Cornea conus severity automatic grading detection method and related device

Abstract

The invention discloses a keratoconus severity automatic grading detection method and a related device, and relates to the technical field of medical data processing, and the method comprises the steps: obtaining cornea image data, extracting cornea biomechanical feature data and cornea morphological feature data, inputting the feature data into a variational auto-encoder, and obtaining a variational auto-encoder; the method comprises the following steps: analyzing cornea feature data distribution by introducing potential variables to obtain a mean value and a variance of the potential variable distribution, finally inputting the mean value and the variance of the potential variable distribution into a Bayesian Gaussian mixture distribution model, carrying out unsupervised clustering on the cornea feature data, and outputting the cornea conus severity level of cornea image data. According to the scheme, the potential variable space of the cornea features is generated through the variational automatic encoder, richer data representation of the keratoconus quantitative features is learned, the potential law or structure of keratoconus severity grading is captured, and the keratoconus severity grade can be accurately detected.

Description

Technical Field

[0001] The present application relates to the field of medical data processing technology, and in particular to a method for automatically grading and detecting the severity of keratoconus and a related device. Background Art

[0002] Keratoconus is an eye disease characterized by corneal dilation, central thinning, and forward protrusion into a cone shape. It often causes highly irregular myopic astigmatism, and in the late stage, acute corneal edema and scarring will occur, with a significant decrease in vision, and it often occurs during adolescence. So far, the cause of keratoconus is still unclear, resulting in an incompletely clear diagnosis and treatment system. However, in the early stages of keratoconus, the biomechanical properties of the cornea undergo subtle changes. When the disease progresses to a certain stage, the corneal morphology will show obvious changes. Therefore, there are many factors that affect the progression of keratoconus, and the relationship is complex and changeable.

[0003] With the development of artificial intelligence technology, applying artificial intelligence technology to the medical field can promote the rapid development of medicine. Relevant studies have shown that the use of AI technology can significantly improve the level of ophthalmic diagnosis and treatment. At present, most of the intelligent detection solutions for keratoconus using artificial intelligence technology are supervised methods, which rely on a large amount of labeled data. However, the cost of manually labeled data is high, and young doctors are also confused about some difficult diseases and it is difficult to accurately manually label categories. Therefore, we urgently hope to use new technologies to further explore the potential application of unsupervised methods in improving the accuracy of intelligent detection of keratoconus, and provide valuable references for the evaluation of the late progression of keratoconus. Summary of the invention

[0004] The purpose of this application is to provide a method and related device for automatic grading detection of keratoconus severity, which can accurately capture the potential laws or characteristics of keratoconus severity grading and improve the accuracy of intelligent keratoconus detection.

[0005] To achieve the above objectives, this application provides the following solutions:

[0006] In a first aspect, the present application provides a method for automatically grading and detecting the severity of keratoconus, comprising the following steps:

[0007] Acquire corneal image data and extract corneal feature data; the corneal feature data includes corneal biomechanical feature data and corneal morphological feature data.

[0008] The corneal feature data is input into the variational autoencoder, and the distribution of the corneal feature data is analyzed by introducing latent variables to obtain the mean and variance of the latent variable distribution.

[0009] The mean and variance of the latent variable distribution are input into the Bayesian Gaussian mixture distribution model, the corneal feature data are clustered unsupervisedly, and the severity level of keratoconus in the corneal image data is output.

[0010] Optionally, acquiring corneal image data and extracting corneal feature data specifically includes the following steps:

[0011] The corneal force deformation video is obtained, and the corneal force deformation video is sampled once at a preset time interval to obtain a plurality of frames of corneal deformation images.

[0012] An unsupervised semantic segmentation algorithm is used to extract the true corneal contour from each frame of corneal deformation image.

[0013] The spline theory is used to fit the changes of corneal contour in each frame of corneal deformation image, and the equation of corneal contour change curve is obtained.

