A dynamic measurement method of diopter based on eye timing image

Abstract

The application provides a dynamic measurement method of diopter based on eye timing images, and relates to the technical field of eye refraction diagnosis, and comprises the following steps: collecting a timing image sequence of a measured eye containing Placido ring projection while presenting programmed switching visual accommodation stimulation to the measured eye, and recording the accommodation stimulation value corresponding to each frame of image; processing each frame of image in the timing image sequence, extracting image features related to the refraction state, and forming timing dynamic features; based on the timing correlation between the timing dynamic features and the accommodation stimulation value, distinguishing and removing the dynamic refraction change component associated with the visual accommodation stimulation in the timing dynamic features, so as to obtain a static diopter estimation value representing the inherent refraction state of the measured eye. The application can realize high-precision static diopter measurement in a non-mydriasis natural state.

Description

Technical Field

[0001] This invention belongs to the field of ocular refractive diagnostic technology, specifically relating to a method for dynamic measurement of refractive power based on temporal images of the eye. Background Technology

[0002] In clinical ophthalmological refractive examinations, obtaining accurate static refractive power (i.e., the refractive state after removing the influence of the eye's accommodative function) is the core basis for refraction and prescription of glasses. Currently, mainstream objective refraction methods, such as computerized refraction, are usually performed in a non-cycloplegic state. However, such methods face a long-standing and insurmountable fundamental technical problem: in a natural (non-cycloplegic) measurement state, the subject's involuntary accommodation, which cannot be completely suppressed, and slight fixational deviations will simultaneously and dynamically interfere with the measurement process.

[0003] Specifically, this manifests as follows: 1) Accommodative interference: Even when patients are asked to fixate on a farsighted target, their ocular accommodation system may still be in an unstable and tense state, or may fluctuate unconsciously during the measurement process, resulting in the measured refractive error value being biased towards myopia (accommodative lag or accommodation lead), and the results have poor repeatability. 2) Fixation deviation interference: It is very difficult to require patients to maintain absolute fixation for a long time. Small eye movements will change the position and angle of the cornea and fundus structures relative to the measurement optical path, introducing geometric optical errors.

[0004] Existing technologies attempt to address this problem, but all have limitations. For example, while measuring by completely paralyzing the accommodative muscle with mydriatic drugs can yield stable static refractive power, it is complex, has side effects, and fails to reflect the visual function of the human eye in its natural state, making it unsuitable for routine screening and rapid refraction. Another approach is to use artificial intelligence to analyze single or multiple static ocular images to predict refractive power, but this essentially involves fitting an apparent refractive state mixed with dynamic interference, failing to fundamentally separate dynamic physiological noise. Therefore, its accuracy and stability significantly decrease when dealing with patients who are accommodatively unstable or poorly cooperative.

[0005] Therefore, how to identify and separate the dynamic optical interference components caused by accommodation and gaze deviation from the measurement signal in real time without relying on mydriatic drugs, so as to stably and accurately obtain the inherent static refractive power of the human eye, has become a technical problem that urgently needs to be solved in this field. Summary of the Invention

[0006] In view of the above-mentioned defects or deficiencies in the prior art, a method for dynamic measurement of refractive power based on temporal images of the eye is provided, comprising the following steps:

[0007] While presenting programmed switching visual accommodation stimuli to the tested eye, a temporal image sequence of the tested eye including the Placido ring projection is acquired, and the accommodation stimulus value corresponding to each frame is recorded.

[0008] Each frame of the temporal image sequence is processed to extract image features related to the refractive state, forming temporal dynamic features;

[0009] Based on the temporal correlation between the temporal dynamic features and the accommodative stimulus value, the dynamic refractive change components associated with the visual accommodative stimulus in the temporal dynamic features are distinguished and removed, thereby obtaining a static refractive power estimate that characterizes the intrinsic refractive state of the tested eye.

[0010] According to the technical solution provided in this application, the presentation of programmed switching visual accommodation stimuli to the tested eye includes the following steps:

[0011] At the beginning of the acquisition of the time-series image sequence, an initial stimulus fixed at optical infinity is presented to the eye being tested and maintained for a first preset duration.

[0012] During the acquisition of the time-series image sequence, the visual accommodation stimulus is switched from an initial stimulus to a near stimulus with a preset refractive power. The near stimulus is used to actively induce the accommodation response of the tested eye.

[0013] After maintaining the near stimulus for the second preset duration, switch the visual accommodation stimulus again;

[0014] The precise time point of each stimulus switch is recorded and synchronously associated with the frame image at the corresponding acquisition time in the time-series image sequence.

[0015] According to the technical solution provided in this application, the image features related to refractive state include geometric features obtained through image processing that characterize the eye's gaze direction in each frame of the image; the gaze direction is obtained by calculating the positional offset of the center of the Placido ring and / or the corneal reflection point relative to the center of the pupil in each frame of the image.

[0016] According to the technical solution provided in this application, before distinguishing and removing the dynamic refractive change component associated with the visual accommodation stimulus in the temporal dynamic features based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value, the method further includes the following steps:

[0017] Based on the geometric features of the gaze direction corresponding to each frame of the image, a predefined gaze direction-refractive deviation mapping relationship is queried to obtain the corresponding refractive measurement deviation;

[0018] If the refractive measurement deviation is greater than or equal to a first preset threshold, the refractive measurement deviation is subtracted from the temporal dynamic features in real time to obtain the corrected temporal dynamic features.

[0019] According to the technical solution provided in this application, the step of distinguishing and removing the dynamic refractive change component associated with the visual accommodation stimulus in the temporal dynamic features based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value includes the following steps:

[0020] Identify the timing of each switching of the programmed visual modulation stimulus;

[0021] Analyze the changes in the temporal dynamic features before and after the switching time point, and extract the dynamic response components that are temporally locked to the stimulus switching event and whose change patterns conform to the physiological laws of eye accommodation.

[0022] The dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component.

