A reading disorder screening system, method and terminal based on eye tracking technology

Abstract

The application provides a dyslexia screening system and method based on eye tracking technology, which matches the optimal eye movement test parameters according to the basic information of the target subject based on the pre-constructed eye movement test parameter norm table, and collects multiple frames of eye movement image data of the target subject in real time based on the eye tracking technology, obtains the original eye movement data and eye movement feature data including multi-dimensional eye movement features of the target subject, inputs the eye movement feature data into the pre-trained eye movement dysfunction recognition model, evaluates the eye movement function of the target subject, identifies the dyslexia risk of the target subject, and generates an eye movement test report in real time; thereby, various eye movement abnormalities can be finely identified, the accuracy and reliability of the eye movement test and the dyslexia screening can be effectively improved, and the eye movement test results of different target subjects and the multiple eye movement test results of the same target subject are comparable.

Description

Technical Field

[0001] This application relates to the field of smart healthcare technology, and in particular to a reading disorder screening system, method and terminal based on eye-tracking technology. Background Technology

[0002] The Developmental Eye Movement Test (DEM), proposed by Garzia et al. in the 1980s, is a standardized test used to assess eye movement and visual-related abilities. Its initial design aimed to differentiate the causes of reading difficulties, such as poor eye movement control or insufficient rapid naming speed. The DEM test generally consists of three parts: measuring naming speed by having the subject read a vertical sub-table; simulating the eye movement pattern of actual reading by having the subject read a horizontal sub-table; and differentiating the relative contribution of naming speed to eye movement control by calculating the ratio of time spent reading the horizontal to the vertical sub-tables. The DEM test is related to automatic digit naming, visual processing speed, and reading ability, and helps identify whether a subject is at risk of dyslexia. Due to its ease of use and wide applicability, the DEM test has become one of the commonly used eye movement and dyslexia screening tools in clinical and educational settings.

[0003] However, existing DEM testing methods still have significant shortcomings in clinical applications.

[0004] ① The test-retest consistency of DEM is not ideal. In a clinical setting, subjects' scores fluctuate significantly between different tests, resulting in poor reliability and stability of the DEM test, which limits its application in long-term follow-up and efficacy monitoring.

[0005] ② The DEM test lacks sensitivity in certain populations. A study by Tanke et al. showed that although the DEM test involves digit naming and horizontal eye movement tasks, it failed to accurately reflect the characteristics of eye movement disorders in patients with nystagmus. This indicates that the DEM test is not applicable to certain groups with severe or specific eye movement abnormalities.

[0006] ③ The performance of the DEM test not only reflects eye movement control ability, but is also often influenced by naming speed and visual processing speed. Gené-Sampedro et al. pointed out that for groups with poor reading ability, the results of the DEM test reflect more the limitations of language naming and visual processing speed than simple eye movement defects, thus reducing its independent validity as an eye movement diagnostic tool.

[0007] ④ Traditional paper-based DEM test data has a large granularity and cannot record details of eye movement trajectories such as fixation count, fixation duration, and saccade amplitude.

[0008] ⑤ The theoretical framework of DEM suggests that the horizontal sub-table mainly reflects saccade and retrace control. However, the study by Tanke et al. pointed out that the time difference between the horizontal and vertical sub-tables mainly depends on fixation time rather than the expected number of saccades or retraces. This makes it difficult for a single "horizontal-vertical ratio" indicator to accurately distinguish between naming speed and eye movement control, thus affecting the accuracy of dyslexia screening. Summary of the Invention

[0009] In view of the shortcomings of the prior art described above, the purpose of this application is to provide a reading disorder screening system, method and terminal based on eye-tracking technology, to solve the technical problems of insufficient reliability and stability, limited sensitivity, low data granularity and poor validity of the existing DEM test and its application in reading disorder screening.

[0010] To achieve the above and other related objectives, a first aspect of this application provides a reading disorder screening system based on eye-tracking technology. The system includes: an eye-tracking device; and a computer device communicatively connected to the eye-tracking device. The computer device acquires basic information of a target subject and, based on a pre-constructed norm table of eye-tracking test parameters, obtains eye-tracking test parameters for the target subject and sends them to the eye-tracking device. The eye-tracking device initializes according to the eye-tracking test parameters and performs an eye-tracking test on the target subject, generating multi-frame eye-tracking image data of the target subject. The computer device collects raw eye-tracking data of the target subject based on each frame of eye-tracking image data acquired from the eye-tracking device, and preprocesses and extracts features from the raw eye-tracking data to generate eye-tracking feature data of the target subject. Based on a pre-trained eye-tracking function abnormality recognition model and the eye-tracking feature data, the computer device assesses the eye-tracking function of the target subject, identifies the reading disorder risk of the target subject, and generates an eye-tracking test report for the target subject.

[0011] In some embodiments of the first aspect of this application, the raw eye-tracking data includes: pupil center coordinate sequence data, pupil diameter sequence data, corneal reflective point coordinate sequence data, fixation point coordinate sequence data, and eye-closing event sequence data; wherein, the method of acquiring the raw eye-tracking data based on each frame of eye-tracking image data includes: acquiring the sampling time of each frame of eye-tracking image data, and performing image preprocessing and image segmentation operations on each frame of eye-tracking image data to obtain the pupil region and corneal reflective region at each sampling time; if the pupil region and corneal reflective region are not obtained, the eye movement behavior at that sampling time is identified as an eye-closing event, and eye-closing event sequence data is generated; otherwise, pupil edge detection operations are performed on each pupil region. Ellipse fitting was performed on the obtained pupil edge point data to obtain pupil fitted ellipses for multiple sampling times. The pupil center coordinates and pupil diameter were calculated for each sampling time, and pupil center coordinate sequence data and pupil diameter sequence data were generated respectively. Reflection edge detection was performed on each corneal reflective area, and corneal reflective points in each corneal reflective area were identified to obtain the coordinates of corneal reflective points for each sampling time, and corneal reflective point coordinate sequence data were generated. Based on the pupil center coordinates and the coordinates of each corneal reflective point, the pupil-corneal reflection vector for each sampling time was calculated. Based on the pre-built coordinate transformation model, the fixation point was located, the fixation point coordinates for each sampling time were obtained, and fixation point coordinate sequence data were generated.

[0012] In some embodiments of the first aspect of this application, the preprocessing and feature extraction of the raw eye-tracking data includes: performing anomaly identification and correction operations on each time-series data in the raw eye-tracking data to obtain corrected pupil center coordinate sequence data, corrected pupil diameter sequence data, corrected corneal reflection point coordinate sequence data, corrected fixation point coordinate sequence data, and corrected eye-closing event sequence data to generate corrected eye-tracking data; calculating the eye movement velocity at each sampling time based on the fixation point coordinates of two consecutive sampling times in the corrected fixation point coordinate sequence data to distinguish the eye movement behavior at each sampling time as a fixation event or a saccade event, and identifying one or more consecutive fixation events and one or more consecutive saccade events; calculating the average fixation point coordinates and fixation duration of each consecutive fixation event based on the fixation point coordinates of multiple sampling times of each consecutive fixation event, and calculating the average fixation duration of all consecutive fixation events accordingly. The system generates gaze feature data by considering gaze dispersion and the proportion of long gazes. Based on the gaze point coordinates at multiple sampling times for each consecutive saccade event, it calculates the saccade amplitude and saccade speed for each consecutive saccade event, and accordingly calculates the number of retraces, average saccade amplitude, standard deviation of saccade speed, retrace path deviation rate, ratio of horizontal to vertical task total time, and skip rate for all consecutive saccade events. Based on the sampling time of each eye-closing event in the corrected eye-closing event sequence data, it identifies one or more consecutive eye-closing events, and calculates the eye-closing time for each consecutive eye-closing event and the average eye-closing time for all consecutive eye-closing events, generating eye-closing feature data. Based on the pupil diameters in the corrected pupil diameter sequence data, it calculates the pupil diameter variation amplitude and generates pupil diameter feature data. Finally, based on the gaze feature data, the saccade feature data, the eye-closing feature data, and the pupil diameter feature data, it generates eye movement feature data.

[0013] In some embodiments of the first aspect of this application, the method of assessing the eye movement function of the target subject and identifying the reading impairment risk of the target subject based on the eye movement function abnormality identification model and the eye movement feature data includes: obtaining the age and historical eye movement feature data of multiple healthy subjects based on a pre-constructed subject eye movement test database, and grouping each healthy subject by age to obtain historical eye movement feature datasets for multiple age groups; performing statistical analysis on the historical eye movement feature datasets for each age group to obtain an eye movement feature norm dataset for each age group, and constructing an eye movement function norm model accordingly; calculating standard scores for multiple eye movement features based on the eye movement function norm model and the eye movement feature data to construct a standard score vector for the eye movement features of the target subject, and filtering one or more abnormal eye movement features to label the eye movement function level of the target subject; and identifying the abnormal eye movement function pattern of the target subject based on the gradient boosting tree, the eye movement feature data, and the standard score vector for the eye movement features, analyzing the reading impairment risk of the target subject, and generating an eye movement test report for the target subject.