[0014] According to the corneal contour change curve equation, a number of corneal biomechanical characteristic data are determined.

[0015] With the corneal vertex as the center, the corneal section Scheimpflug image is collected and analyzed to obtain several corneal morphological feature data.

[0016] Optionally, before inputting the corneal feature data into the variational autoencoder and analyzing the distribution of the corneal feature data by introducing latent variables to obtain the mean and variance of the latent variable distribution, the method for automatically grading the severity of keratoconus further includes the following steps:

[0017] The historical corneal feature data in the historical database is obtained, and the mean and variance of the latent variable distribution are obtained based on the variational autoencoder.

[0018] The latent variables are obtained by sampling according to the mean and variance of the latent variable distribution.

[0019] Through decoding, the latent variables are mapped to the original data space to obtain reconstructed data.

[0020] Based on the ELBO loss function, the parameters of the variational autoencoder are optimized; the ELBO loss function includes reconstruction loss and KL divergence loss; the reconstruction loss is used to measure the difference between the historical corneal feature data and the reconstructed data; the KL divergence loss is used to measure the difference between the mean and logarithmic variance of the latent variable distribution and the mean and logarithmic variance of the historical corneal feature data distribution.

[0021] Optionally, the ELBO loss function is as follows:

[0022]

[0023] in, is the ELBO loss function value, θ is the decoder parameter, is the encoder parameter, and x is the input data of the variational autoencoder; is the reconstruction loss, p θ (x|z) is the data distribution of the latent variable z, is the potential space distribution of the variational autoencoder output, is the KL divergence loss, and p(z) is the prior distribution.

[0024] The reconstruction loss can be calculated as follows:

[0025]

[0026] Where D is the dimension of data x, x i is the i-th dimension feature of data x, To reconstruct data The i-th dimension feature of .

[0027] The KL divergence loss can be calculated as follows:

[0028]

[0029] Among them, μ i and σ i are the mean and variance of the i-th dimension of the latent variable z output by the variational encoder, respectively.

[0030] Optionally, before inputting the mean and variance of the latent variable distribution into the Bayesian-Gaussian mixture distribution model, performing unsupervised clustering on the corneal feature data, and outputting the keratoconus severity level of the corneal image data, the method for automatic grading detection of keratoconus severity further includes the following steps:

[0031] The mean and variance of the latent variable distribution obtained based on the variational autoencoder are used as the input of the Bayesian Gaussian mixture distribution model, and the probability density function parameters of the Bayesian Gaussian mixture distribution model are randomly initialized.

[0032] The EM algorithm is used to estimate the probability density function parameters of the Bayesian Gaussian mixture distribution model and maximize the probability density function of the Bayesian Gaussian mixture distribution model.

[0033] Optionally, the probability density function of the Bayesian Gaussian mixture distribution model is as follows:

[0034]

[0035] Where x is a k-dimensional column vector, representing the input data of the Bayesian Gaussian mixture distribution model; μ is a k-dimensional column vector, representing the mean vector of the Bayesian Gaussian mixture distribution model, Σ is a k×k-dimensional covariance matrix, (2π)k is the volume of the unit sphere in k-dimensional space, |Σ| is the determinant of the covariance matrix, and T represents the transpose operation of the matrix.

[0036] In a second aspect, the present application provides a device for automatically grading and detecting the severity of keratoconus, comprising:

[0037] The corneal feature data extraction module is used to obtain corneal image data and extract corneal feature data; the corneal feature data includes corneal biomechanical feature data and corneal morphological feature data.

[0038] The latent variable distribution analysis module is used to input corneal feature data into the variational autoencoder, analyze the distribution of corneal feature data by introducing latent variables, and obtain the mean and variance of the latent variable distribution.

[0039] The keratoconus severity determination module is used to input the mean and variance of the latent variable distribution into the Bayesian Gaussian mixture distribution model, perform unsupervised clustering on the corneal feature data, and output the keratoconus severity level of the corneal image data.

