Children visual function training system and method based on ciliary muscle adjustment

Abstract

The invention belongs to the technical field of child visual training, and particularly relates to a child visual function training system and method based on ciliary muscle adjustment. The child visual function training system based on ciliary muscle adjustment comprises a collaborative acquisition module and a dynamic training module. The collaborative acquisition module acquires an anterior segment image of an eyeball in real time through a built-in image acquisition unit so as to identify eye features, acquires extraocular muscle pressure deformation data through a built-in pressure acquisition unit, and then associates and analyzes the eye features and the extraocular muscle pressure deformation data through a built-in data fusion unit; a mapping model is established according to the analysis result, the refractive power change of the crystalline lens is calculated, and the calculation result is processed into the ciliary muscle adjustment amount; the dynamic training module determines training parameters in combination with child individual data and historical data acquired by the collaborative acquisition module, and dynamically adjusts training duration and light wave parameters according to the ciliary muscle adjustment amount, thereby solving the problems of poor child training effect and safety risk in the training process.

Description

Technical Field

[0001] This solution belongs to the field of children's vision training technology, specifically involving a children's visual function training system and method based on ciliary muscle accommodation. Background Technology

[0002] In the field of visual function training for children, the real-time and accurate acquisition of ciliary muscle status and the dynamic adaptation of training plans are the core challenges to improving the results, and most current technical solutions have significant shortcomings in this regard.

[0003] The ciliary muscle, a key smooth muscle in the eye that controls the focal length of the lens, directly determines visual accommodation ability through its contraction or relaxation (contraction makes the lens more convex when viewing near objects, and relaxation makes the lens flatter when viewing distant objects). From a physiological perspective, the movement of the ciliary muscle can accurately predict the deformation of the lens because the lens itself changes its curvature by directly pulling on the suspensory ligaments through the contraction and relaxation of the ciliary muscle; there is a direct mechanical linkage between the two. However, conversely, predicting the accommodation of the ciliary muscle through the deformation of the lens is inaccurate. This is because the deformation of the lens can also be affected by other factors such as the lens's own elasticity and the tension of the suspensory ligaments. Single data on lens deformation cannot fully reflect the true contraction or relaxation state of the ciliary muscle.

[0004] The ciliary muscle's anatomical location within the eyeball and its ring-shaped distribution present inherent obstacles to its detection. This ring-shaped structure tightly encloses the lens and is adjacent to the choroid and suspensory ligaments, making it difficult for conventional optical equipment to penetrate the eyeball wall and accurately capture muscle fiber activity. Furthermore, the contraction amplitude of muscle fibers is only at the micrometer level, and subtle movements are easily masked by interference signals such as iris tremors and corneal reflections, making it difficult to extract effective information independently. This leads to two major technical bottlenecks in real-time status acquisition: First, direct measurement is impossible; inference must be made through indirect indicators such as changes in lens curvature and subtle eye movements. However, current acquisition methods are limited, relying solely on images of the anterior segment of the eye to infer the ciliary muscle's regulation of the lens, making it difficult to establish a precise correlation between indirect indicators and the actual state of the ciliary muscle. Second, children's ciliary muscles are not yet fully developed, exhibiting a large accommodative amplitude (e.g., 12-14D in 6-8 year olds, approximately 8D in adults) and are prone to spasms. Their state fluctuates more dramatically with training duration and eye strain intensity, making traditional static acquisition (such as single refraction data) unable to capture dynamic changes, resulting in insufficient data timeliness.

[0005] More importantly, the state of the ciliary muscle plays a decisive role in the formulation and adjustment of training plans. Most current testing methods still revolve around the lens. Whether it's obtaining the refractive power of the lens through optometry equipment or measuring changes in lens curvature using optical instruments, these are essentially tests of visual acuity-related "outcome indicators," reflecting only the current visual level (such as myopia degree or degree of refractive error), but failing to address the dynamic functional state of the ciliary muscle behind the visual results. If the ciliary muscle is in a "spasmodic state" (e.g., the ciliary muscle cannot relax after continuous near work), continued high-intensity training will exacerbate fatigue and even lead to an increase in myopia; if it is in an "active state" (e.g., a sensitive accommodative response), low-intensity training cannot fully activate the accommodative reserve. Most current training programs generally ignore this connection, and most training devices rely on human experience for parameter adjustments (e.g., manually modifying based on visual improvement), failing to acquire the ciliary muscle state in real time through technological means. For example, when a child's ciliary muscle pressure-related signal (extraocular muscle deformation) indicates fatigue, the training device cannot automatically reduce the training intensity, easily causing temporary blurred vision after training. Summary of the Invention

[0006] The purpose of this solution is to provide a visual function training system and method for children based on ciliary muscle accommodation, in order to solve the problems of lack of real-time acquisition of ciliary muscle status and the disconnect between ciliary muscle status and training plan, which leads to poor training results or safety risks for children.

[0007] To achieve the above objectives, this solution provides a visual function training system for children based on ciliary muscle accommodation, comprising: The collaborative acquisition module includes an image acquisition unit, a pressure acquisition unit, and a data fusion unit; The image acquisition unit acquires images of the anterior segment of the eye in real time and identifies ocular feature parameters in the images of the anterior segment of the eye. The pressure acquisition unit includes a pressure sensor located on the skin surface at the attachment point of the extraocular muscles on the outer side of the eye socket, which collects pressure deformation data of the extraocular muscles caused by eyeball convergence movement in real time. The data fusion unit acquires ocular feature parameters and pressure deformation data collected at the same time, and performs correlation analysis. It establishes a mapping model based on the changes in pressure deformation data and lens curvature. The lens refractive power change data is indirectly calculated based on the ocular feature parameters using a thin lens refractive power conversion algorithm as the first data. The pressure deformation data is input into the mapping model to calculate the real-time state of the ciliary muscle, and the lens refractive power change data in the ocular feature parameters is calculated based on the real-time state as the second data. The data fusion unit compares the first data with the second data. If the comparison result falls within the preset error threshold, the real-time state is processed into the real-time adjustment amount of the ciliary muscle according to the acquisition time. If the comparison result exceeds the error threshold, the eye feature parameters and pressure deformation data are obtained again from the image acquisition unit and the pressure acquisition unit, and the first data and the second data are calculated and compared. The dynamic training module acquires individual child data and historical data from the collaborative acquisition module, formulates training parameters according to a preset training cycle, and dynamically adjusts the training parameters based on the real-time adjustment of the ciliary muscle. The real-time state includes fatigue state and active state, and the training parameters include training duration and light wave combination. The module dynamically adjusts the training duration, target frequency, proportion and contrast of different types of light in the light wave combination based on the real-time state, and adjusts the training mode in the parameters based on the continuous real-time state.

[0008] And, a ciliary muscle-based visual function training method for children using a ciliary muscle-based visual function training system.

[0009] The principle and technical effects of this solution are as follows: This solution relies on the dynamic adaptation mechanism between ciliary muscle state and training parameters, constructing a closed loop combining perception, analysis, and control through multi-dimensional acquisition and intelligent regulation. The collaborative acquisition unit, acting as the state perception layer, captures internal features such as lens curvature and iris displacement using an eye image acquisition unit, while simultaneously acquiring external correlation signals of extraocular muscle deformation through an orbital pressure sensor. A mapping model between lens accommodation and extraocular muscle pressure is established through a data fusion unit, overcoming the bottleneck of the inability to directly measure the ciliary muscle and achieving indirect determination of tension, relaxation, and fatigue states. The dynamic training module adjusts training parameters based on real-time state, reducing the target switching frequency and increasing the green light ratio to promote relaxation in fatigue states, and increasing the frequency and red light ratio to strengthen training in active states. Simultaneously, automatic mode switching avoids overtraining. Personalized plans, including duration and light wave combinations, are generated based on individual differences such as the child's age and degree of amblyopia, forming a complete closed loop. This solves the problem that traditional technologies cannot quantify the dynamic changes of the ciliary muscle in children, improves the accuracy of state determination, achieves dynamic and precise training, reduces the risk of ciliary muscle spasm, and increases the amplitude of accommodation.

[0010] This solution further enhances the reliability of state determination through multi-source data fusion. The collaborative acquisition of anterior segment images and pressure deformation data can offset interference from single data sources (such as the influence of lens elasticity on anterior segment images), making ciliary muscle state determination more realistic and maintaining stable judgment even in scenarios involving frequent subtle eye movements in children.

[0011] The dynamic parameter control and multispectral synergy principle in this scheme enhance training synergy. The multispectral module controls the proportion of different types of light in the light wave combination. For example, controlling the proportion of green light in the light wave combination adjusts the pupil diameter and enhances the linkage between the ciliary muscle and the visual system; controlling the proportion of red light in the light wave combination stimulates choroidal blood circulation. These two factors, together with ciliary muscle training, form a synergistic effect, improve the efficiency of ocular blood circulation, provide physiological support for the activation of visual cells in children with amblyopia, and make the training effect more lasting.

[0012] This approach combines contrast adjustment with dynamic training to enhance the accommodative stability of the ciliary muscle in complex visual environments. For example, by dynamically adjusting the image contrast in a simulated complex visual environment, combined with real-time training of the ciliary muscle, the ciliary muscle's adaptability under different visual conditions can be improved, further consolidating its accommodative function.

[0013] This program combines personalized plan generation with dynamic adaptation to accurately match children's developmental differences. Targeting different stages of ciliary muscle development in children aged 6-12, it customizes training plans based on age, amblyopia severity, and historical data, avoiding the limitations of generic models, shortening the amblyopia correction cycle, and achieving better myopia control results.

[0014] The linkage between state perception and mode switching in this solution enhances training safety. By identifying ciliary muscle fatigue in advance (rather than relying solely on external behavioral monitoring), the training mode is automatically switched. For example, if the ciliary muscle is repeatedly in a state of fatigue, the training mode is switched to a brain-based training mode, avoiding temporary vision loss due to overload. This significantly improves the safety and compliance of children's training and reduces training resistance.

[0015] In summary, this solution uses a multi-source closed-loop system to determine ciliary muscle state in real time, dynamically adapts training parameters, and precisely improves the results. It also solves the problems of incomplete real-time state acquisition and training disconnect that lead to poor training results or safety risks for children.

