A refractive screening system for refractive media clarity assessment

Abstract

The application provides a refractive screening system for refractive medium clarity evaluation, and relates to the technical field of ophthalmic medical devices. The refractive screening system for refractive medium clarity evaluation comprises a multi-source acquisition module, through an adaptive eyeball tracking mechanism, a multi-modal synchronous trigger control algorithm is executed, slit lamp front scattering images, optical coherence tomography rear illumination images and polarized light scanning data are synchronously acquired, and an original multi-modal image set is generated; the multi-source acquisition module comprises an optical adaptation sub-module, a dynamic calibration sub-module and a synchronous control sub-module. Through introducing a multi-source information synchronous acquisition and cooperative processing mechanism in the refractive medium evaluation process, the optical data from different imaging modes are kept consistent in the time dimension and the space dimension, the integrity and the stability of the refractive medium information acquisition are effectively improved, and in the data processing stage, the multi-modal images are subjected to feature alignment and weighted fusion.

Description

Technical Field

[0001] This invention relates to the field of ophthalmic medical device technology, specifically to a refractive screening system for assessing the clarity of refractive media. Background Technology

[0002] The field of ophthalmic medical device technology involves the evaluation and testing of the optical properties of the human eye's refractive system, encompassing the analysis of indicators such as transparency, refractive index, scattering characteristics, and structural homogeneity of refractive media such as the cornea, lens, and vitreous humor. Refractive examination technology, through various optical imaging methods, imaging equipment, and image processing techniques, assists ophthalmologists in the early screening, monitoring, and management of refractive errors, cataracts, and other problems in clinical diagnosis and treatment. One such refractive screening system for assessing the clarity of refractive media involves collecting and analyzing the optical information of the human eye's refractive media to objectively evaluate their transparency and scattering characteristics, thereby providing quantitative evidence for refractive screening and visual quality analysis. This system can assist ophthalmologists in assessing the state of refractive media, providing crucial technical support, particularly in applications such as refractive screening and visual function testing, with the ultimate goal of improving the early detection and intervention of eye diseases.

[0003] In practical applications, existing ophthalmic refractive screening technologies often employ single or limited optical imaging methods to observe the state of the refractive media. This results in limited information acquisition, failing to comprehensively reflect the true state of the refractive media under different optical conditions. Consequently, assessment results are highly dependent on imaging conditions and operational experience. During data processing, existing technologies typically rely on two-dimensional images or single parameters for evaluation, lacking a systematic description of the spatial structure and internal changes of the refractive media. This leads to insufficient ability to identify localized opacities, slight scattering changes, and other issues. In real-world screening scenarios, this approach is prone to insufficient sensitivity in early stages of change, affecting the accuracy and stability of the screening. Furthermore, existing technologies rely heavily on manual observation or semi-automated analysis processes during evaluation, making assessment standards highly susceptible to subjective factors and compromising the consistency and repeatability of results. At the information utilization level, existing technologies often focus on single optical indicators, lacking comprehensive analytical capabilities for the multidimensional characteristics related to the refractive media. This results in insufficient information depth in the screening results, making it difficult to provide adequate data support for subsequent follow-up or long-term management. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a refractive screening system for assessing the clarity of refractive media. This system solves the problem that existing ophthalmic refractive screening technologies often rely on single or limited optical imaging methods to observe the state of refractive media in practical applications. The types of information obtained are limited, making it difficult to fully reflect the true state of refractive media under different optical conditions. Consequently, the assessment results are highly dependent on imaging conditions and operational experience.

[0005] To achieve the above objectives, the present invention is implemented through the following technical solution: a refractive screening system for assessing the clarity of refractive media, comprising the following modules: a multi-source acquisition module, which, through an adaptive eye-tracking mechanism, executes a multimodal synchronous triggering control algorithm to simultaneously acquire slit-lamp forward scattering images, optical coherence tomography post-illumination images, and polarized light scanning data, thereby generating an original multimodal image set; The multi-source acquisition module includes an optical adapter submodule, a dynamic calibration submodule, and a synchronization control submodule. The data fusion module, based on the original multimodal image set, executes a cross-modal feature registration algorithm and integrates multi-source features through a channel attention weighted fusion network to generate a registered and fused image; The data fusion module includes a feature extraction submodule, a spatial transformation submodule, and a weighted fusion submodule; The three-dimensional reconstruction module, based on the registered and fused images, executes an improved cube travel algorithm, combines a convolutional neural network to segment the boundaries of the refractive media, and applies the acoustic time-of-flight difference method to map acoustic parameters to generate a three-dimensional sharpness map of the refractive media. The 3D reconstruction module includes a boundary segmentation submodule, a parameter mapping submodule, and a mesh generation submodule; The quantitative assessment module, based on the three-dimensional sharpness map of the refractive medium, executes a multi-scale texture quantization algorithm, and inputs a regression model to calculate the medium scattering index and transparency index, generating a refractive medium sharpness assessment report; The quantitative assessment module includes a texture analysis submodule, an index calculation submodule, and a report generation submodule; The spectral analysis module, based on the refractive media clarity assessment report, executes the Raman spectral analysis algorithm, calculates the relative concentration of metabolic-related components in the media by matching characteristic peak positions, and generates refractive media metabolic characteristic indicators. The spectral analysis module includes a spectral acquisition submodule, a peak position identification submodule, and a concentration calculation submodule; The screening and assessment module, based on the refractive media clarity assessment report and metabolic characteristic indicators, executes a multi-parameter fusion assessment algorithm, outputs the refractive media transparency level and abnormality warning level, and generates refractive screening and assessment results. The screening and assessment module includes a feature fusion submodule, a grade classification submodule, and a result output submodule; The report generation module, based on the refractive screening assessment results, executes the structured report generation engine to automatically output a description of the examination conclusions and a refractive media clarity map annotation, and generates a refractive screening assessment report; The report generation module includes a rule matching submodule, a text generation submodule, and a map annotation submodule; The cloud platform interaction module, based on the refractive screening assessment report, executes an adaptive encrypted transmission protocol and synchronizes the data to the medical information platform through a standardized API interface to generate a cloud-based refractive screening file. The cloud platform interaction module includes a data encryption submodule, an interface management submodule, and a cloud storage submodule.