[0040] In a third aspect, the present application provides a computer device comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the steps of the method for automatically grading the severity of keratoconus described above.

[0041] In a fourth aspect, the present application provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the method for automatically grading the severity of keratoconus described above.

[0042] In a fifth aspect, the present application provides a computer program product, including a computer program, which, when executed by a processor, implements the steps of the method for automatically grading the severity of keratoconus described above.

[0043] According to the specific embodiments provided in this application, this application discloses the following technical effects:

[0044] The present application provides a method for automatic grading detection of the severity of keratoconus and a related device, the method comprising: acquiring corneal image data, extracting corneal biomechanical feature data and corneal morphological feature data, inputting the feature data into a variational autoencoder, analyzing the distribution of corneal feature data by introducing latent variables, obtaining the mean and variance of the latent variable distribution, and finally inputting the mean and variance of the latent variable distribution into a Bayesian-Gaussian mixture distribution model, performing unsupervised clustering on the corneal feature data, and outputting the severity level of keratoconus of the corneal image data. The above-mentioned scheme of the present application is based on the biomechanical characteristics and morphological characteristics of the cornea, generates a latent variable space of corneal features through a variational autoencoder, learns a richer data representation of the quantitative characteristics of keratoconus, captures the potential rules or structure of keratoconus severity grading, and then establishes a keratoconus severity grading model through a Bayesian-Gaussian mixture distribution model, which can accurately detect the severity level of keratoconus, assist doctors in early intervention, and effectively improve the quality of vision of adolescents. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0046] Figure 1 A flowchart of a method for automatically grading and detecting the severity of keratoconus provided in one embodiment of the present application.

[0047] Figure 2 This is a flowchart of step A1 in a method for automatically grading the severity of keratoconus provided in one embodiment of the present application.

[0048] Figure 3 A schematic diagram of the structure of a variational autoencoder in a method for automatically grading the severity of keratoconus provided in one embodiment of the present application.

[0049] Figure 4 A flowchart for optimizing variational autoencoder parameters in a method for automatically grading and detecting the severity of keratoconus provided in one embodiment of the present application.

[0050] Figure 5 A flowchart for estimating the parameters of a Bayesian-Gaussian mixture distribution model in a method for automatically grading the severity of keratoconus provided in one embodiment of the present application.

[0051] Figure 6 A schematic diagram of the functional modules of a device for automatically grading and detecting the severity of keratoconus provided in one embodiment of the present application.

[0052] Figure 7 A schematic diagram of the structure of a computer device provided in one embodiment of the present application. DETAILED DESCRIPTION

[0053] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0054] In order to make the above-mentioned objects, features and advantages of the present application more obvious and easy to understand, the present application is further described in detail below with reference to the accompanying drawings and specific implementation methods.

[0055] The embodiment of the present application provides a method for automatically grading and detecting the severity of keratoconus. The method is executed by a computer device, and can be executed by a computer device such as a terminal or a server alone, or can be executed by a terminal and a server together. In the embodiment of the present application, Figure 1 As shown, the method comprises the following steps:

[0056] A1. Obtain corneal image data and extract corneal feature data; the corneal feature data includes corneal biomechanical feature data and corneal morphological feature data. Figure 2 As shown, step A1 specifically includes the following steps:

[0057] A11. Obtain corneal force deformation video, sample the corneal force deformation video once at a preset time interval, and obtain several frames of corneal deformation images.

[0058] A12. Use an unsupervised semantic segmentation algorithm to extract the true corneal contour in each frame of corneal deformation image.

[0059] A13. Use spline theory to fit the changes in corneal contour in each frame of corneal deformation image to obtain the corneal contour change curve equation.

[0060] A14. According to the corneal contour change curve equation, a number of corneal biomechanical characteristic data are determined.

[0061] A15. With the corneal vertex as the center, collect corneal section Scheimpflug images and analyze to obtain several corneal morphological feature data.

