A keratoconus analysis system based on dynamic corneal texture

Abstract

The present invention requests protection for a keratoconus analysis system based on dynamic corneal texture, which includes an image acquisition unit, an image preprocessing unit, a feature extraction unit, and a feature screening unit connected in sequence; a multi-time-point corneal dynamic image texture analysis method is proposed. To ensure that the selected features have the optimal diagnostic ability, the most diagnostically valuable feature set is screened through the recursive feature elimination method, laying a foundation for the high-precision identification of FFKC. The machine learning classification system includes a segmentation module, a model training module, a model testing module, and a performance evaluation module, which are used to obtain a machine learning classifier; the performance of the model is evaluated through the ROC curve and the confusion matrix to ensure high sensitivity and high specificity. A multi-modal recognition platform integrating dynamic texture and corneal biomechanical parameters is constructed, which sequentially includes a data input module, a processing module, a model module, and an output module, realizing efficient FFKC identification and auxiliary diagnosis.

Description

Technical Field

[0001] The present invention relates to the field of medical technologies, and more specifically, to a keratoconus analysis system based on dynamic corneal texture. Background Art

[0002] Keratoconus (KC) is a progressive corneal ectatic disease, characterized by progressive thinning and bulging of the cornea, resulting in irregular astigmatism and significant visual acuity decline, seriously affecting the quality of life of patients. If not treated in time, it may even lead to blindness. According to research, the global prevalence of keratoconus is about 1 / 2000 to 1 / 1000, and it is particularly common in the young population, bringing a significant burden to society and the economy. With the rapid popularization of refractive surgery and the surge in the number of surgeries, some patients have developed iatrogenic corneal ectasia complications after surgery, and the failure to timely identify occult keratoconus is considered one of the main reasons for its occurrence.

[0003] Fruste form keratoconus (FFKC) is defined as the normal contralateral eye of a patient with clinically diagnosed keratoconus in one eye. The eye appears normal under slit-lamp examination and no corneal topographic abnormalities are detected. According to the 2015 Global Expert Consensus on Keratoconus, "there is no true unilateral keratoconus". Although FFKC currently has no clinical symptoms and signs, it is at a high risk of developing into keratoconus to a large extent. Research shows that although the cause of keratoconus has not been fully clarified, the current consensus is that its occurrence and development are closely related to changes in the regional biomechanical properties of the cornea. Even though no abnormal signs have been found clinically in FFKC, subtle biomechanical changes may have already begun inside the cornea.

[0004] The early identification of FFKC is of great significance for preventing disease progression and avoiding inappropriate refractive surgery. On the one hand, early intervention can delay or prevent disease progression through techniques such as corneal collagen cross-linking; on the other hand, accurately identifying FFKC can prevent high-risk patients from undergoing corneal refractive surgery that may induce corneal ectasia. However, due to the subtle abnormal changes in FFKC, its detection and diagnosis remain a major challenge.

[0005] Existing technologies mainly rely on morphological and biomechanical evaluations. Among them, the main corneal morphological evaluation devices are Pentacam and optical coherence tomography (OCT). This type of method identifies abnormalities by measuring static parameters such as corneal curvature and thickness. Specifically, it includes corneal topography analysis: used to evaluate the anterior surface morphology and provide curvature distribution and asymmetry information; keratoconus percentage index (KISA%): comprehensive parameters such as corneal curvature asymmetry and central curvature value; wavefront aberration analysis: measuring high-order aberrations and evaluating optical quality changes; and OCT imaging: evaluating the integrity and regularity of each layer structure, especially the Bowman layer. These technologies mainly focus on the morphological characteristics of the cornea, and assist in diagnosis by identifying morphological abnormalities with diagnostic significance. However, due to the extremely small morphological changes in the early stage of FFKC, the sensitivity of existing morphological parameters to its detection is limited, and there is not enough sensitivity and specificity to capture early subtle changes.

[0006] It can be seen that the current FFKC diagnostic methods mainly rely on morphological analysis and simple biomechanical parameters, which have many limitations. Morphological methods only focus on the surface characteristics of the cornea and lack sensitivity to changes in the internal structure of the cornea. FFKC lesions are caused by internal collagen fiber disorder, which first manifests as changes in internal microlesions, while there are no obvious abnormalities in the existing morphological parameters. Although the traditional biomechanical parameters provided by the visual corneal biomechanical analyzer (Corvis ST, CVS) can evaluate the dynamic response of the cornea, they are mainly affected by central corneal thickness and intraocular pressure, and are mainly global parameters. They lack the ability to finely analyze local changes in the cornea, resulting in limited diagnostic accuracy.