[0016] Furthermore, the data fusion unit is also used to verify the accuracy of the AC / A constant based on synchronously acquired pressure deformation data and ocular characteristic parameters. During accuracy verification, when the individual error of AC / A is ≤ ±0.3Δ / D and the error of the mapping model after calibration by UBM and open-field refractometer is ≤ ±0.25D, the objective accommodative response is first recorded using an infrared video refractometer or open-field automatic refractometer as the gold standard for accommodative amount. Simultaneously, the instantaneous convergence is measured using the prism cover method or synoptophore as the gold standard for convergence, ensuring that the gold standards for accommodative amount and convergence are obtained. The difference between the AC / A and the synchronously acquired pressure deformation data and ocular characteristic parameters is ≤ ±0.3Δ / D. Then, accommodative stimulation is applied at different viewing distances and repeated multiple times to plot the stimulus-response curve. If the linear fitting R² of the stimulus-response curve is <0.9 or the intercept is >0.5Δ, piecewise fitting or quadratic curve correction is performed. The amount of far and near phoria is measured by the occlusion and unocclusion method. Individual models are built for those with phoria >8Δ. For those with unequal refractive power between the two eyes >1D, the AC / A of the left and right eyes is calibrated and the average value is taken. The measurement is repeated every other day to ensure that the difference between the two AC / A measurements is <±0.2Δ / D.

[0017] This solution employs a data fusion unit to precisely verify the AC / A constant, forming a multi-stage closed-loop verification mechanism. The AC / A error is controlled within ±0.3Δ / D, and the mapping model error is locked within ±0.25D, directly improving the accuracy of ciliary muscle state estimation and providing reliable data support for the dynamic training module. Simultaneously, considering the large accommodative fluctuations in children, automatic correction is implemented when stimulus-response curves become abnormal, preventing the linear assumption from failing and enhancing the stability of this solution. Separate models are created for special children with hysteria >8Δ or refractive power difference >1D, with retesting every other day to ensure consistency, achieving individualized adaptation and addressing the problem of insufficient adaptation for special groups in traditional methods. Furthermore, the verified AC / A can refresh training parameters in real time, forming a zero-delay closed loop combining verification and regulation in collaboration with the dynamic training module. This improves training accuracy and reduces deviations caused by individual differences through multi-stage verification. This multi-effect synergistic closed-loop design is difficult to achieve with a single verification method.

[0018] Furthermore, the dynamic training module includes a dynamic adjustment unit and a multispectral adjustment unit. When the dynamic adjustment unit detects that the ciliary muscle is in a fatigued state in real time, it reduces the target switching frequency to 8-12 times / minute, while increasing the proportion of 500-550nm green light to 60% to promote relaxation. When an active state is detected, it increases the target switching frequency to 15-20 times / minute, while increasing the proportion of 620-660nm red light to 50% to strengthen training. The multispectral adjustment unit is used to emit specific light waves of 430-680nm, where green light adjusts the pupil diameter by alternating brightness, synchronously training the synergy between the ciliary muscle and the pupil, and red light is used for photonutrient supplementation to stimulate accelerated blood circulation in the choroid. This unit achieves training enhancement through the characteristics of light waves and enhances the linkage of the visual system by adjusting the pupil.

[0019] The dynamic adjustment unit of this solution precisely adjusts the target switching frequency and light wave ratio based on the real-time state of the ciliary muscle. This promotes relaxation when fatigued and strengthens training when active, ensuring the training's relevance and effectiveness. The specific light waves emitted by the multispectral adjustment unit not only utilize the characteristics of green and red light to achieve coordinated training of the ciliary muscle and pupil, and stimulation of choroidal blood circulation, but also enhance the linkage of the visual system through pupillary adjustment. The unit achieves the dual functions of training enhancement and linkage. For example, when children perform near-field vision training, the alternating brightness of green light causes the pupil to continuously contract and dilate. During this process, the ciliary muscle also adjusts accordingly, thereby strengthening the cooperation between the two. This coordinated training allows children to achieve smoother eye accommodation when rapidly switching viewing distances, reducing the likelihood of temporary blurred vision. This differs from previous methods that only focused on adjusting a single parameter, significantly improving the comprehensiveness and practicality of the training.

[0020] Furthermore, the dynamic training module also includes a contrast control unit, a mode switching unit, and a scheme generation unit. The contrast control unit dynamically reduces the contrast of the training image in the training parameters to enhance the ciliary muscle's regulatory stability in complex visual environments. The mode switching unit supports automatic switching between physical training and brain-based training. When the data fusion unit determines that the ciliary muscle is fatigued for three consecutive times, it switches to brain-based training for 1-2 minutes. The scheme generation unit combines the child's age, amblyopia degree, and historical data from the collaborative acquisition module to output the daily training duration and light wave combination mode, achieving precise control of training and recovery effects. It also works in conjunction with the dynamic adjustment unit to ensure that the generated scheme is updated in real time according to the ciliary muscle's state. The amblyopia degree is obtained by matching ocular feature parameters with stored amblyopia features.

[0021] The contrast control unit dynamically reduces the contrast of the training screen to simulate a complex visual environment, enhancing the ciliary muscle's accommodative stability under different lighting and clarity conditions. The mode switching unit automatically switches to brain-based training when the ciliary muscle is continuously fatigued, avoiding ciliary muscle damage caused by overtraining and relaxing the eye muscles through short-term brain-based training, achieving a dynamic balance between training and recovery. The plan generation unit generates personalized plans based on the child's age, degree of amblyopia, and historical data, and updates them in real time in conjunction with the dynamic adjustment unit. The degree of amblyopia is determined by matching eye feature parameters to ensure plan suitability. At the same time, the plan generation unit integrates and analyzes multi-dimensional data, making the plan both targeted and flexible.

[0022] Of particular note is the automatic switching mechanism of the mode switching unit, which is not just a simple training mode conversion, but also builds an active defense system for eye protection. For example, after a child has undergone 20 minutes of object-based training, the data fusion unit determines that the ciliary muscle is fatigued three times in a row. The mode switching unit immediately starts a 1-minute brain-based training, which replaces physical object focusing with visual imagination training, allowing the ciliary muscle to recover its vitality in a stress-free state. This mode, which combines predictive protection with alternating training, reduces the risk of ciliary muscle spasm compared to the traditional manual switching mode. This effect is particularly significant in training scenarios after children have been studying at close range for a long time. For example, when conducting visual function training after school, this solution can detect eye fatigue in time and switch modes to avoid increasing the burden on the eyes due to continuous training.

[0023] Furthermore, it also includes a training device, which comprises a housing, a display component, and a fastening component. The display component is located inside the housing and fixedly connected to it, while the fastening component is fixedly connected to the outside of the housing to secure the housing of the training device around the eye. The housing has an orbital hole adapted to the eye socket, and the outer side of the housing around the orbital hole has a pressure detection area on the skin surface corresponding to the attachment point of the extraocular muscles on the lateral side of the eye socket. Several pressure sensors of the pressure acquisition unit are fixedly connected to the pressure detection area. The image acquisition unit is a binocular miniature camera integrated above the orbital hole inside the training device, and the binocular miniature camera is equipped with an infrared fill light. The display component is also fixedly connected inside the housing, and the display component is used to receive training parameters provided by the dynamic training module and display the training image.

[0024] The orbital openings on the outer shell, which are adapted to the eye socket, and the surrounding pressure detection area allow the pressure sensor to accurately correspond to the attachment point of the extraocular muscles, improving the accuracy of pressure deformation data acquisition and providing a reliable foundation for data fusion unit analysis. The binocular miniature camera, paired with an infrared fill light, ensures clear acquisition of images of the anterior segment of the eyeball under different lighting conditions, guaranteeing the stability of eye feature parameter extraction. The display component is fixedly connected to the outer shell, which can stably receive training parameters and display images. With the fastening components, the outer shell is fixed around the eye, ensuring the stability of the device position during training and reducing training interruptions caused by shaking.

[0025] The design of the pressure detection area, perfectly adapted to the eye socket opening, not only accurately collects pressure data but also maintains good contact between the sensor and skin during light activities such as children's movements. For example, if a child sits for a long time and becomes uncomfortable while training at home, turning or swaying in the chair, the pressure sensor will not dislodge due to movement, ensuring continuous data collection and improving the device's flexibility. Furthermore, the integrated design of the display component, camera, pressure sensor, and casing makes the training device a single unit, requiring no additional equipment. Children can quickly begin training at home without complicated installation and setup, even without professional guidance. The coordinated work of all components increases the effective training percentage within the same training time, achieving better training results.

[0026] Furthermore, the pressure acquisition unit integrates a contact sensing component for synchronously acquiring intraocular pressure data; the image acquisition unit analyzes corneal reflection images to determine the tear film breakup time, forming a collaborative verification method for intraocular pressure and tear film status; the collaborative verification method uses intraocular pressure fluctuation thresholds and tear film breakup time thresholds as auxiliary verification conditions, and correlates them with pressure deformation data acquired by the pressure sensor and ocular feature parameters acquired by the image acquisition unit; when the intraocular pressure exceeds a preset fluctuation range, the data fusion unit corrects the weights of the correlation parameters in the mapping model, where the correlation parameters are weight coefficients used in the mapping model to correlate lens curvature and extraocular muscle pressure; when the tear film breakup time is shorter than a preset breakup threshold, the multispectral adjustment unit adjusts the light wave ratio to improve the ocular surface status.

[0027] This solution employs a collaborative verification method that integrates contact sensing components to collect intraocular pressure data in a pressure acquisition unit and analyzes tear film breakup time using an image acquisition unit. This method can correlate and analyze pressure deformation data with ocular feature parameters. When intraocular pressure exceeds a preset fluctuation range, the data fusion unit corrects the weights of the correlated parameters in the mapping model, improving the accuracy of the correlation between lens curvature and extraocular muscle pressure, and providing a more reliable basis for estimating ciliary muscle accommodation. When tear film breakup time is shorter than a preset threshold, the multispectral adjustment unit adjusts the light wave ratio to improve the ocular surface condition and ensure image acquisition clarity. The dynamic correction of correlated parameter weights and the linkage with ocular surface condition improvement are achieved in real-time through the data fusion unit. This ensures the accuracy of the correlation between pressure deformation data and ocular feature parameters while rapidly optimizing the ocular surface environment through light wave adjustment, reducing training interruptions caused by fluctuations in ocular surface condition. For example, when children train in a dry environment, the tear film breakup time is shortened to 3 seconds. After the multispectral adjustment unit increases the proportion of green light, the tear film condition improves. At the same time, the data fusion unit automatically corrects the weights of the associated parameters due to changes in the tear film, avoiding the miscorrelation between pressure and optical data under the influence of the tear film. Even if children blink frequently, this solution can still output training parameters stably, ensuring the continuity of training and improving training accuracy. This real-time linkage adjustment of multi-dimensional data allows training to be carried out efficiently even in complex environments.

