A system and method for detecting color vision function integrating VR and artificial intelligence

Abstract

The present invention discloses a system and method for detecting eye color vision function integrating VR and artificial intelligence, including: VR terminal equipment, data storage module, artificial intelligence processing module, power supply module; wherein, the VR terminal equipment includes a VR head mounted display, an infrared eye tracking camera, a host control system and an interactive input handle; the data storage module is used to store detection data; the artificial intelligence processing module is used to perform qualitative analysis of color vision disorder, quantitative analysis of color vision disorder and correlation evaluation of color vision disorder and fundus problems on the detection data to obtain evaluation results; the power supply module is connected to the VR terminal equipment, the data storage module and the artificial intelligence processing module to provide power supply guarantee. The present invention reduces the interference of subjective factors, improves the detection accuracy, the system's automated detection process and intelligent data analysis capabilities, and also greatly shortens the detection time.

Description

Technical Field

[0001] The present invention belongs to the technical field of ophthalmic detection, and in particular relates to an eye color vision function detection system and method integrating VR and artificial intelligence. Background Art

[0002] Color vision testing plays a key role in ophthalmic clinical diagnosis, occupational health screening, and visual science research. However, traditional color vision testing methods, such as pseudoisochromatic test, color vision testing atlas, and color chess, have significant limitations. Taking the pseudoisochromatic test as an example, its detection principle is based on the subject's recognition and judgment of a pattern composed of points of different colors and brightness. However, in practical applications, this method exposes many problems. According to data from multi-center ophthalmic clinical research, the misdiagnosis rate in the detection of early color-blind patients can be as high as about 35%. This is mainly because the pattern is fixed and lacks dynamic interaction, making it difficult to accurately capture mild color vision abnormalities. At the same time, the test results are easily affected by changes in ambient light and subjective cognitive differences of the subjects. For example, in a dimly lit environment, the subject may misjudge the color of the pattern due to decreased visual resolution; and different subjects may have different understandings and judgments of the pattern due to differences in cultural background, cognitive habits, and other factors. Methods such as color vision testing atlas and color chess also have inherent defects. They are highly dependent on the subject's subjective comparison and judgment of color, and lack objective and accurate quantitative indicators. During the color chess test, the subject's judgment of color arrangement is often affected by multiple factors such as personal experience, visual preference, and psychological state at the time, making it difficult to accurately assess the degree of color vision impairment. More importantly, traditional detection methods cannot deeply explore the potential association between color vision impairment and fundus diseases. In ophthalmic clinical practice, many fundus diseases are often accompanied by color vision abnormalities in the early stages, but traditional detection methods are difficult to detect this inherent connection in the early stages of the disease due to technical limitations, thus missing the best time for diagnosis and treatment. For example, patients with macular degeneration may only show a slight decrease in color sensitivity in the early stages of the disease, but traditional detection methods are difficult to establish an effective correlation diagnosis between this subtle color vision change and macular degeneration, so that the disease may gradually worsen without being discovered in time. With the rapid development of science and technology, VR technology and artificial intelligence technology are increasingly widely used in the medical field, but there is no report on a technical solution that organically combines the two and specifically applies them to the field of color vision detection, and can effectively solve the shortcomings of the above-mentioned traditional detection methods. Summary of the invention

[0003] The present invention aims to construct a high-precision, high-efficiency and comprehensive color vision function detection system and method based on VR and artificial intelligence, to achieve accurate evaluation of color vision function, accurately determine the type and degree of color vision disorder, and deeply explore the potential relationship between color vision disorder and fundus diseases, so as to provide strong technical support for clinical diagnosis, treatment and disease prevention in ophthalmology.

[0004] To achieve the above objectives, the present invention provides a system and method for detecting eye color vision function integrating VR and artificial intelligence. The system and method for detecting eye color vision function integrating VR and artificial intelligence include:

[0005] VR terminal equipment, data storage module, artificial intelligence processing module, power module;

[0006] The VR terminal device includes a VR head mounted display, an infrared eye tracking camera, a host control system and an interactive input handle;

[0007] The data storage module is used to store detection data;

[0008] The artificial intelligence processing module is used to perform qualitative analysis of color vision disorder, quantitative analysis of color vision disorder and correlation evaluation of color vision disorder and fundus problems on the detection data to obtain evaluation results;

[0009] The power supply module is connected to the VR terminal device, the data storage module and the artificial intelligence processing module to provide power supply guarantee.

[0010] The present invention also provides a method for detecting eye color vision function integrating VR and artificial intelligence, comprising:

[0011] After the subject wears the detection system, the control terminal sends an initialization command to the central command control module of the host control system, and the central command control module activates the infrared eye tracking camera and starts the eye fixation point analysis module, calculates the initial eye fixation point position, and instructs the color vision image display module to present the initial guide pattern, and starts the detection process;

[0012] The data collection module collects the subject's reaction data to the colors in the virtual environment, the eye movement data and the related fixation center point analysis data, and sends them to the image recognition and processing module;

[0013] The image recognition and processing module pre-processes the color response data, performs positioning calibration, feature extraction and quantitative analysis on the eye movement data and fixation center point related data, deeply fuses the color response data and the eye movement data, and sends them to the artificial intelligence analysis module;

[0014] The artificial intelligence analysis module uses a deep learning algorithm to perform qualitative analysis, quantitative analysis and corresponding analysis of color vision disorders on the processed data, and sends the results to the data analysis module;

[0015] The data analysis module performs statistics and interpretation on the analysis results and sends the results to the feedback module;

[0016] The feedback module displays the statistical results to the user in the form of a report, and simultaneously sends them to the data storage module for storage.

[0017] Compared with the prior art, the present invention has the following advantages and technical effects:

[0018] Improve detection accuracy: Compared with traditional color vision detection methods, this invention reduces the interference of subjective factors. Through comprehensive analysis of multi-source data and precise judgment of artificial intelligence, it can more accurately detect the existence, type and degree of color vision disorders, as well as potential fundus disease risks, providing a reliable basis for the early diagnosis and treatment of ophthalmic diseases, and helping to improve the overall quality of ophthalmic medical care.

[0019] Improve detection efficiency: The automated detection process and intelligent data analysis capabilities of the system of the present invention greatly shorten the detection time. Traditional detection methods may require the examinee to undergo multiple tedious tests and manual judgments, while the present invention can complete comprehensive color vision function detection and analysis in a relatively short time and generate detailed reports, thereby improving the efficiency of ophthalmic diagnosis and reducing the waiting time for patients.

[0020] Expanded detection function: This invention can not only realize conventional color vision disorder detection, but also deeply explore the relationship between color vision disorder and fundus diseases, providing a new means and perspective for the comprehensive diagnosis of ophthalmic diseases. This helps ophthalmologists to have a more comprehensive understanding of the patient's eye health status, timely discover potential fundus problems, take corresponding treatment measures, and prevent the further development of ophthalmic diseases.

[0021] Personalized medical services: The present invention can meet the special needs of different examinees and provide customized medical services through personalized testing solutions and intelligent decision-making support. Whether it is for ordinary color vision testing needs or for patients with complex eye diseases or special visual problems, the present invention can provide more accurate and targeted testing and diagnostic suggestions, promoting the personalized development of ophthalmic medical services. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] The drawings constituting a part of the present application are used to provide a further understanding of the present application. The illustrative embodiments and descriptions of the present application are used to explain the present application and do not constitute an improper limitation on the present application. In the drawings:

[0023] Figure 1 This is a VR terminal device control system mode diagram of an embodiment of the present invention;

[0024] Figure 2 A schematic diagram of the system structure of an embodiment of the present invention;

[0025] Figure 3 Schematic diagram of the artificial intelligence color analysis process of an embodiment of the present invention. DETAILED DESCRIPTION

[0026] It should be noted that, in the absence of conflict, the embodiments and features in the embodiments of the present application can be combined with each other. The present application will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0027] It should be noted that the steps shown in the flowcharts of the accompanying drawings can be executed in a computer system such as a set of computer executable instructions, and that, although a logical order is shown in the flowcharts, in some cases, the steps shown or described can be executed in an order different from that shown here.

[0028] This embodiment is dedicated to providing an innovative eye color vision function detection system and method based on the deep integration of virtual reality (VR) technology and artificial intelligence technology, aiming to solve many problems faced by traditional color vision detection technology and open a new era of color vision detection.

