A method and device for screening refraction of children based on multi-modal eye movement patterns

Abstract

The application discloses a kind of children refraction screening method and device based on multi-modal eye movement mode, it is related to ophthalmic refraction screening technical field.The method includes: data acquisition, complete noise suppression, extract the fixation stability index of user, accommodation function index, vergence function index, reading efficiency index, eye movement range index;Based on the index calculation model of preset and combined with age stratification weight, the refraction abnormality risk index and visual function comprehensive score are obtained, and the visual function abnormal type is output, and the referral suggestion is generated and the screening report is generated.By introducing multi-modal eye movement characteristics and age stratification weighted fusion, the consistency and discrimination accuracy of the screening result can be improved, the sensitivity of the model to age difference is improved, and the risk assessment and referral basis that can be explained are provided for clinic.

Description

Technical Field

[0001] This invention relates to the field of ophthalmic refractive screening technology, and in particular to a method and device for pediatric refractive screening based on multimodal eye movement patterns. Background Technology

[0002] Refractive errors in children (including myopia, hyperopia, and astigmatism) are a major problem affecting children's visual health. Globally, the prevalence of myopia among children and adolescents continues to rise, exceeding 80% in some East Asian regions. Refractive errors not only affect children's learning and quality of life but can also lead to secondary visual impairments such as amblyopia and strabismus. Traditional refractive screening mainly relies on equipment such as computerized refractometers and retinoscopes for static refractive power measurement. While this provides accurate refractive parameters, it has the following limitations: traditional methods can only measure static refractive power and cannot assess dynamic visual function. During actual eye use, children's eyes constantly need to adjust (changing the curvature of the lens to focus on objects at different distances) and converge (turning both eyes inward to fixate on near targets). Abnormalities in these dynamic functions are often important factors in eye strain, reading difficulties, and myopia progression, but traditional equipment cannot capture this information. Existing refractive screening procedures rarely include fixation stability assessment. Eye movement range testing is easily overlooked in routine refractive examinations. In summary, the five traditional clinical visual function testing techniques—accommodation, convergence, fixation stability, reading eye movement, and eye movement range—each have significant limitations: they rely on the child's subjective cooperation, assessment is highly subjective, they lack dynamic quantitative indicators and their procedures are not interchangeable, and they suffer from significant systematic errors and poor consistency and longitudinal comparison capabilities. Summary of the Invention

[0003] To address the technical problems of existing technologies, such as reliance on children's subjective cooperation in detection, high subjectivity in assessment, lack of dynamic quantitative indicators, incompatible processes, significant systematic errors, and poor consistency and longitudinal comparison capabilities, this invention provides a method and device for pediatric refractive screening based on multimodal eye movement patterns. The technical solution is as follows:

[0004] On the one hand, a method for pediatric refractive screening based on multimodal eye-tracking patterns is provided. This method is implemented by a pediatric refractive screening device based on multimodal eye-tracking patterns, and includes: S1: Based on the visual presentation device, the user's continuous pupil position is collected, and a multi-dimensional fixation feature vector is extracted and weighted fused to obtain the fixation stability index. The multi-dimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microslap video rate, the fixation retention rate, and the binocular fixation coordination. The weighted fusion is based on the index calculation model of children's age stratification. The index calculation model of children's age stratification adopts a two-layer design. Each layer uses different weights to process the feature vector input for a specific age group and outputs a fixation stability index of 0-100. S2: Using a binocular eye-tracking device, visual targets at varying distances are presented, the user's visual accommodation response is collected, and a weighted fusion is performed to obtain the accommodation function index; S3: Using a binocular eye-tracking device, convergence-induced tasks are conducted to collect convergence motion eye-tracking data streams, which are then weighted and fused to obtain the convergence function index; S4: Based on standardized reading materials that match the reading ability of the test children, real-time eye movement signals of the test children during the execution of the standardized reading task are collected through an eye-tracking device to obtain the reading task eye movement data stream, and then weighted and fused to obtain the reading efficiency index; S5: Presents multi-directional guidance. Through a binocular eye-tracking device, it collects eye movement data of the test children during the tracking and fixation process, and obtains the eye movement limit angle, horizontal eye movement range, vertical eye movement range and eye movement symmetry index in each direction. These data are then weighted and fused to obtain the eye movement range index. S6: Input the fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters into a preset refractive state and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, the preset referral rule engine performs logical judgment to obtain a referral suggestion.

[0005] Preferably, the visual presentation device of S1 collects the user's continuous pupil positions, extracts multi-dimensional fixation feature vectors, and performs weighted fusion to obtain a fixation stability index. The multi-dimensional fixation feature vectors include the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microsplice video rate, the fixation retention rate, and binocular fixation coordination. The weighted fusion is based on an age-stratified index calculation model for children. This model employs a two-layer design, with each layer targeting a specific age group and using different weights to process the feature vector input, outputting a fixation stability index of 0-100, including: S11: Generate fixation task instructions to obtain a multi-stage fixation task sequence, the fixation task sequence including a circular fixation target presented in the center, left and / or right of the screen through a visual presentation device, the circular fixation target including a cartoonish visual design. # According to the fixation task sequence, obtain the first stage visual stimulus, the first stage visual stimulus including the fixation target continuously presented on the left side of the screen. S12: Based on the fixation task sequence, obtain the visual stimulus for the second stage, which includes continuously presenting the fixation target on the right side of the screen. S13: Based on the fixation task sequence, obtain the visual stimulus for the third stage, wherein the visual stimulus for the third stage includes continuously presenting the fixation target in the center of the screen. S14: Using a binocular eye-tracking device, the continuous pupil position of the user during the execution of the fixation task sequence is collected according to a preset sampling frequency to obtain the raw eye-tracking data for the fixation task. S15: Input the raw eye movement data for fixation task into a data calibration algorithm to obtain standardized eye movement data mapped to a standard physical space. The data calibration algorithm includes data mapping based on a personalized affine transformation matrix. S16: Using a feature extraction algorithm, fixation features are calculated based on the standardized eye movement data to obtain a multidimensional fixation feature vector. The multidimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microslap video rate, the fixation retention rate, and the binocular fixation coordination. S17: The multidimensional fixation feature vector is input into a preset index calculation model and weighted fusion is performed to obtain a fixation stability index. The weighted fusion is based on an age-stratified index calculation model for children. The age-stratified index calculation model for children adopts a two-layer design. Each layer processes the feature vector input with different weights for a specific age group and outputs a fixation stability index of 0-100. The index calculation model is expressed as follows: in, For fixed-view stability index, For children's age, This is a multidimensional fixation feature vector, containing five feature values ​​extracted from standardized eye-tracking data. For the weight vector, For the younger age group Each weight, For the older age group Each weight, For the first Each feature value.

[0006] Age group 1: 3-7 years old (young developmental stage): Children in early development have poor attention span, are easily distracted, and have weak micro-jump control. The assessment focuses on attention span and micro-jump frequency control.

[0007] Age group 2: 8-12 years old (older developmental stage): In the later stages of development, children's basic attention is stable, and they enter the stage of fine visual control. The assessment focuses on fixation accuracy and binocular coordination control.

