A distributed corneal surface morphology measurement method and device

Abstract

This invention proposes a distributed corneal surface morphology measurement method and device. The method involves acquiring dedicated videos of the left and right corneas reflecting a Placido disk pattern; performing image fusion on the videos and filtering out several clear images that meet analysis requirements based on deep learning models; calculating the positions of the corneal vertex and each reflection point from the filtered images, constructing a point cloud model covering the entire cornea, and calculating key corneal parameters; and statistically analyzing the key corneal parameters from multiple images to select the comprehensive result with the lowest dispersion. Through a distributed architecture and the application of deep learning models, this invention can improve the accuracy and reliability of corneal surface morphology measurement.

Description

Technical Field

[0001] The present invention relates to the technical field of ophthalmic medical detection, and in particular to a distributed corneal surface morphology measurement method and device. Background Art

[0002] Corneal topography is a non-invasive technique that obtains data on the corneal surface morphology and curvature. It provides key parameters such as corneal curvature (K value), astigmatism, and refractive power. It is widely used in preoperative evaluation of refractive surgery (e.g., ruling out keratoconus), contact lens fitting, diagnosis of corneal lesions (ulcers / scars, etc.), monitoring of postoperative outcomes, and precise fitting of eyeglasses. Among the currently used measurement principles, Placido disk reflection imaging has become the dominant technology in the low- and mid-range markets due to its low cost and simple structure. While Scheimpflug imaging and OCT can obtain data on the anterior and posterior corneal surfaces and thickness, their application is limited by their high cost, slow measurement speed, and complex structure.

[0003] Existing devices based on the Placido disc principle have significant drawbacks: the large disc design results in long measurement distances, which can lead to missed monitoring areas and data errors due to insufficient patient cooperation; they cannot process image blur caused by micro-movements or blinks, resulting in poor dynamic adaptability; measurement results rely on manual screening, with a low degree of automation (a single measurement takes approximately three minutes); and the devices are large (often exceeding 50 cm), requiring professional operation, and lacking portability and ease of use. Existing technologies urgently need to improve data accuracy, efficiency, and ease of use to meet clinical needs for accurate and efficient corneal topography measurement. Summary of the Invention

[0004] In response to the above problems, the present invention proposes a distributed corneal surface morphology measurement method to solve the shortcomings of traditional corneal topography equipment in data acquisition speed, integrity, dynamic adaptability and automated analysis.

[0005] The specific scheme of the present invention is as follows:

[0006] A distributed corneal surface morphology measurement method comprises the following steps:

[0007] S1, obtain videos of the left and right corneas reflecting the Placido disk pattern;

[0008] S2, performing image fusion on the video, and screening out several images that meet the analysis requirements from the continuous video based on the deep learning model. The specific operations are as follows:

[0009] S21, converting the video into an image sequence;

[0010] S22, performing homogenization processing on the image;

[0011] S23, denoising the image using a Gaussian filtering algorithm;

[0012] S24, using the Tenengrad algorithm to calculate the fuzzy coefficient of the denoised image, and preliminarily screening the image according to the fuzzy coefficient;

[0013] S25, input the initially screened images into the CNN network for scoring, and screen the images again based on the scoring results;

[0014] S3, calculating the positions of the corneal vertex and each reflection point of the filtered image, constructing a point cloud model covering the entire cornea using the vertex and each reflection point, and further calculating key parameters of the cornea based on the point cloud model;

[0015] S4, by statistically analyzing the key corneal parameters of multiple images, the arithmetic mean and standard deviation methods were used to screen out the comprehensive results with the smallest dispersion.

[0016] Extracting each frame from the video, converting it to grayscale, performing noise reduction, and calculating fuzzy coefficients effectively removes noise and interference, improving image quality and clarity, and providing more accurate image data for subsequent calculations of the corneal vertex and reflection point positions. The synergistic effect of the Tenengrad algorithm and the CNN network automatically selects the clearest images that best meet the analysis requirements, eliminating the subjectivity and errors of manual screening. This improves the automation and efficiency of the measurement, ensures the quality and representativeness of the selected images, and provides a strong foundation for subsequent corneal parameter calculations.

