An ophthalmic child visual development monitoring and intervention system

Abstract

The application discloses an ophthalmic child visual development monitoring and intervention system, which comprises a multi-dimensional synchronous screening sensor array integrated on a head-mounted frame, a synchronous acquisition and analysis controller, an intervention task matching and adaptive game trainer, a continuous monitoring and hierarchical early warning device and a data management and remote calibrator. The sensor array synchronously acquires multi-dimensional data such as eye movement, visual evoked potential, diopter, eye axial length and eye use behavior, and outputs development deviation types and severity grades through a classifier; a double-loop adaptive controller dynamically adjusts game parameters and levels according to real-time eye movement and visual evoked potential signals, and realizes closed-loop training; a continuous monitoring module analyzes time sequence indexes such as eye axial change rate and proportion of weak eye fixation and executes hierarchical early warning. The application realizes multi-dimensional synchronous screening, adaptive intervention and active early warning of child visual development in a family scene, and has the advantages of continuous monitoring, timely intervention and high compliance.

Description

Technical Field

[0001] This invention belongs to the field of medical device technology, and in particular relates to an ophthalmic pediatric vision development monitoring and intervention system. Background Technology

[0002] There is a critical period for children's visual development (3-8 years old). Early screening and timely intervention during this stage for abnormalities such as amblyopia, myopia, insufficient accommodative function, and stereopsis deficits can yield significant rehabilitation results. Currently, commonly used ophthalmic examination equipment includes phoropter, fundus camera, optical coherence tomography (OCT), and visual evoked potential (VEP) analyzer. These devices can accurately measure visual function indicators such as refractive error, axial length, and VEP waveform. In the home setting, some wearable devices (such as smart glasses) are beginning to be used to monitor children's eye-use behavior (viewing distance, duration, ambient light, etc.), and there are also standalone amblyopia training devices or mobile app games that provide visual training functions. Furthermore, some studies are attempting to combine virtual reality / augmented reality technology with visual training, improving children's training adherence through gamification. Hospitals typically recommend that children have a comprehensive eye examination every 3-6 months and prescribe training based on the examination results.

[0003] However, existing technologies still have the following shortcomings: First, home monitoring devices have limited functionality and cannot simultaneously collect multidimensional visual function indicators such as refractive error, axial length, and VEP. Screening still relies on large-scale hospital equipment, making continuous, daily developmental tracking difficult. Second, there is a serious disconnect between screening and intervention. Parents need to manually interpret hospital reports and manually set training parameters, a cumbersome and error-prone process that cannot achieve zero-delay targeted intervention after screening. Third, existing home training programs lack objective physiological feedback, relying mostly on game scores or parental observation to adjust difficulty. They cannot dynamically adjust training parameters based on real-time signals such as eye movement fixation position and VEP amplitude, making them particularly unsuitable for young children who cannot accurately express themselves. Fourth, existing systems are mostly passive monitoring or passive training, lacking an active early warning mechanism based on time-series trend analysis. They cannot promptly alert when axial length accelerates or the proportion of fixation in the amblyopic eye decreases, often missing the optimal intervention window. Fifth, data security and telemedicine collaboration of home devices are insufficient. Directly uploading raw data poses privacy risks, and device calibration requires returning to the factory, making it difficult for doctors to remotely adjust parameters. Therefore, there is an urgent need for a children's vision development monitoring and intervention system that integrates multidimensional synchronous screening, adaptive game intervention, continuous proactive early warning, and family-medical collaborative management. Summary of the Invention

[0004] To overcome the aforementioned deficiencies of the prior art, this invention provides an ophthalmic pediatric vision development monitoring and intervention system, which solves the problems of limited home monitoring functions, disconnect between screening and intervention, lack of physiological feedback loop, absence of proactive early warning, and insufficient data security and remote collaboration in the prior art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: An ophthalmic pediatric vision development monitoring and intervention system, comprising: A multidimensional synchronous screening sensor array integrated on a head-mounted frame includes: an eye-tracking sensor, a flexible dry electrode visual evoked potential acquisition electrode, a refractive power sensor, an axial length estimation sensor, an ambient light sensor, a distance sensor, and a posture sensor. The eye-tracking sensor acquires eye position and motion parameters; the flexible dry electrode visual evoked potential acquisition electrode acquires electrical signals from the occipital region without conductive gel; the refractive power sensor measures the refractive state of the eye; the axial length estimation sensor measures the distance from the corneal apex to the retinal pigment epithelium; the ambient light sensor measures ambient illuminance; the distance sensor measures the distance between the eye and the gaze target; and the posture sensor measures the head posture angle. A synchronous acquisition and analysis controller controls all sensors in the sensor array to synchronously acquire data within the same detection time window. The controller generates a multidimensional feature vector based on the acquired data and inputs this vector into a classifier, which outputs a developmental deviation type and severity rating. The classifier outputs a category label and severity score based on the input vector. The developmental deviation type includes at least one of amblyopia, myopia, insufficient accommodative function, or stereopsis deficit. The severity rating includes mild, moderate, or severe levels. An intervention task matching and adaptive game trainer is provided, comprising: an intervention matching unit and a dual-loop adaptive controller; the intervention matching unit selects a corresponding intervention game from a game library according to the developmental deviation type and sets game parameters according to the severity level; the dual-loop adaptive controller includes a first loop and a second loop, the first loop adjusting the game parameters based on real-time acquired eye-tracking fixation position and visual evoked potential signals, and the second loop adjusting the game level based on fixation duration, visual evoked potential amplitude change rate, and task accuracy; wherein, the intervention game is aimed at visual training. The video game program, wherein the game parameters include contrast, spatial frequency, stimulation duration or game difficulty level, the eye-tracking fixation position is the coordinate of the point on the display screen where the fixation direction of a single eye falls, the visual evoked potential signal is the EEG waveform signal caused by visual stimulation, the fixation duration is the duration of a single fixation behavior, the visual evoked potential amplitude change rate is the percentage obtained by dividing the difference between the current amplitude value and the baseline amplitude value by the baseline amplitude value, the task accuracy rate is the ratio of the number of times the game operation is correctly completed per unit time to the total number of operations, and the game level is an integer representing the level of training difficulty; A continuous monitoring and tiered early warning system is provided. During non-screening and non-training periods, the system collects eye-use behavior data at a reduced sampling frequency. Based on this data, it calculates the axial length change rate, the proportion of fixation time in amblyopic eyes, and the refractive error change slope. When any of these indicators exceeds a preset threshold, a Level 1, Level 2, or Level 3 early warning is issued. The eye-use behavior data includes eye distance, ambient illuminance, head posture angle, blink count, or continuous eye use duration. The axial length change rate is the relative change in axial length per unit time. The amblyopic eye fixation time proportion is the percentage of total fixation time in amblyopic eyes relative to the total fixation time in both eyes. The refractive error change slope is the rate of change of refractive error measurement over time. The data management and remote calibrator stores raw data in a built-in encrypted chip, uploads anonymized feature data to the cloud, performs remote online calibration when sensor baseline drift is detected, and provides a data interface for doctors to remotely adjust parameters. The raw data consists of unprocessed signal sequences directly collected by each sensor; the anonymization process removes personally identifiable information and assigns random codes; sensor baseline drift is the deviation of the sensor's output value from its calibration zero point when there is no input signal; and the remote online calibration is a process where the device's built-in program automatically guides the monitored subject to complete a standard fixation task and re-collects sensor data for correction.