[0028] Furthermore, a flexible Ag / AgCl electrode is fixedly connected to the outer shell of the training device at the corresponding position of the tragus and the corresponding position of the corneal limbus at the edge of the orbital aperture. The flexible Ag / AgCl electrode has a sampling frequency of 250Hz and is used to collect changes in corneal surface potential. When the ciliary muscle contracts, the epithelial ion pump activity is enhanced, causing the corneal surface potential to drift towards a negative value by 0.5-1.2mV. The data fusion unit correlates and analyzes the corneal surface potential change data with intraocular pressure data, tear film breakup time, and pressure deformation data. When the linear deviation between the potential drift amplitude and the adjustment amount exceeds the preset potential adjustment deviation, the correlation parameter weights of the mapping model are simultaneously corrected. The correlation parameters are the weight coefficients in the mapping model used to correlate lens curvature and extraocular muscle pressure. At the same time, the dynamic training module controls the multispectral adjustment unit to adjust the proportion of red light to enhance ion pump activity.

[0029] This scheme collects corneal surface potential changes by placing flexible Ag / AgCl electrodes at corresponding positions on the outer shell of the training device. It then correlates these changes with data on intraocular pressure, tear film breakup time, and pressure deformation. When the linear deviation between the potential drift amplitude and the accommodation amount exceeds the preset potential accommodation deviation, the scheme simultaneously corrects the weights of the mapping model's correlation parameters and adjusts the proportion of red light. This approach not only accurately reflects the ciliary muscle contraction state through potential changes, providing a reliable basis for accommodation measurement and improving the accuracy of data fusion unit analysis, but also enhances ion pump activity and improves ciliary muscle function by adjusting the proportion of red light. The variable of corneal surface potential change can simultaneously verify the ciliary muscle state and assist in optimizing training interventions. For example, when children experience decreased epithelial ion pump activity due to eye fatigue, and the linear deviation between the corneal surface potential drift amplitude and the accommodation amount exceeds the preset value, this scheme corrects the correlation parameter weights to ensure measurement accuracy while increasing the proportion of red light to enhance ion pump activity, promoting the recovery of ciliary muscle contraction function. Even if children do not show obvious fatigue symptoms during training, early intervention through potential changes can prevent further decline in accommodation function, making training more proactive and targeted.

[0030] Furthermore, the data fusion unit performs four-dimensional correlation analysis on corneal surface potential change data, intraocular pressure data, tear film breakup time, and pressure deformation data to establish a dynamic coupling model combining potential, intraocular pressure, tear film, and pressure. When the linear deviation between the potential drift amplitude and the adjustment amount exceeds the preset potential adjustment deviation, and the intraocular pressure exceeds the fluctuation range or the tear film breakup time is shorter than the preset breakup threshold, the weights of the mapping model correlation parameters are simultaneously corrected, the proportion of red and green light in the multispectral adjustment unit is adjusted, and the operation of the contrast control unit is paused.

[0031] This solution utilizes a data fusion unit to perform four-dimensional correlation analysis on corneal surface potential changes, intraocular pressure, tear film breakup time, and pressure deformation data, establishing a dynamic coupling model. When specific conditions are met, multiple operations are triggered simultaneously, improving the comprehensiveness and accuracy of data correlation and providing a more reliable basis for correcting the weights of the mapping model's correlation parameters. Furthermore, a multispectral adjustment unit adjusts the light ratio and pauses contrast modulation, ensuring the effectiveness and comfort of training. The dynamic coupling model enables cross-dimensional data verification. For example, when children train in a dry environment, tear film breakup time is shortened, and abnormal potential drift occurs. The dynamic coupling model can use tear film data to determine whether the potential drift is caused by ciliary muscle contraction, avoiding misjudgment based on single data. If a tear film issue is confirmed, adjusting the green light ratio improves the ocular surface condition while pausing contrast modulation to prevent further decline in visual acuity. Adjusting the red light ratio enhances ion pump activity and synergistically promotes ciliary muscle function recovery. This multi-data-driven collaborative intervention allows training to be conducted accurately and efficiently even in complex ocular conditions, significantly improving the adaptability and training effectiveness of this solution.

[0032] Furthermore, the dynamic coupling model introduces a dynamic baseline calibration coefficient K. K is calculated based on the initial drift value of corneal surface potential, the baseline value of intraocular pressure, and the tear film breakup reference time according to the formula K = 0.3 × potential baseline + 0.2 × intraocular pressure baseline + 0.5 × tear film baseline, and is automatically updated every 30 seconds. The K value is embedded in the reliability score calculation of the dynamic coupling model. When the K fluctuation is ≤5%, the score weight is allocated according to the original ratio. When the K fluctuation exceeds 5%, the score weight of intraocular pressure and tear film data is increased by 20%. At the same time, the K value is linked with the multispectral adjustment unit. When the K fluctuation exceeds 5%, except for the adjustment of red light and green light, the supplementary light intensity increases linearly with the K value deviation, and triggers the data fusion unit to shrink the error threshold to ±0.2Δ / D.

[0033] By introducing a dynamic baseline calibration coefficient K into the dynamic coupling model, dynamic calibration of multi-dimensional data and cross-unit collaborative intervention are realized. The K value is automatically updated every 30 seconds, which can adapt to changes in children's eye status in real time. When the K fluctuation is ≤5%, the original scoring weight allocation is maintained to ensure the stability of data correlation analysis. When the K fluctuation exceeds 5%, the scoring weight of intraocular pressure and tear film data is increased and the multispectral adjustment unit is linked to enhance supplementary lighting and shrink the error threshold. This not only improves the influence of key data under abnormal conditions, but also ensures the image acquisition quality through supplementary lighting enhancement, while shrinking the error threshold improves the strictness of data verification. Among them, the K value, as a unified variable, can simultaneously achieve multiple functions such as data benchmark calibration, scoring weight adjustment, supplementary light intensity control, and error threshold adjustment. For example, when children are training at home in the evening, the indoor light gradually dims, causing tear film data fluctuations and K value fluctuations of up to 6%. This solution immediately increases the scoring weight of tear film data, increases supplementary light intensity by 12%, and shrinks the error threshold to ±0.2Δ / D. This not only avoids insufficient light affecting image acquisition, but also ensures data reliability through stricter error standards. At the same time, the adjustment of red and green light continues normally to improve the ocular surface condition and ciliary muscle function, so that the training accuracy in complex environments is not affected, and the proportion of effective training data is increased to a high level. Attached Figure Description

[0034] Figure 1 This is a schematic diagram of the functional modules of the children's visual function training system based on ciliary muscle accommodation in an embodiment of the present invention.

[0035] Figure 2 This is a flowchart illustrating the accuracy verification process of the AC / A constant by the data fusion unit in this embodiment of the invention. Detailed Implementation

[0036] The following will describe the concept and technical effects of the present invention clearly and completely with reference to embodiments, so as to fully understand the purpose, features and effects of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are all within the scope of protection of the present invention. like Figure 1 As shown, the children's visual function training system based on ciliary muscle accommodation includes: The collaborative acquisition module includes an image acquisition unit, a pressure acquisition unit, and a data fusion unit; The image acquisition unit acquires images of the anterior segment of the eye in real time and identifies ocular feature parameters in the images of the anterior segment of the eye. The pressure acquisition unit includes a pressure sensor located on the skin surface at the attachment point of the extraocular muscles on the outer side of the eye socket, which collects pressure deformation data of the extraocular muscles caused by eyeball convergence movement in real time. The data fusion unit acquires ocular feature parameters and pressure deformation data collected at the same time, and performs correlation analysis. It establishes a mapping model based on the changes in pressure deformation data and lens curvature. The lens refractive power change data is indirectly calculated based on the ocular feature parameters using a thin lens refractive power conversion algorithm as the first data. The pressure deformation data is input into the mapping model to calculate the real-time state of the ciliary muscle, and the lens refractive power change data in the ocular feature parameters is calculated based on the real-time state as the second data. The data fusion unit compares the first data with the second data. If the comparison result falls within the preset error threshold, the real-time state is processed into the real-time adjustment amount of the ciliary muscle according to the acquisition time. If the comparison result exceeds the error threshold, the eye feature parameters and pressure deformation data are obtained again from the image acquisition unit and the pressure acquisition unit, and the first data and the second data are calculated and compared. The dynamic training module acquires individual child data and historical data from the collaborative acquisition module, formulates training parameters according to a preset training cycle, and dynamically adjusts the training parameters based on the real-time adjustment of the ciliary muscle. The real-time state includes fatigue state and active state, and the training parameters include training duration and light wave combination. The module dynamically adjusts the training duration, target frequency, proportion and contrast of different types of light in the light wave combination based on the real-time state, and adjusts the training mode in the parameters based on the continuous real-time state.

[0037] Specifically, the thin lens refractive power conversion algorithm calculates the change in lens refractive power through three steps: parameter acquisition, modeling calculation, and dynamic output. First, the image acquisition unit non-invasively captures images of the anterior segment of the eye, extracting morphological parameters such as the radius of curvature of the anterior and posterior surfaces of the lens, and the central thickness *d*. Simultaneously, it calls upon the child-specific media refractive index database (including the refractive index of the lens, aqueous humor, and vitreous humor) built into the data fusion unit. Next, based on the "thin lens assumption" (e.g., the thickness of a child's lens is 3-4 mm, much smaller than the focal length of 10-20 mm in accommodative state), the modeling is simplified. The basic value of the lens's optical refractive power is first calculated using the marker formula (thin lens refractive power calculation formula), and then... The axial length L measured by the binocular parallax algorithm is used to compensate for the differences in children's axial length development through a correction formula to obtain the actual refractive power. Finally, the sampling frequency is synchronized with the collaborative acquisition module to output the refractive power change data, and filter out abnormal values ​​such as "three consecutive sampling deviations > ±0.1D" (for example, the minimum single accommodation amplitude of ciliary muscle in children aged 6-12 years is about 0.2D, and ±0.1D is less than the minimum physiological accommodation amount, so it avoids filtering the real accommodation signal; and the refractive power fluctuations caused by interference such as blinking and eye movement are instantaneous and discontinuous, and the probability of exceeding ±0.1D three times in a row is extremely low, while the real ciliary muscle accommodation fluctuations, such as fatigue spasm, are continuous) to ensure the contrast accuracy of the input data fusion unit.