[0023] According to the technical solution provided in this application, before removing the dynamic response component as the dynamic refractive change component from the temporal dynamic features, the following steps are further included:

[0024] Before and after the switching time point, a first time window and a second time window are defined respectively;

[0025] Calculate the first statistical feature value of the time-series dynamic feature within the first time window and the second statistical feature value within the second time window;

[0026] The difference between the first statistical feature value and the second statistical feature value is compared with a second preset threshold determined based on the level of physiological nystagmus noise.

[0027] The step of removing the dynamic response component as the dynamic refractive change component from the temporal dynamic features includes the following steps:

[0028] If the difference is greater than the second preset threshold, the dynamic response component is determined to be valid, and the valid dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component.

[0029] According to the technical solution provided in this application, the acquisition of a time-series image sequence of the tested eye including Placido ring projection includes the following steps:

[0030] When acquiring each frame of image, the clarity and integrity of the Placido ring in the current frame image are analyzed in real time.

[0031] Based on real-time analysis results, the acquisition parameters of the imaging device are dynamically adjusted to ensure that the subsequently acquired images meet the preset image quality standards; the acquisition parameters include focus position, illumination intensity, or exposure time.

[0032] According to the technical solution provided in this application, the following steps are also included:

[0033] The tension maintenance state of the tested eye's accommodation system is determined based on the temporal stability characteristics of the dynamic refractive change components extracted during the maintenance of the near stimulus.

[0034] The process of obtaining a static refractive power estimate characterizing the intrinsic refractive state of the tested eye includes the following steps:

[0035] Based on the determination of the state of tension maintenance, data from the maintenance phase of the proximal stimulus, or data from the transition from the initial stimulus to the proximal stimulus, are selectively used for the calculation of the static refractive error estimate.

[0036] According to the technical solution provided in this application, the time-domain stability characteristics include fluctuation amplitude and / or trend indicators;

[0037] The determination result based on the tension maintenance state selectively uses data from the maintenance phase of the near stimulus, or data from the transition from the initial stimulus to the near stimulus, for the calculation of the static refractive error estimate, including the following steps:

[0038] If the fluctuation amplitude is less than the third preset threshold and the trend indicator indicates that the state is stable or tending to be stable, it is determined to be a stable maintenance state, and the data of the stable interval within the maintenance phase is used preferentially for calculation.

[0039] According to the technical solution provided in this application, the step of selectively using data from the maintenance phase of the near stimulus, or data from the transition from the initial stimulus to the near stimulus, for calculating the static refractive error estimate based on the determination result of the tension maintenance state, further includes the following steps:

[0040] If the fluctuation amplitude is greater than or equal to the third preset threshold, or if the trend indicator indicates that the state continues to drift, it is determined to be an unstable maintenance state, and the data during the switching process from the initial stimulus to the nearby stimulus, within the time window in which the dynamic response component is determined to be valid, is used for calculation.

[0041] Compared with the prior art, the beneficial effects of this application are as follows:

[0042] I. Achieving High-Precision Static Refractive Power Measurement in Non-Mydriatic Natural State: By synchronously switching programmed visual accommodation stimuli and acquiring temporal images, an analyzable dynamic observation system was constructed. Utilizing the temporal correlation between temporal dynamic characteristics and accommodation stimulus values, the dynamic refractive change component caused by actively induced accommodation responses can be effectively distinguished and removed. This allows the final static refractive power estimate to be largely free from the interference of involuntary accommodation, thus achieving near-mydriatic refraction accuracy and stability without the need for mydriasis.

[0043] Second, it enhances the anti-interference capability and robustness of the measurement process: actively applying stimuli with known patterns and collecting temporal responses. This active detection method allows the system to distinguish real accommodation signals from random noise (such as physiological tremors, accidental blinks, etc.) based on the time-locking characteristics of physiological responses. Even in cases of slight patient non-cooperation or natural eye movements, effective signal change patterns can be extracted through temporal analysis, thereby significantly improving the measurement success rate and the reliability of the results.

[0044] Third, it provides additional physiological information beyond a single refractive power value: Because the method tracks the dynamic optical response of the eye to programmed stimuli throughout the process, it naturally obtains temporal data on the functional state of the human eye's accommodative system (such as the sensitivity, speed, and stability of the accommodative response) while outputting static refractive power. This provides a potential objective quantitative indicator for the clinical assessment of visual function (such as accommodative insufficiency and accommodative lag), achieving multiple technical benefits. Attached Figure Description

[0045] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0046] Figure 1 A flowchart illustrating the steps of the dynamic refractive power measurement method based on temporal images of the eye provided in this application. Detailed Implementation

[0047] The present application will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0048] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0049] As mentioned in the background section, this application proposes a method for dynamic measurement of refractive power based on temporal images of the eye, such as... Figure 1 As shown, it includes the following steps:

[0050] S1. While presenting programmed switching visual accommodation stimuli to the tested eye, acquire a temporal image sequence of the tested eye including the Placido ring projection, and record the accommodation stimulus value corresponding to each frame.

[0051] S2. Process each frame of the temporal image sequence to extract image features related to the refractive state and form temporal dynamic features;

[0052] S3. Based on the temporal correlation between the temporal dynamic features and the accommodative stimulus value, the dynamic refractive change components associated with the visual accommodative stimulus in the temporal dynamic features are distinguished and removed, thereby obtaining a static refractive power estimate that characterizes the intrinsic refractive state of the tested eye.

[0053] Specifically, the implementation method is as follows: First, a measurement system needs to be built, which includes at least a visual stimulus presentation unit, an eye image acquisition unit, and a processing control unit. The visual stimulus presentation unit is usually an electronically controllable optotype display, capable of displaying optotypes at different optical distances as needed. The image acquisition unit is a corneal topography instrument or similar device equipped with a Placido ring projection module, used to project concentric ring patterns onto the corneal surface and take pictures.

[0054] Programmed switching of visual accommodation stimulation refers to the process control unit controlling the optotype display to switch between different optotypes according to a predetermined time and diopters sequence. For example, the program can be set to first display an optotype simulating optical infinity (0D) for 2 seconds, then instantly switch to a near vision optotype simulating a distance of 33 cm in front of the eyes (+3D) for another 3 seconds. The entire switching logic, duration, and stimulation diopters can all be pre-programmed.