[0014] In some embodiments of the first aspect of this application, the construction of the eye-tracking test parameter norm table includes: based on a pre-built subject eye-tracking test database, obtaining the age and historical eye-tracking test parameters of multiple healthy subjects, and grouping each healthy subject by age to obtain historical eye-tracking test parameter sets for multiple age groups; performing statistical analysis on the historical eye-tracking test parameter sets for each age group to obtain eye-tracking test parameter norm data for each age group, and constructing the eye-tracking test parameter norm table accordingly.

[0015] In some embodiments of the first aspect of this application, the computer device updates the target subject's basic information, eye movement test parameters, raw eye movement data, and eye movement test report to a pre-built subject eye movement test database.

[0016] In some embodiments of the first aspect of this application, the eye-tracking device includes: a display unit, including a stimulus presentation area, for dynamically displaying multiple characters in a preset stimulus table at a preset stimulus speed; wherein the stimulus table includes a vertical number table, a horizontal number table, and reading material; an infrared emitting unit, for emitting invisible infrared light of a preset wavelength to irradiate the eyes of the target subject to generate corneal reflection points on the corneas of the target subject; a dual-camera unit, including two cameras, for simultaneously acquiring eye movement images of the target subject's eyes while the target subject reads the stimulus table based on a preset eye movement sampling frequency, and generating multi-frame eye movement image data of the target subject; and a voice guidance unit, for voice guidance of the target subject to complete the eye-tracking test.

[0017] In some embodiments of the first aspect of this application, before the eye-tracking device performs an eye-tracking test on the target subject, it further includes: sequentially presenting calibration points at multiple designated locations in the stimulus presentation area through the display unit, and having the target subject sequentially gaze at each calibration point, thereby collecting multiple eye-tracking calibration image data of the target subject's eyes; sending each eye-tracking calibration image data and the screen coordinates of each calibration point to the computer device, so that the computer device can calculate the pupil-corneal reflection vector of each calibration point based on each eye-tracking calibration image data, and construct a coordinate transformation model based on the screen coordinates of each calibration point.

[0018] To achieve the above and other related objectives, a second aspect of this application provides a method for screening reading difficulties based on eye-tracking technology, applied to a computer device. This computer device is communicatively connected to an eye-tracking device. The method includes: acquiring basic information of a target subject; obtaining eye-tracking test parameters of the target subject based on a pre-constructed norm table of eye-tracking test parameters and sending these parameters to the eye-tracking device for initialization; performing an eye-tracking test on the target subject to generate multi-frame eye-tracking image data of the target subject; collecting raw eye-tracking data of the target subject based on each frame of eye-tracking image data acquired from the eye-tracking device; preprocessing and extracting features from the raw eye-tracking data to generate eye-tracking feature data of the target subject; and evaluating the eye-tracking function of the target subject based on a pre-trained eye-tracking function abnormality recognition model and the eye-tracking feature data, identifying the risk of reading difficulties in the target subject, and generating an eye-tracking test report for the target subject.

[0019] To achieve the above and other related objectives, a third aspect of this application provides a reading disorder screening terminal based on eye-tracking technology. The reading disorder screening terminal based on eye-tracking technology includes: a memory and a processor; the memory is used to store a computer program; the processor is used to execute the computer program stored in the memory, so that the terminal performs the reading disorder screening method based on eye-tracking technology as described in any of the above embodiments.

[0020] As described above, this application provides a reading disorder screening system, method, and terminal based on eye-tracking technology. By using a pre-constructed norm table of eye-tracking test parameters and matching the optimal eye-tracking test parameters according to the target subject's basic information, and by using eye-tracking technology to collect multi-frame eye-tracking image data of the target subject in real time, the system obtains the target subject's raw eye-tracking data and eye-tracking feature data including multi-dimensional eye-tracking features. This eye-tracking feature data is then input into a pre-trained eye-tracking function abnormality identification model to assess the target subject's eye-tracking function, identify their reading disorder risk, and generate an eye-tracking test report in real time. Therefore, this application has the following beneficial effects: it can accurately identify various eye-tracking abnormalities, effectively improve the accuracy and reliability of eye-tracking tests and reading disorder screening, and make the eye-tracking test results of different target subjects and multiple eye-tracking test results of the same target subject comparable. This solves the technical problems of insufficient reliability and stability, limited sensitivity, low data granularity, and poor validity in existing DEM tests and their applications in reading disorder screening. Attached Figure Description

[0021] Figure 1 The diagram shown is a schematic representation of a reading disorder screening system based on eye-tracking technology according to an embodiment of this application.

[0022] Figure 2 The diagram shows a flowchart of a reading disorder screening method applied to an eye-tracking device and a computer device in one embodiment of this application.

[0023] Figure 3 The diagram shown is a flowchart illustrating the original eye-tracking data acquisition method in one embodiment of this application.

[0024] Figure 4 The diagram shown is a flowchart illustrating the method for preprocessing and extracting features from raw eye-tracking data in one embodiment of this application.

[0025] Figure 5 The diagram shown is a flowchart illustrating an eye movement dysfunction identification method according to an embodiment of this application. Detailed Implementation

[0026] The following specific examples illustrate the implementation of this application. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be noted that, unless otherwise specified, the following embodiments and features in the embodiments can be combined with each other.

[0027] To address the problems mentioned above, this application provides a reading disorder screening system, method, and terminal based on eye-tracking technology. The system aims to acquire multi-frame eye-tracking image data of the target subject in real time using eye-tracking technology, and through multi-dimensional eye-tracking feature analysis, input the obtained eye-tracking feature data into a pre-trained eye-tracking function abnormality identification model to assess the target subject's eye-tracking function and identify potential reading disorder risks. This solves the technical problems of insufficient reliability and stability, limited sensitivity, low data granularity, and poor validity in existing DEM tests and their applications in reading disorder screening.

[0028] Meanwhile, to make the objectives, technical solutions, and advantages of this application clearer, the technical solutions in the embodiments of this application are further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only for explaining this application and are not intended to limit the invention.

[0029] like Figure 1 The diagram illustrates the structure of a reading disorder screening system based on eye-tracking technology according to an embodiment of this application. In this embodiment, the reading disorder screening system based on eye-tracking technology includes an eye-tracking device and a computer device. The computer device is communicatively connected to the eye-tracking device.

[0030] It should be understood that computer equipment refers to electronic devices used for processing data. Specifically, it can be a complete computer system, such as a personal computer, server, or host computer, or it can be an embedded computer integrated into other devices, such as an eye-tracking device. Preferably, an FPGA chip can be used and integrated into the eye-tracking device to implement a reading disorder screening method based on eye-tracking technology.

[0031] like Figure 2 As shown, the reading disorder screening method applied to eye-tracking devices and computer devices mainly includes the following steps.

[0032] Step S1: The computer device acquires the basic information of the target subject and, based on the pre-constructed eye movement test parameter norm table, obtains the eye movement test parameters of the target subject and sends them to the eye movement tracking device.

[0033] The basic information includes, but is not limited to: the target subject's age, gender, corrected visual acuity, and history of reading difficulties or poor attention. The eye-tracking test parameters include, but are not limited to: stimulus presentation area size, stimulus table pattern, stimulus speed, stimulus table font size, and eye-tracking sampling rate.

[0034] The eye-tracking test parameter norm table is a reference standard table established based on eye-tracking test parameters from multiple healthy subjects, including eye-tracking test parameter norm data for multiple groups. Specifically, the eye-tracking test parameter norm data includes the mean and standard deviation of each eye-tracking test parameter set by the eye-tracking device during eye-tracking tests for the corresponding group of healthy subjects: historical stimulus presentation area size, historical stimulus table pattern, historical stimulus speed, historical stimulus table font size, and historical eye-tracking sampling rate.

[0035] In one embodiment, the grouping method of the eye movement test parameter norm table includes: grouping based on one or more pieces of information in the basic information, such as grouping based on age information, grouping based on age information and gender information, or grouping based on age information and corrected visual acuity.

[0036] Preferably, the eye-tracking test parameter norm table can be grouped based on age information, such as 6-6.5 years old, 6.5-7 years old, 7-7.5 years old, 7.5-8 years old, 8-8.5 years old, 8.5-9 years old, 9-9.5 years old, 9.5-10 years old, 10-10.5 years old, 10.5-11 years old, 11-11.5 years old, 11.5-12 years old, 12-12.5 years old, 12.5-13 years old, and other age groups. The eye-tracking test parameter norm table includes eye-tracking test parameter norm data for multiple age groups, so as to match the optimal eye-tracking test parameters according to the age of the target subject.

[0037] In this embodiment, the construction of the eye-tracking test parameter norm table includes the following steps.

[0038] ① Based on a pre-built subject eye-tracking test database, the ages and historical eye-tracking test parameters of multiple healthy subjects were obtained, and the healthy subjects were grouped by age to obtain historical eye-tracking test parameter sets for multiple age groups.