[0038] The marker formula is: F=(n-1)×[(1 / R1)-(1 / R2)+(n-1)×d / (n×R1×R2)]; where F is the refractive power of the lens (unit: D); n is the refractive index of the lens (e.g., 1.41-1.42 for children, and then slightly adjusted with age); R1 is the radius of curvature of the anterior surface of the lens; R2 is the radius of curvature of the posterior surface of the lens; d is the central thickness of the lens).

[0039] From the perspective of optical theory, the essence of refractive power F is the lens's ability to deflect light. It follows two major laws: "focal length is inversely proportional to refractive power" and "the deflection of light increases with the addition of the deflection of light." For thin lenses, since the thickness is negligible, the total refractive power is the sum of the deflection of the anterior and posterior surfaces. The Marker formula is derived based on this characteristic and can directly establish the relationship between lens morphology parameters and refractive power. For example, when the ciliary muscle contracts and makes the lens more convex (R1 decreases), F increases, which is consistent with the physiological regulatory mechanism of "the lens's refractive power increases when looking at near objects." From the perspective of ocular physiological adaptation, children's lenses naturally meet the thin lens assumption, with a thickness of only 1 / 2 to 1 / 5 of the focal length (in the fields of pediatric ocular physiology and ophthalmic optics, the proportional relationship that "the thickness of children's lenses is only 1 / 2 to 1 / 5 of their focal length in the accommodative state" is a conclusion drawn from actual measurements of children's lens physiological parameters, optical focal length calculations, and industry consensus, and is common knowledge in the professional field). Furthermore, they do not exhibit the nuclear hardening phenomenon seen in adults, and their refractive index distribution is approximately uniform, eliminating the need for complex gradient refractive index modeling. This further verifies the rationality of the thin lens simplification and ensures that the algorithm's calculation logic conforms to both physical laws and the structural characteristics of children's eyes.

[0040] The data fusion unit is also used to verify the accuracy of the AC / A constant based on synchronously collected pressure deformation data and eye feature parameters. The AC / A constant is the ratio of accommodative convergence to accommodation, with the unit "Δ / D". It is a core physiological parameter of optometry, representing "the amount of binocular convergence function that accompanies each 1 diopter (D) of accommodation produced by the eye", and is used to reflect the synergistic relationship between accommodation and convergence (the stronger the accommodation, the more obvious the convergence of the binocular visual axes). When verifying accuracy, if the individual error of AC / A is ≤ ±0.3Δ / D (multiple ophthalmological studies have indicated that an AC / A measurement error exceeding ±0.5Δ / D will significantly affect the diagnostic accuracy of strabismus, heterophoria, etc., this scheme sets a stricter ±0.3Δ / D, higher than the conventional diagnostic accuracy requirement, to ensure that the training intervention is based on a reliable physiological model) and the mapping model, after calibration by UBM (ultrasound biomicroscope) and open-field optometry, has an error ≤ ±0.25D (the ciliary muscle accommodation amplitude in children aged 6-12 years can reach 12-14D, and the minimum effective accommodation is approximately 0.25D), then, Figure 2As shown, objective accommodative response was first recorded using an infrared video refractometer or an open-field automatic refractometer as the gold standard for accommodative quantity. Simultaneously, instantaneous convergence was measured using the prism cover method or synoptophore as the gold standard for convergence. This ensured that the difference between the AC / A ratio obtained from the gold standard for accommodative quantity and the gold standard for convergence (also known as "binocular convergence function," referring to the physiological response intensity when the visual axes of both eyes converge from infinity to a near target; essentially, it is the amount of visual axis deflection caused by extraocular muscle contraction, measured in prism diopters Δ) and the synchronously acquired pressure deformation data and ocular characteristic parameters was ≤ ±0.3Δ / D. Then, accommodative stimuli were applied at different visual distances and repeated multiple times, and stimulus-response curves were plotted. If the linear fit of the stimulus-response curve was R² < 0.9 (R² is the goodness of fit of the regression model; under normal circumstances, the accommodation and convergence of the human eye have an approximately linear relationship, especially in children, where the AC / A ratio is stable and the linear fit is good), the desired result was considered. In visual science and strabismus / latent strabismus assessment, R² ≥ 0.9 is a general statistical threshold for judging whether the stimulus-response relationship is stable. If R² < 0.9, it indicates the presence of a nonlinear accommodative response, such as accommodative spasm, excessive / insufficient convergence (linear models will mislead the setting of training parameters) or an intercept > 0.5Δ (Δ is the unit symbol for prism diopters), in which case piecewise fitting or quadratic curve correction is performed. The amount of near and far latent strabismus is measured by the occlusion and de-occlusion method. Individuals with latent strabismus > 8Δ are modeled separately. For individuals with unequal refractive power with a difference of > 1D between the two eyes, the AC / A of the left and right eyes is calibrated and the average value is taken. The measurement is repeated every other day to ensure that the difference between the two AC / A measurements is < ± 0.2Δ / D (multiple studies have shown that the natural fluctuation range of AC / A in normal children or adults in the short term (e.g., 24–48 hours) is ≤ ± 0.2Δ / D; exceeding this range suggests possible accommodative spasm, fatigue, measurement error, or latent strabismus fluctuation).

[0041] Specifically, when the data fusion unit starts the AC / A constant accuracy verification, it first calls the image acquisition unit and the pressure acquisition unit to synchronously acquire 3 sets of eye feature parameters and pressure deformation data, with a 2-second interval between each set of acquisitions. During this period, abnormal data caused by blinking or violent eye movement is automatically removed. After ensuring that the individual AC / A error is ≤ ±0.3Δ / D and the calibration error of the mapping model is ≤ ±0.25D, the infrared video refractometer is controlled to complete the adjustment response recording within 5 seconds. At the same time, the prism cover method is triggered to automatically switch the prism power, recording the convergence value once per second. After 10 seconds, the average value is taken as the gold standard for convergence value (assuming that 6-8 switching cycles are needed to eliminate single-point error, and the head position deviation is <2° and the adjustment drift is <0.1D within 10 consecutive seconds, the average of 10 points can reduce the standard error of convergence value from ±0.3Δ to ±0.1Δ, which just meets the total budget of ±0.3Δ / D). This ensures that the difference between the AC / A obtained from the gold standard of adjustment value and the synchronously acquired data is ≤ ±0.3Δ / D.

[0042] Different viewing distances were set to 20cm (corresponding to 5D), 33cm (approximately equal to 3D), 50cm (corresponding to 2D), and 100cm (corresponding to 1D). 1D, 2D, 3D, and 5D adjustable stimuli were applied at each viewing distance. Each stimulus level was repeatedly sampled 5 times. During the sampling process, the image contrast was fixed at 50% through the contrast control unit to reduce the interference of ambient light on data stability.

[0043] The FCC specifies that 20cm is a common testing distance for electromagnetic exposure (RF) exposure of mobile devices, especially for devices worn on the body or close to the head (such as mobile phones and Bluetooth headsets). The "Guidelines for Appropriate Technologies for Myopia Prevention and Control in Children and Adolescents (2021)" clearly states that the eyes should be one foot (approximately 33 cm) away from the book when reading and writing. ISO 9241-5:2022 (Ergonomics of human-system interaction — Part 5: Workstation layout and postural requirements) recommends a viewing distance range of 50cm to 100cm (adjusted according to screen size, resolution, and user's vision). Based on the formula: Accommodation requirement (D) = 1 / viewing distance (m), the corresponding accommodation requirements for 20cm, 33cm, 50cm, and 100cm can be obtained.

[0044] Children aged 6-12 years have a higher eye movement frequency (approximately 3-5 times / second) than adults. A single data acquisition (e.g., 300ms / frame) may capture "blinking moments" (leading to missing eye feature parameters) and "eye deviation from the target" (leading to deviation in refractive power measurement). Such errors can reach ±0.15D in a single acquisition. By processing 5 acquisitions using the "Grubbs criterion" (outlier detection algorithm), 1-2 outliers (such as deviations caused by blinking) can be effectively eliminated, and the mean error of the remaining 3-4 valid data points can be reduced to within ±0.05D. Fixing the image contrast at 50% in the center of the "adjustment plateau area" is the most economical integer value with the smallest error and the lowest individual variability.

[0045] When measuring heterophoria, the multispectral adjustment unit first emits white light to calibrate the eye position, then performs occlusion and de-occlusion operations with a 3-second interval between each operation. The average value is taken after three consecutive measurements. Individuals with heterophoria >8Δ are modeled separately. For individuals with unequal refractive power (between eyes >1D), the AC / A ratio of the left and right eyes is calibrated separately, and the average value is taken. Measurements are repeated every other day to ensure the difference in AC / A ratio is <±0.2Δ / D. For individuals requiring separate modeling, the dynamic training module temporarily saves their historical training parameters. After AC / A verification is completed, the new parameters are fused with historical data to generate a transitional training scheme.

[0046] In one specific implementation of this solution, the data fusion unit significantly improves the accuracy and reliability of AC / A constant verification through precise multi-stage operations, providing a solid data foundation for subsequent training. Synchronous acquisition and removal of outlier data ensures the quality of initial data and avoids interference from outliers in the verification results; repeated acquisition at multiple viewing distances and stimulus levels, along with fixed contrast, reduces the impact of environmental factors on the data, making the stimulus-response curve more realistic; calibration of eye position using a multispectral adjustment unit and averaging of multiple measurements improve the accuracy of data such as phoria; and separate modeling for specific individuals and fusing historical parameters to generate transitional schemes ensures the continuity and adaptability of training.

[0047] For example, a 7-year-old child with moderate esophageal deviation (esophageal deviation of 9Δ; clinical data shows that esophageal deviation exceeding 9Δ may cause symptoms such as eye strain, reading difficulties, or intermittent diplopia). During validation, the eye position was first precisely calibrated using white light emitted by the multispectral adjustment unit. Then, occlusion and de-occlusion operations were performed, and the average value was taken for three consecutive measurements to accurately obtain the esophageal deviation, after which a separate model was created for the child. When collecting data at different viewing distances, the contrast control unit fixed the image contrast at 50% to avoid interference from changes in external light, resulting in a more accurate AC / A constant. After validation, the dynamic training module integrated the child's historical training parameters to generate a transition scheme, ensuring that the child would not experience discomfort due to excessive changes in the initial application of new parameters. This ensured both training effectiveness and improved the child's cooperation, something that traditional systems relying solely on single data acquisition and general models struggle to achieve.

