Corneal effective defocus quantification method and system based on pupil dynamics coupling

Abstract

The application provides a corneal effective defocus quantification method and system based on pupil dynamics coupling, and relates to the technical field of ophthalmic examination. The method comprises the following steps: acquiring corneal multi-modal image data of a target object; the corneal multi-modal image data comprises a tangential refractive power map and a corneal height map; constructing a radial distance field based on the corneal multi-modal image data, inputting the radial distance field and the tangential refractive power map into a pre-trained physical perception depth neural network model to obtain a full-defocus ring area mask; extracting a pupil area boundary based on the tangential refractive power map to obtain a pupil area mask; calculating the intersection of the full-defocus ring area mask and the pupil area mask to obtain an intra-pupil effective defocus area mask; mapping the intra-pupil effective defocus area mask to the corneal height map, and calculating the three-dimensional physical surface area and effective defocus dose of the intra-pupil effective defocus area based on a differential geometry curved surface integral algorithm. The method can quantize the effective defocus amount that truly enters the eye and plays a role in myopia control.

Description

Technical Field

[0001] The embodiments of the present invention relate to the field of ophthalmic examination technology, and in particular to a method and system for quantifying effective corneal defocus based on pupil dynamic coupling. Background Technology

[0002] Myopia among adolescents has become a global public health problem. Orthokeratology (Ortho-K) lenses, with their dual benefits of overnight wear, daytime uncorrected visual acuity restoration, and myopia progression control, have become a mainstream non-surgical intervention in clinical practice. Their core mechanism of action lies in the mechanical reshaping of the corneal epithelium through reverse geometry lens design, flattening the central corneal curvature to correct central refractive error. Simultaneously, a myopic defocus ring (PlusPower Ring, PPR) is formed in the mid-peripheral cornea. This myopic defocus signal from the peripheral retina inhibits excessive axial elongation, thus achieving myopia control.

[0003] Current clinical assessments of orthokeratology lens fit, shaping effect, and myopia control efficacy heavily rely on corneal topography for morphological acquisition and quantitative analysis. Corneal topography can non-invasively and rapidly acquire three-dimensional morphological data such as corneal anterior surface curvature, height, and refractive power distribution, making it the gold standard tool for defocus ring feature extraction and efficacy evaluation.

[0004] Routine clinical assessment and quantification methods generally focus on the overall geometric characteristics of the defocus ring as the core calculation object. Key parameters include: total area of ​​the defocus ring, ring width, inner and outer diameters, eccentricity, roundness, and radial symmetry. These macroscopic geometric indicators are extracted through threshold segmentation and boundary fitting of corneal curvature difference maps and height difference maps before and after orthokeratology. These indicators are used to determine whether lens positioning, treatment zone size, and defocus ring formation have met the standards. Some protocols further integrate lens design parameters such as treatment zone diameter and posterior optical zone diameter (BOZD) to establish a correlation model between geometric characteristics and axial elongation and refractive error changes, guiding fitting adjustments and long-term follow-up.

[0005] However, existing technologies, which focus on the overall geometric features of the defocus ring, struggle to accurately assess effective corneal defocus, hindering the clinical implementation of orthokeratology lens fitting optimization, efficacy prediction, and individualized design. Therefore, developing a novel method for quantifying effective corneal defocus is a crucial technological requirement for addressing the shortcomings in clinical assessment and improving the accuracy of myopia control. Summary of the Invention

[0006] This invention provides a method and system for quantifying effective corneal defocus based on pupil dynamic coupling. This method overcomes the shortcomings of traditional two-dimensional full-loop measurement, which lacks physiological significance and has large projection errors, by using the technical path of "physical perception segmentation + pupil coupling + three-dimensional quantification". It can quantify the effective defocus amount that actually enters the eye and plays a role in myopia control, and provides a reference for personalized diagnosis and treatment of orthokeratology lenses in clinical practice.

[0007] The first aspect of this invention provides a method for quantifying effective corneal defocus based on pupil dynamic coupling, the method comprising: Acquire corneal multimodal image data of the target object; the corneal multimodal image data includes tangential refractive power map and corneal height map; A radial distance field is constructed based on the corneal multimodal image data, and the radial distance field and the tangential refractive power map are input into a pre-trained physical perception deep neural network model to obtain a full defocus ring region mask; Based on the tangential refractive power map, the pupil region boundary is extracted to obtain a pupil region mask; Calculate the intersection of the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil; The effective defocus region within the pupil is masked onto the corneal height map, and the three-dimensional physical surface area and effective defocus dose of the effective defocus region within the pupil are calculated based on the differential geometric surface integral algorithm.