[0039] It should be noted that the subject eye-tracking test database stores eye-tracking test data from multiple subjects, including historical eye-tracking test parameters, historical raw eye-tracking data, historical eye-tracking feature data, and historical eye-tracking test reports from healthy subjects and subjects with abnormal eye-tracking function. The historical eye-tracking test parameters include, but are not limited to: historical stimulus presentation area size, historical stimulus table pattern, historical stimulus velocity, historical stimulus table font size, and historical eye-tracking sampling rate. Therefore, the historical eye-tracking test parameter set for each age group includes: historical stimulus presentation area size, historical stimulus table pattern, historical stimulus velocity, historical stimulus table font size, and historical eye-tracking sampling rate for all healthy subjects in the corresponding age group.

[0040] ② Statistical analysis was performed on the historical eye movement test parameter sets for each age group to obtain the eye movement test parameter norm data for each age group, and the eye movement test parameter norm table was constructed accordingly.

[0041] Specifically, the statistical analysis of the historical eye-tracking test parameter set includes: standardizing multiple historical stimulus presentation area sizes, multiple historical stimulus table patterns, multiple historical stimulus velocities, multiple historical stimulus table font sizes, and multiple historical eye-tracking sampling rates in the historical eye-tracking test parameter set to obtain the mean and variance of each historical stimulus presentation area size, each historical stimulus table pattern, each historical stimulus velocity, each historical stimulus table font size, and each historical eye-tracking sampling rate, so as to obtain the norm data of eye-tracking test parameters for the corresponding age group.

[0042] Therefore, based on the eye-tracking test parameter norm table, and according to the age of the target subject, a corresponding age group is matched to obtain eye-tracking test parameter norm data for that age group. Based on this eye-tracking test parameter norm data, the eye-tracking test parameters of the target subject are determined. Preferably, the mean value in the eye-tracking test parameter norm data can be used as the eye-tracking test parameter of the target subject; for example, the mean value of the historical stimulus presentation area size of healthy subjects in the corresponding age group can be used as the stimulus presentation area size of the target subject.

[0043] After obtaining the eye movement test parameters of the target subject, the computer device sends them to the eye movement tracking device that is connected to it.

[0044] Step S2: The eye-tracking device is initialized according to the eye-tracking test parameters and performs an eye-tracking test on the target subject to generate multi-frame eye-tracking image data of the target subject.

[0045] In one embodiment, such as Figure 1 As shown, the eye-tracking device includes, but is not limited to, a display unit, an infrared emitting unit, a dual-camera unit, and a voice guidance unit, all of which are communicatively connected to the computer device.

[0046] The display unit includes a stimulus presentation area for dynamically displaying multiple characters from a preset stimulus table at a preset stimulus speed.

[0047] It should be noted that the stimulus presentation area is the central area of ​​the display unit, including the display screen. In a preferred embodiment, the stimulus presentation area uses an LCD or OLED screen with a resolution of up to 1920*1080, which can clearly display the stimulus table to present the three types of eye-tracking test stimuli.

[0048] The stimulus table includes a vertical number table, a horizontal number table, and reading material. The vertical number table comprises multiple numbers arranged vertically to assess the target subject's number naming speed; preferably, the vertical number table typically includes 20 to 40 non-repeating random numbers, ranging from 1 to 99, in Microsoft YaHei font, with a font size of 28pt, and the number color is black (screen background is white). The horizontal number table comprises multiple numbers arranged in horizontal rows and columns to assess the target subject's saccade switching ability; preferably, the horizontal number table typically includes 20 to 40 non-repeating random numbers, ranging from 1 to 99, in Microsoft YaHei font, with a font size of 28pt, and the number color is black (screen background is white), arranged in 4 to 6 rows and 5 to 8 columns, with a row spacing of 1.2cm and a column spacing of 0.8cm. The reading material includes multiple meaningless syllables (such as "ba-da-ka-la", "mi-ni-pi") or pseudowords (such as "wug", "blick") to simulate natural saccades during reading. Preferably, the reading material has 8 to 10 characters per line, 3 to 5 lines in total, with a line spacing of 1.2 cm and a column spacing of 0.8 cm. The task sequence of the stimulus table is preferably set as the vertical number table, the horizontal number table, and the reading material to avoid interference from the task order on the eye movement test results and improve the reliability of the eye movement test.

[0049] In one embodiment, the size of the stimulus presentation area, the stimulus table pattern, the stimulation speed of each character in the stimulus table, and the font size are set according to the eye-tracking test parameters obtained from the computer device during the initialization of the eye-tracking device. The purpose of this design in this embodiment is to personalize the stimulus presentation area and the stimulus table based on the target subject's basic information, especially age, thereby adapting the stimulus range to the visual development level of subjects of different ages and facilitating concentration for subjects of different ages, thus improving the reliability of the eye-tracking test. For example, when conducting eye-tracking tests on subjects aged 6-8, the stimulus presentation area size can be set to 10cm*10cm; the stimulus table pattern can be set to a child-friendly mode, replacing numbers or letters with cartoon images (e.g., displaying the number "5" as a panda holding bamboo, and the number "8" as a snowman), and flashing colored dots at the edge of the stimulus presentation area after each correct identification by the young subject to alleviate test anxiety and improve cooperation. When subjects aged 9-13 or adult subjects undergo eye-tracking tests, the size of the stimulus presentation area can be set to 15cm*15cm, and the stimulus table pattern can be set to the standard pattern.

[0050] The infrared emitting unit is used to emit invisible infrared light of a preset wavelength to illuminate the eyes of the target subject, thereby creating corneal reflection points on the corneas of the target subject. Preferably, the emitted infrared light wavelength is set to 850 to 940 nm, and the infrared light is evenly illuminating the eyes of the target subject to avoid strong light stimulation.

[0051] The dual-camera unit includes two cameras used to simultaneously acquire eye-tracking images of the target subject's eyes while the target subject reads the stimulus table, based on a preset eye-tracking sampling frequency, and generate multi-frame eye-tracking image data of the target subject. Each frame of eye-tracking image data includes: the target subject's left eye-tracking image, right eye-tracking image, and sampling time; the left and right eye-tracking images respectively show the left pupil and left corneal reflective point, and the right pupil and right corneal reflective point. However, if the target subject closes their eyes, the corresponding pupil and corneal reflective point cannot be observed; this is identified as an eye-closing event.

[0052] It should be noted that the two cameras are set to the same eye-tracking sampling frequency and are respectively aimed at the left and right eyes of the target subject to simultaneously acquire left-eye and right-eye eye-tracking images. Preferably, the two cameras can be CMOS (Complementary Metal-Oxide-Semiconductor) cameras with a resolution of 1280*720, an eye-tracking sampling frequency of 120Hz to 240Hz, and a sampling time timestamp accuracy of ≤ 1ms.

[0053] In one embodiment, the preset eye-tracking sampling frequency of the dual-camera unit is set by initializing the eye-tracking device based on the eye-tracking sampling rate in the eye-tracking test parameters obtained from the computer device.

[0054] The voice guidance unit is used to guide the target subject to complete the eye movement test, including guiding the target subject to read the corresponding stimulus table before each type of eye movement stimulus task (e.g., "Next, please find the vertical number friends and read them out in order") and providing voice encouragement when the target subject pauses for a long time (e.g., "Keep going, you'll find the next one soon"), thereby alleviating the target subject's test anxiety and improving the target subject's cooperation.

[0055] This application initializes the eye-tracking device with personalized eye-tracking test parameters, enabling the formation of targeted eye-tracking test conditions for each target subject. This effectively ensures the comparability of eye-tracking test results from different subjects, strengthens the reference significance of the subject eye-tracking test database and the eye-tracking test parameter norm table, and improves the reliability and stability of eye-tracking tests. It overcomes the limitations of eye-tracking tests in long-term tracking of eye movements and visual abilities and in monitoring treatment efficacy, thereby enhancing the accuracy and reliability of eye-tracking tests in screening for reading disorders.

[0056] Furthermore, this application optimizes the stimulus table pattern for younger subjects and incorporates voice guidance, which can significantly reduce eye-tracking test anxiety in younger subjects. Experimental results show that the interruption rate of eye-tracking tests in younger subjects can be controlled to within 10%, enhancing the feasibility of large-scale screening.

[0057] In one embodiment, such as Figure 1 As shown, the eye-tracking device further includes a user interaction unit. Specifically, the user interaction unit can use a visual interactive interface (such as a touch screen) or an external input device (such as an external keyboard) to achieve human-computer interaction.

[0058] The user interaction unit is communicatively connected to the display unit, the infrared emitting unit, the dual-camera unit, the voice guidance unit, and the computer device. On one hand, it collects basic information of the target subject input by the user and sends it to the computer device, allowing the computer device to automatically match the target subject's eye-tracking test parameters. On the other hand, it collects eye-tracking test parameters of the target subject input by the user and sends them to the display unit, the infrared emitting unit, the dual-camera unit, and the voice guidance unit to initialize each unit. This allows users to manually input the target subject's basic information and manually adjust its eye-tracking test parameters, enhancing the intelligence of the eye-tracking device.

[0059] The user interaction unit can also be used to receive the user's control command to start eye movement testing and send it to the display unit, the infrared emitting unit, the dual camera unit and the voice guidance unit to start eye movement testing on the target subject.