[0008] Preferably, step S2 uses a binocular eye-tracking device to present visual targets at alternating distances, collects the user's visual accommodation response, and performs weighted fusion to obtain an accommodation function index, including: S21: Generate adjustment function instructions to obtain an adjustment function sequence. The adjustment function sequence includes presenting alternating near and far visual targets through a visual presentation device. The visual targets include a far target simulating a distance of more than 5 meters and a near target simulating a distance of 33 cm. The presentation of alternating near and far visual targets includes the alternating presentation of the far target and the near target. Each target lasts for 2-3 seconds, and a total of 8-10 cycles are completed. S22: Based on the aforementioned adjustment function sequence, the user's visual adjustment response is collected using a binocular eye-tracking device to obtain the raw adjustment response data stream; S23: Based on the original regulatory response data stream, a set of time-dimensional features are obtained through time series analysis, including the regulatory response latency and / or regulation completion time; S24: Based on the original regulatory response data stream, a set of non-time dimension features are obtained through response morphology analysis. The non-time dimension features include the regulatory response amplitude and the regulatory flexibility index. S25: Divide the time-dimensional features and the non-time-dimensional features into efficiency indicators and capability indicators. The efficiency indicators include the adjustment flexibility index, adjustment completion time, and adjustment response latency. The capability indicators include the adjustment response amplitude. S26: Based on the efficiency index and the efficiency age coefficient corresponding to the user's age, a weighted calculation is performed to obtain the non-normalized dynamic weight of the efficiency index. The efficiency age coefficient is less than 1 when the user's age is less than a certain threshold and greater than 1 when the user's age is greater than a certain threshold. S27: Based on the capability category indicators and the capability category age coefficient corresponding to the user's age, a weighted calculation is performed to obtain the non-normalized dynamic weight of the capability category indicators; S28: Input the non-normalized dynamic weights of the efficiency index and the non-normalized dynamic weights of the capability index into the normalization processing module for remapping to obtain a set of normalized dynamic weights, the sum of the normalized dynamic weights being 1. S29: Based on the normalized dynamic weights, efficiency indicators, and capability indicators, a regulatory function index is obtained by weighted fusion through a preset index calculation model. The value range of the regulatory function index is 0-100.

[0009] Preferably, in step S3, a binocular eye-tracking device is used to collect convergence motion eye-tracking data streams through a convergence-induced task, and these data are weighted and fused to obtain a convergence function index, including: S31: Generate a convergence function instruction to obtain a convergence function sequence, wherein the convergence function sequence includes a visual target that gradually approaches from a distance through a visual presentation device, wherein the dynamic visual target includes a visual stimulus that simulates an object gradually approaching from 5 meters to 10 centimeters. S32: Based on the convergence function sequence, the convergence motion is collected by a binocular eye-tracking device to obtain a convergence motion eye-tracking data stream, which includes the position coordinates of both eyes at continuous time points and / or pupil information. S33: Based on the original eye movement data stream, after filtering and / or calibration processing, the convergence trajectory is obtained, which characterizes the convergence process of the visual axes of both eyes over time; S34: Using the limit analysis algorithm, the convergence trajectory is analyzed to obtain the convergence near point parameters, which are used to characterize the closest distance that can maintain binocular single vision; S35: Based on the starting position of the convergence trajectory and the convergence proximal point, the convergence amplitude is obtained through spatial amplitude calculation. The convergence amplitude is used to characterize the total amplitude of the inward rotation of the eyes. S36: The convergence trajectory is processed by differential operation and / or spectrum analysis to obtain the dynamic parameters of the convergence motion, wherein the dynamic parameters include the peak convergence velocity and / or the convergence smoothness index; S37: Input the convergence proximity parameters, the convergence amplitude, and the dynamic parameters into the feature integration model to obtain a multi-dimensional convergence feature parameter set, which is used to comprehensively characterize the steady-state and / or transient characteristics of the convergence motion. S38: Based on the convergence feature parameter set, a convergence function index is obtained by weighted fusion through a preset index calculation model. The value range of the convergence function index is 0-100.

[0010] Preferably, in step S4, based on standardized reading materials matched to the reading abilities of the test children, an eye-tracking device is used to collect real-time eye-tracking signals during the standardized reading task performed by the test children, obtaining a reading task eye-tracking data stream, which is then weighted and fused to obtain a reading efficiency index, including: S41: Collect the age and grade information of the test children to obtain user physiological parameters, which are used to match the difficulty level of the reading material; S42: Based on the user's physiological parameters, standardized reading materials matching the reading ability of the tested children are obtained through a preset difficulty matching rule library; S43: Using the standardized reading material, an eye-tracking device is used to collect real-time eye movement signals of the test children during the standardized reading task, and the reading task eye movement data stream is obtained; S44: Based on the eye-tracking data stream of the reading task, a structured eye-tracking event sequence is obtained after preprocessing and an event recognition algorithm. The eye-tracking event sequence includes a fixation point sequence and a saccade path sequence. S45: Using the structured eye movement event sequence, eye movement features are extracted to obtain initial eye movement feature data, which includes the duration of each fixation, the amplitude and direction of the saccade, and the number of characters read per unit time. S46: Based on the initial eye movement feature data, a quantized eye movement feature set is obtained through statistical calculation. The quantized eye movement feature set includes average fixation duration, forward saccade amplitude, retrograde ratio, and reading speed. S47: The quantified eye-tracking feature set is weighted and fused using a preset index calculation model to obtain a reading efficiency index, wherein the reading efficiency index ranges from 0 to 100.

[0011] Preferably, the S5 presentation involves multi-directional guidance, which uses a binocular eye-tracking device to collect eye movement data of the child during the tracking and fixation process. This data yields the eye movement limit angles, horizontal eye movement range, vertical eye movement range, and eye movement symmetry index in each direction. These are then weighted and fused to obtain the eye movement range index, which includes: S51: Generate a tracking gaze process instruction to obtain a tracking gaze process sequence, wherein the tracking gaze process sequence includes multi-directional guidance presented by the system through a visual presentation device, wherein the multi-directional guidance includes sequentially guiding the subject child to track gaze in eight directions: up, down, left, right, upper left, lower left, upper right, and / or lower right. S52: Based on the aforementioned gaze tracking process sequence, eye movement data of the test child during the gaze tracking process is collected using a binocular eye tracking device to obtain an eye movement trajectory data stream of the gaze tracking process; S53: Based on the eye movement trajectory data stream of the following gaze process, the eye movement limit angles in each direction are obtained through the angle extraction algorithm. The eye movement limit angles in each direction include the maximum amplitude values ​​of eye movements corresponding to the eight guiding directions. S54: Calculate the horizontal eye movement range based on the eye movement limit angles in the left and right directions, wherein the horizontal eye movement range is the sum of the eye movement limit angles in the left and right directions; S55: Calculate the vertical eye movement range based on the eye movement limit angles in the upward and downward directions, wherein the vertical eye movement range is the sum of the eye movement limit angles in the two directions; S56: Based on the eye movement limit angle in the symmetrical direction, the eye movement symmetry index is calculated using the symmetry evaluation model. The eye movement symmetry index is used to characterize the consistency of the eye movement range in the symmetrical direction. S57: Input the eye movement limit angles, horizontal eye movement range, vertical eye movement range, and eye movement symmetry index in each direction into the feature integration unit to obtain a comprehensive eye movement feature set; S58: Based on the comprehensive eye movement feature set, an eye movement range index is obtained by weighted fusion through a preset index calculation model. The value range of the eye movement range index includes 0-100.

[0012] Preferably, in step S6, the fixation stability index, accommodative function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters are input into a preset refractive state and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, a preset referral rule engine performs logical judgment to obtain referral suggestions, including: S61: Based on the fixation stability index, the accommodation function index, the convergence function index, the reading efficiency index, the eye movement range index, visual acuity data, refractive error data, and user physiological parameters, a comprehensive dataset is obtained; S62: Based on a preset feature fusion model, feature fusion is performed on the comprehensive dataset to obtain fused features. The fused features are used to characterize the intrinsic relationship between refractive state and visual function abnormality. The preset feature fusion model adopts a multi-task deep learning framework. S63: Input the fused features into the preset refractive state and visual function abnormality screening model to obtain the refractive abnormality risk index, visual function comprehensive score and visual function abnormality type. The refractive abnormality risk index is used to quantitatively assess the risk level of refractive error and its related visual function abnormalities. The visual function comprehensive score includes a weighted comprehensive score of five dimensions of visual function index. The preset refractive state and visual function abnormality screening model adopts an attention mechanism. S64: Based on the refractive error risk index, visual function comprehensive score, and visual function abnormality type, a referral rule engine performs logical judgment to obtain a referral suggestion, which includes referral priority and / or recommended further examination items.