[0017] Furthermore, the S22 specifically includes:

[0018] S221, calculating the mean brightness, minimum brightness and maximum brightness of each image;

[0019] S222, calculating an adaptive brightness adjustment parameter for each image based on the calculation result of S221 and a preset brightness parameter;

[0020] S223: Performing a uniform brightness process on each image.

[0021] By calculating the mean brightness, minimum brightness, and maximum brightness of each image, and calculating the adaptive brightness adjustment parameters for each image based on preset brightness parameters, the brightness of each image is homogenized. This can effectively solve the problem of large brightness differences between different images, making the image sequence have more uniform and stable brightness characteristics, facilitating subsequent image processing and analysis, further improving the accuracy and reliability of corneal measurement, and providing a higher-quality image data foundation for subsequent calculation of the corneal vertex and reflection point position.

[0022] Furthermore, the S24 specifically includes:

[0023] S241, using the Sobel operator to calculate the horizontal gradient and vertical gradient of the image respectively, and calculating the square of the gradient magnitude of each pixel in the image based on the horizontal gradient and the vertical gradient of the image;

[0024] S242 , taking the central 50% area of the image as the central area, accumulating the square values of the gradient amplitudes of all pixels in the central area to obtain the blur coefficient of the image.

[0025] The Sobel operator is used to calculate the horizontal and vertical gradients of the image. Based on this, the square of the gradient amplitude for each pixel is calculated. The central 50% of the image is then accumulated to obtain the image's blur coefficient, enabling a more accurate assessment of image clarity. This method focuses on the central area of the image, avoiding the influence of edge interference factors such as eyelashes and eyelids. This makes the calculation of the blur coefficient more accurate and reliable, enabling more effective screening of clear images. This provides high-quality image data for subsequent corneal parameter calculations, improving the accuracy of measurement results.

[0026] Furthermore, the S3 specifically includes:

[0027] S31, estimation of corneal vertex position using Klyce method;

[0028] S32, based on the geometric relationship between the corneal vertex position and adjacent reflection points, calculating the axial curvature radius of each angle meridian, and then determining the three-dimensional coordinates of all reflection points;

[0029] S33, establishing a one-to-one mapping relationship between the feature points on the Placido disk image and the reflection points on the corneal surface, and constructing a point cloud model covering the entire cornea using the three-dimensional coordinates of all reflection points;

[0030] S34, calculates the key parameters of the cornea based on the point cloud model, including corneal curvature, corneal astigmatism, corneal refractive power, corneal irregularity index, corneal height map, corneal diameter, corneal eccentricity, corneal surface topography, and corneal vertex position.

[0031] By estimating the corneal vertex position using the Klyce method and combining the geometric relationships between adjacent reflection points to determine the position of each individual reflection point, key geometric features of the corneal surface can be accurately determined. By establishing a one-to-one mapping between feature points on the Placido disk image and reflection points on the corneal surface, a point cloud model covering the entire cornea is constructed. This model is then used to calculate a variety of key parameters, including corneal curvature, corneal astigmatism, and corneal refractive power.

[0032] A distributed corneal surface morphology measurement device, comprising:

[0033] The acquisition device is used to collect special videos of the Placido disk pattern reflected by the left and right corneas and transmit the collected videos to the terminal;

[0034] The terminal is used to upload the video received from the acquisition device to the server and display the data received from the server;

[0035] The server is used to perform image fusion on the received video and, based on the deep learning model, screen out multiple clear images that meet the analysis requirements from the continuous video; calculate the position of the corneal vertex and each reflection point for the screened images, use these points to construct a point cloud model covering the entire cornea, and further calculate the key parameters of the cornea based on the point cloud model; perform statistical analysis on the key corneal parameters of multiple images to screen out the comprehensive result with the smallest discreteness; and send the result to the terminal.

[0036] This distributed corneal surface morphology measurement device, equipped with an acquisition device, a terminal, and a server, automates and streamlines the corneal surface morphology measurement process. The acquisition device captures dedicated videos of the Placido disk pattern reflected from the left and right corneas and transmits them to the terminal. The terminal is responsible for uploading the video and displaying the data, while the server performs core tasks such as image fusion, clear image screening, calculation of the corneal vertex and reflection point positions, and key parameter calculations. This modular design not only improves the stability and reliability of the device but also enables the modules to work together, leveraging their respective strengths. This effectively enhances the efficiency and accuracy of corneal surface morphology measurement, providing a convenient and efficient corneal morphology measurement solution for clinical applications.

