Cornea surface shape construction method and system
By combining optical coherence tomography (OCT) with corneal anatomical structure segmentation and Gaussian smoothing algorithm, the high-precision problem of corneal surface shape construction in existing technologies has been solved, achieving accurate construction of corneal surface shape and improving feature point extraction and model accuracy.
Patent Information
- Application Number
- CN202511649090.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-11-12
- Publication Date
- 2026-02-06
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
Existing technologies cannot meet the high-precision requirements when constructing corneal surface shapes. They do not take into account the corneal anatomy for targeted division, making it difficult to accurately reflect the morphological characteristics of different regions. They also fail to distinguish invalid data such as noise and motion artifacts, and image parameter adjustments rely on human experience, resulting in limited accuracy in modeling and smoothing.
Raw scan data is acquired using optical coherence tomography (OCT) equipment, divided according to the preset corneal anatomical structure, invalid data points are removed, valid data are marked based on the spatial coordinate system, image parameters are adjusted, region segmentation and feature recognition are performed, and a three-dimensional mesh model is constructed using triangular mesh generation algorithm and Gaussian smoothing algorithm.
It achieves high-precision construction of corneal surface shape, ensures the anatomical relevance of data, improves the accuracy of feature point extraction and model precision, eliminates irregular protrusions or depressions, and obtains a surface shape that conforms to the actual physiological morphology of the cornea.
Smart Images

Figure CN121482329A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of constructing corneal surface shapes, and more particularly to a method and system for constructing corneal surface shapes. Background Technology
[0002] In ophthalmological clinical diagnosis and treatment, the development and adaptation of customized corneal devices, and research on corneal biomechanical properties, the accuracy of the three-dimensional morphology of the corneal surface directly affects the accuracy of treatment plans, the fit of corrective devices, and the reliability of research data. Especially when there are subtle abnormalities in the corneal morphology, it is necessary to accurately capture the curvature trend and local details of the corneal surface, thus placing high demands on the precision and stability of corneal surface shape construction methods.
[0003] Currently, existing technologies typically involve scanning the corneal surface with ordinary optical scanning equipment to obtain raw image data containing corneal morphology information. The raw image data is then subjected to simple invalid data filtering to obtain pre-purified image data. Based on operator experience, parameters such as brightness and contrast are adjusted on the pre-purified image data to obtain a corneal image with acceptable visual clarity. Edge detection tools are used to extract feature points from the corneal image, and the two-dimensional coordinates of these feature points are recorded and converted to three-dimensional coordinates using empirical formulas. Based on the converted three-dimensional coordinates, a corneal surface model is constructed using a simple mesh stitching method, and the model surface is smoothed using mean filtering to finally obtain the corneal surface shape.
[0004] However, existing technologies have several core drawbacks, the most significant being: First, the raw scan data is not specifically segmented based on corneal anatomy, resulting in subsequent corneal data processing lacking anatomical specificity and failing to accurately reflect the morphological characteristics of different corneal regions. Second, the processing of invalid data is limited to basic filtering, failing to distinguish between different types of invalid data such as noise, motion artifacts, and signal loss, easily leaving invalid information that affects the accuracy of subsequent feature extraction. Third, image parameter adjustment relies on manual experience and does not incorporate the spatial location information of corneal tomographic data, making it impossible to guarantee the matching degree between image parameters and the actual corneal morphology. Fourth, the modeling uses simple mesh stitching, and the smoothing process is limited to basic mean filtering, making it difficult to balance the accuracy of local details and overall morphology of the corneal surface. The final constructed corneal surface shape deviates significantly from the actual corneal morphology, failing to meet the requirements of high-precision scenarios. Summary of the Invention
[0005] The purpose of this application is to provide a method and system for constructing corneal surface shapes to solve the problem that existing technologies cannot meet the requirements of high-precision scenarios when constructing corneal surface shapes.
[0006] To address the aforementioned technical problems, in a first aspect, this application provides a method for constructing a corneal surface shape, comprising: The cornea is scanned using an optical coherence tomography (OCT) device to obtain raw corneal scan data. The raw corneal scan data is then divided and processed according to a preset corneal anatomical structure to obtain corneal tomographic scan data. Invalid data points caused by noise, motion artifacts, or signal loss are removed from the corneal tomography data to obtain valid corneal tomography data. Based on the spatial coordinate system of the optical coherence tomography device, the spatial position of each valid data point in the valid corneal tomography data is marked to obtain valid corneal tomography data with position information. Preliminary corneal optical image data is acquired. Based on the effective corneal tomographic scan data with location information, the image parameters of the preliminary corneal optical image data are adjusted to obtain the final corneal optical image data. The final corneal optical image data is processed by region segmentation to obtain a corneal image. Feature recognition is performed on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea. Based on the three-dimensional coordinates, a three-dimensional model of the corneal surface is generated using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then smoothed using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
[0007] Optionally, based on the three-dimensional coordinates, a three-dimensional model of the corneal surface is created using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then smoothed using a Gaussian smoothing algorithm to obtain the shape of the corneal surface, including: Distance calculation is performed on all the three-dimensional coordinates to obtain the distance values between each feature point. The range of the distance values is then defined to obtain the criteria for determining adjacent feature points. Based on the aforementioned criteria, a three-dimensional model of the corneal surface is obtained by using a triangular mesh generation algorithm. The three-dimensional mesh model is subjected to shape detection processing to mark local areas with protrusions or depressions in the mesh, thereby obtaining the area to be smoothed; The Gaussian smoothing algorithm is used to perform local adjustment processing on the region to be smoothed so that the feature point positions of the region to be smoothed are closer to the feature point positions of the surrounding regions, thus obtaining a locally adjusted mesh model. The locally adjusted mesh model is fused to obtain an overall mesh model. The overall mesh model is then fitted with a surface to obtain the corneal surface shape.
[0008] Optionally, the step of performing local adjustment processing on the region to be smoothed using a Gaussian smoothing algorithm, so that the feature point positions of the region to be smoothed approach the feature point positions of the surrounding regions, to obtain a locally adjusted mesh model, includes: Boundary recognition processing is performed on the region to be smoothed to obtain the edge feature points of the region to be smoothed. Feature point selection processing is performed on the surrounding region outside the edge feature points to obtain the surrounding reference feature points. The positional relationship between the feature points in the region to be smoothed and the surrounding reference feature points is calculated using the Gaussian smoothing algorithm. The vector features between the feature points and each of the surrounding reference feature points are obtained. All the vector features are integrated to obtain the positional association information of each feature point. The location association information is divided into ranges to obtain the adjustment weight value of each feature point. Based on the location association information and the adjustment weight value, the feature points are moved to obtain moved feature points. The moved feature points are then connected to obtain a locally adjusted mesh model.
[0009] Secondly, this application provides a system for constructing corneal surface shapes, comprising: The segmentation module is used to scan the cornea using an optical coherence tomography (OCT) device to obtain raw corneal scan data, and to segment the raw corneal scan data according to a preset corneal anatomical structure to obtain corneal tomographic scan data. The marking module is used to remove invalid data points in the corneal tomography data caused by noise, motion artifacts or signal loss, so as to obtain valid corneal tomography data. Based on the spatial coordinate system of the optical coherence tomography device, the module marks the spatial position of each valid data point in the valid corneal tomography data to obtain valid corneal tomography data with position information. The adjustment module is used to acquire preliminary corneal optical image data, and based on the effective corneal tomographic scan data with position information, adjust the image parameters of the preliminary corneal optical image data to obtain the final corneal optical image data. The recognition module is used to perform region segmentation processing on the final corneal optical image data to obtain a corneal image, and to perform feature recognition on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea; The processing module is used to perform three-dimensional modeling of the corneal surface based on the three-dimensional coordinates using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface, and to smooth the three-dimensional mesh model using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
[0010] Thirdly, this application provides an electronic device, comprising: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the method for constructing the corneal surface shape as described in the first aspect above.
[0011] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, can implement the steps of the method for constructing the corneal surface shape as described in the first aspect above.
[0012] The method for constructing the corneal surface shape provided in this application involves scanning the cornea using an optical coherence tomography (OCT) device to obtain raw corneal scan data. This raw corneal scan data is then divided and processed according to a preset corneal anatomical structure to obtain corneal tomographic scan data. Invalid data points caused by noise, motion artifacts, or signal loss are removed from the corneal tomographic scan data to obtain valid corneal tomographic scan data. Based on the spatial coordinate system of the OCT device, the spatial positions of each valid data point in the valid corneal tomographic scan data are marked to obtain valid corneal tomographic scan data with positional information. Preliminary corneal optical image data is acquired. Based on the valid corneal tomographic scan data with location information, the image parameters of the preliminary corneal optical image data are adjusted to obtain final corneal optical image data. The final corneal optical image data is then processed by region segmentation to obtain a corneal image. Feature recognition is performed on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea. Based on the three-dimensional coordinates, a three-dimensional model of the corneal surface is generated using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then smoothed using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
[0013] The technical solution of this application has the following beneficial effects: This application utilizes raw corneal scan data acquired through optical coherence tomography (OCT) equipment, which boasts high resolution. By dividing the data according to predefined corneal anatomical structures, the obtained corneal tomographic data accurately corresponds to different anatomical regions of the cornea, providing a targeted anatomical basis for subsequent precise data processing of specific corneal areas. Targeted removal of three types of invalid data points—noise, motion artifacts, and missing signals—directly improves the purity and reliability of the corneal tomographic data, preventing invalid data from interfering with subsequent processing. Marking the positions of valid data points using the spatial coordinate system of the OCT equipment allows these points to carry precise spatial positioning information, providing an objective spatial reference for subsequent image parameter adjustments. Adjusting preliminary corneal optical image parameters based on valid corneal tomographic data with spatial positioning information avoids relying on... By eliminating the blindness of manual experience-based adjustments, the brightness, clarity, and other parameters of the final corneal optical image data are precisely matched with the actual spatial morphology of the cornea, improving the image's ability to reproduce the subtle morphology of the cornea. Region segmentation of the final corneal optical image data can accurately separate the corneal region from the image, eliminating interference from non-corneal regions such as eyelids and irises. Feature recognition and acquisition of 3D coordinates based on pure corneal images can provide accurate and non-redundant feature point data support for subsequent 3D modeling. Modeling based on accurate feature point 3D coordinates using a triangular mesh generation algorithm can construct a 3D mesh model that fully reflects the topological structure of the corneal surface. Processing the model with a Gaussian smoothing algorithm can eliminate irregular protrusions or depressions on the model surface caused by minor deviations in feature points, resulting in a corneal surface shape that conforms to the natural physiological morphology of the cornea.