[0006] Preferably, the optical adapter submodule adjusts the output wavelength and light intensity distribution of the multispectral light source according to the preset slit lamp optical parameters through a wavelength adaptive control algorithm to generate a standardized illumination light field; The dynamic calibration submodule, based on the standardized illumination light field, uses a real-time eye movement compensation algorithm to drive the fine-tuning actuator to correct the optical path offset and generate calibrated optical path parameters. The synchronization control submodule executes a multimodal temporal synchronization trigger control algorithm based on the calibrated optical path parameters, synchronously starting optical coherence tomography imaging and polarization-sensitive imaging to generate the original multimodal image set.

[0007] Preferably, the feature extraction submodule performs a scale-invariant feature transformation algorithm on the original multimodal image set to extract cross-modal keypoint feature descriptors and generate a multimodal feature point description set; The spatial transformation submodule, based on the multimodal feature point description set, uses a thin-plate spline spatial transformation algorithm to establish a nonlinear mapping relationship between images of different modalities and generate a cross-modal registration transformation matrix. The weighted fusion submodule integrates feature maps of different modalities based on the cross-modal registration transformation matrix and generates a registration fusion image through a channel attention weighted fusion network.

[0008] Preferably, the boundary segmentation submodule uses a convolutional neural network segmentation algorithm to identify the anatomical boundary region of the refractive medium based on the registered and fused image, and generates a refractive medium region segmentation mask; The parameter mapping submodule performs an acoustic time-of-flight difference calculation algorithm based on the refractive medium region segmentation mask to quantify the acoustic propagation characteristics inside the refractive medium and generate an acoustic parameter distribution matrix. The mesh generation submodule uses the acoustic parameter distribution matrix and an improved cube traversal algorithm to construct a three-dimensional voxel model and generate a three-dimensional sharpness map of the refractive medium.

[0009] Preferably, the texture analysis submodule performs a multi-scale gray-level co-occurrence matrix analysis algorithm on the three-dimensional sharpness map of the refractive medium to extract texture feature vectors that reflect scattering characteristics and structural uniformity, and generates a sharpness texture feature vector set; The index calculation submodule inputs the sharpness texture feature vector set into the regression analysis model to calculate the refractive medium scattering index and transparency index, and generates a dataset of refractive medium sharpness quantification indexes. The report generation submodule generates a refractive media clarity assessment report based on the refractive media clarity quantification index dataset and through a structured data encapsulation algorithm.

[0010] Preferably, the spectral acquisition submodule uses Raman spectroscopy focusing scanning strategy to locate the target detection area based on the refractive media clarity assessment report, acquires the spectral data of the corresponding area of ​​the refractive media, and generates a target area spectral dataset. The peak position identification submodule performs a Gaussian fitting peak position analysis algorithm on the target region spectral dataset to identify characteristic peak positions related to the metabolic state of the refractive media and generate a set of metabolic characteristic peak position coordinates. The concentration calculation submodule uses the set of metabolic characteristic peak coordinates to calculate the relative concentration ratios of metabolic-related components using a regression analysis algorithm, thereby generating refractive media metabolic characteristic indicators.

[0011] Preferably, the feature fusion submodule performs a multi-parameter fusion analysis algorithm on the refractive medium clarity quantification index and metabolic characteristic index to generate a comprehensive evaluation feature vector; The grading submodule, based on the comprehensive evaluation feature vector, grades the transparency of the refractive media using preset threshold rules or statistical classification algorithms, generating a refractive media clarity grade and anomaly warning grade. The result output submodule is used to output refractive screening assessment results. These assessment results are for screening and assessment purposes and do not constitute a disease diagnosis conclusion.

[0012] Preferably, the rule matching submodule calls a preset screening prompt rule library to generate corresponding examination prompt information based on the refractive media clarity level and abnormal prompt level; Based on the examination prompt information, the text generation submodule executes a natural language generation algorithm to output a descriptive text of the refractive screening result. The atlas annotation submodule, in conjunction with the description text of the refractive screening results, annotates abnormal areas in the three-dimensional clarity atlas of the refractive media, and generates a refractive screening assessment report.

[0013] Preferably, the data encryption submodule performs encryption algorithm processing on the refractive screening assessment report to generate an encrypted data packet; The interface management submodule encapsulates the encrypted data packet using a standardized application programming interface protocol to generate cloud transmission instructions. The cloud storage submodule synchronously stores the refractive screening assessment report to the information platform according to the cloud transmission instructions, generating a cloud-based refractive screening file.

