A method for the preparation of a personalized eye mask for corneal cross-linking surgery

Abstract

The application discloses a preparation method of a personalized eye membrane for corneal cross-linking surgery, and comprises the following steps: obtaining corneal medical image data of a patient and establishing a three-dimensional digital model of the cornea; establishing a corneal reference model library, generating a corneal local anomaly score map based on a selected reference corneal model, and determining a target treatment area; performing contour extraction, outward expansion, smoothing and manufacturing constraint correction on the target treatment area to generate a penetration hole contour; establishing an eye membrane body model according to corneal front surface data, mapping the penetration hole contour to the eye membrane body model to form a through penetration hole; and generating a manufacturing model and preparing a personalized eye membrane. The abnormal morphology reflected by the corneal image data of the patient is converted into a processable penetration hole structure on the eye membrane, and a personalized eye membrane preparation method based on the individual morphology of the patient is provided for the regionalized guidance of riboflavin liquid in the corneal cross-linking surgery.

Description

Technical Field

[0001] This invention belongs to the field of ophthalmic medical device technology, specifically relating to a method for preparing a personalized eyepiece for corneal cross-linking surgery. Background Technology

[0002] Corneal cross-linking is a routine clinical treatment used to curb the progression of keratoconus. It works by using riboflavin to penetrate the corneal stroma and triggering a photochemical reaction under ultraviolet light, causing new chemical cross-links to form between corneal collagen fibers. This strengthens the cornea and slows down corneal ectasia. Because keratoconus often presents as localized thinning and bulging of the cornea rather than a uniform change across the entire cornea, personalized corneal cross-linking techniques have gained increasing attention in recent years. The core concept is to spatially selectively control the extent of cross-linking treatment based on the morphology of the abnormal corneal area, thus avoiding unnecessary cross-linking of still-normal corneal tissue.

[0003] Currently, the main technical means to achieve personalized corneal crosslinking focus on spatial modulation during the ultraviolet (UV) irradiation stage. Common approaches involve placing a programmable liquid crystal mask or digital micromirror device between the UV light source and the cornea to generate an illumination pattern corresponding to abnormal corneal areas, or using a contact lens with both transparent and shielding areas as a physical photomask to control the UV light transmission range. These approaches share the characteristic that riboflavin is still delivered via full corneal immersion, and the crosslinking area is only indirectly defined by controlling the UV irradiation range. This approach requires the irradiation device to have high spatial resolution and the ability to precisely align with abnormal corneal areas, increasing system complexity and equipment cost. Furthermore, because riboflavin has already infiltrated the entire cornea before irradiation, true regional control cannot be achieved from the drug delivery stage from a treatment logic perspective. In addition, existing technologies lack a method to systematically transform the abnormal morphology reflected in a patient's corneal medical images into a customizable drug-guided membrane structure. Therefore, there is an urgent need to improve existing technologies to address the lack of drug infiltration area guidance structures and their preparation methods based on the individual morphology of the patient's cornea in personalized corneal cross-linking therapy. Summary of the Invention

[0004] The purpose of this invention is to overcome the shortcomings of the prior art and provide a method for preparing a personalized eyepiece for corneal cross-linking surgery, so as to solve the problem that existing corneal cross-linking auxiliary devices are difficult to generate manufacturable and positionable drug permeation pore structures according to the abnormal morphology of the patient's cornea.

[0005] To achieve the above objectives, the present invention adopts the following technical solution: A method for preparing a personalized ocular transluminal membrane for corneal cross-linking surgery includes the following steps: S100: Acquire medical imaging data of the patient's cornea, wherein the medical imaging data includes at least corneal topography data, corneal tomography data, and corneal thickness data; S200: Preprocess the medical image data and establish a three-dimensional digital model of the patient's cornea based on the preprocessed data. The three-dimensional digital model of the cornea includes a three-dimensional mesh of the anterior surface of the cornea, a three-dimensional mesh of the posterior surface of the cornea, and a corneal thickness distribution determined by the three-dimensional mesh of the anterior surface of the cornea and the three-dimensional mesh of the posterior surface of the cornea. S300: Establish a corneal reference model library, generate a corneal local abnormality scoring map based on the corneal three-dimensional digital model and a reference corneal model selected from the corneal reference model library, and determine at least one target treatment area from the corneal local abnormality scoring map; S400: Perform contour extraction, expansion and contour smoothing on each target treatment area to generate a permeation pore contour corresponding to the shape of the target treatment area. The permeation pore contour is a single closed curve. S500: Based on the corneal anterior surface data in the three-dimensional digital model of the cornea, establish an ocular body model that matches the curvature of the patient's cornea. The ocular body model is an arc-shaped sheet structure with an inner surface, an outer surface, and a peripheral edge. S600: Map the permeation pore outline from the corneal 3D digital model to the eye membrane body model, and generate a permeation pore penetrating the inner surface and the outer surface in the eye membrane body model; S700: Generate a manufacturing model based on the ocular body model with the permeation pores, and use the manufacturing model to prepare a personalized ocular membrane for corneal cross-linking surgery.

[0006] Preferably, step S200 includes: removing abnormal data points in the medical image data caused by eye movement or tear film instability; unifying image data from different acquisition devices to the same coordinate system through spatial registration with the corneal vertex and pupil center as reference markers, and performing orientation correction with the corneal vertex or pupil center as the coordinate origin; using the corneal topography data to fit and generate the three-dimensional mesh of the anterior corneal surface, and using the corneal tomography data to fit and generate the three-dimensional mesh of the posterior corneal surface; calculating the corneal thickness distribution based on the corneal thickness map data, or based on the sag difference of corresponding points on the three-dimensional mesh of the anterior corneal surface and the three-dimensional mesh of the posterior corneal surface.

[0007] Further, in step S300, establishing a corneal reference model library includes: performing surface fitting based on the peripheral region with regular curvature in the three-dimensional mesh of the anterior corneal surface to generate a first reference corneal model; performing mirror processing based on the three-dimensional digital model of the patient's contralateral cornea to generate a second reference corneal model; matching a model from a healthy corneal database that is similar to the patient's corneal basic parameters as a third reference corneal model; in step S300, selecting a reference corneal model from the corneal reference model library includes: calculating the morphological similarity between the three-dimensional digital model of the cornea and the first reference corneal model, the second reference corneal model, and the third reference corneal model using a similarity matching algorithm, and selecting the model with the highest morphological similarity as the reference corneal model.

[0008] Preferably, step S300, which generates a local corneal anomaly scoring map based on the three-dimensional digital model of the cornea and a reference corneal model selected from the corneal reference model library, and determines at least one target treatment area from the local corneal anomaly scoring map, includes: comparing the three-dimensional digital model of the cornea with the selected reference corneal model point by point; calculating the anterior surface curvature anomaly deviation using the corneal topography data; calculating the corneal thickness anomaly deviation using the corneal thickness map data; calculating the posterior surface height anomaly deviation using the corneal tomography data; weightedly fusing at least two of the anterior surface curvature anomaly deviation, the corneal thickness anomaly deviation, and the posterior surface height anomaly deviation to generate the local corneal anomaly scoring map; binarizing the local corneal anomaly scoring map according to a preset scoring threshold; extracting continuous regions with scores higher than the preset scoring threshold as initial target regions; and performing connected component analysis on the initial target regions to determine spatially separated connected components as independent target treatment areas.

[0009] Furthermore, step S300 also includes: displaying the corneal local abnormality scoring map superimposed on the corneal three-dimensional digital model or corneal two-dimensional projection map on the display interface; receiving the operator's closed outline input for the target treatment area; and performing union or intersection fusion of the area defined by the closed outline input with the automatically segmented initial target area to finally determine the target treatment area.

[0010] Preferably, step S400 includes: extracting the closed boundary of the target treatment area on the surface of the three-dimensional digital model of the cornea as an initial contour; expanding the initial contour outward at equal intervals along its normal to obtain an expanded contour; smoothing the expanded contour to eliminate sharp corners and jagged edges, wherein the smoothing is performed using any one of B-spline curve fitting, Bezier curve fitting, or multi-segment circular arc fitting; and applying manufacturing constraint correction to the smoothed contour, wherein the manufacturing constraint correction includes limiting the minimum local radius of curvature of the contour, limiting the minimum area enclosed by the contour, and limiting the minimum distance from any point on the contour to the peripheral edge of the corneal body model, to generate the permeation pore contour.

[0011] Further, step S500 includes: using the three-dimensional mesh of the anterior corneal surface as a base surface, offsetting a preset fitting gap along the normal direction of the anterior corneal surface to obtain the inner surface of the eye membrane body model; using the inner surface as a base, offsetting a preset thickness outward along the normal direction to obtain the outer surface of the eye membrane body model; and smoothly connecting the boundaries of the inner surface and the outer surface to form the peripheral edge.