[0013] On the other hand, a pediatric refractive screening device based on multimodal eye-tracking patterns is provided. This device is applied to a pediatric refractive screening method based on multimodal eye-tracking patterns. The device includes: Fixation Feature Module: Based on the visual presentation device, this module collects the user's continuous pupil position, extracts multi-dimensional fixation feature vectors, and performs weighted fusion to obtain a fixation stability index. The multi-dimensional fixation feature vectors include the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microscratching video rate, the fixation retention rate, and binocular fixation coordination. The weighted fusion is based on an index calculation model for children's age stratification. This model adopts a two-layer design, with each layer targeting a specific age group and using different weights to process the feature vector input, outputting a fixation stability index of 0-100. The accommodation function module is used to present visual targets that are either near or far through a binocular eye-tracking device, collect the user's visual accommodation response, and perform weighted fusion to obtain the accommodation function index. Convergence Function Module: Used to collect convergence motion eye-tracking data streams through convergence-induced tasks using binocular eye-tracking devices, and perform weighted fusion to obtain the convergence function index; Reading efficiency module: Based on standardized reading materials that match the reading ability of the test children, it uses an eye-tracking device to collect real-time eye movement signals during the test children's performance of the standardized reading tasks, obtains the reading task eye movement data stream, and performs weighted fusion to obtain the reading efficiency index; Eye movement range module: used to present multi-directional guidance. Through binocular eye movement tracking device, it collects eye movement data of the test children during the following and fixation process, and obtains the eye movement limit angle, horizontal eye movement range, vertical eye movement range and eye movement symmetry index in each direction. The data are then weighted and fused to obtain the eye movement range index. Risk integration module: This module is used to input the fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters into a preset refractive status and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, it performs logical judgment through a preset referral rule engine to obtain a referral recommendation.

[0014] On the other hand, a pediatric refractive screening device based on multimodal eye-tracking patterns is provided. The pediatric refractive screening device based on multimodal eye-tracking patterns includes: a processor; a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method described in any one of the above-described methods for pediatric refractive screening based on multimodal eye-tracking patterns is implemented.

[0015] On the other hand, a computer-readable storage medium is provided, characterized in that the computer-readable storage medium stores program code, which can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

[0016] The beneficial effects of the technical solution provided by this invention include at least the following: First, it achieves integrated assessment of multi-dimensional visual function, integrating and monitoring five dimensions of visual function assessment: fixation stability, accommodation function, convergence function, reading pattern, and eye movement range. Second, it provides objective and quantitative visual function assessment indicators. This invention objectively records eye movement data using eye-tracking technology, employs standardized feature extraction and machine learning assessment algorithms, and outputs quantitative functional indices, reducing reliance on examiner experience and subject subjective responses, and improving the consistency and repeatability of results. Third, it provides denoising methods to complement the algorithm and eliminate systematic errors. Finally, it improves age sensitivity. By using a hierarchical modeling approach for different age groups, the accuracy of the model can be further improved.

[0017] Specifically, this invention integrates visual function assessment across five dimensions—fixation stability, accommodation, convergence, reading patterns, and eye movement range—on a single wearable platform. Combined with visual acuity and refractive error data, it upgrades the process from a single refractive examination to a comprehensive multimodal visual function assessment. Children only need to wear the device once to complete a comprehensive visual function screening within 3-4 minutes, significantly improving screening efficiency and information density. Compared to traditional subjective examination methods, this invention objectively records eye movement data using eye-tracking technology, employs standardized feature extraction and machine learning evaluation algorithms, and outputs quantitative functional indices. This reduces reliance on examiner experience and subject subjective responses, improving the consistency and repeatability of results. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0019] Figure 1 This is a flowchart of a pediatric refractive screening method based on multimodal eye movement patterns provided by an embodiment of the present invention; Figure 2 This is a flowchart of a comprehensive risk assessment method provided by an embodiment of the present invention; Figure 3 This is a block diagram of a pediatric refractive screening device based on multimodal eye movement patterns provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the structure of a pediatric refractive screening device based on multimodal eye movement patterns provided in an embodiment of the present invention. Detailed Implementation

[0020] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0021] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0022] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0023] In this embodiment of the invention, sometimes a subscript such as W1 may be written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0024] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0025] This invention provides a method for pediatric refractive error screening based on multimodal eye-tracking patterns. This method can be implemented using a pediatric refractive error screening device based on multimodal eye-tracking patterns, which can be a terminal or a server. For example... Figure 1 The flowchart shown is for a pediatric refractive error screening method based on multimodal eye movement patterns. The processing flow of this method may include the following steps:

[0026] Based on the visual presentation device, the user's continuous pupil position is collected, and a multidimensional fixation feature vector is extracted and weighted fused to obtain the fixation stability index. The multidimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microscratching video rate, the fixation retention rate, and the binocular fixation coordination. The weighted fusion is based on the index calculation model of children's age stratification. The index calculation model of children's age stratification adopts a two-layer design. Each layer uses different weights to process the feature vector input for a specific age group and outputs a fixation stability index of 0-100. Preferably, based on a visual presentation device, the user's continuous pupil positions are collected, and a multidimensional fixation feature vector is extracted and weighted fused to obtain a fixation stability index. The multidimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microsaccade rate, the fixation retention rate, and binocular fixation coordination. The weighted fusion is based on an age-stratified index calculation model for children. This model employs a two-layer design, with each layer targeting a specific age group and using different weights to process the feature vector input, outputting a fixation stability index of 0-100, including: Generate fixation task instructions to obtain a multi-stage fixation task sequence, the fixation task sequence including a circular fixation target presented in the center, left and / or right of the screen through a visual presentation device, the circular fixation target including a cartoonish visual design. # According to the fixation task sequence, obtain the first stage visual stimulus, the first stage visual stimulus including the fixation target being continuously presented on the left side of the screen. Based on the fixation task sequence, a second-stage visual stimulus is obtained, which includes a fixation target continuously presented on the right side of the screen. Based on the fixation task sequence, a third-stage visual stimulus is obtained, which includes a fixation target continuously presented in the center of the screen. Using a binocular eye-tracking device, the continuous pupil position of the user during the fixation task sequence is collected according to a preset sampling frequency to obtain raw eye-tracking data for the fixation task. The raw eye-tracking data for fixation tasks is input into a data calibration algorithm to obtain standardized eye-tracking data mapped to a standard physical space. The data calibration algorithm includes data mapping based on a personalized affine transformation matrix. A feature extraction algorithm is used to calculate fixation features based on the standardized eye movement data to obtain a multidimensional fixation feature vector. The multidimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microsaccade rate, the fixation retention rate, and the binocular fixation coordination. The multidimensional fixation feature vector is input into a preset index calculation model and weighted and fused to obtain a fixation stability index. The weighted fusion is based on an age-stratified index calculation model for children. This age-stratified index calculation model adopts a two-layer design, with each layer processing the feature vector input with different weights for a specific age group, and outputting a fixation stability index of 0-100. The index calculation model is expressed as follows: in, For fixed-view stability index, For children's age, This is a multidimensional fixation feature vector, containing five feature values ​​extracted from standardized eye-tracking data. For the weight vector, For the younger age group Each weight, For the older age group Each weight, For the first Each feature value.