[0014] A refractive screening refractometer for assessing the clarity of refractive media includes a body. An image capture module is fixedly connected to one side of the body, and light source modules are fixedly connected to both sides of the image capture module. A display module is fixedly connected to the other side of the body for displaying a refractive media clarity map and screening assessment results. A control module is fixedly connected to the upper surface of the body for controlling the multimodal acquisition, data fusion, and screening assessment process. The refractometer is used to acquire multimodal images of refractive media and output refractive media clarity assessment results and screening prompts, but does not have disease diagnosis functions.

[0015] In summary, this application includes at least one of the following beneficial technical effects: This invention introduces a multi-source information synchronous acquisition and collaborative processing mechanism during the refractive media evaluation process. This ensures consistency of optical data from different imaging methods in both temporal and spatial dimensions, effectively improving the completeness and stability of refractive media information acquisition. In the data processing stage, feature alignment and weighted fusion of multimodal images allow for complementary utilization of structural information, scattering characteristics, and polarization properties contained in various images, avoiding evaluation biases caused by insufficient information from a single imaging source. Furthermore, by refining the boundary of the refractive media and combining it with multi-dimensional parameter mapping, a three-dimensional expression of the internal structure and sharpness characteristics of the refractive media is achieved. This expands the evaluation results from a two-dimensional representation to a three-dimensional spatial level, enhancing the overall quality and efficiency of the assessment. This approach enhances the intuitiveness and interpretability of the results. Furthermore, by conducting multi-scale quantitative analysis of the internal texture characteristics of refractive media and combining this with a statistical model to output standardized transparency and scattering indices, the assessment of refractive media clarity is transformed from empirical judgment into quantifiable and comparable data results. The introduction of spectral information analysis establishes a correlation between the optical performance of the refractive media and its internal component characteristics, expanding the assessment dimension from simple optical properties to a multi-parameter comprehensive analysis level. By fusing and judging multiple assessment results and forming a structured output, it not only improves the consistency and repeatability of screening results but also enhances the automation level and information utilization efficiency of the overall screening process. This approach is suitable for large-scale refractive screening and long-term data management applications. Attached Figure Description

[0016] Figure 1 This is a system block diagram of this application; Figure 2 This is a schematic diagram of the multi-source acquisition module of this application; Figure 3 This is a schematic diagram of the data fusion module of this application; Figure 4 This is a schematic diagram of the three-dimensional reconstruction module of this application; Figure 5 This is a schematic diagram of the quantitative evaluation module in this application; Figure 6 This is a schematic diagram of the spectral analysis module of this application; Figure 7 This is a schematic diagram of the screening and assessment module in this application; Figure 8 This is a schematic diagram of the report generation module for this application; Figure 9 This is a schematic diagram of the cloud platform interaction module of this application; Figure 10 This is a three-dimensional structural diagram of this application.

[0017] The components are: 1. Body; 2. Control module; 3. Display module. Detailed Implementation

[0018] The following is in conjunction with the appendix Figure 1-10 This application will be described in further detail.

[0019] like Figure 1-10 As shown, this embodiment of the invention provides a refractive screening system for assessing the clarity of refractive media, including the following modules: a multi-source acquisition module, which executes a multimodal synchronous triggering control algorithm through an adaptive eye-tracking mechanism to simultaneously acquire slit-lamp forward scattering images, optical coherence tomography post-illumination images, and polarized light scanning data to generate an original multimodal image set; The multi-source acquisition module includes an optical adapter submodule, a dynamic calibration submodule, and a synchronization control submodule; The data fusion module, based on the original multimodal image set, executes a cross-modal feature registration algorithm and integrates multi-source features through a channel attention weighted fusion network to generate a registered and fused image; The data fusion module includes a feature extraction submodule, a spatial transformation submodule, and a weighted fusion submodule; The 3D reconstruction module, based on the registered and fused images, executes an improved cube travel algorithm, combines a convolutional neural network to segment the boundary of the refractive medium, and applies the acoustic time-of-flight difference method to map acoustic parameters to generate a 3D sharpness map of the refractive medium. The 3D reconstruction module includes a boundary segmentation submodule, a parameter mapping submodule, and a mesh generation submodule; The quantitative assessment module, based on the three-dimensional clarity map of the refractive media, executes a multi-scale texture quantization algorithm and inputs a regression model to calculate the media scattering index and transparency index, generating a refractive media clarity assessment report; The quantitative assessment module includes a texture analysis submodule, an index calculation submodule, and a report generation submodule; The spectral analysis module, based on the refractive media clarity assessment report, executes the Raman spectral analysis algorithm, calculates the relative concentration of metabolic-related components in the media by matching characteristic peak positions, and generates refractive media metabolic characteristic indicators. The spectral analysis module includes a spectral acquisition submodule, a peak position identification submodule, and a concentration calculation submodule; The screening and assessment module, based on the refractive media clarity assessment report and metabolic characteristic indicators, executes a multi-parameter fusion assessment algorithm to output the refractive media transparency level and abnormality warning level, and generates refractive screening and assessment results. The screening and assessment module includes a feature fusion submodule, a grading submodule, and a result output submodule; The report generation module, based on the refractive screening assessment results, executes a structured report generation engine to automatically output a description of the examination conclusions and a refractive media clarity map annotation, generating a refractive screening assessment report; The report generation module includes a rule matching submodule, a text generation submodule, and a map annotation submodule; The cloud platform interaction module, based on the refractive screening assessment report, executes an adaptive encrypted transmission protocol and synchronizes to the medical information platform through a standardized API interface to generate cloud-based refractive screening records. The cloud platform interaction module includes a data encryption submodule, an interface management submodule, and a cloud storage submodule.

