An intelligent screening system for Alzheimer's disease based on eye tracking

Abstract

The present invention relates to an intelligent screening system for Alzheimer's disease based on eye movement tracking. The system includes: a data capture module, which uses a webcam or a mobile phone camera to capture the eye movement data of the subject in a daily environment, while ensuring the time synchronization of the eye movement video and the viewing video; a data preprocessing and analysis module, which cleans the original eye movement data, extracts features, performs similarity analysis and Alzheimer's disease risk assessment; a user interface module, which provides an intuitive and easy-to-use interface that allows users to start detection, view results and receive feedback. The present invention can accurately identify abnormal eye movement patterns related to Alzheimer's disease by deeply analyzing eye movement features and combining comparisons of key visual moments, thereby achieving high-precision Alzheimer's disease risk assessment, and has low cost, user-friendly interface and strong adaptability.

Description

Technical Field

[0001] The present invention belongs to the field of medical equipment and information technology, and specifically relates to an intelligent screening system for Alzheimer's disease based on eye tracking. Background Art

[0002] Alzheimer's disease (AD) is a progressive neurodegenerative disease characterized by memory loss and cognitive impairment. Traditional diagnostic methods rely on clinical evaluation, neuropsychological testing, and biomarker testing, but require professionals and equipment to diagnose by detecting biomarkers such as β-amyloid and tau proteins in cerebrospinal fluid and plasma. [3] .

[0003] Traditional diagnostic methods are difficult to implement in resource-limited environments, are time-consuming and costly, and are not suitable for large-scale health screening. Eye tracking technology shows potential in AD diagnosis. Studies have shown that eye tracking parameters can distinguish patients with cognitive impairment and AD from normal elderly people. [1] AD is associated with oculomotor impairments, characterized by delayed saccadic movements, slow vertical saccadic movements, low accuracy, long latency, and low saccadic suppression. [2] Eye tracking technology has the potential to be an effective tool for early diagnosis, prompting early intervention and treatment.

[0004] Although eye tracking technology has shown potential in diagnosing Alzheimer's disease, traditional eye tracker equipment is expensive and may cause psychological stress to subjects, limiting its application in large-scale screening. Specifically, an eye tracker is a device used to capture and analyze human eye movements. Its principle is mainly based on tracking the position changes of the eyeball or pupil during visual tracking. Traditional eye tracker equipment often requires subjects to wear specific headgear or face the camera at a very close distance. This physical proximity may make people feel uncomfortable or anxious. Secondly, when most studies use eye trackers, the test method is to ask subjects to complete specific tasks or observe specific objects, which can also cause psychological stress to the subjects. In addition, most eye tracking devices require specific settings and calibration, which further limits their application in daily environments and large-scale public health projects. Summary of the invention

[0005] Based on the problems and challenges in the prior art, this paper proposes an AD intelligent screening system based on eye tracking for efficient and low-cost screening of AD, which can be used in scenarios such as family screening. The system is designed to address the limitations of traditional diagnostic methods in terms of resources, time and early detection, especially in the context of large-scale public health screening.

[0006] The technical solution adopted by the present invention is as follows:

[0007] An intelligent screening system for Alzheimer's disease based on eye tracking, comprising:

[0008] A data capture module, used to capture the eye movement data of the subjects when they are watching the video;

[0009] Data preprocessing and analysis module, used to preprocess and analyze the subjects' eye movement data and conduct risk assessment for Alzheimer's disease;

[0010] A user interface module is used to display the risk assessment results of Alzheimer's disease and interact with the user.

[0011] Furthermore, the data capture module captures the eye movement data of the subject when watching the video in the following manner:

[0012] Adding timestamps to eye tracking data and videos, said timestamps being based on the same system clock or synchronization source to ensure consistency;

[0013] A data synchronization strategy is used to compare and align the timestamps of the subject's eye movement data and the video, and synchronization is achieved by finding matching time points. In the case of time deviation, the time points are aligned by interpolation or adjusting the sampling frequency.