[0014] Furthermore, based on the three-dimensional coordinates of corneal feature points, this application first calculates the distance between all three-dimensional coordinates to obtain the distance values between each feature point, and defines the range of the distance values to obtain the judgment criteria for adjacent feature points. Based on the judgment criteria, a three-dimensional mesh model of the corneal surface is constructed using a triangular mesh generation algorithm. The three-dimensional mesh model is then subjected to morphological detection to mark the areas to be smoothed that have protrusions or depressions. The area to be smoothed is then locally adjusted using a Gaussian smoothing algorithm to make the feature points in the area to be smoothed approach the feature points in the surrounding areas to obtain a locally adjusted mesh model. The locally adjusted mesh model is then fused to obtain an overall mesh model. Finally, the overall mesh model is subjected to surface fitting to obtain the shape of the corneal surface.
[0015] This application uses distance calculation and range definition to determine the criteria for adjacent feature points, which ensures the rationality of feature point connections in the 3D mesh model constructed by the triangular mesh generation algorithm, avoids model shape deviations caused by misconnected feature points, and improves the basic accuracy of the model. Targeted marking and local adjustment of areas to be smoothed can eliminate irregular shapes such as protrusions or depressions while preserving details of normal corneal morphology areas, avoiding over-smoothing. Mesh fusion can achieve seamless connection between locally adjusted and unadjusted areas, avoiding shape breaks at the connection points. Surface fitting further optimizes the continuity and smoothness of the overall shape, ultimately resulting in a corneal surface shape that is highly accurate, natural in shape, and conforms to the actual physiological structure of the cornea. Attached Figure Description
[0016] To more clearly illustrate the technical solutions of the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0017] Figure 1 A schematic flowchart illustrating a method for constructing a corneal surface shape according to an embodiment of this application; Figure 2 A schematic diagram illustrating a specific implementation of a method for constructing a corneal surface shape according to an embodiment of this application; Figure 3 A schematic diagram of a corneal surface shape construction system provided in an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0018] In ophthalmological diagnosis and treatment, and corneal custom-made instrument fitting, the accuracy of corneal surface shape is crucial. However, existing methods for constructing corneal surface shapes have significant shortcomings: the raw data is not specifically segmented based on corneal anatomy, making it difficult to accurately reflect the morphological characteristics of different corneal regions; invalid data such as noise, motion artifacts, and signal loss are only filtered in a basic manner, easily leaving behind interference information; adjusting image parameters relies on human experience and lacks objective spatial references; the modeling and smoothing processes have limited precision, and the final results often deviate from the actual corneal shape, failing to meet the needs of high-precision applications.
[0019] To address the aforementioned issues, this application provides a method for constructing the corneal surface shape: First, raw corneal scan data is acquired using an optical coherence tomography (OCT) device, and corneal tomographic scan data is obtained by dividing the data according to a preset corneal anatomical structure; three types of invalid data points are selectively removed, and the spatial positions of valid data points are marked; preliminary optical image parameters are adjusted based on the tomographic data with positional information, and then the three-dimensional coordinates of feature points are extracted from the optimized image; finally, a three-dimensional mesh model is constructed based on the coordinates and smoothed using Gaussian to obtain a precise corneal surface shape. This solution addresses the shortcomings of existing technologies one by one through anatomical structure-guided data division, classified invalid data processing, spatially referenced image optimization, and high-precision modeling and smoothing, providing a reliable basis for corneal morphology in ophthalmic applications.
[0020] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. Based on the embodiments in this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0021] The core of this application is to provide a method for constructing the shape of the corneal surface, and a flowchart of one specific implementation is shown below. Figure 1 As shown, the method includes: S101. The cornea is scanned using an optical coherence tomography (OCT) device to obtain raw corneal scan data. The raw corneal scan data is then divided and processed according to a preset corneal anatomical structure to obtain corneal tomographic scan data.
[0022] In the above scheme, the optical coherence tomography (OCT) device refers to a professional medical device that utilizes the principle of low-coherence light interference to perform non-invasive scanning of the cornea and generate images and corresponding data of the fine internal structure of the cornea. It can capture the morphological and structural information of the cornea from the surface to the deep layers. The raw corneal scan data refers to the initial data set directly obtained after scanning the cornea with the above device, which includes the gray values, preliminary spatial location information, and structural contour data of each region of the cornea, without any screening or segmentation processing. The preset corneal anatomical structure refers to the pre-set structural division criteria based on the standard anatomical characteristics of the human cornea, including different tissue layers such as the corneal epithelium, corneal stroma, and corneal endothelium, as well as the definition standards of different regions such as the central, paracentral, and peripheral corneas. The corneal tomography data refers to the targeted data obtained after dividing the raw corneal scan data according to the preset corneal anatomical structure. Each data subset corresponds to a specific anatomical layer or region of the cornea and can clearly reflect the morphological characteristics of the corresponding part.
[0023] In this application example, corneal scanning is performed using an optical coherence tomography (OCT) device to obtain raw corneal scanning data. That is, optical coherence tomography technology is used. The OCT device first emits a low-coherence light signal to the cornea to be tested. When the light signal comes into contact with different tissue layers of the cornea, reflected light is generated. After receiving the reflected light from each layer, the device calculates the optical path difference of the reflected light through the principle of optical signal interference. The optical path difference is converted into the corresponding corneal tissue layer thickness, contour and grayscale information. Finally, this information is integrated to form raw corneal scanning data.
[0024] Next, the raw corneal scan data is segmented based on the preset corneal anatomy structure to obtain corneal tomographic scan data. First, the preset corneal anatomy structure standard is called. This standard contains the typical characteristics of different tissue layers of the cornea and the definition range of different regions. Then, each data point in the raw corneal scan data is matched with this standard to determine the corneal anatomy site corresponding to each data point. Then, all data points are classified and collected according to the anatomy site to form data subsets corresponding to different anatomy layers or regions, which are the corneal tomographic scan data.
[0025] In practical applications, medical institutions use optical coherence tomography (OCT) to diagnose corneal surface shape. First, the patient's head is kept still. The OCT lens is aimed at the patient's cornea and the scanning program is initiated. The OCT emits a low-coherence light signal into the cornea. Upon contact with the corneal epithelium, stroma, and endothelium, the light signal generates reflected light. The device receives these reflected lights and calculates the optical path difference using the formula... Calculate the optical path difference of the reflected light, where Let c be the optical path difference, and c be the speed of light, with a value of 3 × 10⁻⁶. 8 meters per second Let be the time difference between the device transmitting and receiving the reflected light signal. Assume the device detects the reflected light from the epithelial layer. 1.33×10 -11 seconds, substituting into the formula, we get Meters, or 1.995 millimeters, correspond to the thickness of the epithelial layer; similarly, the detected light reflected from the stromal layer... It is 8.67×10 -11 seconds, substituting into the formula, we get Meters, or 13.005 millimeters, correspond to information related to the thickness of the matrix layer; light reflected from the endothelial layer was detected. 9.00×10 -11 seconds, calculated Meters, or 13.5 millimeters, correspond to information related to the endothelium. These optical path differences are converted into thickness, contour, and grayscale information for each layer of the cornea. This data is then integrated to form raw corneal scan data containing 5000 data points, each containing a grayscale value and corresponding... And the initial spatial x-axis coordinates.
[0026] Then, the preset corneal anatomy standard is invoked. In this standard, the x-axis coordinate range of the central corneal area is -2.5 mm to 2.5 mm, the paracentral area is 2.5 mm to 5.5 mm, the peripheral area is 5.5 mm to 8 mm, the grayscale range of the epithelium is 75 to 125, the stroma is 35 to 75, and the endothelium is 100 to 140. By iterating through 5000 data points of the original corneal scan data, if a data point has an x-axis coordinate of 1.8 mm and a gray value of 90, it is determined to belong to the central epithelium and be classified into the "central epithelium data subset" because the x-axis coordinate is between -2.5 mm and 2.5 mm and the gray value is between 75 and 125. If a data point has an x-axis coordinate of 3.2 mm and a gray value of 55, it is determined to belong to the paracentral stroma and be classified into the "paracentral stroma data subset" because the x-axis coordinate is between 2.5 mm and 5.5 mm and the gray value is between 35 and 75. If a data point has an x-axis coordinate of 6.0 mm and a gray value of 110, it is determined to belong to the peripheral endothelial layer and be classified into the "peripheral endothelial layer data subset" because the x-axis coordinate is between 5.5 mm and 8 mm and the gray value is between 100 and 140. All data points are classified according to this rule, resulting in 8 data subsets corresponding to different anatomical regions, namely corneal tomography data. This corneal tomography data will be used in subsequent steps to remove invalid data points caused by noise, motion artifacts, or signal loss.
[0027] The S101 overall solution described above acquires raw corneal scan data using professional optical coherence tomography (OCT) equipment, ensuring that the data completely covers the entire corneal structure. Simultaneously, the raw data is divided according to a pre-defined corneal anatomical structure, giving the obtained corneal tomographic scan data a clear anatomical correspondence. This avoids the problem of data mixing between different anatomical layers or regions in subsequent processing, allowing subsequent steps to precisely target specific areas of the cornea. This effectively improves the targeting and reliability of the entire corneal surface shape construction process, laying the foundation for obtaining a final result that conforms to the actual shape of the cornea.
[0028] S102. Remove invalid data points from the corneal tomography data caused by noise, motion artifacts or signal loss to obtain valid corneal tomography data. Based on the spatial coordinate system of the optical coherence tomography device, mark the spatial position of each valid data point in the valid corneal tomography data to obtain valid corneal tomography data with position information.