[0060] In one embodiment, before the eye-tracking device performs an eye-tracking test on the target subject, it further includes: sequentially presenting calibration points at multiple designated locations in the stimulus presentation area through the display unit, and having the target subject sequentially gaze at each calibration point to collect multiple eye-tracking calibration image data; sending each eye-tracking calibration image data and the screen coordinates of each calibration point to the computer device, so that the computer device can calculate the pupil-corneal reflection vector of each calibration point based on each eye-tracking calibration image data, and construct a coordinate transformation model based on the screen coordinates of each calibration point.

[0061] Preferably, the eye-tracking device has a built-in 9-point calibration template, which specifies 9 calibration points to be presented sequentially at the four corners and the center of the stimulus presentation area, thereby performing 9-point calibration and establishing a spatial mapping relationship between screen coordinates and eye-tracking image pixel coordinates.

[0062] Step S3: The computer device collects the raw eye movement data of the target subject based on the eye movement image data of each frame obtained from the eye movement tracking device, and performs preprocessing and feature extraction on the raw eye movement data to generate eye movement feature data of the target subject.

[0063] The raw eye movement data includes, but is not limited to: pupil center coordinate sequence data, pupil diameter sequence data, corneal reflective point coordinate sequence data, fixation point coordinate sequence data, and eye closure event sequence data.

[0064] In one embodiment, such as Figure 3 As shown, the method for acquiring the raw eye-tracking data based on each frame of eye-tracking image data includes the following steps.

[0065] Step S31: Obtain the sampling time of each frame of eye-tracking image data, and perform image preprocessing and image segmentation operations on each frame of eye-tracking image data to obtain the pupil region and corneal reflective region at each sampling time.

[0066] Each frame of eye-tracking image data includes: the left eye-tracking image, the right eye-tracking image, and the sampling time of the target subject. The method for performing image preprocessing and image segmentation operations on the eye-tracking image data for a sampling time includes: performing image preprocessing and image segmentation operations on the left and right eye-tracking images for that sampling time respectively to obtain the corresponding left pupil region, left corneal reflective region, right pupil region, and right corneal reflective region.

[0067] Specifically, the image preprocessing operation for the left-eye or right-eye eye-tracking image includes: converting the left-eye or right-eye eye-tracking image into a grayscale image, and smoothing the grayscale image using a Gaussian blur algorithm to reduce image noise. The image segmentation operation for the preprocessed left-eye or right-eye eye-tracking image includes: using an adaptive thresholding segmentation algorithm to binarize the image and segment the pupil region and the corneal reflective region.

[0068] Step S32: If the pupil area and corneal reflective area are not obtained, the eye movement behavior at that sampling time is identified as an eye-closing event, and eye-closing event sequence data is generated.

[0069] It should be understood that during eye-tracking testing, when the target subject closes their eyes, the acquired eye-tracking images cannot detect the pupil and cornea. Therefore, if the pupil region and the corneal reflective region are not obtained after image preprocessing and image segmentation of the eye-tracking image data, the target subject can be defined as having closed their eyes at that sampling time, and the eye-tracking behavior at that sampling time can be identified as an eye-closing event. When multiple eye-closing events are identified, the eye-closing event sequence data can be generated based on the sampling time of each eye-closing event.

[0070] Step S33: Otherwise, perform pupil edge detection on each pupil region and perform ellipse fitting on the obtained multiple pupil edge point data to obtain pupil fitting ellipses for multiple sampling times, so as to calculate the pupil center coordinates and pupil diameter for each sampling time, and generate pupil center coordinate sequence data and pupil diameter sequence data respectively.

[0071] Specifically, if the pupil region has been obtained, pupil edge detection is performed on each obtained pupil region using an edge detection algorithm to obtain pupil edge point data for each pupil region. The pupil edge point data includes: left eye pupil edge point data and right eye pupil edge point data. It should be understood that edge detection algorithms are used in image processing to identify points with significant brightness changes in an image, including: Sobel operator, Prewitt operator, Canny operator, Laplacian operator, and LoG operator. Users can choose a specific edge detection algorithm according to their needs; this application is not limited to this.

[0072] Based on the obtained pupil edge point data for each pupil region, an ellipse fitting algorithm is used to perform ellipse fitting operations to obtain pupil fitting ellipses at each sampling time, which are then used to fit the pupils of the target subject. The pupil fitting ellipses include: a left eye pupil fitting ellipse and a right eye pupil fitting ellipse.

[0073] It should be understood that an ellipse fitting algorithm is a mathematical method for finding the best-fitting ellipse model from discrete data points, typically based on the general quadratic equation of an ellipse: Iterative search for the optimal The parameters are made to satisfy the constraints. And the equation of the quadratic curve. Mainstream ellipse fitting algorithms include: least squares method, minimum distance method, RANSAC (random sample consensus), and eigenvalue decomposition-based methods. Users can choose a specific ellipse fitting algorithm according to their needs; this application does not limit the choice.

[0074] In one embodiment, after obtaining the fitted ellipse for each pupil, the coordinates of the center point of each fitted ellipse are used as the pupil center coordinates for each sampling time. It should be noted that the pupil center coordinates, including the left eye pupil center coordinates and the right eye pupil center coordinates, are pixel coordinates, that is, they represent the position of the pupil center in the corresponding eye-tracking image.

[0075] In one embodiment, after obtaining the fitted ellipse for each pupil, the average value of the major and minor axes of each fitted ellipse is calculated as the pupil diameter for each sampling time. In another embodiment, the area equivalent diameter of each fitted ellipse may be used as the pupil diameter for each sampling time; this application is not specifically limited.

[0076] Based on the pupil center coordinates and pupil diameter at each sampling time, corresponding pupil center coordinate sequence data and pupil diameter sequence data are generated respectively.

[0077] Step S34: Perform reflective edge detection on each corneal reflective area and identify corneal reflective points in each corneal reflective area to obtain the coordinates of corneal reflective points at each sampling time and generate corneal reflective point coordinate sequence data.

[0078] Specifically, if the corneal reflective areas have been obtained, edge detection is performed on each obtained corneal reflective area using an edge detection algorithm to identify corneal reflective points within each area and locate them to obtain the coordinates of each corneal reflective point. These corneal reflective points include the left and right eye corneal reflective points, both represented by pixel coordinates, indicating their positions in the corresponding eye-tracking images. It should be understood that edge detection algorithms are used in image processing to identify points with significant brightness changes in an image, including operators such as Sobel, Prewitt, Canny, Laplacian, and LoG. Users can choose specific edge detection algorithms according to their needs; this application is not limited to any particular algorithm.

[0079] Based on the corneal reflective point coordinates at each sampling time, a corresponding sequence of corneal reflective point coordinates is generated.

[0080] Step S35: Based on the coordinates of each pupil center and the coordinates of each corneal reflective point, calculate the pupil-corneal reflection vector at each sampling time, locate the fixation point based on the pre-built coordinate transformation model, obtain the fixation point coordinates at each sampling time, and generate fixation point coordinate sequence data.

[0081] In this embodiment, the application employs a pupillary corneal reflection algorithm to locate the fixation point. The basic idea of ​​the pupillary corneal reflection algorithm is to estimate the direction of gaze by utilizing the relative positional relationship between the pupil center and the corneal reflective point, that is, to locate the fixation point of the target subject using the pupillary corneal reflection vector. The pupillary corneal reflection vector is a vector pointing from the coordinates of the pupil center to the coordinates of the corneal reflective point.

[0082] When locating the gaze point, a pre-constructed coordinate transformation model is used to obtain the screen coordinates of the gaze point, i.e., the gaze point coordinates, based on the pupil center coordinates and corneal reflection point coordinates of the pixel coordinates. The coordinate transformation model can be obtained from eye-tracking calibration image data acquired from the eye-tracking device and the screen coordinates of each calibration point. In one embodiment, the coordinate transformation model is constructed by: performing image preprocessing and image segmentation operations on each eye-tracking calibration image data to obtain multiple pupil regions and corneal reflective regions; performing pupil edge detection on each pupil region and ellipse fitting on the obtained multiple pupil edge point data to obtain multiple pupil fitted ellipses, so as to calculate the pupil center coordinates when the target subject gazes at each calibration point; performing reflective edge detection on each corneal reflective region and identifying corneal reflective points in each corneal reflective region to obtain the corneal reflective point coordinates when the target subject gazes at each calibration point; calculating multiple corresponding calibration pupil corneal reflection vectors based on the pupil center coordinates and the corneal reflective point coordinates; establishing a spatial mapping relationship between screen coordinates and eye-tracking image pixel coordinates based on the calibration pupil corneal reflection vectors and the screen coordinates of each calibration point, and constructing the coordinate transformation model, so as to calculate the gaze point coordinates at each sampling time based on the pupil corneal reflection vectors at each sampling time.

