Multi-mode ophthalmic data fusion refractive error parameter analysis system
Through the multimodal ophthalmic data fusion system, a variety of ophthalmic equipment and machine learning algorithms are used to solve the problem of insufficient detection of single modal data, and efficient diagnosis and personalized correction of refractive errors are achieved, which improves diagnostic accuracy and management efficiency.
Patent Information
- Application Number
- CN202510777253.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-11
- Publication Date
- 2025-07-29
AI Technical Summary
In the prior art, single-modal ophthalmic data detection cannot fully reflect the overall refractive state of the eyeball, and there are problems of insufficient diagnostic accuracy and poor effectiveness of correction schemes. Multimodal data lacks systematic fusion analysis and fails to fully utilize the synergistic advantages.
A multimodal ophthalmic data fusion system, including optical coherence tomography, fundus camera, corneal topography and wavefront aberration measuring instrument, is used to process data through a convolutional neural network and support vector machine, and combine weighted fusion and nonlinear regression models to generate personalized correction plans and health education content.
It significantly improves the accuracy of refractive error diagnosis and the effectiveness of personalized correction plans, enhances the efficiency and quality of visual health management, and promotes early intervention and long-term management.
Smart Images
Figure CN120381236A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of medical informatization and artificial intelligence technology, and specifically to a refractive error parameter analysis system for multimodal ophthalmic data fusion. Background Technique
[0002] Traditional detection relies on single-modal data. For example, a refractometer measures refractive power and a corneal topographer measures corneal morphology. Although the operation is simple, there are obvious limitations. On the one hand, single data only reflects local information of the eye, and it is a challenge to comprehensively present the overall refractive state of the eyeball and changes in tissue structure, and to reveal potential factors affecting refraction such as retinal lesions and corneal abnormalities. On the other hand, it is easily interfered by detection errors and individual differences, resulting in insufficient diagnostic accuracy and having a certain impact on the effectiveness of correction plans.
[0003] With the development of ophthalmic imaging technology, the popularization of multimodal detection technologies such as optical coherence tomography (OCT), fundus photography, corneal topography, and wavefront aberration measurement can obtain eye information from multiple dimensions. However, at present, multimodal data are mostly independently applied, lacking systematic fusion analysis methods, and the synergistic advantages have not been fully utilized. There are still certain difficulties in exploring the internal relationship between eye information and refractive errors. Existing fusion technologies stay at the level of data superposition, with information redundancy and feature conflicts, and lack dedicated analysis models and algorithms. The ability to accurately calculate refractive parameters and generate personalized correction plans is still insufficient, and it is difficult to fully meet the clinical needs of efficient and accurate diagnosis and treatment. Therefore, it is of great significance to develop a refractive error analysis system based on deep fusion of multimodal data. Summary of the Invention
[0004] The purpose of the present invention is to provide a refractive error parameter analysis system for multimodal ophthalmic data fusion to solve the problems raised in the above background technique.
[0005] To achieve the above purpose, the present invention provides the following technical solution: A refractive error parameter analysis system for multimodal ophthalmic data fusion, including a data acquisition layer, a data fusion layer, a data analysis layer, and an application service layer, where:
[0006] The data acquisition layer includes an optical coherence tomography scanner, a fundus camera, a corneal topographer, and a wavefront aberration measurer, which are used to obtain three-dimensional retinal structure information, fundus images, corneal morphology parameters, and wavefront aberration distribution data;
[0007] The data fusion layer includes a feature extraction module and a weighted fusion module. The feature extraction module processes image-type data through a convolutional neural network and processes numerical-type data through a support vector machine. The weighted fusion module dynamically adjusts the weights of each modal data according to the application scenario and generates a unified feature vector;
[0008] The data analysis layer includes a refractive parameter calculation module that maps multimodal fusion data to refractive power, axial length, and corneal curvature parameters based on a non-linear regression model, and generates stratified statistical analysis results in combination with the individual factors of the patient.
[0009] The application service layer includes a personalized correction plan generator and a health education database, which are used to generate correction plans and provide health education content.
[0010] Preferably, the optical coherence tomography scanner uses high-speed swept-source technology, the wavelength range of the light source is 800 to 1000 nanometers, and the scanning speed is not less than 70 kHz.