[0014] Furthermore, the data preprocessing and analysis module includes:

[0015] Anomaly detection submodule, used to identify abnormal frames in video sequences;

[0016] A data cleaning submodule, used to mark the abnormal frames identified by the abnormal detection submodule as invalid frames;

[0017] A feature extraction submodule is used to segment the pupil from the video frame and extract eye movement features, including eye movement trajectory, fixation duration, eye movement velocity and eye movement acceleration;

[0018] The standardization submodule is used to normalize the extracted features to the data standard;

[0019] The similarity analysis submodule is used to compare the eye movement trajectories of the subjects with those of normal people and identify abnormal eye movement patterns related to Alzheimer's disease;

[0020] The trend analysis submodule is used to monitor and analyze the changing trends of the subjects' eye movement data over a long period of time.

[0021] Furthermore, the feature extraction submodule includes:

[0022] A pupil segmentation unit is used to extract frames from the video as input to a pupil segmentation network, wherein the pupil segmentation network is a network trained using a data set containing various pupil sizes and shapes, and outputs pupil segmentation results for each frame in the time domain;

[0023] An eye movement trajectory calculation unit, used to calculate the eye movement trajectory, that is, the path of the subject's pupil movement while watching the video;

[0024] A fixation point duration calculation unit, used to calculate the fixation point duration, that is, the length of time the eyes fixate on a specific point;

[0025] An eye movement velocity calculation unit, used to calculate the eye movement velocity, which refers to the rate of change of pupil position over time;

[0026] The eye movement acceleration calculation unit is used to calculate the eye movement acceleration, that is, the rate of change of the eye movement velocity over time.

[0027] Furthermore, the normalization submodule uses a minimum-maximum normalization method to unify the data and scale the data to between 0 and 1.

[0028] Furthermore, the processing steps of the similarity analysis submodule include:

[0029] Similarity calculation of eye movement trajectories: The subject's eye movement trajectory and the predefined standard eye movement trajectory are input into the DTW algorithm, and the best alignment between the two time series is found by elastically stretching or compressing the time series, thereby calculating the similarity of the eye movement trajectories;

[0030] Similarity calculation of fixation point duration: set a threshold T, calculate whether the duration t of each fixation point is within the range of T, and take the proportion of fixation point duration within the threshold range as the similarity of fixation point duration;

[0031] Similarity calculation of eye movement velocity and acceleration: Calculate the velocity and acceleration from one frame to the next frame, and use cosine similarity to calculate the similarity of two time series;

[0032] By comprehensively considering the similarity of eye movement trajectories, the similarity of fixation duration, and the similarity of eye movement velocity and acceleration, a comprehensive similarity assessment score is obtained, which is used to evaluate the degree of proximity between the subjects' eye movement data and Alzheimer's disease-related features, thereby identifying abnormal eye movement patterns related to Alzheimer's disease.

[0033] Furthermore, the processing steps of the trend analysis submodule include:

[0034] Eye tracking tests were performed regularly to collect the subjects’ eye movement data;

[0035] Extract eye movement features from the eye movement data of each test, including eye movement trajectory, fixation duration, eye movement velocity, and eye movement acceleration;

[0036] Conduct statistical analysis to determine the trends of eye movement characteristics over time;

[0037] Set thresholds for different eye movement features based on clinical data to determine whether changes in eye movement features over time are significant;

[0038] The results of the analysis were compared with known patterns of cognitive disease progression to determine if there were signs of cognitive decline;

[0039] Provides easy-to-understand reports for medical professionals and subjects, including trends of eye movement characteristics and potential health effects.

[0040] Furthermore, the user interface module provides an intuitive and easy-to-use interface that allows users to initiate testing, view results, and receive feedback.

[0041] Furthermore, the data capture module is a webcam, a mobile phone camera or a wearable device, etc.; the user interface module is a portable device such as a smart phone.

[0042] Furthermore, the data preprocessing and analysis module includes: a digital video conversion box for receiving, preprocessing and storing eye movement data collected by the data capture module, including anomaly detection, data cleaning and feature extraction; a server for performing data standard normalization, similarity analysis and trend analysis.