[0037] Furthermore, the acquisition device specifically includes:

[0038] Projection unit, including two Placido disks, used to measure the corneal topography data of the left and right eyes;

[0039] An imaging unit, used to collect video of light points reflected from the corneal surface and detect pupil position in real time;

[0040] an adjustment unit for automatically adjusting the position of the Placido disk according to the position of the pupil so as to achieve center alignment with the pupil;

[0041] The communication unit is used to transmit the collected video to the terminal.

[0042] The projection unit, imaging unit, adjustment unit, and communication unit within the acquisition device work together to efficiently collect and transmit corneal topography data for both eyes. The projection unit's two Placido disks measure the left and right corneal topography. The imaging unit captures video of light reflected from the corneal surface and detects pupil position in real time. The adjustment unit automatically adjusts the Placido disk's position based on pupil position to ensure center alignment, ensuring accurate and reliable measurement. The communication unit transmits the captured video to the terminal, ensuring timely data transmission and subsequent processing. This design makes the data acquisition process more automated and intelligent, improving measurement efficiency and accuracy, and providing high-quality data support for the entire corneal surface morphology measurement device.

[0043] Furthermore, the two Placido disks both use conical Placido disks with a diameter of 50mm-80mm, containing at least 16 annular bands, each of which is composed of light-transmitting micropores. Each annular band contains 16 to 36 micropores, and the micropores are distributed in concentric circles, and the spacing between adjacent annular bands gradually increases.

[0044] A conical Placido disk with a diameter of 50mm-80mm is used, and each ring contains 16 to 36 light-transmitting micropores. The micropores are distributed in concentric circles, and the spacing between adjacent rings gradually increases. This design can effectively reduce the size of the Placido disk while ensuring measurement accuracy, making it more portable and easier to operate. In addition, this structure helps to improve the quality and stability of the corneal reflection image, providing a better foundation for subsequent image processing and parameter calculation. By optimizing the structure and parameters of the Placido disk, not only the cost and complexity of the equipment are reduced, but also the efficiency and reliability of the measurement are improved, making corneal surface morphology measurement more convenient and quick, providing strong support for clinical applications and daily monitoring.

[0045] Furthermore, the imaging unit includes an LED light board. When the pupil is centered, the adjustment unit activates the LED light board to emit white light. The white light passes through the micropores into the Placido disk, forming a reflected image composed of light spots on the human cornea.

[0046] The imaging unit includes an LED light panel. After the pupil is centered, the adjustment unit activates the LED light panel to emit white light. This white light passes through the micropores into the Placido disk, forming a reflected image composed of light spots on the human cornea. This design provides a uniform and stable lighting source, ensuring the quality and clarity of the corneal reflection image. By controlling the start and stop and light intensity of the LED light panel, it can better adapt to different eye conditions and measurement requirements, improving the flexibility and adaptability of the measurement. At the same time, this lighting method also helps to improve the contrast of the corneal reflection image and the accuracy of feature point extraction, providing a higher-quality data foundation for subsequent image processing and corneal parameter calculation, further improving the accuracy and reliability of corneal surface morphology measurement.

[0047] Compared with existing technologies, the present invention offers the following advantages: the method and device can effectively improve the accuracy and reliability of corneal surface morphology measurement. By acquiring dedicated videos of the left and right corneas reflecting the Placido disk pattern and performing multi-step processing, clear images that meet the requirements can be precisely selected. A deep learning model is then used to ensure accuracy. Subsequently, a point cloud model is constructed based on the corneal vertex and various reflection points, enabling comprehensive and accurate calculation of key corneal parameters. The distributed architecture enables efficient data processing, with collaborative work between the acquisition device, terminal, and server, improving overall measurement efficiency. The use of a specific Placido disk and related algorithms further enhances the scientific and practical nature of the measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the embodiments of the present drawings or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present drawings. For ordinary technicians in this field, other drawings can be obtained based on the structures shown in these drawings without paying any creative work.