[0055] While presenting the switching stimulus to the tested eye, the image acquisition unit needs to simultaneously capture images of the eye including the Placido ring projection, forming a temporal image sequence. This requires a sufficiently high frame rate for image acquisition (e.g., 30 frames per second or higher) to capture the dynamic changes during the accommodative response. Recording the accommodative stimulus value corresponding to each frame is a crucial synchronization step. This can be achieved by embedding a timestamp in the hardware trigger signal of image acquisition and aligning it with the internal clock or external synchronization clock of the stimulus control program. Ultimately, each frame image file is associated with metadata recording the accommodative stimulus value (in diopters, D) represented by the optotype displayed on the screen at the moment the frame was acquired.

[0056] After obtaining the temporal image sequence, the processing stage begins. Processing each frame of the temporal image sequence involves using image processing algorithms to extract information reflecting the current refractive state of the eye from each image. Since the images contain Placido rings, these features may include, but are not limited to: the shape, curvature, and spacing of the Placido rings, as well as the degree of defocus blur that may be analyzed from fundus reflections. This extracted information is quantified into numerical values, called image features related to the refractive state. Combining all these features arranged in chronological order forms the temporal dynamic features, which are multi-dimensional signals that change over time.

[0057] The next step is the differentiation and removal process. The temporal correlation between the temporal dynamic features and the accommodative stimulus values ​​implies analyzing whether the changing patterns of the temporal dynamic features are related to known accommodative stimulus switching points. For example, when the stimulus switches from 0D to +3D, theoretically, the eye will initiate accommodation, the lens will become more convex, and the refractive power of the eye will dynamically increase. This process will be reflected in the temporal dynamic features, manifested as a specific pattern of change in the feature value after the stimulus switching point. Through signal processing or modeling methods (such as temporal domain analysis locking the stimulus switching point, system identification, etc.), this part of the temporal dynamic features that is temporally locked to the stimulus switching and morphologically conforms to the accommodative physiological response model can be identified. This part is the "dynamic refractive change component associated with the visual accommodative stimulus." Subtracting this dynamic component from the original temporal dynamic features (i.e., "removal"), the remaining signal component is considered to mainly reflect the inherent refractive characteristics of the eye that are not affected by the current programmed stimulus. The final value calculated is the "static refractive power estimate characterizing the inherent refractive state of the tested eye."

[0058] This implementation aims to address the fundamental problem of inaccurate and unstable refractive power measurements caused by involuntary accommodative activity of the tested eye in non-mydriatic states. Its technical principle lies in artificially creating an stimulus-response observation system by actively applying a known and precisely controllable dynamic visual stimulus (programmed switching) and simultaneously acquiring the eye's optical response (temporal image). Then, utilizing the known temporal causal relationship between the response signal (temporal dynamic characteristics) and the stimulus signal (accommodative stimulus value), signal processing techniques are used to separate and eliminate causally related dynamic response components (i.e., induced accommodative responses) from the mixed signal. Ultimately, the retained signal is considered to more purely reflect the eye's static refractive properties, thus achieving high-precision measurement of static refractive power in a natural state. This approach avoids the use of mydriatic drugs, improving the safety and convenience of measurement. Furthermore, by actively separating interference, it theoretically yields more accurate and stable results than passively receiving a single measurement.

[0059] In a preferred embodiment, presenting the programmed switching visual accommodation stimulus to the tested eye includes the following steps:

[0060] At the beginning of the acquisition of the time-series image sequence, an initial stimulus fixed at optical infinity is presented to the eye being tested and maintained for a first preset duration.

[0061] During the acquisition of the time-series image sequence, the visual accommodation stimulus is switched from an initial stimulus to a near stimulus with a preset refractive power. The near stimulus is used to actively induce the accommodation response of the tested eye.

[0062] After maintaining the near stimulus for the second preset duration, switch the visual accommodation stimulus again;

[0063] The precise time point of each stimulus switch is recorded and synchronously associated with the frame image at the corresponding acquisition time in the time-series image sequence.

[0064] Specifically, at the beginning of the acquisition of the time-series image sequence, an initial stimulus fixed to an optical infinity target is presented to the tested eye and maintained for a first preset duration. This means that at the start of the measurement, the tested eye is first given a stimulus that relaxes accommodation. An optical infinity target typically refers to a pattern optically designed to simulate infinity (0 diopter accommodation requirement). Maintaining the target for the first preset duration (e.g., 1.5 to 3 seconds) allows the tested eye's accommodation system sufficient time to relax and stabilize in the initial state, while ensuring that the image acquisition system can acquire a sufficient number of baseline images for subsequent comparison.

[0065] During the acquisition of the time-series image sequence, the visual accommodation stimulus is switched from an initial stimulus to a near stimulus with a preset refractive power. This near stimulus is used to actively induce an accommodation response in the tested eye. This is a crucial step in actively inducing accommodation. While image acquisition continues uninterrupted, the processing control unit issues a command to instantly change the pattern on the optotype display from a distant optotype to a near optotype. This near optotype has a preset refractive power, such as +2.00D, +3.00D, or +4.00D, representing a clear accommodation demand sufficient to stimulate ciliary muscle contraction and lens convexity, thereby actively inducing an observable and significantly significant accommodation response.

[0066] After maintaining the near stimulus for the second preset duration, the visual accommodation stimulus is switched again. After inducing accommodation, the near stimulus needs to be maintained for a period of time (the second preset duration, e.g., 2 to 4 seconds) to observe the process of the accommodation response reaching a steady state and possible fluctuations. Afterward, the stimulus can be switched again, for example, switching back to the distant stimulus to observe the accommodation relaxation process, or switching to another near stimulus of a different degree to conduct a multi-level test. This constitutes a complete stimulation cycle.