[0029] Optionally, step S102, removing invalid data points from the corneal tomographic scan data caused by noise, motion artifacts, or signal loss to obtain valid corneal tomographic scan data, includes: Step 1021: Perform signal stability analysis on the corneal tomography data to obtain the signal fluctuation amplitude, perform numerical statistical processing on the signal fluctuation amplitude to obtain the signal stability standard, and regard the data points whose signal fluctuation amplitude exceeds the signal stability standard as noise invalid data points.
[0030] Step 1022: Perform continuity analysis on the positional relationship of adjacent data points in the corneal tomography data to obtain the positional deviation value between adjacent data points. Perform distribution statistics on all the positional deviation values to obtain the positional continuity standard. Data points whose positional deviation values exceed the positional continuity standard are regarded as invalid data points of motion artifacts.
[0031] Step 1023: Perform signal intensity detection processing on each data point in the corneal tomography data to obtain the signal intensity value of each data point. Perform range statistics on all the signal intensity values to obtain the signal intensity standard. Data points with signal intensity values lower than the signal intensity standard are regarded as invalid data points with missing signals.
[0032] Step 1024: Remove the noise-type invalid data points, motion artifact-type invalid data points, and signal-deficient invalid data points from the corneal tomography data to obtain valid data points. Integrate the valid data points according to the original order of the corneal tomography scan to obtain valid corneal tomography data.
[0033] In the above scheme, invalid data points caused by noise, motion artifacts, or signal loss refer to data points in corneal tomography data that do not conform to the normal corneal morphological characteristics. Noise-related data points are caused by irregular fluctuations in grayscale values due to electromagnetic interference from the equipment or ambient light; motion artifact-related data points are caused by slight eye movements of the subject, resulting in spatial deviations from the normal continuous trajectory; and signal loss-related data points are caused by insufficient signal reception from the equipment, resulting in excessively low grayscale values or incomplete information. The spatial coordinate system of the optical coherence tomography (OCT) equipment is a preset three-dimensional coordinate system, including x, y, and z axes, used to locate the actual spatial position of corneal data points. Valid corneal tomography data with positional information refers to valid corneal tomography data... Based on this, a data set is formed by assigning specific x, y, and z coordinates in a spatial coordinate system to each valid data point; signal fluctuation amplitude refers to the difference between the gray value of a single data point and the average gray value of its neighboring data points, used to determine whether the data point is affected by noise; signal stability standard refers to the normal fluctuation range threshold obtained by statistically analyzing the signal fluctuation amplitude of all data points, used to define invalid data points due to noise; position deviation value refers to the distance difference between the three-dimensional coordinates of adjacent data points in the spatial coordinate system, used to determine whether the data point deviates from the normal trajectory due to motion artifacts; position continuity standard refers to the normal distance range threshold obtained by statistically analyzing the position deviation values of all adjacent data points, used to define invalid data points due to motion artifacts.
[0034] In this application example, firstly, signal stability analysis is performed on the corneal tomography data using a 1021-step process. A sliding window algorithm is used to select a predetermined number of neighboring data points around each data point, and the average gray value of these neighboring data points is calculated. Then, the signal fluctuation amplitude is calculated using the formula... Calculate the signal fluctuation amplitude of a single data point, where For signal fluctuation amplitude, For the grayscale value of a single data point, The average gray value of the surrounding adjacent data points is used; then, the signal fluctuation amplitude of all data points is statistically analyzed, and the mean and standard deviation of the fluctuation amplitude are calculated. The sum of the mean and one standard deviation is set as the signal stability standard, and data points whose signal fluctuation amplitude exceeds this standard are marked as noise invalid data points.
[0035] Next, a continuity analysis of the positional relationship between adjacent data points in the corneal tomography data was performed using a 1022-step process. Adjacent data point pairs were determined according to the original order of the corneal scans, and the Euclidean distance formula was applied. Calculate the positional deviation value for each pair of adjacent data points, where d is the positional deviation value. and These are the three-dimensional coordinates of two adjacent data points in the spatial coordinate system of the optical coherence tomography (OCT) device. The distribution of all position deviation values is statistically analyzed, and the deviation value corresponding to the 95th percentile is taken as the position continuity standard. Data points whose position deviation values exceed this standard are marked as invalid data points of motion artifact type.
[0036] Then, signal intensity detection is performed on each data point in the corneal tomography data through 1023 steps. The gray value of each data point is converted into the corresponding signal intensity value using a signal intensity detection algorithm. The gray value and the signal intensity value are positively correlated. The range of all signal intensity values is statistically analyzed, and the mean of the signal intensity values is calculated. 60% of the mean is set as the signal intensity standard, and data points with signal intensity values lower than this standard are marked as invalid data points with missing signals.
[0037] Finally, through 1024 steps, marked invalid data points of noise, motion artifacts, and signal loss were removed from the corneal tomography data. The remaining data points were then arranged and integrated according to the original order of the corneal tomography scans to obtain valid corneal tomography data. Subsequently, based on the spatial coordinate system of the optical coherence tomography (OCT) device, each valid data point in the valid corneal tomography data was assigned a unique three-dimensional coordinate label along the x, y, and z axes, forming valid corneal tomography data with positional information.
[0038] The aforementioned S102 overall solution effectively removes interference information and improves data purity and reliability by specifically identifying and eliminating three types of invalid data points in corneal tomography data caused by noise, motion artifacts, and signal loss. This ensures that the data used for subsequent processing truly reflects the actual morphological characteristics of the cornea. Simultaneously, by assigning positional markers to valid data points based on the spatial coordinate system of the optical coherence tomography (OCT) device, each valid data point carries clear spatial positioning information. This provides an objective and accurate spatial reference for adjusting preliminary corneal optical image parameters based on this data, avoiding the blindness of image parameter adjustments. Furthermore, it ensures the accuracy of subsequent corneal image extraction, feature point recognition, and 3D modeling, laying a high-quality data foundation for the overall corneal surface shape construction process.
[0039] S103. Obtain preliminary corneal optical image data. Based on the effective corneal tomographic scan data with location information, adjust the image parameters of the preliminary corneal optical image data to obtain the final corneal optical image data.
[0040] Optionally, step S103, based on the valid corneal tomographic data with location information, adjusts the image parameters of the preliminary corneal optical image data to obtain the final corneal optical image data, including: Step 1031: Perform region delineation processing on the effective corneal tomographic data with location information to obtain the spatial coordinate range of key corneal regions. The key corneal regions refer to specific physical regions in the corneal solid structure that have a core influence on the precise construction of corneal surface morphology, ophthalmic diagnosis and treatment, and the adaptation of corneal customized instruments. Specifically, these include the central optical zone of the cornea, which is located at and around the geometric center of the cornea and directly determines visual acuity; the limbal region, which is located at the junction of the cornea and conjunctiva and is crucial for distinguishing corneal and non-corneal tissues; and the paracentral region of the cornea, which is located outside the central optical zone and inside the limbus and affects the continuity of the overall corneal morphology.
[0041] Step 1032: Match the spatial coordinate range with the preliminary corneal optical image data to obtain the image region, and perform pixel feature analysis processing on the image region to obtain the degree of brightness variation and edge continuity of the pixels.
[0042] Step 1033: Based on the degree of brightness change, adjust the exposure parameters of the preliminary corneal optical image data to obtain the adjusted exposure parameters.
[0043] Step 1034: Based on the edge continuity, adjust the image resolution parameters of the preliminary corneal optical image data to obtain the adjusted image resolution parameters.
[0044] Step 1035: Based on the adjusted exposure parameters and the adjusted image resolution parameters, the preliminary corneal optical image data is re-encoded to obtain the final corneal optical image data.
[0045] In the above scheme, preliminary corneal optical image data refers to the initial image data obtained by initially capturing the cornea through optical imaging equipment. This data includes the basic contour and pixel information of the corneal surface and has not undergone parameter optimization, potentially resulting in uneven brightness and blurred edges. Effective corneal tomographic data with location information refers to the data set after invalid data points have been removed in the previous steps, assigning three-dimensional coordinate labels to the remaining effective data points in the optical coherence tomography (OCT) equipment's spatial coordinate system. This data accurately reflects the spatial location and morphological characteristics of various corneal regions. The spatial coordinate range of key corneal regions refers to the spatial coordinate intervals defined from the effective corneal tomographic data with location information, corresponding to regions crucial for corneal morphological analysis. This includes the maximum and minimum values of this region on the x, y, and z axes. The image region refers to the area where the spatial coordinate range of the key corneal regions is compared with the preliminary corneal optical image data. After coordinate matching, the pixel region corresponding to the key corneal area is located in the image. The degree of brightness variation of a pixel refers to the difference in grayscale values of different pixels within the image region, reflecting the uniformity of brightness within the image region. The edge continuity of a pixel refers to the smoothness of the connection between pixels at the edge of the corneal contour within the image region, reflecting the clarity of edge details. Exposure parameters are parameters that affect the brightness of the initial corneal optical image data; adjusting these parameters can change the overall brightness of the image. Image resolution parameters are parameters that affect the clarity of details in the initial corneal optical image data; adjusting these parameters can change the pixel density in the image, thus affecting the presentation of edge details. The final corneal optical image data refers to the image data with uniform brightness and clear edges obtained after re-encoding the initial corneal optical image data based on the adjusted exposure and image resolution parameters.
[0046] The aforementioned S103 overall solution defines the spatial coordinate range of key corneal regions using effective corneal tomographic data with location information, providing a precise spatial reference for subsequent image parameter adjustments and avoiding the blindness of relying on experience-based adjustments. By matching the spatial coordinate range to obtain the image region and analyzing its pixel brightness and edge features, the optimization direction of the initial corneal optical image data can be accurately located. Adjusting the exposure parameters based on the degree of brightness variation ensures uniform brightness in the final image data, avoiding local overexposure or underexposure that could affect morphological recognition. Adjusting the image resolution parameters based on the degree of edge continuity ensures clear edges in the final image data, guaranteeing the accurate presentation of corneal contour details. Finally, the re-encoded final corneal optical image data provides a high-quality image foundation for subsequent corneal image extraction and feature point recognition, further improving the accuracy of corneal surface shape construction.
