinal images.

[0115] During the prediction phase, the model uses only a shared encoder module and a baseline age prediction head module. The specific process is as follows: A single retinal image is input; firstly, the shared encoder extracts the image's feature vector; then, this feature vector is input into the baseline age prediction head, which directly outputs the predicted age value. In this process, the age interval prediction head module and the temporal consistency constraint module do not participate in the calculation; the entire prediction process follows a single forward propagation path, and age estimation is completed solely based on information from a single image.

[0116] In an exemplary embodiment, the initial age recognition model includes an initial first model and an initial second model. Based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial age recognition model is trained using fundus image groups, acquired information and age information for each target to obtain the age recognition model. The process includes: training the initial first model and the initial second model using fundus image groups, acquired information and age information for each target to obtain the first model and the second model, and determining the age recognition model based on the first model and the second model.

[0117] In one exemplary embodiment, such as Figure 6 As shown, another method for determining eye age is provided, which includes the following steps:

[0118] Step 601: Obtain multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group. Each training fundus image group includes a first training fundus image and a second training fundus image. The acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition times of the second training fundus image.

[0119] Step 602: Obtain the initial eye age recognition model, which includes an initial first model and an initial second model. Then, perform image processing on the first and second training fundus images in multiple training fundus image groups to obtain multiple target training fundus image groups. Each target training fundus image group includes a target training fundus image and a second target training fundus image.

[0120] Step 603: Based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial first model and the initial second model are trained using the fundus image group trained by each target, the collected information and the eye age information to obtain the first model and the second model, and the eye age recognition model is determined according to the first model and the second model. The eye age recognition model includes the first model and the second model.

[0121] Step 604: Obtain the user's original fundus image, and perform a first image processing on the original fundus image based on the contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image; perform a second image processing on the first fundus image based on the color normalization algorithm to obtain a target fundus image;

[0122] Step 605: Input the target fundus image into the first model of the eye age recognition model so that the first model can extract features from the target fundus image to obtain target features, which are features of the natural aging type; input the target features into the second model of the eye age recognition model to obtain the user's eye age information output by the second model.

[0123] It should be understood that although the steps in the flowcharts of the embodiments described above are shown sequentially according to the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the embodiments described above may include multiple steps or multiple stages. These steps or stages are not necessarily completed at the same time, but can be executed at different times. The execution order of these steps or stages is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the steps or stages in other steps. It is understood that the steps in different embodiments can be freely combined as needed, and all non-contradictory solutions formed by such combinations are within the scope of protection of this application.

[0124] Based on the same inventive concept, this application also provides an eye age determination device for implementing the eye age determination method described above. The solution provided by this device is similar to the solution described in the above method. Therefore, the specific limitations of one or more eye age determination device embodiments provided below can be found in the limitations of the eye age determination method above, and will not be repeated here.

[0125] In one exemplary embodiment, such as Figure 7 As shown, an eye age determination device 700 is provided, including: an acquisition module 701 and an execution module 702, wherein:

[0126] The acquisition module is used to acquire the user's original fundus image and perform image processing on the original fundus image to obtain the target fundus image;

[0127] The execution module is used to input the target fundus image into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model;

[0128] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0129] In one embodiment, the eye age recognition model includes a first model and a second model. The execution module 702 is specifically used to input a target fundus image into the first model of the eye age recognition model so that the first model extracts features from the target fundus image to obtain target features, which are features of the natural aging type; and input the target features into the second model of the eye age recognition model to obtain the user's eye age information output by the second model.

[0130] In one embodiment, the acquisition module 701 is specifically used to perform a first image processing on the original fundus image based on a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image; and to perform a second image processing on the first fundus image based on a color normalization algorithm to obtain a target fundus image.

[0131] In one embodiment, the execution module 702 is further configured to acquire multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group, wherein each training fundus image group includes a first training fundus image and a second training fundus image, and the acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition time of the second training fundus image; acquire an initial eye age recognition model, and train the initial eye age recognition model based on multiple training fundus image groups, acquisition information, and eye age information to obtain an eye age recognition model.

[0132] In one embodiment, the execution module 702 is specifically used to perform image processing on the first training fundus image and the second training fundus image in a plurality of training fundus image groups to obtain a plurality of target training fundus image groups, each target training fundus image group including a target training fundus image and a second target training fundus image; based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial eye age recognition model is trained using each target training fundus image group, the acquired information and the eye age information to obtain the eye age recognition model.

[0133] In one embodiment, the initial age recognition model includes an initial first model and an initial second model. The execution module 702 is specifically used to train the initial first model and the initial second model using a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework, by using each target to train fundus image group, collected information and age information, so as to obtain the first model and the second model, and to determine the age recognition model based on the first model and the second model.