[0027] In some embodiments, the wearable eye-tracking platform includes a binocular near-infrared camera (frame rate 60-120fps, resolution ≥640×480 pixels), an infrared LED fill light source (wavelength 850nm), a visual stimulus display screen (refresh rate ≥60Hz), and a main control computing unit.

[0028] It should be noted that the original eye-tracking images undergo preprocessing, including pupil detection, pupil center localization, canthal point recognition, and eye position vector calculation. An optimized pupil-canthal vector algorithm is employed, using the inner canthus-outer canthus line as a reference, to calculate the normalized position of the pupil center, eliminating the influence of individual differences in palpebral fissure size and eyeball size.

[0029] It should be further explained that the system presents a circular fixation target with a diameter of approximately 0.5° (which can be cartoonish to attract children's attention) in the center of the screen, and the children are required to fixate on the target for 10-15 seconds. During the task, the system continuously records the position of both pupils at a sampling rate of 60-120Hz.

[0030] Preferably, feature extraction: the following features are extracted from the eye-tracking data of the fixation task: Binocular fixation ellipse area (BFEA): The area of ​​fixation points covered by the 95% confidence ellipse, measured in square degrees.

[0031] Fixation position standard deviation (FPSD): The standard deviation of fixation position in the horizontal and vertical directions.

[0032] Microspar video rate (MSF): The frequency of microspar video (rapid eye movements with an amplitude of <1°) during fixation.

[0033] Fixation retention rate (FMR): The proportion of time the fixation point is within 2° of the center of the target.

[0034] Binocular fixation coordination (BFC): the correlation coefficient and phase difference between the fixation positions of the two eyes.

[0035] Using a binocular eye-tracking device, visual targets at varying distances are presented, the user's visual accommodation response is collected, and the data is weighted and fused to obtain the accommodation function index. Preferably, a binocular eye-tracking device is used to present visual targets at alternating distances, collect the user's visual accommodation response, and perform weighted fusion to obtain an accommodation function index, including: An adjustment function instruction is generated to obtain an adjustment function sequence. The adjustment function sequence includes presenting alternating visual targets at far and near distances through a visual presentation device. The visual targets include a far target simulating a distance of more than 5 meters and a near target simulating a distance of 33 cm. The presentation of the alternating visual targets at far and near distances includes the alternating presentation of the far target and the near target. Each target lasts for 2-3 seconds, and a total of 8-10 cycles are completed. Based on the aforementioned accommodation function sequence, the user's visual accommodation response is collected using a binocular eye-tracking device to obtain the raw accommodation response data stream; Based on the original regulatory response data stream, a set of time-dimensional features are obtained through time series analysis, including the regulatory response latency and / or regulation completion time. Based on the original regulatory response data stream, a set of non-time dimension features are obtained through response morphology analysis. The non-time dimension features include the regulatory response amplitude and the regulatory flexibility index. The time-dimensional features and the non-time-dimensional features are divided into efficiency indicators and capability indicators. The efficiency indicators include the adjustment flexibility index, adjustment completion time, and adjustment response latency. The capability indicators include the adjustment response amplitude. Based on the efficiency index and the efficiency age coefficient corresponding to the user's age, a weighted calculation is performed to obtain the non-normalized dynamic weight of the efficiency index. The efficiency age coefficient is less than 1 when the user's age is less than a certain threshold, and greater than 1 when the user's age is greater than a certain threshold. The non-normalized dynamic weights of the capability indicators are obtained by weighting the capability indicators and the capability age coefficient corresponding to the user's age. The non-normalized dynamic weights of the efficiency indicators and the non-normalized dynamic weights of the capability indicators are input into the normalization processing module for remapping to obtain a set of normalized dynamic weights, the sum of which is 1. Based on the normalized dynamic weights, efficiency indicators, and capability indicators, a regulatory function index is obtained by weighted fusion through a preset index calculation model. The value range of the regulatory function index is 0-100.

[0036] In some embodiments, the system presents visual targets at alternating distances, with distant targets simulating a distance of more than 5 meters (adjustment requirement approximately 0D) and near targets simulating a distance of 33cm (adjustment requirement approximately 3D). The distant and near targets are presented alternately, with each target lasting 2-3 seconds, for a total of 8-10 cycles.

[0037] It should be noted that feature extraction: Accommodation response amplitude (ARA): The magnitude of change in pupil diameter when switching between near and far fixation.

[0038] Regulation response latency (ARL): The time interval from the change in target distance to the onset of the regulation response.

[0039] Adjustment completion time (ACT): The time from the start of the adjustment reaction to reaching steady state.

[0040] Adjustment Flexibility Index (AFX): The speed and smoothness of continuous switching between near and far adjustment.

[0041] Using a binocular eye-tracking device, convergence-induced tasks are conducted to collect convergence motion eye-tracking data streams, which are then weighted and fused to obtain the convergence function index. Preferably, using a binocular eye-tracking device, convergence motion eye-tracking data streams are collected through a convergence-induced task, and weighted fusion is performed to obtain a convergence function index, including: Generate a convergence function instruction to obtain a convergence function sequence. The convergence function sequence includes a visual target that gradually approaches from a distance through a visual presentation device. The dynamic visual target includes a visual stimulus that simulates an object gradually approaching from 5 meters to 10 centimeters. According to the convergence function sequence, the convergence motion is collected by a binocular eye-tracking device to obtain a convergence motion eye-tracking data stream, which includes the position coordinates of both eyes at continuous time points and / or pupil information. Based on the original eye movement data stream, after filtering and / or calibration, the convergence trajectory is obtained, which characterizes the convergence process of the visual axes of both eyes over time. The convergence trajectory is analyzed using a limit analysis algorithm to obtain convergence near point parameters, which are used to characterize the closest distance that can maintain binocular single vision. Based on the starting position of the convergence trajectory and the convergence proximal point, the convergence amplitude is obtained through spatial amplitude calculation. The convergence amplitude is used to characterize the total amplitude of the inward rotation of the eyes. Differential operations and / or spectral analysis are used to process the convergence trajectory to obtain the dynamic parameters of the convergence motion, including the peak convergence velocity and / or the convergence smoothness index. The convergence proximity parameters, convergence amplitude, and dynamic parameters are input into the feature integration model to obtain a multi-dimensional convergence feature parameter set, which is used to comprehensively characterize the steady-state and / or transient characteristics of the convergence motion. Based on the convergence feature parameter set, a convergence function index is obtained by weighted fusion through a preset index calculation model. The value range of the convergence function index is 0-100.

[0042] In some embodiments, the system presents a visual target that gradually approaches from a distance (simulating an object that gradually approaches from 5 meters to 10 cm), inducing convergence movements in the children being tested.

[0043] It should be noted that feature extraction: Convergence Point (NPC): The closest distance at which binocular single vision can be maintained.

[0044] Convergence Amplitude (VA): The total inward rotation of the eyes when shifting from distant to near fixation.

[0045] Peak convergence velocity (PVV): The maximum angular velocity during convergence.

[0046] Convergence Smoothness Index (VSI): The smoothness of the convergence trajectory.