[0047] S104. Perform region segmentation processing on the final corneal optical image data to obtain a corneal image, and perform feature recognition on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea.
[0048] Optionally, step S104 involves performing region segmentation processing on the final corneal optical image data to obtain a corneal image, and performing feature recognition on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea, including: Step 1041: Perform pixel feature analysis on the final corneal optical image data to obtain the pixel feature range of the corneal region. Based on the pixel feature range, perform region segmentation on the final corneal optical image data to obtain a corneal image.
[0049] Step 1042: Perform surface morphology recognition processing on the corneal image to mark the locations where the corneal surface exhibits bending changes and the locations of the corneal edges, thereby obtaining the feature points to be confirmed on the corneal surface.
[0050] Step 1043: Associate the feature points to be confirmed with the valid corneal tomography data to obtain the three-dimensional coordinates of each corneal feature point.
[0051] In the above scheme, the final corneal optical image data refers to the image data re-encoded based on the adjusted exposure and image resolution parameters in the preceding steps. It is characterized by uniform brightness and clear edges, and includes pixel information of the cornea and other ocular surface tissues. Region segmentation processing refers to the technique used to separate the corneal region from the non-corneal region from the final corneal optical image data, achieving region division by selecting pixels that conform to corneal characteristics. Pixel feature analysis processing refers to the technique of detecting and statistically analyzing the pixel grayscale, texture, and other features of the final corneal optical image data to determine the unique pixel feature patterns of the corneal region. The pixel feature range of the corneal region refers to the range of corneal features obtained through pixel feature analysis processing. The normal value range of regional pixels in dimensions such as grayscale and texture is the basis for distinguishing corneal and non-corneal areas; surface morphology recognition processing refers to the technology of scanning corneal images to identify morphological features such as changes in corneal surface curvature and boundary positions; feature points to be confirmed refer to the pixels corresponding to the locations of corneal surface curvature changes and corneal edges marked by surface morphology recognition processing, which are the basis for obtaining three-dimensional coordinates; the three-dimensional coordinates of each corneal feature point refer to the x-axis, y-axis, and z-axis coordinates assigned to each feature point in the spatial coordinate system of the optical coherence tomography (OCT) scanner after associating the feature point to be confirmed with valid corneal tomographic scan data, which are the core data for subsequent three-dimensional modeling.
[0052] In this application example, firstly, pixel feature analysis processing is performed on the final corneal optical image data through 1041 steps. A pixel feature analysis algorithm is used to extract the grayscale values and texture features of all pixels in the image. The texture features are obtained by statistically analyzing the grayscale variation patterns between adjacent pixels. Subsequently, the extracted grayscale values and texture features are classified and statistically analyzed to filter out the grayscale value distribution range and texture variation range corresponding to the corneal region. The normal value ranges of these two dimensions are then integrated to obtain the pixel feature range of the corneal region. Based on the obtained pixel feature range of the corneal region, a region segmentation algorithm is used to perform pixel-by-pixel judgment on the final corneal optical image data. Pixels that conform to the feature range are retained, while pixels that do not conform are removed, i.e., pixels outside the corneal region are eliminated, resulting in a corneal image containing only corneal region pixel information.
[0053] Next, after obtaining the corneal image, the corneal image is processed through 1042 steps to perform surface morphology recognition. The surface morphology recognition algorithm scans the corneal image row by row and column by column to detect the trend of pixel grayscale changes on the corneal surface. That is, when the grayscale change trend shows a clear inflection point, the pixel at that position is marked. At the same time, the positions of abrupt changes in pixel grayscale at the edge of the corneal image are detected and the pixels at that position are marked. All marked pixels are integrated to obtain the feature points to be confirmed on the corneal surface.
[0054] Then, after obtaining the feature points to be confirmed, the feature points to be confirmed are associated with the valid corneal tomographic scan data through step 1043. The correspondence between the pixel coordinate system of the corneal image and the spatial coordinate system of the valid corneal tomographic scan data is established using data association technology. According to the correspondence, the pixel coordinates of each feature point to be confirmed are converted into x-axis and y-axis values of spatial coordinates. In the valid corneal tomographic scan data, valid data points that match the converted x-axis and y-axis values are found, and the z-axis coordinates of the valid data points are extracted. The x-axis, y-axis and z-axis coordinates are combined to obtain the three-dimensional coordinates of each feature point of the cornea.
[0055] In practical applications, firstly, a pixel feature analysis algorithm is used to extract the grayscale values and texture features of all pixels in the final corneal optical image data. Statistical analysis reveals that the grayscale values of pixels belonging to the corneal region are concentrated between 100 and 200, and the grayscale difference between adjacent pixels is generally less than 10. Based on this, "grayscale value 100-200 and grayscale difference between adjacent pixels <10" is determined as the pixel feature range of the corneal region. Then, a region segmentation algorithm is used to judge the final corneal optical image data pixel by pixel, retaining pixels with grayscale values between 100-200 and grayscale difference between adjacent pixels <10, and removing eyelid region pixels with grayscale values below 100 and iris region pixels with grayscale values above 200, finally obtaining a corneal image containing only the corneal region.
[0056] Secondly, a surface morphology recognition algorithm was used to scan the corneal image row by row and column by column. When the scan reached pixel x coordinate 220 and pixel y coordinate 220, the trend of pixel grayscale change was detected to change from "slowly rising" to "rapidly falling". Therefore, this position was marked as a feature point to be confirmed at the curvature of the corneal surface. When the scan reached pixel x coordinate 480, pixel y coordinate 480, pixel x coordinate 480, pixel y coordinate 520, pixel x coordinate 520, pixel y coordinate 480, and pixel x coordinate 520, pixel y coordinate 520, the pixel grayscale was detected to drop sharply from 180 to 20. These positions were marked as feature points to be confirmed at the edge of the cornea. Finally, 30 feature points to be confirmed were obtained.
[0057] Finally, data association technology was used to establish the correspondence between the corneal image pixel coordinate system and the effective corneal tomographic scan data spatial coordinate system. A scaling factor k was set to 0.01 mm / pixel, meaning one pixel corresponds to a spatial distance of 0.01 mm. This was achieved using the formula... and The pixel coordinates of the feature points to be confirmed are converted into spatial x-axis and y-axis coordinates, where X is the spatial x-axis coordinate. Here, x is the pixel x-coordinate of the feature point to be confirmed, k is a scaling factor, and y is the y-axis coordinate in space. Let X be the pixel y-coordinate of the feature point to be confirmed. For example, for the feature point to be confirmed, i.e., pixel x-coordinate 220 and pixel y-coordinate 220, substituting into the formula, we get X = 220 × 0.01 = 2.2 mm and Y = 220 × 0.01 = 2.2 mm. In the valid corneal tomographic scan data, find the valid data point corresponding to 2.2 mm on the spatial x-axis and 2.2 mm on the spatial y-axis, and extract the spatial z-axis coordinate of 0.43 mm for this data point to obtain the three-dimensional coordinates of the feature point (2.2 mm, 2.2 mm, 0.43 mm). Process the remaining 29 feature points to be confirmed in the same way to obtain the three-dimensional coordinates of 30 corneal feature points. These three-dimensional coordinates will be used in subsequent steps to construct a three-dimensional mesh model of the corneal surface using a triangular mesh generation algorithm.
[0058] The aforementioned S104 overall solution, through pixel feature analysis and region segmentation, can accurately separate the corneal region from the final corneal optical image data, eliminating interference from non-corneal tissues such as eyelids and irises. This ensures that subsequent processing is only applied to the cornea itself, avoiding the influence of non-corneal region pixels on morphological analysis. Through surface morphology recognition processing, it can accurately mark the feature points to be confirmed regarding corneal surface curvature and edge positions. These points cover key details of corneal morphology, providing accurate target objects for subsequent acquisition of three-dimensional coordinates. By associating the feature points to be confirmed with valid corneal tomographic scan data, it can assign realistic spatial three-dimensional coordinates to each feature point. This ensures that the coordinate data not only conforms to the morphological features of the corneal image but also relies on the spatial information of the tomographic data to guarantee accuracy. The resulting three-dimensional coordinates of each corneal feature point provide reliable core data support for subsequent three-dimensional modeling, further ensuring the accuracy of corneal surface shape construction.
[0059] S105. Based on the three-dimensional coordinates, a three-dimensional model of the corneal surface is generated using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then smoothed using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
[0060] Optionally, step S105 involves using the three-dimensional coordinates to perform a three-dimensional model of the corneal surface through a triangular mesh generation algorithm, obtaining a three-dimensional mesh model of the corneal surface, and then smoothing the three-dimensional mesh model using a Gaussian smoothing algorithm to obtain the shape of the corneal surface, including: Step 1051: Perform distance calculation processing on all the three-dimensional coordinates to obtain the distance value between each feature point, and perform range definition processing on the distance value to obtain the judgment criteria for adjacent feature points.
[0061] Step 1052: Based on the judgment criteria, the corneal surface is modeled in three dimensions using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface.
[0062] Step 1053: Perform shape detection processing on the three-dimensional mesh model to mark local areas with protrusions or depressions in the mesh, and obtain the area to be smoothed.
[0063] Step 1054: Using the Gaussian smoothing algorithm, perform local adjustment processing on the area to be smoothed so that the feature point positions of the area to be smoothed are closer to the feature point positions of the surrounding areas, thus obtaining the locally adjusted mesh model.
[0064] Step 1054 may specifically include the following steps: performing boundary identification processing on the region to be smoothed to obtain edge feature points of the region to be smoothed; performing feature point selection processing on the surrounding region outside the edge feature points to obtain surrounding reference feature points; calculating the positional relationship between the feature points in the region to be smoothed and the surrounding reference feature points using a Gaussian smoothing algorithm to obtain vector features between the feature points and each of the surrounding reference feature points; integrating all the vector features to obtain positional association information for each feature point; performing range division processing on the positional association information to obtain an adjustment weight value for each feature point; performing movement processing on the feature points based on the positional association information and the adjustment weight value to obtain moved feature points; and performing connection processing on the moved feature points to obtain a locally adjusted mesh model.