[0012] Preferably, step S600 includes: establishing a coordinate correspondence based on a preset mapping relationship between the corneal vertex coordinate system of the three-dimensional digital model of the cornea and the geometric center coordinate system of the ocular body model; projecting the closed curve of the permeation pore outline along the local normal direction of the cornea onto the inner surface of the ocular body model to form an inner surface pore boundary curve; extending the inner surface pore boundary curve along the thickness direction of the ocular body model to the outer surface to form an outer surface pore boundary curve on the outer surface; generating a continuous surface between the inner surface pore boundary curve and the outer surface pore boundary curve as a pore wall, wherein the pore wall, the opening on the inner surface, and the opening on the outer surface together constitute the permeation pore.

[0013] Further, step S700 includes: converting the eye membrane body model with the permeation holes into a manufacturing model file containing three-dimensional surface information and hole boundary information, wherein the manufacturing model file is at least one of STL format, STEP format, IGES format, DXF hole contour format, CNC turning machining code, or laser hole opening path code; inputting the manufacturing model file into a CNC machining equipment, and manufacturing an arc-shaped eye membrane with the permeation holes by any one of the following methods: turning and then laser hole opening, turning and then micro-milling hole opening, mold forming, or three-dimensional printing mold followed by casting, thereby obtaining the personalized eye membrane.

[0014] Preferably, the method further includes: generating a directional positioning structure on the outer surface or peripheral edge of the ocular body model, wherein the directional positioning structure is at least one of a laser marking at the 12 o'clock position, a nasal edge notch, a temporal edge mark, or an asymmetrical edge contour, to indicate the orientation of the permeation hole relative to the corneal abnormality area when worn.

[0015] This invention discloses a method for preparing a personalized ocular membrane for corneal cross-linking surgery. The core of this method lies in transforming the local abnormal morphology reflected in the patient's corneal medical imaging data into a processable through-hole structure on the ocular membrane itself through 3D modeling, abnormality scoring, region segmentation, contour generation, surface mapping, and manufacturing constraint correction. This establishes a complete data link from individual patient corneal data to personalized drug-guided membrane manufacturing documents. Compared to existing technologies that primarily rely on ultraviolet light patterns to control the cross-linking area, this invention starts from the preparation stage of the drug delivery device, enabling selective control of the riboflavin infiltration area at the ocular structure level. This avoids the problem of non-targeted photosensitizer distribution caused by limiting the area through light after full corneal infiltration. By generating an abnormality scoring map through point-by-point comparison between the 3D digital corneal model and a reference corneal model, and allowing the operator to perform closed-line correction on the automatic segmentation results, the determination of the target treatment area combines data objectivity and clinical intervention, reducing morphological deviations caused by judging the pore position solely based on 2D images or human experience. During the generation process, the permeation pore contour undergoes outward expansion, curve smoothing, and manufacturing constraint correction. This transforms irregular lesion boundaries into a single closed curve that meets processing requirements such as minimum radius of curvature, minimum pore area, and pore edge distance. This allows the design results to be directly output as manufacturing files such as STL, STEP, DXF, CNC turning code, or laser drilling paths. This aligns with common processing techniques such as laser drilling after turning, micro-milling, mold forming, or casting after 3D printing, improving the industrial manufacturability of personalized eye masks. Furthermore, by generating a directional positioning structure on the eye mask model with a fixed orientation to the permeation pore location, the manufactured eye mask can be quickly aligned with the corneal abnormality area based on visible markings during clinical wear, reducing the risk of pore area displacement due to rotational misalignment. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of the present invention, 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.

[0017] Figure 1This is a schematic diagram of the overall process of a method for preparing a personalized eye mask for corneal cross-linking surgery according to the present invention; Figure 2 A three-dimensional digital model of the patient's cornea, in which Figure 2 (a) A three-dimensional digital model of the anterior and posterior surfaces of the patient's cornea. Figure 2 (b) is a three-dimensional visualization of the corneal thickness distribution superimposed on the anterior surface of the cornea; Figure 3 This is a scoring chart for localized corneal abnormalities. Figure 4 A schematic diagram is generated for the outlines of multiple permeable pores; Figure 5 A 3D manufacturing model of a personalized eye mask with through-holes; Figure 6 A simplified diagram illustrating different personalized eye masks generated based on the abnormal corneal areas of different patients. Detailed Implementation

[0018] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0019] In the description of this invention, it should be noted that the terms "center," "upper," "lower," "left," "right," "vertical," "horizontal," "inner," and "outer," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limiting this invention.

[0020] Existing personalized corneal crosslinking technology can only achieve spatial selective control during the light exposure phase. Riboflavin still uses a whole-corneal immersion method, which cannot achieve true regionalized treatment from the drug delivery stage. Furthermore, there is a lack of a systematic method to transform the abnormal corneal morphology of patients into a manufacturable drug-guided structure.

[0021] Based on this, and to improve upon the problems in related technologies, embodiments of this application provide a method for preparing a personalized ocular membrane for corneal cross-linking surgery, such as... Figure 1 As shown, it includes the following steps: S100, acquire medical imaging data of the patient's cornea, including at least corneal topography data, corneal tomography data, and corneal thickness data; S200 preprocesses medical image data and establishes a three-dimensional digital model of the patient's cornea based on the preprocessed data. The three-dimensional digital model of the cornea includes a three-dimensional mesh of the anterior surface of the cornea, a three-dimensional mesh of the posterior surface of the cornea, and the corneal thickness distribution determined by the three-dimensional mesh of the anterior surface of the cornea and the three-dimensional mesh of the posterior surface of the cornea. S300: Establish a corneal reference model library, generate a corneal local abnormality scoring map based on the corneal three-dimensional digital model and the reference corneal model selected from the corneal reference model library, and determine at least one target treatment area from the corneal local abnormality scoring map; S400 performs contour extraction, expansion and contour smoothing on each target treatment area to generate a permeation pore contour corresponding to the shape of the target treatment area. The permeation pore contour is a single closed curve. S500 establishes an ocular body model that matches the patient's corneal curvature based on the corneal anterior surface data in the three-dimensional digital model of the cornea. The ocular body model is an arc-shaped sheet structure with an inner surface, an outer surface, and a peripheral edge. S600 maps the permeation pore outline from the corneal 3D digital model to the ocular body model, generating permeation pores that penetrate the inner and outer surfaces in the ocular body model. S700 generates a manufacturing model based on an ocular body model with permeable pores, and uses the manufacturing model to prepare a personalized ocular membrane for corneal cross-linking surgery.

[0022] Medical imaging data is a dataset used to express the geometry and local abnormalities of a patient's cornea. Corneal topography data, acquired through Placido disc projection or Scheimpflug imaging techniques, provides height information for tens of thousands to hundreds of thousands of sampling points on the anterior corneal surface within a range of approximately 10 mm from the center to the periphery, allowing for the calculation of curvature, refractive power, and axial distribution at each point. Corneal tomography data, obtained through optical coherence tomography (OCT), displays cross-sectional structures of the cornea from the epithelium to the endothelium at micron-level resolution, clearly showing the morphological characteristics of the posterior corneal surface and the interlayer interfaces within the cornea. Corneal thickness mapping data can be directly output from a corneal topography instrument or OCT device, or calculated from the height difference between the anterior and posterior surfaces, reflecting the thickness variations at different locations of the cornea. A 3D digital model of the cornea can be established using point clouds, triangular mesh surfaces, or parametric surfaces. Point cloud models are directly composed of original sampling points, preserving the most original data information. Triangular mesh surfaces form triangular patches by connecting adjacent sampling points, efficiently representing complex surfaces. Parametric surfaces, such as non-uniform rational B-spline (NURBS) surfaces, are defined using control points and basis functions, exhibiting good continuity and editability. The goal is to represent the anterior and posterior corneal surfaces and corneal thickness distribution within the same coordinate system. The corneal local abnormality scoring map is not used to output disease diagnostic conclusions but rather to generate design data for personalized corneal pore structures. The results are used to define the position and extent of the target treatment area on the corneal surface; a higher score indicates a more severe abnormality in that area, requiring more cross-linking enhancement treatment. The permeation pore contour is a single closed curve, meaning that the pore contour corresponding to each target treatment area forms a continuous, uninterrupted, and non-self-intersecting closed boundary. This can be subsequently converted into a pore wall surface penetrating the entire corneal body, ensuring that riboflavin solution can accurately penetrate the target treatment area below through this pore.

[0023] Through the above technical solution, the abnormal corneal areas of patients are no longer merely confined to examination images or operator annotations, but are transformed into processable corneal pore structures through modeling, scoring, contour generation, and surface mapping. Since the pore contour originates from the target treatment area on the 3D digital model of the cornea, the position and shape of the pores correspond to the individual corneal data of the patient. Furthermore, because the corneal body model is based on anterior corneal surface data, the final manufacturing model can be directly used by subsequent processing equipment. This process connects medical image processing, personalized corneal modeling, and manufacturing data output, reducing surface distortion caused by directly converting 2D images into pores, and providing a clear data source and processing target for the production of personalized corneal membranes.