[0020] The optical adapter submodule adjusts the output wavelength and light intensity distribution of the multispectral light source according to the preset optical parameters of the slit lamp through a wavelength adaptive control algorithm to generate a standardized illumination light field. Based on the wavelength range and light intensity distribution requirements in the preset slit lamp optical parameters, the system calls the corneal and lens tissue optical response characteristic comparison table stored in the system, selects the target wavelength combination and corresponding light intensity ratio, for example, when evaluating corneal scattering characteristics, the 460-480nm blue light band is selected as the main light source, and when evaluating the lens, the 520-540nm green light band is selected as the main light source. The filter wheel is driven by a stepper motor to switch to the target filter group, and the driving current value of different wavelength light sources in the LED array is adjusted at the same time to make the output spectral distribution match the preset requirements, forming a standardized irradiation light field that covers the target refractive medium area and has a uniform light intensity distribution.

[0021] The dynamic calibration submodule is based on a standardized illumination light field and uses a real-time eye movement compensation algorithm to drive the fine-tuning actuator to correct the optical path offset and generate calibrated optical path parameters. Based on real-time eye image sequences acquired under a standardized illumination field, a feature point tracking algorithm is used to locate the displacement changes of the pupil edge and corneal limbus. When the movement of the eyeball in the XY plane exceeds the preset tolerance threshold, the compensation angle and displacement required by the piezoelectric ceramic micro-motion platform are calculated according to the displacement direction and amplitude. For example, if the eyeball is detected to shift to the right by 0.5 degrees, the platform is driven to translate 50 micrometers to the left and rotate counterclockwise by 0.3 degrees. The optical path offset is continuously corrected through a closed-loop feedback mechanism to ensure that the imaging optical axis is always perpendicular to the surface of the refractive medium, and the calibrated optical path parameters are generated.

[0022] Based on the calibrated optical path parameters, the synchronization control submodule executes a multimodal timing synchronization trigger control algorithm to simultaneously initiate optical coherence tomography imaging and polarization-sensitive imaging, generating the original multimodal image set.

[0023] Based on the spatial coordinates and optical focal length data in the calibrated optical path parameters, a multi-device synchronous trigger clock signal is generated. Within 5 milliseconds after the start of slit lamp illumination exposure, a scan start command is sent to the optical coherence tomography scanner. At the same time, the polarization camera is triggered to complete image acquisition within a specific exposure cycle. For example, the A-scan sequence of OCT is started at the 10th millisecond of the slit lamp exposure cycle, and the synchronous exposure acquisition of polarization images is completed within the 15th-25th millisecond cycle. This ensures that the three imaging modalities are strictly aligned in the time domain, generating an original multimodal image set with consistent timestamps.

[0024] The feature extraction submodule performs a scale-invariant feature transformation algorithm on the original multimodal image set to extract cross-modal keypoint feature descriptors and generate a multimodal feature point description set; Based on the original multimodal imaging, including slit-lamp forward scattering images, OCT post-illumination images, and polarization scan data, key point locations are detected at specific scales. For example, feature points with edge curvature changes exceeding a preset threshold are selected in the corneal region, and local extreme points of scattering intensity distribution gradients are selected in the lens region. For each key point, gradient direction histogram feature vectors are extracted from its surrounding neighborhood, forming a feature point record containing spatial coordinates, scale levels, and 128-dimensional descriptive vectors. This generates a multimodal feature point description set covering the main anatomical structures of the refractive media.

[0025] The spatial transformation submodule is based on a multimodal feature point description set and uses a thin plate spline spatial transformation algorithm to establish a nonlinear mapping relationship between images of different modalities, generating a cross-modal registration transformation matrix. Based on the coordinates and description vectors recorded in the multimodal feature point description set, a corresponding point pair relationship is established between the slit lamp image and the OCT image. For example, the center point of the corneal limbus is selected as the reference control point. When the distance error of this point in the two modes exceeds 0.5 mm, it is marked as an abnormal point pair. The coefficient matrix of the thin plate spline function is solved by the iterative weighted least squares method, where the weight of the bending energy term is set as the logarithm of the inverse of the spatial distance. After three iterations of optimization, the control point displacement mapping function is obtained, and a cross-modal registration transformation matrix is ​​generated to map the spatial coordinates of the polarization image to the space of the OCT image.

[0026] The weighted fusion submodule integrates feature maps from different modalities through a channel attention weighted fusion network based on the cross-modal registration transformation matrix to generate a registered and fused image.

[0027] A cross-modal registration transformation matrix is ​​applied to transform the slit lamp feature map, OCT feature map, and polarization feature map to a unified coordinate system. The importance weight of each feature map is calculated in the channel dimension. For example, the edge information channel of the slit lamp feature map is assigned a higher weight, and the depth information channel of the OCT feature map is assigned a medium weight. The global average and maximum values ​​of the feature maps are processed by a two-layer fully connected network. The attention weight coefficients of each channel are output by the Sigmoid activation function. The weighted feature maps are superimposed and fused in the channel dimension to generate a registered and fused image containing multi-source information.