[0029] like Figure 1-2 As shown, this embodiment provides an eye color vision function detection system integrating VR and artificial intelligence, including:

[0030] VR terminal equipment, data storage module, artificial intelligence processing module, power module;

[0031] Among them, VR terminal equipment includes VR head-mounted display, infrared eye tracking camera, host control system and interactive input handle;

[0032] The data storage module is used to store the detection data;

[0033] The artificial intelligence processing module is used to perform qualitative analysis of color vision impairment, quantitative analysis of color vision impairment, and correlation evaluation of color vision impairment and fundus problems on the test data to obtain evaluation results;

[0034] The power module is connected to the VR terminal device, data storage module and artificial intelligence processing module to provide power supply guarantee.

[0035] Furthermore, the VR head-mounted display of this embodiment adopts ultra-high-definition organic light-emitting diode (OLED) screen technology, with ultra-high resolution of 4K and above, and a refresh rate of up to 120Hz and above, which can create an extremely clear, smooth and realistic immersive visual inspection environment for the examinee. Its display color depth reaches 10 bits and above, and it can accurately present more than 1 billion different colors, providing a solid guarantee for the accurate display of color vision test patterns.

[0036] Furthermore, the host control system includes a central command control module, an eye fixation point analysis module, a color image display module, a data processing module and a data transmission module;

[0037] The central command control module is the core command unit of the entire system. It is connected to the eye fixation point analysis module, the color vision image display module, the interactive input handle and the data processing module respectively. It is used to control the eye tracking and analysis process, transmit the eye fixation point position information, control the timing and sequence of color vision test pattern display, interactive test process management, and data collection and transmission scheduling.

[0038] More specifically, the central command control module uses a specific custom command encoding format (based on binary and hexadecimal mixed encoding) to communicate efficiently with each module, and the data transmission frequency is as high as once every 10 milliseconds. When the detection process starts, it sends a precise startup command to the eye fixation point analysis module, receives and processes the eye fixation point data in real time, and accurately transmits the processed eye fixation point position information to the color vision image display module in a three-dimensional space coordinate data format (accurate to 0.05 pixels). At the same time, according to the preset detection algorithm and process logic, it accurately controls the key parameters such as the display timing (accurate to milliseconds), display order, and display duration of the color vision test pattern. In terms of interactive test process management, by real-time monitoring of the operation feedback information (such as color selection, pattern matching, etc.) input by the examinee through the interactive input handle during the detection process, the interrupt trigger mechanism (using a combination of hardware interrupts and software interrupts) quickly responds to and records these operations, coordinates the data flow and interactive operations between modules, and ensures the smooth progress of the entire detection process.

[0039] The color vision image display module is connected to the central command control module, and is used to present the color vision test pattern according to the eye fixation point position information, and to dynamically adjust the display parameters and optimize and calibrate the display effect;

[0040] The data processing module is connected to various parts of the VR terminal device and is used for data collection and organization, data feature extraction and conversion, and data transmission and storage preparation;

[0041] The data transmission module is connected with various modules of VR terminal equipment and cloud data storage and artificial intelligence processing modules to establish and maintain data transmission channels, ensure the efficiency and accuracy of data transmission, and protect the security and privacy of data transmission;

[0042] Furthermore, the infrared eye tracking camera is integrated into the VR head mounted display, connected to the central command control module and the eye fixation point analysis module, and uses advanced high-resolution infrared imaging sensors to collect the subject's eye movement data. The sampling frequency can be as high as 250Hz and above, which can capture the subject's subtle eye movements with extremely high accuracy and speed, and accurately collect rich eye movement data including iris position, pupil size changes, and eye rotation angles.

[0043] Furthermore, the eye fixation point analysis module includes:

[0044] The fixation point positioning and tracking unit is used to accurately locate the initial fixation point based on eye movement data and continuously track the dynamic changes of the fixation point. The positioning accuracy can reach 0.05 pixels, and the fixation point position information can be updated at a frequency of 250 times per second.

[0045] More specifically, after receiving the start command from the central command control module, the fixation point positioning and tracking unit immediately activates the infrared eye tracking camera to start collecting the subject's eye movement data. The collected eye movement data is processed in real time by combining a feature point matching algorithm based on deep learning (such as an improved SURF algorithm based on a deep convolutional neural network) with a high-precision geometric calculation model. The initial eye fixation point position coordinates of the subject are determined in a very short time (usually within 50 milliseconds), and the positioning accuracy can reach within 0.05 pixels. During the entire color vision detection process, real-time eye movement data from the infrared eye tracking camera is continuously received at an ultra-high frequency of 250 times per second or more. By constructing a hybrid filtering model based on the combination of the Kalman filter algorithm and the particle filter algorithm, the fixation point position information is updated in real time with high precision, and the position changes of the fixation point at different time points are accurately recorded, including key parameters such as horizontal and vertical displacement (accurate to 0.05 pixels), moving speed (in pixels / second, accurate to 0.1 pixels / second) and acceleration (in pixels / second², accurate to 0.01 pixels / second²).

[0046] The fixation point stability and feature analysis unit is used to evaluate the fixation point stability index, including calculating the fluctuation range and jitter frequency of the fixation point within 200-500 milliseconds, and extracting the fixation point feature information, such as the time and number of times the fixation point stays on a specific color area or pattern shape, and the change of movement mode when different colors transition, etc.;

[0047] More specifically, the fixation point stability and feature analysis unit conducts an in-depth time series analysis of the position data of the fixation point, and uses a sliding window algorithm (the window size can be dynamically adjusted according to the detection requirements, usually between 200-500 milliseconds) to calculate the stability parameters such as the fluctuation range (in pixels, accurate to 0.05 pixels) and jitter frequency (in Hertz, accurate to 0.1 Hertz) of the fixation point within a specific time window. By accurately comparing with the fixation point stability data model built based on a large amount of normal people's eye movement data (the model is built by combining big data clustering analysis with a deep learning classification model, with a sample size of more than 100,000 people with normal color vision), it can accurately determine whether the subject's fixation point stability is within the normal range. At the same time, we deeply explore the rich feature information in the movement trajectory of the fixation point, such as the time the fixation point stays in a specific color area or pattern shape (accurate to 0.1 milliseconds), the number of times it stays, and the changes in movement patterns when different colors transition (such as the sudden change angle of the movement direction, the sudden change amplitude of the movement speed, etc., with angles accurate to 0.1 degrees and amplitudes accurate to 0.05 pixels / second).

[0048] The data transmission and interactive collaboration unit is used to feed back data to the central command control module and work with the color image display module. The data transmission adopts a custom encrypted transmission protocol based on the SSL / TLS protocol, and the feedback data format is a custom extended format based on the JSON format.

[0049] More specifically, the data transmission and interactive collaboration unit will feed back the fixation point position information, stability parameters, and feature information obtained through calculation and analysis to the central command control module in real time using a specific encrypted data transmission protocol (such as a custom encrypted transmission protocol based on the SSL / TLS protocol), and will also transmit the fixation point-related data to the color vision image display module so that it can make intelligent and adaptive adjustments to the color vision test pattern based on the fixation point position and stability. For example, when the fixation point drifts slightly (the drift range is between 0.5-2 pixels), the color image display module can use a micro-displacement adjustment algorithm (an algorithm based on the combination of image pixel interpolation and geometric transformation) to fine-tune the pattern position with an adjustment accuracy of up to 0.05 pixels, thereby ensuring the relative stability of the pattern within the subject's visual range; when the stability of the fixation point is poor (such as the jitter frequency exceeds 10 Hz and the fluctuation range is greater than 2 pixels), the color image display module can appropriately reduce the display complexity of the pattern (such as reducing the number of colors in the pattern, reducing the color contrast, and the color contrast adjustment range can be dynamically adjusted between 10%-30% according to the specific situation) or extend the display time of the pattern according to the preset intelligent adjustment strategy (the extension of the display time is determined by a dynamic calculation model based on the instability of the fixation point, generally between 0.5-2 seconds).

[0050] Furthermore, the color image display module includes:

[0051] Color vision test pattern presentation unit, used to accurately display color vision test patterns according to the fixation point and display diversified patterns and colors. The color display accuracy reaches ΔE<1. The pattern library contains at least 5,000 different types of patterns. The patterns can be dynamically changed, such as color gradient, pattern flashing or movement, etc.

