Retina peripheral defocus detection method and system

Abstract

The invention relates to a retina peripheral defocus detection method and system in the technical field of optical detection, and the method comprises the following steps: driving a moving lens to a preset fogging position, and guiding a user to perform fogging relaxation; keeping a fogging relaxed state, and controlling the mobile lens to execute diopter scanning from a preset fogging position according to a preset step length; controlling the movable lens to axially move within the diopter measurement range at a preset step length, acquiring an image stack after each axial movement to obtain a second image sequence, performing registration processing, identifying center coordinates of all projection light spots, and delimiting a region of interest; calculating a definition evaluation value in each region of interest, and calculating a spherical power; according to the method, astigmatism degree calculation and astigmatism angle calculation are carried out to obtain astigmatism degrees and astigmatism axial positions, the astigmatism degrees and the astigmatism axial positions are mapped to a retina coordinate system, a continuous retina peripheral defocus topographic map is output, and the problem that an existing peripheral defocus detection method excessively depends on fundus natural textures, so that peripheral measurement precision is poor is solved.

Description

Technical Field

[0001] This invention relates to the field of optical detection technology, and specifically to a method and system for detecting peripheral defocus of the retina. Background Technology

[0002] With the continued rise in the prevalence of myopia among adolescents, myopia prevention and control has become an important global public health issue. Existing research indicates that peripheral retinal defocus (i.e., whether the peripheral retina exhibits myopic or hyperopic defocus) plays a crucial role in the occurrence and development of myopia.

[0003] When the peripheral retina exhibits hyperopic defocus, it may promote axial elongation through visual signal modulation mechanisms, thereby exacerbating myopia progression; conversely, myopic defocus is thought to inhibit axial elongation. Therefore, accurate and quantitative detection of the refractive state of the peripheral retina (especially the equatorial region) is of significant clinical importance for the formulation of myopia control strategies, the development of personalized interventions, and the evaluation of clinical efficacy.

[0004] Current methods for peripheral defocus detection include: using traditional computer refractometers and spot scanning technology; using Hartmann-Shack wavefront sensing technology; and using fundus image clarity analysis technology (such as MRT).

[0005] The traditional method of peripheral defocus detection using computerized refractometers and point scanning technology is problematic because clinically used computerized refractometers are typically designed only to measure the refractive power of the fovea (central visual acuity). To obtain defocus information of the peripheral retina, it is necessary to guide the subject's eye movements using different fixation targets for point-by-point testing. Therefore, this process is time-consuming and requires a high degree of cooperation and fixation stability from the subject (especially children and adolescents). Inaccurate eye movements can easily introduce measurement errors due to accommodative fluctuations or fixation deviations. On the other hand, traditional single-point mechanical scanning technology has a slow scanning speed and requires complex motion control mechanisms, making it difficult to complete high-density data acquisition of the entire field of vision in a short time.

[0006] Peripheral defocus detection is achieved using Hartmann-Shack wavefront sensing technology. Some high-end devices utilize wavefront phase aberration meters to measure all-eye aberrations and estimate peripheral refractive status. Although this technology offers high measurement accuracy, it has significant limitations in peripheral defocus detection scenarios: First, there are limitations in cost and size. To cover a wide field of view (e.g., ±40° or even larger), extremely expensive large-area microlens arrays and large-target detectors are required, resulting in high equipment costs and hindering its widespread adoption in primary healthcare institutions. Second, there is the issue of limited range in winter. When detecting eyes with high myopia or drastic changes in peripheral aberrations (e.g., high peripheral astigmatism), the Hartmann spot is prone to crossover, leading to algorithm failure and limiting its dynamic measurement range.

[0007] In recent years, several defocus detection systems based on fundus image sharpness analysis techniques (such as MRT) have emerged, including multispectral refractive topography (MRT) and the technology described in patent CN111358421B. These systems acquire multiple fundus images under different refractive compensations and construct a refractive matrix based on image sharpness. However, this approach has fundamental flaws: First, it heavily relies on the contrast of natural fundus textures (blood vessels, pigments) to calculate the sharpness evaluation function. However, due to the sparse blood vessels and uniform pigment distribution in the peripheral retina (especially the equator), high-contrast texture features are lacking. Second, under infrared illumination, the contrast of peripheral fundus images is already extremely low, and the lack of texture makes it difficult for the algorithm to capture accurate sharpness peaks, resulting in a poor signal-to-noise ratio and significantly reduced repeatability and accuracy of the measurement results.

[0008] In summary, current peripheral defocus detection methods generally suffer from high requirements for the cooperation of the test subjects, expensive equipment that is difficult to promote, and poor peripheral measurement accuracy due to over-reliance on the natural texture of the fundus. Summary of the Invention

[0009] This invention addresses the shortcomings of existing technologies by providing a method and system for detecting peripheral defocus of the retina, thus solving the problem of poor peripheral measurement accuracy caused by excessive reliance on the natural texture of the fundus in existing peripheral defocus detection methods.

[0010] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0011] A method for detecting peripheral retinal defocus includes the following steps:

[0012] Drive the moving lens to the preset fogging position so that the projected infrared dot array forms a normal defocus in front of the retina of the tested eye, guiding the user to relax the fogging.