[0028] The boundary segmentation submodule is based on the registered and fused images and uses a convolutional neural network segmentation algorithm to identify the anatomical boundary region of the refractive medium and generate a segmentation mask for the refractive medium region. Based on the structural information of the refractive media in the registered and fused images, a pre-trained convolutional neural network model is used to encode the images. The encoding path contains multiple convolutional and pooling layers to extract feature information at different scales. The decoding path restores spatial resolution through upsampling operations. An activation function is applied to the output layer to generate a pixel-level classification probability map. For example, the boundary between the anterior elastic layer and the endothelial cell layer is identified in the corneal region, and the interface between the capsule and the cortex is identified in the lens region. By setting a probability threshold, the classification results are converted into a binary mask image to generate a region segmentation mask covering the anatomical structure of each layer of the refractive media.

[0029] The parameter mapping submodule segments the refractive medium region mask, executes the sound wave time-of-flight difference calculation algorithm, quantifies the acoustic propagation characteristics inside the refractive medium, and generates an acoustic parameter distribution matrix. Based on the boundary range defined by the refractive medium region segmentation mask, an array of sound wave transmitting and receiving points is set up inside the mask. Broadband pulse signals are emitted to the target area through an ultrasonic transducer array. The time series of sound wave signals captured by each receiving point is recorded, and the time difference of sound wave flight from the transmitting point to the receiving point is calculated. Combined with known medium thickness data, the local sound velocity value is estimated based on the formula for the propagation speed of sound waves in a homogeneous medium. For example, if the time difference of flight is detected to be longer in the lens nucleus region than in the cortical region, the sound velocity value in this region is calculated to be higher than that in the surrounding region. At the same time, the sound absorption coefficient is calculated based on the signal attenuation degree, generating a two-dimensional distribution matrix containing sound velocity and sound attenuation parameters.

[0030] The mesh generation submodule uses an improved cube traversal algorithm based on the acoustic parameter distribution matrix to construct a three-dimensional voxel model and generate a three-dimensional sharpness map of the refractive medium.

[0031] The acoustic parameter distribution matrix is ​​used as the input of three-dimensional scalar field data. An improved cube traversal algorithm is used to traverse each cube cell in the regular grid. The traversal of isosurfaces is judged based on the acoustic parameter values ​​at the vertices of the cells. When the difference between the values ​​at the vertices of the cells exceeds a preset threshold, the coordinates of the intersection points of the isosurfaces and the edges of the cubes are calculated by linear interpolation. The intersection points are connected to form triangular patches. For example, a high-density triangular mesh is generated in the opaque region of the lens where the sound velocity value changes abruptly, and a sparse mesh is generated in the vitreous region where the sound attenuation coefficient is stable. Finally, a three-dimensional voxel model reflecting the spatial distribution of the acoustic properties inside the refractive medium is constructed, and a three-dimensional sharpness map of the refractive medium is generated.

[0032] The texture analysis submodule performs a multi-scale gray-level co-occurrence matrix analysis algorithm on the three-dimensional sharpness map of the refractive medium to extract texture feature vectors that reflect scattering characteristics and structural uniformity, and generates a sharpness texture feature vector set; Based on the gray value distribution information in the three-dimensional sharpness map of the refractive medium, gray-level co-occurrence matrices in different directions are constructed in multiple scale spaces. The frequency of pixel value pairs appearing under specific distance and angle combinations in the matrix is ​​calculated. For example, the contrast feature of gray-level difference between adjacent pixels is calculated in the 0-degree direction, the energy feature of homogeneous regions is calculated in the 45-degree direction, and the entropy feature of gray-level change is calculated in the 90-degree direction. By combining the feature calculation results of different scales and directions, a multi-dimensional feature vector sequence containing contrast, energy, entropy and homogeneity is formed, generating a texture feature vector set that reflects the heterogeneity of the internal structure of the refractive medium.

[0033] The index calculation submodule inputs the sharpness texture feature vector set into the regression analysis model to calculate the refractive medium scattering index and transparency index, and generates a dataset of refractive medium sharpness quantification indexes. The texture feature vector set is input into a pre-trained random forest regression model, which contains multiple decision tree structures. Each decision tree makes node splitting decisions based on texture feature values. For example, when the contrast feature value of the lens region exceeds a preset splitting threshold, it enters the left subtree branch; otherwise, it enters the right subtree branch. By integrating the predicted values ​​output by the leaf nodes of all decision trees and taking the average, the estimated value of the scattering index of the corresponding region is obtained. At the same time, the transparency index is calculated based on energy and entropy features, forming a set of quantitative index data containing scattering index and transparency index.

[0034] The report generation submodule generates a refractive media clarity assessment report based on a dataset of quantitative indicators for refractive media clarity and through a structured data encapsulation algorithm.

[0035] Key values ​​from the dataset of quantitative indicators for refractive media clarity are extracted, including the corneal central scattering index and the lens nucleus transparency index. These values ​​are compared with preset grading standard ranges. For example, when the lens nucleus transparency index is in the medium range, it is marked as a level 2 transparent state. The numerical results are converted into natural language descriptions through field mapping relationships in a predefined report template. At the same time, a 3D atlas rendering engine is called to generate a cross-sectional image of the refractive media with transparency annotations. The text description and the visual atlas are integrated to form a structured document, generating a refractive media clarity assessment report that includes quantitative indicators and visual analysis results.