[0049] Figure 1 is a flow chart of the method of the present invention;

[0050] Figure 2 It is the projection of light spot on human cornea;

[0051] Figure 3 is a schematic diagram of the point cloud;

[0052] Figure 4 This is a schematic diagram of the acquisition equipment structure;

[0053] Figure 5 This is a schematic diagram of the Placido disk + infrared camera structure. DETAILED DESCRIPTION

[0054] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention is described and illustrated below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention. Based on the embodiments provided by the present invention, all other embodiments obtained by those of ordinary skill in the art without creative work are within the scope of protection of the present invention.

[0055] Example 1:

[0056] like Figure 1 As shown, the present invention provides a distributed corneal surface morphology measurement method, which specifically includes the following steps:

[0057] S1, acquire dedicated videos of the left and right corneas reflecting the Placido disk pattern.

[0058] A specially designed binocular, conical, dual-Placido disk acquisition device with apertures allows simultaneous acquisition of specialized videos of the Placido disk pattern reflected from the left and right corneas. This compact, lightweight, and easy-to-use device can be used by the patient themselves. The captured video is then transmitted in real time to a common terminal (such as a mobile phone or PC) via a USB or WiFi interface.

[0059] S2, performing image fusion on the video, and screening out a number of images that meet the analysis requirements, i.e., clear and correct images, from the continuous video based on a deep learning model.

[0060] The terminal then uploads the video to the server for further processing. The server first fuses the left and right eye videos. Then, based on a deep learning model optimized by the Tenengrad algorithm, it automatically selects multiple clear images from the continuous video that meet the analysis requirements, ensuring the validity and integrity of the data.

[0061] Specifically, S2 includes the following steps:

[0062] S21, video image extraction: extract each frame from the video and convert the video into a lossless image sequence.

[0063] S22, image dimensionality reduction processing: performing homogenization processing on the image.

[0064] Raw video image data is illuminated by an infrared light source and captured by an infrared camera. The captured image is a monochrome grayscale image containing only the infrared light channel. To uniformly brighten all infrared images, each image must be normalized. This involves calculating the mean, maximum, and minimum brightness of the image sequence. Based on these values, the brightness of each image is adaptively adjusted to ensure that the brightness of each image is close to or equal to the average brightness.

[0065] The specific operations are as follows:

[0066] 1) Calculate the mean brightness, minimum brightness, and maximum brightness of each image:

[0067] Mean brightness : ;

[0068] Minimum brightness : ;

[0069] Maximum brightness : ;

[0070] Where H and W represent the length and width of the image respectively;

[0071] (x, y) represents the coordinate position;

[0072] Representing an image The pixel value at (x, y) in the middle;

[0073] Indicates the sum of all pixel values.

[0074] 2) Based on the above calculation results and the brightness parameters preset by the acquisition device, calculate the adaptive brightness adjustment parameters for each image:

[0075] Set the target brightness fixed value according to the preset brightness parameters ;

[0076] Calculate the image brightness shift: ;

[0077] Determine the brightness shift range: , ;

[0078] in, Indicates the brightness shift amount;

[0079] Indicates the minimum translation amount;

[0080] Indicates the maximum translation amount.

[0081] 3) Normalize the brightness of each image: ;

[0082] in, Indicates the brightness after translation at (x, y) in the image;

[0083] The clamp function is used to limit pixel values to the range of 0 to 255.

[0084] S23, image noise reduction processing: The gray image is denoised using a Gaussian filtering algorithm to remove noise that may be generated during the video compression and transmission process, preserving as many Placido disk image details as possible.

[0085] The noise generated by compression in general images belongs to normal distribution (Gaussian distribution), so Gaussian filtering is used. The core algorithm is as follows:

[0086] ;

[0087] Among them, σ is the standard deviation, which is used to control the width of the Gaussian curve and is an important parameter for controlling the smoothness of the filter;

[0088] μ is the mean of the Gaussian function;

[0089] p(z) represents the probability density at position z;

[0090] e is the base of natural logarithms;

[0091] π is the ratio of a circle to its circumference.

[0092] S24, using the Tenengrad algorithm to calculate the fuzzy coefficient of the denoised image, and preliminarily screening the image based on the fuzzy coefficient. Specifically, the steps include:

[0093] 1) Use the Sobel operator to calculate the horizontal gradient of the image and vertical gradient ;

[0094] 2) Calculate the square of the gradient magnitude of each pixel in the image : ;

[0095] 3) Take the central 50% of the image as the center area to avoid interference from eyelashes, eyelids, etc.