[0013] While maintaining the fogging relaxation state, the moving lens is controlled to start from the preset fogging position and perform refractive power scanning towards the near or far end at preset step lengths, while simultaneously acquiring image sequence one;

[0014] The moving lens is controlled to move axially within the refractive power measurement range by a preset step size, and the image stack after each axial movement of the moving lens is acquired. The real-time refractive compensation value corresponding to each frame image is recorded to form an image sequence two with refractive power labels.

[0015] The image sequence 2 is registered, and the center coordinates of all projected light spots are identified. A region of interest is defined based on each of the center coordinates.

[0016] Calculate the sharpness evaluation value of the light spot in each region of interest as a function of refractive power, and calculate the sharpness peak position of the infrared dot pattern based on the sharpness evaluation value to obtain the spherical power.

[0017] Astigmatism degree and astigmatism angle are calculated based on the sharpness peak position to obtain astigmatism degree and astigmatism axis.

[0018] The spherical power, astigmatism power, and astigmatism axis of all sampling points are mapped to the retinal coordinate system, and a continuous peripheral defocus topographic map of the retina is output.

[0019] Optionally, the image sequence two is registered, including the following steps:

[0020] The frame with the highest overall contrast in image sequence two is selected as the reference frame.

[0021] The displacement of image sequence two in other frames relative to the reference frame is calculated based on the rigid body transformation algorithm, and translation correction is performed on image sequence two in other frames to make the retinal features of all image sequence twos spatially aligned.

[0022] Optionally, identify the center coordinates of all projected light spots, and delineate the region of interest based on each of the center coordinates, including the following steps:

[0023] An adaptive threshold segmentation algorithm is used on the reference frame to separate the bright dot matrix spots from the dark background, resulting in a binarized mask.

[0024] Based on connected component analysis, the connected regions of each independent light spot are identified, and the centroid coordinates of each connected region are calculated;

[0025] Using each centroid coordinate as the center, a rectangular region of fixed size is cropped as the region of interest for that sampling point.

[0026] Optionally, the sharpness peak position of the infrared dot pattern is calculated based on the sharpness evaluation value to obtain the spherical power, including the following steps:

[0027] A discrete curve is constructed based on the sharpness evaluation value and the corresponding refractive power.

[0028] Curve fitting is performed on the discrete data points on the discrete curve, and the refractive power corresponding to the peak value of the curve fitting is taken as the actual spherical refractive power of the retinal position to obtain the spherical power.

[0029] Optionally, astigmatism calculation is performed, including the following steps:

[0030] Extract the curves showing the change in sharpness of mutually perpendicular first-direction lines and second-direction lines in the composite geometric target within the region of interest as a function of refractive power;

[0031] Identify the position of the moving lens when the first direction line reaches its clearest peak, and convert it into the first meridian diopter;

[0032] Identify the position of the moving lens when the second-direction line reaches its sharpest peak and convert it into the second meridian diopter;

[0033] The astigmatism is calculated based on the absolute value of the difference between the refractive power of the first meridian and the refractive power of the second meridian.

[0034] Optionally, perform astigmatism angle calculation, including the following steps:

[0035] Identify closed ring structures within each region of interest, and extract the elliptical contours formed by the closed ring structures on the imaging sensor based on the magnification differences of the astigmatic eye in different meridian directions.

[0036] The major axis angle of the elliptical contour is determined using a fitting algorithm, and the major axis angle is mapped to the astigmatic axis of the sampling point.

[0037] Optionally, each frame in the second image sequence corresponds to a known refractive compensation value.

[0038] A peripheral retinal defocus detection system, the peripheral retinal defocus detection system being used to perform the peripheral retinal defocus detection method as described in any of the above, including a dot matrix light source projection unit, a fixation guidance and fogging unit, a fundus imaging acquisition unit, and a refractive compensation scanning unit.

[0039] The dot matrix light source projection unit is used to project an infrared dot matrix pattern onto the retina and form a high-contrast artificial texture within the retina.

[0040] The fixation guidance and fogging unit is used to make the projected infrared dot array form a normal defocus in front of the retina of the tested eye, guiding the user to perform fogging relaxation.

[0041] The fundus imaging acquisition unit is used to acquire the image stack after each axial movement of the moving lens, record the real-time descaling compensation value corresponding to each frame image, and form an image sequence with refractive power label.

[0042] The refractive compensation scanning unit is used to employ a telecentric Badal optical structure, including a movable lens, and the movable lens is configured to maintain a constant lateral magnification of the dot pattern projected onto the retina during axial movement, so as to ensure that the field-of-view coordinates of each sampling point in the periphery of the retina do not scale or shift with refractive power scanning.

[0043] Optionally, the dot matrix light source projection unit includes a first polarizer, and the fundus imaging acquisition unit includes a second polarizer, wherein the polarization directions of the first polarizer and the second polarizer are configured to be orthogonal to each other.