[0036] Based on the refractive media clarity assessment report, the spectral acquisition submodule uses a Raman spectroscopy focusing scanning strategy to locate the target detection area, acquires the spectral data of the corresponding area of ​​the refractive media, and generates a target area spectral dataset. Based on the coordinates of the abnormal areas identified in the refractive media clarity assessment report, the laser focusing system is controlled to position the probe spot to the target location. For example, a circular scanning path is set around the center point of the cloudy area in the lens nucleus. The focal length of the objective lens is adjusted by the piezoelectric ceramic actuator so that the depth of the laser focus matches the thickness of the target area. The target area is illuminated by a light source with a specific excitation wavelength. The reflected spectral signal is collected by a spectrometer. Each sampling point obtains a spectral curve containing wavelength and intensity information, generating a spatial-spectral correlation dataset covering the target area.

[0037] The peak position identification submodule performs a Gaussian fitting peak position analysis algorithm on the target region spectral dataset to identify characteristic peak positions related to the metabolic state of the refractive media and generate a set of metabolic characteristic peak position coordinates. Baseline correction is performed on individual spectral curves in the target region spectral dataset to eliminate background fluorescence interference. Local intensity maxima are searched within a specific band range. When the peak intensity exceeds the average intensity of the neighboring region by a certain multiple, it is marked as a candidate peak. A Gaussian function is applied to fit the candidate peaks, and the center position, height, and width parameters of the Gaussian function are adjusted to minimize the mean square error between the fitted curve and the original spectral data. For example, peaks that conform to the Gaussian distribution are identified in the characteristic bands of characteristic metabolites. The center wavelength coordinates of the peaks are recorded to generate a set of characteristic peak coordinates that reflect the distribution of refractive media metabolites.

[0038] The concentration calculation submodule uses a set of metabolic characteristic peak coordinates and a regression analysis algorithm to calculate the relative concentration ratios of metabolic-related components, generating metabolic characteristic indicators of refractive media.

[0039] Based on the peak wavelength information recorded in the set of metabolic characteristic peak coordinates, a predefined metabolite characteristic peak reference table is matched. For example, a specific wavelength peak corresponding to glutathione is identified, and the integrated intensity data of the peak is extracted. At the same time, the characteristic peak intensity of the reference substance is selected as the benchmark, and the ratio of the peak intensity of the target metabolite to the peak intensity of the reference peak is calculated. This ratio is input into a pre-trained linear regression model. The model parameters are obtained based on known concentration samples and output the relative concentration estimate of the target metabolite in the refractive medium, generating a set of characteristic indicators containing the relative concentrations of various metabolites.

[0040] The feature fusion submodule performs a multi-parameter fusion analysis algorithm on the refractive medium clarity quantification index and metabolic characteristic index to generate a comprehensive evaluation feature vector; Based on the scattering index, transparency index, and relative concentration data of metabolites in the quantification indicators of refractive media clarity, the indicators are standardized to convert indicators of different dimensions to the same numerical range. Weight coefficients are assigned according to the type of indicator. For example, the scattering index, which has a greater impact on transparency assessment, is given a higher weight, while changes in metabolite concentration are given a medium weight. The values ​​of the various indicators are integrated by weighted summation to form a comprehensive evaluation feature vector containing multidimensional features.

[0041] The grading submodule is based on a comprehensive evaluation feature vector and uses preset threshold rules or statistical classification algorithms to grade the transparency of refractive media, generating a refractive media clarity grade and anomaly warning grade. Based on the comprehensive evaluation of the numerical distribution of each dimension in the feature vector, the vector is input into a pre-trained support vector machine classification model. This model constructs an optimal classification hyperplane in a high-dimensional feature space and determines the category based on the distance of the feature vector to the hyperplane. For example, when the feature vector of the corneal region is located in the negative direction of the hyperplane and the distance exceeds a preset threshold, it is classified as a low transparency level. When the feature vector of the lens region is located in the positive direction and the distance is relatively close, it is classified as a medium transparency level. At the same time, when abnormal shifts in metabolic indicators are detected, a secondary abnormality prompt is triggered, generating a classification result that includes the transparency level and the abnormality prompt level.

[0042] The results output submodule is used to output the refractive screening assessment results. The assessment results are used for screening and assessment purposes and do not constitute a disease diagnosis conclusion.

[0043] The system integrates the clarity level and abnormality indication level information of the refractive media, organizes the data structure according to a predefined output format, including the examinee number, examination timestamp, transparency level code of each anatomical region, and abnormality marker bits. For example, the central corneal area is marked as T2 transparency and the lens nucleus area is marked as A1 abnormality indication. The structured data package is transmitted to the display terminal and printing device through a standard data interface to generate standardized output results for refractive screening and status assessment.

[0044] The rule matching submodule calls the preset screening prompt rule library based on the refractive media clarity level and abnormality prompt level to generate corresponding examination prompt information; Based on the specific values ​​of the refractive media clarity level and the abnormality warning level, matching entries are retrieved from the preset rule base. For example, when the lens nucleus transparency level is in the medium range and accompanied by metabolic abnormality warning, the corresponding rule entry is triggered, the standard description template and suggested content stored in the entry are extracted, and examination prompt information containing abnormality location description and screening suggestions is generated.

[0045] The text generation submodule, based on the examination prompts, executes a natural language generation algorithm to output a descriptive text of the refractive screening results. Based on the key elements in the inspection prompt information, including the name of the anatomical location, the abnormality type code, and the suggested measures identifier, a basic subject-verb-object sentence structure is selected from the preset sentence template library. The location name is filled into the subject field, the abnormality type code is converted into an adjective phrase and filled into the modifier field, and the suggested measures identifier is converted into a verb phrase and filled into the predicate field. The descriptive text that conforms to the natural language expression norm is generated through sentence structure recombination.