[0096] 4) Accumulate the squared values of the gradient amplitudes of all pixels in the central area to obtain the Tenengrad score, which is the blur coefficient of the image. The higher the score, the clearer the image. It is calculated using the following formula:

[0097] .

[0098] S25, the images after the initial screening are input into the CNN network for scoring, and the images are screened again according to the scoring results.

[0099] The Tenengrad algorithm only judges image clarity, not image content or quality. To prevent subsequent algorithms using Tenengrad alone from mistakenly using incorrect images (such as closed eyes, squinting, eyelash occlusion, equipment failure, etc.) as input and affecting the calculation results, a small convolutional neural network (CNN) is introduced for training to work with the Tenengrad algorithm to eliminate clear but incorrect images.

[0100] 1) Using a pre-trained lightweight CNN as the backbone network (ResNet-18);

[0101] 2) Design a binary classification network that outputs a single probability value (the probability that the image is qualified);

[0102] 3) Acquire 120fps (frames per second) Placido disk projection video to ensure coverage of different ages, genders, and corneal conditions;

[0103] 4) Collect 120,000 frames of raw images containing samples of typical quality issues, such as blink occlusion, motion blur, ring fragmentation, and abnormal reflections;

[0104] 5) 70% of the images are manually annotated as a training set, and the remaining 30% of the images are used as a validation test set. After training, they work in conjunction with the Tenengrad algorithm.

[0105] To speed up the calculation, the CNN network is slower than the Tenengrad algorithm. Therefore, the images are first scored using the Tenengrad algorithm, and the top 10% of the images are selected and input into the CNN network. Finally, the 10 images with higher scores in the CNN network are selected as the basis for subsequent calculations.

[0106] S3, calculating the positions of the corneal vertex and each reflection point based on the filtered image, using these points to construct a point cloud model covering the entire cornea, and further calculating the key parameters of the cornea based on the point cloud model.

[0107] For each selected image, the device automatically calculates the position of the corneal vertex and each reflection point, using these points to construct a point cloud model covering the entire cornea. Based on this model, key parameters including corneal curvature, corneal astigmatism, corneal refractive power, corneal irregularity index, corneal height map, corneal diameter, corneal eccentricity, and corneal surface topography are further calculated.

[0108] The Klyce method is currently used to estimate the position of the corneal vertex. For the first ring (innermost ring) of the Placido disk, the distance from the outermost ring end face of the Placido disk to the camera CCD is basically equal to the distance from the Placido disk surface to the corneal vertex. Therefore, the radius of the first ring on the Placido image directly corresponds to the radius of the first reflection point on the cornea. The position of the first reflection point on the cornea can be used to estimate the radius of the corneal vertex curvature. :

[0109] ;

[0110] in, is the radius of the first reflection point on the i-th meridian;

[0111] n is the number of meridians.

[0112] After determining the corneal vertex curvature radius according to Klyce, the corneal vertex position is obtained according to the following formula :

[0113] ;

[0114] in, is the distance from the outermost ring surface of the Placido disk to the camera CCD;

[0115] is the radius of the innermost ring of the Placido disk.

[0116] Based on the geometric relationship between the corneal vertex position and adjacent reflection points, the axial curvature radius of the meridian at each angle is calculated, and then the three-dimensional coordinates of all reflection points are determined.

[0117] 1) Axial curvature radius : Reflects the curvature of the corneal surface along the axis, defined as the distance from the corneal reflection point to the intersection of the normal line at that point and the axis:

[0118] ;

[0119] ;

[0120] ;

[0121] ;

[0122] ;

[0123] ;

[0124] in, is the reflection point to be calculated;

[0125] Reflection point The corresponding annular radius of the Placido disk;

[0126] 、 are the lengths of the incident ray and the reflected ray respectively;

[0127] is the angle between the incident light and the reflected light;

[0128] Reflection point The slope angle of the tangent line.

[0129] 2) Instantaneous curvature radius: refers to the curvature radius of a local area on the corneal surface. The curvature characteristics of this local area are determined by fitting three adjacent reflection points into the same arc. Therefore, we can obtain:

[0130] ;

[0131] in, 、 、 are three adjacent reflection points;

[0132] is the instantaneous curvature radius;

[0133] is the local center of curvature.