[0065] Step 1055: Perform mesh fusion on the locally adjusted mesh model to obtain an overall mesh model, and perform surface fitting on the overall mesh model to obtain the corneal surface shape.
[0066] Step 1055 may specifically include the following steps: performing boundary traversal processing on the locally adjusted mesh model to obtain the boundary feature points between the locally adjusted region and the unadjusted region, and calculating the gap distance and angular deviation between adjacent boundary feature points; based on the gap distance and angular deviation, fine-tuning the position of the boundary feature points to obtain adjusted boundary feature points; connecting the adjusted boundary feature points, feature points within the locally adjusted region, and feature points within the unadjusted region to obtain the overall mesh model; performing region division processing on the overall mesh model to obtain sub-region meshes; performing surface feature analysis on the sub-region meshes to obtain surface feature parameters; adjusting the sub-region meshes according to the surface feature parameters to obtain a coordinated mesh model; and performing surface fitting processing on the coordinated mesh model to obtain the corneal surface shape.
[0067] The unadjusted regions refer to areas in the 3D corneal mesh model that have not undergone Gaussian smoothing adjustments in step 1054. These correspond to parts of the corneal structure that are morphologically stable and without obvious protrusions or depressions. In the initially constructed 3D mesh model, the surface morphology of these regions already conforms to the natural physiological characteristics of the cornea, such as the gentle curvature of the central cornea and the gradual curvature of the paracentral cornea, requiring no local adjustment. Functionally and structurally, the unadjusted regions typically include areas of good morphological continuity and without abrupt changes in local morphology, such as most of the central cornea and the peripheral cornea without abnormalities. The normal vector angles and feature point distributions of their mesh units conform to the morphological patterns of a normal cornea, accurately reflecting the actual morphology of the corresponding parts of the cornea. Therefore, they can serve as a reliable basis for subsequent mesh fusion and surface fitting without additional adjustment. The identification method first uses the surface morphology detection algorithm in step S1053 to mark the areas in the 3D mesh model that have protrusions or depressions and need to be smoothed. The region is defined, and the coordinates of its boundary feature points are recorded. Then, during the local adjustment in step S1054, the positions of the feature points are fine-tuned only for the region to be smoothed, and the boundary of the locally adjusted region is completely consistent with the boundary of the region to be smoothed marked in step S1053. By scanning the locally adjusted mesh model through the boundary traversal algorithm, the coordinates of the boundary feature points of the locally adjusted region can be reconfirmed. Then, the coordinates of all feature points in the locally adjusted mesh model are compared with the coordinates of the confirmed boundary feature points of the locally adjusted region. If the coordinates of a feature point are outside the boundary of the locally adjusted region, and the mesh cell where the feature point is located was not marked as a region to be smoothed in step S1053, then the feature point and the mesh cell where it is located together constitute an unadjusted region.
[0068] The division of the regional grid follows the rule of "matching corneal anatomical location with physiological function." Combining the spatial coordinate system of the optical coherence tomography (OCT) equipment from the previous steps and the preset corneal anatomical structure standards, the specific division logic is as follows: Taking the geometric center of the cornea as the origin, the radial range is defined based on the x-axis and y-axis plane coordinates of the spatial coordinate system. At the same time, considering the functional differences in corneal surface morphology, the overall grid model is divided into sub-grids that correspond one-to-one with the anatomical regions of the cornea. The surface feature parameters refer to the indicators that quantify the morphological characteristics of the corneal surface in that region by analyzing the surface morphology of the corresponding sub-grids. These indicators directly correspond to the physical properties of the corneal surface, such as curvature, smoothness, and morphological trend. They are the core basis for judging whether the shape of the sub-grid conforms to the natural physiological state of the cornea, and at the same time provide a quantitative standard for subsequent adjustments to the sub-grids to achieve overall coordination.
[0069] In the above scheme, the triangular mesh generation algorithm refers to a modeling technique that connects adjacent feature points to form triangular units, which are then pieced together to form a mesh structure covering the corneal surface. The distance between each feature point refers to the straight-line distance between any two feature points in three-dimensional space, used to determine whether feature points are adjacent. The criterion for determining adjacent feature points is a threshold obtained by defining the range of distance values; feature points with distance values less than this threshold are determined to be adjacent. The three-dimensional mesh model of the corneal surface refers to a mesh structure formed by connecting adjacent feature points through the triangular mesh generation algorithm, which can initially reflect the topological morphology of the corneal surface. Morphological detection processing... This refers to a technique that scans a 3D mesh model to identify protrusions or depressions that deviate from the normal corneal curvature. The region to be smoothed refers to a localized area within the mesh that contains protrusions or depressions, marked through shape detection processing. Gaussian smoothing is a smoothing technique that calculates the positional relationship between feature points in the region to be smoothed and surrounding reference feature points, adjusting the positions of the feature points to eliminate protrusions or depressions. Peripheral reference feature points are feature points selected from the surrounding area outside the edge feature points and used as adjustment benchmarks. Vector features refer to the relationship between the feature points in the region to be smoothed and the surrounding reference feature points along the x, y, and z axes. The following are the key terms and their corresponding functions: position difference; position association information refers to the comprehensive information obtained by integrating all vector features, reflecting the positional relationship between feature points in the area to be smoothed and surrounding reference feature points; adjustment weight value refers to the value used to determine the movement range of feature points in the area to be smoothed, based on position association information. The larger the weight, the closer the movement range is to the corresponding surrounding reference feature points; moving feature point refers to the feature point in the area to be smoothed after position adjustment; locally adjusted mesh model refers to the mesh structure obtained by connecting moving feature points, eliminating local protrusions or depressions; mesh fusion refers to the boundary connection optimization of the adjusted and unadjusted areas in the locally adjusted mesh model; overall mesh model refers to the complete mesh structure obtained through mesh fusion, where the adjusted and unadjusted areas are seamlessly connected; surface fitting refers to the technique of optimizing the surface of the overall mesh model to fit the natural physiological morphology of the cornea; sub-regional mesh refers to the sub-mesh obtained by dividing the overall mesh model according to the anatomical regions of the cornea; surface feature parameters refer to the parameters obtained by analyzing sub-regional meshes, reflecting the curvature and smoothness of the surface in the corresponding region; coordinated mesh model refers to the mesh structure obtained by adjusting the sub-regional mesh based on surface feature parameters, where the surface morphology of each region is coordinated.
[0070] In this application example, firstly, distance calculation is performed on all three-dimensional coordinates through step 1051. The distance between any two feature points is calculated using the Euclidean distance formula, which is: Where d is the distance between the two feature points. and These are the x-axis coordinates of the two feature points. and These are the y-axis coordinates of the two feature points. and The z-axis coordinates of the two feature points are given. Statistical analysis is performed on all calculated distance values, and the 70th percentile of the distance value distribution is taken as the criterion for determining adjacent feature points. Feature points with distance values less than this criterion are determined to be adjacent, thus obtaining the criterion for determining adjacent feature points.
[0071] Next, after obtaining the criteria for determining adjacent feature points, a triangular mesh generation algorithm is used to perform three-dimensional modeling of the corneal surface based on the criteria. First, adjacent feature points that meet the criteria are selected for each feature point. Then, triangular units are constructed with each feature point and its adjacent feature points as vertices. All triangular units are spliced together according to the overall shape trend of the corneal surface to form a mesh structure covering the entire corneal surface, thus obtaining a three-dimensional mesh model of the corneal surface.
[0072] Then, the 3D mesh model is subjected to shape detection processing. The surface shape detection algorithm is used to calculate the normal vector of each triangular unit in the mesh, and then the angle between the normal vector and the normal vector of the surrounding adjacent triangular units is calculated. If the angle exceeds the preset normal range, it indicates that there is an abrupt change in the surface shape of the region, and there is a problem of convexity or concavity. The region is marked as the region to be smoothed, and the region to be smoothed is obtained.
[0073] Next, local adjustments are made to the region to be smoothed. First, a boundary recognition algorithm is used to scan the contour of the region to be smoothed, marking the feature points at the region boundary as edge feature points. From the surrounding area outside the edge feature points, feature points within a range of 1 to 2 triangular units away from the edge feature points are selected as surrounding reference feature points. Then, a Gaussian smoothing algorithm is used to calculate the positional differences between each feature point in the region to be smoothed and each surrounding reference feature point in the x, y, and z axes, obtaining vector features. Subsequently, all vector features corresponding to each feature point to be smoothed are integrated to form positional association information reflecting the positional relationship between the feature point and the surrounding reference feature points. According to the rule that the closer the surrounding reference feature point is to the feature point to be smoothed, the larger the corresponding weight value, the weight values are adjusted based on the positional association information. Then, the target position of the feature point to be smoothed is calculated based on the positional association information and the adjusted weight values. The feature point to be smoothed is moved to the target position to obtain the moved feature point, and all moved feature points and the unadjusted feature points around the region to be smoothed are connected to obtain the locally adjusted mesh model.
[0074] Finally, mesh fusion and surface fitting are performed on the locally adjusted mesh model. The first step uses a boundary traversal algorithm to scan the locally adjusted mesh model, marking the feature points at the boundary between the adjusted and unadjusted regions as boundary feature points. The second step uses the Euclidean distance formula to calculate the gap distance between adjacent boundary feature points and calculates the angle between the line connecting adjacent boundary feature points and the tangent to the corneal surface as the angular deviation. Based on the gap distance and angular deviation, the positions of the boundary feature points are fine-tuned to eliminate gaps and correct angular deviations, resulting in adjusted boundary feature points. These adjusted boundary feature points, the moving feature points within the adjusted region, and the feature points within the unadjusted region are then connected. The first step involves creating a complete, unbroken grid structure, resulting in an overall grid model. The second step divides the overall grid model into regions according to the anatomical division of the central, paracentral, and peripheral corneal regions, resulting in sub-regional grids. A surface feature analysis algorithm is used to calculate the curvature, torsion, and other parameters of each sub-regional grid, which are then used as surface feature parameters. The shape of the sub-regional grids is adjusted based on these surface feature parameters to ensure that the surface shapes of each region are consistent and harmonious, resulting in a coordinated grid model. The third step uses a surface fitting algorithm to optimize the coordinated grid model as a whole, making the grid surface continuous, smooth, and conforming to the natural physiological shape of the cornea, thus obtaining the corneal surface shape.