[0008] A second aspect of this invention provides a corneal effective defocus quantification system based on pupil dynamic coupling, the system comprising: The data acquisition module is used to acquire corneal multimodal image data of the target object; the corneal multimodal image data includes tangential refractive power maps and corneal height maps; The first mask determination module is used to construct a radial distance field based on the corneal multimodal image data, and input the radial distance field and the tangential refractive power map into a pre-trained physical perception deep neural network model to obtain a full defocus ring region mask; The second mask determination module is used to extract the pupil region boundary based on the tangential refractive power map to obtain a pupil region mask. The first calculation module is used to calculate the intersection of the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil. The second calculation module is used to map the mask of the effective defocus area within the pupil onto the corneal height map, and calculate the three-dimensional physical surface area and effective defocus dose of the effective defocus area within the pupil based on the differential geometric surface integral algorithm.

[0009] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes a method for quantifying effective corneal defocus based on pupil dynamic coupling as described in the first aspect of the present invention.

[0010] A fourth aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the corneal effective defocus quantification method based on pupil dynamic coupling as described in the first aspect of the present invention.

[0011] A fifth aspect of the present invention provides a computer program product, including a computer program / instructions, which are implemented by a processor as the steps in the corneal effective defocus quantification method based on pupil dynamic coupling described in the first aspect of the present invention.

[0012] Compared with the prior art, the present invention has the following beneficial effects: (1) The present invention provides a method for quantifying effective defocus of the cornea based on pupil dynamic coupling. Through the technical path of "physical perception segmentation + pupil coupling + three-dimensional quantization", it overcomes the defects of traditional two-dimensional full-ring measurement that lacks physiological significance and has large projection error. It can accurately segment the defocus ring and, combined with the dynamic range of the pupil area, realize the accurate calculation of the "effective three-dimensional physical defocus amount" that actually enters the eye. That is, it can accurately quantify the effective defocus amount that actually enters the eye and plays a role in myopia control, providing a scientific basis for personalized diagnosis and treatment of clinical corneal reshaping lenses.

[0013] (2) By introducing pupil mask constraints, the present invention eliminates the defocus area located outside the pupil light transmission range, and the calculated pupil defocus amount has higher clinical relevance.

[0014] (3) The present invention uses differential geometric surface integral to replace two-dimensional projected area calculation, which reduces the measurement error caused by corneal morphological differences and makes the quantification result closer to the true surface area of ​​the cornea.

[0015] (4) This invention integrates optical physical priors through a physical perception deep neural network model (PI-CorneaNet), which can maintain high-precision segmentation performance in complex clinical images such as low contrast and eyelid occlusion, and the algorithm is robust. Attached Figure Description

[0016] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0017] Figure 1 This is a flowchart of the steps of the corneal effective defocus quantification method based on pupil dynamic coupling provided in the embodiments of the present invention; Figure 2 This is a flowchart illustrating the corneal effective defocus quantification method based on pupil dynamic coupling provided in this embodiment of the invention. Figure 3 A schematic diagram of the pupil coupling principle of the corneal effective defocus quantification method based on pupil dynamic coupling provided in an embodiment of the present invention; Figure 4 This is a schematic diagram of the differential geometric three-dimensional quantization principle in the corneal effective defocus quantization method based on pupil dynamic coupling provided in an embodiment of the present invention. Detailed Implementation

[0018] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] Currently, the clinical evaluation of the wearing effect of orthokeratology lenses mainly relies on the analysis of corneal topography; conventional methods usually calculate the geometric characteristics of the entire "defocus ring" (such as total area, width, and off-center distance).

[0020] The inventors discovered through exploration and analysis that existing evaluation methods mainly suffer from the following technical defects: 1. “Whole-loop analysis” lacks physiological significance: Only light entering the pupil zone can form an image on the retina and generate the “peripheral myopia defocus signal” required for myopia control; existing whole-loop measurement methods include a large number of ineffective defocus areas that are blocked by the iris and cannot enter the eye, resulting in a mismatch between the measured “anatomical quantity” and the actual effective “functional quantity”.

[0021] 2. Two-dimensional projection error: The cornea is a complex aspherical surface. Existing technologies mostly calculate based on the two-dimensional projection area of ​​topographic maps, ignoring the geometric projection error introduced by the difference in corneal curvature (K value).

[0022] 3. Insufficient segmentation accuracy: The defocus ring has an irregular shape and blurred edge gradient, making it difficult for traditional threshold segmentation methods to accurately extract the boundary.

[0023] Specifically, such as Figure 1 As shown, the corneal effective defocus quantification method based on pupil dynamic coupling provided in this embodiment of the invention includes the following steps: S101, acquire corneal multimodal image data of the target object; the corneal multimodal image data includes tangential refractive power map and corneal height map.