[0067] Finally, the precise time point of each stimulus switch is recorded and synchronously correlated with the corresponding frame image at the acquisition time in the time-series image sequence. This is the foundation for realizing time-series correlation analysis. A high-precision timing mechanism (such as the high-precision clock of the operating system or a dedicated hardware clock) must be used to mark the absolute time or relative time to the start of acquisition for each stimulus switch. Simultaneously, each frame of the acquired image also needs to be accurately timestamped. Through timestamp alignment, it can be clearly determined whether the Nth frame image was acquired before, at the moment of, or after the stimulus switch, thus providing the possibility of pinpointing the stimulus switch event in subsequent analysis.

[0068] This implementation specifies that the measurement begins with establishing a baseline (distant stimulation), followed by the application of a defined accommodative load (proximal stimulation), and may include multiple phases. This design simulates the "accommodative stimulus-response" curve determination in clinical accommodative function testing, but seamlessly integrates it into the refractive measurement workflow. Accurate recording of switching time points and synchronization with image frames is a prerequisite for all subsequent event-locked temporal analyses. It ensures that the dynamic response signal has a clear reference origin on the time axis, making it possible to extract effective physiological responses from noise, thereby greatly enhancing the method's ability to distinguish real accommodative signals from random interference.

[0069] In a preferred embodiment, the image features related to the refractive state include geometric features acquired through image processing that characterize the eye's gaze direction in each frame of the image; the gaze direction is obtained by calculating the positional offset of the center of the Placido ring and / or the corneal reflective point relative to the center of the pupil in each frame of the image.

[0070] Specifically, in practice, from each frame of an eye image containing the Placido rings, image analysis is first required to locate several key anatomical landmarks. The first is the pupil center. The pupil's outline can be located in the image using image processing algorithms (such as edge detection and ellipse fitting), and its geometric center calculated. The second is the center of the Placido rings. Since the Placido rings are a set of concentric rings (or partial rings) projected onto the corneal surface, their common center can be fitted by detecting the pattern of these rings in the image. This center theoretically corresponds to the apex of the cornea or the approximate direction of the visual axis. The third is the corneal reflection point, usually referring to the first Purkinje image, i.e., the high-brightness point formed by the reflection of the measuring device's illumination source on the anterior surface of the cornea. This point can also be detected.

[0071] The gaze direction is obtained by calculating the positional offset of the center of the Placido ring and / or the corneal reflector relative to the pupil center in each frame of the image. This means that the gaze direction feature is quantized as one or two two-dimensional vectors. Under ideal emmetropia and fixation on the target optical axis of the device, the pupil center, the center of the Placido ring, and the corneal reflector should be nearly coincident. When the eyeball rotates (gaze direction changes), the pupil moves relative to the eye, while the Placido ring pattern and the corneal reflector, due to their fixed connection to the eyeball, will have relative displacement with respect to the pupil center. Therefore, by calculating the coordinate offsets of the Placido ring center relative to the pupil center (ΔX_ring, ΔY_ring), and / or the coordinate offsets of the corneal reflector relative to the pupil center (ΔX_purkinje, ΔY_purkinje), the angle and direction of eyeball rotation can be deduced. These offsets are the geometric features characterizing the gaze direction of the eye at that moment in the image frame.

[0072] This feature is related to refractive state because changes in gaze direction directly affect the measurement optical path. When the eye rotates, the camera no longer captures the central region of the eye's visual axis, but rather an off-center region. This leads to at least two effects: 1) The Placido ring morphology obtained from the off-center region cannot represent the central corneal curvature, and directly using it for refractive analysis will introduce errors; 2) The eye is a complex optical system, and off-axis imaging introduces aberrations, affecting refractive power estimation based on image analysis. Therefore, quantifying changes in gaze direction is crucial for subsequent correction of these geometric deviations.

[0073] This implementation aims to address the geometrical optical errors introduced into refractive power measurements by the unavoidable minute rotations of the tested eye (fixational instability) during natural measurement processes. Its technical principle is based on the geometric relationships between the eye's anatomical structure and optical imaging. By utilizing multiple naturally occurring positioning landmarks (pupil, ring center, reflection point) in the Placido ring image, high-frame-rate, non-contact estimation of the gaze direction is achieved without adding additional hardware (such as a dedicated eye tracker). This feature can then be used in subsequent steps to model and compensate for systematic measurement biases caused by gaze deviation, thereby effectively improving measurement accuracy and repeatability even with generally low patient cooperation (in the presence of natural minute eye movements). This is a significant supplement and enhancement to existing refractive analysis techniques that rely solely on image photometric or texture information.

[0074] In a preferred embodiment, before distinguishing and removing the dynamic refractive change component associated with the visual accommodation stimulus in the temporal dynamic features based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value, the method further includes the following steps:

[0075] Based on the geometric features of the gaze direction corresponding to each frame of the image, a predefined gaze direction-refractive deviation mapping relationship is queried to obtain the corresponding refractive measurement deviation;

[0076] If the refractive measurement deviation is greater than or equal to a first preset threshold, the refractive measurement deviation is subtracted from the temporal dynamic features in real time to obtain the corrected temporal dynamic features.

[0077] Specifically, the first step is to establish a "predefined gaze direction-refractive deviation mapping relationship." This mapping relationship describes the systematic bias that occurs when the eye's gaze direction deviates from the standard measurement position (usually directly facing the device's optical axis) by a specific angle (quantified by geometric features), affecting the estimated "apparent refractive power" obtained from image analysis. This relationship can be obtained through theoretical modeling or experimental calibration. Theoretical modeling can be based on an optical model of the eye (such as the Gullstrand model) and geometric optics principles to calculate the impact of changes in the measurement area of ​​the anterior corneal surface (the plane containing the Placido ring) and the alteration of the light path on the final refractive power calculation after gaze deviation. Experimental calibration involves having a model eye with a known refractive power or a real subject intentionally perform a series of gaze deviations at known angles on the device, while recording the deviation amount and the deviation between the device-measured refractive power and the true value, thereby fitting the mapping relationship. This relationship can be a lookup table or a mathematical function (such as a polynomial).