[0075] In practical applications, firstly, the Euclidean distance formula is used. Calculate the distance between any two feature points, where d is the distance between the two feature points. and These are the x-axis coordinates of the two feature points. and These are the y-axis coordinates of the two feature points. and These are the z-axis coordinates of two feature points; for example, to calculate the distance between feature point (2.2, 2.2, 0.43) and feature point (2.5, 2.5, 0.45), Substituting into the formula, we get d=0.424; we count the distance values between all 30 feature points, resulting in 435 distance values. After sorting these distance values from smallest to largest, we take 0.5, which corresponds to the 70th quantile, as the criterion for determining adjacent feature points. Feature points with a distance value less than 0.5 are considered adjacent.
[0076] Secondly, based on the adjacent feature point judgment standard of 0.5, a triangular mesh generation algorithm is used for modeling. For example, for the feature point (2.2, 2.2, 0.43), adjacent feature points (2.5, 2.5, 0.45) and (2.0, 2.3, 0.42) with a distance value less than 0.5 are selected, and triangular units are constructed with these three feature points as vertices. Triangular units are constructed for the remaining 29 feature points in the same way. All triangular units are spliced together according to the morphological trend of the corneal surface from the center to the periphery to form a three-dimensional mesh model covering the entire corneal surface.
[0077] Then, the surface shape detection algorithm is used to calculate the normal vector of each triangular unit in the 3D mesh model. For example, the normal vector of a certain triangular unit is (0.1, 0.2, 0.3), and the normal vector of its surrounding triangular units is (0.3, 0.4, 0.5). The angle between the two is calculated to be 15°, which exceeds the preset normal range of 10°. The area where the triangular unit is located is marked as the area to be smoothed.
[0078] Next, a boundary recognition algorithm is used to scan the region to be smoothed, marking the feature points (2.3, 2.4, 0.44) and (2.4, 2.3, 0.43) at the region boundary as edge feature points; from one to two triangular units outside the edge feature points, feature points (2.1, 2.2, 0.42) and (2.5, 2.6, 0.46) are selected as surrounding reference feature points; the feature point (2.35) within the region to be smoothed is calculated. The vector features of (2.1, 2.2, 0.42) and its surrounding reference feature point (2.35, 0.46) are (2.1-2.35, 2.2-2.35, 0.42-0.46), i.e. (-0.25, -0.15, -0.04). The vector features of (2.5, 2.6, 0.46) and its surrounding reference feature point (2.5, 2.6, 0.46) are (2.5-2.35, 2.6-2.35, 0.46-0.46), i.e. (-0.25, -0.15, -0.04). The two vector features are integrated to obtain the positional association information of the feature point to be smoothed; the weight values are adjusted based on the distance. The weight value of the surrounding reference feature point (2.1, 2.2, 0.42) is closer, so the weight value is set to 0.6, and the weight value of the surrounding reference feature point (2.5, 2.6, 0.46) is set to 0.4; then the target position is calculated as (2.35+(-0.25)×0.6+0.15×0.4, 2.35+(-0.15)×0.6+0.25×0.4, 0.46+(-0.04)×0.6+0×0.4), that is, (2.26, 2.36, 0.436), and the feature point to be smoothed (2.35, 2.35, 0.46) is moved to the target position to obtain the moved feature point; all moved feature points and surrounding unadjusted feature points are connected to obtain the locally adjusted mesh model.
[0079] Finally, the boundary feature point (2.3, 2.4, 0.44) between the adjusted and unadjusted regions is marked using a boundary traversal algorithm. The gap distance between adjacent boundary feature points is calculated to be 0.05, and the angle deviation is 8°. Based on this gap distance and angle deviation, the boundary feature point (2.3, 2.4, 0.44) is fine-tuned to (2.32, 2.41, 0.442) to obtain the adjusted boundary feature point. Connecting the adjusted boundary feature point, the moving feature point within the adjusted region, and the feature point within the unadjusted region yields the overall mesh model. The model is then constructed according to the corneal central area (x-axis -3 to 3...). The overall mesh model is divided into three regions: the central region (x-axis 3 to 5) and the peripheral region (x-axis 5 to 8), resulting in sub-regional meshes. The curvature of the sub-regional mesh in the central region is calculated to be 0.02, that in the central region is 0.05, and that in the peripheral region is 0.08 as surface feature parameters. Based on these parameters, the sub-regional meshes are adjusted to make the surface of the central region smoother and the surface of the peripheral region steeper, resulting in a coordinated mesh model. The coordinated mesh model is optimized using a surface fitting algorithm to obtain a corneal surface shape that conforms to the natural shape of the cornea. This corneal surface shape will be used to develop refractive surgery plans or customize orthokeratology lenses for patients.
[0080] The S105 overall solution described above ensures that the triangular mesh generation algorithm can accurately connect adjacent feature points by calculating the distance between feature points and defining adjacent standards. This avoids deviations in the 3D mesh model caused by misconnected feature points, laying an accurate foundation for subsequent modeling. The constructed 3D mesh model can initially restore the topological morphology of the corneal surface, providing a complete object for morphological detection. By accurately marking the areas to be smoothed, subsequent smoothing processing becomes more targeted, avoiding over-adjustment of normal morphological areas. By using a Gaussian smoothing algorithm combined with surrounding reference feature points to adjust the areas to be smoothed, abnormal morphologies such as protrusions or depressions can be eliminated, while conforming to the surrounding normal morphology, ensuring the naturalness of the local corneal morphology. Mesh fusion achieves seamless connection between the adjusted and unadjusted areas, avoiding boundary breakage issues. Surface fitting further optimizes the continuity and smoothness of the overall morphology. The final corneal surface shape can accurately conform to the actual physiological morphology of the cornea, providing a reliable morphological basis for ophthalmic diagnosis and treatment, corneal device customization, and other scenarios.
[0081] The following is a complete example for steps S101 to S105, such as Figure 2 As shown, when medical institutions construct the corneal surface shape for a patient, they first use an optical coherence tomography (OCT) scanner to scan the patient's cornea. The scanner emits low-coherence light signals into the cornea and receives reflected light from the corneal epithelium, stroma, and endothelium. The optical path difference is then calculated using the formula... Calculate the optical path difference of each layer, where Let c be the optical path difference, and c be the speed of light, with a value of 3 × 10⁻⁶. 8 meters per second The time difference between the emitted light signal and the received reflected light signal from the device; for example, detecting light reflected from the epithelial layer. 1.33×10 -11 seconds, substituting into the formula, we get Meters, or 1.995 millimeters, can be used to calculate the light reflected by the matrix and endothelial layers similarly. The corresponding optical path difference is obtained, and these differences are converted into the thickness, contour, and grayscale information of each layer of the cornea. This information is then integrated to form raw corneal scan data containing 5000 data points. Next, according to a preset corneal anatomy structure—namely, the central region x-axis range of -2.5 mm to 2.5 mm, the epithelial layer grayscale range of 75 to 125, the paracentral region x-axis range of 2.5 mm to 5.5 mm, and the stromal layer grayscale range of 35 to 75—the raw corneal scan data is divided. For example, the x-axis coordinate is 1.8 mm, and the grayscale value is... 90% of the data points were assigned to the central epithelial layer subset, and data points with an x-axis coordinate of 3.2 mm and a gray value of 55 were assigned to the paracentral stromal layer subset. This resulted in corneal tomography data containing eight anatomical region subsets, which will be used for subsequent invalid data point removal. After removing invalid data points from the corneal tomography data, a sliding window algorithm was used to perform signal stability analysis on each data point. The average gray value of the five neighboring data points around each data point was calculated, and the signal fluctuation amplitude formula was applied. Calculate the signal fluctuation amplitude, where For signal fluctuation amplitude, For the grayscale value of a single data point, It is the average gray value of the five surrounding data points; for example, a certain data point The gray values of its five neighboring data points are 92, 93, 94, 95, and 96, respectively. The calculation yields... Substituting into the formula, we get Subsequently, the signal fluctuation amplitude of all 5000 data points was statistically analyzed, and the mean was calculated to be 4, with a standard deviation of 1.5. The signal stability standard was then set to 4 + 1.5 = 5.5, and 100 data points were marked. Invalid data points exceeding 5.5 (noise class).
[0082] Secondly, adjacent data point pairs are determined according to the original order of corneal scans, and the Euclidean distance formula is used. Calculate the position deviation value, where d is the position deviation value. and These are the three-dimensional coordinates of two adjacent data points; for example, the coordinates of a pair of adjacent data points are (1.8, 2.0, 0.5) and (2.1, 2.3, 0.52), substituting them into the formula yields... Then, the positional deviation values of all adjacent data point pairs were calculated, and 0.2, corresponding to the 95th quantile, was taken as the positional continuity standard. Eighty invalid data points with positional deviation values exceeding 0.2 were marked as motion artifacts. Next, a signal strength detection algorithm was used to convert the grayscale value of each data point into a signal strength value at a ratio of 1:1.2. That is, for every increase of 1 in grayscale value, the signal strength value increases by 1.2. For example, a data point with a grayscale value of 50 would have a converted signal strength value of 50 × 1.2 = 60. The signal strength values of all data points were then calculated, and the mean was found to be 110. The signal strength standard was set to 110 × 60% = 66, and 50 invalid data points with signal strength values below 66 were marked as signal missing data points. Then... From 5000 data points, 100 invalid data points (noise-related, motion artifact-related, and signal-missing) were removed, leaving 4770 valid data points. These were then integrated into valid corneal tomographic data according to the original order of the corneal tomography scan. Based on the x, y, and z axes of the optical coherence tomography (OCT) equipment, each valid data point was assigned a three-dimensional coordinate label. For example, a valid data point in the central epithelial layer was labeled as (1.5, 1.8, 0.48), and a valid data point in the paracentral stromal layer was labeled as (3.0, 2.5, 0.6). This yielded valid corneal tomographic data with positional information, which will be used to adjust the parameters of the initial corneal optical image data.