[0006] Preferably, the eye-tracking sensor is a binocular infrared eye-tracking camera with a sampling frequency greater than or equal to 120 Hz; the refractive power detection sensor is a Shaker-Hartmann wavefront sensor; and the axial length estimation sensor is an optical coherent ranging sensor based on the principle of low-coherence optical interference.

[0007] Preferably, the flexible dry electrode visual evoked potential acquisition electrode has 3 to 5 dry electrode contacts, which are installed on the nose pad and the inner side of the temple. The contact material is a spring-loaded microneedle or a conductive fabric. The spring-loaded microneedle is a miniature needle-shaped conductor supported by an elastic element. The conductive fabric is a flexible fabric woven from conductive fibers.

[0008] Preferably, in the dual-loop adaptive controller: the delay of the first closed loop in adjusting game parameters does not exceed 50 milliseconds, and the adjusted game parameters include target brightness, contrast, target size, or background noise; the second closed loop adjusts the game level every 30 seconds, and the game level ranges from 1 to 10; wherein, the target brightness is the display brightness of the interactive game object, the contrast is the brightness ratio between the target and the background, the target size is the pixel diameter of the interactive game object in the display screen, and the background noise is the intensity of random pixel disturbance in the non-target area of ​​the screen.

[0009] Preferably, in the graded warning device: the first-level warning is a prompt sound emitted through a bone conduction speaker or a flashing light emitted through a light-emitting diode; the second-level warning is a semi-transparent dynamic image superimposed on the display screen or the activation of a blink training task; the third-level warning is a text message sent to the guardian's terminal and a prompt to seek medical attention; wherein, the bone conduction speaker is an electroacoustic transducer that transmits sound through skull vibration, the blink training task is a program that guides the monitored person to blink continuously 5 to 10 times, and the guardian's terminal is a smartphone, tablet computer, or personal computer.

[0010] Preferably, the remote online calibration process is as follows: when sensor baseline drift is detected, a set of standard fixation points are presented on the display screen, and the child is guided to fixate on each fixation point in turn. At the same time, sensor data is re-acquired for correction. The entire calibration process does not exceed 60 seconds. The standard fixation points are five circular markers located in the center and four corners of the display screen. The sequential fixation means that each fixation point is displayed for 2 to 3 seconds and the monitored person needs to focus their gaze on the point and keep it stable.

[0011] Preferably, the intervention games include: monocular stimulation games for amblyopia, near-far alternating focusing games for accommodative insufficiency, and random dot stereogram games for stereoscopic vision impairment; wherein, the monocular stimulation games are games in which only the amblyopic eye can see a clear game screen while the screen of the healthy eye is obscured or blurred, the near-far alternating focusing games are games in which the game target appears alternately at near and far distances, and the random dot stereogram games are games composed of randomly distributed dots and require binocular fusion to identify the hidden stereoscopic shapes.

[0012] Preferably, the system further includes a binocular differential stimulator, which enhances the display of the image corresponding to the amblyopic eye and blurs the image corresponding to the healthy eye during gameplay. The degree of blurring is updated according to the real-time gaze position. The enhanced display increases the brightness or contrast of the image to 120% to 150% of the base value, the blurring display applies Gaussian filtering to the image, and the update according to the real-time gaze position involves recalculating the gaze point coordinates and blurring the image of the healthy eye outside a certain area centered on the gaze point when each frame is refreshed. The certain area is a circular area with a radius of 50 to 100 pixels.

[0013] Preferably, the head-mounted frame is augmented reality glasses, which are equipped with binocular see-through waveguide displays, and the brightness of each eye is independently adjustable; wherein, the binocular see-through waveguide display is an optical module that allows some external light to pass through while displaying virtual images, and the independent brightness adjustment of each eye means that the brightness of the left and right eye displays is set to different values ​​and the adjustment range is 0 to 2000 nits.

[0014] Preferably, the classifier is a random forest classifier or a support vector machine classifier.

[0015] The technical effects and advantages of the ophthalmic pediatric vision development monitoring and intervention system of the present invention are as follows: 1. This invention integrates multiple sensors, including eye-tracking, flexible dry electrode visual evoked potential acquisition, refractive error detection, axial length estimation, ambient light, viewing distance, and head posture, onto a wearable headband. A synchronous acquisition and analysis controller completes all data acquisition within the same detection time window, generating multidimensional feature vectors and automatically outputting the type and severity grade of developmental deviations. Users can obtain comprehensive and objective visual function assessment results in their home environment without needing to visit professional institutions or operate large desktop equipment, significantly reducing the threshold and cost of monitoring children's visual development.