[0078] During real-time measurement, for each acquired image frame, the geometric features (i.e., offset vector) of its gaze direction are first calculated. Then, a predefined gaze direction-refractive deviation mapping relationship is queried, and a corresponding refractive measurement deviation value is read or calculated from the mapping relationship based on the offset of the current frame. This deviation value is an estimate, representing the magnitude of the error that might occur if the refractive error is analyzed solely based on this image frame without correcting for the gaze offset.

[0079] Next, a judgment is made: "If the refractive measurement deviation is greater than or equal to a first preset threshold." Setting a "first preset threshold" (e.g., 0.12D) is to avoid unnecessary corrections for negligible small offsets, thereby introducing additional computational noise or rounding errors. Correction is only initiated when the estimated deviation is significant enough to potentially affect the accuracy of the final result.

[0080] If the conditions are met, the refractive measurement bias is subtracted from the temporal dynamic features in real time. The temporal dynamic features here are the original features extracted from the image for subsequent separation of dynamic refractive components. Since gaze offset bias is an error term superimposed on these features, the feature vector corresponding to the frame image is adjusted inversely (subtracted) based on the calculated bias value, thus obtaining the corrected temporal dynamic features. This corrected feature sequence theoretically eliminates the influence of geometric errors caused by gaze direction changes, making it more purely reflect the true refractive state changes of the eyeball, laying a better data foundation for subsequent accurate separation of dynamic components caused by accommodation.

[0081] This implementation directly addresses the specific interference source of fixational micromotion in natural measurements, providing an active and quantitative correction method. Its technical principle lies in converting the physical shift in the gaze direction into an estimation error of the refractive measurement value through a pre-established mapping model, and then compensating for this error in real time at the feature level. This is equivalent to adding a geometric distortion correction step at the front end of the data processing chain. Its beneficial effects are twofold: First, it improves the accuracy of temporal dynamic features, ensuring that any subsequent analysis based on these features (including the separation of accommodative components) is grounded in more reliable data. Second, it reduces the dependence on perfect fixation in the subject, improving the tolerance and robustness of the measurement method. Even if the patient exhibits slight, unconscious eye movements during the measurement, the system can compensate to some extent, potentially yielding more stable and repeatable measurement results, thus broadening the applicability of this technology in clinical practice (especially for groups with potentially poor cooperation, such as children and the elderly).

[0082] In a preferred embodiment, the step of distinguishing and removing the dynamic refractive change component associated with the visual accommodation stimulus in the temporal dynamic features based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value includes the following steps:

[0083] Identify the timing of each switching of the programmed visual modulation stimulus;

[0084] Analyze the changes in the temporal dynamic features before and after the switching time point, and extract the dynamic response components that are temporally locked to the stimulus switching event and whose change patterns conform to the physiological laws of eye accommodation.

[0085] The dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component.

[0086] Specifically, the first step is to identify the timing of each switch in the programmed switching of visual modulation stimuli. This relies on previously established precise timestamp records. The processing unit reads these timing points from the stored synchronization log, such as t_switch1 = 2.000s (switching from the initial stimulus to the near stimulus) and t_switch2 = 5.000s (switching from the near stimulus to the next state). These timing points serve as the absolute time reference for the entire analysis.

[0087] The second step is to analyze the changes in the temporal dynamic features before and after the switching time point. This requires aligning the temporal dynamic feature signal (e.g., a numerical sequence reflecting changes in the overall curvature of the anterior corneal surface) with the switching time point. Taking t_switch1 as an example, the analysis will focus on a time window centered on that point, such as [t_switch1 - 0.2s, t_switch1 + 1.0s]. Within this window, the feature signal is examined at high resolution to find its trajectory of change. Specific algorithms for the analysis may include:

[0088] Differential or derivative calculation: Observe whether the signal slope changes significantly near the switching point.

[0089] Moving average or filtering: Smooth high-frequency noise to observe trend changes.

[0090] Template matching: The correlation between the signal and a standard modulated response waveform template (which has typical characteristics such as rapid rise, peak, slow decay, or stabilization) is calculated.

[0091] The third step is to extract the dynamic response components that are temporally locked to the stimulus switching event and whose change patterns conform to the physiological laws of eye accommodation. This is the key information extraction step.

[0092] Temporal locking means that the extracted signal changes must have a strict temporal relationship with the switching event. For example, the signal should begin to change within 100-300 milliseconds after the switching event (corresponding to the neuromuscular delay in the regulatory response), rather than appearing randomly.

[0093] A morphological conformity to physiological laws is a qualitative requirement for the extracted content. For example, when switching stimuli from far to near, an effective regulatory response component should exhibit a monotonous or near-monotonous change in signal value in a specific direction (representing an increase in refractive power), reaching a new stable plateau or maximum value within a certain time (e.g., 0.5-1.5 seconds), rather than violent oscillations or random walks. This can be determined by examining the sign of the first derivative of the signal within the response window, the number of extreme points, or the goodness of fit with an ideal exponential growth curve.

[0094] The final step is to remove the dynamic response component from the temporal dynamic feature as the dynamic refractive change component. Once the event-locked response waveform A(t) is identified and defined from the original temporal dynamic feature Y(t), a simple vector subtraction is performed: Y_corrected(t) = Y(t) - A(t). Y_corrected(t) is the signal after removing the induced accommodation dynamic component, which is used in the final calculation to obtain an estimate that is closer to the static refractive power.

[0095] In a preferred embodiment, before removing the dynamic response component as the dynamic refractive change component from the temporal dynamic features, the method further includes the following steps:

[0096] Before and after the switching time point, a first time window and a second time window are defined respectively;

[0097] Calculate the first statistical feature value of the time-series dynamic feature within the first time window and the second statistical feature value within the second time window;

[0098] The difference between the first statistical feature value and the second statistical feature value is compared with a second preset threshold determined based on the level of physiological nystagmus noise.

[0099] The step of removing the dynamic response component as the dynamic refractive change component from the temporal dynamic features includes the following steps:

[0100] If the difference is greater than the second preset threshold, the dynamic response component is determined to be valid, and the valid dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component.