[0024] Specifically, corneal topography can be used to obtain ocular image data of the target subject after wearing orthokeratology lenses, including a tangential power map reflecting the instantaneous curvature change of the corneal surface and an elevation map reflecting the physical height information of the corneal surface.

[0025] Corneal height mapping is a core component of corneal topography. It refers to the imaging detection results that use the best-fit sphere (BFS) of the cornea as a reference plane, obtain the three-dimensional coordinates of each point on the corneal surface through detection equipment, calculate and visually present the differences in height and undulation of each point on the cornea relative to the reference plane using color coding. It is a key basic detection method in the fields of corneal morphology assessment, eye disease screening and refractive related technologies.

[0026] Its core features include: 1. Benchmark reference system: The virtual best-fit sphere (BFS) that conforms to the overall shape of the cornea is used as the benchmark for height difference calculation, ensuring the objectivity and consistency of the detection results; 2. Color coding rules: Warm colors (red, yellow, etc.) are used to represent corneal points that are higher than the benchmark (positive height value, i.e., local bulging), green colors represent corneal points that are at the same height as the benchmark (height value close to 0), and cool colors (blue, cyan, etc.) represent corneal points that are lower than the benchmark (negative height value, i.e., local concavity), clearly distinguishing height differences through color gradients; 3. Dual surface detection dimensions: The anterior and posterior corneal height maps can be obtained separately, with the posterior surface height map having higher detection sensitivity for early corneal morphological abnormalities (such as keratoconus).

[0027] The Best Fit Sphere (BFS) refers to the virtual sphere that best fits the overall surface morphology of the cornea. It serves as the reference surface for calculating the height difference in the corneal height map, and its parameter settings directly affect the accuracy of the corneal height detection results. The anterior corneal height map refers to the detection image of the height difference between each point on the anterior corneal surface and the BFS. The posterior corneal height map refers to the detection image of the height difference between each point on the posterior corneal surface and the BFS. It is a key detection basis for the early detection of corneal morphological abnormalities.

[0028] During the fitting of orthokeratology lenses, corneal elevation mapping can assess corneal asymmetry, guide lens design and positioning, and optimize the fitting effect.

[0029] Tangential refractive power mapping (also known as tangential curvature mapping or instantaneous curvature mapping) is a type of corneal topography used to accurately represent the true local curvature and refractive power distribution at various points on the corneal surface, and is highly sensitive to subtle morphological changes. Its core principle is to calculate the instantaneous curvature of the meridian where each point is located, with the center of curvature being the true geometric center of the arc segment at that point, without forcibly aligning with the corneal optical axis.

[0030] Tangential refractive power imaging has the following key characteristics: High sensitivity: It is more effective at capturing subtle local changes than axial plots, making "steep areas steeper and flat areas flatter".

[0031] Precisely reproduces the shape: accurately reflects the local concavity and convexity of the cornea, especially suitable for peripheral areas and post-orthokeratology evaluation.

[0032] Prone to noise: Point-to-point independent calculations may result in slight "irregular" noise even in a normal cornea.

[0033] During the fitting of orthokeratology lenses (OK lenses), tangential refractive power mapping can visually display the lens positioning, flattening zone, and reversal zone morphology, guiding parameter adjustments.

[0034] In this embodiment of the invention, the tangential refractive power map of the target subject after wearing orthokeratology lenses can accurately display the magnitude and symmetry of the tangential refractive power in the central, paracentral, and peripheral areas of the cornea, and determine whether the cornea has a regular spherical / aspherical shape after reshaping; it can identify the range of the reshaping flattening area, centering offset, and off-center fit, and evaluate whether the lens positioning is centered, whether there is any offset, or whether the reshaping is off-center; it can quantify the changes in peripheral corneal refractive power, intuitively reflect the shape of the myopia defocus ring and the distribution of defocus, and be used to evaluate the myopia control optical effect of orthokeratology lenses; it can guide the adjustment of lens parameters (base curve, reversal curve, edge curve, diameter), and provide topological basis for personalized trial fitting and adaptation iteration.

[0035] In this embodiment of the invention, the corneal height map of the target subject after wearing orthokeratology lenses can intuitively present the boundaries, shape, and degree of eccentricity of the central flattened and concave area and the peripheral raised and arched area of ​​the cornea after lens wearing, quantifying the geometric deformation of the reshaping process; it can separately assess the height of the anterior and posterior corneal surfaces, screen for over-shaping, local abnormal bulging, and corneal bulging risks, and screen for subclinical keratoconus and unstable corneal morphology; it can monitor the corneal deformation recovery pattern and rebound speed after wearing lenses at night, and assess the tightness of the lens fit, tear film distribution, and adhesion effect; it can perform quantitative analysis of height difference to objectively determine whether the reshaping effect meets the standards and whether there are any local height abnormalities, avoiding contraindications to refractive surgery and corneal safety risks.