[0046] The atlas annotation submodule combines the description text of the refractive screening results to annotate abnormal areas in the three-dimensional clarity atlas of the refractive media, generating a refractive screening assessment report.

[0047] Based on the abnormal area names marked in the description text of the refractive screening results, the corresponding anatomical structure coordinate range is located in the three-dimensional clarity atlas spatial coordinate system. For example, the lens nucleus corresponds to the Z-axis depth range and the XY plane coordinate region. Text labels and indicator arrows are added to the surface of this region. The relevant statements in the description text are converted into labeled text. The labeled information is superimposed onto the atlas surface through the three-dimensional rendering engine to generate a visual report document with structured annotations.

[0048] The data encryption submodule processes the refractive screening assessment report using an encryption algorithm to generate an encrypted data packet; Based on the data structure characteristics of the refractive screening assessment report, byte stream information of text and image data is extracted, the byte stream is divided into fixed-length data blocks, each data block is appended with a unique sequence identifier, and each data block is independently encrypted using a symmetric encryption algorithm to generate an encrypted data packet containing encrypted data blocks and sequence identifiers.

[0049] The interface management submodule encapsulates encrypted data packets using a standardized application programming interface (API) protocol to generate cloud transmission instructions. The sequence identifier and data block length information in the encrypted data packet are parsed. The data frame header is constructed according to the requirements of the medical information exchange standard protocol, including the protocol version number, the total data packet length field and the target platform address code. The encrypted data blocks are filled into the effective payload area of ​​the data frame in sequence, a frame check sequence field is attached, and a cloud transmission instruction conforming to the standard transmission format is generated.

[0050] The cloud storage submodule synchronously stores the refractive screening assessment report to the information platform according to the cloud transmission instructions, generating a cloud-based refractive screening file.

[0051] Based on the target platform address code in the cloud transmission command, connect to the storage service interface of the corresponding medical information platform, parse the encrypted data block sequence in the command data frame, reassemble the original byte stream according to the sequence identifier order, write the reassembled data stream into the designated database partition of the platform storage system, and generate a cloud refractive screening file associated with the examinee's identity.

[0052] A refractive screening refractometer for assessing the clarity of refractive media includes a body 1. An image capture module is fixedly connected to one side of the body 1, and light source modules are fixedly connected to both sides of the image capture module. A display module 3 is fixedly connected to the other side of the body for displaying the refractive media clarity map and screening assessment results. A control module 2 is fixedly connected to the upper surface of the body for controlling the multimodal acquisition, data fusion, and screening assessment process. The refractometer is used to acquire multimodal images of refractive media and output the refractive media clarity assessment results and screening prompts, but does not have disease diagnosis functions.

[0053] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. A refractive screening system for assessing the clarity of refractive media, characterized in that, It includes the following modules: a multi-source acquisition module, which uses an adaptive eye-tracking mechanism to execute a multi-modal synchronous triggering control algorithm to simultaneously acquire slit lamp forward scattering images, optical coherence tomography post-illumination images, and polarized light scanning data to generate the original multi-modal image set; The multi-source acquisition module includes an optical adapter submodule, a dynamic calibration submodule, and a synchronization control submodule. The data fusion module, based on the original multimodal image set, executes a cross-modal feature registration algorithm and integrates multi-source features through a channel attention weighted fusion network to generate a registered and fused image; The data fusion module includes a feature extraction submodule, a spatial transformation submodule, and a weighted fusion submodule; The three-dimensional reconstruction module, based on the registered and fused images, executes an improved cube travel algorithm, combines a convolutional neural network to segment the boundaries of the refractive media, and applies the acoustic time-of-flight difference method to map acoustic parameters to generate a three-dimensional sharpness map of the refractive media. The 3D reconstruction module includes a boundary segmentation submodule, a parameter mapping submodule, and a mesh generation submodule; The quantitative assessment module, based on the three-dimensional sharpness map of the refractive medium, executes a multi-scale texture quantization algorithm, and inputs a regression model to calculate the medium scattering index and transparency index, generating a refractive medium sharpness assessment report; The quantitative assessment module includes a texture analysis submodule, an index calculation submodule, and a report generation submodule; The spectral analysis module, based on the refractive media clarity assessment report, executes the Raman spectral analysis algorithm, calculates the relative concentration of metabolic-related components in the media by matching characteristic peak positions, and generates refractive media metabolic characteristic indicators. The spectral analysis module includes a spectral acquisition submodule, a peak position identification submodule, and a concentration calculation submodule; The screening and assessment module, based on the refractive media clarity assessment report and metabolic characteristic indicators, executes a multi-parameter fusion assessment algorithm, outputs the refractive media transparency level and abnormality warning level, and generates refractive screening and assessment results. The screening and assessment module includes a feature fusion submodule, a grade classification submodule, and a result output submodule; The report generation module, based on the refractive screening assessment results, executes the structured report generation engine to automatically output a description of the examination conclusions and a refractive media clarity map annotation, and generates a refractive screening assessment report; The report generation module includes a rule matching submodule, a text generation submodule, and a map annotation submodule; The cloud platform interaction module, based on the refractive screening assessment report, executes an adaptive encrypted transmission protocol and synchronizes the data to the medical information platform through a standardized API interface to generate a cloud-based refractive screening file. The cloud platform interaction module includes a data encryption submodule, an interface management submodule, and a cloud storage submodule.