[0043] Compared with the prior art, the AD intelligent screening system based on eye tracking of the present invention has the following advantages and benefits:

[0044] 1. High efficiency and low cost: Using common devices (such as webcams or smartphone cameras) to capture eye movement data greatly reduces the cost and technical threshold of AD diagnosis. This makes large-scale screening possible in resource-limited environments.

[0045] 2. User-friendly interface: It provides graphical eye movement display, AD risk assessment and interactive education modules to enhance the user experience. The system is designed with non-professionals in mind and can be self-tested without special training.

[0046] 3. High-precision risk assessment: Through in-depth analysis of eye movement characteristics and comparison of key visual moments, the system can accurately identify abnormal eye movement patterns related to AD and achieve high-precision AD risk assessment.

[0047] 4. Strong adaptability: The present invention can automatically adapt to different ambient lighting conditions and the distance between the subject and the camera to ensure accurate data collection; the time synchronization function ensures that the eye movement data matches the timeline of the viewing content, improving the accuracy of the analysis. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is the operation flow chart of the Alzheimer's disease screening system.

[0049] Figure 2 This is the overall architecture diagram of the Alzheimer's disease screening system.

[0050] Figure 3 It is a schematic diagram of the internal functions and component layout of the digital video converter box.

[0051] Figure 4 This is a schematic diagram of the connection between the digital video converter box and other components.

[0052] Figure 5 It is the design layout and functional diagram of the user interface. DETAILED DESCRIPTION

[0053] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below through specific embodiments and drawings.

[0054] At a specific moment in video playback, the eye movement trajectories of normal people show a high degree of consistency. For example, when a key plot or visual stimulus appears in a video (movie or other video), the gaze points and eye movement paths of normal viewers will show obvious similarities. By comparing the eye movement trajectories of the subjects with those of normal people, the present invention can effectively identify abnormal eye movement patterns associated with AD. This method can not only capture the subtle visual and cognitive differences associated with AD, but also greatly reduce the dependence on professional medical equipment and environment due to its non-invasiveness and convenience. The core of the present invention is an eye tracking system based on a webcam or a mobile phone camera. The system is used to assess the risk of AD by capturing and analyzing the eye movement data of the subjects while watching videos (movies or other videos) in real time.

[0055] The Alzheimer's disease screening system of the present invention includes three core modules: a data capture module, a data preprocessing and analysis module, and a user interface module. 1) The data capture module uses a webcam or a mobile phone camera to capture the eye movement data of the subject in a daily environment and stores it in a digital video conversion box. This module will remind the subject to improve the ambient lighting or adjust the distance from the screen to ensure that the conditions for data collection meet the requirements, while ensuring that the eye movement video is synchronized with the viewing video. 2) The data preprocessing and analysis module cleans, extracts features, performs similarity analysis and risk assessment on the original eye movement data. This module deeply analyzes the preprocessed data, including comparing the similarity of eye movement trajectories, analyzing eye movement velocity and acceleration, etc., to assess AD risk and carry out subsequent work. 3) The user interface module provides an intuitive and easy-to-use interface that allows users to start testing, view results and receive feedback.

[0056] Figure 1 is an operation flow chart of an Alzheimer's disease screening system according to an embodiment of the present invention, Figure 2 This is the overall architecture diagram of the system.

[0057] like Figure 1 As shown, the webcam (or mobile phone camera) of the system belongs to the data capture module of the present invention, which is a device for collecting eye movement video data of the subject, with high resolution and high frame rate to ensure clear and smooth video. The automatic light adjustment function ensures high-quality video under different environmental conditions. At least 1080p resolution, at least 30fps frame rate, automatic light adjustment function.

[0058] like Figure 1 As shown, the digital video conversion box of the system belongs to the data preprocessing and analysis module of the present invention. As the core data processing unit of the system, it is responsible for receiving, preprocessing and storing the eye movement video data collected by the camera. Through its high-speed processor and large-capacity storage, it can quickly process and save a large amount of eye movement data, while providing network connection to support data upload. The hardware configuration requirements are as follows: 1) Processor: Equipped with a 2.5GHz quad-core processor, it can efficiently process complex eye movement data; 2) Memory: At least 8GB RAM to ensure fast access and processing of data; 3) Storage: At least 256GB solid-state drive (SSD) to provide fast data reading and writing capabilities; 4) Network interface: Built-in Wi-Fi and Ethernet ports to facilitate data upload and remote access; 5) Video input: Support high-resolution video input, compatible with a variety of webcams and smartphone cameras; 6) Operating system: Built-in efficient and stable operating system to support data processing and analysis software. Figure 3 The figure shows the internal functions and component layout of the digital video conversion box used in one embodiment of the present invention. Figure 4It shows how to connect the digital video converter box with other components (network camera, server, TV, etc.).