[0024] The present invention further proposes that step S200 includes: removing abnormal data points in medical image data caused by eye movement or tear film instability; unifying image data from different acquisition devices to the same coordinate system through spatial registration with the corneal apex and pupil center as reference markers, and performing orientation correction with the corneal apex or pupil center as the origin; generating a three-dimensional mesh of the anterior corneal surface using corneal topographic map data, and generating a three-dimensional mesh of the posterior corneal surface using corneal tomographic data; calculating the corneal thickness distribution based on corneal thickness map data, or based on the sag difference of corresponding points on the three-dimensional mesh of the anterior and posterior corneal surfaces.

[0025] Abnormal data points can manifest as sudden changes in local height, deviations from duplicate points within the same acquisition area, acquisition defects, or discontinuous curvature changes. In preprocessing, low-quality data points can be initially screened using the quality markers built into the acquisition device, and then further identified as abnormal points by assessing the variation in height between adjacent grid points. For example, for a given sampling point, if the difference between its height value and the average height value of sampling points in its surrounding 3×3 or 5×5 neighborhood exceeds three times the standard deviation, it is considered an abnormal data point. Alternatively, judgment can be made using surface fitting residuals. A local quadratic surface is fitted based on the neighboring sampling points, and the vertical distance from the sampling point to the fitted surface is calculated. If the distance exceeds a preset threshold, the sample is discarded. Spatial registration uses the corneal vertex and pupil center as common reference landmarks. These two anatomical landmarks have high recognizability and stability in images from different devices. First, the corneal vertex (the point with the greatest curvature on the anterior surface of the cornea) and the pupil center (the geometric center of the pupillary region) are automatically identified in each image. Then, using these two points as references, rotation and translation transformation parameters are calculated to transform all data into the same coordinate system, avoiding relative offsets between the anterior and posterior corneal surface and corneal thickness data. Orientation correction can be performed based on left and right eye information, the nasotemporal direction, and the vertical direction, ensuring that the subsequent perforation aperture contour does not suffer from mirror errors due to eye differences. For the left and right eyes, corresponding coordinate systems are established separately to ensure that the generated corneal model matches the patient's actual eye type. The three-dimensional meshes of the anterior and posterior corneal surfaces are generated using non-uniform rational B-spline surface fitting, which can accurately describe the complex surface morphology of the cornea, ensuring the continuity and smoothness of the surface. Corneal thickness distribution can be obtained directly from data measured by a corneal thickness mapping device, or calculated from the distance between corresponding points on the anterior and posterior surface three-dimensional meshes. These two methods can be mutually verified to improve the accuracy of the thickness data. When the device directly provides high-quality corneal thickness maps, the measured data is used first; when the measured data contains noise or is missing, it is supplemented by calculation of the height difference between the anterior and posterior surfaces.

[0026] As a specific implementation method, when removing outlier data points, the first... k The output of the first acquisition devicei Each sampling point is denoted as ,in , and These represent the three-dimensional coordinates of the sampling point in the coordinate system of the corresponding acquisition device. For each sampling point, a local reference surface can be obtained by fitting it based on its neighboring sampling points. The height residual of the sampling point is defined as... ,in This indicates the degree of deviation of the sampling point relative to the neighboring reference surface. The average height residual of all sampling points within the same acquisition area is calculated. and standard deviation ,when When this happens, the sampling point is identified as an outlier and removed. This is the outlier screening coefficient, typically set to 2.0-3.0. For spatial registration, data from different acquisition devices can be uniformly mapped to a common corneal coordinate system through a rigid transformation. The transformation relationship is expressed as follows: ,in Indicates the converted sampling points. Indicates the first k The rotation matrix of each acquisition device's coordinate system relative to the common corneal coordinate system. This represents the corresponding translation vector. The common corneal coordinate system can be based on the corneal vertex. Using the origin as the reference point and the distance from the corneal apex to the center of the pupil as the reference point. The directional relationship determines the coordinate axis orientation. When calculating corneal thickness distribution, the first [axis] on the anterior surface of the cornea can be [used to determine the coordinate axis orientation]. i The corresponding points are denoted as The corresponding point on the posterior surface of the cornea is denoted as Place the anterior surface of the cornea on The unit normal vector at point is denoted as The corneal thickness at that location is then expressed as... ,in Indicates the first i Corneal thickness at each grid location, symbol " "" indicates vector dot product. This thickness calculation method is performed along the local normal of the cornea, which can maintain correspondence with the corneal surface morphology and avoid the errors caused by calculations in the vertical direction.

[0027] As an alternative embodiment, when the acquisition device directly outputs registered anterior corneal surface data, posterior corneal surface data, and corneal thickness data, the registration result output by the device can also be read as initial data, and then directional correction and outlier removal can be performed on it. For locally missing sampling areas, interpolation of adjacent valid sampling points, surface fitting to fill points, or directly marking the missing area as a non-design area can be used to ensure that the subsequent perforation contour does not pass through the location of insufficient data quality.

[0028] This setup converts corneal data from multiple sources into a 3D model in a unified coordinate system, allowing subsequent anomaly scoring and contour extraction to be performed based on the same spatial reference. Anomaly point removal reduces the impact of acquisition errors on the lesion scoring map, orientation correction minimizes aperture position errors caused by left and right spectacle images, and the synchronized establishment of the anterior and posterior corneal surface mesh and corneal thickness distribution allows local corneal morphology and thickness information to simultaneously participate in target area determination. This step provides a data foundation for subsequent reference model comparison and aperture contour generation, preventing the final corneal aperture position from deviating from the target area due to inconsistencies in coordinates between different devices.

[0029] The present invention further proposes that the establishment of a corneal reference model library in step S300 includes: A first reference corneal model is generated by fitting the surface of the peripheral region with curvature regularity in the three-dimensional mesh of the anterior corneal surface. A second reference corneal model is generated by mirroring the three-dimensional digital model of the patient's contralateral eye's cornea; a model with similar basic corneal parameters to the patient's cornea is matched from a healthy corneal database and used as a third reference corneal model. Step S300 involves selecting a reference corneal model from the corneal reference model library, including: calculating the morphological similarity between the three-dimensional digital corneal model and the first, second, and third reference corneal models respectively using a similarity matching algorithm, and selecting the model with the highest morphological similarity as the reference corneal model.

[0030] The first reference corneal model is derived from a relatively regular peripheral area of ​​the patient's own cornea, typically a region within 6-8 mm of the limbus with gentle curvature and uniform thickness, for surface fitting. This model reflects the natural curvature characteristics and individual differences of the patient's cornea and is suitable when the contralateral eye cannot be used as a reference, such as when both eyes have lesions but the peripheral areas remain relatively normal. The second reference corneal model utilizes the individual characteristics of the contralateral eye of the same patient, obtained by mirroring the three-dimensional digital model of the contralateral eye's cornea along the sagittal plane. For patients with unilateral disease or significant differences in the severity of lesions between the two eyes, the morphology of the contralateral healthy eye is the ideal reference benchmark, preserving the patient's own corneal diameter, overall curvature, and eye-related morphological information. The third reference corneal model is derived from a healthy corneal database containing corneal data from thousands to tens of thousands of healthy individuals of different ages, sexes, and ethnicities. By matching the patient's age, sex, corneal diameter, average curvature, and other basic parameters, the most similar healthy corneal model is selected from the database as a reference. This model can supplement the first two types of reference models when they are insufficient, such as when both eyes have severe lesions and irregular peripheral areas. The similarity matching algorithm can comprehensively compare information such as overall curvature, corneal diameter, peripheral height distribution, and thickness trends, selecting the model that is closest to the patient's basic corneal morphology from multiple reference sources. The reference corneal model here is used to establish the morphological benchmark for the design of the corneal aperture structure. It is not required to be the only normal model in a medical diagnostic sense, as long as it can reflect the normal morphological trend of the patient's cornea.

[0031] As one specific implementation, the first reference corneal model, the second reference corneal model, and the third reference corneal model can be respectively denoted as... , and The three-dimensional digital model of the patient's cornea is denoted as To avoid interference from the lesion area itself in the selection of the reference model, a relatively regular region can be selected from the three-dimensional digital model of the cornea based on curvature continuity and thickness stability as the similarity calculation region, and this region is denoted as... For any reference corneal model ,in For reference corneal model numbering, it can be found The internal calculation shows the height difference between the patient's three-dimensional digital corneal model and the reference corneal model. Curvature difference and thickness difference .in, The root mean square difference between the patient's corneal height and the reference corneal height. The root mean square difference between the patient's corneal curvature value and the reference corneal curvature value. This represents the root mean square difference between the patient's corneal thickness and the reference corneal thickness. The overall difference can be expressed as... ,in , , These are the weighting coefficients, and In practical applications, the weights for height differences are typically set to 0.4, curvature differences to 0.4, and thickness differences to 0.2, to highlight the overall morphological characteristics of the cornea. Furthermore, morphological similarity is defined as... ,in The larger the value, the higher the value. The closer the reference corneal model and the patient's three-dimensional digital corneal model are within a relatively regular area, the more likely the system will select [the appropriate model]. The largest reference corneal model is used as the reference corneal model for subsequent generation of local corneal abnormality scoring maps.

