A cognitive risk assessment system, method, device and storage medium based on eye movement characteristics

Abstract

The present invention discloses a cognitive risk assessment system, method, device and storage medium based on eye movement features, which relates to the field of cognitive risk assessment, including a learning phase of a visual paired comparison task, a reverse saccade task and a test phase of a visual paired comparison task presented in sequence on a mobile terminal. The above-mentioned task is a task paradigm specially designed to evaluate cognitive functions related to cognitive impairment, and the reverse saccade task is interspersed with the visual paired comparison task. The time to complete the reverse saccade task can be used as the memory time of the visual paired comparison task, thereby optimizing the task paradigm. Quantified eye movement features are determined based on the facial video of the user when completing the cognitive risk assessment task, and the quantitative eye movement features are analyzed using a machine learning model to determine the cognitive risk assessment results. The cognitive risk assessment results determined based on eye movement features are more objective, and cognitive risk assessment can be achieved using a mobile terminal, which is more efficient, convenient, low-cost and applicable to a large population.

Description

Technical Field

[0001] The present invention relates to the field of cognitive risk assessment, and in particular to a cognitive risk assessment system, method, device and storage medium based on eye movement characteristics. Background Art

[0002] With the increasing aging of the global population and the accelerating pace of life, cognitive health issues are receiving increasing attention. Decreased cognitive function (such as memory, attention, and executive function) not only impacts individual quality of life but may also signal the onset of neurodegenerative diseases (such as mild cognitive impairment and Alzheimer's disease). Early detection of cognitive risks and timely intervention are crucial for slowing disease progression and improving prognosis.

[0003] Existing cognitive assessment methods have numerous limitations: cognitive scales are susceptible to subjective influences from the test-taker; biomarker testing is invasive, complex, requires high experimental conditions, and is expensive; and cognitive assessment methods based on brain imaging or infrared eye tracking are expensive and have limited application scenarios. Therefore, the market urgently needs an objective, efficient, convenient, low-cost cognitive risk assessment tool that can be applied to a large population. Summary of the Invention

[0004] The purpose of the present invention is to provide a cognitive risk assessment system, method, device and storage medium based on eye movement characteristics, which can improve the objectivity of cognitive risk assessment results, be more efficient, convenient, low-cost and suitable for a large population.

[0005] To solve the above technical problems, the present invention provides a cognitive risk assessment system based on eye movement characteristics, comprising:

[0006] a cognitive paradigm presentation module, configured to sequentially present on a display screen of a mobile terminal a learning phase of a visual paired comparison task, a reverse saccade task, and a test phase of the visual paired comparison task in a cognitive risk assessment task;

[0007] The learning phase of the visual paired comparison task includes sequentially presenting a first number of pairs of learning pictures, and the display duration of each pair of learning pictures is the first duration; the testing phase of the visual paired comparison task includes sequentially presenting a second number of pairs of test pictures, each pair of test pictures includes one of the learning pictures and a new picture that has not been presented in the learning phase, and the display duration of each pair of test pictures is the second duration; the reverse saccade task includes a third number of groups of stimulation tasks, when each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen, and after the fixed fixation point is displayed for a preset delay time, a target fixation point at a different position from the fixed fixation point is displayed on the display screen, and the display duration of the target fixation point is the third duration, and the preset delay time corresponding to each group of stimulation tasks changes according to a preset rule;

[0008] an eye movement data processing module, configured to obtain a facial video captured by a front-facing camera of the mobile terminal while the user completes the cognitive risk assessment task in accordance with the task rules, determine raw eye movement data based on the facial video, and determine quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task based on the raw eye movement data and cognitive risk assessment task information;

[0009] A cognitive risk assessment module is used to input the quantified eye movement features into a pre-trained machine learning model, and use the machine learning model to determine a cognitive risk assessment result based on the quantified eye movement features.