[0059] like Figure 1 As shown, the server of the system belongs to the data preprocessing and analysis module of the present invention, which is mainly used for complex data analysis and long-term data storage. It has a multi-core processor, at least 32GB RAM, and a large-capacity hard disk.

[0060] like Figure 1 As shown, the smartphone of the system is a portable device, which belongs to the user interface module of the present invention. Its camera can be used to record eye movement data, and its high-performance processor and sufficient storage space support local data processing and storage, which is convenient for users to perform AD screening at any time. Therefore, it is recommended to use a smartphone with a high-performance processor and at least 4GB RAM, support at least 1080p video recording, and at least 64GB of internal storage.

[0061] The specific technical implementation plans involved in each component of the system are as follows:

[0062] 1. Data Capture Module

[0063] The data capture module is the basis of this system and is responsible for capturing the eye movement data of the subjects when watching video content. It includes two implementation methods: 1) When collecting eye movement data, add an accurate timestamp to each frame of eye movement video data. These timestamps should be based on the same system clock or synchronization source to ensure consistency. At the same time, add a timestamp to each frame or key frame of the video (movie or other video). 2) Use a data synchronization strategy to compare and align the timestamps of the eye movement data and the video (movie or other video), and complete the synchronization by finding matching time points in the two time series. In the case of time deviation, the time points can be aligned by interpolation or adjusting the sampling frequency.

[0064] 2. Data preprocessing and analysis module

[0065] The data preprocessing and analysis module contains a series of submodules to complete data processing and analysis and AD risk trend judgment. First, a preliminary check is performed, and the original eye movement video data imported from the data capture module is input into the anomaly detection submodule and the data cleaning submodule. By identifying anomalies in the video sequence and removing the moments when the subject turns his head or his eyes are off, it ensures that the data set contains only valid gaze data, thereby improving the accuracy of the analysis.

[0066] 1) Anomaly detection submodule

[0067] The anomaly detection submodule uses the pupil visibility detection algorithm (Pupil Visibility Detection, PVD), the specific formula is as follows:

[0068]

[0069] Let V i represents the pupil visibility at the tth frame, where V t =1 means pupil is visible, V t =0 means pupil is not visible, defining abnormal index A t is the number of times the pupil is invisible in N consecutive frames. If A t ≥ threshold, then the frames from time t-N+1 to time t are marked as abnormal frames.

[0070] 2) Data cleaning submodule

[0071] The data cleaning submodule marks the abnormal frames identified by the anomaly detection submodule as invalid frames, and the data of these abnormal frames will not be used in subsequent feature extraction and comparison.

[0072] 3) Feature extraction submodule

[0073] After that, the feature extraction submodule uses the U-Net network to segment the pupil and then extract key features such as eye movement trajectory, fixation duration, eye movement velocity and acceleration. The specific feature extraction calculation process and related formulas are described in detail as follows:

[0074] a) Pupil segmentation unit: extract frames from the video as input to the U-Net network, which is pre-trained on a dataset containing various pupil sizes and shapes for pupil segmentation. The U-Net network outputs pupil segmentation results based on each frame in the time domain, usually a pixel-level mask of the pupil area.

[0075] b) Eye movement trajectory calculation unit: The eye movement trajectory refers to the path of the pupil during viewing. The steps for calculating the eye movement trajectory include calibration before data collection, using the ResNet-50 network to predict the eye movement video frame to obtain the pitch and deviation angles of the pupil. Combined with the calibration results, the specific coordinates on the screen are calculated by linear interpolation to confirm the subject's gaze direction. The continuous recording of the eye movement trajectory is achieved by recording the coordinates of the pupil in each frame. If (x t ,y t ) represents the pupil position at time t, then the eye movement trajectory can be represented as a set of coordinate points {(x1,y1)(x2,y2),…,(x n ,y n )}.