[0007] In recent years, with the advancement of medical imaging technology, accurate assessment of corneal morphology and biomechanical properties has become possible. As a new type of corneal biomechanical assessment device, CVS stimulates corneal deformation through airflow pulses and records its dynamic response, providing a new method to quantify corneal biomechanical properties. At the same time, artificial intelligence (AI) technology is increasingly used in the field of medical diagnosis, providing new possibilities for the processing and analysis of complex medical data. Radiomics, as a technology for extracting high-dimensional quantitative features from medical images, has shown great potential in oncology and other medical fields. At present, radiomics still faces challenges in identifying early biomechanical abnormalities in FFKC, and lacks a stable and interpretable modeling method.

[0008] In this context, how to analyze FFKC based on corneal dynamic deformation data is the technical problem to be solved by this application. Summary of the invention

[0009] In view of the deficiencies in the prior art, the present invention proposes a keratoconus analysis system based on dynamic corneal texture, which combines CVS, advanced imaging genomics analysis and machine learning algorithms to extract rich texture features from corneal deformation images at multiple time points, so as to be able to deeply analyze the internal structure and dynamic deformation behavior of the cornea, and contribute to the diagnosis of forme fruste keratoconus.

[0010] On the one hand, the present invention provides a keratoconus analysis system based on dynamic corneal texture, and the technical solution is as follows:

[0011] A keratoconus analysis system based on dynamic corneal texture includes an image acquisition unit, an image preprocessing unit, a feature extraction unit and a feature screening unit connected in sequence, and a machine learning classifier is arranged in the feature screening unit.

[0012] The image acquisition unit acquires the initial moment image, the first flattening moment image and the maximum deformation moment image of the cornea through a visual corneal biomechanics analyzer.

[0013] The image preprocessing unit extracts the region of interest (ROI) from each moment image respectively.

[0014] The feature extraction unit extracts a number of texture features from the region of interest of each moment image respectively.

[0015] The feature screening unit screens out a feature subset from the texture features of each moment image by using the recursive feature elimination method through the machine learning classifier.

[0016] On the second aspect, the present invention provides a machine learning classification system, and the technical solution is as follows:

[0017] The machine learning classification system includes a segmentation module, a model training module, a model testing module and a performance evaluation module connected in sequence.

[0018] Obtain a data set including normal eye samples and forme fruste keratoconus eye samples, and the data segmentation module divides the data set into a training set and a test set.

[0019] The model training module constructs and optimizes different types of machine learning classifiers based on the training set.

[0020] The model testing module uses the trained machine learning classifier to predict the test set, so as to obtain a prediction result.

[0021] The performance evaluation module calculates performance metrics according to the prediction results, so as to obtain the machine learning classifier with the best performance metrics.

[0022] In a third aspect, the present invention provides a keratoconus analysis platform based on the fusion of dynamic corneal texture and biomechanical parameters, and the technical solution is as follows:

[0023] It includes a data input layer, a data processing layer, a model layer and an output layer connected in sequence.

[0024] The image acquisition unit is arranged in the data input layer. The data input layer is used to receive the corneal dynamic images collected by a visual corneal biomechanical analyzer and calculate biomechanical parameters. The image acquisition unit collects the initial moment image, the first flattening moment image and the maximum deformation moment image of the cornea from the corneal dynamic images and then transmits them to the data input layer.

[0025] The image preprocessing unit and the feature extraction unit are arranged in the data processing layer. The data processing layer also includes a biomechanical parameter processing module for processing biomechanical parameters. The data processing layer is used to extract radiomics texture features from the corneal dynamic images and perform standardization processing on the biomechanical parameters.

[0026] The feature screening unit is arranged in the model layer. The model layer includes a first classification model constructed based on texture features, a second classification model constructed based on biomechanical parameters, and a fusion module for fusing the prediction results of the first classification model and the second classification model.

[0027] The output layer outputs the keratoconus diagnosis result, the key feature analysis result and the visual diagnosis report according to the prediction result of the fusion module.

[0028] In summary, the above technical solution has the following beneficial effects: The keratoconus analysis system based on dynamic corneal texture of the present invention can significantly improve the early detection rate of FFKC, reduce the misdiagnosis rate, and provide a reliable and non-invasive diagnostic tool for clinicians. By early identifying high-risk patients with corneal ectasia, the occurrence of serious complications after corneal refractive surgery can be greatly avoided, and clinical decisions can be optimized. At the same time, the simplified process of a single device reduces the examination burden on patients, improves the screening efficiency, provides a scientific basis for the early intervention and personalized treatment plan formulation of FFKC, and is of great significance for improving the diagnosis and treatment level of corneal diseases and the visual health of patients. Brief Description of the Drawings

[0029] Figure 1 It is a schematic diagram of images of three key time points of the cornea collected by a visual corneal biomechanical analyzer;

[0030] Figure 2 It is a schematic diagram of extracting features from the initial moment image, normalizing all features, and outputting a feature matrix;

[0031] Figure 3Schematic diagrams of ROC curves and confusion matrices for three different machine learning models;

[0032] Figure 4 Schematic diagram of the ROC curve of existing CVS biomechanical parameters. Specific implementation manners

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention.