[0052] More specifically, the color vision test pattern presentation unit receives the eye fixation point position information from the central command control module, and presents the color vision test pattern at the corresponding position on the VR head-mounted display with ultra-high precision based on this information. Its color display accuracy reaches the ultra-high standard of ΔE<1, and it can present patterns of different colors, brightness, saturation, and various complex color combinations extremely accurately. The appropriate pattern is selected from the massive color vision test pattern library for display. The pattern library contains a wide variety of patterns, covering single basic colors such as red, green, blue, yellow, cyan, and magenta, as well as thousands of intermediate colors and complex colors mixed by these basic colors in different proportions. It also includes the same color with different brightness levels (such as more than 100 different brightness levels) and saturation changes (such as more than 50 different saturation levels). The shapes of the patterns include circles, squares, bars, triangles, polygons, and various irregular shapes. They can also be highly dynamically changed according to detection needs, such as color gradient (gradient speed can be precisely adjusted between 0.1-10 degrees / second), pattern flashing (flashing frequency can be precisely set between 0.5-20 Hz) or movement (movement speed can be precisely controlled between 0.1-10 pixels / second, and movement direction can be arbitrarily set between 0-360 degrees).

[0053] A display parameter dynamic adjustment unit, used to adjust the display stability according to the stability of the fixation point, such as adjusting the display parameters in a specific way when the fixation point stability index meets a specific condition, and adjusting the display content according to the detection process and the subject's reaction, such as adjusting the pattern difficulty level or color combination according to the subject's reaction time and error rate using a model based on a combination of a machine learning decision tree and a neural network;

[0054] More specifically, the display parameter dynamic adjustment unit works closely with the eye fixation point analysis module to receive information feedback about the stability of the fixation point in real time. When the stability of the fixation point is good (such as the jitter frequency is less than 5 Hz and the fluctuation range is less than 1 pixel), the color vision test pattern is presented with the default optimal display parameters, maintaining a stable refresh rate (such as 120Hz), brightness (such as 100 nits) and contrast (such as 100:1). When the eye fixation point analysis module detects that the fixation point is unstable to a certain extent, the intelligent adjustment mechanism is immediately activated to adjust the display parameters accordingly. For example, by using an adaptive adjustment algorithm based on a visual perception model (combining the human eye visual contrast sensitivity function and the image pyramid algorithm) to fine-tune the display position of the pattern so that it always closely follows the movement of the fixation point, or by using a dynamic time adjustment algorithm (based on a model combining an exponential function and a linear function) according to the degree of instability of the fixation point, the display time of the pattern is appropriately increased, so that the subject has more time to identify the color. According to the complex detection process and intelligent algorithm set by the central command control module, the color vision test pattern content displayed on the VR head-mounted display is dynamically adjusted at different detection stages. For example, at the beginning of the qualitative detection stage of color vision impairment, some simple patterns with distinct color differences are first displayed, such as the red and green circular combination pattern commonly used in red-green color blindness detection (the circular diameter is 100 pixels, and the color codes for red and green are specific and precise values, based on the definition of the CIE1931 color space), to preliminarily determine whether the subject has color vision abnormalities. If the subject shows a reaction suspected of color vision impairment at this stage, then in the subsequent quantitative detection stage, a series of more complex patterns with more subtle color differences are displayed, such as mixed patterns of red and green with different hues and saturations (the mixing ratio can be accurately adjusted between 0.1:9.9-9.9:0.1, using more than 100 different hues and more than 50 different saturations of red and green for combination), and combined with the subject's response data to these patterns, the system further accurately calculates the degree of color vision impairment. In addition, based on the real-time reaction of the subject during the test, such as too long reaction time (more than 3 seconds) or high error rate (error rate more than 30%), the intelligent difficulty adjustment algorithm (based on a model combining machine learning decision tree and neural network) is automatically started to adjust the difficulty level of the displayed pattern (the difficulty level can be precisely adjusted between 1-10) or the color combination (the adjustment strategy of the color combination is determined based on the combination of color psychology principles and data-driven optimization algorithms).

[0055] The display effect optimization and calibration unit is used for color calibration and consistency assurance to control the color deviation within ΔE<0.5, as well as to enhance the display effect and optimize visual comfort. For example, an edge smoothing algorithm based on multi-scale wavelet transform is used to improve pattern clarity and contrast, and an intelligent brightness adjustment algorithm and gamma value correction algorithm are used to adjust the pattern brightness and gamma value according to the subject's visual habits and VR device characteristics.

[0056] More specifically, the display optimization and calibration unit uses professional color calibration equipment (such as a high-precision spectrophotometer) to accurately calibrate the color display parameters of the VR head-mounted display before displaying the color vision test pattern. By comparing and fine-tuning with the system's built-in ultra-precise color standard database (a multi-dimensional color standard that uses a combination of CIE1931 color space and CIELAB color space), the brightness, contrast, and color balance of the red, green, and blue primary color channels are adjusted with high precision, so that the color of the displayed color vision test pattern is accurate and the color deviation is controlled within a very small range of ΔE<0.5. The display effect of the color vision test pattern is enhanced in all aspects. The edge of the pattern is refined and smoothed using advanced edge smoothing algorithms (such as edge smoothing algorithms based on multi-scale wavelet transforms) to effectively reduce the visual interference of jagged edges. The clarity and contrast of the pattern are improved by combining high dynamic range image (HDR) technology with local contrast enhancement algorithms (such as adaptive local contrast enhancement algorithms based on Retinex theory), making the difference between colors more obvious and easier for the examinee to distinguish. According to the hardware characteristics of the VR head-mounted display (such as the optical characteristics of the screen material, brightness curve, etc.) and the subjects, the color of the image is enhanced. Based on the examinee's personalized visual habits (the examinee's preference data for different brightness and contrast are obtained through pre-tests, and a personalized analysis method based on a combination of user portraits and visual perception models is adopted), an intelligent brightness adjustment algorithm (an algorithm based on a combination of an ambient light sensor and a machine learning regression model) and a gamma value correction algorithm (a gamma correction model based on a power function combined with a piecewise linear function) are used to accurately adjust the pattern's brightness (the brightness adjustment range can be dynamically adjusted between 50-200 nits) and gamma value (the gamma value adjustment range can be accurately adjusted between 1.8-2.5) to avoid discomfort to the eyes caused by overly bright or dark displays.

[0057] Furthermore, the data processing module includes:

[0058] Data collection and collation unit, used for multi-source data integration, built data storage format as a custom extended format based on JSON format, as well as data format unification and cleaning, removing abnormal data by setting data thresholds and a hybrid anomaly detection model based on the combination of the isolation forest algorithm and the local anomaly factor algorithm;

[0059] More specifically, the data collection and collation unit is responsible for comprehensively collecting massive amounts of data information from all parts of the VR terminal device. On the one hand, it collects key data points such as the subject's reaction data to the colors in the virtual environment, including the precise time (accurate to 0.1 milliseconds) when the subject reacts to different color and brightness stimuli, the accuracy of the judgment (recorded in detail with binary correct or incorrect identification), and the type of error that occurs (accurately recorded according to a variety of detailed classifications such as red-green confusion, blue-yellow confusion, brightness confusion, saturation confusion, etc.). On the other hand, it receives the eye movement data and fixation center data collected by the infrared eye tracking camera and deeply processed by the eye fixation point analysis module, including the precise position coordinates of the fixation point (recorded in three-dimensional coordinate data with an accuracy of 0.05 pixels), stability parameters (such as the fluctuation range within 200-500 milliseconds with an accuracy of 0.05 pixels, and the jitter frequency with an accuracy of 0.1 Hz), movement trajectory characteristics (such as the movement direction with an accuracy of 0.1 degrees, the speed change curve with an accuracy of 0.05 pixels / second²), and the offset of the fixation point under different color stimulation (the offset direction is expressed in angles with an accuracy of 0.1 degrees, and the offset distance is in pixels with an accuracy of 0.05 pixels) and other information. These multi-source data are efficiently integrated to form a complete and orderly data set. The data storage format adopts a custom extended format based on the JSON format, laying a solid foundation for subsequent in-depth analysis. The collected data is fully formatted and converted into a standard format that the system can uniformly process and analyze. For example, the coordinate data in the eye movement data is converted from the original device-specific coordinate format to the normalized coordinate format based on the CIE1931 color space, and the time data in the color response data is converted from the original device timestamp format to the millisecond time format based on the International Standard Time (UTC). At the same time, the data is strictly cleaned to remove noise data, outliers, and erroneous data caused by factors such as equipment failure or interference. By setting multi-dimensional data thresholds (such as the abnormal threshold of the fixation point coordinates is set to deviate from the normal range by more than 3 pixels, and the abnormal threshold of the reaction time is set to less than 0.1 seconds or greater than 5 seconds) and advanced anomaly detection algorithms (such as a hybrid anomaly detection model based on the combination of the isolation forest algorithm and the local anomaly factor algorithm), these abnormal data points can be accurately identified and eliminated to ensure that the data used for subsequent analysis has extremely high quality and reliability.