[0076] c) Fixation duration calculation unit: The fixation duration refers to the length of time the eyes fixate on a specific point. It can be determined by detecting the stability of the pupil position within a certain time threshold. If the pupil position (xt ,y t ) does not change much within time T, then T is the duration of a gaze point, and Gaze Duration is used to represent the duration of a gaze point.

[0077] Gaze Duration = t end -t start

[0078] Among them, t start and t end are the timestamps of the start and end of fixation, respectively.

[0079] d) Eye movement velocity calculation unit: Eye movement velocity refers to the rate of change of pupil position over time. The specific formula is as follows:

[0080]

[0081] Where Δx and Δy are the horizontal and vertical changes in pupil position between adjacent frames, and Δt is the time interval between adjacent frames.

[0082] e) Eye movement acceleration calculation unit: Eye movement acceleration refers to the rate of change of eye movement velocity over time. The specific calculation formula is as follows:

[0083] a=Δv / Δt

[0084] Among them, Δv is the change in eye movement velocity between adjacent time points, and Δt is the time interval.

[0085] 4) Standardized submodule

[0086] The standardization submodule normalizes the extracted features and uses the minimum-maximum normalization method to unify parameters such as gaze point coordinates, eye movement velocity, and eye movement acceleration. The specific formula is as follows:

[0087]

[0088] Where X is the original data point, X min and X max are the minimum and maximum values ​​of the data column, respectively, norm is the normalized value, which scales the data between 0 and 1, helping to deal with scale differences captured by different cameras.

[0089] 5) Similarity analysis submodule

[0090] Next, the similarity analysis submodule is used to compare the eye movement trajectory of the subject with that of a normal person (the eye movement trajectory of a normal person can be obtained by the same method as above). By comparing the eye movement trajectory of the subject with that of a normal person, the system can effectively identify abnormal eye movement patterns related to AD. The following are the specific implementation steps of the similarity analysis submodule:

[0091] First, calculate the similarity of the eye movement trajectories. The dynamic time warping (DTW) algorithm is applied to the eye movement trajectories. The DTW algorithm is used to calculate the minimum distance between two time series. The subject's eye movement trajectory and the predefined standard eye movement trajectory are input into the DTW algorithm, which elastically stretches or compresses the time series to find the best alignment between the two time series to calculate the similarity. The specific calculation process is as follows:

[0092] a) Assume two time series A = a1, a2, ..., a m and B=b1,b2,…,b m , where m is the length of the sequence.

[0093] b) Construct an m×m matrix D, where D[i][j] represents the DTW distance between the first i elements of sequence A and the first j elements of sequence B.

[0094] c) Initialize the first row and column of matrix D.

[0095] d) For each element D[I][J] in D, the calculation formula is as follows:

[0096] d[I][j]=d(a i ,b i )+min(D[i-1],D[i][j-1],D[i-1][j-1])

[0097] Among them, d(a i ,b i ) is a i and b i The Euclidean distance between them. D[m][m] is the DTW distance between the two sequences.

[0098] Secondly, for features such as fixation duration, eye movement velocity, and acceleration, a standardized method can be used to calculate their similarity.

[0099] a) Similarity calculation of fixation point duration: Set a threshold T, and for each fixation point, calculate whether its duration t is within the range of T. The similarity score can be the proportion of fixation point durations within the threshold range.

[0100] b) Calculation of similarity of eye movement velocity and acceleration: First, calculate the velocity and acceleration from each frame to the next frame, and use cosine similarity to compare the similarity of the two sequences. The specific formula is as follows:

[0101]

[0102] Among them, X i and i are the elements of two sequences, and is their average value.