[0101] Specifically, the definition of time windows needs to be based on physiological knowledge. For example, for the stimulus switching point t_switch: the first time window (baseline window): can be [t_switch - 0.5s, t_switch]. This represents the period before the stimulus occurs when the regulatory system is in a relatively stable or expected state (such as complete relaxation). The second time window (response window): can be [t_switch + 0.3s, t_switch + 1.0s]. This avoids the neural delay period of regulatory initiation (approximately 0.1-0.3s) and covers the main rising or stabilizing phase of the regulatory response.

[0102] Secondly, statistical characteristic values ​​are used to quantify the overall level of the signal within two windows. The most commonly used is the arithmetic mean. The median can also be used to combat outliers, or the root mean square value can be calculated when the signal fluctuates significantly. Assume the baseline window mean is M1 and the response window mean is M2.

[0103] Next, the difference is typically |M2 - M1| or M2 - M1 (direction considered). The "physiological nystagmus noise level" is a key benchmark in this step and needs to be determined experimentally beforehand. It represents the natural range of fluctuations in the temporal dynamic characteristics caused by inherent micro-nystagmus of the eye, heartbeat, and other physiological activities in the absence of external stimulus changes. This level can be estimated by recording the standard deviation or mean absolute difference of the characteristic signal while fixating on a fixed target for an extended period. A second preset threshold is then set based on this, typically 2-3 times the standard deviation of the noise level (corresponding to the statistical significance level). For example, if the measured standard deviation of the noise fluctuation is 0.05 D, the threshold can be set to 0.10 D or 0.15 D.

[0104] Finally, the determination and operation are as follows: If the difference is greater than the second preset threshold, the dynamic response component is determined to be valid, and the valid dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component. This means that only when the signal level difference between the response window and the baseline window is significantly greater than the inherent background noise of the eye itself, is the system certain that the observed change is a real accommodative response caused by external stimuli, rather than random fluctuations. Only in this case will the removal operation be performed. If the difference is not significant, the extracted dynamic response component is considered unreliable (possibly just noise) and will not be adopted, thereby avoiding the introduction of erroneous correction.

[0105] This implementation addresses a very practical problem: in natural measurements, signal fluctuations can occur even near stimulus switching points due to subject blinking, slight head movements, or instrument noise. By setting an objective threshold based on physiological noise levels, these false positive responses are effectively filtered out. The beneficial effect is a significant enhancement of the robustness and anti-interference capability of the entire method, ensuring that only reliable physiological events with a sufficient signal-to-noise ratio affect the final results, thereby improving the accuracy and repeatability of the measurement results.

[0106] In a preferred embodiment, acquiring a temporal image sequence of the tested eye including the Placido ring projection includes the following steps:

[0107] When acquiring each frame of image, the clarity and integrity of the Placido ring in the current frame image are analyzed in real time.

[0108] Based on real-time analysis results, the acquisition parameters of the imaging device are dynamically adjusted to ensure that the subsequently acquired images meet the preset image quality standards; the acquisition parameters include focus position, illumination intensity, or exposure time.

[0109] Specifically, sharpness analysis: Gradient calculation (such as the Sobel operator) can be used to evaluate the sharpness of the Placido ring edges in the image, calculating the average gradient magnitude of the ring edge region as the sharpness score. Alternatively, frequency domain analysis (Fast Fourier Transform) can be used to examine the energy intensity of the image at the spatial frequencies representing the ring structure. Integrity analysis: Connected regions of the Placido ring can be extracted using image segmentation algorithms (such as thresholding and edge detection), and then the area and circularity of the region can be calculated, or the presence of severe breaks in the ring due to eyelid occlusion or tear film rupture can be detected. A simple metric is the percentage of the effective ring region area to the expected area. Preset ring image quality standards: For example, the sharpness score must be higher than a certain threshold Th_clarity, and the integrity percentage must be higher than Th_completeness.

[0110] Dynamic Adjustment Strategy: Focus Position: If the sharpness remains low, the control unit drives the autofocus motor to fine-tune the lens position and search for the focus until the sharpness score meets the standard. Illumination Intensity: If the ring's contrast is insufficient (affecting sharpness detection) or there is overexposure (causing the ring edges to melt), the intensity of the infrared or visible light source is adjusted. Exposure Time: If the overall image is too dark or too bright, the exposure time of the camera sensor is adjusted to obtain a ring image with optimal contrast. Adjustment is an iterative process. For example, if the system detects insufficient sharpness in frame k, it immediately fine-tunes the focus and acquires frame k+1, re-evaluating until the parameters stabilize near their optimal values.

[0111] In a preferred embodiment, the following steps are also included:

[0112] The tension maintenance state of the tested eye's accommodation system is determined based on the temporal stability characteristics of the dynamic refractive change components extracted during the maintenance of the near stimulus.

[0113] The process of obtaining a static refractive power estimate characterizing the intrinsic refractive state of the tested eye includes the following steps:

[0114] Based on the determination of the state of tension maintenance, data from the maintenance phase of the proximal stimulus, or data from the transition from the initial stimulus to the proximal stimulus, are selectively used for the calculation of the static refractive error estimate.

[0115] Specifically, the first step is the establishment and calculation of state criteria. During the maintenance period of the near stimulus (e.g., from 0.3 seconds after the switching is completed until the next switching), the system performs time-domain analysis on the extracted dynamic refractive change component signal A(t) (which essentially represents the change in amplitude of the accommodative response over time) to calculate its time-domain stability characteristics. Key characteristics include:

[0116] Volatility amplitude: Calculate the standard deviation (σ) of A(t) over this period. For example, σ less than 0.12D can be considered low volatility, and greater than 0.25D can be considered high volatility. This is a direct indicator of the ability to regulate and maintain steady state.

[0117] Trend indicator: A linear regression is performed on A(t) to obtain the slope k. The absolute value and sign of k indicate whether the moderating amount remains stable (|k| ≈ 0), gradually relaxes (k is negative), or remains tense (k is positive) during the maintenance period. For example, if |k| > 0.05 D / s, a significant trend drift is considered to exist.