[0010] Optionally, the quantitative eye movement features corresponding to the visual paired comparison task determined based on the raw eye movement data and the cognitive risk assessment task information include: dwell time on the new picture, dwell time on the old picture, number of gaze blinks, duration of the gaze area, dwell ratio on the new picture, and preference ratio between the new and old pictures;

[0011] The quantitative eye movement features corresponding to the reverse saccade task determined based on the original eye movement data and the cognitive risk assessment task information include: maximum saccade velocity, the number of saccade points before stimulation, the number of saccade points after stimulation, gaze stability, the number of gaze blinks, the initial saccade gain, the saccade gain ratio, the saccade accuracy rate, the saccade error rate, the correct saccade delay time, the corrected correct saccade delay time, the correct number of error corrections, and the correction delay.

[0012] Optionally, determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes:

[0013] The saccade gain ratio, the correction delay, the maximum saccade speed, the proportion of stay on the new image, the gaze area duration, and the saccade accuracy are input into the machine learning model, so that the machine learning model determines a cognitive assessment risk value based on an eye movement logistic regression formula, and the cognitive assessment risk value is positively correlated with the probability that the user has Alzheimer's disease; the eye movement logistic regression formula is:

[0014] ;

[0015] in, Assessing a risk value for the cognition, is a constant term, 、 to The independent variables 、 to The coefficient of is the saccadic gain ratio, For the correction delay, is the maximum eye speed, is the retention ratio of the new image, is the duration of the gaze area, is the eye saccade accuracy rate.

[0016] Optionally, determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes:

[0017] Inputting the percentage of new images the user stays on during the visual paired comparison task into the machine learning model, so that the machine learning model determines the user's initial memory ability value based on the memory ability initial value calculation formula, determines a memory ability score based on the memory ability initial value and the memory score calculation formula, and uses the memory ability score as the cognitive risk assessment result;

[0018] The calculation formula for the initial value of the memory capacity is:

[0019] ,in, is the initial value of the memory capacity, is a constant term, is the first coefficient, is the second coefficient, is the retention ratio of the new image, The user's age;

[0020] The memory score calculation formula is:

[0021] ,in, is the memory ability score, is a constant term, is the third coefficient.

[0022] Optionally, determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes:

[0023] Inputting the correct saccade delay time and the maximum saccade speed into the machine learning model, so that the machine learning model determines the user's initial execution ability value based on an execution ability initial value calculation formula, calculates an execution ability score based on the initial execution ability value and an execution ability score calculation formula, and uses the execution ability score as the cognitive risk assessment result;

[0024] The calculation formula for the initial value of the execution capability is:

[0025] ,in, is the initial value of the execution capability, is a constant term, 、 as well as The fourth coefficient, the fifth coefficient and the sixth coefficient are respectively, is the correct saccade delay time, is the maximum saccade velocity, The user's age;

[0026] The execution ability score calculation formula is:

[0027] ,in, is the execution ability score, is a constant term, The seventh coefficient.

[0028] Optionally, determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes:

[0029] The gaze area duration is input into the machine learning model so that the machine learning model determines the user's attention ability score based on the gaze area duration and an attention ability score calculation formula, and uses the attention ability score as the cognitive risk assessment result; the attention ability score calculation formula is:

[0030] ,in, is the attention ability score, is the duration of the gaze area, is a constant term, The eighth coefficient.

[0031] Optionally, the cognitive paradigm presentation module is further configured to:

[0032] Select black and white sketch images from the preset image library, and the images that meet the preset task image requirements as learning images and test images in the visual paired comparison task;

[0033] The method includes determining the screen features of the current mobile terminal, determining the size and display position of each learning image and test image in the visual paired comparison task based on the screen features, and determining the display positions of the fixed fixation point and the target fixation point in the countersaccade task. After the determination, the method then proceeds to the step of sequentially presenting the learning phase of the visual paired comparison task, the countersaccade task, and the test phase of the visual paired comparison task in the cognitive risk assessment task on the display screen of the mobile terminal.