[0011] Preferably, the fundus camera records the distribution of fundus blood vessels and lesion characteristics through high-resolution imaging technology, and the imaging resolution reaches the 5-micron level.
[0012] Preferably, the corneal topographer measures the corneal surface morphology using Placido disk projection technology and outputs corneal curvature radius and astigmatism axis parameters.
[0013] Preferably, the wavefront aberration meter measures the wavefront aberration distribution of the eye's optical system based on the Hartmann-Shack sensor principle and generates Zernike polynomial coefficients as numerical data.
[0014] Preferably, the feature extraction module uses a convolutional neural network to identify local features for image-type data and a support vector machine to screen significant variables for numerical-type data.
[0015] Preferably, the weighted fusion module integrates multimodal data features through an improved weighted algorithm, which introduces an adaptive weight adjustment mechanism and considers the correlation matrix of each modal data.
[0016] Preferably, the refractive parameter calculation module introduces an error correction mechanism based on Bayesian inference and adjusts the parameter estimation value through an iterative optimization algorithm.
[0017] Preferably, the personalized correction plan generator recommends correction methods in combination with the patient's daily living habits and eye use needs, and the correction methods include frame glasses, contact lenses, or surgical treatment.
[0018] Preferably, the health education database contains the causes, prevention methods, and daily care suggestions for refractive errors, which are presented on the screen of digital devices and support communication between doctors and patients.
[0019] Compared with the prior art, the beneficial effects of the present invention are as follows: Through the deep fusion and precise analysis of multi-modal data, the accuracy of refractive error diagnosis and the effectiveness of personalized correction plans are significantly improved; The systematic visual health management model not only enhances the patient's cognitive level of refractive errors, but also promotes the implementation of early intervention and long-term management. In addition, all analysis results can be presented on the digital device screen, facilitating communication between doctors and patients, and further improving the efficiency and quality of medical services. BRIEF DESCRIPTION OF THE DRAWINGS
[0020] Figure 1 It is a schematic structural diagram of the system architecture of the present invention.
[0021] In the figure: 1. Data acquisition layer; 2. Data fusion layer; 3. Data analysis layer; 4. Application service layer; 5. Optical coherence tomography scanner; 6. Fundus camera; 7. Corneal topographer; 8. Wavefront aberration measuring instrument; 9. Feature extraction module; 10. Weighted fusion module; 11. Refractive parameter calculation module; 12. Personalized correction plan generator; 13. Health education database. DETAILED DESCRIPTION OF THE EMBODIMENTS
[0022] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0023] Please refer to Figure 1 , the present invention provides a technical solution: A refractive error parameter analysis system for multi-modal ophthalmic data fusion, including a data acquisition layer 1, a data fusion layer 2, a data analysis layer 3, and an application service layer 4, wherein:
[0024] The data acquisition layer 1 includes an optical coherence tomography scanner 5, a fundus camera 6, a corneal topographer 7, and a wavefront aberration measuring instrument 8, and is used to obtain three-dimensional retinal structure information, fundus images, corneal morphology parameters, and wavefront aberration distribution data;
[0025] The data fusion layer 2 includes a feature extraction module 9 and a weighted fusion module 10. The feature extraction module 9 processes image-type data through a convolutional neural network and processes numerical-type data through a support vector machine. The weighted fusion module 10 dynamically adjusts the weights of each modal data according to the application scenario and generates a unified feature vector;
[0026] The data analysis layer 3 includes a refractive parameter calculation module 11, which maps multi-modal fusion data to refractive power, axial length, and corneal curvature parameters based on a non-linear regression model, and generates a stratified statistical analysis result in combination with patient individual factors;
[0027] The application service layer 4 includes a personalized correction plan generator 12 and a health education database 13, which are used to generate correction plans and provide health education content.
[0028] Specifically, in the data acquisition layer 1, an optical coherence tomography scanner 5, a fundus camera 6, a corneal topographer 7, and a wavefront aberration meter 8 are connected to a central data processing unit through a standardized interface to form a multi-modal data acquisition network.
[0029] The optical coherence tomography scanner 5 uses high-speed swept-source technology to obtain three-dimensional structural information of the retina and the cornea. Its light source wavelength range is 800 to 1000 nanometers, and the scanning speed is not less than 70 kHz to ensure high-precision and consistent data acquisition.