[0083] It should be noted that since the pupil center coordinates include the coordinates of the left and right pupil centers, and the corneal reflection points include the corneal reflection points of the left and right eyes, the calculated pupillary corneal reflection vectors also include the left and right pupillary corneal reflection vectors. Based on the left and right pupillary corneal reflection vectors, two corresponding fixation point coordinates can be obtained, which are used as the left and right eye fixation point coordinates, respectively. The average of the left and right eye fixation point coordinates is used as the fixation point coordinates of the target subject at the corresponding sampling time. This allows for mutual correction of the fixation point coordinates using the left and right eye fixation point coordinates, ensuring consistent line of sight for the target subject at the same sampling time, improving the accuracy of the fixation point, and thus ensuring the accuracy and reliability of the eye movement test results and the reading disorder screening results.

[0084] Furthermore, based on the gaze point coordinates at each sampling time, corresponding gaze point coordinate sequence data is generated. Then, combined with the pupil center coordinate sequence data, the pupil diameter sequence data, the corneal reflective point coordinate sequence data, and the eye closure event sequence data, the raw eye movement data is generated.

[0085] In one embodiment, such as Figure 4 As shown, the preprocessing and feature extraction of the raw eye-tracking data includes the following steps.

[0086] Step S36: Perform anomaly identification and correction operations on each time series data in the original eye movement data to obtain corrected pupil center coordinate sequence data, corrected pupil diameter sequence data, corrected corneal reflection point coordinate sequence data, corrected fixation point coordinate sequence data, and corrected eye closure event sequence data, so as to generate corrected eye movement data.

[0087] In one embodiment, a sliding window filtering algorithm is used to identify and correct outliers in each time series of the original eye movement data. Specifically, a window of a preset size is defined and slid across each time series of data, such as the pupil diameter sequence data or the fixation point coordinate sequence data. Statistical analysis is performed on the local data within the window to identify and correct outliers that deviate from the statistical characteristics of the local data. This eliminates jumps that do not conform to physiological patterns caused by factors such as blinking, temporary signal loss, and head micro-movements, thus smoothing the eye movement trajectory.

[0088] In a preferred embodiment, when the pupil diameter sequence data shows a sudden drop in pupil diameter greater than 50% for multiple consecutive sampling times with a duration ≥100ms, it is determined that the target subject has blinked. The original eye movement data from these multiple sampling times are then identified as outliers, and all original eye movement data from these sampling times are discarded. Conversely, when the fixation point coordinate displacement is ≥1cm for multiple consecutive sampling times (e.g., three consecutive sampling times), it is determined that the target subject has made a head movement. The original eye movement data from these multiple sampling times are then identified as outliers, discarded, and the missing values ​​are supplemented using a linear interpolation algorithm based on the original eye movement data from adjacent sampling times. It should be noted that when supplementing missing values ​​using the sampling interpolation algorithm, all original eye movement data from these multiple sampling times must be processed uniformly.

[0089] Step S37: Based on the gaze coordinates of two consecutive sampling times in the corrected gaze coordinate sequence data, calculate the eye movement velocity for each sampling time to distinguish the eye movement behavior of each sampling time as a gaze event or a saccade event, and identify one or more consecutive gaze events and one or more consecutive saccade events.

[0090] In one embodiment, a velocity threshold method is used to distinguish between fixation events and saccade events. It should be understood that the velocity threshold method refers to setting a velocity threshold based on the velocity characteristics of different eye movement events, classifying low-speed eye movements as fixation events and high-speed eye movements as saccade events.

[0091] Preferably, the speed threshold can be set to 30. If the eye movement speed during sampling time If the eye movement behavior at that sampling time is determined to be a fixation event, and if the eye movement behavior at multiple consecutive sampling times is determined to be a fixation event, it is identified as a continuous fixation event, and the fixation start time and fixation end time of the continuous fixation event are marked; if the eye movement velocity at the sampling time is... If the eye movement behavior at that sampling time is determined to be a saccade event, and if eye movement behavior at multiple consecutive sampling times is determined to be a saccade event with the same saccade direction, it is identified as a continuous saccade event, and the saccade start coordinates, saccade start time, saccade end time, saccade end coordinates, and saccade direction of the continuous saccade event are marked. and This represents the current sampling time and the coordinates of the gaze point at the current sampling time. and This refers to the coordinates of the gaze point at the previous sampling time.

[0092] It should be noted that the speed threshold can be adjusted according to experimental data and user needs, and this application does not limit it.

[0093] Step S38: Based on the gaze coordinates of multiple sampling times for each continuous gaze event, calculate the average gaze coordinates and gaze duration for each continuous gaze event, and calculate the average gaze duration, gaze dispersion, and long gaze percentage for all continuous gaze events to generate gaze feature data.

[0094] Specifically, the average coordinates of the gaze point in the continuous gaze event are the average of the coordinates of the gaze point at multiple sampling times of the continuous gaze event; the gaze duration is the length of time from the gaze start time to the gaze end time of the continuous gaze event.

[0095] The average fixation duration is the average of the fixation duration of all consecutive fixation events.

[0096] The gaze dispersion is the standard deviation or variance of the average coordinates of the gaze points across all consecutive gaze events, used to reflect the spatial dispersion of the gaze points. In one embodiment, the area covered by the gaze point coordinates at multiple sampling times of all consecutive gaze events can also be used, but the specific calculation method is not limited in this application.

[0097] The long fixation percentage is calculated by dividing the number of consecutive fixation events with a fixation duration exceeding a preset fixation duration threshold by the total number of consecutive fixation times, thus reflecting the proportion of long fixation events. In one embodiment, the fixation duration threshold is set to 500ms; when the fixation duration exceeds 500ms, it is identified as a long fixation event.

[0098] Step S39: Based on the gaze point coordinates of multiple sampling times for each continuous saccade event, calculate the saccade amplitude and saccade speed of each continuous saccade event, and accordingly calculate the number of retraces, average saccade amplitude, standard deviation of saccade speed, retrace path deviation rate, ratio of horizontal to vertical task total time and skip rate for all continuous saccade events, and generate saccade feature data.

[0099] Specifically, the saccade amplitude of the continuous gaze event is the distance between the saccade start coordinate and the saccade end coordinate of the continuous gaze event; the saccade speed is the saccade amplitude of the continuous gaze event divided by the difference between the saccade end time and the saccade start time.

[0100] The number of retracements refers to the number of consecutive retrace events where the retrace direction is opposite to the target retrace direction. It should be understood that reading generally involves scanning from left to right, so the target retrace direction is from left to right; while a retrace is generally a retrace from the end of one line to the beginning of the next line, with the retrace direction being from right to left, opposite to the target retrace direction.

[0101] The average saccadic amplitude is the average of the saccadic amplitudes of all consecutive saccadic events.

[0102] The standard deviation of saccade speed is the standard deviation of the maximum or average saccade speed of all consecutive saccade events, used to reflect changes in saccade speed. Specifically, the maximum saccade speed of a consecutive saccade event is the maximum value of the saccade speeds of multiple saccade events within that consecutive saccade event; the average saccade speed is the average value of the saccade speeds of multiple saccade events within that consecutive saccade event.

[0103] The retrace path deviation rate is the average deviation distance between the retrace trajectory and the ideal retrace path of multiple consecutive scanning events whose scanning direction is opposite to the target scanning direction. It is used to reflect the degree of deviation between each retrace trajectory and the ideal retrace path.

[0104] The ratio of the total time for horizontal to vertical tasks is the ratio of the total time spent completing horizontal tasks to the total time spent completing vertical tasks. Specifically, the total time spent completing horizontal tasks includes the total time spent scanning multiple lines of characters while reading the horizontal number table and the total time spent scanning multiple lines of characters while reading the reading material; the total time spent completing vertical tasks includes the total time spent reading the vertical number table, the total time spent scanning the horizontal number table for each iteration, and the total time spent scanning the reading material multiple times.

[0105] The skipping rate is the percentage of times the target character is skipped.

[0106] Step S310: Based on the sampling time of each eye-closing event in the corrected eye-closing event sequence data, identify one or more consecutive eye-closing events, and calculate the eye-closing time of each consecutive eye-closing event and the average eye-closing time of all consecutive eye-closing events to generate eye-closing feature data.

[0107] Specifically, when eye movement behaviors at multiple consecutive sampling times are all eye-closing events, they are identified as continuous eye-closing events, and the start and end times of eye-closing in each continuous eye-closing event are marked. The eye-closing time of a continuous eye-closing event is the duration from the start to the end of the eye-closing event. The average eye-closing time is the average of the eye-closing times of all continuous eye-closing events.

[0108] Step S311: Calculate the pupil diameter variation range based on each pupil diameter in the corrected pupil diameter sequence data, and generate pupil diameter feature data.

[0109] Specifically, the pupil diameter change range is the difference between the maximum and minimum pupil diameter during the eye movement test, used to reflect the degree of change in pupil diameter.

[0110] Step S312: Generate the eye movement feature data based on the gaze feature data, the saccade feature data, the eye closure feature data, and the pupil diameter feature data.