[0034] To solve the above technical problems, the present invention also provides a cognitive risk assessment method based on eye movement characteristics, comprising:

[0035] sequentially presenting a learning phase of a visual paired comparison task, a reverse saccade task, and a test phase of the visual paired comparison task in a cognitive risk assessment task on a display screen of a mobile terminal;

[0036] The learning phase of the visual paired comparison task includes sequentially presenting a first number of pairs of learning pictures, and the display duration of each pair of learning pictures is the first duration; the testing phase of the visual paired comparison task includes sequentially presenting a second number of pairs of test pictures, each pair of test pictures includes one of the learning pictures and a new picture that has not been presented in the learning phase, and the display duration of each pair of test pictures is the second duration; the reverse saccade task includes a third number of groups of stimulation tasks, when each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen, and after the fixed fixation point is displayed for a preset delay time, a target fixation point at a different position from the fixed fixation point is displayed on the display screen, and the display duration of the target fixation point is the third duration, and the preset delay time corresponding to each group of stimulation tasks changes according to a preset rule;

[0037] Obtaining a facial video captured by a front-facing camera of the mobile terminal when the user completes the cognitive risk assessment task in accordance with the task rules, determining raw eye movement data based on the facial video, and determining quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task based on the raw eye movement data and cognitive risk assessment task information;

[0038] The quantified eye movement features are input into a pre-trained machine learning model, and the machine learning model is used to determine a cognitive risk assessment result based on the quantified eye movement features.

[0039] To solve the above technical problems, the present invention further provides a cognitive risk assessment device based on eye movement characteristics, comprising:

[0040] memory for storing computer programs;

[0041] A processor is configured to implement the steps of the above-mentioned cognitive risk assessment method based on eye movement characteristics when executing the computer program.

[0042] In order to solve the above technical problems, the present invention also provides a storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the above-mentioned cognitive risk assessment method based on eye movement characteristics are implemented.

[0043] The beneficial effect of the present invention is to provide a cognitive risk assessment system, method, device and storage medium based on eye movement features, including a learning phase of a visual paired comparison task, a reverse saccade task and a test phase of a visual paired comparison task, which are presented in sequence on a mobile terminal. The above-mentioned task is a task paradigm specially designed to evaluate cognitive functions related to cognitive impairment, and the reverse saccade task is interspersed with the visual paired comparison task. The time to complete the reverse saccade task can be used as the memory time of the visual paired comparison task, thereby optimizing the task paradigm. Quantified eye movement features are determined based on the facial video of the user when completing the cognitive risk assessment task, and the quantitative eye movement features are analyzed using a machine learning model to determine the cognitive risk assessment results. The cognitive risk assessment results determined based on eye movement features are more objective, and cognitive risk assessment can be achieved using a mobile terminal, which is more efficient, convenient, low-cost and applicable to a large population. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the prior art and the drawings required for use in the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0045] Figure 1 A schematic diagram of a cognitive risk assessment system based on eye movement characteristics provided by the present invention;

[0046] Figure 2 A schematic diagram of a cognitive risk assessment task provided by the present invention;

[0047] Figure 3 A flowchart of a cognitive risk assessment method based on eye movement characteristics provided by the present invention;

[0048] Figure 4 This is a structural diagram of a cognitive risk assessment device based on eye movement characteristics provided by the present invention. DETAILED DESCRIPTION

[0049] The core of the present invention is to provide a cognitive risk assessment system, method, device and storage medium based on eye movement characteristics, which can improve the objectivity of cognitive risk assessment results, be more efficient, convenient, low-cost and suitable for a large population.

[0050] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0051] The core of this invention lies in the design of a cognitive risk assessment task paradigm suitable for evaluating specific cognitive functions (including memory ability, attention and executive function) of people with cognitive impairment, combined with a machine learning model to conduct quantitative analysis of facial videos captured by mobile terminals to achieve cognitive risk assessment.

[0052] Please refer to Figure 1 , Figure 1 This is a schematic diagram of a cognitive risk assessment system based on eye movement characteristics provided by the present invention, which includes:

[0053] The cognitive paradigm presentation module 101 is configured to sequentially present the learning phase, the reverse saccade task, and the test phase of the visual paired comparison task in the cognitive risk assessment task on the display screen of the mobile terminal.