[0048] The dynamic training module includes a dynamic adjustment unit and a multispectral adjustment unit. When the dynamic adjustment unit detects fatigue in the ciliary muscle, it reduces the target switching frequency to 8-12 times / minute while increasing the proportion of 500-550nm green light to 60% to promote relaxation. When an active state is detected, it increases the target switching frequency to 15-20 times / minute while increasing the proportion of 620-660nm red light to 50% to enhance training. The multispectral adjustment unit emits specific light waves in the 430-680nm range. Green light adjusts the pupil diameter through alternating brightness, simultaneously training the synergy between the ciliary muscle and pupil. Red light provides photonutrient supplementation to stimulate accelerated choroidal blood circulation. This unit enhances training through photowave characteristics and strengthens the visual system's coordination through pupillary adjustment.

[0049] Specifically, the dynamic adjustment unit incorporates a built-in state recognition method. After receiving the real-time ciliary muscle adjustment data from the data fusion unit, it completes state determination within one second. When switching target frequencies, a step-wise transition is used (e.g., increasing or decreasing by 2 times / minute, with a 5-second interval) to avoid sudden frequency changes that could cause eye discomfort. The green light-dark alternation cycle of the multispectral adjustment unit is linked to the target switching frequency (e.g., 8 times / minute corresponding to a 4-second light-dark cycle). During the alternation, the brightness gradually changes from 300 lux to 800 lux and then falls back, simultaneously triggering the image acquisition unit to record changes in pupil diameter as an evaluation basis for the collaborative training effect. The red light output intensity is dynamically adjusted based on choroidal blood circulation simulation data (indirectly calculated based on pressure deformation data), with an intensity increase of 10%-20% when blood circulation is slower. The dynamic adjustment unit communicates with the data fusion unit in real time, feeding back the light wave action data to the mapping model to optimize the accuracy of ciliary muscle state determination, achieving a closed-loop enhancement that combines training, evaluation, and model optimization.

[0050] For example, in training an 8-year-old child with amblyopia, the dynamic adjustment unit detected an active state, and the target frequency was increased stepwise from 10 times / minute to 15 times / minute, with the proportion of red light simultaneously increasing to 50%. The multispectral adjustment unit alternated green light brightness at a frequency of 15 times / minute, and the image acquisition unit recorded the synchronicity of pupil diameter changes with green light. The data fusion unit then fine-tuned the mapping model accordingly. When the child continuously fixated on the training target, causing choroidal blood circulation to slow down, the red light intensity automatically increased by 15%, which both enhanced the training effect and accelerated eye metabolism, shortening the time to achieve the target in a single training session compared to the traditional fixed light wave scheme.

[0051] The dynamic training module includes a contrast control unit, a mode switching unit, and a scheme generation unit. The contrast control unit dynamically reduces the contrast of the training image in the training parameters to enhance the ciliary muscle's accommodative stability in complex visual environments. The mode switching unit supports automatic switching between object-based training and brain-based training. Object-based training drives ocular physiological responses through the content viewed by the child (set by the administrator), the clarity of the viewed content, and the dynamic changes in the viewed content; for example, using a flip-and-play device with letter cards to train the ciliary muscle's accommodative flexibility, using a convergence / divergence ball to guide the convergence / divergence of the visual axes of both eyes to improve convergence function, and using a moving light spot to train the stability of eye movements. Brain-based training involves having the child gaze at a preset neutral gray screen (set by the administrator) and, under the guidance of a preset voice (set by the administrator), perform mental imagery tasks such as "imagine looking far and near," "imagine tracking a light spot," and "imagine flipping and playing letters." During this process, the dynamic training module collects the child's brain activity in real time through EEG signals and simultaneously implements dual interventions of red light ion pump activation and alpha band neurofeedback. When the data fusion unit determines that the ciliary muscle is fatigued for three consecutive times, it switches to brain-based training for 1-2 minutes. The program generation unit combines the child's age, amblyopia degree, and historical data from the collaborative acquisition module to output the daily training duration and light wave combination mode, achieving precise control of training and recovery effects. It also links with the dynamic adjustment unit so that the generated program can be updated in real time according to the ciliary muscle state. The amblyopia degree is obtained by matching ocular feature parameters with stored amblyopia features.

[0052] More specifically, the contrast control unit reduces the image contrast in stages at 5% increments (e.g., initial 50%, minimum 20%), with each stage lasting 2 minutes. Simultaneously, it receives extraocular muscle pressure data from the pressure acquisition unit. If the pressure increase exceeds 10% (when the adjustment pressure increase exceeds 10%, it may cause excessive tension in the ciliary muscle; maintaining the current contrast at this point can avoid visual function compensation disorder caused by parameter mutations), the contrast reduction is paused and the current level is maintained. The mode switching unit activates a green light warning 10 seconds before switching (e.g., temporarily increasing the proportion of green light in the 500-550nm range to 70%). During the brain training phase, the target switching function of the dynamic adjustment unit is disabled, retaining only the basic red light illumination of the multispectral adjustment unit (e.g., 30% proportion of 620-660nm). The scheme generation unit incorporates an amblyopia feature database. It determines the degree of amblyopia by matching the eye feature parameters from the image acquisition unit (e.g., ≥85% is considered the corresponding level). Daily training duration is age-appropriate (e.g., 20 minutes / day for 6-8 years old, 30 minutes / day for 9-12 years old) and is linked to the real-time status of the dynamic adjustment unit. During active periods, each training session is extended by 1-2 minutes; during fatigue periods, a 30-second green light relaxation session is inserted. The scheme generation unit shares historical data with the data fusion unit, enabling scheme updates to simultaneously optimize the error threshold of the mapping model and improve the collaborative accuracy of status determination and training schemes.

[0053] In one specific implementation of this solution, a 10-year-old child with moderate amblyopia had an eye feature matching rate of 92%. During training, the contrast control unit gradually reduced the contrast from 50% to 30%. During this period, the extraocular muscle pressure increased by 15%, and the contrast reduction was immediately paused. After the data fusion unit determined the fatigue state three times in a row, the mode switching unit first issued a high proportion of green light warning, and then switched to brain training for 1.5 minutes, during which only basic red light irradiation was retained. The scheme generation unit combined the child's 9-year-old age group and historical data to initially set 30 minutes of training per day. Due to the high number of active states, the unit automatically extended the single training session to 3 minutes and inserted a green light relaxation session between two active states. This not only enhanced the child's ability to adjust to complex environments, but also avoided excessive fatigue through mode switching. The training efficiency was significantly improved compared to the fixed scheme.

[0054] The system also includes a training device comprising a housing, a display component, and a fastening component. The display component is located inside the housing and fixedly connected to it, while the fastening component is fixedly connected to the outside of the housing to secure the housing to the area around the eye. The housing has an orbital aperture adapted to the eye socket, and a pressure detection area corresponding to the attachment point of the extraocular muscles on the outer side of the skin around the orbital aperture is located on the outer side of the housing. Several pressure sensors of the pressure acquisition unit are fixedly connected to the pressure detection area. The image acquisition unit is a binocular miniature camera integrated above the orbital aperture inside the training device, equipped with an infrared fill light. The display component is also fixedly connected inside the housing, and is used to receive training parameters provided by the dynamic training module and display the training images.

[0055] A fixed dry electrode EEG sensor is embedded in the area where the shell fits against the forehead to collect EEG signals from the prefrontal cortex and motor cortex in real time. The data fusion unit filters and amplifies the EEG signals and identifies them according to frequency, such as the alpha band (e.g., capturing 8-13Hz as the alpha band). When the system detects that the child has entered a stable visual imagery state (e.g., alpha band power > 5μV² / Hz, where μV² / Hz is a unit of power spectral density, representing the power contained in the EEG signal within a 1Hz frequency bandwidth), an infrared supplemental light emits a preset fixed pulse (set by the administrator) to help relax the ciliary muscle and reduce tension to zero through photobiological regulation. At the same time, the data fusion unit analyzes the alpha band power in real time and converts it into dynamic feedback for the display components.

[0056] Specifically, the training device uses liquid lenses with a diameter of 5-8mm (response time <50ms) or miniature electric focusing lenses. By controlling the curvature of the lens with voltage, it can achieve continuous adjustment of refractive power from 1D to 5D (accuracy up to ±0.1D) to meet the requirements of visual distance simulation from 20cm (5D) to 100cm (1D).

[0057] Specifically, the training device's outer shell is made of food-grade silicone, with a thickness of 2mm. Food-grade silicone is soft and skin-friendly, suitable for children's delicate skin. The 2mm thickness ensures structural support while also providing flexibility, conforming to the different eye contours of children, reducing the feeling of foreign objects when wearing it, improving comfort, and reducing children's resistance.

[0058] The edges of the eye socket holes are rounded by 0.5mm to avoid sharp edges rubbing against the skin around the eyes and causing damage; this further enhances the safety of wearing the device, making it especially suitable for active children.

[0059] The pressure detection area consists of a 5mm wide annular protrusion around the hole, embedding eight flexible pressure sensors (typically 3mm in diameter, but custom diameters can be made for children). These sensors are symmetrically distributed: two for the medial rectus muscle, two for the lateral rectus muscle, two for the superior rectus muscle, and two for the inferior rectus muscle. The annular protrusion enhances the fit between the sensor and the skin, and the symmetrical distribution comprehensively captures pressure deformation of each extraocular muscle, resulting in more comprehensive and accurate pressure data acquisition. This provides a reliable basis for data fusion unit analysis and reduces errors compared to single-point acquisition. The breathable non-woven fabric layer on the skin-contacting side of the pressure sensor is made of bamboo fiber, and each piece is replaceable. Bamboo fiber is breathable and antibacterial; the replaceable design facilitates cleaning and hygiene, reduces skin stuffiness and allergies caused by prolonged wear, and is easy to maintain.

[0060] The binocular miniature camera boasts a 12-megapixel resolution, with the lens axis positioned at a 15° angle to the anterior segment of the eye. An 850nm infrared LED offers three brightness levels (other levels can be customized for children), automatically adjusted by the image acquisition unit based on ambient light intensity (high brightness is activated at ≤300 lux). The 12-megapixel resolution and 15° angle clearly capture details of the anterior segment of the eye, while the three brightness levels adapt to different ambient light conditions, ensuring stable high-quality image acquisition in both bright and dark environments. This effectively extracts eye feature parameters, improving image quality.