[0044] Optionally, the infrared dot pattern is composed of multiple composite geometric targets arranged in an array. Each composite geometric target includes a cross-shaped structure that is perpendicular to each other and a closed ring structure. The cross-shaped structure is used to extract the refractive power of different meridians through the Smith conus principle to calculate the astigmatism degree. The closed ring structure is used to calculate the astigmatism axis by detecting its elliptical deformation characteristics after imaging.

[0045] Compared with the prior art, the technical solution provided by this invention has the following advantages:

[0046] First, it has an extremely high detection signal-to-noise ratio. Compared with MRT technology, this invention actively projects a high-brightness dot matrix, artificially creating a high-contrast texture. Even in the peripheral retinal region where the pigment epithelium is thin and the reflectivity is weak, a high signal-to-noise ratio signal can be obtained, which significantly improves the detection success rate.

[0047] Second, it suppresses stray light (high contrast). Dot illumination is a type of "sparse illumination," with most areas of the fundus in a dark background, effectively reducing intraocular scattered light and decreasing imaging contrast.

[0048] Third, it has the ability to detect astigmatism and axis. Through the design of non-circular (such as cross-shaped) light spots, combined with the Stern light cone principle, the device can not only measure defocus, but also accurately analyze peripheral astigmatism, providing a basis for the design of peripheral astigmatism correction lenses.

[0049] Fourth, the astigmatic axis can be directly mapped by the deformation of the circular structure in the composite geometric target, without the need for complex rotation scanning or large-scale search matching algorithms. This greatly reduces the computational burden on the image processing unit, making it possible to construct high-density peripheral defocus topographic maps on low-power hardware. Attached Figure Description

[0050] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0051] Figure 1 This is a flowchart of a method for detecting peripheral defocus of the retina proposed in Embodiment 1.

[0052] Figure 2 This is a schematic diagram of the imaging of the circular dot matrix light spot and the cross-shaped dot matrix light spot proposed in this embodiment under different refractive states;

[0053] Figure 3 This is a structural diagram of a peripheral retinal defocus detection system proposed in Embodiment 2. Detailed Implementation

[0054] The present invention will be further described in detail below with reference to the embodiments. The following embodiments are explanations of the present invention, but the present invention is not limited to the following embodiments.

[0055] Example 1

[0056] like Figure 1 As shown, a method for detecting peripheral retinal defocus includes step S1: inducing the user to fixate on a specific target and perform fogging relaxation, thereby stabilizing the eye position, eliminating accommodative interference, forcibly relaxing the ciliary muscle, ensuring that the measurement is performed in a naturally relaxed state of the eye, and improving data accuracy.

[0057] Specifically, in order to eliminate the mechanical myopia (accommodative tension) caused by the user's fixation on the device, this device adopts a design with a common optical path for fixation and detection. Before the formal measurement, the system uses the shared Badal zoom mechanism to first execute the automatic fogging logic: control the movement of the moving lens so that the fixation target is preset to a positive defocus state of +2.00D to +3.00D to induce the patient to relax accommodation, and then return to a micro-fog state of +0.25D to +0.50D and keep it locked.

[0058] During this process, the fixation target exhibits a state of "initially significant blurring (forced relaxation), then gradual sharpening, and finally maintaining slight blurring (positive defocus lock)," thereby inducing complete relaxation of the patient's ciliary muscle. The system monitors pupil dynamics in real time using an infrared camera. When the pupil center displacement and pupil diameter change rate are both below preset thresholds over multiple consecutive frames, the system is determined to be in a stable accommodative state. The system automatically triggers infrared dot matrix projection and full-range axial scanning acquisition, ensuring both eye stability and the objectivity and accuracy of peripheral defocus measurement data.

[0059] Step S2: Projection dot matrix: The infrared dot matrix pattern emitted by the infrared dot matrix light source is projected onto the retina through the pupil, forming a high-contrast artificial texture. This solves the problem of focusing in areas with poor fundus features and provides a clear reference for subsequent calculations. Preferably, the infrared dot matrix pattern adopts a composite geometric target (such as a combination of a ring and a crosshair). That is, it is composed of multiple composite geometric targets distributed in an array. Each composite geometric target includes a crosshair structure that is perpendicular to each other and a closed ring structure. The crosshair structure is used to extract the refractive power of different meridians through the Smith light cone principle to calculate the astigmatism degree. The closed ring structure is used to calculate the astigmatism axis by detecting its elliptical deformation characteristics after imaging.

[0060] It should be noted that the high-contrast artificial texture is formed when the near-infrared dot matrix is ​​projected onto the retina at the fundus, creating a luminescent dot matrix on the retina. These luminescent dots are scattered by the retinal tissue, carrying the eye's refractive information, and return along the imaging optical path. They are then captured by the imaging sensor chip, and the resulting image will display these high-contrast dot matrix patterns. If there is no dot matrix projection, the contrast in the obtained fundus image will be very low, leading to low accuracy in data analysis.

[0061] Step S3: Perform sequential scanning acquisition. Control the Badal lens group (i.e., the moving lens) to perform axial scanning at a preset step size (e.g., 0.25D). The area scan camera is triggered synchronously to acquire a series of fundus images (image stack) under different focusing conditions.