[0054] Among them, the learning phase of the visual paired comparison task includes presenting a first number of pairs of learning pictures in sequence, and the display duration of each pair of learning pictures is the first duration; the testing phase of the visual paired comparison task includes presenting a second number of pairs of test pictures in sequence, each pair of test pictures includes a learning picture and a new picture that has not been presented in the learning phase, and the display duration of each pair of test pictures is the second duration; the reverse saccade task includes a third number of groups of stimulation tasks, when each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen, and after the fixed fixation point is displayed for a preset delay time, a target fixation point at a different position from the fixed fixation point is displayed on the display screen, and the display duration of the target fixation point is the third duration. The preset delay time corresponding to each group of stimulation tasks changes according to preset rules.

[0055] The eye movement data processing module 102 is used to obtain a facial video captured by the front camera of the mobile terminal when the user completes the cognitive risk assessment task in accordance with the task rules, determine raw eye movement data based on the facial video, and determine quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task based on the raw eye movement data and the cognitive risk assessment task information.

[0056] The cognitive risk assessment module 103 is used to input the quantified eye movement features into a pre-trained machine learning model, and use the machine learning model to determine the cognitive risk assessment results based on the quantified eye movement features.

[0057] The mobile terminal can be a device with a front-facing camera and display, such as a smartphone or tablet. The present invention pre-deploys a cognitive paradigm presentation module 101 on the mobile terminal. This module is responsible for presenting specific cognitive risk assessment tasks on the mobile terminal's display and guiding the user through the tasks. The present invention carefully designs and selects cognitive risk assessment tasks that are sensitive to memory, attention, and executive function, ensuring that these tasks not only effectively stimulate key eye movement behavioral differences at the cognitive level but are also technically more suitable for mobile devices.

[0058] The cognitive risk assessment task designed by the present invention is introduced in detail below.

[0059] The cognitive risk assessment task includes a visual paired comparison task and a countersaccade task. The visual paired comparison task consists of a learning phase, a delay phase, and a test phase. The visual paired comparison task assesses an individual's ability to distinguish between new and old visual information and is sensitive to memory impairments related to hippocampal function (such as mild cognitive impairment and early Alzheimer's disease). The countersaccade task assesses an individual's ability to inhibit a dominant response (toward a stimulus) and execute a nondominant response (away from the stimulus), reflecting prefrontal cortex-related executive control functions.

[0060] The cognitive paradigm presentation module 101 first displays the learning phase of the visual paired comparison task on the mobile terminal's display screen, guiding the user to observe and memorize the learning images presented during the learning phase. The number of learning images and the display duration of each learning image pair can be adjusted based on actual needs. For example, five learning image pairs can be displayed on the display screen for five seconds each, though this is not specifically limited in the present invention.

[0061] Subsequently, the cognitive paradigm presentation module 101 displays the reverse saccade task on the display screen. On the one hand, the user's executive function and inhibitory control ability can be evaluated based on the user's eye movement performance in the reverse saccade task. On the other hand, the reverse saccade task can also be used as a delay period for the visual paired comparison task, making the overall process of the cognitive risk assessment task relatively fast (for example, completed within 10-15 minutes).

[0062] The reverse saccade task includes a third number of groups of stimulation tasks. When each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen. After the fixed fixation point is displayed for a preset delay time, a target fixation point with a position different from the fixed fixation point is displayed on the display screen. The number of stimulation tasks and the display time of the target fixation point in each group of stimulation tasks can be set according to actual needs. However, it should be noted that in each group of stimulation tasks, after the fixed fixation point is displayed, the target fixation point is displayed after a preset delay time. The preset delay time here can be a randomly varying time length to better evaluate the user's executive function and inhibitory control ability.

[0063] Finally, the cognitive paradigm presentation module 101 displays the test phase of the visual paired comparison task on the display screen. The test phase requires sequentially presenting a second number of pairs of test images, each pair of test images including a learning image that has been presented in the learning phase and a new image that has not been presented in the learning phase. The cognitive paradigm presentation module 101 prompts the user to look at the new image after each pair of test images appears. The number of pairs of test images included in the test phase is the second number, which can be twice the number of learning images, and the present invention does not specifically limit this. The display duration of each pair of test images can also be set according to actual needs.