[0047] Based on standardized reading materials that match the reading abilities of the test children, real-time eye movement signals of the test children during the execution of the standardized reading task are collected through an eye-tracking device to obtain the reading task eye movement data stream, and then weighted and fused to obtain the reading efficiency index; Preferably, based on standardized reading materials that match the reading abilities of the test children, real-time eye movement signals are collected during the test children's performance of the standardized reading task using an eye-tracking device to obtain a reading task eye movement data stream, which is then weighted and fused to obtain a reading efficiency index, including: The age and grade information of the children in the test were collected to obtain user physiological parameters, which were used to match the difficulty level of the reading materials; Based on the user's physiological parameters, standardized reading materials matching the reading ability of the tested children are obtained through a preset difficulty matching rule library. Using the standardized reading materials, real-time eye movement signals of the test children were collected during the standardized reading task using an eye-tracking device to obtain the reading task eye movement data stream; Based on the eye-tracking data stream of the reading task, a structured eye-tracking event sequence is obtained after preprocessing and event recognition algorithms. The eye-tracking event sequence includes a fixation point sequence and a saccade path sequence. Using the structured eye movement event sequence, eye movement features are extracted to obtain initial eye movement feature data, which includes the duration of each fixation, the amplitude and direction of the saccade, and the number of characters read per unit time. Based on the initial eye movement feature data, a quantized eye movement feature set is obtained through statistical calculation. The quantized eye movement feature set includes average fixation duration, forward saccade amplitude, retrograde ratio, and reading speed. The quantified eye-tracking feature set is weighted and fused using a preset index calculation model to obtain a reading efficiency index, the value of which ranges from 0 to 100.

[0048] In some embodiments, feature extraction: Mean fixation duration (MFD): The average duration of each fixation during reading.

[0049] Forward saccadic amplitude (FSA): The average amplitude of saccadic movements in the reading direction.

[0050] Regression Ratio (RR): The proportion of reverse reading direction scans out of the total number of scans.

[0051] Reading speed (RS): The number of characters read per unit of time.

[0052] The system provides multi-directional guidance and collects eye movement data of the children during the tracking and fixation process using a binocular eye-tracking device. This data includes the eye movement limit angle, horizontal eye movement range, vertical eye movement range, and eye movement symmetry index in each direction. The data are then weighted and fused to obtain the eye movement range index. Preferably, multi-directional guidance is presented. Eye movement data of the tested children during the tracking and fixation process is collected using a binocular eye-tracking device. This yields the eye movement limit angles, horizontal eye movement range, vertical eye movement range, and eye movement symmetry index in each direction. These are then weighted and fused to obtain the eye movement range index, which includes: Generate a gaze tracking process instruction to obtain a gaze tracking process sequence. The gaze tracking process sequence includes multi-directional guidance presented by the system through a visual presentation device. The multi-directional guidance includes sequentially guiding the subject child to gaze in eight directions: up, down, left, right, upper left, lower left, upper right, and / or lower right. Based on the aforementioned gaze tracking sequence, eye movement data of the test child during the gaze tracking process is collected using a binocular eye tracking device to obtain an eye movement trajectory data stream of the gaze tracking process; Based on the eye movement trajectory data stream of the following gaze process, the eye movement limit angles in each direction are obtained through angle extraction algorithm processing. The eye movement limit angles in each direction include the maximum amplitude values ​​of eye movement corresponding to eight guiding directions. The horizontal eye movement range is calculated based on the eye movement limit angles in the left and right directions, and the horizontal eye movement range is the sum of the eye movement limit angles in the left and right directions; The vertical eye movement range is calculated based on the eye movement limit angles in the upward and downward directions, and the vertical eye movement range is the sum of the eye movement limit angles in the two directions. Based on the eye movement limit angle in the symmetrical direction, the eye movement symmetry index is calculated using the symmetry evaluation model. The eye movement symmetry index is used to characterize the consistency of the eye movement range in the symmetrical direction. The eye movement limit angles in each direction, the horizontal eye movement range, the vertical eye movement range, and the eye movement symmetry index are input into the feature integration unit to obtain a comprehensive eye movement feature set. Based on the comprehensive eye movement feature set, an eye movement range index is obtained by weighted fusion through a preset index calculation model. The value range of the eye movement range index includes 0-100.

[0053] In some embodiments, feature extraction: Degrees of eye movement (DME): The maximum range of eye movement in eight directions.

[0054] Horizontal eye movement range (HRM): The sum of the limits of eye movement in both the left and right directions.

[0055] Vertical eye movement range (VRM): The sum of the limits of eye movement in both the vertical and vertical directions.

[0056] Eye Movement Symmetry Index (OMSI): Consistency of eye movement range in symmetrical directions.

[0057] The fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters are input into a preset refractive state and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, a preset referral rule engine performs logical judgment to obtain a referral recommendation.

[0058] Preferably, the fixation stability index, accommodative function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters are input into a preset refractive status and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, a preset referral rule engine performs logical judgment to obtain referral suggestions, including: A comprehensive dataset is obtained based on the fixation stability index, the accommodation function index, the convergence function index, the reading efficiency index, the eye movement range index, visual acuity data, refractive error data, and user physiological parameters. Based on a preset feature fusion model, features are fused on a comprehensive dataset to obtain fused features. These fused features are used to characterize the intrinsic relationship between refractive state and visual function abnormalities. The preset feature fusion model adopts a multi-task deep learning framework. The fused features are input into a preset refractive state and visual function abnormality screening model to obtain a refractive abnormality risk index, a comprehensive visual function score, and a visual function abnormality type. The refractive abnormality risk index is used to quantitatively assess the risk level of refractive errors and their related visual function abnormalities. The comprehensive visual function score includes a weighted comprehensive score of five dimensions of visual function indices. The preset refractive state and visual function abnormality screening model adopts an attention mechanism. like Figure 2 As shown, based on the refractive error risk index, visual function comprehensive score, and visual function abnormality type, a pre-set referral rule engine performs logical judgment to obtain referral suggestions, which include referral priority and / or recommended further examination items.

[0059] In some embodiments, the preset refractive state and visual function abnormality screening model adopts a cascaded design, with the core consisting of a multi-task feature fusion network and an attention screening network.

[0060] The first part is a multi-task deep learning feature fusion framework. It maps the input heterogeneous comprehensive dataset (including features such as fixation stability, accommodation / convergence function, reading efficiency, eye movement range, visual acuity, refractive error, and user physiological parameters) into a unified, highly correlated, high-dimensional feature representation, specifically including:

[0061] Feature preprocessing and embedding layer: Various numerical and categorical input features are processed by normalization or embedding layer to transform them into initial feature vectors of uniform dimension.

[0062] Shared Feature Encoder: All initial feature vectors are concatenated and fed into a shared backbone network consisting of multiple fully connected layers. This part is the core of the model, and its parameter updates are jointly driven by multiple auxiliary learning tasks.

[0063] Multi-task learning head: Several parallel, relatively simple task-specific output layers are connected on top of a shared backbone network to form a multi-task learning framework. These auxiliary tasks can be set as preliminary regression predictions or related binary classification tasks for the original visual function indices. The purpose is not to obtain the final output, but to constrain and guide the shared backbone network to learn by backpropagating gradient signals from different tasks, ensuring that the learning generally reflects the fusion characteristics of the visual function system's collaborative and compensatory mechanisms.

[0064] The second part is the screening model based on the attention mechanism. This model takes fused features as input and performs the final refined screening and assessment task. The introduction of the attention mechanism aims to simulate the differentiated attention paid to different signs by clinical experts during comprehensive assessment. Specifically, it includes: an attention weight generation module: this module is a lightweight feedforward neural network that outputs an attention weight vector of the same dimension, while multiplying the fused features element-wise with the attention weights. This operation dynamically scales the original features, strengthening key features and weakening irrelevant or noisy features. Multi-objective output layer: the input is fed to the final multi-branch output layer. Each branch consists of fully connected layers, responsible for generating: refractive error risk index, comprehensive visual function score, and visual function abnormality type.

[0065] The above is an introduction to the method embodiments. The following describes the solution described in this application through device embodiments.

[0066] Figure 3 This is a block diagram illustrating a pediatric refractive screening device based on multimodal eye-tracking patterns, according to an exemplary embodiment. The device is used in a pediatric refractive screening method based on multimodal eye-tracking patterns. (Refer to...) Figure 3 The device includes a fixation feature module, an accommodation function module, a convergence function module, a reading efficiency module, an eye movement range module, and a risk integration module.