[0062] In the process of controlling the Badal lens group to perform axial scanning, the processing unit sends commands to the drive mechanism (such as a stepper motor or piezoelectric motor) in the Badal lens group to control the moving lens to move mechanically along the optical axis. The scanning is performed sequentially according to a preset refractive power measurement range (e.g., -15.00D to +10.00D) and a preset step size (e.g., 0.5D) (equivalent to repeating the "step-pause-acquisition" steps). Specifically, based on the preset refractive power measurement range (e.g., -15.00D to +10.00D) and step size (e.g., 0.5D), the system calculates the corresponding sequence of physical positions that the moving lens needs to move to, and sequentially drives the moving lens to each position node to acquire images. Here, the "fundus image" refers to the dot pattern projected onto the retina, diffusely reflected by the retina, and then passed through the pupil and the system's imaging optical path to finally form an image on the CMOS sensor chip.

[0063] Furthermore, the formation process of the fundus image is as follows: when the moving lens moves to a certain position, the projected light path forms a dot matrix projection with a specific defocus state on the retina; simultaneously, the camera (CMOS / CCD) captures an image through a synchronous trigger signal. Due to the reversibility of the light path, the image sharpness acquired by the camera directly reflects the degree of focus of the projected pattern on the retina under that refractive power. The final output is a sequence of N frames (i.e., an image stack), where the i-th frame corresponds to a known refractive compensation value D. i .

[0064] Next, step S4 is performed: image registration and ROI segmentation are carried out. Using the constant magnification characteristic, the image stack is registered, the center coordinates of all projected light spots in the field of view are identified, and the region of interest (ROI) is defined for each light spot.

[0065] Specifically, the image registration is based on a constant magnification registration strategy: due to the Badal optical path design used in this invention, the magnification of the image remains constant during the movement of the moving lens. Consequently, the spacing between light spots in the image stack does not scale with changes in refractive power, and only the translation or rotation deviation caused by human eye tremors needs to be corrected.

[0066] More specifically, the frame with the highest overall contrast in the image stack is selected as the reference frame. Using rigid body transformation algorithms (such as feature-point matching or gray-level cross-correlation matching), the displacement (d) of each of the other frames relative to the reference frame is calculated. x d y The image is then translated to ensure that the retinal features in all frames are spatially aligned.

[0067] On the other hand, the specific implementation of ROI (Region of Interest) segmentation is as follows: On the reference frame, an adaptive threshold segmentation algorithm (such as the OTSU method) is used to separate the bright dot matrix spots from the dark background to obtain a binarized mask. Connectivity analysis is then used to identify the connected regions of each independent spot and calculate its centroid coordinates (C). x C y For each centroid, a rectangular region of fixed size (e.g., 64×64 pixels) is cropped out as the region of interest (ROI) for that sampling point. The coordinates of this ROI are fixed throughout the image stack, thus extracting a sub-image sequence of each spot as a function of diopter.

[0068] Next, proceed to step S5: perform spot morphology analysis and refractive calculation, including spherical lens calculation, astigmatism calculation, and astigmatism angle calculation.

[0069] In the spherical power calculation, for each spot within a region of interest (ROI), a sharpness evaluation function (such as grayscale gradient or spot diameter) is calculated as a function of refractive power. The refractive compensation value corresponding to the extreme point (the sharpest point) of the function curve is the spherical power at that location.

[0070] The sharpness evaluation function is calculated as follows: for each ROI sub-image sequence, its sharpness evaluation value is calculated. The algorithm for the sharpness evaluation value preferably uses a gray-level gradient function (such as Brenner gradient, Tenengrad function, or Sobel operator energy and Laplacian variance). Taking the Tenengrad function as an example, it calculates the sum of squared gradients of each pixel within the ROI in the horizontal and vertical directions. The larger the sum, the sharper the image edges, i.e., the clearer the image.

[0071] Next, the sharpest point (extreme point) is determined. First, a discrete curve is constructed, with the refractive error D on the horizontal axis and the calculated sharpness evaluation value F on the vertical axis, plotting discrete data points. Then, the curve is fitted. To eliminate noise and obtain finer accuracy than the scanning step size (0.5D), a Gaussian function or a quadratic polynomial is used to fit the discrete data points. Finally, the horizontal coordinate D corresponding to the peak (maximum) of the fitted curve is determined. peak This refers to the actual spherical refractive power at that retinal position.

[0072] When calculating astigmatism (cylindrical lens), if a cross-shaped light spot is used, the peak sharpness positions of the two orthogonal arms (horizontal and vertical) of the cross pattern are calculated separately. Let D1 be the refractive power when the horizontal arm is sharpest; let D2 be the refractive power when the vertical arm is sharpest; then the astigmatism power D at that point is... cyl = |D1 - D2|, equivalent spherical diopter D sphere = (D1+ D2) / 2.

[0073] It should be noted that the astigmatism calculation principle is based on the Schrödinger's light cone principle. Since astigmatism is essentially caused by the eye's optical system having different focal lengths (refractive powers) on two mutually perpendicular principal meridians, a point light source cannot be focused into a single point, instead forming a Schrödinger's light cone. Within this cone, there are two separate focal line positions. When the light path is focused on the focal point of the first principal meridian, lines perpendicular to that meridian are sharp, while parallel lines are blurry; and vice versa. Therefore, the cross-shaped light spot of this invention provides two mutually orthogonal line structures (a horizontal arm and a vertical arm). By tracking the changes in sharpness of these two arms separately, the focal positions of the two principal meridians can be physically separated.