[0111] This application utilizes eye-tracking technology to collect multi-frame eye-tracking image data in real time during the eye-tracking test of the target subject. This allows for the real-time acquisition and recording of the target subject's raw eye-tracking data and eye-tracking feature data, including multiple fixations, saccade trajectories, and retracement counts. Compared to traditional eye-tracking tests that rely on manual timing, this avoids human error and omissions, eliminates subjective judgment, effectively improves the accuracy of eye-tracking tests, and ensures consistency across each test. Furthermore, the eye-tracking feature data generated by this application includes multi-dimensional eye-tracking features, which can be input into a pre-trained eye-tracking dysfunction identification model. This allows for the precise identification of different types of eye-tracking abnormalities, analysis of reading disorder risks, and assistance in analyzing the causes of reading disorders. Unlike traditional eye-tracking tests that only output total time and error rate, this application provides more comprehensive and accurate eye-tracking test results and reading disorder screening results.

[0112] Step S4: The computer device, based on a pre-trained eye movement function abnormality recognition model, assesses the eye movement function of the target subject according to the eye movement feature data, identifies the risk of reading impairment in the target subject, and generates an eye movement test report for the target subject.

[0113] The eye movement disorder (EMD) identification model can be trained based on a pre-built subject EEM test database and a gradient boosting tree. In one embodiment, the training method of the EMD identification model includes: constructing a training sample set based on the pre-built subject EEM test database by selecting eye movement feature data from multiple healthy subjects and multiple subjects with EMD abnormalities, and inputting this data into a gradient boosting training tree to train the EMD identification model; and using five-fold cross-validation to obtain the finally converged EMD identification model. Preferably, the training sample set of the EMD identification model includes eye movement feature data from 10,000 subjects of different age groups, including eye movement feature data from 7,000 healthy subjects and eye movement feature data from 3,000 subjects with EMD abnormalities. The cross-validation accuracy of the trained EMD identification model reaches over 85%, enabling it to learn meaningful abnormal eye movement patterns from the training data and possessing sufficient generalization ability and reliability.

[0114] It should be noted that this application may also employ a federated learning mechanism, allowing multiple participants (such as multiple hospitals) holding subject eye movement feature data to collaboratively train the eye movement dysfunction recognition model, thereby ensuring that the eye movement dysfunction recognition model maintains high consistency and reliability under different regions and different eye movement tracking device conditions. However, this application does not specifically limit this.

[0115] In one embodiment, such as Figure 5As shown, the method for assessing the eye movement function of the target subject and identifying the risk of reading impairment based on the eye movement feature data, according to the eye movement function abnormality identification model, includes the following steps.

[0116] Step S41: Based on the pre-built subject eye movement test database, obtain the age and historical eye movement feature data of multiple healthy subjects, and group each healthy subject by age to obtain historical eye movement feature datasets for multiple age groups.

[0117] As described above, the subject eye movement test database stores eye movement test data from multiple subjects, including historical raw eye movement data and historical eye movement feature data from healthy subjects and subjects with abnormal eye movement function. The historical raw eye movement data includes, but is not limited to: historical pupil center coordinate sequence data, historical pupil diameter sequence data, historical corneal reflection point coordinate sequence data, historical fixation point coordinate sequence data, and historical eye-closing event sequence data. The historical eye movement feature data includes, but is not limited to: historical fixation feature data, historical saccade feature data, historical eye-closing feature data, and historical pupil diameter feature data; that is, the historical eye movement feature data includes historical measured values ​​of multiple eye movement features. Therefore, the historical eye movement feature dataset for each age group includes: historical fixation feature data, historical saccade feature data, historical eye-closing feature data, and historical pupil diameter feature data from all healthy subjects in the corresponding age group. The historical fixation feature data includes, but is not limited to, eye movement features such as historical average fixation duration, historical fixation dispersion, and historical long fixation percentage; the historical saccade feature data includes, but is not limited to, eye movement features such as historical retrace count, historical average saccade amplitude, historical saccade speed standard deviation, historical retrace path deviation rate, historical ratio of total horizontal to vertical task duration, and historical skipping rate; the historical eye-closing feature data includes, but is not limited to, eye movement features such as average eye-closing time; and the historical pupil diameter feature data includes, but is not limited to, eye movement features such as pupil diameter variation.

[0118] Step S42: Perform statistical analysis on the historical eye movement feature datasets for each age group to obtain the eye movement feature norm datasets for each age group, and construct an eye movement function norm model based on it.

[0119] Specifically, the statistical analysis of the historical eye movement feature dataset includes: standardizing multiple eye movement features in the historical eye movement feature dataset to obtain the mean and variance of each eye movement feature, thereby obtaining an eye movement feature norm dataset for the corresponding age group, and constructing an eye movement function norm model to construct the normal eye movement function boundary for multiple age groups, representing the norm range of each eye movement feature.

[0120] In this embodiment, by constructing the eye movement function norm model, the eye movement function abnormality identification model can output the eye movement function level and abnormality prompt of the target subject according to the norm range of the same age group, thereby eliminating the influence of age on eye movement function assessment, significantly improving the scientific nature of eye movement testing, and ensuring the comparability of eye movement test results of subjects of different ages.

[0121] Step S43: Based on the eye movement function norm model, calculate the standard scores of multiple eye movement features according to the eye movement feature data to construct the standard score vector of eye movement features of the target subject, and filter out one or more abnormal eye movement features to mark the eye movement function level of the target subject.

[0122] The formula for calculating the standard score for each eye movement feature is as follows: .in, A standard score for a certain eye movement feature; This represents the measured value of a certain eye movement feature; The mean value of a certain eye movement feature in the eye movement function norm model; denoted as the standard deviation of a certain eye movement feature in the stated eye movement function norm model.

[0123] The purpose of this design in this embodiment is to unify all eye movement features to the same scale by calculating the standard scores of each eye movement feature, thereby making multiple eye movement features comparable and effectively preventing certain eye movement features from dominating the decision-making of the eye movement function abnormality recognition model due to their large dimensions, thus improving the robustness of the model.

[0124] In one embodiment, when the absolute value of the standard score of a certain eye movement feature exceeds a preset fluctuation threshold, the eye movement feature is determined to be an abnormal eye movement feature. Preferably, the fluctuation threshold is set to 1.5, that is, when the absolute value of the standard score of a certain eye movement feature exceeds 1.5, the eye movement feature is determined to be an abnormal eye movement feature.

[0125] Then, based on the number of abnormal eye movement features identified, the eye movement function level of the target subject is determined. It should be noted that the subject's eye movement function can be divided into multiple eye movement function levels, including: normal eye movement function, mild eye movement dysfunction, moderate eye movement dysfunction, and severe eye movement dysfunction.

[0126] In one embodiment, a corresponding threshold for the number of abnormal features can be set for each eye movement function level. When the number of identified abnormal eye movement features meets the threshold for a certain eye movement function level, the eye movement function of the target subject is determined to be at that level. Preferably, when no abnormal eye movement features are identified, or only one abnormal eye movement feature is identified, it is determined to be "normal eye movement function"; when 2-3 abnormal eye movement features are identified, it is determined to be "mild eye movement dysfunction"; when 4-5 abnormal eye movement features are identified, it is determined to be "moderate eye movement dysfunction"; and when 6 or more abnormal eye movement features are identified, it is determined to be "severe eye movement dysfunction".

[0127] In one embodiment, the eye movement function level of the target subject can also be determined based on the standard score of each eye movement feature. That is, when the standard score of one eye movement feature exceeds a preset standard score threshold for a certain eye movement function level, the eye movement function of the target subject is determined to be that eye movement function level. For example, when the standard score of one eye movement feature exceeds 2.5, it is determined to be "severe eye movement dysfunction".

[0128] It should be noted that the specific values ​​of the fluctuation threshold, the threshold for the number of abnormal features for each eye movement function level, and the standard score threshold can be set by the user according to their needs, and this application does not specifically limit them.

[0129] Step S44: Based on the gradient boosting tree, according to the eye movement feature data and the standard score vector of the eye movement feature, identify the abnormal eye movement function pattern of the target subject, analyze the reading difficulty risk of the target subject, and generate the eye movement test report of the target subject.

[0130] It should be understood that the gradient boosting tree algorithm is an ensemble machine learning algorithm that builds a strong learner by combining multiple decision trees. By gradually adding new decision trees, each new tree corrects the residuals of the previous tree and the negative gradient direction of the loss function, thereby iteratively training to achieve classification and continuously reducing classification error. Therefore, it performs excellently in classification and prediction tasks for structured data. Based on the eye movement feature data and the standard score vector of the eye movement features, it can accurately identify abnormal patterns of eye movement function and predict the potential reading difficulties of the target subject.

[0131] This application, based on eye-tracking technology, provides comprehensive and rich eye movement features, offering data support for subsequent eye movement testing research and dyslexia research. Through a pre-trained eye movement dysfunction identification model, it objectively links eye movements with reading ability, accurately predicting the potential dyslexia risk of the target subjects, thereby improving the accuracy and reliability of dyslexia screening. It also helps clinicians differentiate the causes of dyslexia, such as poor eye movement control, developmental oculomotor dysfunction, or other pathological causes, enabling clinicians to provide more accurate medical decisions and recommendations.

[0132] In one embodiment, the generated eye-tracking test report for the target subject includes, but is not limited to: the target subject's basic information, eye-tracking feature radar chart, details of abnormal eye-tracking features, eye-tracking test conclusions, and clinical recommendations. Preferably, the eye-tracking test report can be output in PDF format for easy printing.