[0067] Fixation Feature Module: Based on the visual presentation device, this module collects the user's continuous pupil position, extracts multi-dimensional fixation feature vectors, and performs weighted fusion to obtain a fixation stability index. The multi-dimensional fixation feature vectors include the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microscratching video rate, the fixation retention rate, and binocular fixation coordination. The weighted fusion is based on an index calculation model for children's age stratification. This model adopts a two-layer design, with each layer targeting a specific age group and using different weights to process the feature vector input, outputting a fixation stability index of 0-100. The accommodation function module is used to present visual targets that are either near or far through a binocular eye-tracking device, collect the user's visual accommodation response, and perform weighted fusion to obtain the accommodation function index. Convergence Function Module: Used to collect convergence motion eye-tracking data streams through convergence-induced tasks using binocular eye-tracking devices, and perform weighted fusion to obtain the convergence function index; Reading efficiency module: Based on standardized reading materials that match the reading ability of the test children, it uses an eye-tracking device to collect real-time eye movement signals during the test children's performance of the standardized reading tasks, obtains the reading task eye movement data stream, and performs weighted fusion to obtain the reading efficiency index; Eye movement range module: used to present multi-directional guidance. Through binocular eye movement tracking device, it collects eye movement data of the test children during the following and fixation process, and obtains the eye movement limit angle, horizontal eye movement range, vertical eye movement range and eye movement symmetry index in each direction. The data are then weighted and fused to obtain the eye movement range index. Risk integration module: This module is used to input the fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters into a preset refractive status and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, it performs logical judgment through a preset referral rule engine to obtain a referral recommendation.

[0068] A pediatric refractive screening device based on multimodal eye-tracking patterns, the pediatric refractive screening device based on multimodal eye-tracking patterns includes: a processor; a memory, the memory storing computer-readable instructions, which, when executed by the processor, implement the method described in any of the above-described methods for pediatric refractive screening based on multimodal eye-tracking patterns.

[0069] A computer-readable storage medium, characterized in that the computer-readable storage medium stores program code, the program code being invoked by a processor to execute the method as described in any one of claims 1 to 7.

[0070] Figure 4 This is a schematic diagram of the structure of a pediatric refractive screening device based on multimodal eye movement patterns provided in an embodiment of the present invention, as shown below. Figure 4 As shown, pediatric refractive screening devices based on multimodal eye-tracking patterns may include the above-mentioned... Figure 3 The illustrated device is a pediatric refractive screening device based on multimodal eye movement patterns. Optionally, the pediatric refractive screening device 410 based on multimodal eye movement patterns may include a first processor 2001.

[0071] Optionally, the pediatric refractive screening device 410 based on multimodal eye movement patterns may also include a memory 2002 and a transceiver 2003.

[0072] The first processor 2001, memory 2002, and transceiver 2003 can be connected via a communication bus.

[0073] The following is combined with Figure 4 A detailed description of each component of the pediatric refractive screening device 410 based on multimodal eye movement patterns is provided below: The first processor 2001 is the control center of the pediatric refractive screening device 410 based on multimodal eye movement patterns. It can be a single processor or a collective term for multiple processing elements. For example, the first processor 2001 can be one or more central processing units (CPUs), application-specific integrated circuits (ASICs), or one or more integrated circuits configured to implement embodiments of the present invention, such as one or more digital signal processors (DSPs), or one or more field-programmable gate arrays (FPGAs).

[0074] Optionally, the first processor 2001 can perform various functions of the pediatric refractive screening device 410 based on multimodal eye movement patterns by running or executing software programs stored in the memory 2002 and calling data stored in the memory 2002.

[0075] In a specific implementation, as one example, the first processor 2001 may include one or more CPUs, for example... Figure 4 CPU0 and CPU1 are shown in the diagram.

[0076] In a specific implementation, as one example, the pediatric refractive screening device 410 based on multimodal eye-tracking patterns may also include multiple processors, such as... Figure 4 The first processor 2001 and the second processor 2004 are shown in the diagram. Each of these processors can be a single-core processor or a multi-core processor. Here, a processor can refer to one or more devices, circuits, and / or processing cores used to process data (such as computer program instructions).

[0077] The memory 2002 is used to store the software program that executes the present invention, and is controlled by the first processor 2001 to execute it. The specific implementation method can be referred to the above method embodiment, and will not be repeated here.

[0078] Optionally, the memory 2002 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. The memory 2002 may be integrated with the first processor 2001 or may exist independently, and may be connected via the interface circuit of the multimodal eye-tracking pattern-based pediatric refractive screening device 410. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0079] The transceiver 2003 is used to communicate with network devices or with terminal devices.

[0080] Alternatively, transceiver 2003 may include a receiver and a transmitter. Figure 4 (Not shown separately). The receiver is used to implement the receiving function, and the transmitter is used to implement the transmitting function.

[0081] Optionally, the transceiver 2003 can be integrated with the first processor 2001 or exist independently, and can be connected via the interface circuit of the pediatric refractive screening device 410 based on multimodal eye-tracking patterns. Figure 4 (Not shown in the image) is coupled to the first processor 2001, and this embodiment of the invention does not specifically limit this.

[0082] It should be noted that, Figure 4 The structure of the pediatric refractive screening device 410 based on multimodal eye movement patterns shown in the figure does not constitute a limitation on the router. Actual knowledge structure recognition devices may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0083] Furthermore, the technical effects of the pediatric refractive screening device 410 based on multimodal eye-tracking patterns can be referred to the technical effects of the pediatric refractive screening method based on multimodal eye-tracking patterns described in the above method embodiments, and will not be repeated here.

[0084] It should be understood that the first processor 2001 in the embodiments of the present invention may be a central processing unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0085] It should also be understood that the memory in the embodiments of the present invention can be volatile memory or non-volatile memory, or may include both volatile and non-volatile memory. The non-volatile memory can be read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), or flash memory. The volatile memory can be random access memory (RAM), which is used as an external cache. By way of example, but not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate synchronous DRAM (DDR SDRAM), enhanced synchronous DRAM (ESDRAM), synchronous linked DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0086] The above embodiments can be implemented, in whole or in part, by software, hardware (such as circuits), firmware, or any other combination thereof. When implemented using software, the above embodiments can be implemented, in whole or in part, as a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that includes one or more sets of available media. The available medium can be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium. A semiconductor medium can be a solid-state drive.

[0087] It should be understood that the term "and / or" in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. A and B can be singular or plural. Additionally, the character " / " in this article generally indicates an "or" relationship between the preceding and following related objects, but it can also represent an "and / or" relationship. Please refer to the context for a more accurate understanding.

[0088] In this invention, "at least one" means one or more, and "more than one" means two or more. "At least one of the following" or similar expressions refer to any combination of these items, including any combination of a single item or a plurality of items. For example, at least one of a, b, or c can represent: a, b, c, ab, ac, bc, or abc, where a, b, and c can be a single item or multiple items.

[0089] It should be understood that, in various embodiments of the present invention, the order of the above-mentioned process numbers does not imply the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0090] Those skilled in the art will recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0091] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0092] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another device, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some interfaces; the indirect coupling or communication connection between devices or units may be electrical, mechanical, or other forms.