[0074] Furthermore, astigmatism calculation specifically includes the following steps: First, direction separation is performed. Within the ROI, the crosshair spot is separated into a "horizontal layer" and a "vertical layer" through mask extraction or Radon transform. Then, a bimodal search is performed to calculate the sharpness evaluation curves for the horizontal and vertical arms respectively. When the horizontal arm is sharpest, the corresponding first meridian refractive power D1 represents the refractive power of the vertical meridian. When the vertical arm is sharpest, the corresponding second meridian refractive power D2 represents the refractive power of the horizontal meridian. Finally, the results are calculated. The absolute value of the difference between the first and second meridian refractive powers, |D1 - D2|, is the astigmatism power (cylindrical power), because the definition of astigmatism power is precisely the difference between the refractive powers of the two principal meridians, and half of the sum of the two, D... sphere = (D1 + D2) / 2 is the equivalent spherical power (the position of the circle of least confusion).

[0075] When calculating the astigmatism angle (axis), the edge contour of the light spot within the ROI is extracted using the circular or annular structure in the composite target. Due to beam distortion caused by astigmatism, the circular structure appears as an ellipse on the imaging plane. The contour is fitted using an ellipse fitting algorithm (such as the least squares method), and the rotation angle of the major axis of the fitted ellipse relative to the horizontal axis of the sensor coordinate system is calculated. This angle is then determined as the astigmatism axis.

[0076] Specifically, the specific implementation of astigmatism angle (axis) calculation is as follows: (1) Select the best analysis frame. In the image stack acquired in step S3, not every frame is suitable for shape analysis. The system automatically selects the equivalent spherical power (D) based on the calculation results. sphereThe image frame corresponding to the minimum circle of confusion (i.e., the location of the minimum circle of confusion) is used as the analysis object. At this time, the edge of the light spot is the sharpest, the signal-to-noise ratio is the highest, and the shape feature is closest to the real refractive deformation state. Secondly, image preprocessing and edge extraction are performed. During preprocessing, Gaussian filtering is applied to the ROI image in the selected frame to smooth noise, and then binarization is performed using adaptive thresholding (OTSU). During contour extraction, the Canny edge detection operator or morphological gradient operation is used to extract the sub-pixel level edge contour point set P = {(x i y i If a composite target (cross + ring) is used, the outer ring outline needs to be separated through topological analysis of the connected components (such as hole detection) to eliminate interference from the inner cross structure.

[0077] Then, least squares ellipse fitting is performed: assuming the edge contour point set conforms to the general ellipse equation: Then construct the objective function: .

[0078] The coefficient vector [A, B, C, D, E, F] is solved using the direct least squares method to minimize the objective function E. Compared with conventional fitting, this algorithm has the advantages of high numerical stability and forcibly guarantees that the fitted result is an ellipse (rather than a hyperbola).

[0079] Finally, the axis position angle is calculated. Based on the fitted algebraic coefficients A, B, and C, the rotation angle of the major axis of the ellipse relative to the sensor's horizontal axis (x-axis) is calculated. (i.e., astigmatism axis), the calculation formula is: The final output axis angle needs to be transformed according to the ophthalmic standard (TABO notation), usually with the horizontal to the left as 0. 0 / 180 0 Rotate counterclockwise.

[0080] Finally, step S6 is executed: topographic map generation. The refractive data of all sampling points are mapped to the retinal coordinate system, and after interpolation and smoothing, a continuous peripheral defocus topographic map of the retina is output. The refractive data refers to the refractive parameter vector set calculated in step S5 for each discrete sampling point on the retina (i.e., the position of each projected light spot). This vector set contains three core parameters: spherical power (S), cylindrical power (C), and astigmatic axis (A). Furthermore, depending on actual needs, the equivalent spherical power (SE = S + C / 2) and relative peripheral refractive power (RPR = SE{peripheral} - SE{central}) can also be calculated.

[0081] It should be noted that during coordinate mapping, the system uses a pre-calibrated optical distortion model to convert the pixel coordinates (u, v) on the image sensor into the physical coordinates (α, β) of the retinal field of view (e.g., the angle deflected towards the nasal / temporal side with the fovea as the origin). This ensures that the topographic map can accurately reflect the refractive state of the retina at different eccentric positions.

[0082] Spatial interpolation and smoothing are necessary because the projected data is a dot matrix pattern, and the measurement data is discretely distributed in space. To obtain a continuous topographic map (heat map), interpolation is needed for the data gaps. The specific algorithm uses bicubic interpolation or Kriging interpolation to estimate the refractive values ​​of all pixels in the gridded coordinate system based on the measured discrete refractive data points, generating a smooth, continuous surface.

[0083] The final visualization output generates multiple images, which can be pseudo-color topographic maps, including but not limited to: equivalent spherical power topographic map (showing myopia / hyperopia distribution), equivalent cylindrical power topographic map (showing peripheral astigmatism value distribution), and peripheral defocus topographic map (showing the peripheral hyperopic or myopic defocus gradient relative to the center).