[0016] 2. This invention incorporates an intervention task matching and adaptive game trainer. The intervention matching unit automatically selects corresponding intervention games from the game library based on the developmental deviation type output by the classifier and automatically sets game parameters according to the severity level, eliminating the need for parents to manually interpret reports or configure training plans. This mechanism eliminates the information gap between screening and intervention, ensuring that children can receive targeted training immediately after completing the screening, thus improving the timeliness and accuracy of intervention.

[0017] 3. This invention employs a dual-loop adaptive controller. The first loop acquires eye-tracking fixation position and visual evoked potential signals in real time with low latency and adjusts game parameters accordingly. The second loop dynamically adjusts the game level based on fixation duration, the rate of change of visual evoked potential amplitude, and task accuracy. This design overcomes the limitations of traditional solutions that rely solely on game scores or children's subjective feedback. It uses objective physiological signals to drive real-time matching of training intensity, making it particularly suitable for young children who cannot accurately express their feelings, effectively improving the scientific nature and effectiveness of training.

[0018] 4. This invention incorporates a continuous monitoring and tiered early warning system. During non-screening and non-training periods, it continuously collects eye-use behavior data at a reduced sampling frequency and calculates key time-series indicators such as the rate of change in axial length, the proportion of fixation time in amblyopic eyes, and the slope of refractive error change. When any indicator exceeds a preset threshold, the system issues a Level 1, Level 2, or Level 3 warning (from mild reminders to automatic intervention tasks and then to medical recommendations). This mechanism can identify early trends of developmental deterioration, transforming passive follow-up examinations into proactive interventions, and helping to seize the optimal intervention window during the critical period of visual development.

[0019] 5. This invention stores raw data in a built-in encrypted chip, uploading only anonymized feature data to protect children's privacy. It also provides remote online calibration, correcting sensor baseline drift without requiring the device to return to the factory, and includes a data interface for doctors to remotely adjust parameters, achieving effective integration between home settings and professional medical resources, thus balancing data security and medical collaboration. Attached Figure Description

[0020] Figure 1 This is a system block diagram of an ophthalmic pediatric vision development monitoring and intervention system proposed in this invention. Detailed Implementation

[0021] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0022] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include," "contain," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that includes a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitation, an element defined by the phrase "includes..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.

[0023] refer to Figure 1This invention provides a system for monitoring and intervening in pediatric visual development, comprising: a multidimensional synchronous screening sensor array integrated on a head-mounted frame, the array including an eye-tracking sensor, a flexible dry electrode visual evoked potential acquisition electrode, a refractive power detection sensor, an axial length estimation sensor, an ambient light sensor, a distance sensor, and a posture sensor; a synchronous acquisition and analysis controller, used to control all sensors to synchronously acquire data within the same detection time window and generate multidimensional feature vectors, and output the developmental deviation type and severity level through a classifier; and an intervention task matching and adaptive game trainer, whose intervention matching unit selects the corresponding intervention task according to the developmental deviation type. The invention includes a pre-game setup with pre-defined parameters. A dual-loop adaptive controller adjusts game parameters based on real-time collected eye movement fixation positions and visual evoked potential signals, and adjusts the game level according to fixation duration, visual evoked potential amplitude change rate, and task accuracy. A continuous monitoring and tiered early warning system collects eye behavior data at a reduced sampling frequency during non-screening and non-training periods, calculates the axial length change rate, the proportion of fixation time in amblyopic eyes, and the refractive error change slope, and executes level one, two, or three warnings when indicators exceed thresholds. A data management and remote calibration system is also included, used to encrypt and store raw data, upload anonymized feature data, perform remote online calibration, and provide a remote parameter adjustment interface for doctors. This invention achieves multi-dimensional synchronous screening of children's visual development in a home setting, adaptive game intervention, continuous proactive early warning, and family-medical collaborative management. Overall architecture: This invention provides a system for monitoring and intervening in pediatric vision development. The system uses augmented reality glasses as a head-mounted platform and integrates five functional modules: a multi-dimensional synchronous screening sensor array, a synchronous acquisition and analysis controller, an intervention task matching and adaptive game trainer, a continuous monitoring and graded early warning system, and a data management and remote calibration system. The specific implementation of this invention is described in detail below through five embodiments and one comparative example.

[0024] Example 1: Purpose of implementation: This embodiment aims to provide a head-mounted system suitable for children aged 3-12 years that integrates multidimensional visual function screening and adaptive game intervention, solving the problems of existing home monitoring devices having single functions, disconnect between screening and intervention, and lack of objective physiological feedback loop, and realizing closed-loop management of children's visual development in the home setting.

[0025] Systematization: The system hardware configuration in this embodiment is as follows: Headband Frame: Adopting a lightweight AR glasses design, the main frame is made of medical-grade polycarbonate, with a total weight controlled to under 120 grams. The glasses are equipped with a binocular see-through waveguide display (40° field of view, 1280×720 resolution, independent brightness adjustment for each eye from 0 to 2000 nits, 50% light transmittance), which can overlay virtual images onto the real world.

[0026] Eye-tracking sensor: It adopts a binocular infrared eye-tracking camera (sampling frequency 120Hz). Two miniature cameras are installed on the inside of the left and right frames near the bridge of the nose, respectively, and are equipped with four 850nm infrared LEDs to capture the position of the pupils and the coordinates of the gaze point in real time.

[0027] Flexible dry electrode visual evoked potential acquisition electrode: A reference electrode is set at the nose pad, and two signal electrodes are set on the inner sides of the left and right temples corresponding to the occipital region (O1 / O2 position), for a total of 5 dry electrode contacts. The electrode material adopts a gold-plated microneedle array (needle length 1.2mm, spacing 2mm), which can penetrate the hair and form a stable contact with the scalp without conductive gel.

[0028] Refractive power sensor: Integrated in the center of the frame above the bridge of the nose, it adopts a Shaker-Hartmann wavefront sensor (size 15mm×12mm×8mm), with a built-in 780nm laser diode and microlens array, which can quickly measure the refractive power of both eyes, with a measurement range of -10D to +8D and an accuracy of ±0.25D.

[0029] Axial length estimation sensor: Integrated at the front of the left temple, it is an optical coherent ranging sensor (center wavelength 850nm, bandwidth 50nm) that uses the principle of low coherence light interference to estimate the axial length by measuring the optical path difference from the corneal apex to the retinal pigment epithelium, with a measurement accuracy of ±0.05mm.