[0034] The present invention aims to solve the key problems such as insufficient sensitivity, low specificity and dependence on comprehensive evaluation of multiple devices in the early diagnosis of forme fruste keratoconus (FFKC). The existing FFKC diagnosis methods mainly rely on corneal morphology and corneal biomechanics examinations, lacking in-depth analysis of the internal structure and dynamic deformation process of the cornea, resulting in low early detection rate and high misdiagnosis rate, thus delaying the best treatment opportunity and increasing the risk of serious complications such as corneal ectasia after refractive surgery. In addition, the current diagnosis process is cumbersome, often requiring the combination of multiple examination devices and technologies, and relying on the clinical experience of physicians, which not only increases the burden on patients but also reduces the clinical efficiency, restricting the wide development of early screening for FFKC.

[0035] Research shows that corneal biomechanical changes play a key role in the development of FFKC. Early lesions first manifest as changes in local biomechanical properties rather than obvious morphological abnormalities. The dynamic response process of the cornea when subjected to external forces (such as air pulses) contains rich biomechanical information, which is of great value for the identification of early FFKC. However, traditional diagnosis methods only focus on extracting limited biomechanical parameters from corneal surface deformation, unable to comprehensively capture the subtle changes and dynamic response characteristics of deep corneal tissues, resulting in insufficient sensitivity to early lesions, especially in the early stage of FFKC where the morphology is normal but the biomechanics has changed.

[0036] In view of the above problems, the present invention proposes a keratoconus analysis system based on dynamic corneal texture, which combines a visual corneal biomechanics analyzer (Corvis ST, CVS), advanced imaging omics analysis and machine learning algorithms to extract rich texture features from corneal deformation images at multiple time points and deeply analyze the internal structure and dynamic deformation behavior of the cornea. The present invention screens the most diagnostically valuable features by the recursive feature elimination method, constructs a high-accuracy FFKC intelligent recognition model to achieve precise capture of fine microbiomechanical changes. In addition, the present invention integrates time-domain information, simultaneously analyzes the corneal characteristics at the initial moment, the first flattening moment and the maximum deformation moment, comprehensively evaluates the corneal biomechanical characteristics, transforms from empirical diagnosis to precise quantitative diagnosis, improves the sensitivity and specificity of early screening, and simplifies the diagnostic process.

[0037] According to a first aspect of the present invention, there is provided a keratoconus analysis system based on dynamic corneal texture, which includes an image acquisition unit, an image preprocessing unit, a feature extraction unit and a feature screening unit connected in sequence, and a machine learning classifier is arranged in the feature screening unit.

[0038] As Figure 1 shown, the image acquisition unit acquires the initial moment image, the first flattening moment image and the maximum deformation moment image of the cornea through a visual corneal biomechanics analyzer. The visual corneal biomechanics analyzer uses a standardized air flow pulse to record the dynamic deformation process of the cornea under the action of external force. The specific steps include: fixing the subject's head on the device bracket; adjusting the device to the center of the subject's cornea and focusing; starting the device and applying a standard air flow pulse; recording the corneal deformation process to obtain a high-speed camera image sequence; extracting images at three key time points from the image sequence: the initial moment, the first flattening moment and the maximum deformation moment.

[0039] The image preprocessing unit extracts the region of interest (ROI) from each moment image respectively; the extraction of the region of interest is realized through the following process: obtaining the grayscale image of each moment according to the image of each moment; establishing a coordinate system, automatically detecting the front and back surfaces of the cornea through a visual corneal biomechanics analyzer and obtaining the coordinates; using edge detection and curve fitting methods to segment the corneal region image: removing the noise of the corneal region image and enhancing the contrast, so as to obtain the region of interest of each moment image. The image preprocessing unit realizes the preprocessing steps by using matlab code, and ensures the consistency of the data format by parsing the original 8-bit grayscale image of the visual corneal biomechanics analyzer. Specifically, the edge detection is to detect the edge of the cornea through the Canny edge detection algorithm or the Sobel operator. The corneal boundary is smoothed by polynomial fitting, morphological opening and closing operations are used to remove extra noise, and the corneal region is automatically cropped to output a standardized region of interest.