[0036] In this embodiment of the invention, after wearing orthokeratology lenses, the corneal tangential refractive power, astigmatism distribution, defocus optical effect and lens positioning are evaluated by tangential refractive power mapping; the three-dimensional morphology of the anterior and posterior surfaces of the cornea, deformation geometry and corneal biomorphological safety are evaluated by corneal height mapping; the combination of the two can achieve dual judgment of optical refractive assessment + three-dimensional morphological safety assessment.

[0037] Specifically, such as Figure 2 As shown, the data source for the entire process of the corneal effective defocus quantification method based on pupil dynamic coupling provided in this embodiment of the invention includes three types of key corneal images: Radial physical field: contains physical parameters related to the radial distribution of the cornea (used in conjunction with the subsequent PI-CorneaNet network), which will be described in detail in later steps; Tangential refractive power mapping: Reflects the tangential refractive power distribution on the corneal surface and is the core of evaluating defocus optical performance; Corneal height map: Reflects the three-dimensional undulations of the corneal surface and is the basis for calculating the three-dimensional area; This method achieves an objective quantitative assessment of the effective defocus dose and three-dimensional physical area after wearing orthokeratology lenses by performing parallel processing, mask fusion, and differential geometric integral operations on multimodal image data.

[0038] Specifically, the tangential refractive power map and radial physical field data can be used as inputs, and the following processing steps can be performed sequentially: the tangential refractive power map is converted to HSV color space and morphologically denoised to remove image noise and enhance the refractive power distribution features; combined with the radial physical field data, the processed image is converted into 4-channel tensor data and input into a pre-trained PI-CorneaNet network; the PI-CorneaNet network completes the semantic segmentation of the corneal defocus ring region and outputs a full defocus ring region mask.

[0039] Specifically, the tangential refractive power map is used as input, and the following processes are performed in sequence: the image is converted to grayscale and thresholded to initially locate the pupil region; morphological operations are performed on the segmentation results to optimize the boundary and extract the pupil contour; a pupil region mask is generated to limit the effective optical range.

[0040] In subsequent steps, a logical AND operation is performed between the full defocus ring region mask and the pupil region mask to obtain an effective defocus region mask within the pupil, thereby eliminating invalid defocus regions outside the pupil. Combining the three-dimensional topographic data provided by the corneal height map, differential geometric curve and area integration operations are performed on the effective defocus region within the pupil to calculate the three-dimensional physical area of ​​the effective defocus region. Finally, the effective defocus dose is calculated using the refractive power distribution data from the tangential refractive power map.

[0041] In this embodiment of the invention, after step S101, the method further includes: S101.5, the pseudo-color image corresponding to the tangential refractive power map is converted to the HSV color space, and the defocus ring candidate region is extracted according to the hue interval corresponding to the preset warm hue; wherein, the pixels falling into the hue interval corresponding to the preset warm hue are marked as the defocus ring candidate region; morphological opening operation is performed on the defocus ring candidate region to remove eyelid occlusion and illumination artifact interference.

[0042] In this embodiment of the invention, the tangential refractive power map can be preprocessed based on the above step S101.5 to achieve physical field enhancement.

[0043] For example, corneal multimodal image data was acquired using the Oculus Pentacam HR anterior segment analysis system, and a tangential power matrix with a resolution of 141×141 was exported as the data basis for the construction of the tangential power map and the input of the physical perception deep neural network model. In order to highlight the defocus signal, the pseudo-color image corresponding to the tangential power map was first converted to the HSV color space, and the threshold of the Hue channel corresponding to the warm tone was set to [0,30]∪[150, 180] (corresponding to the red / orange area), and the candidate region of the defocus ring with high refractive power was extracted. Then, morphological opening operation was performed on the candidate region of the defocus ring using a 5×5 elliptical structuring element to remove noise caused by eyelash occlusion and local illumination artifacts.

[0044] S102, construct a radial distance field based on the corneal multimodal image data, and input the radial distance field and the tangential refractive force map into a pre-trained physical perception deep neural network model to obtain a full defocus ring region mask.

[0045] In this embodiment of the invention, constructing a radial distance field (Radial Physics Map) based on the corneal multimodal image data includes: Using the corneal optical center or the center point of the tangential refractive power map as the reference origin, the radial distance from each pixel in the tangential refractive power map to the reference origin is calculated to obtain a two-dimensional distance matrix with the same size as the tangential refractive power map. The radial distance field is obtained by normalizing the two-dimensional distance matrix.