[0030] Ambient light sensor: Located in the center of the front of the frame, it uses a digital light sensor with a measurement range of 0~65535 Lux.

[0031] Distance sensor: Located on both sides of the front of the frame, it uses an infrared time-of-flight sensor to measure distances from 5 to 200 cm and is used to monitor eye distance.

[0032] Attitude sensor: Integrated into the right temple, it uses a nine-axis inertial measurement unit to output three-axis acceleration, angular velocity and magnetometer data, which are then fused by an algorithm to obtain the head pitch angle, roll angle and yaw angle.

[0033] Data processing unit: The glasses have a built-in embedded processor (ARM Cortex-M7, 400MHz, with FPU), 2MB SRAM, 16MB Flash, running a lightweight real-time operating system. All sensor data is collected to the processor via I2C / SPI bus and signal processing is completed at the edge.

[0034] Power supply: Removable lithium polymer battery (1100mAh capacity), located at the end of the temple, supports 2 hours of continuous screening / training or 8 hours of standby monitoring.

[0035] Implementation steps: Wearing and Activation: After the child puts on the glasses, the guardian can activate the "Quick Screening" mode via a mobile app. The system automatically performs sensor self-checks, illuminating the infrared LED and wavefront sensor.

[0036] Simultaneous screening: Within the same detection time window (2 minutes), eye movement trajectory, VEP signal (using checkerboard flip stimulation, frequency 2Hz), refractive error, axial length, ambient illumination, viewing distance, and head posture are collected simultaneously. All data are synchronized with timestamps via hardware synchronization signals to generate a multidimensional vector containing 20 features (such as fixation point spread standard deviation, VEPP100 latency, refractive error difference, absolute value of axial length, etc.).

[0037] Intelligent analysis: Input multidimensional feature vectors into a pre-trained random forest classifier (100 trees, maximum depth 15), and output developmental deviation type (e.g., "left eye amblyopia, mild; accommodation function normal; stereopsis function normal") and severity score (0~100 points).

[0038] Intervention Matching: Based on the output results, the intervention matching unit automatically calls the "monocular stimulation game"—the balloon popping game—from the local game library. Initial game parameters: The amblyopic eye can see a full-color balloon, and the healthy eye's image is superimposed with Gaussian blur (kernel size 9×9, contrast reduced to 60%). The initial difficulty level is set to level 3, the contrast is 100%, the spatial frequency is 2 cycles / deg, and the stimulation duration is 15 minutes.

[0039] Adaptive Training: During game training, the dual-loop adaptive controller collects eye-tracking fixation points and VEP signals in real time with a 40ms delay. If the amblyopic eye's fixation deviates from the center of the balloon by more than 50 pixels, the first loop automatically increases the balloon's brightness by 30% and amplifies it by 10%; if the VEP amplitude decreases by more than 20% from the baseline, the contrast is automatically increased to 120%. Every 30 seconds, a comprehensive feedback index (fixation duration 40%, VEP amplitude change rate 30%, task accuracy 30%) is calculated. When the index is above 0.7 for three consecutive cycles, the game level increases from level 3 to level 4; when it is below 0.3, it decreases to level 2.

[0040] Continuous monitoring and early warning: During non-training periods, the system records eye behavior data at a sampling frequency of once every 2 minutes. The background time-series model uses a sliding window (7 days) to calculate the rate of change in axial length, the proportion of fixation time in amblyopic eyes, and the slope of refractive error change. When the system detects that the proportion of fixation time in amblyopic eyes is less than 10% for 3 consecutive days, it triggers a level 2 warning: a semi-transparent animation of "Take a break and play a blinking game" is overlaid on the display screen, and blinking guidance is initiated for 20 seconds (prompting blinking once every 3 seconds).

[0041] Data Management: All raw data (sensor ADC values, VEP waveforms, game operation logs) are encrypted and stored in the glasses' built-in encrypted chip. When a Level 3 alert is triggered, anonymized feature data (such as daily gaze percentage curves and VEP amplitude trends) is uploaded to the cloud. Doctors can remotely review reports and adjust classifier thresholds online via a dedicated app.

[0042] Implementation results:

[0043] After 4 weeks of continuous use by 50 children with amblyopia (aged 3-8), the average compliance rate reached 92%. The proportion of fixation time in the amblyopic eye increased from 12% at baseline to 41%, the VEP P100 latency was shortened by 18ms, and visual acuity improved by more than 2 lines (Snellen scale). No adverse reactions were reported by any children, and the system was easy to operate according to the feedback from guardians, who were happy to wear it.

[0044] Example 2: Purpose of implementation: This embodiment aims to provide an intervention system that focuses on dynamic monitoring of axial length and refractive power, combined with alternating near and far focusing training, to address the early prevention and control needs of myopia in children, thereby delaying or preventing the occurrence and development of myopia.

[0045] Systematization: It is basically the same as Example 1, except for the following system adjustments: The refractive power sensor adds astigmatism measurement function (outputs cylinder power and axis).

[0046] The axial length estimation sensor increases the sampling frequency to once every 4 hours (automatically wakes up for measurement during non-training periods).

[0047] The intervention game library has been replaced with a "near-far alternating focus game": the game target (cartoon animal) randomly appears in the foreground (a virtual plane 30cm away from the eyes) or the background (5m away), and the child needs to adjust their ciliary muscle to make the target clear and click on it. The system has a built-in adjustment reaction time measurement module.

[0048] The warning logic has been modified: A Level 1 warning (temple vibration alert) is triggered when the axial length change rate is >0.03 mm / day or the continuous screen time is >45 minutes and the ambient illuminance is <300 Lux. A Level 3 warning is triggered when the refractive error deterioration slope exceeds -0.50D / year.

[0049] Remove the binocular differential stimulator associated with amblyopia.

[0050] Implementation steps: The system continuously records the axial length of the eye (every 4 hours), the distance of eye use (real-time), the ambient illuminance (real-time), and the head posture of children who wear glasses daily.