[0064] It should be noted that before presenting the cognitive risk assessment task, the cognitive paradigm presentation module 101 can first display guidance information through the mobile device interface or play voice guidance information. For example, before presenting the learning phase of the visual paired comparison task, the user is instructed to memorize the learning images presented during the learning phase; before presenting the test phase of the visual paired comparison task, the user is instructed to fixate on a new image not presented during the learning phase. Before presenting the reverse saccade task, the user is instructed to shift their gaze to the opposite side of the display screen after the target fixation point appears on the display screen.

[0065] While the cognitive risk assessment task is presented on the mobile device's display, the mobile device's front-facing camera captures real-time facial video (particularly the eye area) of the user completing the cognitive risk assessment task. The facial video records the user's natural eye movements as they respond to the specific task.

[0066] The eye movement data processing module 102 obtains the facial video captured by the camera and uses an artificial intelligence algorithm (such as a deep learning model) to estimate raw eye movement data in real time based on the facial video and current task information. For example, the artificial intelligence algorithm analyzes the facial video in real time, detects key eye points (such as the pupil center) from the facial video, and calculates the coordinate sequence of the user's gaze point on the screen, gaze events (gaze start time, gaze end time, and gaze position), saccade events (saccade start time, saccade end time, saccade starting point, and saccade end point), and pupil diameter changes, and uses these parameters as raw eye movement data.

[0067] Next, based on the raw eye movement data and cognitive risk assessment task information, we determined the quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task. For example, during the visual paired comparison task, we focused on extracting features related to recognition memory for the new and old image stimuli, such as fixation distribution and duration; during the countersaccade task, we focused on extracting features related to inhibitory control, such as saccade latency, error rate, and correction time. The quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task will be described later and are not detailed here.

[0068] Finally, the quantified eye movement features are input into a pre-trained machine learning model through the cognitive risk assessment module 103. This model is pre-trained using a supervised learning method using a large amount of sample data with known cognitive state labels. These cognitive state labels include, but are not limited to, healthy, mild cognitive impairment, etc. The machine learning model analyzes and calculates the input quantified eye movement features to determine the user's cognitive risk assessment result. The cognitive risk assessment result can be a cognitive risk classification result (such as normal, low risk, medium risk, and high risk), a cognitive risk assessment score, or a performance assessment of a specific cognitive domain (such as memory ability and executive function), which is not particularly limited in the present invention.

[0069] It should also be noted that the eye movement data processing module 102 and the cognitive risk assessment module 103 can be deployed on a mobile device or as Figure 1 The system is shown as being deployed on a cloud platform, to which the present invention makes no special limitation.

[0070] Furthermore, the aforementioned machine learning model can be a variety of algorithms, such as Logistic Regression (LR), Extra Trees (ET), and Random Forest (RF). Experiments conducted by the present invention have determined that the Logistic Regression algorithm demonstrates superior performance in cognitive risk assessment, and therefore, the Logistic Regression algorithm can be preferred for cognitive risk assessment.

[0071] Furthermore, the report generation and display module allows users to display cognitive risk assessment results on the mobile device's screen in the form of text reports, radar charts, and score bars. Raw data summaries, features, and results from this cognitive assessment can also be saved locally for historical comparison and longitudinal tracking.

[0072] In summary, this invention specifically designs a visual paired comparison task and a reverse saccade task based on specific cognitive risks (mild cognitive impairment and Alzheimer's disease). This establishes a quantitative relationship between eye movement characteristics under specific cognitive tasks and cognitive risk. By leveraging the sensitivity of eye movement characteristics to changes in cognitive function, this method can identify subtle cognitive abnormalities earlier than traditional methods, providing clues for early intervention. Furthermore, by capturing and analyzing the user's eye movement characteristics during the cognitive risk assessment task, which serve as an objective physiological signal, this invention can reduce the subjective bias of traditional scale assessments.