[0060] Data feature extraction and conversion unit, used for color reaction data feature extraction, such as calculating reaction time statistical features and error type distribution features, as well as eye movement data feature extraction and association, such as extracting fixation point trajectory curvature, bending direction change frequency and other features and performing correlation analysis with color reaction data features, using a correlation analysis algorithm based on a combination of structural equation model and Bayesian network to explore the internal connection between the two;

[0061] More specifically, the data feature extraction and conversion unit performs in-depth feature extraction operations on the subject's reaction data to color. Calculate various statistical features of reaction time, such as average reaction time (accurate to 0.1 milliseconds), standard deviation of reaction time (accurate to 0.1 milliseconds), skewness and kurtosis of reaction time (accurate to 0.01), etc., deeply analyze the distribution characteristics of error types, determine which color discrimination errors the subject is prone to, and whether the error types have a certain regularity (such as whether they are concentrated on certain color pairs or color categories, using a method based on association rule mining algorithm combined with cluster analysis algorithm for analysis). Extract the changing trend characteristics of color recognition accuracy, observe whether the subject's accuracy gradually increases, decreases or fluctuates as different color stimuli are presented throughout the detection process, and intuitively display this feature by establishing a high-precision accuracy change curve (using a curve fitting algorithm based on cubic spline interpolation combined with Gaussian process regression). Through these feature extraction operations, the original color response data is converted into feature vectors with more analytical value. The data conversion algorithm adopts a feature dimension reduction and conversion model based on the combination of principal component analysis (PCA) and linear discriminant analysis (LDA) to provide strong data support for subsequent color vision function evaluation and disorder analysis. Comprehensive and in-depth feature extraction is also carried out for eye movement data and fixation center point related data. Key features are extracted from the moving trajectory of the fixation point, such as the curvature of the trajectory (the calculation method adopts an accurate curvature calculation method based on differential geometry and spline curve fitting), the frequency of change in the bending direction (in Hertz, accurate to 0.1 Hertz), and the fractal dimension of the trajectory (accurately calculated using the box dimension method, accurate to 0.01). These features can deeply reflect the transfer pattern and stability of the subject's visual attention. The proportion of the time the fixation point stays in different color areas or specific patterns is calculated (expressed as a percentage, accurate to 0.1%) to analyze the differences in the subject's visual attention to different colors or patterns. The eye movement data features and color response data features are deeply correlated and analyzed to explore the intrinsic connection between the two. For example, using the association analysis algorithm based on the combination of structural equation model and Bayesian network, we study whether there is a significant correlation between the stability of the fixation point and the color recognition accuracy under specific color stimulation (using the multidimensional correlation analysis method combining Pearson correlation coefficient, Spearman rank correlation coefficient and Kendall harmony coefficient), or whether the movement trajectory characteristics of the fixation point can effectively predict the type of error in the subject's response to subsequent color stimulation (using the prediction model combining the support vector machine regression model and the random forest regression model for analysis), etc. Through this association analysis, we can deeply excavate more information about color vision function and visual behavior hidden behind the data, providing a multidimensional and high-precision data basis for comprehensive and accurate evaluation of the subject's color vision status.

[0062] The data transmission and storage preparation unit is used to transmit data to the cloud system. It adopts the AES-256 encryption algorithm and combines the digital signature technology based on the elliptic curve cryptography (ECC) to encrypt data. The transmission protocol is a custom optimized protocol based on the HTTP / 3 protocol, as well as local storage and cache management. The data cache adopts a hybrid cache strategy based on the LRU-K algorithm and the 2Q algorithm. The local backup frequency is once a day, and the cleanup strategy is an intelligent cleanup strategy based on the combination of data lifecycle management and heat analysis.

[0063] More specifically, after completing a series of complex operations such as data collection, sorting, cleaning, feature extraction and conversion, the data transmission and storage preparation unit transmits the processed data to the cloud data storage and artificial intelligence processing module in a high-speed and secure manner through the data transmission module. Data transmission uses high-intensity encryption technology for encryption processing. The encryption algorithm uses the AES-256 encryption algorithm, combined with the digital signature technology based on the elliptic curve cryptography (ECC) system to ensure the security of the subject's personal privacy information and test data, and prevent the data from being stolen or tampered with during transmission. During the transmission process, advanced transmission protocols (such as custom optimization protocols based on HTTP / 3 protocol) and efficient data encapsulation formats (such as custom extended formats based on ProtocolBuffers) are used to ensure the integrity and accuracy of data transmission and avoid data loss or transmission errors. At the same time, data is intelligently stored and cached locally. During the detection process, some key data is temporarily stored in the local cache. The data cache adopts a hybrid cache strategy based on the LRU-K algorithm combined with the 2Q algorithm so that it can be quickly read and used when needed, significantly improving the system's response speed and processing efficiency. For example, when the recent test data of the examinee needs to be quickly reviewed and analyzed, the data can be read directly from the local cache without downloading it from the cloud system again. The locally stored data is backed up and cleaned up regularly. The backup data is stored in the local redundant hard disk array and the off-site disaster recovery center. The backup frequency is once a day. The cleanup strategy adopts an intelligent cleanup strategy based on data lifecycle management and heat analysis to ensure the security of local data and the effective use of storage space.

[0064] Furthermore, the data transmission module includes:

[0065] The data transmission channel establishment and maintenance unit is used to establish communication links between internal modules to ensure the bidirectionality of data transmission and the connection between terminal devices and cloud systems. It can adapt to different network environments. The network connection detection mechanism sends multiple types of heartbeat packets every 1 second. When the connection is interrupted or abnormally fluctuates, remedial measures such as reconnection or dynamic adjustment of network parameters and switching of network access points are taken;

[0066] More specifically, the data transmission channel establishment and maintenance unit is responsible for building a stable and high-speed data transmission channel between various modules within the system. In the VR terminal device, ensure that the data processing module can efficiently receive data generated by the central command control module, the eye fixation point analysis module and the color vision image display module in the host control system. These data include control instructions from the central command control module (the instruction format is a specific binary and hexadecimal mixed encoding format), eye movement data and fixation center point analysis results generated by the eye fixation point analysis module (the data format is a custom extended format based on the JSON format), and related parameters for the display of color vision test patterns recorded by the color vision image display module (the parameter format is a specific custom parameter format), etc. Ensure the bidirectionality and real-time nature of data transmission, so that efficient information interaction can be carried out between modules. After completing the preliminary processing of the collected data, the data transmission module can promptly feed back the processed results to the central command control module through the data transmission module. The feedback data format is a custom feedback format based on the XML format, so that the central command control module can flexibly adjust the subsequent detection process and command sending according to these results. Establish a reliable connection between the VR terminal device and the cloud data storage and artificial intelligence processing module. This connection can adapt to different network environments, whether in the high-speed LAN environment within the hospital (network bandwidth can reach 10Gbps and above), or connected to a remote cloud server through the Internet (supporting multiple network access methods, such as 5G, wired broadband, satellite communications, etc., and can automatically switch and optimize according to network conditions), the data transmission module can automatically identify and adopt appropriate network protocols and communication methods (such as using a custom optimization protocol based on the TCP / IP protocol in a LAN environment, and automatically switching 5G / wired broadband / satellite communications according to network conditions in an Internet environment, and using corresponding adaptive optimization protocols) to ensure smooth data transmission. The stability of this connection channel is continuously monitored and maintained by adopting advanced network connection detection mechanisms, such as sending multiple types of heartbeat packets every 1 second (including traditional heartbeat packets based on the ICMP protocol and intelligent heartbeat packets based on custom application layer protocols) to detect whether the network connection is normal. Once the connection is interrupted or abnormal fluctuations occur (such as network delay exceeding 50 milliseconds or packet loss rate exceeding 1%), the data transmission module will automatically try to reconnect (the number of reconnection attempts is 10, and the interval between each attempt ranges from 1 to 10 seconds, and is dynamically adjusted according to network conditions) or take other intelligent remedial measures, such as dynamically adjusting network parameters (including network buffer size, transmission window size, routing strategy, etc.), switching network access points (such as switching from 5G to wired broadband or satellite communication, or intelligently switching between multiple available network access points), etc., to ensure the continuity and integrity of data transmission.