[0134] Based on the above formula, solve the problem of each reflection point .

[0135] like Figure 3 As shown in FIG, a one-to-one mapping relationship is established between the feature points on the Placido disk image and the reflection points on the corneal surface, and the three-dimensional coordinates of all reflection points are used to construct a point cloud model covering the entire cornea.

[0136] The number of data points in the point cloud model depends on the number of feature points formed by the microholes on the specially designed Placido disk.

[0137] The key parameters of the cornea are calculated based on the point cloud model, including corneal curvature, corneal astigmatism, corneal refractive power, corneal irregularity index, corneal height map, corneal diameter, corneal eccentricity, corneal surface topography, corneal vertex position and other corneal surface morphological data.

[0138] S4, by statistically analyzing the key corneal parameters of multiple images, the arithmetic mean and standard deviation methods were used to screen out the comprehensive results with the smallest dispersion.

[0139] Example 2:

[0140] The present invention also provides a distributed corneal surface morphology measurement device, which specifically includes the following three modules:

[0141] The acquisition device is used to collect special videos of the Placido disk pattern reflected by the left and right corneas and transmit the collected videos to the terminal.

[0142] Specifically, such as Figure 4 As shown, the acquisition equipment includes:

[0143] The projection unit includes two Placido disks, a first Placido disk 1 and a second Placido disk 11, for measuring the topographic data of the left and right corneas.

[0144] Specifically, such as Figure 5 As shown, both Placido disks use a conical Placido disk with a diameter of 50mm-80mm, which contains at least 16 annular bands. The annular bands are composed of translucent micropores 10. Each annular band contains 16 to 36 micropores 10. The diameter of each micropore 10 is 0.5±0.1mm. The micropores 10 are distributed in concentric circles, and the spacing between adjacent annular bands gradually increases.

[0145] The imaging unit is used to collect videos of light spots reflected from the corneal surface and detect the pupil position in real time.

[0146] Specifically, the imaging unit includes:

[0147] At least two cameras capture video of light spots reflected from the corneal surface at a frame rate of 120 frames per second;

[0148] Infrared guidance device, including infrared illumination light source and pupil detection camera 2, for detecting pupil position in real time;

[0149] LED light board 5, when the pupil is aligned in the center, the adjustment unit starts the LED light board 5 to emit white light, and the white light passes through the microhole 10 into the Placido disk, forming a reflected image composed of light spots on the human cornea.

[0150] The adjustment unit is used to automatically adjust the position of the Placido disk according to the position of the pupil so as to achieve center alignment with the pupil.

[0151] Specifically, the adjustment unit includes a guide rail 3 , a motor 4 and a control board 6 .

[0152] The communication unit is used to transmit the collected video to the terminal.

[0153] The acquisition device is integrated into a handheld portable device. It aligns the subject's eyes through physical contact, can measure the corneal topography of both eyes at the same time, and communicates with the terminal through the USB interface 7 or WiFi interface 8. It does not have any calculation and measurement functions.

[0154] Specifically, the workflow of the acquisition device is as follows:

[0155] 1) First, the person being measured places the measuring hole of the data acquisition device 12 close to their eyes. The eyes of the person being measured are now covered by the first and second Placido plates 1, 11. The person is looking at the infrared camera 9, which is equipped with a gaze-guiding red light with a wavelength of approximately 850nm. This light can be observed by the human eye as a guide for gaze, and can also provide infrared illumination without affecting the pupil image.

[0156] 2) The control board 6 of the acquisition device has a built-in pupil position recognition algorithm, which cyclically detects the images collected by the pupil detection camera 2.

[0157] 3) After the pupil pattern is recognized in the camera image, the control board 6 drives the adjustment motor 4 to adjust the positions of the left and right placido plates through the control rails 3 so that the left and right placido plates are centered with the left and right pupils.

[0158] 4) When the pupil is centered, the control board 6 activates the device's built-in LED panels 5. These panels use white LED light sources, whose wavelengths do not conflict with the red light emitted by the infrared camera 9. Once activated, these panels emit light through micro-holes 10 in the specially designed conical Placido disk.