[0051] Guardians can initiate a "myopia risk assessment" screening once a week via the APP (which takes 1 minute), and the system simultaneously collects refractive error, axial length, and accommodation reaction time.

[0052] When accelerated axial elongation (weekly change rate > 0.03 mm / day) or sluggish accommodation response (near-to-far focus switching time > 1.5 seconds) is detected, the system automatically recommends starting the "Alternating Near and Far Focusing Game". Game parameters are initialized based on age and current accommodation ability: initial difficulty level 2, target appearance interval of 2 seconds, and near-to-far ratio of 1:1.

[0053] During training, the second closed loop adjusts the difficulty based on task accuracy and reaction time: the level is increased when accuracy is >90% and reaction time is <1.0 second, and decreased when accuracy is <70% or reaction time is >2.0 seconds. The first closed loop automatically adjusts the target size based on eye-tracking fixation position (enlarging it when it deviates from the center).

[0054] The early warning and data management are the same as in Example 1, but when a level 3 early warning is triggered, a reminder to the guardian is automatically pushed that "it is recommended to perform cycloplegic refraction".

[0055] Implementation results: In a 3-month trial involving 40 pre-myopic children (hyperopic reserve ≤0.75D, aged 6-9 years), the average axial length increase was 0.08mm, significantly lower than the historical control group (average increase of 0.21mm in children without intervention during the same period). The average amplitude of accommodation increased by 2.5D, and the rate of decline in hyperopic reserve slowed by 60%. Parents reported that their children showed high acceptance of the near-far alternation game and actively asked to "play another round."

[0056] Example 3: Purpose of implementation: This embodiment aims to provide a low-cost, game-free, pure monitoring system suitable for families with limited economic resources or those that only need regular screening and do not require intervention, as well as large-scale vision screening scenarios in kindergartens / schools.

[0057] Systematization: The difference from Example 1 is as follows: Remove AR displays and game training-related hardware (waveguides, GPUs, etc.), and simplify the headband to a lightweight headband (total weight <80g).

[0058] It retains the complete sensor array (eye movement, VEP, diopter, axial length, ambient light, distance, posture) and synchronous acquisition and analysis controller.

[0059] Remove the dual-loop adaptive controller and intervention matching unit, but retain the software interface for intervention task matching (for future upgrades).

[0060] The warning device only retains data analysis and push functions, and does not perform active intervention (such as not overlaying animations or starting blink training).

[0061] Data management and remote calibrators remain unchanged.

[0062] Implementation steps: The child wears a headband, and the guardian initiates a "quick screening" via a mobile app (takes 90 seconds). The system automatically completes the synchronous data collection and analysis of all sensors.

[0063] A "visual function development fingerprint" report is generated, including indicators such as refractive error, axial length, VEP latency / amplitude, fixation stability score, and the percentage of fixation in amblyopic eyes, as well as percentiles compared with age-matched norms.

[0064] The report is automatically pushed to the guardian's app. If all indicators are normal, it is recommended to have a follow-up examination every 3 months; if an abnormality is detected (such as refractive error deviation >1.0D or VEP latency prolongation >2 standard deviations), the system will push "recommend medical treatment" and recommend nearby professional institutions.

[0065] Continuous monitoring mode is available: If the guardian purchases the upgrade service, the system can enable daily eye use behavior recording, but the alerts will only be pushed through the APP (not through the device itself).

[0066] Data storage and remote calibration are the same as in Example 1.

[0067] Implementation results: The cost is reduced by approximately 60% compared to Example 1 (approximately 1200 RMB vs. 3000 RMB). In a trial at a kindergarten (300 children), the average screening time was 82 seconds, and the children's acceptance rate was 98% (no complaints of discomfort). Thirteen suspected cases of amblyopia and 29 cases of pre-myopia were detected, with a 94% accuracy rate in subsequent hospital diagnoses. Guardians generally found the operation simple and the reports easy to understand.

[0068] Example 4: Purpose of implementation: This embodiment addresses the needs of precise clinical diagnosis and scientific research. While ensuring portability, it improves the monitoring accuracy to a level comparable to large desktop devices, making it suitable for use in hospital ophthalmology clinics, clinical trials, and visual development research.

[0069] Systematization: Based on Example 1, the following modules can be added or upgraded: Multi-channel VEP acquisition: The number of dry electrodes is increased to 8 (covering the O1, O2, Oz, P3, P4, Pz, Cz, and Fpz positions of the international 10-20 system), using actively shielded cables and differential amplification, achieving a common-mode rejection ratio >110dB and a 3-fold improvement in signal-to-noise ratio. The sampling rate is 1000Hz, with bandpass filtering from 1 to 100Hz.

[0070] Dynamic measurement of pupil diameter: Using infrared images from an eye-tracking camera, an additional pupil diameter change curve (accuracy 0.1 mm) is output to assess dark adaptation and photoexcitation recovery time.

[0071] Adjustment micro-fluctuation analysis: The refractive power sensor continuously samples at a frequency of 50 Hz for 10 seconds to analyze the adjustment micro-fluctuation spectrum (peak frequency and amplitude) and assess ciliary muscle function.

[0072] Stereoscopic vision quantitative test: A random dot stereoscopic image (RDS, dot density 200 dots / degree², parallax reduced from 400 arcseconds to 20 arcseconds) is presented on an AR display. Stereoscopic vision sharpness is quantified by observing the child's click position and reaction time on the hidden image.

[0073] AI-assisted diagnosis in the cloud: The full multidimensional feature vector is uploaded to the hospital's private cloud, and diagnostic suggestions are generated by a deep neural network (ResNet50), which are then reviewed by ophthalmologists.

[0074] Calibration module upgrade: Automatic temperature and humidity compensation added, calibration cycle extended to 3 months.

[0075] Implementation steps: Children wear specialized glasses with the assistance of doctors or technicians, which connect to a dedicated workstation at the hospital (via Bluetooth or Wi-Fi).