2. The refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The optical adapter submodule adjusts the output wavelength and light intensity distribution of the multispectral light source according to the preset optical parameters of the slit lamp through a wavelength adaptive control algorithm to generate a standardized illumination light field. The dynamic calibration submodule, based on the standardized illumination light field, uses a real-time eye movement compensation algorithm to drive the fine-tuning actuator to correct the optical path offset and generate calibrated optical path parameters. The synchronization control submodule executes a multimodal temporal synchronization trigger control algorithm based on the calibrated optical path parameters, synchronously starting optical coherence tomography imaging and polarization-sensitive imaging to generate the original multimodal image set.

3. The refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The feature extraction submodule performs a scale-invariant feature transformation algorithm on the original multimodal image set to extract cross-modal key point feature descriptors and generate a multimodal feature point description set; The spatial transformation submodule, based on the multimodal feature point description set, uses a thin-plate spline spatial transformation algorithm to establish a nonlinear mapping relationship between images of different modalities and generate a cross-modal registration transformation matrix. The weighted fusion submodule integrates feature maps of different modalities based on the cross-modal registration transformation matrix and generates a registration fusion image through a channel attention weighted fusion network.

4. A refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The boundary segmentation submodule uses the registered and fused image to identify the anatomical boundary region of the refractive medium using a convolutional neural network segmentation algorithm, and generates a refractive medium region segmentation mask. The parameter mapping submodule performs an acoustic time-of-flight difference calculation algorithm based on the refractive medium region segmentation mask to quantify the acoustic propagation characteristics inside the refractive medium and generate an acoustic parameter distribution matrix. The mesh generation submodule uses the acoustic parameter distribution matrix and an improved cube traversal algorithm to construct a three-dimensional voxel model and generate a three-dimensional sharpness map of the refractive medium.

5. A refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The texture analysis submodule performs a multi-scale gray-level co-occurrence matrix analysis algorithm on the three-dimensional sharpness map of the refractive medium to extract texture feature vectors that reflect scattering characteristics and structural uniformity, and generates a sharpness texture feature vector set. The index calculation submodule inputs the sharpness texture feature vector set into the regression analysis model to calculate the refractive medium scattering index and transparency index, and generates a dataset of refractive medium sharpness quantification indexes. The report generation submodule generates a refractive media clarity assessment report based on the refractive media clarity quantification index dataset and through a structured data encapsulation algorithm.

6. A refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The spectral acquisition submodule uses the Raman spectral focusing scanning strategy to locate the target detection area based on the refractive media clarity assessment report, acquires the spectral data of the corresponding area of ​​the refractive media, and generates a target area spectral dataset. The peak position identification submodule performs a Gaussian fitting peak position analysis algorithm on the target region spectral dataset to identify characteristic peak positions related to the metabolic state of the refractive media and generate a set of metabolic characteristic peak position coordinates. The concentration calculation submodule uses the set of metabolic characteristic peak coordinates to calculate the relative concentration ratios of metabolic-related components using a regression analysis algorithm, thereby generating refractive media metabolic characteristic indicators.

7. A refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The feature fusion submodule performs a multi-parameter fusion analysis algorithm on the refractive medium clarity quantification index and metabolic characteristic index to generate a comprehensive evaluation feature vector. The grading submodule, based on the comprehensive evaluation feature vector, grades the transparency of the refractive media using preset threshold rules or statistical classification algorithms, generating a refractive media clarity grade and anomaly warning grade. The result output submodule is used to output refractive screening assessment results. These assessment results are for screening and assessment purposes and do not constitute a disease diagnosis conclusion.

8. A refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The rule matching submodule calls a preset screening prompt rule library to generate corresponding examination prompt information based on the refractive media clarity level and abnormal prompt level. Based on the examination prompt information, the text generation submodule executes a natural language generation algorithm to output a descriptive text of the refractive screening result. The atlas annotation submodule, in conjunction with the description text of the refractive screening results, annotates abnormal areas in the three-dimensional clarity atlas of the refractive media, and generates a refractive screening assessment report.

9. A refractive screening system for assessing the clarity of refractive media according to claim 1, characterized in that: The data encryption submodule performs encryption algorithm processing on the refractive screening assessment report to generate an encrypted data packet; The interface management submodule encapsulates the encrypted data packet using a standardized application programming interface protocol to generate cloud transmission instructions. The cloud storage submodule synchronously stores the refractive screening assessment report to the information platform according to the cloud transmission instructions, generating a cloud-based refractive screening file.

10. A refractive screening refractometer for assessing the clarity of refractive media, characterized in that: The optometry device is based on the refractive media clarity assessment system according to any one of claims 1-9, including a body (1), an image capture module is fixedly connected to one side of the body (1), a light source module is fixedly connected to both sides of the image capture module, a display module (3) is fixedly connected to the other side of the body (1) for displaying the refractive media clarity map and screening assessment results, and a control module (2) is fixedly connected to the upper surface of the body (1) for controlling the multimodal acquisition, data fusion and screening assessment process. The optometry device is used to acquire multimodal images of refractive media and output the refractive media clarity assessment results and screening prompt information, but does not have disease diagnosis function.

Images