[0032] As an alternative embodiment, when multiple reference corneal models with similar morphological similarity exist in the reference corneal model library, these models can be weighted and fused to form a composite reference corneal model. In this case, the weight of each reference corneal model can be set to be proportional to its morphological similarity, thereby allowing models with higher morphological similarity to contribute more to the composite reference corneal model. This alternative embodiment can be retained as an implementation method in the specification without changing the basic scheme of selecting the model with the highest morphological similarity as the reference corneal model in the claims.

[0033] Through the above technical solution, the reference corneal model no longer relies on a single source, but is selected from three types of models: patient-specific fitting, contralateral spectacle image, and database matching. This reduces the bias in abnormal area judgment caused by a single unsuitable reference model and facilitates the implementation of this method under different clinical data conditions. Even when the contralateral eye already has abnormalities or the database matching is insufficient, those skilled in the art can still obtain a reference model by fitting the patient's own regular regions; if the patient's own regular regions are insufficient, they can be supplemented by the contralateral eye or database models. This reference model library provides a comparative benchmark for local abnormality scoring maps, ensuring that the generation of the target treatment area has traceable data support.

[0034] The present invention further proposes that step S300, which generates a corneal local anomaly scoring map based on a three-dimensional digital model of the cornea and a reference corneal model selected from a corneal reference model library, and determines at least one target treatment area from the corneal local anomaly scoring map, includes: comparing the three-dimensional digital model of the cornea with the selected reference corneal model point by point; calculating the anterior surface curvature abnormality deviation using corneal topographic map data; calculating the corneal thickness abnormality deviation using corneal thickness map data; calculating the posterior surface height abnormality deviation using corneal tomographic data; performing weighted fusion on at least two of the anterior surface curvature abnormality deviation, corneal thickness abnormality deviation, and posterior surface height abnormality deviation to generate a corneal local anomaly scoring map; performing binarization segmentation on the corneal local anomaly scoring map according to a preset scoring threshold; extracting continuous regions with scores higher than the preset scoring threshold as initial target regions; and performing connected component analysis on the initial target regions to determine spatially separated connected components as independent target treatment areas.

[0035] Point-by-point comparison refers to comparing the patient's 3D digital corneal model and a reference corneal model under a unified coordinate system or a unified grid node. First, the two models are precisely registered to ensure that corresponding points correspond one-to-one in spatial location. Anterior surface curvature abnormality deviation reflects the degree of local steepening of the anterior corneal surface, calculated as the difference between the anterior surface curvature of the patient's cornea and the reference corneal model. A larger difference indicates a steeper anterior surface in that area and a more severe abnormality. Corneal thickness abnormality deviation reflects the degree of local thinning, calculated as the difference between the thickness of the corresponding location on the reference corneal model and the thickness of the patient's cornea. A larger difference indicates a thinner cornea in that area, requiring more cross-linking enhancement. Posterior surface height abnormality deviation reflects the local bulging of the posterior surface relative to the reference model, calculated as the difference between the posterior surface height of the patient's cornea and the posterior surface height of the reference corneal model. A larger difference indicates a more pronounced posterior surface bulge and a potentially more severe lesion. During weighted fusion, various deviations can be scaled up first, allowing data of different dimensions to be merged under the same scoring scale. Scale-up typically employs Z-score normalization, converting each deviation value into a standard score with a mean of 0 and a standard deviation of 1. Binarization segmentation converts continuous scoring images into target regions, dividing the corneal region into abnormal and normal areas based on a preset scoring threshold. Connected component analysis distinguishes between single lesion regions and multiple separated lesion regions. By identifying interconnected pixel regions in the image, spatially separated abnormal regions are divided into different target treatment regions, each corresponding to an independent perforation pore.

[0036] As one specific implementation method, when generating a corneal local abnormality scoring map, the third-order corneal local abnormality scoring map can be generated from the three-dimensional digital model of the cornea. i The front surface curvature at each sampling location is denoted as . The anterior surface curvature on the corresponding reference corneal model is denoted as The patient's corneal thickness was recorded as The reference corneal thickness is denoted as The height of the patient's posterior surface was recorded as The height of the referenced surface is recorded as For cases involving keratoconus or localized corneal ectasia, the abnormal deviation in anterior surface curvature can be expressed as: Abnormal deviations in corneal thickness can be expressed as The abnormal deviation in rear surface height can be expressed as: ,in , and These are the scaling coefficients for the corresponding parameters within the effective design region. This indicates that only positive abnormal deviations are considered, because we only focus on corneal steepening, thinning, and posterior surface bulging.

[0037] Local corneal abnormality score can be expressed as ,in For the first i Anomaly scores for each sampling location, , and These are the weighting coefficients, and In the diagnosis of keratoconus, anterior surface curvature and corneal thickness are the two most important indicators, therefore they are usually... and Set it to 0.4. Set it to 0.2. To reduce the impact of local extreme values ​​on the threshold, all... The median is denoted as The absolute median difference is denoted as And set the rating threshold to ,in This is the threshold adjustment coefficient, typically set to 2.0-3.0. When When, the sampling location is included in the initial target area; when At that time, the sampling location is designated as a non-target region. Subsequently, connected component labeling is performed based on the grid connectivity between adjacent sampling points, thereby obtaining one or more spatially separated target treatment regions.

[0038] As an alternative embodiment, if a certain data point is missing or its quality does not meet the usage requirements, the corresponding weight can be set to zero, and the remaining weights can be renormalized. For example, when posterior surface height data is missing, a local corneal anomaly scoring map can be generated using anterior surface curvature anomaly deviation and corneal thickness anomaly deviation. , If the operator wishes to incorporate changes in follow-up progression into the score, a progression abnormality bias item can be added and integrated into the corneal local abnormality scoring map using the same scaling method. Progression abnormality bias can be calculated by comparing the change in the patient's current corneal data with the previous follow-up data, reflecting the rate of disease progression.

[0039] Through the above technical solution, the target treatment area is not determined solely by a single image or a single examination indicator, but rather by fusing multiple corneal morphological information such as anterior surface curvature, thickness, and posterior surface height. Continuous scoring maps can express the spatial distribution of abnormality levels, and binarization segmentation and connected component analysis can transform the scoring results into geometric regions that can be used for aperture design. This process is directly linked to the subsequent generation of the permeability aperture contour; high-scoring continuous regions correspond to the target areas where aperture contours need to be formed, while separated regions can correspond to different permeability apertures, thus making the corneal aperture design more consistent with the patient's abnormal corneal morphology.

[0040] The present invention further proposes that step S300 includes: superimposing the corneal local abnormality scoring map on the corneal three-dimensional digital model or the corneal two-dimensional projection map on the display interface, receiving the operator's closed outline input for the target treatment area, and performing union or intersection fusion of the area defined by the closed outline input and the automatically segmented initial target area to finally determine the target treatment area.

[0041] The display interface utilizes 3D visualization technology, allowing observation of corneal morphology and abnormality score distribution from any angle. The interface can simultaneously display a 3D corneal model, a local corneal abnormality score map, the automatically segmented initial target area, and the operator's outline input. The operator can use a mouse, graphics tablet, or touchscreen to draw closed outlines on the display interface, manually defining the area to be treated. The system converts the operator's lines into closed boundaries on a curved surface in real time and calculates the area enclosed by those boundaries. Union fusion is suitable for cases where automatic segmentation misses some abnormal areas, merging the automatically segmented area and the manually outlined area to ensure all areas requiring treatment are included. Intersection fusion is suitable for cases where automatic segmentation includes some normal areas, retaining only the overlapping portion of the automatically segmented and manually outlined areas, excluding normal areas that do not require treatment. The fused target treatment area retains the objective data basis of the automatic segmentation results and also allows the operator to modify it according to actual design needs, making the final target treatment area more consistent with clinical requirements.