[0061] The display unit can be a 2.5-inch OLED screen with a 60Hz refresh rate, or it can be replaced with another screen with a higher refresh rate and better display effect according to the parents' specific requirements. The display unit is electrically connected to the dynamic training module. The OLED screen has vivid colors, fast response, and low latency to ensure that the training screen and parameter adjustments are synchronized, allowing children to view the training screen more clearly and smoothly, improving their concentration and reducing training deviations caused by latency.

[0062] The fastening component can be an elastic headband made of nylon, or other flexible materials can be replaced according to parents' needs. The length of the elastic headband is adjustable from 40-60cm. The pressure feedback module at the connection with the shell has a sampling frequency of 10Hz. When the pressure exceeds 2kPa, in addition to reducing the display brightness by 10%, it also triggers the multispectral adjustment unit to increase the proportion of green light in the 500-550nm range by 5%. (kPa is an abbreviation for "kilopascal," a unit of pressure. 1 mmHg = the pressure generated by a 1 mm high mercury column under standard gravity. 2 kPa is approximately 15 mmHg, corresponding to the peak pressure of extraocular muscle contraction. Exceeding this indicates that the headband is too tight or the child is tense and frowning. It immediately reduces the brightness by 10% and supplements with 5% green light, which can relieve local pressure and relax the adjustment tension, avoiding the vicious cycle of "the tighter it gets, the more tired it becomes"). The adjustable headband adapts to different head circumferences, and the pressure feedback prevents it from being worn too tightly. While ensuring the stability of the device, it alleviates eye discomfort through dual adjustment of reducing brightness and increasing the proportion of green light, allowing children to complete training in a comfortable state. Combined with the dynamic training module, it improves training efficiency.

[0063] In one specific implementation of this solution, when a 6-year-old child with anisometropia wears the training device, the silicone shell fits snugly against the eye socket, and the eight sensors in the pressure detection area precisely correspond to the attachment points of the extraocular muscles. When the ambient light is dim, the infrared supplementary light automatically adjusts to a high level, and the binocular camera clearly captures images of the anterior segment of the eyeball. When the headband is adjusted to a suitable length, the pressure feedback module monitors in real time. When the child frowns due to tension, causing the wearing pressure to reach 2.2 kPa, the display brightness decreases by 10% and the proportion of green light increases, which not only relieves eye discomfort but also ensures clear training images. Combined with the parameter adjustment of the dynamic training module, the effective duration of a single training session is increased.

[0064] Specifically, the pressure acquisition unit integrates a contact sensing component for synchronously acquiring intraocular pressure data; the image acquisition unit analyzes corneal reflection images to determine the tear film breakup time, forming a collaborative verification method for intraocular pressure and tear film status. This collaborative verification method uses intraocular pressure fluctuation thresholds and tear film breakup time thresholds as auxiliary verification conditions, correlated with pressure deformation data acquired by the pressure sensor and ocular feature parameters obtained by the image acquisition unit. When intraocular pressure exceeds a preset fluctuation range, the data fusion unit corrects the weights of the correlation parameters in the mapping model; these correlation parameters are the weighting coefficients in the mapping model used to correlate lens curvature and extraocular muscle pressure. When the tear film breakup time is shorter than a preset breakup threshold, the multispectral adjustment unit adjusts the light wave ratio to improve the ocular surface condition.

[0065] Specifically, the contact sensing component is a miniature intraocular pressure sensor with an accuracy of ±0.5 mmHg. It is integrated next to the sensor corresponding to the superior rectus muscle in the pressure detection area, with a sampling frequency of 5 Hz, and is triggered synchronously with the pressure sensor for acquisition. When analyzing the tear film breakup time, the image acquisition unit captures 3 frames of corneal reflection images every 0.5 seconds, and calculates the breakup time using a grayscale gradient algorithm with an error ≤0.3 seconds.

[0066] The initial values ​​of the associated parameters were set as follows: extraocular muscle pressure influence weight 0.5, lens curvature response coefficient 1.2Δ / D / kPa, and intraocular pressure baseline 15 mmHg. When the intraocular pressure exceeded the range of 12-18 mmHg, the intraocular pressure correction factor was adjusted by ±0.05 for every 1 mmHg deviation, simultaneously correcting the output value of the mapping model. When the tear film breakup time was <5 seconds, in addition to adjusting the pressure influence weight and curvature response coefficient, the multispectral adjustment unit increased the proportion of green light in the 500-550nm range from 30% to 50%, and after 2 minutes, the tear film state was re-detected. At this time, the dynamic training module paused the gradient descent contrast operation of the contrast control unit.

[0067] The above-mentioned content is linked with the scheme generation unit of the dynamic training module. When the associated parameters are corrected three times in a row, the scheme generation unit automatically shortens the single training segment by 1 minute and extends the brain training interval, effectively reducing the rate of visual load accumulation during children's training.

[0068] In the above description, the synchronous triggering and acquisition of data by the miniature intraocular pressure sensor and the pressure sensor is not a simple parallel data acquisition. Instead, it uses a 5Hz sampling frequency as a time anchor to instantaneously verify the mechanical relationship between intraocular pressure fluctuations and extraocular muscle pressure deformation, establishing a connection between the intraocular and extraocular systems. This millisecond-level time-aligned mechanical coupling analysis breaks through the traditional understanding of treating intraocular pressure and extraocular muscle pressure as independent variables. For example, when a child rapidly moves their eyeballs, this approach can distinguish between pressure changes caused by extraocular muscle contraction and passive fluctuations in intraocular pressure, avoiding misjudgment of accommodation by the mapping model.

[0069] Secondly, the synergy between tear film breakup time-triggered multispectral modulation and contrast modulation pause transcends the conventional logic of simply improving the ocular surface condition. When the proportion of green light increases to 50%, it promotes tear secretion through photochemical action and reduces the ciliary muscle accommodation load due to the wavelength characteristics of green light. Meanwhile, pausing gradient contrast reduction avoids a sudden drop in visual clarity when the tear film is unstable, forming a dual protection that combines ocular surface repair and accommodation load buffering.

[0070] Finally, by triggering dynamic adjustments to the training scheme through continuous correlation correction of parameters, predictive intervention for data anomalies and scheme adaptation is achieved. The scheme generation unit does not only adjust based on real-time status, but also predicts the potential risk of load accumulation by adjusting the parameter correction frequency, proactively shortening training segments and extending brain training intervals. This forward-looking regulation based on data trends breaks the traditional model of intervention only after fatigue occurs, improving the accuracy of visual load control.

[0071] Specifically, flexible Ag / AgCl electrodes are fixedly connected to the outer shell of the training device at the corresponding positions of the tragus and the corneal limbus at the edge of the orbital aperture. The electrode area is ≤10mm² and the contact resistance is ≤5kΩ. The sampling frequency of the flexible Ag / AgCl electrode is 250Hz (250Hz is only used for "slow change" ciliary muscle electrocouple indices and not for high-frequency ERP), which is used to collect changes in corneal surface potential.

[0072] When the ciliary muscle contracts, the increased activity of the epithelial ion pump causes the corneal surface potential to drift towards a negative value by 0.5-1.2 mV.

[0073] The data fusion unit correlates and analyzes corneal surface potential change data with intraocular pressure data, tear film breakup time, and pressure deformation data. When the linear deviation between the potential drift amplitude and the adjustment amount exceeds the preset potential adjustment deviation, the weight of the correlation parameters of the mapping model is corrected simultaneously. The correlation parameters are the weight coefficients in the mapping model used to correlate lens curvature and extraocular muscle pressure. At the same time, the dynamic training module controls the multispectral adjustment unit to adjust the proportion of red light to enhance ion pump activity.

[0074] More specifically, the flexible Ag / AgCl electrodes utilize a 0.1mm ultrathin substrate. The electrode at the tragus is circular with a diameter of 5mm, while the one at the limbus is arc-shaped with a diameter of 10mm, respectively matching the contact area requirements of the two anatomical structures to ensure the stability of potential signal acquisition. The flexible Ag / AgCl electrodes at the tragus and limbus are fixed with medical double-sided tape, ensuring a contact resistance with the skin of ≤5kΩ, guaranteeing low-loss transmission of weak potential signals.

[0075] With a preset potential adjustment deviation of 5%, the correlation parameter correction rule is as follows: for every 1% increase in potential drift deviation, the weight of extraocular muscle pressure influence is ±0.02, and the lens curvature response coefficient is ±0.1Δ / D / kPa. This means that for every 1 kPa change in extraocular muscle pressure, the change in the "accommodative set to accommodation ratio (AC / A constant)" associated with lens curvature adjustment is ±0.1 prism power / diopter. Assuming that the extraocular muscle pressure range for children aged 6–12 years is 0.5–2 kPa, and the AC / A constant range for children is 3–5 Δ / D, the pressure-induced AC / A fluctuation is usually ≤0.3 Δ / D. Combining the pressure fluctuation range and the upper limit of AC / A fluctuation, the "AC / A change caused by a unit pressure change" is estimated to be approximately ±0.1 Δ / D / kPa. The lens curvature response coefficient is specifically determined based on the child's actual extraocular muscle pressure range and AC / A constant range. The initial proportion of red light (620-660nm) in the multispectral modulation unit is set at 20%. When the deviation exceeds the limit, the red light proportion increases by 2% for every 1% increase in deviation (up to a maximum of 40%), and then decreases by 5% after 1 minute. The preset 5% potential modulation deviation and related parameter correction rules are based on the fluctuation range of the linear relationship between ciliary muscle modulation and potential drift in clinical data. By quantifying the correspondence between deviation and parameter adjustment, the mapping model is accurately calibrated. The initial 20% red light proportion in the multispectral modulation unit, with a stepwise increase when the deviation exceeds the limit, avoids excessive red light stimulation of the retina while dynamically enhancing intervention according to the degree of insufficient ion pump activity.

[0076] Within the 250Hz interrupt service routine, the training device calculates the first-order difference for each sample: if the absolute value of the slope is <0.5µV or the instantaneous slope exceeds ±40µV within 1 second, it is determined that the electrode has fallen off or drifted abnormally, and an invalid flag is immediately set. At the same time, a 0.5mm flat motor and a red LED are driven to emit a 100ms vibration-light prompt. This dual-condition verification uses integer operations throughout, with an additional CPU load of <2%, and there is no risk of empty sampling at high sampling rates.