[0062] Specifically, the corneal deformation video is collected. For the corneal deformation video, the 31.88ms video data is sampled every 0.23ms to obtain 140 frames of corneal deformation images; the unsupervised semantic segmentation method is used to extract the real corneal contour in the corneal deformation image, and 115200 (576*200) corneal contour pixel points are obtained; the spline theory is used to fit the corneal contour change curve equation, and 23 corneal biomechanical feature data are calculated. With the corneal vertex as the center, the corneal fault Scheimpflug image is collected, and 65 corneal morphological feature data can be obtained by analysis through related equipment.

[0063] The 23 corneal biomechanical characteristic data include: first flattening full contour length change, second flattening full contour length change, maximum indentation full contour length change, first flattening time, second flattening time, maximum indentation time, first flattening depth of thinnest point, second flattening depth of thinnest point, maximum indentation depth of thinnest point, first flattening length, second flattening length, peak distance, relative displacement of thinnest point (1mm), relative displacement of thinnest point (2mm), maximum curvature, maximum indentation curvature radius, curvature integral of thinnest point, first flattening rate, second flattening rate, first flattening deformation area, second flattening deformation area, maximum indentation deformation area and thickness of thinnest point of cornea; 65 corneal morphological characteristic data can be directly obtained according to the output of corneal tomography Scheimpflug image through relevant equipment. The above 23 corneal biomechanical characteristic data and 65 corneal morphological characteristic data together constitute 88 dimensions of corneal characteristic data.

[0064] A2. Input the corneal feature data into the variational autoencoder, analyze the distribution of the corneal feature data by introducing latent variables, and obtain the mean and variance of the latent variable distribution. Figure 3 As shown in the structure, the above variational autoencoder adopts a fully connected structure, and the network nodes of the encoder and decoder are 88-128-256-2 and 2-256-128-88 respectively. Among them, 88 represents the dimension of corneal feature data, 256 and 128 represent the number of hidden neurons in the deep network, and 2 is the dimension of the sampling layer.

[0065] A3. Input the mean and variance of the latent variable distribution into the Bayesian Gaussian mixture distribution model, perform unsupervised clustering on the corneal feature data, and output the severity level of keratoconus in the corneal image data.

[0066] Specifically in this embodiment, before executing step A2, the method for automatically grading the severity of keratoconus also includes a process of optimizing the variational autoencoder parameters, such as Figure 4 As shown, the process includes the following steps:

[0067] B1. Obtain the historical corneal feature data in the historical database, and obtain the mean and variance of the latent variable distribution based on the variational autoencoder. In this step, the historical corneal feature data x is mapped to a low-dimensional latent space, and the features of the input data are gradually extracted through the variational autoencoder network, the mean μ and variance σ of the low-dimensional latent variable are calculated, and random noise ε is generated from the standard normal distribution N(0,1).

[0068] B2. Sampling based on the mean and variance of the latent variable distribution to obtain the latent variable. Using the sampling function, resample the mean μ and variance σ of the above latent variable to obtain the latent variable z.

[0069] B3. Through decoding, the latent variables are mapped to the original data space to obtain reconstructed data. The decoder based on the variational autoencoder reconstructs the vectors in the latent space to obtain reconstructed samples similar to the original input data.

[0070] B4. Based on the ELBO loss function, optimize the parameters of the variational autoencoder; the ELBO loss function includes reconstruction loss and KL divergence loss; the reconstruction loss is used to measure the difference between the historical corneal feature data and the reconstructed data; the KL divergence loss is used to measure the difference between the mean and logarithmic variance of the latent variable distribution and the mean and logarithmic variance of the historical corneal feature data distribution. In this embodiment, the ELBO loss function is shown as follows:

[0071]

[0072] in, is the ELBO loss function value, θ is the decoder parameter, is the encoder parameter, and x is the input data of the variational autoencoder; is the reconstruction loss, p θ (x|z) is the data distribution of the latent variable z, is the potential space distribution of the variational autoencoder output, is the KL divergence loss, and p(z) is the prior distribution.