[0040] As Figure 2 shown, the feature extraction unit extracts a number of texture features from the regions of interest in the images at each moment; the texture features include first-order statistical features, shape features, gray-level co-occurrence matrix features, gray-level run length matrix features, gray-level size zone matrix features, gray-level dependence matrix features, neighborhood gray-level difference matrix features, and wavelet transform features; the texture features at three different moments are all normalized and a feature matrix is output. Specifically, the feature extraction unit uses radiomics analysis technology to extract multi-dimensional texture features from the region-of-interest images. Specifically, symbolic textures are first extracted from the region-of-interest images, denoted by Texture, and then the texture features extracted from the symbolic textures are denoted by Feature. Further, the eight types of texture features extracted from each image include 464 features, and for one person, the 3 images at different moments have 1392 feature elements.

[0041] The feature screening unit uses a machine learning classifier to screen out a feature subset from the texture features of the images at each moment by means of recursive feature elimination. The process of screening out the feature subset by means of recursive feature elimination is as follows: the feature matrix and the label vector of the corresponding samples are input into the feature screening unit; after initializing the machine learning classifier, the importance scores of the features are calculated by the machine learning classifier, and all texture features are arranged in descending order of importance; the range of the number of texture features is set, and five-fold cross-validation is performed to determine the optimal number of features and output the screened texture features, thereby obtaining a feature subset. Specifically, there are 51 feature elements in the feature subset, and only 51 feature elements remain after the 1392 features of the 3 images at different moments of one person are screened by means of recursive feature elimination.

[0042] The beneficial effects of a keratoconus analysis system based on dynamic corneal texture include: different from traditional static morphological analysis or single-time-point biomechanical parameter evaluation, the analysis system of the present invention proposes a multi-time-point corneal dynamic image texture analysis method. The radiomics method is used to extract comprehensive texture features from the images, including first-order statistical features, shape features, gray-level co-occurrence matrix (GLCM), gray-level dependence matrix (GLDM), gray-level run length matrix (GLRLM), gray-level size zone matrix (GLSZM), neighborhood gray-level difference matrix (NGTDM), and wavelet transform features, to comprehensively evaluate the dynamic changes of the corneal microstructure. To ensure that the selected features have the optimal diagnostic ability, the most diagnostically valuable feature set is screened by means of recursive feature elimination (RFE), laying a foundation for the high-precision recognition of FFKC.

[0043] A keratoconus analysis system based on dynamic corneal texture of the present invention can significantly improve the early detection rate of FFKC, reduce the misdiagnosis rate, and provide a reliable and non-invasive diagnostic tool for clinicians. By early identifying high-risk patients with corneal ectasia, the occurrence of severe complications after corneal refractive surgery can be greatly avoided, and clinical decisions can be optimized. At the same time, the simplified process of a single device reduces the examination burden on patients, improves the screening efficiency, provides a scientific basis for early intervention and personalized treatment plan formulation of FFKC, and is of great significance for improving the diagnosis and treatment level of corneal diseases and the visual health of patients.

[0044] Alternative radiomics feature extraction techniques: The present invention originally preferably uses radiomics to diagnose FFKC by extracting texture features of corneal images. As an alternative, deep learning techniques, especially convolutional neural networks (CNNs), can be used to automatically learn features directly from corneal images. This method does not require manual feature design, but automatically extracts deep features of corneal images through multi-layer convolution and pooling operations, and is suitable for large-scale datasets and complex image analysis scenarios. In addition, quantitative ultrasound elastography technology can also be used as an alternative to evaluate corneal elasticity by measuring the propagation characteristics of sound waves in corneal tissue, indirectly reflecting changes in corneal biomechanics, and is suitable for clinical environments that require non-contact evaluation.

[0045] Alternative recursive feature elimination methods: The present invention originally preferably uses recursive feature elimination to screen the most diagnostically valuable texture features. As an alternative, principal component analysis (PCA) technology can be used to retain the maximum variance information of the data through dimensionality reduction and quickly extract the main features. Another alternative is to evaluate the feature importance based on tree models, such as ExtraTrees or LightGBM. These methods evaluate the feature importance by constructing multiple decision trees and are suitable for rapid screening in high-dimensional feature spaces. In addition, LASSO regression can be used to synchronize feature selection and model training, and automatically screen important features through L1 regularization, which is suitable for the case where the number of features is much larger than the number of samples.

[0046] Alternative multi-time point integration strategies: The present invention preferably integrates corneal image features at three time points. As an alternative, time series analysis methods can be used to regard the entire corneal deformation process as a continuous time series and extract dynamic change features, such as time domain information such as deformation rate and rebound characteristics. Another alternative is to use an attention mechanism to assign different weights to features at different time points and automatically learn the most diagnostically valuable time points, which is suitable for the case where the contributions of time points are uneven. The optical flow method can also be used to track the displacement field of each pixel during the corneal deformation process and calculate the local strain distribution to more directly reflect the corneal biomechanical properties.

[0047] According to a second aspect of the present invention, a machine learning classification system is provided for obtaining a machine learning classifier.