[0067] The data transmission efficiency and accuracy unit is used for data packaging and serialization. The data identification adopts a hybrid identification method based on XML tags and binary identification. The order follows multi-dimensional order rules such as data importance, data type and data generation time. The serialization format is a custom extended format based on Protocol Buffers. In addition, transmission error detection and correction adopts multiple error detection mechanisms such as checksum, cyclic redundancy check and message authentication code based on hash function. Error correction adopts error correction code or request to resend data according to different error degrees.

[0068] More specifically, the efficiency and accuracy of data transmission unit carefully packages and serializes the data to be transmitted in order to achieve efficient data transmission. Various types of data (such as numerical data, text data, image data, etc.) collected from different modules are sorted and combined according to strict rules to form data packets that are easy to transmit. For example, the subject's color response data, eye movement data, and related equipment parameters are integrated into a data packet, and each data item has a clear identification (using a hybrid identification method based on XML tags and binary identification) and order (arranged according to multi-dimensional order rules such as data importance, data type, and data generation time). In the serialization process, the data is converted into a format suitable for transmission in the network or other communication media, such as binary format or a specific message format (such as a custom extended format based on ProtocolBuffers). This can significantly reduce the storage space occupied by the data, improve transmission efficiency, and also facilitate the correct parsing and processing of the data by the receiving end. In the data transmission process, a variety of advanced error detection mechanisms are used to ensure the accuracy of the data. For example, comprehensive verification information is embedded in the data packet by adding checksum, cyclic redundancy check (CRC), and hash function-based message authentication code (HMAC). After receiving the data, the receiver can use this verification information to comprehensively check whether there are any errors in the data transmission process. If a data transmission error is detected, the data transmission module will start the intelligent error correction mechanism. For some minor errors, such as the flipping of a single data bit or the loss of a small amount of data, they can be automatically corrected by the error correction code (Error-Correcting Code); for more serious errors, such as data packet loss or severe damage, the sender will be requested to resend the data, and a strategy based on data redundancy transmission combined with a data recovery algorithm is adopted to ensure that the data finally transmitted to the receiving end is complete and correct.

[0069] The data transmission security and privacy protection unit is used for encrypted data transmission. The encryption algorithm is the AES-256 encryption algorithm combined with the digital signature technology based on the elliptic curve cryptography (ECC), with a key length of 256 bits, as well as user authentication and authorization. The transmission request is authenticated by verifying the unique identifier of the device, the user's login credentials or digital certificate.

[0070] More specifically, the data transmission security and privacy protection unit performs high-intensity encryption on the transmitted data in order to fully protect the privacy of the examinee and the security of the test data. The advanced AES-256 encryption algorithm is used, combined with the digital signature technology based on the elliptic curve cryptography (ECC), and the data is encrypted according to the encryption key (key length is 256 bits) pre-set by the system. In the process of data transmission from the VR terminal device to the cloud system, even if the data is intercepted, without the correct decryption key, the third party cannot obtain the true content of the data. When establishing a data transmission connection, strictly participate in the user authentication and authorization process. Ensure that only authorized devices and users can access and transmit test data. For example, the transmission request is authenticated by verifying the unique identifier of the device (such as the device's MAC address, IMEI number or UUID), the user's login credentials (such as username and password, the password is encrypted and stored using a salted hash algorithm, and combined with multi-factor authentication, such as fingerprint recognition, facial recognition, SMS verification code verification, etc.) or digital certificates (using certificates issued by authoritative digital certificate authorities, and regularly verifying and updating certificates). Only requests that have passed verification can establish data transmission channels, preventing unauthorized data access and transmission, further ensuring the data security of the system and the privacy of the examinees.

[0071] Furthermore, the data storage module adopts a distributed storage architecture, such as a hybrid architecture based on the Hadoop distributed file system combined with the Ceph distributed storage system, which has ultra-high capacity storage characteristics and can store massive amounts of test data. The storage capacity can be dynamically and infinitely expanded according to demand, and the reliability of data storage is as high as 99.999% or more, ensuring that the test data will not be lost due to hardware failure or other unexpected situations. The data storage module performs refined classification and storage of data, and indexes multiple dimensions such as the information of the examinee (including name, age, gender, ID number, contact information, etc.), test time, test data type (such as color reaction data, eye movement data, equipment operation data, etc.), and test result classification (such as normal color vision, various types of color vision disorders, fundus disease risk assessment results, etc.), to facilitate ultra-fast retrieval and query of data. For example, when querying the test data of a specific examinee within a certain period of time, these index information can locate and extract relevant data in a very short time (such as within 1 second).

[0072] Furthermore, the artificial intelligence processing module includes:

[0073] The color vision disorder qualitative analysis unit is used for deep data mining and feature learning. It uses a convolutional neural network (CNN), recurrent neural network (RNN) and attention mechanism to extract feature patterns and accurately judge the type of disorder after learning a large amount of sample data. It judges the type of color vision disorder based on an integrated classifier combining support vector machine (SVM), random forest (RF) and deep neural network (DNN), and can identify complex and rare color vision abnormalities.

[0074] More specifically, the color vision disorder qualitative analysis unit uses a deep learning algorithm to deeply mine the massive comprehensive data transmitted from the VR terminal device through the data transmission module. The deep learning algorithm adopts an innovative architecture that combines convolutional neural network (CNN), recurrent neural network (RNN) and attention mechanism. The CNN part is mainly used to learn the features of the color vision test pattern image data and the eye movement trajectory image when the subject observes the pattern. Its network structure includes more than 10 convolution layers (the convolution kernel size includes 3x3, 5x5, 7x7 and other specifications), more than 5 pooling layers (pooling size is 2x2, 3x3, etc.) and more than 3 fully connected layers. It can automatically extract local features in the image (such as color block distribution, edge shape, texture features, etc.), global features (such as color combination features of the entire pattern, overall shape features of the eye movement trajectory, spatial distribution features, etc.) and long-distance dependency features (realized through the attention mechanism, which can pay attention to the correlation between different areas in the image). The RNN part is used to process the sequence information of non-image data (such as reaction time, error type, etc.), and can learn the complex characteristics and laws of these data in time series, such as the changing trend of reaction time, the dynamic changes of the order and frequency of error types, etc. Through deep learning of a large number of sample data of different color vision conditions (including normal color vision and various types of color vision disorders) (the number of samples reaches more than 1 million), highly representative and distinguishing feature patterns are extracted. For example, for patients with red and green color blindness, their reaction time under red and green stimulation may be significantly prolonged (on average by more than 1.5 seconds), and the movement trajectory and stability of the fixation point in the red and green areas show completely different characteristics from those of normal color vision (such as the jitter amplitude of the fixation point in the red area increases by more than 5 pixels on average, and the stay time in the green area is shortened by more than 0.5 seconds on average). The artificial intelligence processing module can accurately learn and identify these feature patterns. When receiving the test data of a new subject, it is accurately matched and compared with the learned feature patterns. Based on advanced classification models, such as the integrated classifier combining support vector machine (SVM), random forest (RF) and deep neural network (DNN), it can accurately determine whether the subject has color vision disorder and the specific type of disorder. It can not only accurately distinguish common types such as red-green color blindness and blue-yellow color blindness, but also some more complex and rare color vision abnormalities, such as different degrees and types of color weakness (such as mild, moderate, severe red weakness, mild, moderate, severe green weakness, etc., with a segmentation accuracy of up to 0.1 level), special color vision disorders caused by acquired factors (such as drug side effects, complications of eye diseases, nervous system lesions, etc.), can also be effectively identified with strong learning ability and accurate model judgment ability, providing key qualitative basis for subsequent targeted diagnosis and treatment.