[0076] When the "Clinical Diagnosis Mode" is activated, the system automatically executes the following sequence: ① Pupil recovery curve measurement after 5 minutes of dark adaptation; ② Multi-channel VEP acquisition (checkerboard flipping, spatial frequency 1 / 2 / 4 cycles / deg repeated 50 times each); ③ Accommodation micro-fluctuation test; ④ Stereoscopic acuity measurement; ⑤ Refractive error and axial length measurement.

[0077] All data is displayed in real time on the workstation screen, and doctors can manually annotate abnormal waveforms or remove artifacts.

[0078] The system automatically generates a structured report and compares it with previous data to plot a developmental trajectory curve. After the doctor confirms and signs the report, it is stored in the hospital's electronic medical record system.

[0079] Remote calibration is performed by hospital engineers every three months using a dedicated calibration target.

[0080] Implementation results: Compared with the Topcon CV-5000 phoropter and the Roland RETI-port clinical VEP, the refractive error in this embodiment is ≤0.12D (correlation r=0.98), the VEP latency error is ≤3ms (within-group correlation coefficient 0.96), and the stereoscopic acuity test repeatability is ≥95% (compared to the Titmus stereoscopic examination). Figure 1 (Conformity 92%). It has passed ethical review at a top-tier hospital in Beijing and has entered the clinical trial phase (120 cases enrolled). Doctors commented that "the procedure is simple, the results are reliable, and children have high cooperation."

[0081] Example 5: Purpose of implementation: This embodiment aims to address the problem of poor training compliance among children with amblyopia. By introducing gamified rewards, parental collaboration, and social competition mechanisms, it improves training motivation and completion rate, and is particularly suitable for children who have previously failed training or have severe resistance.

[0082] Systematization: The following subsystems are added based on Example 1: Reward Mechanism: Each time a training task is completed (reaching the target duration or accuracy), the system awards virtual badges (such as "Focus Star" or "Fast-Eye Shooter") in the AR interface. Accumulated badges can be redeemed for physical toys (ordered through the guardian's app and shipped by the partner).

[0083] Parental Collaboration: Parents can view their child's training status in real time via a mobile app (current level, accuracy, VEP amplitude). Parents can record encouraging voice messages (such as "Come on, baby, look a little closer!"), which the system will automatically play when the child's attention wanes (via bone conduction speakers).

[0084] Gamified social interaction: Supports online multiplayer battles between two children (requires two devices). In battle mode, two children simultaneously engage in the same monocular stimulation game. The system displays the scores of both players in real time, and the winner receives an additional virtual badge at the end of the match.

[0085] Fatigue detection: The system automatically assesses fatigue by using the VEP amplitude decline trend (a decrease of >10% over 5 consecutive cycles) and blink frequency (>25 times / minute). When the fatigue index exceeds the threshold, the system terminates training early and awards a "rest badge," while simultaneously sending a voice prompt saying, "Baby is tired, take a rest."

[0086] Implementation steps: Guardians can register accounts for their children through the app and set reward goals (such as "collect 10 medals to redeem a toy car").

[0087] Before a child begins training, the system plays an encouraging voice message recorded by the parent. During training, if the child's gaze wanders or their reaction slows down, the system automatically switches to "Parent's Encouragement Mode"—playing a personalized voice message.

[0088] A weekly "Battle Day" is set up: two children (from different locations) can log in simultaneously, and the system will match them with opponents of similar skill levels. In battle mode, game parameters (difficulty, target frequency) are the same, and scores are displayed in real time. After the battle, the system analyzes the data of both players (such as fixation time of the amblyopic eye and VEP amplitude improvement rate) and provides improvement suggestions.

[0089] The fatigue detection module continuously monitors physiological signals. If a child shows signs of fatigue (such as a VEP amplitude decrease of >15% for 5 consecutive minutes and a blinking frequency of >28 times / minute), the system will terminate the training early (the minimum training time can be set to 10 minutes) and reward the child with a "rest badge" as encouragement.

[0090] A weekly "compliance report" is generated, including the number of training days, completion rate, number of medals earned, and win rate, and is sent to guardians and doctors. Doctors can adjust the training plan based on the report (such as increasing the difficulty or changing the type of game).

[0091] Implementation results: In a study of 30 amblyopic children with poor compliance (whose completion rate with traditional amblyopia training devices was <40% in the past month), this implementation method improved the training completion rate to 78% after 4 weeks, with an effective vision improvement rate of 83% (visual improvement of ≥2 lines). The social battle function increased the frequency of children actively requesting training by 2.5 times, and parents reported that "their children no longer resisted training and even proactively reminded them to put on their glasses." The fatigue detection function effectively prevented aversion caused by overtraining, with no cases of children quitting due to fatigue.

[0092] Comparative Example 1: Purpose of implementation: This comparative example is used to demonstrate the advantages of the present invention over the prior art. It adopts a standard clinical approach, namely, regular hospital checkups combined with a home-based independent training device.

[0093] Systematization: The solution includes the following components: Hospital screening: Have a comprehensive eye exam at the hospital every 3 months, including cycloplegic refraction (tropicamide), fundus photography, VEP testing (using a desktop electroencephalogram) and axial length measurement (A-scan).

[0094] Home-based training: Based on the examination results, the doctor prescribes a medication, and parents purchase an independent amblyopia training device (such as the BSV-1 grating stimulator or a red flashing device), or download a mobile app (such as "Amblyopia Training Expert"). The training content includes grating stimulation, red light flashing, and fine visual acuity training, 30 minutes a day, without physiological feedback adjustment.

[0095] Compliance management: Parents manually record training logs and bring them to the doctor for review every 3 months during follow-up appointments. There are no proactive warnings or reminders.

[0096] Implementation steps: Initial visit: The child completes cycloplegic refraction, fundus examination, VEP (visual excision procedure), and axial length measurement. The doctor diagnoses amblyopia (e.g., left eye visual acuity 0.3, right eye 0.8). The doctor prescribes an amblyopia training device, and the parents purchase the equipment.

[0097] Home training: Parents should supervise their children's training for 30 minutes daily. The training device has only a fixed program (such as a 5Hz red light flashing frequency) and the difficulty level cannot be adjusted. Children often cry and try to avoid the training due to boredom, requiring parents to forcibly restrain them.