[0048] The machine learning classification system includes a segmentation module, a model training module, a model testing module, and a performance evaluation module that are connected in sequence;

[0049] A data set including normal eye samples and forme fruste keratoconus eye samples is obtained, and the data segmentation module divides the data set into a training set and a test set. The samples are obtained through a visual corneal biomechanics analyzer, and this technology can provide a variety of biomechanical parameters: the first applanation time (A1T) measures the time for the cornea to reach the first applanated state; the Ambrósio-related thickness (ARTh) calculates the spatial distribution ratio of the corneal thickness; the stress-strain index (SSI) reflects the tissue hardness; the stiffness parameter at the first applanation moment (SP-A1) evaluates the anti-deformation ability; the deformation amplitude ratio (DARatio2) measures the ratio of the deformation amplitude between the vertex and 2 mm from the periphery; the integrated reciprocal radius (IIR) evaluates the central deformation curvature; and the CVS biomechanical index (CBI) is used as a multi-parameter comprehensive score. By quantifying the external force response characteristics, these parameters evaluate the biomechanical properties of the corneal tissue such as elasticity, stiffness, and viscoelasticity.

[0050] The model training module constructs and optimizes several machine learning classifiers based on the training set; the construction and optimization of several machine learning classifiers are achieved through the following process: input the feature matrix and label vector of each sample in the training set into the model training module; construct classifiers such as a random forest classifier, a C5.0 decision tree classifier, and an extreme gradient boosting classifier; use five-fold cross-validation to optimize the parameters of each classifier, and determine the optimal parameters according to the five-fold cross-validation results; use the optimal parameters and the training set to retrain and output the trained classifier, thereby improving the model's recognition ability for FFKC;

[0051] The model testing module uses the trained machine learning classifier to predict the test set, thereby obtaining a prediction result; the prediction of the test set includes the following process: input the feature matrix of each sample in the test set and the trained classifier into the model testing module; the trained classifier classifies each sample and outputs a prediction result; compare the prediction result with the label vector of the corresponding sample to obtain a prediction probability value, and output a confusion matrix.

[0052] The performance evaluation module calculates performance metrics based on the prediction results to obtain the machine learning classifier with the best performance metrics. The process of obtaining the machine learning classifier with the best performance metrics is as follows: The performance evaluation module calculates the ROC curve based on the prediction probability values and the confusion matrix to evaluate the machine learning classifier; draws a visualization chart and outputs a performance evaluation report based on the prediction probability values, the confusion matrix, and the ROC curve, so as to obtain the machine learning classifier with the best performance metrics. Specifically, the confusion matrix and the prediction probability values are input into the performance evaluation module; the accuracy, sensitivity, specificity, positive predictive value, negative predictive value, and the ROC curve are calculated; and finally, a visualization chart is drawn and a performance evaluation report is output, so as to obtain the machine learning classifier with the best performance metrics. As Figure 3 shown, a and b are respectively the ROC curve of the random forest model and the confusion matrix made with the test set, c and d are those of the XGBoost model, and e and f are those of the C5.0 model. As Figure 4 shown is the ROC curve of the existing CVS biomechanical parameters.

[0053] The beneficial effects of the machine learning classification system include: On the basis of texture feature screening, a machine learning classifier integrating multi-time point information is constructed, including algorithms such as random forest (RF), C5.0, and extreme gradient boosting (XGBoost). The model parameters are optimized through five-fold cross-validation to improve the model's recognition ability for FFKC. The model outputs the recognition results and probability distribution of FFKC, quantifies the diagnostic confidence, and evaluates the model performance through the ROC curve (Receiver Operating Characteristic Curve) and the confusion matrix (Confusion Matrix) to ensure high sensitivity and high specificity.

[0054] According to the third aspect of the present invention, there is provided a keratoconus analysis platform based on the fusion of dynamic corneal texture and biomechanical parameters, which is used to access a keratoconus analysis system based on dynamic corneal texture and serves as an extended module for multi-modal feature integration.

[0055] It includes a data input layer, a data processing layer, a model layer, and an output layer that are connected in sequence.

[0056] An image acquisition unit is arranged in the data input layer. The data input layer is used to receive the corneal dynamic images collected by a visualization corneal biomechanical analyzer and calculate the biomechanical parameters. The image acquisition unit collects the initial moment image, the first flattening moment image, and the maximum deformation moment image of the cornea from the corneal dynamic images and then transmits them to the data input layer.

[0057] The image preprocessing unit and the feature extraction unit are arranged in the data processing layer, and the data processing layer further includes a biomechanical parameter processing module for processing biomechanical parameters. The data processing layer is used to extract radiomics texture features from the dynamic images of the cornea and perform normalization processing on the biomechanical parameters.