[0133] The basic information is visualized on the basic information page of the eye-tracking test report. This basic information page also includes the eye-tracking test time and a summary of the eye-tracking test results. In a preferred embodiment, the basic information page may anonymously display the target subject's name, such as displaying only the target subject's ID number.

[0134] The eye-tracking feature radar chart refers to a radar-shaped, multi-dimensional icon representing multiple eye-tracking features in the eye-tracking feature data, visualized on the eye-tracking feature page of the eye-tracking test report. The eye-tracking feature radar chart includes a center point and multiple axes radiating outwards from the center point. The center point represents the baseline value for all eye-tracking features; each axis represents an eye-tracking feature, and each axis is marked with the measured value of that feature. Connecting the measured values ​​of each eye-tracking feature forms a closed polygon, reflecting the overall pattern of the target subject across all eye-tracking features. In a preferred embodiment, when displaying the eye-tracking feature radar chart on the eye-tracking feature page, different colors can be used to distinguish abnormal eye-tracking features, such as using red axes to mark abnormal eye-tracking features and blue axes to mark non-abnormal eye-tracking features.

[0135] The detailed abnormal eye movement features are visualized on the abnormality details page of the eye movement test report, including: the measured values, standard scores, and abnormality prompts for one or more identified abnormal eye movement features. These abnormality prompts include, but are not limited to: insufficient saccade amplitude, excessive retracement counts, and fixation instability, used to indicate the target subject's risk of reading difficulties. For example, a standard score of -1.8 for the eye movement feature "average saccade amplitude" indicates insufficient saccade amplitude; a score of 2.1 for the eye movement feature "retracement counts" indicates excessive retracement counts.

[0136] The eye-tracking test results are visualized on the conclusion suggestion page of the eye-tracking test report, including but not limited to: eye-tracking function level and reading disorder risk warnings. The reading disorder risk warnings include, but are not limited to: the probability of reading disorder and its manifestations. Manifestations of reading disorder include, for example, inattention or slow naming speed. For instance, if the eye-tracking feature "pupil diameter change amplitude" is less than 0.5mm, it suggests insufficient attention; if the eye-tracking feature "total vertical task time" is greater than twice the average for the same age, it suggests slow naming speed, etc.

[0137] The clinical recommendations are visualized on the conclusion and recommendations page of the eye movement test report, such as "It is recommended to go to the ophthalmology department for a detailed eye movement function examination" or "A colored guide ruler can be used to assist with daily reading," etc.

[0138] In one embodiment, when outputting the eye movement test report in PDF format, the eye movement test parameters, raw eye movement data, and eye movement feature data of the target subject can be simultaneously stored in CSV format and output to facilitate traceability and further eye movement abnormality analysis.

[0139] In one embodiment, the reading disorder screening method based on eye-tracking technology further includes: updating the target subject's basic information, eye-tracking test parameters, raw eye-tracking data, and eye-tracking test reports to a pre-constructed subject eye-tracking test database, thereby supporting continuous updates to the eye-tracking function abnormality identification model and continuously optimizing its accuracy in abnormality identification and reading disorder risk prediction. The subject eye-tracking test database can be stored offline locally, supporting more than 2000 eye-tracking test data sets. Preferably, it can also be synchronously and encrypted and uploaded to a cloud database, thereby supporting batch statistics and analysis by school, class, community, and other dimensions, meeting the needs of large-scale screening at the grassroots level.

[0140] It should be understood that the division of modules or units in the embodiments of this application is illustrative and only represents a logical functional division. In actual implementation, there may be other division methods. Furthermore, the functions of the computer device in the various embodiments of this application can be integrated into a single processor, exist physically separately, or be implemented as multiple functional modules. Each functional module can be implemented in hardware or as a software functional module. This application does not impose any limitations.

[0141] This application also provides a method for screening reading difficulties based on eye-tracking technology. In this embodiment, the method for screening reading difficulties based on eye-tracking technology is applied to the computer device described in the above embodiments, and includes: acquiring basic information of the target subject, and obtaining eye-tracking test parameters of the target subject based on a pre-constructed eye-tracking test parameter norm table, sending the eye-tracking parameters to the eye-tracking device for initialization, and performing eye-tracking tests on the target subject to generate multi-frame eye-tracking image data of the target subject; collecting raw eye-tracking data of the target subject based on each frame of eye-tracking image data acquired from the eye-tracking device, and preprocessing and extracting features from the raw eye-tracking data to generate eye-tracking feature data of the target subject; evaluating the eye-tracking function of the target subject based on a pre-trained eye-tracking function abnormality recognition model, and identifying the reading difficulty risk of the target subject, and generating an eye-tracking test report of the target subject.

[0142] It should be noted that the reading disorder screening method based on eye-tracking technology provided in the above embodiments and the reading disorder screening system based on eye-tracking technology belong to the same concept. For details of the specific method steps, please refer to the system embodiments, which will not be repeated here.

[0143] The reading disorder screening method based on eye-tracking technology for computer devices provided in this application embodiment can be implemented on the terminal side or the server side. The reading disorder screening terminal based on eye-tracking technology includes a processor and a memory.

[0144] The memory is used to store computer programs to implement the reading disorder screening method based on eye-tracking technology described in the above embodiments. The memory can be volatile or non-volatile, and may include both. Non-volatile memory can be read-only memory (ROM) or programmable read-only memory (PROM), which serves as an external cache. By way of example, but not limitation, many forms of RAM are available, such as static random access memory (SRAM) and synchronous static random access memory (SSRAM). The memories described in the embodiments of this application are intended to include, but are not limited to, these and any other suitable categories of memory.

[0145] The processor executes the computer program stored in the memory to cause the computer device to perform the reading disorder screening method based on eye-tracking technology described in the above embodiments. The processor may be an integrated circuit chip, a general-purpose processor, a digital signal processor (DSP), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The processor can implement or execute the methods, steps, and logic block diagrams disclosed in the embodiments of this application.

[0146] In summary, this application provides a reading disorder screening system, method, and terminal based on eye-tracking technology. By using a pre-constructed norm table of eye-tracking test parameters and matching optimal eye-tracking test parameters according to the target subject's basic information, and by using eye-tracking technology to collect multi-frame eye-tracking image data of the target subject in real time, it obtains the target subject's raw eye-tracking data and eye-tracking feature data including multi-dimensional eye-tracking features. This eye-tracking feature data is then input into a pre-trained eye-tracking function abnormality recognition model to assess the target subject's eye-tracking function, identify their reading disorder risk, and generate an eye-tracking test report in real time. This allows for the precise identification of various eye-tracking abnormalities, effectively improving the accuracy and reliability of eye-tracking testing and reading disorder screening, and making the eye-tracking test results of different target subjects and multiple eye-tracking test results of the same target subject comparable. Therefore, this application effectively overcomes the various shortcomings of existing technologies and has high industrial application value.

[0147] The above embodiments are merely illustrative of the principles and effects of this application and are not intended to limit this application. Any person skilled in the art can modify or alter the above embodiments without departing from the spirit and scope of this application. Therefore, all equivalent modifications or alterations made by those skilled in the art without departing from the spirit and technical concept disclosed in this application should still be covered by the claims of this application.

Claims

1. A reading disorder screening system based on eye-tracking technology, characterized in that, include: Eye-tracking devices; Computer equipment, wherein the computer equipment is communicatively connected to the eye-tracking device; The computer device acquires the basic information of the target subject and, based on a pre-constructed norm table of eye movement test parameters, obtains the eye movement test parameters of the target subject and sends them to the eye movement tracking device. The eye-tracking device is initialized according to the eye-tracking test parameters, and performs an eye-tracking test on the target subject to generate multi-frame eye-tracking image data of the target subject; The computer device collects the raw eye movement data of the target subject based on the eye movement image data of each frame obtained from the eye movement tracking device, and performs preprocessing and feature extraction on the raw eye movement data to generate eye movement feature data of the target subject. The computer device, based on a pre-trained eye movement dysfunction recognition model, assesses the eye movement function of the target subject according to the eye movement feature data, identifies the risk of reading difficulties in the target subject, and generates an eye movement test report for the target subject. The eye-tracking test parameters include: stimulus presentation area size, stimulus table pattern, stimulus speed, stimulus table font size, and eye-tracking sampling rate. The eye-tracking device includes: a display unit, including a stimulus presentation area, for dynamically displaying multiple characters in a preset stimulus table at a preset stimulus speed; wherein the stimulus table includes a vertical number table, a horizontal number table, and reading material; an infrared emission unit, for emitting invisible infrared light of a preset wavelength to illuminate the eyes of the target subject, thereby generating corneal reflection points on the corneas of the target subject; a dual-camera unit, including two cameras, for simultaneously acquiring eye movement images of the target subject's eyes while the target subject reads the stimulus table based on a preset eye movement sampling frequency, and generating multi-frame eye movement image data of the target subject; and a voice guidance unit, for providing voice guidance to the target subject to complete the eye-tracking test; the size of the stimulus presentation area of ​​the display unit, the stimulus table mode, the stimulus speed and font size of each character in the stimulus table, and the preset eye movement sampling frequency of the dual-camera unit are respectively set according to the eye-tracking test parameters during the initialization of the eye-tracking device; Based on the eye movement function abnormality identification model, the method for assessing the eye movement function of the target subject and identifying the risk of reading impairment based on the eye movement feature data includes: obtaining the age and historical eye movement feature data of multiple healthy subjects based on a pre-constructed subject eye movement test database, and grouping each healthy subject by age to obtain historical eye movement feature datasets for multiple age groups; performing statistical analysis on the historical eye movement feature datasets for each age group to obtain eye movement feature norm datasets for each age group, and constructing an eye movement function norm model accordingly; based on the eye movement function norm model, calculating standard scores for multiple eye movement features according to the eye movement feature data to construct a standard score vector for the eye movement features of the target subject, and filtering one or more abnormal eye movement features to label the eye movement function level of the target subject; based on a gradient boosting tree, identifying abnormal eye movement function patterns of the target subject according to the eye movement feature data and the standard score vector for the eye movement features, analyzing the risk of reading impairment of the target subject, and generating an eye movement test report for the target subject.