[0046] In this embodiment of the invention, the corneal optical center or the center point of the tangential refractive power map can be used as the reference origin. The radial distance from each pixel in the image to the reference origin is calculated to obtain a two-dimensional distance matrix with the same size as the tangential refractive power map. This distance matrix is ​​then normalized to form a radial distance field. The radial distance field is denoted as R(x,y), and its value increases monotonically with the increase of the distance between the pixel and the corneal optical center.

[0047] In this embodiment of the invention, the radial distance field is constructed according to explicit spatial geometric rules. Essentially, it is a pixel-by-pixel spatial distribution map representing the distance of each location relative to the corneal center. The construction process is based on a unified reference center, generating a spatial distribution reflecting radial distance pixel-by-pixel, and relating it to tangential refractive power. Figure 1 One-to-one correspondence.

[0048] In this embodiment of the invention, the physical perception deep neural network model (PI-CorneaNet) is trained based on the following steps: S1, obtain the sample tangential refractive power map and its corresponding sample radial distance field; mark the full defocus ring region in the sample tangential refractive power map to obtain the marked tangential refractive power map; S2, the labeled tangential refractive power map and its corresponding sample radial distance field are spliced ​​together to form a multi-channel input tensor; S3, based on the multi-channel input tensor, the convolutional neural network is iteratively trained to obtain the pre-trained physical perception deep neural network model.

[0049] In this embodiment of the invention, a tangential refractive power map and its corresponding radial distance field are obtained; a fully defocused ring region is marked in the tangential refractive power map to obtain a marked tangential refractive power map; the marked tangential refractive power map and its corresponding radial distance field are spliced ​​together to form a multi-channel input tensor; training samples are constructed based on the multi-channel input tensor.

[0050] In this embodiment of the invention, the sample label of the training samples is a pre-labeled defocus ring region. Specifically, samples with high image quality and clear defocus ring boundaries can be selected from the acquired tangential refractive power maps first, and then the defocus ring regions of these samples are pre-labeled to form a label corresponding to the tangential refractive power map. Figure 1 A corresponding supervisory label. The supervisory label is obtained by pre-annotation based on color threshold segmentation and morphological processing, combined with manual verification.

[0051] In this embodiment of the invention, a convolutional neural network is iteratively trained based on the training samples to obtain a pre-trained physical perception deep neural network model; the convolutional neural network includes an encoder and a decoder, the encoder is preferably EfficientNet-B4, and the decoder is preferably U-Net++; during the training process, a hybrid loss function composed of a weighted average of the Dice loss function and the binary cross-entropy loss function is used, and the training is terminated according to whether the segmentation accuracy of the validation set meets the preset requirements or according to the convergence of the hybrid loss function.

[0052] The physical perception deep neural network model constructed in this invention introduces a radial distance field as a physical prior channel, which, together with the tangential refractive force map, serves as the network input. Compared to the traditional 3-channel input, this invention innovatively introduces a fourth channel—the radial distance field—compared to the traditional method of using only the tangential refractive force map. Figure 3In terms of channel input, this invention innovatively introduces a fourth channel, namely the radial distance field channel. During model training, the radial distance field serves as an auxiliary input, jointly inputting with the tangential refractive power map into the segmentation model. Specifically, firstly, the tangential refractive power map is acquired, and a radial distance field corresponding to each pixel is constructed based on the image size; where each pixel value in the radial distance field represents the radial distance from that pixel to the reference center point. Subsequently, the tangential refractive power map and the radial distance field are combined along the channel dimension according to the pixel correspondence to form multi-channel input data, which is then input into the convolutional neural network for training. The radial distance field utilizes the optical principle that 'the defocused ring region is usually distributed around the corneal optical center,' enabling the network to learn the radial spatial distribution characteristics of each pixel relative to the corneal optical center while learning image color, texture, and edge features. This effectively suppresses background artifacts and non-target interference areas far from the corneal optical center, improves the automatic recognition capability of the defocused ring region, and achieves more accurate mask segmentation of the defocused ring region.

[0053] In this embodiment of the invention, channel concatenation of the tangential refractive power map and its corresponding radial distance field specifically includes: combining the three-channel tangential refractive power map and the radial distance field according to pixel correspondence in the channel dimension; wherein, the red, green, and blue channels of the tangential refractive power map constitute the first three channels of the input data, and the radial distance field constitutes the fourth channel, thereby forming an input tensor of size H×W×4. The channel concatenation is not a simple addition or multiplication of the tangential refractive power map and the radial distance field, but rather uses them as parallel feature layers of the same image on different attributes, inputting them into the model. This allows the network to simultaneously utilize refractive power distribution information and radial spatial location information during training to learn the segmentation rules for the fully defocused ring region, thereby outputting the corresponding fully defocused ring region mask.