[0084] Furthermore, addressing the challenges of peripheral retinal astigmatism measurement, this invention optimizes the geometry of the dot matrix target to adapt to varying detection accuracy requirements and computational power environments. In a preferred embodiment, accurate measurement of peripheral retinal astigmatism (including degree and axis) can be achieved.

[0085] First, an analysis of the limitations of conventional circular cursors: traditional dot matrix projection typically uses a single circular light spot (such as...). Figure 2 (As shown in a). In defocused mode, although a circular spot will be distorted into an ellipse due to astigmatism, making it relatively easy to obtain the astigmatic axis through ellipse fitting, when determining the sharpness peak through axial scanning, the lack of clear directional edge features of the circle makes it difficult to accurately separate the focal positions of the two principal meridians in the frequency or spatial domain. Therefore, using only a circular cursor can usually only accurately obtain the equivalent spherical power (SE), but it is difficult to accurately decouple the spherical and cylindrical (astigmatic) components.

[0086] Second, regarding the crosshair cursor design, to address the problem that circular light spots cannot separate the components of the spherical and cylindrical lenses, this invention proposes a crosshair cursor design (such as...). Figure 2As shown in Figure b), by utilizing the two orthogonal arms (horizontal and vertical) of the crosshair pattern, the two separate focal lines of the Stern conic (corresponding to the sharpness peaks of the two principal meridians) can be clearly observed during axial scanning, thus enabling accurate calculation of the spherical power and astigmatism (Cylinder). Although theoretically, astigmatism axis can be estimated using a crosshair cursor through complex image transformation algorithms (such as full-angle Radon transform), the computational load of the data processing unit increases significantly, and the robustness of axis calculation is not as good as that of a closed geometric figure when the signal-to-noise ratio of peripheral retinal imaging is low. For portable applications that do not require high-precision axis data or have limited computing resources, a crosshair cursor alone can meet the basic requirements for measuring astigmatism amplitude and output the corresponding refractive (including astigmatism) topographic map.

[0087] Third, regarding the composite cursor design, in order to ensure the accuracy of all measured parameters (spherical, cylindrical, and axis) while minimizing algorithm complexity, this invention further preferably adopts a composite cursor design of "circular (or circular) plus crosshair". This target consists of a circular structure for indicating direction and a crosshair structure for indicating focal length superimposed (e.g., ...). Figure 2 (As shown in c). Utilizing the dual-peak clarity characteristic of the cross-shaped structure, the refractive power of the two principal meridians can be quickly and accurately located to calculate the astigmatic amplitude. Furthermore, the elliptical deformation characteristic of the circular structure in the astigmatic fundus can be used to directly read the major axis angle using a low-computational-power ellipse fitting algorithm to obtain the astigmatic axis. This composite design combines the advantages of the aforementioned two types of cursors, achieving the acquisition of the most complete refractive parameters with minimal computational cost.

[0088] Furthermore, this invention utilizes the optical properties of different structures within a composite geometric target to achieve decoupled measurement of refractive parameters: the refractive power measurement principle (based on the Smith conus and cross structure) states that when astigmatism exists in the peripheral retina of the tested eye, the ocular optical system has different refractive powers along two mutually perpendicular principal meridians. This causes the projected cross-shaped structure to be unable to focus simultaneously on the same focal plane: on the first focal plane (corresponding to the first meridian refractive power D1), the first direction lines of the cross pattern (such as the horizontal arm) are the clearest, while the second direction lines are blurry; on the second focal plane (corresponding to the second meridian refractive power D2), the second direction lines of the cross pattern (such as the vertical arm) are the clearest. The difference in refractive power between the two focal planes corresponds to the astigmatism degree.

[0089] Axis measurement principle (based on geometric deformation and circular structure): Due to the difference in magnification of the astigmatic eye along different meridians, the circular or annular structure in the composite target will undergo asymmetrical deformation on the imaging plane, appearing as an ellipse. The rotation angle of the major axis of the fitted ellipse relative to the horizontal axis of the sensor coordinate system is calculated, and this angle is determined as the astigmatic axis.

[0090] Example 2

[0091] like Figure 3 As shown, a peripheral retinal defocus detection system includes a dot matrix light source projection unit, a fixation guidance and fogging unit, a fundus imaging acquisition unit, and a refractive compensation scanning unit. The dot matrix light source projection unit projects an infrared dot matrix pattern onto the retina, forming a high-contrast artificial texture within the retina. The fixation guidance and fogging unit ensures that the projected infrared dot matrix forms a symmetric defocus in front of the retina of the tested eye, guiding the user to perform fogging relaxation. The fundus imaging acquisition unit acquires the image stack after each axial movement of the moving lens, records the real-time defocus compensation value corresponding to each frame, and forms an image sequence with refractive power labels. The refractive compensation scanning unit employs a telecentric Badal optical structure, including a moving lens configured to maintain a constant lateral magnification of the dot matrix pattern projected onto the retina during axial movement, ensuring that the field-of-view coordinates of each sampling point in the peripheral retina do not scale or shift with the refractive power scan.