2. The reading disorder screening system based on eye-tracking technology according to claim 1, characterized in that, The raw eye-tracking data includes: pupil center coordinate sequence data, pupil diameter sequence data, corneal reflection point coordinate sequence data, fixation point coordinate sequence data, and eye closure event sequence data; wherein, the raw eye-tracking data is acquired based on each frame of eye-tracking image data in the following ways: The sampling time of each frame of eye-tracking image data is obtained, and image preprocessing and image segmentation operations are performed on each frame of eye-tracking image data to obtain the pupil region and corneal reflective region at each sampling time. If the pupil area and corneal reflective area are not obtained, the eye movement behavior at that sampling time is identified as an eye-closing event, and eye-closing event sequence data is generated; Otherwise, pupil edge detection is performed on each pupil region separately, and ellipse fitting is performed on the obtained multiple pupil edge point data to obtain pupil fitting ellipses at multiple sampling times, so as to calculate the pupil center coordinates and pupil diameter at each sampling time, and generate pupil center coordinate sequence data and pupil diameter sequence data respectively. The reflective edge detection operation is performed on each corneal reflective area, and the corneal reflective points in each corneal reflective area are identified to obtain the coordinates of the corneal reflective points at each sampling time, and generate corneal reflective point coordinate sequence data. Based on the coordinates of each pupil center and each corneal reflective point, the pupil-corneal reflection vectors at each sampling time are calculated. Based on the pre-built coordinate transformation model, the fixation point is located, the fixation point coordinates at each sampling time are obtained, and fixation point coordinate sequence data is generated.

3. The reading disorder screening system based on eye-tracking technology according to claim 2, characterized in that, The methods for preprocessing and feature extraction of the raw eye-tracking data include: Anomaly identification and correction operations are performed on each time series data in the original eye movement data to obtain corrected pupil center coordinate sequence data, corrected pupil diameter sequence data, corrected corneal reflection point coordinate sequence data, corrected fixation point coordinate sequence data, and corrected eye closure event sequence data, so as to generate corrected eye movement data. Based on the gaze coordinates of two consecutive sampling times in the corrected gaze coordinate sequence data, the eye movement velocity at each sampling time is calculated to distinguish the eye movement behavior at each sampling time as a gaze event or a saccade event, and to identify one or more consecutive gaze events and one or more consecutive saccade events. Based on the gaze coordinates of multiple sampling times for each continuous gaze event, the average gaze coordinates and gaze duration of each continuous gaze event are calculated, and the average gaze duration, gaze dispersion, and long gaze ratio of all continuous gaze events are calculated accordingly to generate gaze feature data. Based on the gaze point coordinates of multiple sampling times for each consecutive saccade event, the saccade amplitude and saccade speed of each consecutive saccade event are calculated respectively. Based on this, the number of retraces, average saccade amplitude, standard deviation of saccade speed, retrace path deviation rate, ratio of total horizontal to vertical task time, and skip rate of all consecutive saccade events are calculated to generate saccade feature data. Based on the sampling time of each eye-closing event in the corrected eye-closing event sequence data, one or more consecutive eye-closing events are identified, and the eye-closing time of each consecutive eye-closing event and the average eye-closing time of all consecutive eye-closing events are calculated to generate eye-closing feature data. Based on the pupil diameters of each pupil in the corrected pupil diameter sequence data, the pupil diameter variation range is calculated, and pupil diameter feature data is generated. The eye movement feature data is generated based on the gaze feature data, the saccade feature data, the eye closure feature data, and the pupil diameter feature data.

4. The reading disorder screening system based on eye-tracking technology according to claim 1, characterized in that, The method for constructing the eye-tracking test parameter norm table includes: Based on a pre-built subject eye-tracking test database, the ages and historical eye-tracking test parameters of multiple healthy subjects were obtained, and the healthy subjects were grouped by age to obtain historical eye-tracking test parameter sets for multiple age groups. Statistical analysis was performed on the historical eye movement test parameter sets for each age group to obtain the eye movement test parameter norm data for each age group, and the eye movement test parameter norm table was constructed accordingly.

5. The reading disorder screening system based on eye-tracking technology according to claim 1, characterized in that, The computer device updates the target subject's basic information, eye movement test parameters, raw eye movement data, and eye movement test report to the pre-built subject eye movement test database.

6. The reading disorder screening system based on eye-tracking technology according to claim 1, characterized in that, Before the eye-tracking device performs an eye-tracking test on the target subject, it also includes: Through the display unit, calibration points are sequentially presented at multiple designated locations in the stimulus presentation area, and the target subject is instructed to gaze at each calibration point in sequence, thereby collecting multiple eye movement calibration image data of the target subject's eyes. Each eye-tracking calibration image data and the screen coordinates of each calibration point are sent to the computer device, so that the computer device can calculate the pupil-corneal reflection vector of each calibration point based on each eye-tracking calibration image data, and construct a coordinate transformation model based on the screen coordinates of each calibration point.

7. A method for screening reading disorders based on eye-tracking technology, characterized in that, Applied to a computer device that communicates with an eye-tracking device, including: The basic information of the target subject is obtained, and based on the pre-constructed eye movement test parameter norm table, the eye movement test parameters of the target subject are obtained and sent to the eye movement tracking device for initialization. The eye movement test is then performed on the target subject to generate multi-frame eye movement image data of the target subject. Based on the eye movement image data of each frame obtained from the eye movement tracking device, the raw eye movement data of the target subject is collected, and the raw eye movement data is preprocessed and feature extracted to generate the eye movement feature data of the target subject. Based on a pre-trained eye movement function abnormality recognition model, the eye movement function of the target subject is evaluated according to the eye movement feature data, and the risk of reading impairment of the target subject is identified, and an eye movement test report of the target subject is generated. The eye-tracking test parameters include: stimulus presentation area size, stimulus table pattern, stimulus speed, stimulus table font size, and eye-tracking sampling rate. The eye-tracking device includes: a display unit, including a stimulus presentation area, for dynamically displaying multiple characters in a preset stimulus table at a preset stimulus speed; wherein the stimulus table includes a vertical number table, a horizontal number table, and reading material; an infrared emission unit, for emitting invisible infrared light of a preset wavelength to illuminate the eyes of the target subject, thereby generating corneal reflection points on the corneas of the target subject; a dual-camera unit, including two cameras, for simultaneously acquiring eye movement images of the target subject's eyes while the target subject reads the stimulus table based on a preset eye movement sampling frequency, and generating multi-frame eye movement image data of the target subject; and a voice guidance unit, for providing voice guidance to the target subject to complete the eye-tracking test; the size of the stimulus presentation area of ​​the display unit, the stimulus table mode, the stimulus speed and font size of each character in the stimulus table, and the preset eye movement sampling frequency of the dual-camera unit are respectively set according to the eye-tracking test parameters during the initialization of the eye-tracking device; Based on the eye movement function abnormality identification model, the method for assessing the eye movement function of the target subject and identifying the risk of reading impairment based on the eye movement feature data includes: obtaining the age and historical eye movement feature data of multiple healthy subjects based on a pre-constructed subject eye movement test database, and grouping each healthy subject by age to obtain historical eye movement feature datasets for multiple age groups; performing statistical analysis on the historical eye movement feature datasets for each age group to obtain eye movement feature norm datasets for each age group, and constructing an eye movement function norm model accordingly; based on the eye movement function norm model, calculating standard scores for multiple eye movement features according to the eye movement feature data to construct a standard score vector for the eye movement features of the target subject, and filtering one or more abnormal eye movement features to label the eye movement function level of the target subject; based on a gradient boosting tree, identifying abnormal eye movement function patterns of the target subject according to the eye movement feature data and the standard score vector for the eye movement features, analyzing the risk of reading impairment of the target subject, and generating an eye movement test report for the target subject.

8. A reading disorder screening terminal based on eye-tracking technology, characterized in that, include: Processor and memory; The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory, so that the terminal performs the reading disorder screening method based on eye-tracking technology as described in claim 7.

Images