[0083] Then, preliminary corneal optical image data is acquired. Based on valid corneal tomography data with location information, the spatial coordinate range of key corneal regions is defined. For example, the x-axis range of the central cornea is -3 mm to 3 mm, and the y-axis range is -3 mm to 3 mm; the x-axis range of the limbal region is -5 mm to 5 mm, and the y-axis range is -5 mm to 5 mm. This spatial coordinate range is matched with the preliminary corneal optical image data to establish the correspondence between the spatial coordinate system of the optical coherence tomography (OCT) device and the pixel coordinate system of the preliminary corneal optical image data, i.e., 1 mm corresponds to 100 pixels in the image. The image region corresponding to the key corneal region is then located, and pixel feature analysis is performed on this image region using the formula... Calculate the degree of brightness variation of pixels, where The degree of change in brightness, The maximum gray value of a pixel within the image region. This refers to the minimum grayscale value of a pixel within an image region; for example, it represents the minimum grayscale value of pixels within an image region. , Substituting into the formula, we get Due to significant differences in brightness, the initial exposure parameter was adjusted from 100 to 80, resulting in the adjusted exposure parameter. Simultaneously, the continuity of pixels at the corneal contour edges within the image region was detected, and the average distance deviation between adjacent edge pixels was calculated to be 5. The initial image resolution parameter was then adjusted from 300 dpi to 600 dpi, resulting in the adjusted image resolution parameter. Based on the adjusted exposure parameter and the adjusted image resolution parameter, the initial corneal optical image data was re-encoded to obtain the final corneal optical image data. This data will be used for subsequent region segmentation to obtain the corneal image.
[0084] Next, the final corneal optical image data was processed by region segmentation. The pixel feature range of the corneal region was set to a gray value of 100 to 200 with a gray value difference of less than 10 between adjacent pixels. A region segmentation algorithm was used to determine whether each pixel met this feature range. Pixels that met the feature range were retained, and pixels that did not meet the range were removed, resulting in a corneal image containing only the corneal region. Subsequently, feature recognition was performed on the corneal image. A surface morphology recognition algorithm was used to scan the corneal image row by row and column by column. When the scan reached pixel x coordinate 220 and pixel y coordinate 220, the gray value change trend of the pixel was detected to change from "slowly rising" to "rapidly falling". Therefore, this position was marked as a corneal surface curvature change. The feature points to be confirmed at the corneal edge were identified. When the scan reached pixel x-coordinate 480, pixel y-coordinate 480, pixel x-coordinate 480, pixel y-coordinate 520, pixel x-coordinate 520, pixel y-coordinate 480, and pixel x-coordinate 520, pixel y-coordinate 520, a sudden drop in pixel grayscale from 180 to 20 was detected. These locations were marked as feature points to be confirmed at the corneal edge, resulting in 30 feature points to be confirmed. Then, data association technology was used to establish the correspondence between the corneal image pixel coordinate system and the effective corneal tomographic scan data spatial coordinate system. The scaling factor k was set to 0.01 mm / pixel, meaning one pixel corresponds to a spatial distance of 0.01 mm. This was achieved using the formula... and The pixel coordinates of the feature points to be confirmed are converted into spatial x-axis and y-axis coordinates, where X is the spatial x-axis coordinate. Here, x is the pixel x-coordinate of the feature point to be confirmed, k is a scaling factor, and y is the y-axis coordinate in space. Let X be the pixel y-coordinate of the feature point to be confirmed. For example, for the feature point to be confirmed (pixel x-coordinate 220, pixel y-coordinate 220), substituting into the formula, we get X = 220 × 0.01 = 2.2 mm and Y = 220 × 0.01 = 2.2 mm. In the valid corneal tomographic scan data, find the valid data point corresponding to 2.2 mm on the spatial x-axis and 2.2 mm on the spatial y-axis, and extract the spatial z-axis coordinate of 0.43 mm for this data point to obtain the three-dimensional coordinates of the feature point (2.2, 2.2, 0.43). Then, process the remaining 29 feature points to be confirmed in the same way to obtain the three-dimensional coordinates of 30 corneal feature points. These three-dimensional coordinates will be used for subsequent three-dimensional modeling of the corneal surface.
[0085] Next, the corneal surface shape was constructed based on the three-dimensional coordinates of 30 corneal feature points, first using the Euclidean distance formula. Calculate the distance between any two feature points, where d is the distance between the two feature points. and These are the x-axis coordinates of the two feature points. and These are the y-axis coordinates of the two feature points. and These are the z-axis coordinates of two feature points. For example, calculating the distance between feature points (2.2, 2.2, 0.43) and (2.5, 2.5, 0.45), substituting into the formula, we get d = 0.424. We then count the distances between all 30 feature points, sort these distances from smallest to largest, and use 0.5, corresponding to the 70th quantile, as the criterion for determining adjacent feature points. Feature points with a distance value less than 0.5 are considered adjacent. Based on this criterion, a triangular mesh generation algorithm is used to perform 3D modeling of the corneal surface. For example, for feature point (2.2, 2.2, 0.43), adjacent feature points (2.5, 2.5, 0.45) and (2.0, 2.3, 0.42) are selected. These three feature points are used as vertices to construct triangular units. Triangular units are constructed for the remaining 29 feature points in the same way. All triangular units are spliced together according to the morphological trend of the corneal surface from the center to the periphery to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then subjected to morphological detection processing. A surface morphological detection algorithm is used to calculate the angle between the normal vector of each triangular unit and the normal vector of the surrounding triangular units. For example, if the normal vector of a certain triangular unit is (0.1, 0.2, 0.3) and the normal vector of its surrounding triangular units is (0.3, 0.4, 0.5), the angle between the two is calculated to be 15°, which exceeds the preset normal range of 10°. The area where the triangular unit is located is marked as the area to be smoothed.
[0086] Finally, the Gaussian smoothing algorithm is used to locally adjust the region to be smoothed. First, a boundary recognition algorithm is used to scan the region to be smoothed, marking the feature points (2.3, 2.4, 0.44) and (2.4, 2.3, 0.43) at the region boundary as edge feature points. From one to two triangular units outside the edge feature points, feature points (2.1, 2.2, 0.42) and (2.5, 2.6, 0.46) are selected as surrounding reference feature points. The vector feature of feature point (2.35, 2.35, 0.46) within the region to be smoothed and the surrounding reference feature point (2.1, 2.2, 0.42) is calculated as (2.1...). The vector features of the point (-2.35, 2.2-2.35, 0.42-0.46) are (-0.25, -0.15, -0.04), and the vector features of the surrounding reference feature point (2.5, 2.6, 0.46) are (2.5-2.35, 2.6-2.35, 0.46-0.46), which is (0.15, 0.25, 0). Integrating these two vector features yields the positional association information of the feature point to be smoothed. The weight values are adjusted based on distance. Since the surrounding reference feature point (2.1, 2.2, 0.42) is closer, its weight value is set to 0.6. The weight value is set to 0.4. Based on the location association information and the adjusted weight value, the target position of the feature point to be smoothed is calculated: x-axis coordinate = 2.35 + (-0.25) × 0.6 + 0.15 × 0.4 = 2.35 - 0.15 + 0.06 = 2.26, y-axis coordinate = 2.35 + (-0.15) × 0.6 + 0.25 × 0.4 = 2.35 - 0.09 + 0.1 = 2.36, z-axis coordinate = 0.46 + (-0.04) × 0.6 + 0 × 0.4 = 0.46 - 0.024 = 0.436. The feature point to be smoothed (2.35, 2.35, 0.46) is then moved to the target position. The moving feature point (2.26, 2.36, 0.436) is obtained; all moving feature points and the unadjusted feature points around the area to be smoothed are connected to obtain the locally adjusted mesh model; then, mesh fusion is performed on the locally adjusted mesh model, and the boundary traversal algorithm is used to scan the model to mark the boundary feature point (2.3, 2.4, 0.44) between the adjusted and unadjusted areas, and the gap distance between adjacent boundary feature points is calculated to be 0.05 and the angle deviation is 8° using the Euclidean distance formula; based on this gap distance and angle deviation, the boundary feature point (2.3, 2.4, 0.44) is fine-tuned to (2.32, 2.41, 0.436).442), obtain the adjusted boundary feature points; connect the adjusted boundary feature points, the moving feature points within the adjusted area, and the feature points within the unadjusted area to obtain the overall mesh model; then perform region division processing on the overall mesh model, according to the anatomical division method of the central corneal area (x-axis -3 mm to 3 mm), the paracentral area (x-axis 3 mm to 5 mm), and the peripheral area (x-axis 5 mm to 8 mm), to obtain sub-regional meshes; perform surface feature analysis on the sub-regional meshes, for example, calculate the curvature of the central sub-regional mesh to be 0.02, and the curvature of the paracentral sub-regional mesh to be... The curvature of the peripheral area grid is 0.08, with a value of 0.05. These curvature values are used as surface feature parameters. Based on these parameters, the regional grids are adjusted to make the central area surface smoother, the pericentral area surface gradually transition, and the peripheral area surface steeper, resulting in a harmonious grid model. This harmonious grid model is then subjected to surface fitting processing, and a surface fitting algorithm is used to optimize the grid surface morphology, creating a continuous, smooth surface that conforms to the natural physiological morphology of the cornea. This results in the final corneal surface shape, which can be used to plan refractive surgery or customize orthokeratology lenses for patients.
[0087] Figure 3 This is a schematic diagram of a specific embodiment of a corneal surface shape construction system provided in this application, referring to... Figure 3 The system may include: The segmentation module 31 is used to scan the cornea using an optical coherence tomography (OCT) device to obtain raw corneal scan data, and to segment the raw corneal scan data according to a preset corneal anatomical structure to obtain corneal tomographic scan data.
[0088] The marking module 32 is used to remove invalid data points in the corneal tomography data caused by noise, motion artifacts or signal loss, so as to obtain valid corneal tomography data. Based on the spatial coordinate system of the optical coherence tomography device, the module marks the spatial position of each valid data point in the valid corneal tomography data to obtain valid corneal tomography data with position information.