[0075] The color vision impairment quantitative analysis unit is used to quantitatively integrate multi-source data, establish complex mathematical models of multiple variables, such as a hybrid model combining a deep learning-based regression model with a traditional statistical regression model, and outputs of precision assessment, such as calculating color vision sensitivity scores, which are obtained through a hybrid formula based on a combination of a multivariate linear regression equation and a neural network output;

[0076] More specifically, the color vision disorder quantitative analysis unit comprehensively considers the various data performances of the subjects during the color vision test and conducts quantitative analysis. From the color reaction data, the distribution of reaction time (such as calculating the median, interquartile range, percentile and other statistics of the reaction time), the number and frequency of error types (such as counting the percentage of the number of occurrences of each error type in the total number of tests, and analyzing the trend of its change over time and color stimulation) and other factors are analyzed in detail; combined with the fixation point stability parameters in the eye movement data (such as the average, standard deviation, maximum, minimum value of the fixation point jitter amplitude, etc.), the quantitative value of the offset of the fixation point under different color stimuli (such as calculating the average offset distance, maximum offset distance, and distribution of the offset direction of the fixation point under red stimulation) and other information. By establishing a complex mathematical model with multiple variables, such as a hybrid model combining a deep learning-based regression model (such as a multi-layer perceptron regression model, a long short-term memory network regression model, etc.) with a traditional statistical regression model (such as a multivariate linear regression model, a ridge regression model, a Lasso regression model, etc.), these seemingly scattered data are organically integrated and converted into a quantitative indicator that can intuitively reflect the degree of color vision impairment. For example, a special color vision sensitivity scoring algorithm is designed and applied. Based on the comprehensive analysis results of the above multi-source data, a complex scoring calculation model formula (such as a hybrid formula based on a combination of a multivariate linear regression equation and a neural network output) is used to generate a color vision sensitivity score accurate to 0.1 points for the subject. This score can clearly indicate the severity of the subject's color vision impairment. The lower the score, the more severe the color vision impairment. Whether it is used to judge the severity of the disease in clinical diagnosis or in long-term color vision function tracking and monitoring, it can provide doctors and researchers with accurate and objective quantitative evaluation basis, so as to formulate personalized treatment plans, rehabilitation training plans or conduct further medical research.

[0077] The color vision disorder and fundus problem association assessment unit is used for big data association learning. It uses a combination of long short-term memory network (LSTM) and graph neural network (GNN) to process the association between color vision detection data and fundus disease clinical data, as well as potential risk prediction and judgment. Based on the detected color vision disorder characteristics, the trained model is used to output fundus disease risk assessment results, including disease type and risk probability;

[0078] More specifically, the color vision disorder and fundus problem association assessment unit uses deep learning algorithms to conduct deep association learning on massive amounts of color vision detection data and rich clinical data on fundus diseases. The deep learning algorithm uses an innovative architecture that combines long short-term memory networks (LSTM) with graph neural networks (GNN). Since the development of fundus diseases often has a time series and is closely related to the physiological structure and functional network of the eye, the LSTM network can effectively process the time series characteristics of color vision detection data and learn the potential relationship between color vision disorder characteristics and the development process of fundus diseases, while the GNN network can build a network model of the physiological structure and function of the eye, and conduct association analysis between fundus diseases and color vision disorders at the network level. For example, for patients with macular degeneration, their color vision impairment may gradually show specific color recognition abnormal patterns as the fundus lesions worsen (such as the ability to distinguish blue and yellow gradually decreases, the type of color recognition errors gradually changes from occasional confusion to frequent misjudgment, and the error rate increases exponentially over time) and the change pattern of fixation point (such as the stability of the fixation point in the visual field area corresponding to macular degeneration gradually decreases, the jitter amplitude gradually increases, and the movement trajectory becomes irregular and has specific fractal characteristics). The LSTM-GNN network can capture the long-term dependencies between these changes and the associated propagation paths in the eye physiological network. When predicting and judging, when it is detected that the subject's color vision impairment characteristics conform to a certain pattern related to fundus disease, the system outputs the fundus disease risk assessment results based on the trained model, including the types of fundus diseases that may be involved (such as macular degeneration, retinopathy, glaucoma, diabetic retinopathy, etc., with a subdivision accuracy of specific lesion subtypes) and the risk probability of disease development (expressed as a percentage, accurate to 0.1%). For example, if it is detected that the subject's color sensitivity is severely reduced under blue and yellow stimuli (color sensitivity score is less than 20 points), and the fixation point shows unstable jumping or deviation in the corresponding visual field area (the jitter amplitude of the fixation point exceeds 25 pixels and the irregularity index of the movement trajectory exceeds 0.9), and combined with other relevant data features, the system may predict that the subject is at risk of macular degeneration, and give a corresponding risk probability assessment based on the model (such as a 65.5% risk of macular degeneration), providing ophthalmologists with extremely valuable clues and auxiliary decision-making basis for early screening and diagnosis of fundus diseases, thereby achieving early detection and early treatment of ophthalmic diseases, and effectively improving the prevention and treatment effects of ophthalmic medical care.

[0079] The result output and feedback support unit is used to integrate and transmit the analysis results, integrate the qualitative and quantitative analysis of color vision impairment and the assessment results associated with fundus problems into a detailed report and transmit it to the feedback module, as well as intelligent decision-making assistance support, and recommend further examination items, treatment plan suggestions and rehabilitation guidance suggestions based on the analysis results.

[0080] More specifically, the result output and feedback support unit

[0081] The results of qualitative and quantitative analysis of color vision impairment and fundus problem-related assessment are fully integrated to form a complete, detailed, highly professional and readable analysis report. The report includes the basic information of the examinee (such as name, age, gender, ID number, etc.), an overview of color vision status (such as whether there is color vision impairment, the type and degree of impairment), the results of potential fundus disease risk assessment (such as whether there is risk, possible disease type and risk probability), and related data support and analysis basis (such as specific values ​​of various test data, feature analysis results, model prediction process and basis, etc.). Then, this analysis report is accurately transmitted to the feedback module through the data transmission module for subsequent display and storage to the user. In addition to generating analysis reports, it can also provide intelligent decision-making support for ophthalmologists or other professional medical personnel. For example, based on the analysis results, further examination items are recommended (such as specific fundus examinations, such as fundus fluorescein angiography, optical coherence tomography, electroretinogram, etc.; genetic testing, such as testing for genes related to certain hereditary color vision disorders; blood tests, such as testing blood sugar, blood lipids and other indicators related to fundus diseases), treatment plan suggestions (such as visual training methods for color vision disorders, such as color discrimination training, visual attention training, visual memory training, etc.; drug treatment for potential fundus diseases, such as the use of drugs to improve retinal microcirculation, anti-vascular endothelial growth factor drugs, neurotrophic drugs, etc., or surgical treatment methods) It provides treatment options, such as laser treatment for macular degeneration, vitrectomy, surgical repositioning of retinal detachment, etc.) and rehabilitation guidance suggestions (such as dietary adjustments, such as recommending the intake of more foods rich in vitamins A, C, E and antioxidants such as lutein and zeaxanthin; lifestyle improvements, such as avoiding prolonged use of the eyes, regular eye examinations, and controlling risk factors for chronic diseases such as blood pressure, blood sugar, and blood lipids; psychological counseling, such as providing psychological support and guidance for the psychological pressure and anxiety caused by color vision disorders or fundus diseases), to help medical personnel make more scientific, reasonable, and personalized medical decisions, and improve the quality and efficiency of ophthalmic medical services.