[0058] The feature screening unit is arranged in the model layer. The model layer includes a first classification model constructed based on texture features, a second classification model constructed based on biomechanical parameters, and a fusion module for fusing the prediction results of the first classification model and the second classification model. The input biomechanical parameters include stress and strain index and its improved version, stiffness parameter at the first applanation moment, Ambrósio relative thickness, deformation amplitude ratio, CVS biomechanical index, etc.; the data input layer is also used to input and manage the basic information of patients, etc. Specifically, the first classification model is a random forest model constructed based on texture features, and the second classification model is a linear discriminant analysis (LDA) model constructed based on corneal biomechanical parameters. Weight coefficients are assigned to them according to the performance indexes of the first classification model and the second classification model on the test set; the fusion module fuses the prediction results of the two classification models according to the weighted voting mechanism or the weighted probability mechanism; the performance indexes include but are not limited to the area under the ROC curve, accuracy rate or F1 score; the weight coefficients are calculated according to the following normalization formula ; is the weight coefficient of the i-th classification model, is the area under the ROC curve of the i-th classification model, is the sum of the areas under the ROC curves of the two classification models. Preferably, the two machine learning classifiers are a random forest classifier based on radiomics features and a linear discriminant analysis classifier based on radiomics features; the fusion module adopts a weighted voting mechanism and sets weight coefficients according to the relative performance of the models. The F1 score is the harmonic mean of precision and recall, and is used to measure the comprehensive performance of the model in positive class prediction.

[0059] Alternative to the random forest model: Besides the random forest model, various machine learning algorithms can be used as alternative solutions. The support vector machine (SVM) achieves high-precision classification by finding the optimal classification hyperplane; the deep neural network can handle more complex non-linear relationships; ensemble learning methods such as AdaBoost and Stacking can integrate the advantages of multiple classifiers; the Bayesian network is suitable for probability reasoning and uncertainty quantification, providing risk assessment for clinical decision-making.

[0060] The output layer outputs the keratoconus diagnosis result, key feature analysis result, and visual diagnosis report according to the prediction result of the fusion module. The output content of the output layer includes FFKC analysis result, risk assessment, key feature analysis, visual report, clinical suggestions, etc. This platform supports local deployment or cloud service mode. The system workflow includes data acquisition, preprocessing, feature extraction, model prediction, and result output. The platform adopts a modular design, which is convenient for maintenance and upgrade, and supports continuous learning of new data to ensure the continuous optimization of analysis performance.

[0061] The beneficial effects of a keratoconus analysis platform based on the fusion of dynamic corneal texture and biomechanical parameters include: combining the CVS texture analysis result with traditional biomechanical parameters to establish an FFKC comprehensive analysis platform. This platform integrates the internal texture features of the cornea and the global biomechanical deformation characteristics through the model layer to form a multi-dimensional FFKC evaluation system, effectively making up for the limitations of single-modal methods. Verify the analysis accuracy of the system through clinical data and continuously optimize the algorithm model to improve clinical applicability. Finally, realize the transformation from surface morphology analysis to deep tissue property analysis, establish a scientific and accurate FFKC early analysis process, and provide an objective basis for clinical decision-making and personalized treatment plan formulation.

[0062] In summary, the beneficial effects of this application include: 1. The present invention first applies the radiomics technology to the analysis of corneal dynamic deformation images, filling the gap in the existing FFKC analysis technology. By extracting deep texture features from CVS images, the present invention can capture the tiny internal corneal structure changes that cannot be recognized by traditional parameters, greatly improving the early detection rate of FFKC and reducing the risk of missed diagnosis.

[0063] 2. The present invention uses CVS to obtain high-resolution images of the cornea at three critical moments, the initial moment, the first flattening moment, and the maximum deformation moment, establishes a classifier integrated with multiple time points, and extracts the texture features inside the cornea through radiomics technology, breaking through the limitations of traditional static morphological analysis or global biomechanical parameter evaluation, breaking through the limitations of the existing technology that only focuses on a single time point or static morphology. Experiments have proved that the multi-time point model is significantly better than the single-time point model, providing a more comprehensive evaluation of the corneal dynamic deformation process.

[0064] 3. The present invention applies the radiomics analysis method to extract eight types of features from corneal images: first-order statistical features, shape features, gray-level co-occurrence matrix (GLCM) features, gray-level run length matrix (GLRLM) features, gray-level size zone matrix (GLSZM) features, gray-level dependence matrix (GLDM) features, neighborhood gray-level difference matrix (NGTDM) features, and wavelet transform features, which can comprehensively capture the tiny changes in corneal structure.

[0065] 4. The present invention screens the most analytically valuable feature subset through the Recursive Feature Elimination (RFE) method, reducing the number of features from 1,392 to 51, significantly improving the algorithm efficiency and generalization ability, while ensuring the analysis accuracy.