[0134] Each module in the aforementioned age determination device can be implemented entirely or partially through software, hardware, or a combination thereof. These modules can be embedded in or independent of the processor in a computer device, or stored in the memory of a computer device as software, so that the processor can call and execute the corresponding operations of each module.

[0135] In one exemplary embodiment, a computer device is provided, which may be a server, and its internal structure diagram may be as follows: Figure 8 As shown, this computer device includes a processor, memory, input / output (I / O) interfaces, and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is also connected to the system bus via the I / O interfaces. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system, computer programs, and a database. The internal memory provides the environment for the operating system and computer programs stored in the non-volatile storage media. The database stores data. The I / O interfaces are used for exchanging information between the processor and external devices. The communication interface is used for communicating with external terminals via a network. When executed by the processor, the computer program implements a method for determining eye age.

[0136] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 9As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, Near Field Communication (NFC), or other technologies. When executed by the processor, the computer program implements a method for determining eye age. The display unit is used to form a visually visible image and can be a display screen, projection device, or virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0137] Those skilled in the art will understand that Figure 8 or Figure 9 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0138] In one exemplary embodiment, a computer device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to perform the following steps:

[0139] The system acquires the user's original fundus image and performs image processing on the original fundus image to obtain the target fundus image.

[0140] The target fundus image is input into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model;

[0141] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0142] In one embodiment, when the processor executes the computer program, it further performs the following steps: inputting a target fundus image into a first model in the eye age recognition model, so that the first model extracts features from the target fundus image to obtain target features, wherein the target features are features of the natural aging type; inputting the target features into a second model in the eye age recognition model to obtain the user's eye age information output by the second model.

[0143] In one embodiment, when the processor executes the computer program, it further performs the following steps: performing a first image processing on the original fundus image based on a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image; and performing a second image processing on the first fundus image based on a color normalization algorithm to obtain a target fundus image.

[0144] In one embodiment, when the processor executes the computer program, it further performs the following steps: acquiring multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group, wherein each training fundus image group includes a first training fundus image and a second training fundus image, and the acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition time of the second training fundus image; acquiring an initial eye age recognition model, and training the initial eye age recognition model based on the multiple training fundus image groups, the acquisition information, and the eye age information to obtain an eye age recognition model.

[0145] In one embodiment, when the processor executes the computer program, it further performs the following steps: image processing on a first training fundus image and a second training fundus image in a plurality of training fundus image groups to obtain a plurality of target training fundus image groups, each target training fundus image group including a target training fundus image and a second target training fundus image; and training an initial eye age recognition model using each target training fundus image group, the acquired information, and the eye age information based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to obtain an eye age recognition model.

[0146] In one embodiment, when the processor executes the computer program, it further implements the following steps: a retinal image age causal feature extraction algorithm based on a deformation field algorithm and a dual-task learning framework, training an initial first model and an initial second model using fundus image groups trained by each target, collected information and eye age information, to obtain a first model and a second model, and determining an eye age recognition model based on the first model and the second model.

[0147] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon, the computer program performing the following steps when executed by a processor:

[0148] The system acquires the user's original fundus image and performs image processing on the original fundus image to obtain the target fundus image.

[0149] The target fundus image is input into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model;

[0150] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0151] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: inputting a target fundus image into a first model in the eye age recognition model, so that the first model extracts features from the target fundus image to obtain target features, wherein the target features are features of the natural aging type; inputting the target features into a second model in the eye age recognition model to obtain the user's eye age information output by the second model.

[0152] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: performing a first image processing on the original fundus image based on a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image; and performing a second image processing on the first fundus image based on a color normalization algorithm to obtain a target fundus image.

[0153] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: acquiring multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group, wherein each training fundus image group includes a first training fundus image and a second training fundus image, and the acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition time of the second training fundus image; acquiring an initial eye age recognition model, and training the initial eye age recognition model based on the multiple training fundus image groups, the acquisition information, and the eye age information to obtain an eye age recognition model.

[0154] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: image processing on a first training fundus image and a second training fundus image in a plurality of training fundus image groups to obtain a plurality of target training fundus image groups, each target training fundus image group including a target training fundus image and a second target training fundus image; and training an initial eye age recognition model using each target training fundus image group, the acquired information, and the eye age information based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to obtain an eye age recognition model.

[0155] In one embodiment, when the computer program is executed by the processor, it further implements the following steps: a retinal image age causal feature extraction algorithm based on a deformation field algorithm and a dual-task learning framework, training an initial first model and an initial second model using fundus image groups trained by each target, collected information and eye age information, to obtain a first model and a second model, and determining an eye age recognition model based on the first model and the second model.

[0156] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, performs the following steps:

[0157] The system acquires the user's original fundus image and performs image processing on the original fundus image to obtain the target fundus image.

[0158] The target fundus image is input into a pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model;

[0159] The eye age recognition model is trained based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework.