[0077] The flexible Ag / AgCl electrodes are linked to the protocol generation unit. When the red light percentage is ≥30%, the protocol generation unit automatically adds a 5-minute brain training session to the daily training plan, utilizing the synergistic effect of red light and brain training to enhance ion pump repair. Simultaneously, the electrode data acquisition works in conjunction with the mode switching unit of the dynamic training module. If the potential drift remains unchanged for 30 consecutive seconds, it is determined to be a poor electrode contact, triggering a vibration alert (0.5mm amplitude) to prevent data loss. This linkage with the protocol generation unit and vibration alert design transforms the electrode data acquisition into an active intervention strategy: adding brain training when the red light percentage is ≥30%, utilizing the ion pump activation effect of red light and the neuromodulatory effect of brain training to form a synergy; and triggering a vibration alert after 30 seconds of no potential drift, quickly identifying data acquisition anomalies such as electrode detachment and avoiding ineffective training.

[0078] The ultra-thin substrate and irregular shape design improve electrode wearing comfort and extend children's tolerance time; contact resistance control ensures improved potential signal-to-noise ratio, and the detection accuracy of weak drift at the 0.5mV level reaches a high level.

[0079] The parameter correction rule reduces the calculation error of the adjustment amount to ±0.15Δ / D when the potential deviation exceeds the limit, which improves the accuracy compared with the fixed parameter model; the red light step adjustment accelerates the recovery speed of ion pump activity and shortens the tear film breakup time.

[0080] The linkage with the scheme generation unit improves the training efficiency for children with severe regulatory dysfunction, while vibration reminders reduce the proportion of ineffective training caused by missing data, realizing a closed-loop optimization of the entire chain of signal acquisition, model calibration and training intervention.

[0081] Specifically, the data fusion unit performs four-dimensional correlation analysis on corneal surface potential change data, intraocular pressure data, tear film breakup time and pressure deformation data to establish a dynamic coupling model combining potential, intraocular pressure, tear film and pressure.

[0082] When the linear deviation between the potential drift amplitude and the adjustment amount exceeds the preset potential adjustment deviation, and the intraocular pressure exceeds the fluctuation range or the tear film rupture time is shorter than the preset rupture threshold, the weights of the associated parameters of the mapping model are simultaneously corrected, the proportion of red and green light in the multispectral adjustment unit is adjusted, and the operation of the contrast control unit is paused.

[0083] The training device continuously captures micro-deformations of the outer wall of the eye using a 5Hz flexible pressure sensor, and simultaneously outputs a proximal pressure of 5–25 mmHg from a micro-strain bridge chip packaged in the same area. After Goldmann calibration, a linear mapping of "outer wall stress + corneal apex micro-displacement" is established, with individual three-point anchoring (8 / 15 / 22 mmHg) and zero drift correction every 30 seconds using the tear film stabilization period. The IOP_est is calculated in real time and used as the result of intraocular pressure acquisition. It is refreshed every 1 second with an error ≤ ±0.5 mmHg. When the estimated value is >18 mmHg or <1 mmHg, the training intensity is automatically reduced to ensure training safety under high intraocular pressure.

[0084] The device captures blinking behavior in real time using an infrared-illuminated anterior segment camera (30fps). Starting from the last blink frame, it identifies the first connected dry spot with ≥3 pixels on the cornea frame by frame using grayscale gradient. The tear film breakup time (TBUT) is obtained by dividing the frame number difference by 30. The device is automatically updated every 30 seconds. ≥10 seconds is considered normal, and ≤5 seconds indicates dry eye. This is used for intraocular pressure zero drift correction and linkage with training parameters.

[0085] More specifically, the dynamic coupling model employs a multilayer perceptron algorithm. The input layer contains four dimensions of data: potential drift amplitude, intraocular pressure, tear film breakup time, and peak pressure deformation. The hidden layer has two neurons, and the output layer is a data reliability score (0-100 points). The preset potential adjustment deviation is 5%, the intraocular pressure fluctuation range is 12-18 mmHg, and the tear film breakup threshold is 5 seconds. The multilayer perceptron algorithm can efficiently handle multi-dimensional nonlinear data associations. By balancing computational complexity and fitting accuracy through two hidden layer neurons, it is suitable for real-time computation in the embedded mode of this solution. The output data reliability score (0-100 points) quantifies data quality, providing a clear basis for subsequent intervention. The dynamic coupling model's computational latency is controlled within 50ms, and the data association analysis achieves high accuracy. Compared with traditional linear models, its accuracy is significantly improved, and the reliability score allows this solution to quickly identify invalid data, avoiding interventions based on erroneous data.

[0086] The preset parameters—potential adjustment deviation of 5%, intraocular pressure of 12-18 mmHg, and tear film threshold of 5 seconds—are based on statistical values ​​of children's ocular physiological data. This not only covers almost all normal fluctuation ranges but also allows for timely detection of abnormal states. The effect is to reduce most false triggering interventions while maintaining a high level of sensitivity in identifying genuine abnormalities, ensuring that this solution can respond promptly even to minor abnormalities in children's eyes.

[0087] When the triple adjustment is triggered, the correction rules are as follows: the weight of extraocular muscle pressure in the associated parameter weights is adjusted by ±0.03 for every 1% change in the deviation between the actual value of extraocular muscle pressure and the benchmark value, and the lens curvature response coefficient is adjusted by ±0.15Δ / D / kPa; the proportion of red light (620-660nm) in the multispectral adjustment unit is increased from 20% to 22%-24%, and the proportion of green light (500-550nm) is increased from 30% to 50%, and after 2 minutes, the green light drops back to 40%; during the pause of the contrast control unit, the contrast of the training screen is maintained at the current value.

[0088] In the aforementioned triple regulation rules, the adjustment range of the associated parameter weights (i.e., for every 1% change in the corresponding deviation, the weight is adjusted by ±0.03) is based on fitting of previous clinical data to ensure a balance between correction speed and stability. The stepwise adjustment of the proportion of red and green light avoids visual stimulation from sudden changes in light waves while allowing for precise intervention according to the severity of the pathology (red light enhances the ion pump, and green light improves the tear film). After the associated parameter correction, the calculation error of the regulation amount in the mapping model is reduced to ±0.1Δ / D, which improves the ion pump activity of light wave regulation, prolongs the tear film breakup time, and reduces children's discomfort to changes in light waves.

[0089] The above-mentioned method is linked to the mode switching unit. If the dynamic coupling model score is less than 60 points and lasts for 10 seconds, the mode switching unit will force a switch to brain-based training. At the same time, the scheme generation unit will record this abnormal data for use in optimizing the training scheme for the next day. In addition, this scheme has a built-in data self-calibration module that automatically collects three sets of basic data (including resting intraocular pressure, tear film breakup time, and corneal surface potential baseline) before the first use each day as the benchmark values ​​for the model parameters for that day, reducing the impact of individual daily physiological fluctuations.

[0090] The system is linked to the mode switching unit, setting a score below 60 points and triggering a switch after 10 seconds to prevent accidental operation due to instantaneous data fluctuations. The 10-second delay balances response speed and judgment accuracy. The plan generation unit records abnormal data, which can be used to optimize the next day's plan through machine learning, enabling personalized training iterations. This reduces the rate of accidental mode switching, improves the adaptability of the next day's training plan, and makes the training more suitable for the child's eye condition on that day.

[0091] The data self-calibration module collects three sets of basic data daily. Since children's ocular physiological indicators fluctuate throughout the day (e.g., higher intraocular pressure in the morning), using the daily baseline value instead of a fixed standard reduces the impact of individual physiological rhythms on model accuracy, making parameter corrections more closely reflect real-time conditions. This reduces model errors caused by daily physiological fluctuations, improves the accuracy of parameter corrections, and further ensures the precision of training.

[0092] Specifically, the dynamic coupling model introduces a dynamic baseline calibration coefficient K. K is calculated based on the initial drift value of corneal surface potential, the baseline value of intraocular pressure, and the tear film breakup reference time according to the formula K=0.3×potential baseline+0.2×intraocular pressure baseline+0.5×tear film baseline, and is automatically updated every 30 seconds. The K value is embedded in the reliability score calculation of the dynamic coupling model. When the K fluctuation is ≤5%, the score weight is allocated according to the original ratio. When the K fluctuation exceeds 5%, the score weight of intraocular pressure and tear film data is increased by 20%. At the same time, the K value is linked with the multispectral adjustment unit. When the K fluctuation exceeds 5%, except for the adjustment of red light and green light, the supplementary light intensity increases linearly with the K value deviation, and triggers the data fusion unit to shrink the error threshold to ±0.2Δ / D.

[0093] More specifically, the initial value of the dynamic baseline calibration coefficient K is calculated using 10 sets of basic data collected continuously for the first 30 seconds of training. The electrical potential baseline is taken as the average value within the range of 0.5-1.2mV, the intraocular pressure baseline is taken as the average value within the range of 12-18mmHg, and the tear film baseline is taken as the breakup time of the first test. When the fluctuation of K is ≤5%, the weights of the four dimensions of data in the confidence score are: electrical potential 25%, intraocular pressure 20%, tear film 30%, and pressure 25%; when the fluctuation exceeds 5%, the weights of intraocular pressure and tear film are each increased by 20% (i.e., intraocular pressure 24%, tear film 36%). The supplementary light intensity adjustment step size is 1% supplementary light power for every 0.1 unit deviation of K value. After the error threshold is shrunk to ±0.2Δ / D, the image acquisition unit is triggered to take two additional frames of corneal reflection images every 0.3 seconds.

[0094] The weighting of K values ​​(i.e., tear film 0.5 > intraocular pressure 0.2 > electrical potential 0.3) is based on the ranking of the impact of ocular surface condition on data acquisition. Tear film stability has the most significant impact on image quality, so it is given the highest weight. A 30-second update cycle balances real-time performance and data smoothness, avoiding frequent adjustments caused by instantaneous fluctuations. In the linkage between the dynamic baseline calibration coefficient K and dual verification (i.e., comparison verification between the first and second data), secondary verification of data quality is achieved through dynamic shrinkage of the error threshold, forming a closed loop of calibration, verification, and re-acquisition.