[0066] 5. Based on the screened features, the present invention constructs a random forest classifier integrated with multiple time points. By integrating the information of three key corneal deformation moments, it realizes high-precision recognition of FFKC, with an AUC of 0.989, significantly superior to the existing evaluation parameter (SP-A1 parameter, with an AUC of only 0.728).

[0067] 5. Combining artificial intelligence and radiomics to achieve intelligent screening of FFKC and promote the development of precision medicine. Existing FFKC screening technologies mostly rely on empirical judgments with fixed thresholds, and the diagnostic results are greatly affected by the operator's subjectivity. The present invention combines radiomics and machine learning methods to achieve data-driven intelligent FFKC diagnosis, reducing human intervention and improving consistency. The continuously optimizable AI model can continuously improve the algorithm performance with the accumulation of data, promoting the development of FFKC diagnosis towards precision medicine. 6. The analysis platform is easy to operate and promote, and is suitable for large-scale clinical screening. Traditional FFKC detection methods may need to combine multiple devices (such as Pentacam + Corvis ST + OCT), with complex inspection processes and high costs. The present invention only requires one CVS device to complete the detection, reducing the patient's examination time and economic burden, improving the screening efficiency, and making the early screening of FFKC more feasible: applicable to preoperative FFKC risk screening to reduce corneal ectasia complications after refractive surgery. It lowers the clinical operation threshold, facilitates the promotion in primary medical institutions, and improves the early diagnosis rate of FFKC.

[0068] Based on the previous research, the project team of this application has established a corneal dynamic image and patient database for the FFKC diagnosis model, and completed the complete process from data collection to model verification. The study included a sufficient number of normal eyes and FFKC eyes, used CVS to obtain high-resolution images, and appropriately allocated the training set and test set.

[0069] Through radiomics feature extraction and recursive feature elimination, we screened out the most diagnostically valuable texture features from the images of three time points of each eye. The random forest model performed well on the test set, and the diagnostic accuracy was significantly higher than that of the traditional CVS parameters. We conducted comparative experiments on the multi-time point integrated model and the single time point model. The results showed that the performance of the multi-time point model was significantly better than that of the single time point model, verifying the superiority of multi-time point information integration. In addition, the fusion of texture features and traditional biomechanical parameters further improved the diagnostic efficiency. The analysis of texture features during corneal deformation showed that there were significant differences in texture features between normal eyes and FFKC eyes at three time points, especially at the moment of maximum deformation. This shows that although FFKC has no obvious morphological abnormalities, the response pattern of its internal structure has changed after being stressed, confirming the ability of the present invention to capture subtle structural changes. Preliminary results of clinical application show that the present invention not only improves the detection rate of FFKC, but also reduces the time required for the diagnostic process, especially in pre-refractive surgery screening, showing great application value. The model's ability to identify high-risk groups early helps prevent complications of postoperative corneal ectasia and improve the safety of refractive surgery.

[0070] In general, the experimental and preliminary clinical application results fully demonstrated the feasibility and effectiveness of the present invention, provided a new technical solution for the early and accurate diagnosis of FFKC, and had broad prospects for clinical application.

[0071] The above are only preferred embodiments of the present invention. The protection scope of the present invention is not limited to the above embodiments. All technical solutions under the concept of the present invention belong to the protection scope of the present invention. It should be pointed out that for ordinary technicians in this technical field, some improvements and modifications without departing from the principle of the present invention should also be regarded as the protection scope of the present invention.

Claims

1. A keratoconus analysis system based on dynamic corneal texture, characterized in that, It includes an image acquisition unit, an image preprocessing unit, a feature extraction unit, and a feature screening unit connected in sequence. A machine learning classifier is provided in the feature screening unit; Apply an air flow pulse to the cornea. The image acquisition unit acquires the initial moment image, the first flattening moment image, and the maximum deformation moment image of the cornea through a visual corneal biomechanics analyzer; The image preprocessing unit extracts the region of interest from each moment image respectively; The feature extraction unit extracts a number of texture features from the region of interest of each moment image respectively; The feature screening unit uses the recursive feature elimination method to screen out the feature subset from the texture features of each moment image through the machine learning classifier; Normalize the texture features at three different moments and output the feature matrix; Input the feature matrix and the label vector of the corresponding sample into the feature screening unit; After initializing the machine learning classifier, calculate the importance score of the features through the machine learning classifier and arrange all texture features in descending order of importance; Set the range of the number of texture features, perform five-fold cross-validation to determine the optimal number of features and output the screened texture features, so as to obtain the feature subset; Receive the biomechanical parameters calculated by the visual corneal biomechanics analyzer, construct a first classification model based on the texture features, construct a second classification model based on the biomechanical parameters, fuse the prediction results of the first classification model and the second classification model, and output the keratoconus diagnosis result.