[0082] The implementation process and effects of the present invention are further illustrated below through actual cases:

[0083] Case 1: Ordinary color vision test

[0084] A 30-year-old male patient was tested for color vision in the ophthalmology clinic of a certain hospital. After the patient put on the VR and AI-based color vision function detection system, the detection process started. The central command control module quickly activated the infrared eye tracking camera, and the eye fixation point analysis module accurately located the patient's initial eye fixation point within 80 milliseconds, with a positioning accuracy of 0.05 pixels, and fed back the information to the central command control module. Subsequently, the color vision image display module presented a series of basic color vision test patterns on the VR head-mounted display according to the eye fixation point position, such as circular blocks of different colors, simple color line combinations, etc. During the detection process, the data processing module collected the patient's response data to color stimuli in real time (including reaction time accurate to 0.1 milliseconds, judgment accuracy and error type) and eye movement data (fixation point position accuracy of 0.05 pixels, stability and movement trajectory). For example, when a red circular color block is presented, the patient accurately judges the color within 0.45 seconds, and the fixation point is stable in the center area of ​​the color block, with a fluctuation range of 0.2 pixels; but when a green and blue mixed pattern is presented, the patient's reaction time is extended to 1.1 seconds, and the fixation point is slightly offset by 2 pixels. The data processing module organizes, cleans and extracts features from these data, and then transmits them to the cloud data storage and artificial intelligence processing module. After in-depth analysis, the artificial intelligence processing module determines that the patient's color vision is normal, but there is a certain reaction delay in distinguishing the mixed color of green and blue, which may be related to individual differences in visual attention allocation or color cognitive processing. After the test is completed, the feedback module generates a detailed report, including the conclusion that the patient's color vision is normal, details of each test data, and analysis of the delay in distinguishing the mixed color of green and blue. The report is also stored in the data storage module for subsequent reference by doctors or for medical research.

[0085] Case 2: Color vision impairment detection and analysis

[0086] A 45-year-old female patient came to see a doctor because she suspected she had a color vision problem. When using the detection system of the present invention, it was found during the detection process that the patient's reaction time was significantly prolonged when facing the red and green contrast patterns, with an average reaction time of 2.5 seconds, and the movement trajectory of the fixation point in the red and green areas was chaotic, with poor stability, and the fixation point jitter frequency reached 15 Hz, with a fluctuation range of 10 pixels. After the data was transmitted to the artificial intelligence processing module, after comparing and learning with a large number of color vision disorder characteristic patterns, the system determined that the patient had a tendency to red and green color blindness. Subsequently, the system entered the quantitative analysis stage, comprehensively considering the patient's color reaction data (such as the number of incorrect judgments for different red and green combinations reached 15 times, and the standard deviation of the reaction time distribution was 1.2 seconds) and eye movement data (the average jitter amplitude of the fixation point in the red and green areas was 8 pixels, the average offset was 6 pixels, etc.), and through a special color vision sensitivity scoring algorithm, it was calculated that the patient's color vision sensitivity score was low, 20 points, indicating that the degree of color vision disorder was more obvious. In addition, the AI ​​processing module further analyzed the relationship between the patient's color vision impairment characteristics and fundus diseases, combined with the abnormal changes in the fixation point in a specific area in the patient's eye movement data and the specific patterns in the color response data, predicted that the patient might be at risk of retinopathy, and gave a corresponding risk probability assessment of 40%. Based on the test report, the doctor arranged further fundus examinations for the patient, and ultimately diagnosed the patient with red-green color blindness and early retinopathy, and promptly developed a targeted treatment and rehabilitation plan.

[0087] Case 3: Screening for potential risk of fundus disease

[0088] A 60-year-old male patient used this detection system during a routine eye examination. During the color vision test, the patient's color sensitivity to blue and yellow stimuli decreased, with a color sensitivity score of only 15 points, and the fixation point showed unstable jumps and offsets in the corresponding visual field area. The jitter amplitude of the fixation point exceeded 20 pixels and the irregularity index of the moving trajectory reached 0.8. After processing, the data was transmitted to the artificial intelligence processing module, and the system analyzed it based on the established big data association model. Because the patient's color vision disorder characteristics and the characteristic patterns of macular lesions have a high degree of match, the system predicts that the patient has a risk of macular lesions and assesses the risk probability as a moderate risk of about 50%. Based on the test results, the doctor added special examination items related to macular lesions for the patient, such as fundus fluorescein angiography. The examination results confirmed that the patient had signs of early macular lesions. Due to timely discovery, the doctor took corresponding treatment measures to delay the progression of the disease and effectively protect the patient's vision.

[0089] This embodiment deeply integrates virtual reality (VR) technology and artificial intelligence technology and applies them to the field of eye color vision function detection. VR technology provides the subject with an immersive color vision detection environment, simulates the visual scenes of the real world, and enhances the authenticity and effectiveness of the detection; artificial intelligence technology achieves accurate qualitative and quantitative analysis of color vision disorders and correlation assessment with fundus problems through learning and analysis of massive data, overcomes the limitations of traditional detection methods that rely on subjective judgment, and greatly improves the accuracy and scientificity of the detection.

[0090] This embodiment uses an infrared eye tracking camera to accurately obtain the subject's eye movement data, and combines advanced algorithms to achieve accurate positioning, tracking, stability assessment and feature extraction of the eye fixation point. Through comprehensive analysis of these data and the subject's response data to colors in the virtual environment, it is possible to fully and deeply understand the subject's color vision function and potential visual problems, not only accurately determine the type of color vision disorder, but also accurately quantify its degree, and further explore the relationship with fundus diseases, providing richer and more valuable information for ophthalmic diagnosis.

[0091] The intelligent personalized detection system of this embodiment can dynamically adjust the display content, parameters and detection process of the color vision test pattern according to the real-time reaction and eye movement characteristics of the subject during the detection process. For example, the stability of the pattern display is adjusted according to the stability of the subject's fixation point, and the difficulty level or color combination of the pattern is adjusted according to the reaction time and error rate, so as to realize a personalized detection plan, improve the detection efficiency and accuracy, and also enhance the detection experience of the subject.

[0092] The artificial intelligence processing module of this embodiment is based on big data learning and complex model construction, and can provide intelligent decision-making support for ophthalmologists or other professional medical personnel. According to the color vision test results, it recommends further examination items, treatment plan suggestions, and rehabilitation guidance suggestions, helping medical personnel to make more scientific, reasonable, and personalized medical decisions, and promoting the development of ophthalmic medical services in the direction of intelligence and precision.

[0093] like Figure 3 As shown, based on the same inventive concept, this embodiment also provides a method for detecting eye color vision function integrating VR and artificial intelligence, comprising the following steps:

[0094] S1: Detection preparation and start: After the subject wears the detection system, the control terminal sends an initialization command to the central command control module of the host control system. The central command control module activates the infrared eye tracking camera and starts the eye fixation point analysis module, calculates the initial eye fixation point position, and instructs the color vision image display module to present the initial guidance pattern, and starts the detection process;

[0095] S2: Data collection and transmission: The data collection module collects the subject's reaction data to the colors in the virtual environment, eye movement data and related fixation center analysis data, and sends them to the image recognition and processing module;

[0096] S3: Image data processing and fusion: The image recognition and processing module pre-processes the color response data, performs positioning calibration, feature extraction and quantitative analysis on the eye movement data and fixation center point related data, deeply fuses the color response data and eye movement data, and sends them to the artificial intelligence analysis module;

[0097] S4: AI deep analysis: The AI ​​analysis module uses deep learning algorithms to perform qualitative and quantitative analysis of color vision disorders and analysis of fundus diseases on the processed data, and sends the results to the data analysis module;

[0098] S5: Data statistics and interpretation: The data analysis module counts and interprets the analysis results and sends the results to the feedback module;

[0099] S6: Result feedback and storage: The feedback module displays the statistical results to the user in the form of a report and simultaneously sends it to the data storage module for storage.

[0100] In summary, the eye color vision function detection system and method based on VR and artificial intelligence disclosed in the present invention represent a major innovation in the field of ophthalmic color vision detection technology. By cleverly integrating VR technology and artificial intelligence technology, many disadvantages of traditional color vision detection methods have been overcome. Its precise detection capabilities, efficient data processing and analysis mechanisms, and personalized detection solutions not only provide ophthalmologists with more reliable and comprehensive diagnostic tools, but also help to early detect and intervene in color vision disorders and related fundus diseases, and improve the quality of ophthalmic medical care; and also bring more convenient, comfortable and accurate detection experience to the examinees. In the future field of ophthalmic medical care, the present invention is expected to be widely used and further developed, and will play an important role in protecting people's visual health. At the same time, it will also promote the continuous innovation and integration of related technologies in the field of medical testing, inject new vitality into the development of intelligent medical care, and have extremely broad application prospects and significant social and economic benefits.