[0042] As one specific implementation method, when merging the automatically segmented region with the operator's closed outline region, the initial target region obtained by automatic segmentation can be denoted as... The area defined by the closed stroke input is denoted as The system first calculates the overlap ratio between the two. ,in Indicates the area of ​​the region. This represents the intersection of two regions. Represents the union of two regions. Overlap ratio. This reflects the degree of consistency between the automatic segmentation results and the operator's judgment. A value closer to 1 indicates higher consistency. If the operator chooses to expand the coverage area, the final target treatment area will be... If the operator chooses conservative coverage, the final target treatment area will be [missing information]. As another fusion method, the boundaries of automatically segmented regions and closed outlined regions can be converted into symbolic distance fields, respectively. and In this system, points within the region are represented by negative values, points outside the region by positive values, and the boundary points by zero values. Then, according to... Generate a fusion distance field, where The fusion weights range from 0 to 1, and the final weight is... The closed curve is used as the boundary of the final target treatment area. This method can create a continuous transition boundary between the automatic segmentation result and the manual outline result, avoiding obvious boundary abrupt changes.

[0043] As an alternative embodiment, closed stroke input can be used directly as the final target treatment area, or it can be used only to correct the local boundaries of the automatically segmented area. If the operator only adjusts part of the boundary, the system can retain most of the boundaries of the automatically segmented area, replacing only the boundary segments modified by the operator with the corresponding closed stroke boundary segments. This reduces the amount of repeated redrawing of the area and maintains the continuity of the target treatment area boundaries. The system also provides undo, redo, and erase editing functions, facilitating the operator's modification and adjustment of the stroke results.

[0044] The above technical solution combines automatic identification with manual verification in determining the target treatment area. Automatic segmentation quickly obtains the initial area corresponding to the anomaly scoring map, while closed outlining incorporates the operator's judgment of the patient's specific condition into the orifice design process. Fusion processing avoids the situation where manual outlining is completely detached from image data, and also avoids the situation where automatic segmentation results cannot adapt to actual operational judgments. The final target treatment area has a clear source of image data and can be corrected and verified by the operator, providing a relatively stable input boundary for the subsequent generation of personalized perforation orifice contours.

[0045] The present invention further proposes that step S400 includes: extracting the closed boundary of the target treatment area on the surface of the three-dimensional digital model of the cornea as the initial contour; expanding the initial contour outward at equal intervals along its normal to obtain the expanded contour; smoothing the expanded contour to eliminate sharp corners and jagged edges on the expanded contour, wherein the smoothing is performed using any one of B-spline curve fitting, Bezier curve fitting, or multi-segment circular arc fitting; and applying manufacturing constraint correction to the smoothed contour, wherein the manufacturing constraint correction includes limiting the minimum local radius of curvature of the contour, limiting the minimum area enclosed by the contour, and limiting the minimum distance from any point on the contour to the peripheral edge of the corneal body model, so as to generate the permeation hole contour.

[0046] The initial contour is derived from the boundary of the target treatment area on the corneal surface, obtained by extracting the boundary line between the target and background areas from the binarized segmentation results. Since automatic segmentation or manual outlining may result in jagged boundaries, the initial contour is often not smooth enough and requires further processing. Equidistant outward expansion is used to geometrically cover the edge of the target treatment area with the aperture contour. The outward expansion direction can be determined based on the local tangent plane of the corneal surface and proceeds along the outer edge of the boundary. The purpose of outward expansion is to compensate for minor displacements that may occur during membrane wear, ensuring that the target treatment area is still completely covered by the permeation aperture even with slight membrane shifts. Smoothing processing converts the jagged boundaries generated in automatic segmentation or manual outlining into a continuous, machinable curve, eliminating sharp corners and irregular details, making the permeation aperture shape more regular and facilitating subsequent processing and use. Manufacturing constraint correction is a crucial step in transforming the medical area boundary into a machinable aperture shape. By constraining the minimum radius of curvature, minimum aperture area, and distance from the aperture to the edge of the cornea, the permeation aperture contour is made suitable for laser, milling, or mold core machining. If the local radius of curvature of the contour is too small, overheating may occur during laser processing, the tool may not be able to reach it during milling, and the mold core may easily break. If the hole area is too small, the penetration of riboflavin solution will be insufficient, affecting the treatment effect. If the hole is too close to the edge of the eye membrane, it will reduce the mechanical strength of the eye membrane and make it easy to break during wearing and removal.

[0047] As one specific implementation method, when expanding the closed boundary of the target treatment area, the target treatment area can be denoted as... The average abnormal score of the target treatment area is recorded as The scoring threshold is denoted as The outward expansion distance can be set to... ,in The base extension is typically 0.2-0.5mm. The outward expansion adjustment coefficient corresponding to the abnormal score is typically taken as 0.1-0.3 mm / score. This means that the more severe the abnormality, the greater the outward expansion distance, resulting in more comprehensive treatment coverage. The initial contour is then adjusted accordingly. i The boundary points are denoted as Let the unit normal vector of this point pointing outward from the target treatment area within the local tangent plane of the corneal surface be denoted as... Then the expanded boundary points are represented as After obtaining the set of points on the outer boundary, a B-spline curve can be used to fit it. The fitted curve is represented as follows: ,in These are curve parameters, with values ​​ranging from 0 to 1. As control points, Let p be the B-spline basis functions. To determine the number of control points, a cubic B-spline curve is typically used for fitting, ensuring the curve's second-order continuity. When correcting the fitted contour for manufacturing constraints, the curvature of the contour at parameter u can be calculated. And express the radius of curvature as .when At that time, the control point was smoothly adjusted, among which... The minimum permissible radius of curvature is set according to the processing method; for laser processing, it is typically 0.1-0.2 mm, and for micro-milling, it is typically 0.2-0.3 mm. The area enclosed by the contour can also be calculated. ,when At that time, the outline is expanded outwards either as a whole or in parts. The minimum area threshold for the processable permeable pores is typically taken as 0.5-1.0 mm². Further, the peripheral edge curve of the eye membrane is denoted as... Calculate the minimum distance from each point on the contour to the surrounding edges. ,when At that time, the contour segments near the peripheral edges are corrected by inward shrinkage, among which... This is the minimum physical distance that needs to be maintained between the edge of the hole and the peripheral edge of the eye membrane, usually taken as 0.5-1.0mm.

[0048] As an alternative embodiment, smoothing can be achieved using Bézier curve fitting or multi-segment circular arc fitting. When using Bézier curve fitting, the control points can be adjusted to make the curve pass through or approach the key points of the outward expansion contour; when using multi-segment circular arc fitting, the outward expansion contour can be decomposed into several arc segments, and adjacent arc segments can be kept tangentially continuous. If the target treatment area is small and the area of ​​the hole after expansion is still below the processable threshold, the system can merge the target treatment area with neighboring areas, or prompt the operator to reconfirm the target treatment area.

[0049] Through the above technical solution, the boundary of the lesion area is transformed into a suitable permeation pore contour for processing, rather than directly using the irregular medical segmentation boundary as the processing path. Outward contour expansion ensures coverage of the target treatment area's edge, smooth curves reduce the impact of sharp corners and serrations on processing quality, and manufacturing constraint corrections prevent production difficulties caused by pores that are too small, have excessively sharp edges, or are too close to the outer edge of the eye. This step combines the clinical target area with industrial manufacturing conditions, ensuring that the resulting permeation pore contour corresponds to the patient's lesion area and can be incorporated into subsequent eye surface mapping and actual processing procedures.

[0050] The present invention further proposes that step S500 includes: using the three-dimensional mesh of the anterior corneal surface as the basic curved surface, offsetting a preset bonding gap along the normal direction of the anterior corneal surface to obtain the inner surface of the eye membrane body model; based on the inner surface, offsetting a preset thickness outward along the normal direction to obtain the outer surface of the eye membrane body model; and smoothly connecting the boundaries of the inner surface and the outer surface to form the peripheral edge.

[0051] The ocular body model serves to support the subsequently generated perforation aperture. Its overall shape matches the anterior surface of the patient's cornea, ensuring a tight fit during wear. A pre-set fit gap is maintained between the inner surface and the anterior corneal surface, allowing the ocular body to cover the corneal surface in a form similar to a contact lens or therapeutic ocular body. The size of the pre-set fit gap needs to be moderate; too small a gap will result in insufficient tear film between the ocular body and the cornea, affecting wearing comfort and the flow of riboflavin solution; too large a gap will lead to unstable wearing of the ocular body, making it prone to displacement and affecting the alignment accuracy of the perforation aperture with the target treatment area. The outer surface is formed by offsetting the inner surface along the normal direction, giving the ocular body a certain thickness. The thickness of the ocular body needs to strike a balance between mechanical strength and flexibility; too thin a thickness will make the ocular body prone to breakage, while too thick a thickness will reduce the flexibility of the ocular body, affecting the fit. The peripheral edge connects the inner and outer surfaces, and the edge should avoid forming sharp transitions to facilitate subsequent wearing, removal, and processing. Sharp edges can not only irritate the surface tissues of the eye and cause discomfort when worn, but are also prone to chips and cracks during processing and transportation.