[0054] In this embodiment of the invention, the radial distance field and the tangential refractive power map are input into the pre-trained physical perception deep neural network model to obtain a full defocus ring region mask, including: The tangential refractive force map and the radial distance field are concatenated to obtain a multi-channel input tensor. The encoder in the convolutional neural network extracts multi-scale texture features from the multi-channel input tensor to obtain features at different levels. The decoder fuses the features at different levels and outputs a binarized full defocus ring region mask (Mring) after orthokeratology lens surgery. The convolutional neural network uses EfficientNet-B4 as the encoder to extract multi-scale texture features and uses a U-Net++ architecture as the decoder to fuse deep semantics and shallow edge features in the multi-scale texture features to obtain the full defocus ring region mask.

[0055] In this embodiment of the invention, during the deep neural network segmentation process, not only are the color, texture, and boundary information of the tangential refractive power map itself utilized, but also physical prior information reflecting the spatial geometric distribution of the cornea is introduced. This allows the model's segmentation process to be simultaneously constrained by both image appearance features and the optical structure of the cornea. Specifically, the tangential refractive power map and the radial physical field correspond one-to-one in pixel coordinates, but they express different information: the tangential refractive power map characterizes the refractive power distribution at each location, while the radial physical field characterizes the radial positional relationship of each location relative to the corneal optical center. Since the defocused ring region is typically distributed in a ring around the corneal optical center, the radial physical field can assist the model in determining whether the target region is within a reasonable radial range, thereby suppressing background artifacts and non-target interference regions, improving the accuracy of mask segmentation of the defocused ring region, and achieving physical perception.

[0056] S103, based on the tangential refractive force map, extract the pupil region boundary to obtain a pupil region mask.

[0057] Specifically, based on the tangential refractive power map, an image processing algorithm based on grayscale conversion, adaptive threshold segmentation, morphological operations, contour extraction, and center constraint screening can be used to extract the closed boundary surrounding the corneal optical center and fill it to generate a binarized pupil region mask. This mask represents the optical window through which light enters the eye and is projected onto the retina.

[0058] S104, calculate the intersection of the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil.

[0059] In this embodiment of the invention, pupil coupling is as follows: Figure 3 As shown, a Boolean intersection operation is performed on the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask (Meffective) within the pupil. Figure 3 In the diagram, the first image from the left is a tangential refractive power map before wearing the orthokeratology lens, used to characterize the distribution of refractive power on the corneal surface before wearing the lens; the second image from the left is a schematic diagram of the full defocus ring region obtained after identifying the tangential refractive power map after wearing the lens using the physical perception deep neural network model in the method of this invention, wherein the model automatically segments the full defocus ring region from the corneal topography map after wearing the lens; the third image from the left is a schematic diagram of pupil region boundary extraction, which further identifies the blue ring-shaped boundary region corresponding to the pupil in the tangential refractive power map after wearing the lens, used to determine the pupil region range; the fourth image from the left is a schematic diagram of the effective defocus region within the pupil, which is obtained by intersecting the full defocus ring region identified in the second image and the pupil region extracted in the third image, and the highlighted part is the effective defocus region within the pupil range.

[0060] The colors in the image represent different refractive power levels. Blue to green represent areas with lower refractive power, yellow represents areas with intermediate or transitional refractive power, and orange to red represents areas with higher refractive power. The orange to red areas distributed in a ring shape correspond to the main distribution range of the defocus ring. The specific numerical correspondence is based on the preset color scale of the corneal topography instrument. The bright highlighted parts in the last image represent the target analysis area obtained after intersection screening.

[0061] The full defocus ring region mask is usually a complete "O" shape, while the pupil mask is circular; the intersection of the two is the "effective defocus area within the pupil". Light in this area can enter the eye through the pupil and form a myopic defocus spot on the periphery of the retina, thereby generating a biological signal that inhibits axial elongation; the part blocked by the pupil (outside the intersection) is considered ineffective defocus; therefore, the effective defocus area within the pupil excludes the ineffective defocus part blocked by the iris and only retains the defocus area within the pupil diameter that can generate actual myopia control signals.

[0062] Boolean intersection is expressed by the following formula: M effective =M ring ∩M pupil In the formula, M effective A mask for the effective defocus area within the pupil; M ring For the entire defocus ring region mask; M pupil A mask for the pupil area.

[0063] S105, the effective defocus area mask within the pupil is mapped onto the corneal height map, and the three-dimensional physical surface area and effective defocus dose of the effective defocus area within the pupil are calculated based on the differential geometric surface integral algorithm.