2. The keratoconus analysis system based on dynamic corneal texture according to claim 1, characterized in that The extraction of the region of interest is realized through the following process: Obtain the grayscale image at each moment according to each moment image; Establish a coordinate system, and automatically detect the front and back surfaces of the cornea through a visual corneal biomechanics analyzer to obtain coordinates; Use edge detection and curve fitting methods to segment the corneal region image: Remove the noise of the corneal region image and enhance the contrast, so as to obtain the region of interest of each moment image.

3. The keratoconus analysis system based on dynamic corneal texture according to claim 1, wherein, The texture features include first-order statistical features, shape features, gray-level co-occurrence matrix features, gray-level run length matrix features, gray-level size zone matrix features, gray-level dependence matrix features, neighborhood gray-level difference matrix features, and wavelet transform features.

4. A machine learning classification system, characterized in that, For obtaining the machine learning classifier according to any one of claims 1-3, the machine learning classification system includes a segmentation module, a model training module, a model testing module, and a performance evaluation module connected in sequence; Obtain a data set including normal eye samples and incipient keratoconus eye samples. The segmentation module divides the data set into a training set and a test set; The model training module constructs and optimizes a number of machine learning classifiers based on the training set; The model testing module uses the trained machine learning classifier to predict the test set, so as to obtain the prediction result; The performance evaluation module calculates the performance index according to the prediction result, so as to obtain the machine learning classifier with the best performance index.

5. A machine learning classification system according to claim 4, wherein, The construction and optimization of a number of machine learning classifiers are realized through the following process: Input the feature matrix and the label vector of each sample in the training set into the model training module; The parameters of each classifier are optimized using five-fold cross-validation, and the optimal parameters are determined according to the results of five-fold cross-validation; the optimal parameters and the training set are used to retrain and output the trained classifier.

6. A machine learning classification system according to claim 5, characterized in that, Predicting the test set includes the following processes: Input the feature matrices of each sample in the test set and the trained classifier into the model test module; The trained classifier classifies each sample and outputs the prediction results; The prediction results are compared with the label vectors of the corresponding samples to obtain the prediction probability values, and the confusion matrix is output.

7. A machine learning classification system according to claim 6, wherein Obtaining the machine learning classifier with the best performance metrics is achieved through the following process: The performance evaluation module calculates the ROC curve based on the prediction probability values and the confusion matrix for evaluating the machine learning classifier; Visualization charts are drawn based on the prediction probability values, the confusion matrix, and the ROC curve, and a performance evaluation report is output to obtain the machine learning classifier with the best performance metrics.

8. A keratoconus analysis platform based on the fusion of dynamic corneal texture and biomechanical parameters, for accessing the keratoconus analysis system based on dynamic corneal texture according to any one of claims 1-3, characterized in that It includes a data input layer, a data processing layer, a model layer, and an output layer connected in sequence; The image acquisition unit is arranged in the data input layer. The data input layer is used to receive the corneal dynamic images collected by the visual corneal biomechanical analyzer and calculate the biomechanical parameters. The image acquisition unit collects the initial moment image, the first flattening moment image, and the maximum deformation moment image of the cornea from the corneal dynamic images and transmits them to the data input layer; The image preprocessing unit and the feature extraction unit are arranged in the data processing layer. The data processing layer further includes a biomechanical parameter processing module for processing biomechanical parameters. The data processing layer is used to extract radiomics texture features from the corneal dynamic images and perform standardization processing on the biomechanical parameters; The feature screening unit is arranged in the model layer. The model layer includes a first classification model constructed based on texture features, a second classification model constructed based on biomechanical parameters, and a fusion module for fusing the prediction results of the first classification model and the second classification model; The output layer outputs the keratoconus diagnosis results, key feature analysis results, and visual diagnosis reports according to the prediction results of the fusion module.

9. The keratoconus analysis platform based on the fusion of dynamic corneal texture and biomechanical parameters according to claim 8, characterized in that, The first classification model is a random forest model constructed based on texture features, and the second classification model is a linear discriminant analysis model constructed based on corneal biomechanical parameters. Weight coefficients are assigned to them according to the performance metrics of the first classification model and the second classification model on the test set; The fusion module fuses the prediction results of the two classification models according to the weighted voting mechanism or the weighted probability mechanism; The performance metrics include but are not limited to the area under the ROC curve, accuracy, or F1 score; The weight coefficient is calculated according to the following normalization formula: ; Among them, is the weight coefficient of the i-th classification model, is the area under the ROC curve of the i-th classification model, is the sum of the areas under the ROC curves of two classification models.

Images