[0073] The reconstruction loss can be calculated as follows:

[0074]

[0075] Where D is the dimension of data x, x i is the i-th dimension feature of data x, To reconstruct data The i-th dimension feature of .

[0076] The KL divergence loss can be calculated as follows:

[0077]

[0078] Among them, μ i and σ i are the mean and variance of the i-th dimension of the latent variable z output by the variational encoder, respectively.

[0079] Specifically in this embodiment, before executing step A3, the method for automatically grading the severity of keratoconus also includes a process of estimating the parameters of the Bayesian-Gaussian mixture distribution model, such as Figure 5 As shown, the process includes the following steps:

[0080] C1. Use the mean and variance of the latent variable distribution obtained based on the variational autoencoder as the input of the Bayesian Gaussian mixture distribution model, and randomly initialize the probability density function parameters of the Bayesian Gaussian mixture distribution model.

[0081] C2, using the EM algorithm to estimate the probability density function parameters of the Bayesian Gaussian mixture distribution model, and maximize the probability density function of the Bayesian Gaussian mixture distribution model. In this step, the probability density function of the model is maximized by alternating the expectation step (E step) and the maximization step (M step) until the model parameters converge or reach a predetermined number of iterations. In this embodiment, the probability density function of the Bayesian Gaussian mixture distribution model is shown below:

[0082]

[0083] Where x is a k-dimensional column vector, representing the input data of the Bayesian Gaussian mixture distribution model; μ is a k-dimensional column vector, representing the mean vector of the Bayesian Gaussian mixture distribution model, Σ is a k×k-dimensional covariance matrix, (2π) k is the volume of the unit sphere in k-dimensional space, |Σ| is the determinant of the covariance matrix, T represents the transpose operation of the matrix, (x-μ) T Σ -1 (x-μ) is the square of the distance of the vector (x-μ) relative to the covariance matrix Σ.

[0084] In the above embodiments of the present application, the biomechanical properties and morphological characteristics of keratoconus are comprehensively considered, a latent variable space is generated through a variational autoencoder, the structural changes in the feature data are accurately captured, and the relevant features of the severity of keratoconus are identified. Then, based on the Bayesian-Gaussian mixture distribution model, a precise detection model for the severity of keratoconus is established through probabilistic modeling, which realizes automatic and accurate grading detection of keratoconus, shows good performance and accuracy, and can assist doctors in early intervention and improve visual quality.

[0085] The present application also provides an application scenario, which applies the above-mentioned automatic grading detection method for the severity of keratoconus. Specifically: The automatic grading detection method for the severity of keratoconus provided in this embodiment can be applied in the intelligent grading scenario of keratoconus. The intelligent grading scenario of keratoconus includes a data acquisition link, a feature extraction link and a grading prediction link; the corneal image data enters the feature extraction link from the data acquisition link, and the potential feature representation of the keratoconus sample data is learned by introducing latent variables through the variational autoencoder, and the mean and variance of the distribution of the latent variables are extracted, and then enter the downstream grading prediction link based on the Bayesian Gaussian mixture distribution model.

[0086] Based on the same inventive concept, the embodiment of the present application also provides a device for implementing the above-mentioned method for automatically grading and detecting the severity of keratoconus. The implementation scheme for solving the problem provided by the device is similar to the implementation scheme recorded in the above-mentioned method, so the specific limitations in the embodiments of one or more devices for automatically grading and detecting the severity of keratoconus provided below can refer to the limitations of the method for automatically grading and detecting the severity of keratoconus above, and will not be repeated here.

[0087] In an exemplary embodiment, Figure 6 As shown, a device for automatically grading and detecting the severity of keratoconus is provided, comprising the following modules:

[0088] The corneal feature data extraction module is used to obtain corneal image data and extract corneal feature data; the corneal feature data includes corneal biomechanical feature data and corneal morphological feature data.