[0093] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0094] In addition, the functional units in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0095] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0096] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for screening refractive errors in children based on multimodal eye-tracking patterns, characterized in that, The method includes: S1: Based on the visual presentation device, the user's continuous pupil position is collected, and a multi-dimensional fixation feature vector is extracted and weighted fused to obtain the fixation stability index. The multi-dimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microslap video rate, the fixation retention rate, and the binocular fixation coordination. The weighted fusion is based on the index calculation model of children's age stratification. The index calculation model of children's age stratification adopts a two-layer design. Each layer uses different weights to process the feature vector input for a specific age group and outputs a fixation stability index of 0-100. S2: Using a binocular eye-tracking device, visual targets at varying distances are presented, the user's visual accommodation response is collected, and a weighted fusion is performed to obtain the accommodation function index; S3: Using a binocular eye-tracking device, convergence-induced tasks are conducted to collect convergence motion eye-tracking data streams, which are then weighted and fused to obtain the convergence function index; S4: Based on standardized reading materials that match the reading ability of the test children, real-time eye movement signals of the test children during the execution of the standardized reading task are collected through an eye-tracking device to obtain the reading task eye movement data stream, and then weighted and fused to obtain the reading efficiency index; S5: Presents multi-directional guidance. Through a binocular eye-tracking device, it collects eye movement data of the test children during the tracking and fixation process, and obtains the eye movement limit angle, horizontal eye movement range, vertical eye movement range and eye movement symmetry index in each direction. These data are then weighted and fused to obtain the eye movement range index. S6: Input the fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters into a preset refractive state and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, the preset referral rule engine performs logical judgment to obtain a referral suggestion.

2. The method for pediatric refractive error screening based on multimodal eye movement patterns according to claim 1, characterized in that, The S1-based visual presentation device collects the user's continuous pupil positions, extracts multi-dimensional fixation feature vectors, and performs weighted fusion to obtain a fixation stability index. The multi-dimensional fixation feature vectors include the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microsaccade rate, the fixation retention rate, and binocular fixation coordination. The weighted fusion is based on an age-stratified index calculation model for children. This model employs a two-layer design, with each layer targeting a specific age group and using different weights to process the feature vector input, outputting a fixation stability index of 0-100, including: S11: Generate fixation task instructions to obtain a multi-stage fixation task sequence, the fixation task sequence including a circular fixation target presented in the center, left and / or right of the screen through a visual presentation device, the circular fixation target including a cartoonish visual design. # According to the fixation task sequence, obtain the first stage visual stimulus, the first stage visual stimulus including the fixation target continuously presented on the left side of the screen. S12: Based on the fixation task sequence, obtain the visual stimulus for the second stage, which includes continuously presenting the fixation target on the right side of the screen. S13: Based on the fixation task sequence, obtain the visual stimulus for the third stage, wherein the visual stimulus for the third stage includes continuously presenting the fixation target in the center of the screen. S14: Using a binocular eye-tracking device, the continuous pupil position of the user during the execution of the fixation task sequence is collected according to a preset sampling frequency to obtain the raw eye-tracking data for the fixation task. S15: Input the raw eye movement data for fixation task into a data calibration algorithm to obtain standardized eye movement data mapped to a standard physical space. The data calibration algorithm includes data mapping based on a personalized affine transformation matrix. S16: Using a feature extraction algorithm, fixation features are calculated based on the standardized eye movement data to obtain a multidimensional fixation feature vector. The multidimensional fixation feature vector includes the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microslap video rate, the fixation retention rate, and the binocular fixation coordination. S17: The multidimensional fixation feature vector is input into a preset index calculation model and weighted fusion is performed to obtain a fixation stability index. The weighted fusion is based on an age-stratified index calculation model for children. The age-stratified index calculation model for children adopts a two-layer design. Each layer processes the feature vector input with different weights for a specific age group and outputs a fixation stability index of 0-100. The index calculation model is expressed as follows: in, For fixed-view stability index, For children's age, This is a multidimensional fixation feature vector, containing five feature values ​​extracted from standardized eye-tracking data. For the weight vector, For the younger age group Each weight, For the older age group Each weight, For the first Each feature value.

3. The method for pediatric refractive screening based on multimodal eye movement patterns according to claim 1, characterized in that, The S2 uses a binocular eye-tracking device to present visual targets at alternating distances, collects the user's visual accommodation response, and performs weighted fusion to obtain an accommodation function index, including: S21: Generate adjustment function instructions to obtain an adjustment function sequence. The adjustment function sequence includes presenting alternating near and far visual targets through a visual presentation device. The visual targets include a far target simulating a distance of more than 5 meters and a near target simulating a distance of 33 cm. The presentation of alternating near and far visual targets includes the alternating presentation of the far target and the near target. Each target lasts for 2-3 seconds, and a total of 8-10 cycles are completed. S22: Based on the aforementioned adjustment function sequence, the user's visual adjustment response is collected using a binocular eye-tracking device to obtain the raw adjustment response data stream; S23: Based on the original regulatory response data stream, a set of time-dimensional features are obtained through time series analysis, including the regulatory response latency and / or regulation completion time; S24: Based on the original regulatory response data stream, a set of non-time dimension features are obtained through response morphology analysis. The non-time dimension features include the regulatory response amplitude and the regulatory flexibility index. S25: Divide the time-dimensional features and the non-time-dimensional features into efficiency indicators and capability indicators. The efficiency indicators include the adjustment flexibility index, adjustment completion time, and adjustment response latency. The capability indicators include the adjustment response amplitude. S26: Based on the efficiency index and the efficiency age coefficient corresponding to the user's age, a weighted calculation is performed to obtain the non-normalized dynamic weight of the efficiency index. The efficiency age coefficient is less than 1 when the user's age is less than a certain threshold and greater than 1 when the user's age is greater than a certain threshold. S27: Based on the capability category indicators and the capability category age coefficient corresponding to the user's age, a weighted calculation is performed to obtain the non-normalized dynamic weight of the capability category indicators; S28: Input the non-normalized dynamic weights of the efficiency index and the non-normalized dynamic weights of the capability index into the normalization processing module for remapping to obtain a set of normalized dynamic weights, the sum of the normalized dynamic weights being 1. S29: Based on the normalized dynamic weights, efficiency indicators, and ability indicators, a regulatory function index is obtained by weighted fusion through a preset index calculation model. The value range of the regulatory function index is 0-100, and the weighted fusion is based on an index calculation model for children's age stratification.

4. The method for pediatric refractive screening based on multimodal eye movement patterns according to claim 1, characterized in that, The S3 method uses a binocular eye-tracking device to collect convergence motion eye-tracking data streams through a convergence-induced task, and performs weighted fusion to obtain a convergence function index, including: S31: Generate a convergence function instruction to obtain a convergence function sequence, wherein the convergence function sequence includes a visual target that gradually approaches from a distance through a visual presentation device, wherein the dynamic visual target includes a visual stimulus that simulates an object gradually approaching from 5 meters to 10 centimeters. S32: Based on the convergence function sequence, the convergence motion is collected by a binocular eye-tracking device to obtain a convergence motion eye-tracking data stream, which includes the position coordinates of both eyes at continuous time points and / or pupil information. S33: Based on the original eye movement data stream, after filtering and / or calibration processing, the convergence trajectory is obtained, which characterizes the convergence process of the visual axes of both eyes over time; S34: Using the limit analysis algorithm, the convergence trajectory is analyzed to obtain the convergence near point parameters, which are used to characterize the closest distance that can maintain binocular single vision; S35: Based on the starting position of the convergence trajectory and the convergence proximal point, the convergence amplitude is obtained through spatial amplitude calculation. The convergence amplitude is used to characterize the total amplitude of the inward rotation of the eyes. S36: The convergence trajectory is processed by differential operation and / or spectrum analysis to obtain the dynamic parameters of the convergence motion, wherein the dynamic parameters include the peak convergence velocity and / or the convergence smoothness index; S37: Input the convergence proximity parameters, the convergence amplitude, and the dynamic parameters into the feature integration model to obtain a multi-dimensional convergence feature parameter set, which is used to comprehensively characterize the steady-state and / or transient characteristics of the convergence motion. S38: Based on the convergence feature parameter set, a convergence function index is obtained by weighted fusion through a preset index calculation model. The value range of the convergence function index is 0-100, and the weighted fusion is based on the index calculation model of children's age stratification.