[0159] 5) The white light emitted by the LED light board 5 enters the Placido disk through the microhole 10, and eventually forms a reflected image composed of light spots on the human cornea.

[0160] The Placido disk structure of this device differs from traditional Placido disk designs. Traditional Placido disk optical designs typically employ either a reflective design (with a highly reflective coating on the white area of the disk) or a transmissive design (with a backlight source and a transparent material covered with an opaque coating). This traditional design aims to ensure uniform brightness distribution across the projection rings during projection, preventing uneven light intensity from causing a decrease in the image signal-to-noise ratio, which could affect corneal curvature calculations. However, due to the need to maintain the machining accuracy of the concentric rings (which prevents processing of very thin lines), the lines in traditional designs are relatively wide. This large line width prevents the Placido disk from being reduced in size or closer to the pupil to maintain line density when projected onto the cornea. This is one of the reasons why traditional corneal topographers cannot be reduced in size.

[0161] Because the circular ring design of the traditional Placido disc is ultimately intended to extract feature points on the corneal projection within the ring to calculate corneal parameters. This device directly simplifies the ring into light points to form feature points. It then uses the feature points formed by the light points to fit a single ring through ellipse calculation. Multiple fitted rings are then combined to form corneal surface parameters, which are ultimately used for corneal morphology calculation. This not only replaces the circular ring design of the traditional Placido disc, but also significantly reduces processing difficulty and precision. It no longer requires the use of translucent materials and can be molded with ordinary ABS injection molding or 3D printing, which are impossible with traditional Placido disc designs. It also significantly reduces the size of the Placido disc.

[0162] 6) After the projection images of the corneas of both eyes are formed in the Placido dish, the imaging unit records a 5-10 second video at a speed of 120 frames per second, and then transmits the video to the terminal via the USB interface 7 or the WiFi interface 8.

[0163] The terminal is used to upload the video received from the acquisition device to the server and display the data received from the server.

[0164] The terminal can be any device that supports USB communication or Wi-Fi. After downloading the video generated by the acquisition device through a standard protocol, it is uploaded to the server. After the server completes the video analysis, it sends the corneal morphology calculation results, which the terminal then displays. Compared with traditional corneal topography measurement equipment, the terminal does not require any major computing work during this process, and there are almost no hardware requirements. An ordinary mobile phone or computer can be used. In this way, corneal topography measurement no longer requires the computing and display devices required by traditional corneal topography measurement equipment. This change can not only effectively reduce the size of the corneal topography measurement equipment, but also greatly reduce the cost and size of the corneal topography measurement equipment.

[0165] The server is used to perform image fusion on the received video and, based on the deep learning model, screen out multiple clear images that meet the analysis requirements from the continuous video; calculate the position of the corneal vertex and each reflection point for the screened images, use these points to construct a point cloud model covering the entire cornea, and further calculate the key parameters of the cornea based on the point cloud model; perform statistical analysis on the key corneal parameters of multiple images to screen out the comprehensive result with the smallest discreteness; and send the result to the terminal.

[0166] To reduce the difficulty of corneal topography measurement, the present invention places the calculation part of traditional corneal topography measurement into the server side and abandons the traditional measurement method of manually screening qualified images. The server side performs measurement data quality analysis, vertex fitting, and corneal topography parameter calculation, thus solving the shortcomings of traditional corneal topography equipment such as large data errors, poor dynamic adaptability, and low degree of automation.

[0167] Thanks to its dual-disc structure and innovative automatic image screening mechanism, monocular measurement speeds are increased by 20%, and binocular measurement speeds by over 50%. Furthermore, the new Placido disk structure reduces manufacturing costs by 90% compared to traditional Placido disks. The newly designed Placido disk also reduces machining errors, improving data measurement reliability by 5-10%.

[0168] It should be noted that the present invention is not limited to the above-mentioned embodiments. The above-mentioned embodiments are merely examples, and any embodiments having substantially the same structure and effect as the technical concept within the scope of the technical solution of the present invention are all included in the technical scope of the present invention. In addition, without departing from the scope of the present invention, other embodiments that can be conceived by those skilled in the art and that combine some of the constituent elements in the embodiments are also included in the scope of the present invention.