[0092] Specifically, the dot matrix light source projection unit includes, in sequence along the optical path propagation direction, an infrared light source (Source), a collimating lens L4, a dot matrix mask (Mask), and a first polarizer P1. The collimating lens L4 is placed in front of the infrared light source to convert the diverging light emitted by the light source into parallel light to uniformly illuminate the mask. The dot matrix mask (Mask) is placed at the front focal plane (optical distance) of the auxiliary lens L1.

[0093] The fixation guidance and fogging unit includes a visible light fixation source and a fixation lens L5. The beam generated by this unit is coupled to the main optical path through a dichroic mirror, which is configured to reflect visible light and transmit infrared light.

[0094] The refractive compensation scanning unit (common main optical path) is located behind the dichroic mirror and includes, in sequence, a beam splitter 1 (BS1), a fixed auxiliary lens L1, a movable lens L2 that can move along the optical axis, and a fixed objective lens L3. The auxiliary lens L1 and the objective lens L3 are both fixed elements. The movable lens L2 is located between the auxiliary lens and the objective lens and is configured to be driven by a drive mechanism to move back and forth along the optical axis to change the convergence and divergence (i.e., diopter) of the beam incident on the human eye.

[0095] The fundus imaging acquisition unit is located in the direction of the reflected light path of the beam splitter 1. Along the light path, it includes a second polarizer P2, an imaging lens group L6 and an image sensor (CMOS / CCD). The polarization directions of the first polarizer (P1) and the second polarizer (P2) are configured to be orthogonal to each other (90°).

[0096] The system in this embodiment achieves high-precision, interference-free detection of peripheral retinal defocus through the following core mechanism:

[0097] The Badal constant magnification scanning principle employs an improved Badal optical path design. The dot matrix mask is optically located on the object focal plane of the auxiliary lens L1, forming an image-side telecentric optical path. This means that the spatial position information at different heights on the mask is converted into parallel beams at different angles (angle encoding) after passing through L1. The moving lens L2 moves within these parallel beams to perform a through-focus scan. According to the image-side telecentric optical path principle, the axial movement of L2 only changes the convergence and divergence of the beams (i.e., the front and back positions of the focal plane), without changing the angle at which the main ray enters the eye. Therefore, throughout the entire refractive scan process, the lateral magnification of the dot matrix pattern projected on the retina remains constant, ensuring the accuracy of the spatial correspondence in the calculation of the peripheral defocus topographic map.

[0098] Pupil conjugation and Maxwell's view: The system is configured to satisfy the aperture conjugation condition, that is, after the infrared light source passes through the lens group (L4, L1, L2, L3), its image is formed at the pupil of the subject's eye, such as... Figure 3 As shown in the light trajectory diagram, the principal rays of the beam from the center of the mask (red dashed line) and the peripheral field-of-view beam from the edge of the mask (red dashed line edge) converge at the pupil. This nodal coupling design ensures that the measurement beam can pass through the pupil without obstruction, and even peripheral detection beams with a large field of view are not blocked by the iris, thus achieving effective illumination and measurement of the peripheral retina in a wide field of view.

[0099] Orthogonal polarization stray light elimination principle: In order to solve the problem of low contrast in fundus images, this system introduces polarization gating technology. Incident light path: Infrared light becomes linearly polarized light after passing through P1; Reflection process: When light shines on the cornea and lens surface, specular reflection occurs, and the reflected light maintains its original polarization state; while the light shining on the retina undergoes diffuse reflection, and due to the depolarization effect of biological tissue, the polarization state of the reflected light changes.

[0100] Imaging optical path: A polarizer P2, orthogonal to P1, is placed before imaging. P2 can effectively block strong reflected light (stray light) from the cornea / lens that maintains the original polarization state, and only allow diffuse reflection signals from the retina that have depolarized to pass through, thereby significantly improving the signal-to-noise ratio and contrast of the fundus dot matrix image and ensuring the accuracy of the algorithm in analyzing the spot morphology.

[0101] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any form or substance. It should be noted that those skilled in the art can make various improvements and additions without departing from the method of the present invention, and these improvements and additions should also be considered within the scope of protection of the present invention. Any modifications, alterations, and equivalent changes made by those skilled in the art based on the above-disclosed technical content without departing from the spirit and scope of the present invention are equivalent embodiments of the present invention. Furthermore, any modifications, alterations, and evolutions made to the above embodiments based on the essential technology of the present invention still fall within the scope of the technical solution of the present invention.