[0073] The present invention also utilizes common mobile devices commonly owned by users (such as smartphones and tablets) and their existing front-facing cameras to capture facial videos of users completing cognitive risk assessment tasks. The facial videos are then analyzed using artificial intelligence algorithms to implement cognitive risk assessments of users. By utilizing widely available smartphones or tablets as the hardware platform, there is no need to purchase expensive, dedicated eye trackers. Users can perform assessments at any time, anywhere with a suitable mobile device (home, community, clinic). The assessment process typically takes only a few minutes to over ten minutes, greatly facilitating user convenience. The system is particularly suitable for large-scale population screening, applications in remote areas, and scenarios requiring frequent monitoring. Machine learning algorithms are used to overcome the challenges posed by these tasks (such as lighting changes, head posture, and device diversity). This represents a deeply customized design for the mobile ecosystem.

[0074] Based on the above embodiment:

[0075] As an optional embodiment, the quantitative eye movement features corresponding to the visual paired comparison task determined based on the original eye movement data and cognitive risk assessment task information include: dwell time on new pictures, dwell time on old pictures, number of gaze blinks, duration of gaze area, dwell ratio on new pictures, and preference ratio between new and old pictures.

[0076] The quantitative eye movement features corresponding to the countersaccade task, determined based on the original eye movement data and cognitive risk assessment task information, include: maximum saccade velocity, number of saccade points before stimulation, number of saccade points after stimulation, gaze stability, number of gaze blinks, initial saccade gain, saccade gain ratio, saccade accuracy, saccade error rate, correct saccade latency, corrected correct saccade latency, number of corrected errors, and correction latency.

[0077] In this embodiment, after selecting the most effective machine learning model, eye movement features are further selected. For example, the cumulative AUC (Area Under Curve) is calculated by the stepwise regression method to select some eye movement features to further optimize the cognitive risk assessment effect.

[0078] After screening, this embodiment extracts quantitative eye movement features corresponding to the visual paired comparison task based on the raw eye movement data and cognitive risk assessment task information, including: dwell time on the new image, dwell time on the old image, number of fixations and blinks, gaze area duration, dwell ratio on the new image, and preference ratio between the new and old images. The dwell time on the new image (milliseconds) indicates the time the center of the gaze tuple remains on the new image. The dwell time on the old image (milliseconds) indicates the time the center of the gaze tuple remains on the learning image. The number of fixations and blinks indicates the number of fixations that may contain blinks. The gaze area duration (milliseconds) indicates the sum of the durations of all gaze tuples. The dwell ratio on the new image indicates the proportion of the dwell time on the new image to the gaze area duration. The preference ratio between the new and old images indicates the ratio of the dwell time on the new image to the dwell time on the old image.

[0079] Based on the original eye movement data and cognitive risk assessment task information, the quantitative eye movement features corresponding to the countersaccade task were extracted, including: maximum saccade velocity, number of saccade points before stimulation, number of saccade points after stimulation, gaze stability, number of gaze blinks, initial saccade gain, saccade gain ratio, saccade accuracy, saccade error rate, correct saccade delay time, corrected correct saccade delay time, number of corrected errors, and correction delay.

[0080] The maximum saccade velocity (horizontally) (pixels / ms) represents the maximum saccade velocity in the horizontal direction during the countersaccade task.

[0081] The number of pre-stimulus saccades refers to the number of saccades that occurred before the target fixation point appeared.

[0082] The number of post-stimulus saccades refers to the number of saccades that occurred after the target fixation point appeared.

[0083] Gaze stability refers to the number of gaze tuples before the target point appears.

[0084] The number of gaze-blinks indicates the number of gaze tuples that may contain blinks.

[0085] The initial saccade gain represents the horizontal change in distance of the first saccade after the target fixation point appears. A positive value indicates a saccade toward the target, while a negative value indicates a saccade away from the target. In a countersaccade task, a negative initial saccade gain indicates a correct countersaccade.

[0086] The saccade gain ratio represents the amplitude of the first saccade after the target fixation point appears. It is calculated as the ratio of the change in horizontal distance between the first saccade and the first fixation point in the first fixation tuple (delta) to the horizontal distance between the last fixation point in the last fixation tuple before the target fixation point appears and the task target point (ref), i.e., delta / ref.

[0087] The saccade accuracy rate represents the proportion of the number of stimulation tasks in which the user performed the correct reverse saccade task among all stimulation tasks.