[0089] The latent variable distribution analysis module is used to input corneal feature data into the variational autoencoder, analyze the distribution of corneal feature data by introducing latent variables, and obtain the mean and variance of the latent variable distribution.

[0090] The keratoconus severity determination module is used to input the mean and variance of the latent variable distribution into the Bayesian Gaussian mixture distribution model, perform unsupervised clustering on the corneal feature data, and output the keratoconus severity level of the corneal image data.

[0091] certainly, Figure 6 The architecture shown is only exemplary and can be omitted according to actual needs when implementing different functions. Figure 6 One or at least two components of the system shown.

[0092] In an exemplary embodiment, a computer device is provided. The computer device may be a server or a terminal. The internal structure diagram thereof may be as follows: Figure 7As shown. The computer device includes a processor, a memory, an input / output interface (Input / Output, referred to as I / O) and a communication interface. The processor, the memory and the input / output interface are connected through a system bus, and the communication interface is connected to the system bus through the input / output interface. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program and a database. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The database of the computer device is used to store historical corneal feature data. The input / output interface of the computer device is used to exchange information between the processor and an external device. The communication interface of the computer device is used to communicate with an external terminal through a network connection. When the computer program is executed by the processor, the automatic grading detection method for the severity of keratoconus proposed above can be implemented.

[0093] Those skilled in the art will understand that Figure 7 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0094] In an exemplary embodiment, a computer device is further provided, including a memory and a processor, wherein a computer program is stored in the memory, and the processor implements the steps in the above-mentioned method embodiments when executing the computer program.

[0095] In an exemplary embodiment, a computer-readable storage medium is provided, storing a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0096] In an exemplary embodiment, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the steps in the above method embodiments are implemented.

[0097] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant regulations.

[0098] Those of ordinary skill in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to the memory, database or other medium used in the embodiments provided in the present application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM may be in various forms, such as static random access memory (SRAM) or dynamic random access memory (DRAM).

[0099] The database involved in each embodiment provided in this application may include at least one of a relational database and a non-relational database. The non-relational database may include a distributed database based on blockchain, etc., but is not limited thereto. The processor involved in each embodiment provided in this application may be a general-purpose processor, a central processing unit, a graphics processor, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, etc., but is not limited thereto.

[0100] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0101] This article uses specific examples to illustrate the principles and implementation methods of this application. The description of the above embodiments is only used to help understand the method and core ideas of this application. At the same time, for those skilled in the art, according to the ideas of this application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting this application.

Claims

1. A method for automatically grading the severity of keratoconus, characterized in that: include: Acquire corneal image data and extract corneal feature data; the corneal feature data includes corneal biomechanical feature data and corneal morphological feature data; Inputting the corneal feature data into a variational autoencoder, analyzing the distribution of the corneal feature data by introducing latent variables, and obtaining the mean and variance of the latent variable distribution; The mean and variance of the latent variable distribution are input into a Bayesian-Gaussian mixture distribution model, unsupervised clustering is performed on the corneal feature data, and the severity level of corneal keratoconus in the corneal image data is output.

2. The method for automatically grading the severity of keratoconus according to claim 1, characterized in that: Acquire corneal image data and extract corneal feature data, including: Acquire a corneal force deformation video, and sample the corneal force deformation video once at a preset time interval to obtain a plurality of frames of corneal deformation images; An unsupervised semantic segmentation algorithm is used to extract the true corneal contour in each frame of corneal deformation image. The spline theory is used to fit the changes of corneal contour in each frame of corneal deformation image to obtain the equation of corneal contour change curve; According to the corneal contour change curve equation, a plurality of corneal biomechanical characteristic data are determined; With the corneal vertex as the center, the corneal section Scheimpflug image is collected and analyzed to obtain several corneal morphological feature data.