[0052] As a specific implementation method, when generating the corneal body model, the three-dimensional mesh of the anterior corneal surface can be represented as follows: Where u and v are the parametric coordinates of the corneal surface, the unit normal vector of the anterior corneal surface at the corresponding position is expressed as... The preset fitting gap is expressed as The inner surface of the ocular body model is represented as... ,in These are points on the inner surface of the eyelid. If a uniform value is used for the adhesion gap, then... Use a constant amount, typically 0.02-0.05mm; if local adjustments are needed for different areas of the cornea, then... The gap is set as a function that varies with the position of the curved surface; for example, the gap is slightly smaller in the central region of the cornea and slightly larger in the peripheral region to improve wearing stability. Furthermore, the corneal thickness is expressed as... The unit normal vector of the inner surface of the eye membrane at the corresponding position is expressed as The outer surface of the eye membrane is represented as ,in These are points on the outer surface of the eye membrane. The thickness of the eye membrane is typically a uniform value of 0.2-0.3 mm, but can be adjusted locally as needed. For the peripheral edges, the inner surface boundary curve can be used. and outer surface boundary curve Generate transition surfaces between them Where s represents the parameter along the edge direction, and r represents the parameter that transitions from the inner surface boundary to the outer surface boundary, with a value ranging from 0 to 1. It is a smooth transition function, and , In one implementation, Can be taken This is to make the boundary transition more continuous and avoid obvious sharp corners.

[0053] As an alternative embodiment, the corneal body model can also be established using a combination of curved surfaces commonly used in contact lens manufacturing, comprising a base curve region, a transition region, and a peripheral region. In this case, the central region can be generated based on a three-dimensional mesh offset from the anterior corneal surface, and the peripheral region can be generated based on a preset edge shape. The central region and the peripheral region are connected by a continuous transition surface. This alternative embodiment still retains the basic structural relationships of the inner surface, outer surface, and peripheral edges of the cornea, while better adapting to the process requirements of contact lens manufacturing equipment.

[0054] Through the above technical solution, the ocular membrane model is directly derived from the anterior surface data of the patient's cornea, thus matching the overall curvature with the patient's corneal morphology. The creation of the inner surface, outer surface, and peripheral edges transforms the ocular membrane from a two-dimensional perforated pattern into a three-dimensional solid model with thickness and curved surface structure. This model provides a carrier for the subsequent generation of through-holes and also provides the geometric basis for manufacturing methods such as turning, mold forming, or 3D printing. Because the peripheral edges are smoothly transitioned, the resulting ocular membrane sheet is less prone to local interference caused by abrupt edge changes during handling and application.

[0055] The present invention further proposes that step S600 includes: establishing a coordinate correspondence based on a preset mapping relationship between the corneal vertex coordinate system of the three-dimensional digital model of the cornea and the geometric center coordinate system of the ocular body model; projecting the closed curve of the permeation pore outline along the local normal direction of the cornea onto the inner surface of the ocular body model to form an inner surface pore boundary curve; extending the inner surface pore boundary curve along the thickness direction of the ocular body model to the outer surface to form an outer surface pore boundary curve on the outer surface; generating a continuous surface between the inner surface pore boundary curve and the outer surface pore boundary curve as a pore wall, and the pore wall together with the opening on the inner surface and the opening on the outer surface constitute the permeation pore.

[0056] The corneal vertex coordinate system is used to express the position of the target treatment area on the corneal surface, while the geometric center coordinate system of the ocular body model is used to express the geometric position in the ocular processing model. Preset mapping relationships can include translation, rotation, and surface correspondence relationships, enabling accurate transfer of the permeation aperture contour from the corneal model to the ocular model. Typically, the corneal vertex is aligned with the ocular geometric center, and the optical axis of the cornea is aligned with the central axis of the ocular model to ensure overall positional correspondence. Local normal projection is used to maintain the surface correspondence between the permeation aperture contour and the corneal lesion area. Directly projecting the corneal surface contour onto a plane and then mapping it onto the ocular surface will result in significant shape and positional distortions, while projection along the local normal direction of the cornea can preserve the original shape of the contour to the greatest extent. The aperture wall is a continuous curved surface connecting the inner and outer surface openings. Its shape can be generated along the thickness direction or set as a surface with a slight taper depending on the processing method. A slightly taper aperture wall is more conducive to demolding and drug flow; the taper is typically 1-3 degrees.

[0057] As one specific implementation, when mapping the permeation pore contour from the corneal 3D digital model to the ocular body model, any boundary point on the permeation pore contour can be denoted as... ,in The closed curve generated in step S400 is represented by u, where u is the curve parameter. This point is transformed to the coordinate system of the eye membrane model using a preset coordinate transformation relationship, expressed as follows: ,in For the transformed point, Let be the rotation matrix from the corneal vertex coordinate system to the corneal geometric center coordinate system. This is the translation vector. In an ideal situation, It is the identity matrix. The zero vector is defined as the coordinate system completely coinciding with the zero vector. Subsequently, using... Starting from the cornea, establish a projection ray along the corresponding local normal direction. ,in For ray parameters, The local normal vector of the cornea, ray With the inner surface of the eye membrane The intersection point is used as the boundary point of the inner surface hole. After all boundary points have been projected, the boundary curve of the inner surface hole is obtained. Then, generate the surface boundary curve along the thickness direction of the eye membrane. ,in This represents the thickness of the cornea at the corresponding boundary point. Let be the unit normal vector of the inner surface of the eye membrane at that location. The curved surface of the aperture wall is represented as... Where r is a parameter representing the transition from the inner surface hole boundary to the outer surface hole boundary, with a value ranging from zero to one. If a chamfer or fillet is required, it can be applied in the region where r is close to zero or one. Local smoothing is performed to create a continuous transition at the junction of the pore wall and the inner and outer surfaces of the eye membrane, avoiding sharp edges.

[0058] As an alternative embodiment, the permeation pore contour can also be mapped to the ocular body model using surface parametric coordinates. Specifically, the boundary points of the target treatment area are represented as parametric coordinates in the three-dimensional digital model of the cornea. Then, points with the same or corresponding parameter coordinates are read from the inner surface of the ocular body model as the boundary points of the inner surface holes. This method is suitable for cases where the inner surface of the ocular body is generated by offsetting the anterior surface of the cornea and both have the same parameter domain. It is faster to calculate and avoids the problem of non-unique projection intersection points. For regions where local normal projection may produce non-unique intersection points, parameter coordinate mapping can be used preferentially.

[0059] The above technical solution avoids simply creating perforations in a planar pattern. Instead, it generates perforations by establishing a coordinate correspondence between the curved surfaces of the cornea and the eye. The inner surface perforation boundary curve corresponds to the target treatment area on the cornea, while the outer surface perforation boundary curve corresponds to the outer opening on the eye. The perforation walls connect the two to form a true three-dimensional through-structure. This processing method reduces positional and shape deviations caused by directly copying two-dimensional image contours onto the curved eye, and also gives the perforations a clear three-dimensional geometric definition, facilitating output as a solid model, processing path, or quality inspection standard.

[0060] The present invention further proposes that step S700 includes: converting the ocular membrane body model with permeable pores into a manufacturing model file containing three-dimensional surface information and pore boundary information, wherein the manufacturing model file is in STL format, STEP format, etc. I The manufacturing model file is input into a CNC machining equipment and manufactured into an arc-shaped eyepiece with permeable pores by any of the following methods: laser drilling after turning, micro-milling drilling after turning, mold forming, or casting after 3D printing mold, to obtain a personalized eyepiece.

[0061] The manufacturing model file uses different formats depending on the processing method. STL format is currently the most commonly used file format in the 3D printing and rapid prototyping field. It represents 3D solid models using triangular facets and is suitable for 3D printing molds or 3D contour detection. STEP format and... I GES format is an internationally recognized product data exchange standard that can completely preserve the geometric information and topological relationships of 3D models. It is suitable for general 3D CAD / CAM systems, facilitating subsequent editing and processing programming. DXF hole contour format is a 2D vector graphics format suitable for extracting 2D or unfolded hole contours for use by laser cutting or drilling equipment. CNC turning code is the instruction code that directly controls the movement of a CNC lathe, used to form the inner and outer curved surfaces of the eye membrane. Laser drilling path code is the instruction code that directly controls the movement of laser processing equipment and the laser switch, used to process through holes according to the permeation hole contour. Laser drilling after turning is a commonly used process in the contact lens manufacturing field. First, a non-porous eye membrane substrate is machined by CNC turning, and then permeation holes are machined on the substrate using a laser. This method has high precision and is suitable for small-batch personalized production. Micro-milling drilling after turning is suitable for hole shapes with complex boundaries or requiring physical tool finishing, achieving better hole wall quality. Mold forming and 3D printing followed by casting are suitable for batch or semi-batch production of eye membrane sheets based on manufacturing models, reducing unit production costs.