[0103] Then, the DTW distance and the similarity scores of other features are taken into consideration to obtain a comprehensive similarity assessment. This comprehensive score can be used to evaluate the closeness of the subject's eye movement data to AD-related features, thereby identifying abnormal eye movement patterns associated with Alzheimer's disease. Abnormal eye movement patterns can be derived based on the eye movement norms of healthy people (normal eye movement data of a large number of healthy people) and the comprehensive evaluation of clinical experts.

[0104] To give a specific example, there are the following features and weights:

[0105] DTW distance similarity score: 0.8, weight: 0.4;

[0106] Fixation duration similarity score: 0.7, weight: 0.3;

[0107] Eye movement speed similarity score: 0.6, weight: 0.2;

[0108] Eye movement acceleration similarity score: 0.5, weight: 0.1.

[0109] After weighted calculation, the comprehensive score is 0.70. Because the comprehensive score is obtained by comparing the subject's eye movement data with the eye movement trajectories of the normal population. In this case, the higher the score, the more similar the subject's eye movement data is to the eye movement trajectories of the normal population. Therefore, a higher score indicates that the subject's eye movement characteristics are closer to those of the normal population, thereby indicating a lower risk of AD. The lower the score, the greater the difference between the subject's eye movement data and the eye movement trajectories of the normal population. Therefore, a lower score indicates that the subject's eye movement characteristics are significantly different from those of the normal population, which may indicate a higher risk of AD.

[0110] In the above example, the composite score is 0.70, but whether it is considered a "high" or "low" score may depend on specific evaluation criteria or thresholds. These criteria need to be determined based on a large amount of experimental data and clinical research as well as comprehensive evaluation by experts in related industries to ensure the accuracy and reliability of the evaluation results.

[0111] 6) Trend analysis submodule

[0112] Through the above comprehensive evaluation method, the similarity between the subject's eye movement data and AD-related characteristics can be accurately evaluated in a single test. However, to fully understand the cognitive health status of an individual, it is crucial to monitor and analyze the changing trends of the subject's eye movement data over a long period of time. At this time, the trend analysis function plays a key role. The specific steps of the trend analysis link are as follows:

[0113] a) Collect eye movement data from subjects on a regular basis. This may involve performing eye tracking tests at regular intervals, for example every few months.

[0114] b) Extract key eye movement features from the data of each test, such as fixation duration, eye movement velocity, acceleration, eye movement trajectory, etc.

[0115] c) Perform statistical analysis to determine how these features change over time. This may include calculating the mean, standard deviation, etc. of the features and observing how they change over time.

[0116] d) Based on clinical data, thresholds are set for different eye movement features to determine whether the changes are significant.

[0117] e) Compare the results of the analysis with known patterns of cognitive disease progression to determine if there are signs of cognitive decline.

[0118] f) Provide easy-to-understand reports for medical professionals and subjects, including trends of eye movement characteristics and potential health effects.

[0119] Through trend analysis, subtle changes in cognitive function can be detected in a timely manner, which is crucial for early diagnosis and timely intervention in disease development.

[0120] In the data preprocessing and analysis module, considering the computing power and storage limitations of digital video converters and smartphones, this patent divides the task modules for local devices and servers. Modules running locally on the mobile phone: 1) Anomaly detection and data cleaning: These tasks usually do not require a lot of computing resources and can be performed efficiently on the mobile phone to remove invalid data and noise; 2) Feature extraction: such as eye movement trajectory, gaze point, eye movement speed and acceleration. These features can be extracted locally and stored in a compressed format to reduce the amount of data transmission. Tasks recommended for uploading to the server: 1) Data standard normalization: The server's high-performance computing resources are required to process data from different users and devices to ensure a unified analysis standard; 2) Similarity analysis and risk assessment: These tasks involve complex algorithms, such as the dynamic time warping (DTW) algorithm, which usually requires more powerful computing resources, so it is best to perform them on the server side; 3) Trend analysis: This requires long-term data accumulation and processing, so it is suitable to be performed on the server to track and analyze cognitive change trends over time.

[0121] 3. User interface module

[0122] The patented user interface design focuses on providing a clear, interactive and informative user experience. Figure 5 It is a user interface design diagram, showing the layout and some functions of the user interface. The following is a detailed description of the specific implementation of the user interface:

[0123] 1) Graphical eye movement trajectory display: An intuitive graphical interface is used to display the user's eye movement trajectory, including the path of the pupil moving on the screen, the gaze point, and the changes in eye movement speed and acceleration.