[0118] Step 2: Determining the State of Tension Maintenance. The system compares the calculated fluctuation amplitude and trend indicators with a preset threshold set and performs a logical judgment. A typical judgment rule could be:

[0119] Stable state is maintained if and only if the fluctuation amplitude is less than the threshold σ and the trend slope is less than the threshold k. For example, threshold σ = 0.15 D, threshold k = 0.05 D / s.

[0120] Unstable maintenance state: This state is determined when any of the above conditions is not met. For example, excessive fluctuation (σ ≥ 0.15D) or obvious drift (|k| ≥ 0.05D / s).

[0121] Step 3: Adaptive selection and calculation instructions for the data source. This determination will directly determine the calculation path for the final static refractive error estimate:

[0122] If the state is determined to be stable, the system instructs subsequent computational units to use data from the maintenance phase of the proximal stimulus. Typically, this is specified as using data from the later part of the maintenance phase, within the interval where A(t) is most stable.

[0123] If the state is determined to be unstable, the system instructs subsequent computational units to use data from the transition from the initial stimulus to the proximal stimulus. This means abandoning the use of chaotic steady-state data and instead analyzing transient response data at the moment of stimulation.

[0124] This implementation method ensures that the selection of data sources matches the current physiological state, avoiding systematic errors that may be introduced by using low-quality data from the source, and providing key adaptability to cope with individual differences (such as the strength of regulatory function).

[0125] Furthermore, the time-domain stability characteristics include fluctuation amplitude and / or trend indicators;

[0126] The determination result based on the tension maintenance state selectively uses data from the maintenance phase of the near stimulus, or data from the transition from the initial stimulus to the near stimulus, for the calculation of the static refractive error estimate, including the following steps:

[0127] If the fluctuation amplitude is less than the third preset threshold and the trend indicator indicates that the state is stable or tending to be stable, it is determined to be a stable maintenance state, and the data of the stable interval within the maintenance phase is used preferentially for calculation.

[0128] Specifically, once the system determines that it is in a stable state, the following operations are performed:

[0129] Determining the stable interval for calculation: Instead of using data from the entire maintenance phase, a sub-interval of optimal quality needs to be identified. The system scans the dynamic component A(t) within the maintenance phase (e.g., starting 0.5 seconds after stimulus switching) to find a continuous period in which the instantaneous value or short-term moving average of A(t) consistently and simultaneously satisfies the following two conditions:

[0130] Fluctuation amplitude < third preset threshold: The fluctuation amplitude here can refer to the local standard deviation within the sub-interval. The third preset threshold is a more stringent value than the judgment threshold (e.g., 0.10 D) to ensure that the segment used for calculation is extremely stable.

[0131] Trend indicators suggest that the state is stable or tending to be stable: when a linear fit is made to A(t) for the sub-interval, the absolute value of the slope should be close to zero, or at least less than a very small threshold (such as 0.02 D / s), indicating that there is no drift trend.

[0132] Calculation execution: Once this stable interval is determined, the system extracts all corrected temporal dynamic features (i.e., data after gaze bias correction and initial separation of dynamic components) within that interval. Since A(t) is nearly constant within this interval, subtracting it from the features yields a very stable residual signal. Ultimately, the static refractive power estimate can be obtained simply by averaging the residual signals of all frames within this interval, or by using the more robust statistical median. This method fully leverages the high signal-to-noise ratio of high-quality steady-state data.

[0133] When the subject's accommodation function is good, this embodiment can automatically lock and utilize the data that achieves a perfect steady state through accommodation response, thereby calculating the theoretically most accurate and reliable static refractive power value, and fully utilizing the upper limit of the performance of this method under ideal conditions.

[0134] Furthermore, the step of selectively using data from the maintenance phase of the near stimulus, or data from the transition from the initial stimulus to the near stimulus, for the calculation of the static refractive error estimate based on the determination result of the tension maintenance state, further includes the following steps:

[0135] If the fluctuation amplitude is greater than or equal to the third preset threshold, or if the trend indicator indicates that the state continues to drift, it is determined to be an unstable maintenance state, and the data during the switching process from the initial stimulus to the nearby stimulus, within the time window in which the dynamic response component is determined to be valid, is used for calculation.

[0136] Specifically, when the system determines that it is in an unstable state, the following operations are performed:

[0137] Enable alternative path: The system abandons the use of data from the entire near-field stimulus maintenance phase.

[0138] Locating and verifying transient data: The focus then shifts to data during the transition from the initial stimulus to the proximal stimulus. Specifically, the system will invoke all the mechanisms established above to handle this particular stimulus transition event:

[0139] Define time windows (such as baseline window and response window) before and after the switching point.

[0140] Calculate the difference ΔM between the statistical characteristic values ​​of the time series dynamics within the two windows.

[0141] ΔM is compared with a second preset threshold determined based on the physiological noise level.

[0142] Key determination: Only when ΔM > the second preset threshold is it determined that this switch has induced a valid adjustment response, and the data time period (especially the response window) covered by the response is marked as the dynamic response component and determined to be a valid time window.

[0143] Calculation Execution: If the above verification passes, the system uses the data within the effective time window for the final calculation. The calculation method at this point may differ from the steady-state path. Since transient response data is used, A(t) is variable. One implementation approach is to employ a system identification strategy: inputting the original temporal dynamics characteristics within the effective time window, along with the known modulating stimulus input (step change), into a pre-defined modulating dynamics differential equation model (such as a first-order hysteresis model). The model parameters (including static refractive power R_static and dynamic parameters) are then simultaneously estimated using fitting techniques. In this way, even if steady-state data is unavailable, the static refractive power can still be solved from a clear excitation transient.

[0144] Handling invalid transients: If verification shows that ΔM is not greater than the second preset threshold, it means that this stimulus switch failed to elicit even a significant regulatory response. In this case, the data from this round will be deemed invalid, triggering a remeasurement or marking the measurement as having extremely low confidence.