[0062] As a specific implementation method, when generating the manufacturing model file, the permeation hole boundary can be compensated for according to different processing methods. Since the processing tool has a certain size, the actual processed hole shape will be larger than the designed hole shape by one tool size; therefore, pre-compensation is necessary. The designed external surface hole boundary curve is denoted as... The equivalent radius of action of the machining tool is denoted as Let the unit normal vector of the hole boundary pointing towards the solid material of the eye membrane within the local machining plane be denoted as... When the machining path is the tool centerline or the spot centerline, and the machining equipment removes material along the inside of the hole, the compensation curve used to generate the machining path is represented as follows: ,in This represents the actual path curve executed by the processing equipment. For laser drilling, This corresponds to an equivalent radius of the laser spot, typically 0.05-0.1mm; for micro-milling openings, This corresponds to a tool radius, typically 0.1-0.2 mm. After manufacturing, the actual hole boundary curve can be obtained through contour scanning. And calculate the maximum hole shape deviation. and root mean square hole shape deviation Where N is the number of sampling points, Let be the q-th sampling parameter. When... and If all deviations do not exceed the preset manufacturing tolerance, the hole shape test is passed; if any deviation exceeds the preset manufacturing tolerance, the processing compensation curve is adjusted or the processing path is regenerated according to the direction of the deviation. Typically, the maximum hole shape deviation is required to be no more than 0.1 mm, and the root mean square hole shape deviation is required to be no more than 0.05 mm.

[0063] As an alternative embodiment, when preparing personalized eye masks using a mold-forming method, a positive mold and a negative mold can be generated based on the eye mask body model, and a core for forming holes corresponding to the permeation holes can be set in the mold. After molding, an arc-shaped eye mask sheet with through permeation holes is formed by demolding. If a 3D printing mold followed by casting molding method is used, the eye mask body model can be reverse-converted into a mold model, and the hole boundary and hole wall corresponding structures can be retained in the mold model. For the turning molding followed by laser drilling method, the inner and outer surfaces of the eye mask can be machined first, and then the through holes can be formed according to the laser drilling path code.

[0064] Through the aforementioned technical solution, the 3D modeling results are converted into manufacturing files that can be recognized by processing equipment, eliminating the need for manually redrawing the pore shapes in the production of personalized eye masks. Different file formats correspond to different processing techniques, enabling this method to be adapted to existing contact lens processing, laser drilling, micro-milling, or mold manufacturing equipment. Since the manufacturing model simultaneously contains information on the eye mask surface and pore boundaries, the production end can process the eye mask body and permeable pores using the same data source, reducing repetitive positioning between surface forming and pore processing. This step transforms the personalized eye mask from a design result into a practical device, meeting the requirements for output, processability, and testability necessary for industrial-scale production.

[0065] The present invention further proposes that the method also includes: generating a directional positioning structure on the outer surface or peripheral edge of the ocular body model, wherein the directional positioning structure is at least one of a laser marking at the 12 o'clock position, a nasal edge notch, a temporal edge mark, or an asymmetrical edge contour, to indicate the orientation of the permeation hole relative to the corneal abnormal area when worn.

[0066] The orientation positioning structure is used to identify the placement orientation of the eyepiece during clinical use. Since the perforation contour is typically generated based on the patient's corneal abnormality area, its shape and position may be eccentric or directional. If the eyepiece rotates during wear, the correspondence between the perforation and the target treatment area may be affected, leading to insufficient treatment of some lesion areas or unnecessary cross-linking of healthy areas. A laser marking at the 12 o'clock position can serve as a superior positioning reference; the marking is aligned with the patient's 12 o'clock position during wear. Nasal edge notches and temporal edge markings can be used to distinguish the nasotemporal direction; the notch or marking is aligned with the patient's nasal or temporal side during wear. Asymmetrical edge contours can indicate the wearing direction through shape differences; for example, designing the nasal edge of the eyepiece to be slightly flatter than the temporal edge allows the operator to visually determine the correct wearing direction. As an alternative embodiment, the orientation positioning structure can be generated in conjunction with manufacturing quality control standards. Specifically, the orientation positioning structure can be used as one of the coordinate references in the manufacturing model, detecting the relative positional relationship between the geometric center of the perforation, the perforation boundary, and the orientation positioning structure during quality control. If the angular deviation between the actual manufactured orientation structure and the permeation hole exceeds the preset allowable error, the orientation marking of the eyepiece can be determined to not meet the design requirements. Typically, the angular deviation is required to be no more than 5 degrees.

[0067] Through the above technical solution, the personalized eyepiece not only has permeation holes corresponding to the patient's lesion area, but also structural features for identifying the wearing direction. The orientation positioning structure extends the coordinate relationships in the manufacturing model to the actual device use, enabling the operator to determine the orientation of the permeation holes relative to the corneal abnormality area when placing the eyepiece. This structure does not change the basic formation logic of the permeation holes, but reduces the risk of hole misalignment due to unclear rotation direction identification, and provides a visible benchmark for alignment confirmation in subsequent clinical use.

[0068] In practice, to more intuitively illustrate the generation process of the personalized eye membrane of this invention, a three-dimensional digital model of the patient's cornea can be established based on examination data such as corneal topography, corneal tomography, OCT images, or corneal thickness maps. See also... Figure 2 In this embodiment, a three-dimensional curved surface model of the cornea is first constructed based on the anterior and posterior surface data of the patient's cornea, and corneal thickness distribution information is superimposed on the model. Figure 2 (a) shows the spatial morphological relationship between the anterior and posterior surfaces of the patient’s cornea, which can reflect structural changes such as local corneal anterior surface bulge and posterior surface anterior convexity; Figure 2 (b) The corneal thickness distribution is further superimposed on the anterior corneal surface to visually identify areas of localized corneal thinning. This three-dimensional digital model allows for the acquisition of the required surface parameters for the inner surface of the cornea, ensuring that the final manufactured cornea adheres well to the patient's corneal surface.

[0069] After obtaining a three-dimensional digital model of the patient's cornea, further quantitative analysis was performed on local abnormalities of the cornea. (See also...) Figure 3 This embodiment calculates a local corneal abnormality score based on parameters such as changes in anterior corneal curvature, corneal thickness, and posterior corneal height. Specifically, the height difference of the patient's anterior corneal surface relative to a reference corneal model, the increase in local curvature, the decrease in corneal thickness, and the posterior corneal bulge are normalized separately and then fused according to preset weights to form a local corneal abnormality score map. A higher score indicates a more significant corneal morphological abnormality or treatment need in that area. By setting an abnormality score threshold, the boundary of the target treatment area can be automatically extracted from the score map, thus providing a basis for determining the location, shape, and size of the permeation hole.

[0070] See Figure 4 After determining the target treatment area, this embodiment does not simply set through-holes of fixed positions or sizes on the cornea. Instead, it determines the contours of multiple perforations based on a local anomaly scoring map. For areas with high anomaly scores and large areas, larger elliptical or near-elliptical perforations can be generated; for areas with low anomaly scores or small areas, smaller perforations can be generated. The center position, major axis, minor axis, and rotation angle of each perforation can be determined according to the position, shape, and direction of the corresponding anomaly area. To improve manufacturing feasibility and wearing safety, the initial treatment core area can be expanded, smoothed, and edge-constrained to match the perforation contours with the corneal anomaly area, while preventing the perforations from being too close to the outer edge of the cornea. As a result, the cross-linked drug solution or light-treated area can more accurately correspond to the abnormal area of ​​the patient's cornea.

[0071] Furthermore, based on the patient's three-dimensional digital corneal model and pore contour data, a final personalized eyepiece manufacturing model is generated. See also Figure 5 In this embodiment, the anterior surface of the patient's cornea is used as a base, a predetermined fitting gap is set on its outer side, and the inner and outer surfaces of the cornea are generated according to the preset corneal thickness; simultaneously, Figure 4 The obtained pore contours are projected onto the curved surface of the eye membrane to form a pore structure that runs through the inner and outer surfaces of the eye membrane. Figure 5 The pore wall structure shown represents the penetration boundary of each permeation pore in the thickness direction of the ocular membrane, ensuring that medication, oxygen, or cross-linking aids enter the target treatment area through the corresponding pore region. Continuous edge sidewalls are also formed around the periphery of the ocular membrane to ensure the overall structural strength and edge stability of the membrane.

[0072] To facilitate clinical wearing and manufacturing positioning, this embodiment can also set directional positioning marks in a preset direction on the membrane, such as positioning marks at the 12 o'clock position. These positioning marks help doctors quickly determine the orientation of the membrane during wear, preventing rotational misalignment between the permeation area and abnormal areas of the patient's cornea. Through continuous data transfer between the patient's three-dimensional digital corneal model, corneal local abnormality scoring map, permeation area contour map, and the final three-dimensional manufacturing model, this invention can achieve automated or semi-automated generation of personalized membrane structural parameters from patient corneal examination data.