[0098] Follow-up visit: A second check-up at the hospital is scheduled 3 months later. The doctor compares the data before and after the check-up and adjusts the training plan (such as increasing the duration or changing the stimulation frequency). Due to the lack of daily data, the doctor cannot determine the authenticity of compliance (parents may misrepresent training time).

[0099] If the effect is not good (e.g., visual improvement of less than 1 line), the doctor may recommend additional occlusion therapy or surgery.

[0100] Implementation results: Forty children with mild to moderate amblyopia, aged 3-6 years and with matched initial visual acuity (0.3-0.5), were randomly divided into the invention group (n=20, using Example 1) and the control group (n=20, using the comparative scheme). After a 3-month treatment period: In this invention group: the average visual acuity improved by 3.1 lines (from 0.4 to 0.7), the proportion of fixation time in amblyopic eyes increased from 13% to 44%, the training completion rate was 91%, and there were no dropouts.

[0101] Control group: Average visual acuity improved by 1.4 lines (from 0.4 to 0.55), the percentage of amblyopic eyes with longer fixation time increased from 14% to 22%, the training completion rate was 38%, and 5 cases dropped out (parents gave up due to "children's resistance"). During follow-up visits, 3 parents admitted to not persisting with the training.

[0102] This comparative example demonstrates the following significant shortcomings of the existing technical solution: Discontinuous monitoring: Spot checks every 3 months fail to capture daily fluctuations and early deterioration trends.

[0103] Screening and intervention are disconnected: Parents need to manually interpret the reports and set training parameters, which is prone to errors and time-consuming.

[0104] The training lacks adaptability: it is either fixed in difficulty or adjusted only based on game scores, lacks objective physiological feedback, and cannot match the child's real-time state.

[0105] Poor compliance: The training is tedious, lacks motivation and social mechanisms, and children strongly resist it.

[0106] No proactive warning: If the intervention window is missed, the ineffectiveness is often only discovered during follow-up visits.

[0107] Through its integrated, closed-loop, adaptive, and proactive design, this invention achieves significantly better results than existing technologies in all of the above aspects.

[0108] Compared to Examples 1-5 and Comparative Example 1, the traditional approach represented by Comparative Example 1 (regular hospital screening + home amblyopia training device) has significant limitations. First, monitoring is scattered and fragmented; a check-up every three months fails to capture daily developmental fluctuations, leaving doctors without continuous data to support accurate decision-making. Second, screening and intervention are completely disconnected: parents must interpret hospital reports and manually set the training device parameters, a cumbersome and error-prone process. Third, training lacks adaptability: traditional training devices only provide fixed-frequency grating or red light stimulation, with no adjustable difficulty, and no real-time physiological feedback (such as eye movement or VEP) to dynamically adjust training intensity, resulting in uncontrollable training effects. Fourth, compliance is extremely poor: the monotonous training content causes strong resistance in children, requiring parents to forcefully supervise; in this study, the completion rate in the comparative example group was only 38%, with a dropout rate as high as 25%. Finally, there is no proactive warning mechanism; when a child's developmental indicators deteriorate, the system cannot proactively alert the parent, often only discovering intervention failure during a follow-up visit, missing the optimal intervention period.

[0109] In comparison, the five embodiments of this invention demonstrate significant advantages in different scenarios. Embodiment 1 constructs a basic integrated head-mounted system that integrates multidimensional synchronous screening, adaptive game intervention, and continuous monitoring and early warning, improving the visual acuity of amblyopic children by an average of 3.1 lines within 4 weeks, with a training completion rate of 91%. Embodiment 2, targeting myopia prevention, uses dynamic axial length monitoring and alternating near and far focusing games to slow the rate of axial length growth in pre-myopia children by 62% and increase the accommodative amplitude by 2.5D. Embodiment 3's simplified screening device reduces costs by 60%, with a single screening taking only 90 seconds, suitable for large-scale community screening, and achieving a detection accuracy rate of 94%. Embodiment 4's high-precision medical version has measurement errors comparable to large desktop devices and has entered clinical trials in tertiary hospitals. Embodiment 5 integrates behavioral therapy and social competition, increasing the training completion rate of children with previously poor compliance from below 40% to 78%, and increasing the frequency of proactive training requests by 2.5 times. All embodiments achieve closed-loop management of "screening-intervention-monitoring-early warning," utilizing eye-tracking and VEP physiological signals to drive dual-loop adaptive training in real time, and achieving proactive health management through tiered early warning. In comprehensive comparison, this invention surpasses traditional solutions in terms of monitoring continuity, intervention timeliness, training adaptability, and child compliance, providing a complete technical path for the home-based, precise, and intelligent management of children's visual development.

[0110] The above are merely specific embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of protection of the claims.