[0145] This implementation method, when faced with unstable accommodative function (such as patients with under-accommodation) or poor steady-state data due to inattention during measurement, does not fail outright. Instead, it automatically and intelligently switches to an alternative analysis scheme (transient analysis), thus successfully obtaining valid measurement results even under more challenging conditions. This greatly broadens the applicable population and scenarios for the method and improves the overall success rate.

[0146] The above description is merely a preferred embodiment of this application and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in this application is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the inventive concept. For example, technical solutions formed by substituting the above features with (but not limited to) technical features with similar functions disclosed in this application.

Claims

1. A method for dynamic measurement of refractive power based on temporal images of the eye, characterized in that, Includes the following steps: While presenting programmed switching visual accommodation stimuli to the tested eye, a temporal image sequence of the tested eye including the Placido ring projection is acquired, and the accommodation stimulus value corresponding to each frame is recorded; the precise time point of each stimulus switching is recorded and a synchronous association is established with the frame image at the corresponding acquisition time in the temporal image sequence. Each frame of the temporal image sequence is processed to extract image features related to the refractive state, forming temporal dynamic features; Based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value, the dynamic refractive change components associated with the visual accommodation stimulus in the temporal dynamic features are distinguished and removed. The average or median of the remaining signal components is calculated, or the remaining signal components and the accommodation stimulus value are substituted into a preset accommodation dynamics differential equation model for parameter fitting, thereby obtaining a static refractive power estimate that characterizes the intrinsic refractive state of the tested eye. The step of distinguishing and removing the dynamic refractive change components associated with the visual accommodation stimulus in the temporal dynamic features based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value includes the following steps: Identify the timing of each switching of the programmed visual modulation stimulus; Analyze the changes in the temporal dynamic features before and after the switching time point, and extract the dynamic response components that are temporally locked to the stimulus switching event and whose change patterns conform to the physiological laws of eye accommodation. The dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component.

2. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 1, characterized in that, The presentation of programmed switching visual accommodation stimuli to the tested eye includes the following steps: At the beginning of the acquisition of the time-series image sequence, an initial stimulus fixed at optical infinity is presented to the eye being tested and maintained for a first preset duration. During the acquisition of the time-series image sequence, the visual accommodation stimulus is switched from an initial stimulus to a near stimulus with a preset refractive power. The near stimulus is used to actively induce the accommodation response of the tested eye. After maintaining the near-field stimulus for the second preset duration, the visual accommodation stimulus is switched again.

3. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 1, characterized in that, The image features related to the refractive state include geometric features acquired through image processing that characterize the eye's gaze direction in each frame of the image; the gaze direction is obtained by calculating the positional offset of the center of the Placido ring and / or the corneal reflective point relative to the center of the pupil in each frame of the image.

4. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 3, characterized in that, Before distinguishing and removing the dynamic refractive change components associated with the visual accommodation stimulus in the temporal dynamic features based on the temporal correlation between the temporal dynamic features and the accommodation stimulus value, the method further includes the following steps: Based on the geometric features of the gaze direction corresponding to each frame of the image, a predefined gaze direction-refractive deviation mapping relationship is queried to obtain the corresponding refractive measurement deviation; If the refractive measurement deviation is greater than or equal to a first preset threshold, the refractive measurement deviation is subtracted from the temporal dynamic features in real time to obtain the corrected temporal dynamic features.

5. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 4, characterized in that, Before removing the dynamic response component as the dynamic refractive change component from the temporal dynamic features, the method further includes the following steps: Before and after the switching time point, a first time window and a second time window are defined respectively; Calculate the first statistical feature value of the time-series dynamic feature within the first time window and the second statistical feature value within the second time window; The difference between the first statistical feature value and the second statistical feature value is compared with a second preset threshold determined based on the level of physiological nystagmus noise. The step of removing the dynamic response component as the dynamic refractive change component from the temporal dynamic features includes the following steps: If the difference is greater than the second preset threshold, the dynamic response component is determined to be valid, and the valid dynamic response component is removed from the temporal dynamic features as the dynamic refractive change component.

6. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 1, characterized in that, The acquisition of the temporal image sequence of the tested eye, including the Placido ring projection, includes the following steps: When acquiring each frame of image, the clarity and integrity of the Placido ring in the current frame image are analyzed in real time. Based on real-time analysis results, the acquisition parameters of the imaging device are dynamically adjusted to ensure that the subsequently acquired images meet the preset image quality standards; the acquisition parameters include focus position, illumination intensity, or exposure time.

7. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 2, characterized in that, It also includes the following steps: The tension maintenance state of the tested eye's accommodation system is determined based on the temporal stability characteristics of the dynamic refractive change components extracted during the maintenance of the near stimulus. The process of obtaining a static refractive power estimate characterizing the intrinsic refractive state of the tested eye includes the following steps: Based on the determination of the state of tension maintenance, data from the maintenance phase of the proximal stimulus, or data from the transition from the initial stimulus to the proximal stimulus, are selectively used for the calculation of the static refractive error estimate.

8. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 7, characterized in that, The time-domain stability characteristics include fluctuation amplitude and / or trend indicators; The determination result based on the tension maintenance state selectively uses data from the maintenance phase of the near stimulus, or data from the transition from the initial stimulus to the near stimulus, for the calculation of the static refractive error estimate, including the following steps: If the fluctuation amplitude is less than the third preset threshold and the trend indicator indicates that the state is stable or tending to be stable, it is determined to be a stable maintenance state, and the data of the stable interval within the maintenance phase is used preferentially for calculation.

9. The method for dynamic measurement of refractive power based on temporal images of the eye according to claim 8, characterized in that, The method of selectively using data from the maintenance phase of the near stimulus, or data from the transition from the initial stimulus to the near stimulus, for calculating the static refractive error estimate based on the determination result of the tension maintenance state, further includes the following steps: If the fluctuation amplitude is greater than or equal to the third preset threshold, or if the trend indicator indicates that the state continues to drift, it is determined to be an unstable maintenance state, and the data during the switching process from the initial stimulus to the nearby stimulus, within the time window in which the dynamic response component is determined to be valid, is used for calculation.

Images