[0088] The saccade error rate represents the percentage of tasks in which the first saccade direction is wrong after the target fixation point appears, and the result is still wrong after correction.

[0089] Corrected correct saccade latency represents the latency to the first (correct) saccade in tasks where the first saccade was in the correct direction, or in tasks where the first saccade was in the wrong direction but subsequently corrected. If the first saccade was correct, this value is equal to the saccade latency (first_saccade); if correction is required, this value is the delay in the timing of the final correct saccade relative to the stimulus onset time. Corrected correct saccade latency is the average of these two cases.

[0090] Correct saccade latency refers to the latency of the first correct saccade in a task where the first saccade is in the correct direction. It refers to the latency of the first saccade, but only has a value when the first saccade is in the correct direction.

[0091] The number of corrected errors refers to the number of tasks in which the first saccade direction was wrong but was subsequently corrected and successfully corrected.

[0092] Correction delay refers to the time from the first incorrect saccade to the completion of the first correct saccade. It refers to the time delay of the correction saccade relative to the first saccade, rather than the time relative to the stimulus appearance.

[0093] In summary, selecting the above-mentioned quantitative eye movement features for subsequent cognitive risk assessment through experiments and analysis can further improve the accuracy and reliability of cognitive risk assessment.

[0094] The following describes an implementation method for using a machine learning model to determine cognitive risk assessment results based on quantified eye movement features.

[0095] As an optional embodiment, a machine learning model is used to determine a cognitive risk assessment result based on quantified eye movement features, including:

[0096] The saccade gain ratio, correction delay, maximum saccade speed, proportion of dwelling on new images, duration of gaze area, and saccade accuracy are input into the machine learning model so that the machine learning model can determine the cognitive assessment risk value based on the eye movement logistic regression formula, and the cognitive assessment risk value is positively correlated with the probability that the user has Alzheimer's disease; the eye movement logistic regression formula is:

[0097] ;

[0098] in, To assess risk value, is a constant term, 、 to The independent variables 、 to The coefficient of is the saccadic gain ratio, To correct for the delay, is the maximum saccade velocity, is the percentage of new pictures remaining. is the gaze area duration, is the saccade accuracy.

[0099] In this embodiment, the eye movement logistic regression formula is used to calculate the cognitive assessment risk value, which is directly related to the degree of verification of Alzheimer's disease. The eye movement logistic regression formula is obtained through machine learning methods and trained based on data from normal and diseased groups (of varying severity). The higher the cognitive assessment risk value, the higher the probability that the user has Alzheimer's disease. In some embodiments, the above eye movement logistic regression formula can be expressed as:

[0100] .

[0101] In addition to calculating the cognitive assessment risk value using the eye movement logistic regression formula, the present invention also calculates the memory ability score, the executive ability score and the attention ability score as the cognitive risk assessment results.

[0102] Memory, executive function, and attention scores are all calculated using machine learning models based on quantified eye movement features. The eye movement features used to assess memory, executive function, and attention were selected based on a combination of assessment paradigms (visual paired comparison task and countersaccade task) and previous research experience. The coefficients for each eye movement feature and their relationship with age were quantitatively calculated using collected clinical data from subjects with varying disease severity and age groups.

[0103] As an optional embodiment, a machine learning model is used to determine a cognitive risk assessment result based on quantitative eye movement features, including:

[0104] The proportion of new image stays during the user's completion of the visual paired comparison task is input into the machine learning model, so that the machine learning model determines the user's initial memory ability value based on the initial memory ability value calculation formula, determines the memory ability score based on the initial memory ability value and the memory score calculation formula, and uses the memory ability score as the cognitive risk assessment result.

[0105] The calculation formula for the initial value of memory capacity is:

[0106] ,in, is the initial value of memory capacity, is a constant term, is the first coefficient, is the second coefficient, is the percentage of new pictures remaining. The user's age.

[0107] The formula for calculating the memory score is:

[0108] ,in, is the memory ability score, is a constant term, is the third coefficient.

[0109] In some embodiments, the calculation formula for the initial value of the memory capacity can be expressed as: ; The calculation formula of memory score can be expressed as: .