3. The method for automatically grading the severity of keratoconus according to claim 1, characterized in that: Before inputting the corneal feature data into the variational autoencoder and analyzing the distribution of the corneal feature data by introducing latent variables to obtain the mean and variance of the latent variable distribution, the method for automatically grading the severity of keratoconus further includes: Obtain historical corneal feature data from the historical database, and obtain the mean and variance of the latent variable distribution based on the variational autoencoder; Sampling according to the mean and variance of the latent variable distribution to obtain a latent variable; Through decoding, the latent variables are mapped to the original data space to obtain reconstructed data; Based on the ELBO loss function, the parameters of the variational autoencoder are optimized; the ELBO loss function includes reconstruction loss and KL divergence loss; the reconstruction loss is used to measure the difference between the historical corneal feature data and the reconstructed data; the KL divergence loss is used to measure the difference between the mean and logarithmic variance of the latent variable distribution and the mean and logarithmic variance of the historical corneal feature data distribution.

4. The method for automatically grading the severity of keratoconus according to claim 3, characterized in that: The ELBO loss function is shown as follows: in, is the ELBO loss function value, θ is the decoder parameter, is the encoder parameter, and x is the input data of the variational autoencoder; is the reconstruction loss, p θ (x|z) is the data distribution of the latent variable z, is the potential space distribution of the variational autoencoder output, is the KL divergence loss, p(z) is the prior distribution; The reconstruction loss can be calculated by the following formula: Where D is the dimension of data x, x i is the i-th dimension feature of data x, To reconstruct data The i-th dimension feature of The KL divergence loss can be calculated by the following formula: Among them, μ i and σ i are the mean and variance of the i-th dimension of the latent variable z output by the variational encoder, respectively.

5. The method for automatically grading the severity of keratoconus according to claim 3, characterized in that: Before inputting the mean and variance of the latent variable distribution into the Bayesian-Gaussian mixture distribution model, performing unsupervised clustering on the corneal feature data, and outputting the keratoconus severity level of the corneal image data, the keratoconus severity automatic grading detection method further includes: Using the mean and variance of the latent variable distribution obtained based on the variational autoencoder as input to the Bayesian Gaussian mixture distribution model, and randomly initializing the probability density function parameters of the Bayesian Gaussian mixture distribution model; The EM algorithm is used to estimate the probability density function parameters of the Bayesian Gaussian mixture distribution model and maximize the probability density function of the Bayesian Gaussian mixture distribution model.

6. The method for automatically grading the severity of keratoconus according to claim 4, characterized in that: The probability density function of the Bayesian Gaussian mixture distribution model is shown as follows: Where x is a k-dimensional column vector, representing the input data of the Bayesian Gaussian mixture distribution model; μ is a k-dimensional column vector, representing the mean of the Bayesian Gaussian mixture distribution model, Σ is a k×k-dimensional covariance matrix, (2π) k is the volume of the unit sphere in k-dimensional space, |Σ| is the determinant of the covariance matrix, and T represents the transpose operation of the matrix.

7. An automatic grading detection device for keratoconus severity, characterized in that: include: A corneal feature data extraction module, used to obtain corneal image data and extract corneal feature data; the corneal feature data includes corneal biomechanical feature data and corneal morphological feature data; A latent variable distribution analysis module, used for inputting the corneal feature data into a variational autoencoder, analyzing the distribution of the corneal feature data by introducing latent variables, and obtaining the mean and variance of the latent variable distribution; The keratoconus severity determination module is used to input the mean and variance of the latent variable distribution into the Bayesian-Gaussian mixture distribution model, perform unsupervised clustering on the corneal feature data, and output the keratoconus severity level of the corneal image data.

8. A computer device comprising: A memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the computer program to implement the method for automatically grading the severity of keratoconus according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method for automatically grading and detecting the severity of keratoconus according to any one of claims 1 to 6 is implemented.

10. A computer program product, comprising a computer program, characterized in that When the computer program is executed by a processor, the method for automatically grading and detecting the severity of keratoconus according to any one of claims 1 to 6 is implemented.

Images