[0030] The fundus camera 6 records the distribution of fundus blood vessels and lesion characteristics through high-resolution imaging technology. Its imaging resolution reaches the 5-micron level, and it can clearly display tiny lesion areas of the retina.
[0031] The corneal topographer 7 uses Placido disk projection technology to measure the surface morphology of the cornea and outputs key parameters such as corneal radius of curvature and astigmatism axis.
[0032] The wavefront aberration meter 8 is based on the Hartmann-Shack sensor principle, measures the wavefront aberration distribution of the eye's optical system, and generates Zernike polynomial coefficients as the basis for numerical data.
[0033] These devices are connected to the central data processing unit through dedicated data transmission cables.
[0034] In the embodiment, specifically, in the data fusion layer 2, the feature extraction module 9 and the weighted fusion module 10 jointly act on the deep processing of multi-modal data.
[0035] The feature extraction module 9 uses a convolutional neural network and traditional machine learning algorithms to extract features for different types of input data respectively. For image-type data such as optical coherence tomography images and fundus photos, the convolutional neural network identifies local features through multi-layer convolutional kernel operations. For numerical-type data such as wavefront aberration measurement results, the support vector machine selects variables that significantly affect the refractive state through kernel function mapping.
[0036] Furthermore, after the feature extraction is completed, the weighted fusion module 10 dynamically adjusts the weight allocation strategy for each modality data according to the application scenario. The weighted fusion module 10 integrates the extracted features into a unified feature vector through an improved weighted algorithm, which takes into account the correlation matrix of each modality data and introduces an adaptive weight adjustment mechanism.
[0037] In the data analysis layer 3, the refractive parameter calculation module 11 constructs a refractive error parameter calculation model based on the nonlinear regression analysis method.
[0038] Specifically, this model maps the multi-modal fusion data to core parameters such as refractive power, axial length, and corneal curvature, and generates a stratified statistical analysis result in combination with individual factors such as the patient's age and gender. For example, for a 10-year-old child, the refractive parameter calculation module 11 predicts the possible refractive changes in the next five years based on the current trends of corneal curvature and axial length, and outputs a corresponding risk assessment report. To improve the accuracy of the model, the refractive parameter calculation module 11 also introduces an error correction mechanism based on Bayesian inference, and continuously adjusts the parameter estimation value through an iterative optimization algorithm.
[0039] Specifically, the initial parameter estimation value is obtained by the least squares method, and then corrected by the maximum a posteriori probability estimation. The finally output parameter estimation value has a higher confidence level.
[0040] In the application service layer 4, the personalized correction plan generator 12 and the health education database 13 provide diverse functional supports for doctors and patients.
[0041] Specifically, the personalized correction plan generator 12 recommends the most suitable correction method based on the output result of the aforementioned data analysis layer 3, combined with the patient's daily living habits and eye use needs. For example, for an office worker who needs to use an electronic screen for a long time, the system will give priority to recommending frame glasses or contact lenses with blue light blocking function. The treatment effect monitoring tool evaluates the actual effect of the correction measure through the comparative analysis of regularly updated multi-modal data, and prompts whether the treatment plan needs to be adjusted. The health education database 13 contains rich popular science content, covering the causes of refractive errors, prevention methods, and daily care suggestions, to help patients and their families better understand relevant knowledge. All analysis results are presented on the digital device screen, and the interface design follows the user-friendly principle to ensure smoother communication between doctors and patients.
[0042] The selection of the system's hardware equipment adheres to the principle of balancing high performance and compatibility. The optical coherence tomography scanner 5 uses a model with high-speed data transmission capabilities and an internal storage capacity of no less than 512GB to meet the needs of large-scale data acquisition. The fundus camera 6 is equipped with a high-resolution CMOS sensor with a pixel size of 2.4 microns, ensuring that the imaging quality meets clinical diagnostic requirements. The software platform is developed based on a cloud computing architecture, using a distributed file system to store multimodal data and support real-time analysis of petabyte-level data. The network architecture adopts a distributed deployment model, with the main server and backup server connected by a dedicated fiber optic line to ensure system stability and scalability. The data processing process begins with multimodal data acquisition and goes through steps such as preprocessing, feature extraction, data fusion, and parameter calculation.
[0043] In practical application scenarios, the system demonstrates strong flexibility and practicality. For example, in the prevention and control of myopia in adolescents, the system regularly monitors changes in corneal morphology and axial length to promptly identify potential risks and formulate intervention measures.