[0089] The adjustment module 33 is used to acquire preliminary corneal optical image data and, based on the effective corneal tomographic scan data with location information, adjust the image parameters of the preliminary corneal optical image data to obtain the final corneal optical image data.
[0090] The recognition module 34 is used to perform region segmentation processing on the final corneal optical image data to obtain a corneal image, and to perform feature recognition on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea.
[0091] The processing module 35 is used to perform three-dimensional modeling of the corneal surface based on the three-dimensional coordinates using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface, and to smooth the three-dimensional mesh model using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
[0092] The corneal surface shape construction system of this application embodiment is used to implement the aforementioned corneal surface shape construction method. Therefore, the specific implementation of the corneal surface shape construction system can be found in the embodiment section of the corneal surface shape construction method above. The specific implementation can be referred to the description of the corresponding embodiment, which will not be repeated here.
[0093] like Figure 4 As shown, this application also provides an electronic device, including: a memory 41 for storing a computer program; and a processor 42 for executing the computer program to implement the steps of any of the above-described methods for constructing the corneal surface shape.
[0094] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of any of the methods for constructing the corneal surface shape described above.
[0095] In one exemplary embodiment, the aforementioned computer-readable storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory, random access memory, portable hard drives, magnetic disks, or optical disks.
[0096] Embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements the steps in any of the above embodiments of the method for constructing a corneal surface shape.
[0097] Those skilled in the art will further recognize that the units and algorithm steps of the various examples described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of both. To clearly illustrate the interchangeability of hardware and software, the components and steps of the various examples have been generally described in terms of functionality in the foregoing description. Whether these functions are implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.
[0098] The foregoing has provided a detailed description of a method and system for constructing a corneal surface shape. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the embodiments above are merely for the purpose of helping to understand the method and its core ideas. It should be noted that those skilled in the art can make various improvements and modifications to this application without departing from its principles, and these improvements and modifications also fall within the protection scope of this application.
Claims
1. A method for constructing a corneal surface shape, characterized in that, include: The cornea is scanned using an optical coherence tomography (OCT) device to obtain raw corneal scan data. The raw corneal scan data is then divided and processed according to a preset corneal anatomical structure to obtain corneal tomographic scan data. Invalid data points caused by noise, motion artifacts, or signal loss are removed from the corneal tomography data to obtain valid corneal tomography data. Based on the spatial coordinate system of the optical coherence tomography device, the spatial position of each valid data point in the valid corneal tomography data is marked to obtain valid corneal tomography data with position information. Preliminary corneal optical image data is acquired. Based on the effective corneal tomographic scan data with location information, the image parameters of the preliminary corneal optical image data are adjusted to obtain the final corneal optical image data. The final corneal optical image data is processed by region segmentation to obtain a corneal image. Feature recognition is performed on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea. Based on the three-dimensional coordinates, a three-dimensional model of the corneal surface is generated using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then smoothed using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
2. The method according to claim 1, characterized in that, Based on the three-dimensional coordinates, a three-dimensional model of the corneal surface is generated using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface. The three-dimensional mesh model is then smoothed using a Gaussian smoothing algorithm to obtain the shape of the corneal surface, including: Distance calculation is performed on all the three-dimensional coordinates to obtain the distance values between each feature point. The range of the distance values is then defined to obtain the criteria for determining adjacent feature points. Based on the aforementioned criteria, a three-dimensional model of the corneal surface is obtained by using a triangular mesh generation algorithm. The three-dimensional mesh model is subjected to shape detection processing to mark local areas with protrusions or depressions in the mesh, thereby obtaining the area to be smoothed; The Gaussian smoothing algorithm is used to perform local adjustment processing on the region to be smoothed so that the feature point positions of the region to be smoothed are closer to the feature point positions of the surrounding regions, thus obtaining a locally adjusted mesh model. The locally adjusted mesh model is fused to obtain an overall mesh model. The overall mesh model is then fitted with a surface to obtain the corneal surface shape.
3. The method according to claim 2, characterized in that, The process involves using a Gaussian smoothing algorithm to locally adjust the region to be smoothed, so that the feature point positions in the region to be smoothed approach the feature point positions in the surrounding regions, resulting in a locally adjusted mesh model. Boundary recognition processing is performed on the region to be smoothed to obtain the edge feature points of the region to be smoothed. Feature point selection processing is performed on the surrounding region outside the edge feature points to obtain the surrounding reference feature points. The positional relationship between the feature points in the region to be smoothed and the surrounding reference feature points is calculated using the Gaussian smoothing algorithm. The vector features between the feature points and each of the surrounding reference feature points are obtained. All the vector features are integrated to obtain the positional association information of each feature point. The location association information is divided into ranges to obtain the adjustment weight value of each feature point. Based on the location association information and the adjustment weight value, the feature points are moved to obtain moved feature points. The moved feature points are then connected to obtain a locally adjusted mesh model.
4. The method according to claim 2, characterized in that, The process of fusing the locally adjusted mesh model to obtain an overall mesh model, and then performing surface fitting on the overall mesh model to obtain the corneal surface shape includes: The locally adjusted mesh model is subjected to boundary traversal processing to obtain the boundary feature points between the locally adjusted region and the unadjusted region, and the gap distance and angle deviation between adjacent boundary feature points are calculated. Based on the gap distance and angle deviation, the position of the boundary feature points is finely adjusted to obtain the adjusted boundary feature points. The adjusted boundary feature points, feature points in the local adjustment area, and feature points in the unadjusted area are connected to obtain the overall mesh model. The overall mesh model is divided into regions to obtain sub-regional meshes. Surface feature analysis is performed on the sub-regional meshes to obtain surface feature parameters. Based on the surface feature parameters, the sub-regional meshes are adjusted to obtain a coordinated mesh model. The corneal surface shape is obtained by performing surface fitting on the coordinated mesh model.
5. The method according to claim 1, characterized in that, The process of performing region segmentation on the final corneal optical image data to obtain a corneal image, and then performing feature recognition on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea, includes: The final corneal optical image data is subjected to pixel feature analysis to obtain the pixel feature range of the corneal region. Based on the pixel feature range, the final corneal optical image data is subjected to region segmentation to obtain the corneal image. The corneal image is subjected to surface morphology recognition processing to mark the locations where the corneal surface exhibits curvature changes and the locations of the corneal edges, thereby obtaining the feature points on the corneal surface to be confirmed; The feature points to be confirmed are associated with the valid corneal tomography data to obtain the three-dimensional coordinates of each corneal feature point.
6. The method according to claim 1, characterized in that, The step of adjusting the image parameters of the preliminary corneal optical image data based on the valid corneal tomographic data with location information to obtain the final corneal optical image data includes: The effective corneal tomographic scan data with location information is subjected to region delimitation processing to obtain the spatial coordinate range of key corneal regions. The spatial coordinate range is matched with the preliminary corneal optical image data to obtain an image region. Pixel feature analysis is performed on the image region to obtain the degree of brightness variation and edge continuity of the pixels. Based on the degree of brightness change, the exposure parameters of the preliminary corneal optical image data are adjusted to obtain the adjusted exposure parameters; Based on the degree of edge continuity, the image resolution parameters of the preliminary corneal optical image data are adjusted to obtain the adjusted image resolution parameters; Based on the adjusted exposure parameters and the adjusted image resolution parameters, the preliminary corneal optical image data is re-encoded to obtain the final corneal optical image data.
7. The method according to claim 1, characterized in that, The process of removing invalid data points from the corneal tomographic data caused by noise, motion artifacts, or signal loss to obtain valid corneal tomographic data includes: Signal stability analysis is performed on the corneal tomography data to obtain the signal fluctuation amplitude. Numerical statistical processing is performed on the signal fluctuation amplitude to obtain the signal stability standard. Data points whose signal fluctuation amplitude exceeds the signal stability standard are regarded as noise invalid data points. A continuity analysis is performed on the positional relationship of adjacent data points in the corneal tomography data to obtain the positional deviation value between adjacent data points. The distribution statistics of all the positional deviation values are performed to obtain the positional continuity standard. Data points whose positional deviation values exceed the positional continuity standard are regarded as invalid data points of motion artifacts. Signal intensity detection processing is performed on each data point in the corneal tomography data to obtain the signal intensity value of each data point. Range statistics are performed on all the signal intensity values to obtain the signal intensity standard. Data points whose signal intensity values are lower than the signal intensity standard are regarded as invalid data points with missing signals. The invalid data points are obtained by removing noise-related invalid data points, motion artifact-related invalid data points, and signal-deficient invalid data points from the corneal tomography data. The valid data points are then integrated according to the original order of the corneal tomography scan to obtain valid corneal tomography data.
8. A system for constructing corneal surface shapes, characterized in that, include: The segmentation module is used to scan the cornea using an optical coherence tomography (OCT) device to obtain raw corneal scan data, and to segment the raw corneal scan data according to a preset corneal anatomical structure to obtain corneal tomographic scan data. The marking module is used to remove invalid data points in the corneal tomography data caused by noise, motion artifacts or signal loss, so as to obtain valid corneal tomography data. Based on the spatial coordinate system of the optical coherence tomography device, the module marks the spatial position of each valid data point in the valid corneal tomography data to obtain valid corneal tomography data with position information. The adjustment module is used to acquire preliminary corneal optical image data, and based on the effective corneal tomographic scan data with position information, adjust the image parameters of the preliminary corneal optical image data to obtain the final corneal optical image data. The recognition module is used to perform region segmentation processing on the final corneal optical image data to obtain a corneal image, and to perform feature recognition on the corneal image to obtain the three-dimensional coordinates of each feature point of the cornea; The processing module is used to perform three-dimensional modeling of the corneal surface based on the three-dimensional coordinates using a triangular mesh generation algorithm to obtain a three-dimensional mesh model of the corneal surface, and to smooth the three-dimensional mesh model using a Gaussian smoothing algorithm to obtain the shape of the corneal surface.
9. An electronic device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the method for constructing the corneal surface shape as described in any one of claims 1 to 7.
10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the method for constructing the corneal surface shape as described in any one of claims 1 to 7.