[0064] In this embodiment of the invention, three-dimensional manifold quantization is as follows: Figure 4 As shown, the two-dimensional mask of the effective defocus area within the pupil is mapped back to the original corneal height map. Since the cornea is a curved surface, the traditional pixel counting method calculates the projected area, which underestimates the actual area of ​​the surrounding steep regions. This invention utilizes the principle of surface integral of the first kind in differential geometry, introduces a gradient compensation factor, calculates the corresponding three-dimensional surface element area for each pixel within the effective defocus area of ​​the pupil, and integrates and accumulates the three-dimensional surface element area of ​​each pixel to finally obtain the true three-dimensional physical surface area and weighted effective defocus dose of the effective defocus area within the pupil.

[0065] The area of ​​each pixel's three-dimensional surface element is calculated using the following formula: ; in, The area of ​​a three-dimensional curved surface element for each pixel; Let be the gradient vector of the height field.

[0066] The three-dimensional physical surface area of ​​the effective defocus region within the pupil is calculated using the following formula: ; in, It is the three-dimensional physical surface area; This represents the height value of the corneal elevation map at coordinates (x, y). , These are the partial derivatives of the height value in the x and y directions, respectively. This refers to the effective defocus area within the pupil.

[0067] With a preoperative tangential refractive power map obtained, the effective defocus dose of the effective defocus region within the pupil is calculated using the following formula: ; in, Effective defocus dose; A mask for the effective defocus area within the pupil; The difference between the postoperative and preoperative tangential refractive power maps at coordinates x, y is given. Let be the gradient vector of the height field.

[0068] Based on the same inventive concept, embodiments of the present invention also provide a corneal effective defocus quantification system based on pupil dynamic coupling, the system comprising: The data acquisition module is used to acquire corneal multimodal image data of the target object; the corneal multimodal image data includes tangential refractive power maps and corneal height maps; The first mask determination module is used to construct a radial distance field based on the corneal multimodal image data, and input the radial distance field and the tangential refractive power map into a pre-trained physical perception deep neural network model to obtain a full defocus ring region mask; The second mask determination module is used to extract the pupil region boundary based on the tangential refractive power map to obtain a pupil region mask. The first calculation module is used to calculate the intersection of the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil. The second calculation module is used to map the mask of the effective defocus area within the pupil onto the corneal height map, and calculate the three-dimensional physical surface area and effective defocus dose of the effective defocus area within the pupil based on the differential geometric surface integral algorithm.

[0069] Optionally, the system further includes: The preprocessing module is used to convert the pseudo-color image corresponding to the tangential refractive power map to the HSV color space, and extract the defocus ring candidate region according to the hue interval corresponding to the preset warm hue; wherein, the pixels falling into the hue interval corresponding to the preset warm hue are marked as the defocus ring candidate region; and perform morphological opening operation on the defocus ring candidate region to remove eyelid occlusion and illumination artifact interference.

[0070] Optionally, the first mask determination module is used to: Using the corneal optical center or the center point of the tangential refractive power map as the reference origin, the radial distance from each pixel in the tangential refractive power map to the reference origin is calculated to obtain a two-dimensional distance matrix with the same size as the tangential refractive power map. The radial distance field is obtained by normalizing the two-dimensional distance matrix.

[0071] Optionally, the physical perception deep neural network model is trained based on the following steps: Obtain the sample tangential refractive power map and its corresponding sample radial distance field; mark the full defocus ring region in the sample tangential refractive power map to obtain the marked tangential refractive power map; The labeled tangential refractive power map and its corresponding sample radial distance field are spliced ​​together to form a multi-channel input tensor. The convolutional neural network is iteratively trained based on the multi-channel input tensor to obtain the pre-trained physical perception deep neural network model.

[0072] Optionally, the second mask determining module is used for: Based on the tangential refractive power map, an image processing algorithm based on grayscale conversion, adaptive threshold segmentation, morphological operation, contour extraction, and center constraint screening is used to extract the closed boundary around the corneal optical center and fill it to generate a binarized pupil region mask.

[0073] Optionally, the second computing module is used for: The effective defocus dose in the effective defocus area within the pupil is calculated using the following formula:

[0074] in, Effective defocus dose; A mask for the effective defocus area within the pupil; The difference between the postoperative and preoperative tangential refractive power maps at coordinates x, y is given. Let be the gradient vector of the height field.

[0075] The first calculation module is used for: A Boolean intersection operation is performed on the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil.

[0076] The embodiment of the corneal effective defocus quantification system based on pupil dynamic coupling provided by this invention can be applied to any device with data processing capabilities, such as a computer or other similar device. The system embodiment can be implemented through software, hardware, or a combination of both. Taking software implementation as an example, as a logical device, it is formed by the processor of any data processing device loading the corresponding computer program instructions from non-volatile memory into memory for execution.

[0077] Based on the same inventive concept, embodiments of the present invention also provide an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the steps of the corneal effective defocus quantification method based on pupil dynamic coupling as described in any of the above embodiments.