[0044] During the specific operation, the corneal topographer 7 is first used to measure the patient's corneal curvature and astigmatism, while the optical coherence tomography scanner 5 is used to obtain the three-dimensional structural information of the retina and cornea. The wavefront aberration meter 8 records the aberration distribution of the optical system of the eye, and the fundus camera 6 takes fundus photos to assess the health of the retina. After all the raw data are uploaded to the central data processing unit, the feature extraction module 9 automatically identifies the key features and integrates them through the weighted fusion module 10. The refractive parameter calculation module 11 generates a refractive status assessment report based on the fused feature vector, and the personalized correction scheme generator 12 recommends the appropriate correction method based on this. The entire process realizes automated processing from data acquisition to result output, significantly improving the accuracy and efficiency of refractive error diagnosis.
[0045] Although the present invention has been described in detail with reference to the aforementioned embodiments, it is still possible for those skilled in the art to modify the technical solutions described in the aforementioned embodiments, or to make equivalent substitutions for some of the technical features therein. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.
Claims
1. A refractive error parameter analysis system for multimodal ophthalmic data fusion, characterized in that: It includes a data acquisition layer (1), a data fusion layer (2), a data analysis layer (3), and an application service layer (4), where: The data acquisition layer (1) includes an optical coherence tomography scanner (5), a fundus camera (6), a corneal topographer (7), and a wavefront aberration meter (8), which are used to obtain three-dimensional retinal structure information, fundus images, corneal morphology parameters, and wavefront aberration distribution data; The data fusion layer (2) includes a feature extraction module (9) and a weighted fusion module (10). The feature extraction module (9) processes image-type data through a convolutional neural network and processes numerical-type data through a support vector machine. The weighted fusion module (10) dynamically adjusts the weights of each modality data according to the application scenario and generates a unified feature vector; The data analysis layer (3) includes a refractive parameter calculation module (11), which maps multi-modal fusion data to refractive power, axial length, and corneal curvature parameters based on a non-linear regression model, and generates a stratified statistical analysis result in combination with the patient's individual factors; The application service layer (4) includes a personalized correction plan generator (12) and a health education database (13), which are used to generate correction plans and provide health education content.
2. The refractive error parameter analysis system according to claim 1, wherein The optical coherence tomography scanner (5) adopts high-speed swept-source technology, with a light source wavelength range of 800 to 1000 nanometers and a scanning speed of not less than 70 kHz.
3. The refractive error parameter analysis system according to claim 1, wherein: The fundus camera (6) records the distribution of fundus blood vessels and lesion characteristics through high-resolution imaging technology, and the imaging resolution reaches the 5-micron level.
4. The refractive error parameter analysis system according to claim 1, wherein: The corneal topographer (7) measures the corneal surface morphology using Placido disk projection technology and outputs corneal radius of curvature and astigmatism axis parameters.
5. The refractive error parameter analysis system according to claim 1, wherein: The wavefront aberration meter (8) measures the wavefront aberration distribution of the eye's optical system based on the Hartmann-Shack sensor principle and generates Zernike polynomial coefficients as numerical-type data.
6. The refractive error parameter analysis system according to claim 1, wherein: The feature extraction module (9) uses a convolutional neural network to identify local features for image-type data and uses a support vector machine to screen significant variables for numerical-type data.
7. The refractive error parameter analysis system according to claim 1, wherein: The weighted fusion module (10) integrates multi-modal data features through an improved weighted algorithm, which introduces an adaptive weight adjustment mechanism and considers the correlation matrix of each modality data.
8. The refractive error parameter analysis system according to claim 1, characterized in that: The refractive parameter calculation module (11) introduces an error correction mechanism based on Bayesian inference and adjusts the parameter estimation value through an iterative optimization algorithm.
9. The refractive error parameter analysis system according to claim 1, wherein: The personalized correction plan generator (12) recommends correction methods in combination with the patient's daily living habits and eye use needs. The correction methods include frame glasses, contact lenses, or surgical treatment.
10. The refractive error parameter analysis system according to claim 1, wherein: The health education database (13) contains content on the causes, prevention methods, and daily care suggestions for refractive errors, which is presented on the digital device screen and supports communication between doctors and patients.