Claims

1. A distributed corneal surface morphology measurement device, characterized in that: include: The acquisition device collects special videos of the Placido disk pattern reflected by the left and right corneas and transmits the collected videos to the terminal. Specifically, it includes: The projection unit includes two Placido disks for measuring corneal topography data of the left and right eyes. The two Placido disks are conical Placido disks with a diameter of 50 mm to 80 mm and at least 16 annular zones. The annular zones are composed of light-transmitting micropores. Each annular zone contains 16 to 36 micropores. The micropores are distributed in concentric circles, and the spacing between adjacent annular zones gradually increases. An imaging unit is used to capture video of light spots reflected from the corneal surface and detect pupil position in real time. The imaging unit includes an LED light board. When the pupil is centered, the adjustment unit activates the LED light board to emit white light. The white light passes through the micropores into the Placido disk, forming a reflected image composed of light spots on the human cornea. an adjustment unit for automatically adjusting the position of the Placido disk according to the position of the pupil so as to achieve center alignment with the pupil; A communication unit, used to transmit the collected video to the terminal; The terminal is used to upload the video received from the acquisition device to the server and display the data received from the server; The server is used to perform image fusion on the received video, use the Tenengrad algorithm to calculate the fuzzy coefficient of the denoised image, preliminarily screen the image based on the fuzzy coefficient, input the preliminarily screened image into the CNN network for scoring, and screen the image again based on the scoring result; calculate the position of the corneal vertex and each reflection point for the screened image, use these corneal vertices and each reflection point to construct a point cloud model covering the entire cornea, and further calculate the key parameters of the cornea based on the point cloud model; through statistical analysis of the key corneal parameters of multiple images, screen out the comprehensive result with the smallest discreteness; and send the result to the terminal.

2. A distributed corneal surface morphology measurement method, using the distributed corneal surface morphology measurement device according to claim 1, characterized in that: The following steps are involved: S1, obtain videos of the left and right corneas reflecting the Placido disk pattern; S2, performing image fusion on the video, and screening out several images that meet the analysis requirements from the continuous video based on the deep learning model. The specific operations are as follows: S21, converting the video into an image sequence; S22, performing homogenization processing on the image; S23, denoising the image using a Gaussian filtering algorithm; S24, using the Tenengrad algorithm to calculate the fuzzy coefficient of the denoised image, and preliminarily screening the image according to the fuzzy coefficient; S25, input the initially screened images into the CNN network for scoring, and screen the images again based on the scoring results; S3, calculating the positions of the corneal vertices and reflection points of the filtered images, constructing a point cloud model covering the entire cornea using the vertices and reflection points, and further calculating key parameters of the cornea based on the point cloud model, specifically including: S31, estimation of corneal vertex position using Klyce method; S32, based on the geometric relationship between the corneal vertex position and adjacent reflection points, calculating the axial curvature radius of each angle meridian, and then determining the three-dimensional coordinates of all reflection points; S33, establishing a one-to-one mapping relationship between the feature points on the Placido disk image and the reflection points on the corneal surface, and constructing a point cloud model covering the entire cornea using the three-dimensional coordinates of all reflection points; S34, calculates key corneal parameters based on the point cloud model, including corneal curvature, corneal astigmatism, corneal refractive power, corneal irregularity index, corneal height map, corneal diameter, corneal eccentricity, corneal surface topography, and corneal vertex position; S4, by statistically analyzing the key corneal parameters of multiple images, the arithmetic mean and standard deviation methods were used to screen out the comprehensive results with the smallest dispersion.

3. A distributed corneal surface morphology measurement method according to claim 2, characterized in that: The S22 specifically includes: S221, calculating the mean brightness, minimum brightness and maximum brightness of each image; S222, calculating an adaptive brightness adjustment parameter for each image based on the calculation result of S221 and a preset brightness parameter; S223: Performing a uniform brightness process on each image.

4. The distributed corneal surface morphology measurement method according to claim 2, characterized in that: The S24 specifically includes: S241, using the Sobel operator to calculate the horizontal gradient and vertical gradient of the image respectively, and calculating the square of the gradient magnitude of each pixel in the image based on the horizontal gradient and the vertical gradient of the image; S242 , taking the central 50% area of the image as the central area, accumulating the square values of the gradient amplitudes of all pixels in the central area to obtain the blur coefficient of the image.

Images