[0124] 2) AD risk level display: Display the AD risk level based on eye movement data analysis, using an intuitive rating system, such as color coding or percentage indicators, to indicate the risk level.

[0125] 3) User interaction and personalization: Allow users to enter personal information (such as age, gender, health status) and viewing preferences to personalize the detection process.

[0126] 4) Result export function: allows users to export AD risk assessment reports into PDF or other electronic formats, making it easier for users to record, archive and share with medical professionals.

[0127] 5) Interactive Education Module: Provides basic knowledge, preventive measures and health advice about AD. Enhances users’ health awareness through interactive educational tools such as infographics, videos and quizzes.

[0128] 6) User feedback mechanism: Collect user feedback on system usage experience and understanding of results to improve system design and functions and increase user satisfaction.

[0129] 7) Multi-language support: The design interface supports multiple languages ​​to meet the language needs of different users.

[0130] Through each of these features, the user interface is designed to provide an interactive, educational, and easy-to-use experience, while ensuring that users can fully understand and utilize eye tracking data to assess their own AD risk.

[0131] The innovation of this invention is focused on the intelligent eye tracking system, which provides a novel, efficient and low-cost early AD identification method. The key innovations include:

[0132] 1. Intelligent eye tracking system: Use a high-resolution webcam or smartphone camera to capture eye movement data in real time, including data capture, preprocessing and analysis, and user interface modules to accurately capture and analyze eye movement characteristics for early risk assessment of AD.

[0133] 2. Non-invasive eye tracking technology: Use ordinary webcams to replace traditional invasive eye tracker equipment to reduce the psychological pressure of subjects and improve test acceptability and comfort.

[0134] 3. Data preprocessing and analysis: Design a data preprocessing and analysis module to convert the raw eye movement data into a format suitable for analysis and extract key indicators such as eye movement trajectory, velocity, acceleration, etc. for risk assessment.

[0135] 4. User-friendly interface: Provides an intuitive and easy-to-use user interface that enables non-professional users to easily operate the system and understand the results, lowering the technical threshold, including detection startup, real-time data viewing and analysis reports.

[0136] 5. Daily environment adaptability and timing synchronization: The system can automatically detect the lighting and distance of the subjects to ensure stable data capture, while ensuring that the eye movement video is synchronized with the video (movie or other video) viewing time to ensure the accuracy of data analysis.

[0137] The intelligent eye tracking Alzheimer's disease diagnosis system of the present invention may also be implemented in other ways:

[0138] 1. Data capture: In addition to the current solutions, wearable devices (such as smart glasses) can also be used to replace traditional cameras to provide a more stable data collection environment suitable for long-term monitoring.

[0139] 2. Data processing and analysis: In addition to the current solution, deep learning technology can also be introduced to extract and analyze the features of eye movement data. The neural network model can be used to improve the accuracy of the system for AD-related abnormal eye movement patterns and enhance risk assessment.

[0140] These alternatives can further improve system adaptability, accuracy, and user experience.

[0141] The specific embodiments of the present invention disclosed above are intended to help understand the content of the present invention and implement it accordingly. It can be understood by those skilled in the art that various replacements, changes and modifications are possible without departing from the spirit and scope of the present invention. The present invention should not be limited to the contents disclosed in the embodiments of this specification, and the scope of protection of the present invention shall be subject to the scope defined in the claims.