[0095] Dynamic adjustment of the K value reduces invalid data caused by environmental interference during home training. In conjunction with the mode switching unit, it automatically switches to low-intensity mode when the K value exceeds the limit three times in a row, improving training safety. The additional image frames reduce the error in tear film breakup time calculation. Together with the supplementary lighting enhancement of multispectral adjustment, it improves image clarity in low-light environments, achieving a triple optimization that combines environmental adaptation, data accuracy, and training safety.

[0096] This specific embodiment also includes a ciliary muscle-based visual function training method for children using a ciliary muscle-based visual function training system.

[0097] The above descriptions are merely embodiments of the present invention, and common knowledge regarding specific structures and characteristics is not elaborated upon here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the structure of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A visual function training system for children based on ciliary muscle accommodation, characterized in that, include: The collaborative acquisition module includes an image acquisition unit, a pressure acquisition unit, and a data fusion unit; The image acquisition unit acquires images of the anterior segment of the eye in real time and identifies ocular feature parameters in the images of the anterior segment of the eye. The pressure acquisition unit includes a pressure sensor located on the skin surface at the attachment point of the extraocular muscles on the outer side of the eye socket, which collects pressure deformation data of the extraocular muscles caused by eyeball convergence movement in real time. The data fusion unit acquires ocular feature parameters and pressure deformation data collected at the same time, and performs correlation analysis. It establishes a mapping model based on the changes in pressure deformation data and lens curvature. The lens refractive power change data is indirectly calculated based on the ocular feature parameters using a thin lens refractive power conversion algorithm as the first data. The pressure deformation data is input into the mapping model to calculate the real-time state of the ciliary muscle, and the lens refractive power change data in the ocular feature parameters is calculated based on the real-time state as the second data. The data fusion unit compares the first data with the second data. If the comparison result falls within the preset error threshold, the real-time state is processed into the real-time adjustment amount of the ciliary muscle according to the acquisition time. If the comparison result exceeds the error threshold, the eye feature parameters and pressure deformation data are obtained again from the image acquisition unit and the pressure acquisition unit, and the first data and the second data are calculated and compared. The dynamic training module acquires individual data of children and historical data from the collaborative acquisition module, formulates training parameters according to the preset training cycle, and dynamically adjusts the training parameters according to the real-time adjustment of the ciliary muscle. The real-time state includes fatigue state and active state, and the training parameters include training duration and light wave combination. The training duration, target frequency, proportion of different types of light in the light wave combination and contrast are dynamically adjusted according to the real-time state. The training mode is adjusted based on the parameters of continuous real-time state control.

2. The children's visual function training system based on ciliary muscle accommodation according to claim 1, characterized in that: The data fusion unit is also used to verify the accuracy of the AC / A constant based on synchronously acquired pressure deformation data and ocular characteristic parameters. During accuracy verification, when the individual error of AC / A is ≤ ±0.3Δ / D and the error of the mapping model after calibration by UBM and open-field refractometer is ≤ ±0.25D, the objective accommodative response is first recorded using an infrared video refractometer or open-field automatic refractometer as the gold standard for accommodative amount. Simultaneously, the instantaneous convergence is measured using the prism cover method or synoptophore as the gold standard for convergence, ensuring that the AC / A constant obtained from the gold standard for accommodative amount and the gold standard for convergence are consistent. The difference between the AC / A ratio and the pressure deformation data and ocular characteristic parameters acquired simultaneously is ≤ ±0.3Δ / D. Accommodation stimuli are then applied at different viewing distances and repeated multiple times to plot stimulus-response curves. If the linear fit of the stimulus-response curve is R² < 0.9 or the intercept is > 0.5Δ, piecewise fitting or quadratic curve correction is performed. The amount of phoria at near and far distances is measured by occlusion and deocclusion methods. Individual models are built for those with phoria > 8Δ. For those with unequal refractive power with a difference of > 1D between the two eyes, the AC / A ratio of the left and right eyes is calibrated and the average value is taken. The measurement is repeated every other day to ensure that the difference between the two AC / A ratios is < ±0.2Δ / D.

3. The children's visual function training system based on ciliary muscle accommodation according to claim 2, characterized in that: The dynamic training module includes a dynamic adjustment unit and a multispectral adjustment unit; When the dynamic adjustment unit detects that the ciliary muscle is in a fatigued state in real time, it reduces the target switching frequency to 8-12 times / minute and increases the proportion of 500-550nm green light to 60% to promote relaxation; when an active state is detected, it increases the target switching frequency to 15-20 times / minute and increases the proportion of 620-660nm red light to 50% to strengthen training. The multispectral adjustment unit is used to emit specific light waves of 430-680nm. Green light adjusts the pupil diameter by alternating between brightness and darkness, and simultaneously trains the coordination between the ciliary muscle and the pupil. Red light is used for photonutrient supplementation to stimulate the acceleration of blood circulation in the choroid. This unit achieves training and strengthening through the characteristics of light waves, and enhances the linkage of the visual system by adjusting the pupil.

4. The children's visual function training system based on ciliary muscle accommodation according to claim 3, characterized in that: The dynamic training module also includes a contrast adjustment unit, a mode switching unit, and a scheme generation unit; The contrast control unit is used to dynamically reduce the contrast of the training image in the training parameters in order to enhance the regulatory stability of the ciliary muscle in complex visual environments. The mode switching unit supports automatic switching between physical training and brain training. When the data fusion unit determines that the ciliary muscle is in a state of fatigue for 3 consecutive times, it switches to brain training for 1-2 minutes. The scheme generation unit combines the child's age, degree of amblyopia, and historical data from the collaborative acquisition module to output the daily training duration and light wave combination mode, achieving precise control of training and recovery effects. It also works in conjunction with the dynamic adjustment unit to allow the generated scheme to be updated in real time according to the ciliary muscle state. The degree of amblyopia is obtained by matching eye feature parameters with stored amblyopia features.

5. The children's visual function training system based on ciliary muscle accommodation according to claim 4, characterized in that: It also includes a training device, which includes a housing, a display component, and a fastening component. The display component is located inside the housing and is fixedly connected to the housing, and the fastening component is fixedly connected to the outside of the housing, for fixing the housing of the training device around the eyes. The outer shell has an orbital opening that fits the eye socket. Around the orbital opening, the outer shell has a pressure detection area on the skin surface corresponding to the attachment point of the extraocular muscles on the lateral side of the eye socket. Several pressure sensors of the pressure acquisition unit are fixedly connected to the pressure detection area. The image acquisition unit is a binocular miniature camera integrated above the orbital opening inside the training device. The binocular miniature camera is equipped with an infrared fill light. A display component is also fixedly connected inside the outer shell. The display component is used to receive training parameters provided by the dynamic training module and display the training image.

6. The children's visual function training system based on ciliary muscle accommodation according to claim 5, characterized in that: The pressure acquisition unit integrates a contact sensing component for synchronously acquiring intraocular pressure data; the image acquisition unit analyzes corneal reflection images to determine tear film breakup time, forming a collaborative verification method for intraocular pressure and tear film status. The collaborative verification method uses the intraocular pressure fluctuation threshold and tear film breakup time threshold as auxiliary verification conditions, and performs correlation analysis with the pressure deformation data collected by the pressure sensor and the ocular feature parameters acquired by the image acquisition unit. When the intraocular pressure exceeds the preset fluctuation range, the data fusion unit corrects the weights of the correlation parameters in the mapping model. The correlation parameters are the weight coefficients in the mapping model used to correlate lens curvature and extraocular muscle pressure. When the tear film breakup time is shorter than the preset breakup threshold, the multispectral adjustment unit adjusts the light wave ratio to improve the ocular surface condition.

7. The children's visual function training system based on ciliary muscle accommodation according to claim 6, characterized in that: A flexible Ag / AgCl electrode is fixedly connected to the outer shell of the training device at the corresponding position of the tragus and the corresponding position of the corneal limbus at the edge of the orbital opening. The flexible Ag / AgCl electrode has a sampling frequency of 250Hz and is used to collect changes in the corneal surface potential. When the ciliary muscle contracts, the increased activity of the epithelial ion pump causes the corneal surface potential to shift to a negative value by 0.5-1.2 mV; The data fusion unit correlates and analyzes corneal surface potential change data with intraocular pressure data, tear film breakup time, and pressure deformation data. When the linear deviation between the potential drift amplitude and the adjustment amount exceeds the preset potential adjustment deviation, the weight of the correlation parameters of the mapping model is corrected simultaneously. The correlation parameters are the weight coefficients in the mapping model used to correlate lens curvature and extraocular muscle pressure. At the same time, the dynamic training module controls the multispectral adjustment unit to adjust the proportion of red light to enhance ion pump activity.

8. The children's visual function training system based on ciliary muscle accommodation according to claim 7, characterized in that: The data fusion unit performs four-dimensional correlation analysis on corneal surface potential change data, intraocular pressure data, tear film breakup time and pressure deformation data to establish a dynamic coupling model combining potential, intraocular pressure, tear film and pressure. When the linear deviation between the potential drift amplitude and the adjustment amount exceeds the preset potential adjustment deviation, and the intraocular pressure exceeds the fluctuation range or the tear film rupture time is shorter than the preset rupture threshold, the weights of the associated parameters of the mapping model are simultaneously corrected, the proportion of red and green light in the multispectral adjustment unit is adjusted, and the operation of the contrast control unit is paused.

9. The children's visual function training system based on ciliary muscle accommodation according to claim 8, characterized in that: The dynamic coupling model introduces a dynamic baseline calibration coefficient K, which is calculated based on the initial drift value of corneal surface potential, the baseline value of intraocular pressure, and the tear film breakup reference time according to the formula K=0.3×potential baseline+0.2×intraocular pressure baseline+0.5×tear film baseline, and is automatically updated every 30 seconds. The reliability score calculation for the K-value embedded in the dynamic coupling model is as follows: when the K fluctuation is ≤5%, the score weight is allocated according to the original proportion; when the K fluctuation exceeds 5%, the score weight of intraocular pressure and tear film data is increased by 20%. Meanwhile, the K value is linked with the multispectral adjustment unit. When the K fluctuation exceeds 5%, in addition to the adjustment of red and green light, the supplementary light intensity increases linearly with the K value deviation, and triggers the data fusion unit to shrink the error threshold to ±0.2Δ / D.

10. A method for training children's visual function based on ciliary muscle accommodation, characterized in that, The ciliary muscle-based visual function training system for children, as described in any one of the claims, was used.

Images