[0073] See further Figure 6 Because the location, number, extent and morphology of abnormal corneal areas vary among different patients, the resulting corneal permeability pore structure can also be different. Figure 6 Examples of various personalized ocular membrane structures are illustrated, including single large permeable pores, multi-porous distributed permeable pores, lattice-type permeable pores, and radial or strip-shaped permeable pores. All of these structures can be generated using the same technical process: determining the ocular membrane adhesion surface through a three-dimensional digital model of the patient's cornea, and determining the shape, location, and number of permeable pores based on corneal local abnormality scoring results.

[0074] The overall working principle of this invention is as follows: First, a three-dimensional digital model of the patient's cornea is established using data such as corneal topography, corneal tomography, and corneal thickness mapping. Then, a reference corneal model is used to score local abnormalities in the patient's cornea to determine the target treatment area. Subsequently, the target treatment area is converted from the medical imaging area into a single closed-curve permeable pore outline, and this outline is made feasible for processing through expansion, smoothing, and manufacturing constraint correction. Next, an ocular body model is established based on the anterior surface of the patient's cornea, and the permeable pore outline is mapped from the corneal surface to the ocular surface, generating a pore structure that runs through the inner and outer surfaces of the ocular membrane. Finally, a manufacturing file is generated based on the perforated ocular body model and the membrane is processed. This process forms a continuous conversion relationship between patient corneal data and a personalized ocular body.

[0075] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for preparing a personalized ocular membrane for corneal cross-linking surgery, characterized in that, Includes the following steps: Step S100: Acquire medical imaging data of the patient's cornea, wherein the medical imaging data includes at least corneal topography data, corneal tomography data, and corneal thickness data; Step S200: Preprocess the medical image data and establish a three-dimensional digital model of the patient's cornea based on the preprocessed data. The three-dimensional digital model of the cornea includes a three-dimensional mesh of the anterior surface of the cornea, a three-dimensional mesh of the posterior surface of the cornea, and a corneal thickness distribution determined by the three-dimensional mesh of the anterior surface of the cornea and the three-dimensional mesh of the posterior surface of the cornea. Step S300: Establish a corneal reference model library, generate a corneal local abnormality scoring map based on the corneal three-dimensional digital model and a reference corneal model selected from the corneal reference model library, and determine at least one target treatment area from the corneal local abnormality scoring map; Step S400: Perform contour extraction, expansion and contour smoothing on each target treatment area to generate a permeation pore contour corresponding to the shape of the target treatment area. The permeation pore contour is a single closed curve. Step S500: Based on the corneal anterior surface data in the three-dimensional digital model of the cornea, establish an ocular body model that matches the curvature of the patient's cornea. The ocular body model is an arc-shaped sheet structure with an inner surface, an outer surface, and a peripheral edge. Step S600: Map the permeation pore outline from the corneal 3D digital model to the eye membrane body model, and generate a permeation pore penetrating the inner surface and the outer surface in the eye membrane body model; Step S700: Generate a manufacturing model based on the ocular body model with the permeation pores, and use the manufacturing model to prepare a personalized ocular membrane for corneal cross-linking surgery.

2. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 1, characterized in that, Step S200 includes: Remove abnormal data points from the medical imaging data caused by eye movement or tear film instability; Image data from different acquisition devices are unified into the same coordinate system through spatial registration with the corneal apex and pupil center as reference markers, and the orientation is corrected with the corneal apex or pupil center as the coordinate origin. The corneal topography data is used to fit and generate a three-dimensional mesh of the anterior corneal surface, and the corneal tomography data is used to fit and generate a three-dimensional mesh of the posterior corneal surface. The corneal thickness distribution is calculated based on the corneal thickness map data, or based on the sag difference between corresponding points on the three-dimensional grid of the anterior corneal surface and the three-dimensional grid of the posterior corneal surface.

3. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 1, characterized in that, The step S300 of establishing a corneal reference model library includes: A first reference corneal model is generated by fitting the surface of the peripheral region with regular curvature in the three-dimensional mesh of the anterior corneal surface. A second reference corneal model is generated by mirroring the three-dimensional digital model of the patient's contralateral eye's cornea. A model that closely matches the patient's corneal basic parameters from a healthy corneal database is used as a third reference corneal model. Step S300, which involves selecting a reference corneal model from the corneal reference model library, includes: calculating the morphological similarity between the three-dimensional digital corneal model and the first reference corneal model, the second reference corneal model, and the third reference corneal model using a similarity matching algorithm, and selecting the model with the highest morphological similarity as the reference corneal model.

4. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 3, characterized in that, Step S300 involves generating a corneal local anomaly scoring map based on the three-dimensional digital model of the cornea and a reference corneal model selected from the corneal reference model library, and determining at least one target treatment area from the corneal local anomaly scoring map, including: The three-dimensional digital model of the cornea is compared point by point with the selected reference corneal model. The anterior surface curvature abnormality deviation is calculated using the corneal topography data, the corneal thickness abnormality deviation is calculated using the corneal thickness map data, and the posterior surface height abnormality deviation is calculated using the corneal tomography data. At least two of the abnormal deviations in anterior surface curvature, corneal thickness, and posterior surface height are weighted and fused to generate a local corneal abnormality scoring map. The corneal local abnormality scoring map is binarized and segmented according to a preset scoring threshold. Continuous regions with scores higher than the preset scoring threshold are extracted as initial target regions. Connectivity analysis is then performed on the initial target regions, and spatially separated connected regions are determined as independent target treatment regions.

5. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 4, characterized in that, Step S300 further includes: displaying the corneal local abnormality scoring map superimposed on the corneal three-dimensional digital model or corneal two-dimensional projection map on the display interface; receiving the operator's closed outline input for the target treatment area; and performing union or intersection fusion of the area defined by the closed outline input with the automatically segmented initial target area to finally determine the target treatment area.

6. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 1, characterized in that, Step S400 includes: The closed boundary of the target treatment area on the surface of the three-dimensional digital model of the cornea is extracted as the initial contour; The initial contour is expanded outward at equal intervals along its normal to obtain the expanded contour. The outer contour is smoothed to eliminate sharp corners and jagged edges. The smoothing process is performed using any one of B-spline curve fitting, Bezier curve fitting, or multi-segment circular arc fitting. Manufacturing constraints are applied to the smoothed contour, including limiting the minimum local radius of curvature of the contour, limiting the minimum area enclosed by the contour, and limiting the minimum distance from any point on the contour to the peripheral edge of the ocular body model, to generate the permeation pore contour.

7. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 1, characterized in that, Step S500 includes: Using the three-dimensional mesh of the anterior corneal surface as the base surface, a preset fitting gap is offset along the normal direction of the anterior corneal surface to obtain the inner surface of the eye membrane body model. Based on the inner surface, a preset thickness is offset outward along the normal direction to obtain the outer surface of the eye membrane body model; The boundaries between the inner surface and the outer surface are smoothly connected to form the peripheral edge.

8. The method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 1, characterized in that, Step S600 includes: A coordinate correspondence is established based on a preset mapping relationship between the corneal vertex coordinate system of the three-dimensional digital model of the cornea and the geometric center coordinate system of the corneal body model; The closed curve of the permeation pore outline is projected along the local normal direction of the cornea onto the inner surface of the eye membrane body model to form the inner surface pore boundary curve; The inner surface hole boundary curve is extended along the thickness direction of the eye membrane body model to the outer surface, forming an outer surface hole boundary curve on the outer surface. A continuous curved surface is generated between the inner surface hole boundary curve and the outer surface hole boundary curve to serve as the hole wall. The hole wall, the opening on the inner surface, and the opening on the outer surface together constitute the permeation hole.

9. A method for preparing a personalized ocular membrane for corneal cross-linking surgery according to claim 1, characterized in that, Step S700 includes: The eye membrane body model with the permeation pores is converted into a manufacturing model file containing three-dimensional surface information and pore boundary information. The manufacturing model file is at least one of the following formats: STL, STEP, IGES, DXF pore contour, CNC turning code, or laser drilling path code. The manufacturing model file is input into a CNC machining equipment, and an arc-shaped eye membrane with the permeation holes is manufactured by any of the following methods: turning and forming followed by laser drilling, turning and forming followed by micro-milling drilling, mold forming, or 3D printing mold followed by casting, thereby obtaining the personalized eye membrane.

10. A method for preparing a personalized ocular membrane for corneal cross-linking surgery according to any one of claims 1 to 9, characterized in that, The method further includes generating a directional positioning structure on the outer surface or peripheral edge of the ocular body model, wherein the directional positioning structure is at least one of a laser marking at the 12 o'clock position, a nasal edge notch, a temporal edge mark, or an asymmetrical edge contour, to indicate the orientation of the permeation hole relative to the corneal abnormality area when worn.

Images