Claims

1. A method for detecting peripheral defocus of the retina, characterized in that, Includes the following steps: Drive the moving lens to the preset fogging position so that the projected infrared dot array forms a normal defocus in front of the retina of the tested eye, guiding the user to relax the fogging. While maintaining the fogging relaxation state, the moving lens is controlled to start from the preset fogging position and perform refractive power scanning towards the near or far end at preset step lengths, while simultaneously acquiring image sequence one; The moving lens is controlled to move axially within the refractive power measurement range by a preset step size, and the image stack after each axial movement of the moving lens is acquired. The real-time refractive compensation value corresponding to each frame image is recorded to form an image sequence two with refractive power labels. The image sequence 2 is registered, and the center coordinates of all projected light spots are identified. A region of interest is defined based on each of the center coordinates. Calculate the sharpness evaluation value of the light spot in each region of interest as a function of refractive power, and calculate the sharpness peak position of the infrared dot pattern based on the sharpness evaluation value to obtain the spherical power. Astigmatism degree and astigmatism angle are calculated based on the sharpness peak position to obtain astigmatism degree and astigmatism axis. The spherical power, astigmatism power, and astigmatism axis of all sampling points are mapped to the retinal coordinate system, and a continuous peripheral defocus topographic map of the retina is output.

2. The method for detecting peripheral defocus of the retina according to claim 1, characterized in that, The image sequence two is registered, including the following steps: The frame with the highest overall contrast in image sequence two is selected as the reference frame. The displacement of image sequence two in other frames relative to the reference frame is calculated based on the rigid body transformation algorithm, and translation correction is performed on image sequence two in other frames to make the retinal features of all image sequence twos spatially aligned.

3. The method for detecting peripheral retinal defocus according to claim 2, characterized in that, Identify the center coordinates of all projected light spots, and delineate the region of interest based on each of the center coordinates, including the following steps: An adaptive threshold segmentation algorithm is used on the reference frame to separate the bright dot matrix spots from the dark background, resulting in a binarized mask. Based on connected component analysis, the connected regions of each independent light spot are identified, and the centroid coordinates of each connected region are calculated; Using each centroid coordinate as the center, a rectangular region of fixed size is cropped as the region of interest for that sampling point.

4. The method for detecting peripheral retinal defocus according to claim 1, characterized in that, The sharpness peak position of the infrared dot pattern is calculated based on the sharpness evaluation value to obtain the spherical power, including the following steps: A discrete curve is constructed based on the sharpness evaluation value and the corresponding refractive power. Curve fitting is performed on the discrete data points on the discrete curve, and the refractive power corresponding to the peak value of the curve fitting is taken as the actual spherical refractive power of the retinal position to obtain the spherical power.

5. The method for detecting peripheral defocus of the retina according to claim 4, characterized in that, Calculating astigmatism involves the following steps: Extract the curves showing the change in sharpness of mutually perpendicular first-direction lines and second-direction lines in the composite geometric target within the region of interest as a function of refractive power; Identify the position of the moving lens when the first direction line reaches its clearest peak, and convert it into the first meridian diopter; Identify the position of the moving lens when the second-direction line reaches its sharpest peak and convert it into the second meridian diopter; The astigmatism is calculated based on the absolute value of the difference between the refractive power of the first meridian and the refractive power of the second meridian.

6. The method for detecting peripheral retinal defocus according to claim 5, characterized in that, Calculating the astigmatism angle includes the following steps: Identify closed ring structures within each region of interest, and extract the elliptical contours formed by the closed ring structures on the imaging sensor based on the magnification differences of the astigmatic eye in different meridian directions. The major axis angle of the elliptical contour is determined using a fitting algorithm, and the major axis angle is mapped to the astigmatic axis of the sampling point.

7. A method for detecting peripheral retinal defocus according to any one of claims 1-6, characterized in that, Each frame in the second image sequence corresponds to a known refractive compensation value.

8. A peripheral retinal defocus detection system, characterized in that, The peripheral retinal defocus detection system is used to perform the peripheral retinal defocus detection method as described in any one of claims 1-7, including a dot matrix light source projection unit, a fixation guidance and fogging unit, a fundus imaging acquisition unit, and a refractive compensation scanning unit. The dot matrix light source projection unit is used to project an infrared dot matrix pattern onto the retina and form a high-contrast artificial texture within the retina. The fixation guidance and fogging unit is used to make the projected infrared dot array form a normal defocus in front of the retina of the tested eye, guiding the user to perform fogging relaxation. The fundus imaging acquisition unit is used to acquire the image stack after each axial movement of the moving lens, record the real-time descaling compensation value corresponding to each frame image, and form an image sequence with refractive power label. The refractive compensation scanning unit is used to employ a telecentric Badal optical structure, including a movable lens, and the movable lens is configured to maintain a constant lateral magnification of the dot pattern projected onto the retina during axial movement, so as to ensure that the field-of-view coordinates of each sampling point in the periphery of the retina do not scale or shift with refractive power scanning.

9. A peripheral retinal defocus detection system according to claim 8, characterized in that, The dot matrix light source projection unit includes a first polarizer, and the fundus imaging acquisition unit includes a second polarizer. The polarization directions of the first polarizer and the second polarizer are configured to be orthogonal to each other.

10. A peripheral retinal defocus detection system according to claim 8 or 9, characterized in that, The infrared dot pattern is composed of multiple composite geometric targets arranged in an array. Each composite geometric target includes a cross-shaped structure that is perpendicular to each other and a closed ring structure. The cross-shaped structure is used to extract the refractive power of different meridians through the Smith conus principle to calculate the astigmatism degree. The closed ring structure is used to calculate the astigmatism axis by detecting its elliptical deformation characteristics after imaging.

Images