[0101] The above are only preferred specific implementations of the present application, but the protection scope of the present application is not limited thereto. Any changes or substitutions that can be easily thought of by a person skilled in the art within the technical scope disclosed in the present application should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A color vision function detection system integrating VR and artificial intelligence, characterized in that: include: VR terminal equipment, data storage module, artificial intelligence processing module, power module; The VR terminal device includes a VR head mounted display, an infrared eye tracking camera, a host control system and an interactive input handle; The data storage module is used to store detection data; The artificial intelligence processing module is used to perform qualitative analysis of color vision disorder, quantitative analysis of color vision disorder and correlation evaluation of color vision disorder and fundus problems on the detection data to obtain evaluation results; The power supply module is connected to the VR terminal device, the data storage module and the artificial intelligence processing module to provide power supply guarantee; The host control system includes a central command control module, an eye fixation point analysis module, a color image display module, a data processing module and a data transmission module; The central command control module is respectively connected to the eye fixation point analysis module, the color vision image display module, the interactive input handle and the data processing module, and is used to control the eye movement tracking and analysis process, transmit the eye fixation point position information, control the color vision test pattern display timing and sequence, interactive test process management and data collection and transmission scheduling; The color vision image display module is connected to the central command control module, and is used to present the color vision test pattern according to the eye fixation point position information, and to dynamically adjust the display parameters and optimize and calibrate the display effect; The data processing modules are respectively connected to various parts of the VR terminal device and are used for data collection and organization, data feature extraction and conversion, and data transmission and storage preparation; The data transmission module is respectively connected to the VR terminal device, the data storage module, and the artificial intelligence processing module, and is used for establishing and maintaining the data transmission channel, the efficiency and accuracy of data transmission, and the security and privacy protection of data transmission; The eye fixation point analysis module includes a fixation point positioning and tracking unit, a fixation point stability and feature analysis unit, and a data transmission and interactive collaboration unit; The fixation point positioning and tracking unit is used to locate the initial fixation point based on the eye movement data and continuously track the dynamic changes of the fixation point; The fixation point stability and feature analysis unit is used to evaluate the fixation point stability index and extract the fixation point feature information; wherein the fixation point stability index includes the fluctuation range and the jitter frequency of the fixation point, and the fixation point feature information includes the stay time and the number of stays on a specific color area or pattern shape, and the change of the movement mode when different colors transition; The data transmission and interactive collaboration unit is used to feed back data to the central command control module and work in coordination with the color image display module; wherein, data transmission adopts a custom encrypted transmission protocol based on the SSL / TLS protocol, and the feedback data format is a custom extended format based on the JSON format.

2. The system according to claim 1, characterized in that The infrared eye tracking camera is integrated into the VR head mounted display, connected to the central command control module and the eye fixation point analysis module, and is used to collect the eye movement data of the subject.

3. The system according to claim 1, characterized in that The color vision image display module includes a color vision test pattern presentation unit, a display parameter dynamic adjustment unit, and a display effect optimization and calibration unit; The color vision test pattern presentation unit is used to display the color vision test pattern according to the fixation point and to display diversified patterns and colors; The display parameter dynamic adjustment unit is used to adjust the display stability according to the stability of the fixation point, adjust the display parameters in a specific way when the fixation point stability index meets a specific condition, and adjust the display content according to the detection process and the subject's reaction, and adjust the pattern difficulty level or color combination according to the subject's reaction time and error rate using a model based on a combination of a machine learning decision tree and a neural network; The display effect optimization and calibration unit is used for color calibration and consistency assurance, controls color deviation, and adopts an edge smoothing algorithm based on multi-scale wavelet transform. It adjusts the pattern brightness and gamma value based on an intelligent brightness adjustment algorithm and a gamma value correction algorithm according to the visual habits of the subject and the characteristics of the VR device.

4. The system according to claim 1, characterized in that The data processing module includes a data collection and arrangement unit, a data feature extraction and conversion unit, and a data transmission and storage preparation unit; The data collection and sorting unit is used for multi-source data integration, data format unification and cleaning, and abnormal data is removed by setting data thresholds and a hybrid anomaly detection model based on a combination of an isolation forest algorithm and a local anomaly factor algorithm; The data feature extraction and conversion unit is used for color reaction data feature extraction and eye movement data feature extraction and association, and then uses a correlation analysis algorithm based on a combination of a structural equation model and a Bayesian network to explore the inherent relationship between the two; The data transmission and storage preparation unit is used to transmit data to the cloud system, and then encrypt the data using the AES-256 encryption algorithm combined with the digital signature technology based on the elliptic curve cryptography system.

5. The system according to claim 1, characterized in that The data transmission module includes a data transmission channel establishment and maintenance unit, a data transmission efficiency and accuracy unit, and a data transmission security and privacy protection unit; The data transmission channel establishment and maintenance unit is used to establish a communication link between internal modules, and connects the terminal device to the cloud system through two-way data transmission. The network connection detection mechanism sends multiple types of heartbeat packets every 1 second. When a connection is interrupted or abnormally fluctuates, reconnection or dynamic adjustment of network parameters and switching of network access points are adopted for remediation; The data transmission efficiency and accuracy unit is used for data packaging and serialization, wherein data identification adopts a hybrid identification method based on a combination of XML tags and binary identification, and the order follows the multi-dimensional order rules of data importance, data type and data generation time. The serialization format is a custom extended format based on Protocol Buffers, and transmission error detection and correction adopts an error detection mechanism of checksum, cyclic redundancy check and message authentication code based on hash function, and error correction adopts error correction code or request to resend data according to different error degrees; The data transmission security and privacy protection unit is used for data encryption transmission, as well as user authentication and authorization, and authenticates the transmission request by verifying the unique identifier of the device, the user's login credentials or digital certificate; wherein the encryption algorithm is the AES-256 encryption algorithm combined with the digital signature technology based on the elliptic curve cryptography system, and the key length is 256 bits.

6. The system according to claim 1, characterized in that The artificial intelligence processing module includes a color vision disorder qualitative analysis unit, a color vision disorder quantitative analysis unit, a color vision disorder and fundus problem correlation assessment unit, and a result output and feedback support unit; The color vision disorder qualitative analysis unit is used for deep data mining and feature learning. It adopts a convolutional neural network, a recurrent neural network and an attention mechanism to extract feature patterns after learning sample data, and determines the type of color vision disorder based on an integrated classifier combining a support vector machine, a random forest and a deep neural network to identify color vision abnormalities. The color vision disorder quantitative analysis unit is used for quantitative integration of multi-source data, establishing a hybrid model combining a deep learning-based regression model with a traditional statistical regression model, and obtaining a color vision sensitivity score by calculating a hybrid formula based on a combination of a multivariate linear regression equation and a neural network output; The color vision disorder and fundus problem association assessment unit is used to process the association between color vision detection data and fundus disease clinical data, as well as potential risk prediction and judgment, using an architecture that combines a long short-term memory network with a graph neural network, and output fundus disease risk assessment results using a trained model based on the detected color vision disorder characteristics; The result output and feedback support unit is used to analyze the risk assessment results and then integrate and transmit them, integrate the qualitative and quantitative analysis of color vision impairment and the assessment results associated with fundus problems into a detailed report and transmit it to the feedback module, as well as intelligent decision-making assistance support, and recommend further examination items, treatment plan suggestions and rehabilitation guidance suggestions based on the analysis results.

7. A method for detecting eye color vision function integrating VR and artificial intelligence, characterized in that: include: After the subject wears the detection system, the control terminal sends an initialization command to the central command control module of the host control system, and the central command control module activates the infrared eye tracking camera and starts the eye fixation point analysis module, calculates the initial eye fixation point position, and instructs the color vision image display module to present the initial guide pattern, and starts the detection process; The data collection module collects the subject's reaction data to the colors in the virtual environment, the eye movement data and the related fixation center point analysis data, and sends them to the image recognition and processing module; The image recognition and processing module pre-processes the color response data, performs positioning calibration, feature extraction and quantitative analysis on the eye movement data and fixation center point related data, deeply fuses the color response data and the eye movement data, and sends them to the artificial intelligence analysis module; The artificial intelligence analysis module uses a deep learning algorithm to perform qualitative analysis, quantitative analysis and corresponding analysis of color vision disorders on the processed data, and sends the results to the data analysis module; The data analysis module counts and interprets the analysis results and sends the results to the feedback module; The feedback module displays the statistical results to the user in the form of a report, and simultaneously sends them to the data storage module for storage.

Images