[0160] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: inputting a target fundus image into a first model in the eye age recognition model, so that the first model extracts features from the target fundus image to obtain target features, wherein the target features are features of the natural aging type; inputting the target features into a second model in the eye age recognition model to obtain the user's eye age information output by the second model.

[0161] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: performing a first image processing on the original fundus image based on a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image; and performing a second image processing on the first fundus image based on a color normalization algorithm to obtain a target fundus image.

[0162] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: acquiring multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group, wherein each training fundus image group includes a first training fundus image and a second training fundus image, and the acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition time of the second training fundus image; acquiring an initial eye age recognition model, and training the initial eye age recognition model based on the multiple training fundus image groups, the acquisition information, and the eye age information to obtain an eye age recognition model.

[0163] In one embodiment, when the computer program is executed by the processor, it further performs the following steps: image processing on a first training fundus image and a second training fundus image in a plurality of training fundus image groups to obtain a plurality of target training fundus image groups, each target training fundus image group including a target training fundus image and a second target training fundus image; and training an initial eye age recognition model using each target training fundus image group, the acquired information, and the eye age information based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework to obtain an eye age recognition model.

[0164] In one embodiment, when the computer program is executed by the processor, it further implements the following steps: a retinal image age causal feature extraction algorithm based on a deformation field algorithm and a dual-task learning framework, training an initial first model and an initial second model using fundus image groups trained by each target, collected information and eye age information, to obtain a first model and a second model, and determining an eye age recognition model based on the first model and the second model.

[0165] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of the relevant data must comply with relevant regulations.

[0166] Those skilled in the art will understand that all or part of the processes in the methods of the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile memory and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, artificial intelligence (AI) processors, etc., and are not limited to these.

[0167] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this application.

[0168] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this patent application. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this application should be determined by the appended claims.

Claims

1. A method for determining visual age, characterized in that, The method includes: The user's original fundus image is acquired, and the original fundus image is processed to obtain the target fundus image; The target fundus image is input into a pre-trained eye age recognition model to obtain the eye age information of the user output by the eye age recognition model; The age recognition model is trained based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework.

2. The method according to claim 1, characterized in that, The eye age recognition model includes a first model and a second model. The step of inputting the target fundus image into the pre-trained eye age recognition model to obtain the user's eye age information output by the eye age recognition model includes: The target fundus image is input into the first model in the eye age recognition model, so that the first model extracts features from the target fundus image to obtain target features, which are features of natural aging. The target features are input into the second model in the eye age recognition model to obtain the user's eye age information output by the second model.

3. The method according to claim 1, characterized in that, The step of processing the original fundus image to obtain the target fundus image includes: The original fundus image is processed by a contrast-limited adaptive histogram equalization algorithm to obtain a first fundus image. The first fundus image is processed using a color normalization algorithm to obtain the target fundus image.

4. The method according to claim 1, characterized in that, The training process of the eye age recognition model includes: Multiple training fundus image groups, acquisition information of each training fundus image group, and eye age information of each training fundus image group are acquired. Each training fundus image group includes a first training fundus image and a second training fundus image. The acquisition information of each training fundus image group includes the acquisition time of the first training fundus image and the interval between the acquisition time of the second training fundus image. An initial eye age recognition model is obtained, and the initial eye age recognition model is trained based on the multiple training fundus image groups, the acquired information, and the eye age information to obtain the eye age recognition model.

5. The method according to claim 4, characterized in that, The step of training the initial age recognition model based on the multiple sets of training fundus images, the acquired information, and the age information to obtain the age recognition model includes: Image processing is performed on the first training fundus image and the second training fundus image in the plurality of training fundus image groups to obtain a plurality of target training fundus image groups, each of the target training fundus image groups including a target training fundus image and a second target training fundus image; Based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial eye age recognition model is trained using the target training fundus image group, the acquired information and the eye age information to obtain the eye age recognition model.

6. The method according to claim 5, characterized in that, The initial age recognition model includes an initial first model and an initial second model. The retinal image age causal feature extraction algorithm based on the deformation field algorithm and the dual-task learning framework uses the target training fundus image groups, the acquired information, and the age information to train the initial age recognition model to obtain the age recognition model, including: Based on the deformation field algorithm and the retinal image age causal feature extraction algorithm based on the dual-task learning framework, the initial first model and the initial second model are trained using the target training fundus image groups, the acquired information and the eye age information to obtain the first model and the second model, and the eye age recognition model is determined based on the first model and the second model.

7. A device for determining eye age, characterized in that, The device includes: The acquisition module is used to acquire the user's original fundus image and perform image processing on the original fundus image to obtain the target fundus image; The execution module is used to input the target fundus image into a pre-trained eye age recognition model to obtain the eye age information of the user output by the eye age recognition model; The age recognition model is trained based on a deformation field algorithm and a retinal image age causal feature extraction algorithm based on a dual-task learning framework.

8. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1 to 6.

Images