5. The method for pediatric refractive screening based on multimodal eye movement patterns according to claim 1, characterized in that, The S4 method, based on standardized reading materials matched to the reading abilities of the test children, uses an eye-tracking device to collect real-time eye-tracking signals during the standardized reading task, obtaining a reading task eye-tracking data stream. This data is then weighted and fused to obtain a reading efficiency index, including: S41: Collect the age and grade information of the test children to obtain user physiological parameters, which are used to match the difficulty level of the reading material; S42: Based on the user's physiological parameters, standardized reading materials matching the reading ability of the tested children are obtained through a preset difficulty matching rule library; S43: Using the standardized reading material, an eye-tracking device is used to collect real-time eye movement signals of the test children during the standardized reading task, and the reading task eye movement data stream is obtained; S44: Based on the eye-tracking data stream of the reading task, a structured eye-tracking event sequence is obtained after preprocessing and an event recognition algorithm. The eye-tracking event sequence includes a fixation point sequence and a saccade path sequence. S45: Using the structured eye movement event sequence, eye movement features are extracted to obtain initial eye movement feature data, which includes the duration of each fixation, the amplitude and direction of the saccade, and the number of characters read per unit time. S46: Based on the initial eye movement feature data, a quantized eye movement feature set is obtained through statistical calculation. The quantized eye movement feature set includes average fixation duration, forward saccade amplitude, retrograde ratio, and reading speed. S47: The quantified eye-tracking feature set is weighted and fused using a preset index calculation model to obtain a reading efficiency index. The reading efficiency index ranges from 0 to 100. The weighted fusion is based on an index calculation model that stratifies children by age.

6. The method for pediatric refractive screening based on multimodal eye movement patterns according to claim 1, characterized in that, The S5 provides multi-directional guidance by collecting eye movement data of the tested children during the tracking and fixation process using a binocular eye-tracking device. This data yields the eye movement limit angles, horizontal eye movement range, vertical eye movement range, and eye movement symmetry index in each direction. These are then weighted and fused to obtain the eye movement range index, which includes: S51: Generate a tracking gaze process instruction to obtain a tracking gaze process sequence, wherein the tracking gaze process sequence includes multi-directional guidance presented by the system through a visual presentation device, wherein the multi-directional guidance includes sequentially guiding the subject child to track gaze in eight directions: up, down, left, right, upper left, lower left, upper right, and / or lower right. S52: Based on the aforementioned gaze tracking process sequence, eye movement data of the test child during the gaze tracking process is collected using a binocular eye tracking device to obtain an eye movement trajectory data stream of the gaze tracking process; S53: Based on the eye movement trajectory data stream of the following gaze process, the eye movement limit angles in each direction are obtained through the angle extraction algorithm. The eye movement limit angles in each direction include the maximum amplitude values ​​of eye movements corresponding to the eight guiding directions. S54: Calculate the horizontal eye movement range based on the eye movement limit angles in the left and right directions, wherein the horizontal eye movement range is the sum of the eye movement limit angles in the left and right directions; S55: Calculate the vertical eye movement range based on the eye movement limit angles in the upward and downward directions, wherein the vertical eye movement range is the sum of the eye movement limit angles in the two directions; S56: Based on the eye movement limit angle in the symmetrical direction, the eye movement symmetry index is calculated using the symmetry evaluation model. The eye movement symmetry index is used to characterize the consistency of the eye movement range in the symmetrical direction. S57: Input the eye movement limit angles, horizontal eye movement range, vertical eye movement range, and eye movement symmetry index in each direction into the feature integration unit to obtain a comprehensive eye movement feature set; S58: Based on the comprehensive eye movement feature set, an eye movement range index is obtained by weighted fusion through a preset index calculation model. The value range of the eye movement range index includes 0-100. The weighted fusion is based on an index calculation model for children's age stratification.

7. The method for screening refractive errors in children based on multimodal eye movement patterns according to claim 1, characterized in that, The S6 process inputs the fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters into a preset refractive status and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, a preset referral rule engine performs logical judgment to obtain referral suggestions, including: S61: Based on the fixation stability index, the accommodation function index, the convergence function index, the reading efficiency index, the eye movement range index, visual acuity data, refractive error data, and user physiological parameters, a comprehensive dataset is obtained; S62: Based on a preset feature fusion model, feature fusion is performed on the comprehensive dataset to obtain fused features. The fused features are used to characterize the intrinsic relationship between refractive state and visual function abnormality. The preset feature fusion model adopts a multi-task deep learning framework. S63: Input the fused features into the preset refractive state and visual function abnormality screening model to obtain the refractive abnormality risk index, visual function comprehensive score and visual function abnormality type. The refractive abnormality risk index is used to quantitatively assess the risk level of refractive error and its related visual function abnormalities. The visual function comprehensive score includes a weighted comprehensive score of five dimensions of visual function index. The preset refractive state and visual function abnormality screening model adopts an attention mechanism. S64: Based on the refractive error risk index, visual function comprehensive score, and visual function abnormality type, a referral rule engine performs logical judgment to obtain a referral suggestion, which includes referral priority and / or recommended further examination items.

8. A pediatric refractive screening device based on multimodal eye-tracking patterns, wherein the pediatric refractive screening device based on multimodal eye-tracking patterns is used to implement the pediatric refractive screening method based on multimodal eye-tracking patterns as described in any one of claims 1-7, characterized in that, The device includes: Fixation Feature Module: Based on the visual presentation device, this module collects the user's continuous pupil position, extracts multi-dimensional fixation feature vectors, and performs weighted fusion to obtain a fixation stability index. The multi-dimensional fixation feature vectors include the area of ​​the binocular fixation ellipse, the standard deviation of fixation position, the microscratching video rate, the fixation retention rate, and binocular fixation coordination. The weighted fusion is based on an index calculation model for children's age stratification. This model adopts a two-layer design, with each layer targeting a specific age group and using different weights to process the feature vector input, outputting a fixation stability index of 0-100. The accommodation function module is used to present visual targets that are either near or far through a binocular eye-tracking device, collect the user's visual accommodation response, and perform weighted fusion to obtain the accommodation function index. Convergence Function Module: Used to collect convergence motion eye-tracking data streams through convergence-induced tasks using binocular eye-tracking devices, and perform weighted fusion to obtain the convergence function index; Reading efficiency module: Based on standardized reading materials that match the reading ability of the test children, it uses an eye-tracking device to collect real-time eye movement signals during the test children's performance of the standardized reading tasks, obtains the reading task eye movement data stream, and performs weighted fusion to obtain the reading efficiency index; Eye movement range module: used to present multi-directional guidance. Through binocular eye movement tracking device, it collects eye movement data of the test children during the following and fixation process, and obtains the eye movement limit angle, horizontal eye movement range, vertical eye movement range and eye movement symmetry index in each direction. The data are then weighted and fused to obtain the eye movement range index. Risk integration module: This module is used to input the fixation stability index, accommodation function index, convergence function index, reading efficiency index, eye movement range index, visual acuity data, refractive error data, and user physiological parameters into a preset refractive status and visual function abnormality screening model to obtain a refractive error risk index, a comprehensive visual function score, and a visual function abnormality type. Finally, it performs logical judgment through a preset referral rule engine to obtain a referral recommendation.

9. A pediatric refractive screening device based on multimodal eye-tracking patterns, characterized in that, The pediatric refractive screening processor based on multimodal eye movement patterns; a memory storing computer-readable instructions, which, when executed by the processor, implement the method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium contains program code that can be invoked by a processor to execute the method as described in any one of claims 1 to 7.

Images