[0078] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps in the corneal effective defocus quantification method based on pupil dynamic coupling described in any of the above embodiments.

[0079] Based on the same inventive concept, embodiments of the present invention provide a computer program product, including a computer program / instruction, which, when executed by a processor, implements the steps in the corneal effective defocus quantification method based on pupil dynamic coupling described in any of the above embodiments.

[0080] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0081] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, apparatus, or computer program products. Therefore, embodiments of the present invention can take the form of entirely hardware embodiments, entirely software embodiments, or embodiments combining software and hardware aspects. Furthermore, embodiments of the present invention can take the form of computer program products implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0082] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods, terminal devices (apparatus), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable terminal device, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0083] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable terminal device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0084] These computer program instructions can also be loaded onto a computer or other programmable terminal device to cause a series of operational steps to be performed on the computer or other programmable terminal device to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal device for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0085] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0086] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0087] The present invention provides a detailed description of a method and system for quantifying effective corneal defocus based on pupil dynamic coupling. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A corneal effective defocus quantification system based on pupil dynamic coupling, characterized in that, The system includes: The data acquisition module is used to acquire corneal multimodal image data of the target object; the corneal multimodal image data includes tangential refractive power maps and corneal height maps; The first mask determination module is used to construct a radial distance field based on the corneal multimodal image data, and input the radial distance field and the tangential refractive power map into a pre-trained physical perception deep neural network model to obtain a full defocus ring region mask; The second mask determination module is used to extract the pupil region boundary based on the tangential refractive power map to obtain a pupil region mask. The first calculation module is used to calculate the intersection of the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil. The second calculation module is used to map the mask of the effective defocus area inside the pupil onto the corneal height map, and calculate the three-dimensional physical surface area and effective defocus dose of the effective defocus area inside the pupil based on the differential geometric surface integral algorithm. The first mask determination module is used for: Using the corneal optical center or the center point of the tangential refractive power map as the reference origin, the radial distance from each pixel in the tangential refractive power map to the reference origin is calculated to obtain a two-dimensional distance matrix with the same size as the tangential refractive power map. The radial distance field is obtained by normalizing the two-dimensional distance matrix.

2. The corneal effective defocus quantification system based on pupil dynamic coupling according to claim 1, characterized in that, The system also includes: The preprocessing module is used to convert the pseudo-color image corresponding to the tangential refractive power map to the HSV color space, and extract the defocus ring candidate region according to the hue interval corresponding to the preset warm hue; wherein, the pixels falling into the hue interval corresponding to the preset warm hue are marked as the defocus ring candidate region; and perform morphological opening operation on the defocus ring candidate region to remove eyelid occlusion and illumination artifact interference.

3. The corneal effective defocus quantification system based on pupil dynamic coupling according to claim 1, characterized in that, The physical perception deep neural network model is trained based on the following steps: Obtain the sample tangential refractive power map and its corresponding sample radial distance field; mark the full defocus ring region in the sample tangential refractive power map to obtain the marked tangential refractive power map; The labeled tangential refractive power map and its corresponding sample radial distance field are spliced ​​together to form a multi-channel input tensor. The convolutional neural network is iteratively trained based on the multi-channel input tensor to obtain the pre-trained physical perception deep neural network model.

4. The corneal effective defocus quantification system based on pupil dynamic coupling according to claim 1, characterized in that, The second mask determination module is used for: Based on the tangential refractive power map, an image processing algorithm based on grayscale conversion, adaptive threshold segmentation, morphological operation, contour extraction, and center constraint screening is used to extract the closed boundary around the corneal optical center and fill it to generate a binarized pupil region mask.

5. The corneal effective defocus quantification system based on pupil dynamic coupling according to any one of claims 1-4, characterized in that, The second calculation module is used for: The effective defocus dose in the effective defocus area within the pupil is calculated using the following formula: ； in, Effective defocus dose; A mask for the effective defocus area within the pupil; The difference between the postoperative and preoperative tangential refractive power maps at coordinates x, y is given. Let be the gradient vector of the height field.

6. The corneal effective defocus quantification system based on pupil dynamic coupling according to claim 1, characterized in that, The first calculation module is used for: A Boolean intersection operation is performed on the full defocus ring region mask and the pupil region mask to obtain the effective defocus region mask within the pupil.

7. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the corneal effective defocus quantification system based on pupil dynamic coupling as described in any one of claims 1-6.

8. A readable storage medium having a computer program stored thereon, characterized in that, When executed by a processor, the computer program implements the corneal effective defocus quantization system based on pupil dynamic coupling as described in any one of claims 1-6.

9. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the corneal effective defocus quantization system based on pupil dynamic coupling as described in any one of claims 1-6.

Images