Claims

1. An intelligent screening system for Alzheimer's disease based on eye tracking, characterized in that: include: A data capture module, used to capture the eye movement data of the subjects when they are watching the video; Data preprocessing and analysis module, used to preprocess and analyze the subjects' eye movement data and conduct risk assessment for Alzheimer's disease; A user interface module, for displaying the risk assessment results of Alzheimer's disease and interacting with the user; The data preprocessing and analysis module includes: Anomaly detection submodule, used to identify abnormal frames in video sequences; A data cleaning submodule, used to mark the abnormal frames identified by the abnormal detection submodule as invalid frames; A feature extraction submodule is used to segment the pupil from the video frame and extract eye movement features, including eye movement trajectory, fixation duration, eye movement velocity and eye movement acceleration; The standardization submodule is used to normalize the extracted features to the data standard; The similarity analysis submodule is used to compare the eye movement trajectories of the subjects with those of normal people and identify abnormal eye movement patterns related to Alzheimer's disease; The trend analysis submodule is used to monitor and analyze the change trend of the subjects' eye movement data over a long period of time; The processing steps of the similarity analysis submodule include: Similarity calculation of eye movement trajectories: The subject's eye movement trajectory and the predefined standard eye movement trajectory are input into the DTW algorithm, and the best alignment between the two time series is found by elastically stretching or compressing the time series, thereby calculating the similarity of the eye movement trajectories; Similarity calculation of fixation point duration: set a threshold T, calculate whether the duration t of each fixation point is within the range of T, and take the proportion of fixation point duration within the threshold range as the similarity of fixation point duration; Similarity calculation of eye movement velocity and acceleration: Calculate the velocity and acceleration from one frame to the next frame, and use cosine similarity to calculate the similarity of two time series; By comprehensively considering the similarity of eye movement trajectories, the similarity of fixation duration, and the similarity of eye movement velocity and acceleration, a comprehensive similarity assessment score is obtained, which is used to evaluate the degree of proximity between the subjects' eye movement data and Alzheimer's disease-related features, thereby identifying abnormal eye movement patterns related to Alzheimer's disease.

2. The system according to claim 1, characterized in that The data capture module captures the eye movement data of the subject when watching the video in the following manner: Adding timestamps to eye tracking data and videos, said timestamps being based on the same system clock or synchronization source to ensure consistency; A data synchronization strategy is used to compare and align the timestamps of the subject's eye movement data and the video, and synchronization is achieved by finding matching time points. In the case of time deviation, the time points are aligned by interpolation or adjusting the sampling frequency.

3. The system according to claim 1, characterized in that The feature extraction submodule includes: A pupil segmentation unit is used to extract frames from the video as input to a pupil segmentation network, wherein the pupil segmentation network is a network trained using a data set containing various pupil sizes and shapes, and outputs pupil segmentation results for each frame in the time domain; An eye movement trajectory calculation unit, used to calculate the eye movement trajectory, that is, the path of the subject's pupil movement while watching the video; A fixation point duration calculation unit, used to calculate the fixation point duration, that is, the length of time the eyes fixate on a specific point; An eye movement velocity calculation unit, used to calculate the eye movement velocity, which refers to the rate of change of pupil position over time; The eye movement acceleration calculation unit is used to calculate the eye movement acceleration, that is, the rate of change of the eye movement velocity over time.

4. The system according to claim 1, characterized in that The standardization submodule uses the minimum-maximum normalization method to unify the data and scale the data to between 0 and 1.

5. The system according to claim 1, characterized in that The processing steps of the trend analysis submodule include: Eye tracking tests were performed regularly to collect the subjects’ eye movement data; Extract eye movement features from the eye movement data of each test, including eye movement trajectory, fixation duration, eye movement velocity, and eye movement acceleration; Conduct statistical analysis to determine the trends of eye movement characteristics over time; Set thresholds for different eye movement features based on clinical data to determine whether changes in eye movement features over time are significant; The results of the analysis were compared with known patterns of cognitive disease progression to determine if there were signs of cognitive decline; Provides easy-to-understand reports for medical professionals and subjects, including trends of eye movement characteristics and potential health effects.

6. The system according to claim 1, characterized in that The user interface module provides an intuitive, easy-to-use interface that allows users to initiate testing, view results, and receive feedback.

7. The system according to claim 1, characterized in that The data capture module is a webcam, a mobile phone camera or a wearable device; the user interface module is a smart phone.

8. The system according to claim 1, characterized in that The data preprocessing and analysis module includes: A digital video conversion box, used to receive, pre-process and store the eye movement data collected by the data capture module, including anomaly detection, data cleaning and feature extraction; Server, used for data standard normalization, similarity analysis and trend analysis.

Images