[0111] In conclusion, the above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A system for monitoring and intervening in pediatric visual development, characterized in that, include: A multidimensional synchronous screening sensor array integrated on a head-mounted frame includes: an eye-tracking sensor, a flexible dry electrode visual evoked potential acquisition electrode, a refractive power sensor, an axial length estimation sensor, an ambient light sensor, a distance sensor, and a posture sensor. The eye-tracking sensor acquires eye position and motion parameters; the flexible dry electrode visual evoked potential acquisition electrode acquires electrical signals from the occipital region without conductive gel; the refractive power sensor measures the refractive state of the eye; the axial length estimation sensor measures the distance from the corneal apex to the retinal pigment epithelium; the ambient light sensor measures ambient illuminance; the distance sensor measures the distance between the eye and the gaze target; and the posture sensor measures the head posture angle. A synchronous acquisition and analysis controller controls all sensors in the sensor array to synchronously acquire data within the same detection time window. The controller generates a multidimensional feature vector based on the acquired data and inputs this vector into a classifier, which outputs a developmental deviation type and severity rating. The classifier outputs a category label and severity score based on the input vector. The developmental deviation type includes at least one of amblyopia, myopia, insufficient accommodative function, or stereopsis deficit. The severity rating includes mild, moderate, or severe levels. An intervention task matching and adaptive game trainer is provided, comprising: an intervention matching unit and a dual-loop adaptive controller; the intervention matching unit selects a corresponding intervention game from a game library according to the developmental deviation type and sets game parameters according to the severity level; the dual-loop adaptive controller includes a first loop and a second loop, the first loop adjusting the game parameters based on real-time acquired eye-tracking fixation position and visual evoked potential signals, and the second loop adjusting the game level based on fixation duration, visual evoked potential amplitude change rate, and task accuracy; wherein, the intervention game is aimed at visual training. The video game program, wherein the game parameters include contrast, spatial frequency, stimulation duration or game difficulty level, the eye-tracking fixation position is the coordinate of the point on the display screen where the fixation direction of a single eye falls, the visual evoked potential signal is the EEG waveform signal caused by visual stimulation, the fixation duration is the duration of a single fixation behavior, the visual evoked potential amplitude change rate is the percentage obtained by dividing the difference between the current amplitude value and the baseline amplitude value by the baseline amplitude value, the task accuracy rate is the ratio of the number of times the game operation is correctly completed per unit time to the total number of operations, and the game level is an integer representing the level of training difficulty; A continuous monitoring and tiered early warning system is provided. During non-screening and non-training periods, the system collects eye-use behavior data at a reduced sampling frequency. Based on this data, it calculates the axial length change rate, the proportion of fixation time in amblyopic eyes, and the refractive error change slope. When any of these indicators exceeds a preset threshold, a Level 1, Level 2, or Level 3 early warning is issued. The eye-use behavior data includes eye distance, ambient illuminance, head posture angle, blink count, or continuous eye use duration. The axial length change rate is the relative change in axial length per unit time. The amblyopic eye fixation time proportion is the percentage of total fixation time in amblyopic eyes relative to the total fixation time in both eyes. The refractive error change slope is the rate of change of refractive error measurement over time. The data management and remote calibrator stores raw data in a built-in encrypted chip, uploads anonymized feature data to the cloud, performs remote online calibration when sensor baseline drift is detected, and provides a data interface for doctors to remotely adjust parameters. The raw data consists of unprocessed signal sequences directly collected by each sensor; the anonymization process removes personally identifiable information and assigns random codes; sensor baseline drift is the deviation of the sensor's output value from its calibration zero point when there is no input signal; and the remote online calibration is a process where the device's built-in program automatically guides the monitored subject to complete a standard fixation task and re-collects sensor data for correction.

2. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The eye-tracking sensor is a binocular infrared eye-tracking camera with a sampling frequency greater than or equal to 120 Hz; the refractive power detection sensor is a Shaker-Hartmann wavefront sensor; and the axial length estimation sensor is an optical coherent ranging sensor based on the principle of low-coherence optical interference.

3. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The flexible dry electrode visual evoked potential acquisition electrode has 3 to 5 dry electrode contacts, which are installed on the nose pad and the inner side of the temple. The contact material is spring-loaded microneedles or conductive fabric. The spring-loaded microneedles are miniature needle-shaped conductors supported by elastic elements. The conductive fabric is a flexible fabric woven from conductive fibers.

4. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, In the dual-loop adaptive controller: the delay of the first closed loop in adjusting game parameters does not exceed 50 milliseconds, and the adjusted game parameters include target brightness, contrast, target size, or background noise; the second closed loop adjusts the game level every 30 seconds, and the game level ranges from 1 to 10; wherein, the target brightness is the display brightness of the interactive game object, the contrast is the brightness ratio between the target and the background, the target size is the pixel diameter of the interactive game object in the display screen, and the background noise is the intensity of random pixel disturbance in the non-target area of ​​the screen.

5. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, In the graded early warning device: the first-level early warning is to emit a prompt sound through a bone conduction speaker or to emit a flashing light through a light-emitting diode; the second-level early warning is to overlay a semi-transparent dynamic image on the display screen or to start a blink training task; the third-level early warning is to send a text message to the guardian's terminal and prompt them to seek medical attention; wherein, the bone conduction speaker is an electroacoustic transducer that transmits sound through skull vibration, the blink training task is a program that guides the monitored person to blink continuously 5 to 10 times, and the guardian's terminal is a smartphone, tablet computer or personal computer.

6. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The remote online calibration process is as follows: when sensor baseline drift is detected, a set of standard fixation points are presented on the display screen, and the child is guided to fixate on each fixation point in turn. At the same time, sensor data is re-acquired for correction. The entire calibration process does not exceed 60 seconds. The standard fixation points are five circular markers located in the center and four corners of the display screen. The sequential fixation means that each fixation point is displayed for 2 to 3 seconds and the monitored person needs to focus their gaze on the point and keep it stable.

7. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The intervention games include: monocular stimulation games for amblyopia, near-far alternating focusing games for insufficient accommodation, and random dot stereogram games for stereoscopic vision impairment; wherein, the monocular stimulation games are games in which only the amblyopic eye can see the clear game screen while the screen of the healthy eye is obscured or blurred, the near-far alternating focusing games are games in which the game target appears alternately at near and far distances, and the random dot stereogram games are games composed of randomly distributed dots and require binocular fusion to identify the hidden stereoscopic shapes.

8. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The system also includes a binocular differential stimulator, which enhances the display of the image corresponding to the amblyopic eye during gameplay and blurs the image corresponding to the healthy eye. The degree of blurring is updated according to the real-time gaze position. The enhanced display increases the brightness or contrast of the image to 120% to 150% of the base value, the blurring display applies Gaussian filtering to the image, and the update according to the real-time gaze position involves recalculating the gaze point coordinates and blurring the healthy eye's image outside a certain area centered on the gaze point when each frame is refreshed. This certain area is a circular area with a radius of 50 to 100 pixels.

9. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The head-mounted frame is augmented reality glasses, which are equipped with binocular see-through waveguide displays, and the brightness of each eye is independently adjustable. The binocular see-through waveguide displays are optical modules that allow some external light to pass through while displaying virtual images. The independent brightness adjustment of each eye means that the brightness of the left and right eye displays is set to different values, with an adjustment range of 0 to 2000 nits.

10. The ophthalmic pediatric vision development monitoring and intervention system as described in claim 1, characterized in that, The classifier is either a random forest classifier or a support vector machine classifier.

Images