[0110] As an optional embodiment, a machine learning model is used to determine a cognitive risk assessment result based on quantitative eye movement features, including:

[0111] The correct saccade delay time and the maximum saccade speed are input into the machine learning model so that the machine learning model can determine the user's initial execution ability value based on the execution ability initial value calculation formula, calculate the execution ability score based on the execution ability initial value and the execution ability score calculation formula, and use the execution ability score as the cognitive risk assessment result.

[0112] The calculation formula for the initial value of execution capability is:

[0113] ,in, is the initial value of execution capability, is a constant term, 、 as well as The fourth coefficient, the fifth coefficient and the sixth coefficient are respectively, To determine the correct saccade delay time, is the maximum saccade velocity, The user's age.

[0114] The formula for calculating the execution ability score is:

[0115] ,in, is the execution ability score, is a constant term, The seventh coefficient.

[0116] In some embodiments, the above calculation formula for the initial value of execution capability can be expressed as: ; The above execution capability score calculation formula can be expressed as: .

[0117] As an optional embodiment, a machine learning model is used to determine a cognitive risk assessment result based on quantified eye movement features, including:

[0118] The gaze area duration is input into the machine learning model so that the machine learning model can determine the user's attention ability score based on the gaze area duration and the attention ability score calculation formula, and use the attention ability score as the cognitive risk assessment result; the attention ability score calculation formula is:

[0119] ,in, is the attention ability score, is the gaze area duration, is a constant term, The eighth coefficient.

[0120] In some embodiments, the above attention ability score calculation formula can be expressed as .

[0121] As an optional embodiment, the cognitive paradigm presentation module 101 is further configured to:

[0122] Sketch images in black and white are selected from the preset image library, and images whose content meets the preset task image requirements are used as learning images and test images in the visual paired comparison task.

[0123] The method includes determining the screen features of the current mobile terminal, determining the size and display position of each learning image and test image in the visual paired comparison task based on the screen features, and determining the display positions of the fixed fixation point and the target fixation point in the countersaccade task. After the determination, the method then proceeds to the step of sequentially presenting the learning phase of the visual paired comparison task, the countersaccade task, and the test phase of the visual paired comparison task in the cognitive risk assessment task on the display screen of the mobile terminal.

[0124] In this embodiment, to prevent users from being affected by image color and novelty biases, the present invention selects black and white sketches from a preset image library, whose content meets the preset task requirements, as learning and test images for the visual paired comparison task. The preset task requirements can include images of everyday objects such as transportation, fruit, clothing, household appliances, and birds, and can be set based on actual needs.

[0125] Furthermore, to ensure standardization and effectiveness of cognitive risk assessments across different mobile devices, this embodiment first determines the current mobile terminal's screen characteristics. It then optimizes and adapts the visual stimuli (graphics, color, and size), timing parameters, and interaction logic for all tasks to the current mobile device's screen characteristics and computing power. This includes adjusting the size and position of the learning and test images to the current mobile device's screen size (this can be achieved by adjusting the coordinates of designated points on the learning and test images on the screen), and adjusting the display positions of the fixed and target fixation points (this can be achieved by adjusting the radius and on-screen coordinates of the fixed and target fixation points).

[0126] Please refer to Figure 2 , Figure 2 This is a schematic diagram of a cognitive risk assessment task provided by the present invention. Figure 2 A specific implementation method of the cognitive risk assessment task is described.

[0127] The learning phase of the visual paired comparison task included five pairs of learning pictures, each presented for five seconds. All the learning pictures were black and white sketch images.

[0128] After all five pairs of learning pictures were presented, the reverse saccade task began. The reverse saccade task included 12 sets of stimulation tasks. At the beginning of each stimulation task, a fixed fixation point (a dot of a specific color, color information such as RGB (252, 248, 230)) was displayed at the center of the screen. Then, after a randomly varying time delay (such as 1-2 seconds), a target fixation point (such as a red dot, color information is RGB (250, 0, 0)) was presented at a random position on the left, right, or periphery of the screen. Figure 2 The target fixation point is indicated by a triangle in the figure), and the user is clearly instructed to move their gaze to the other side of the display screen where the target fixation point does not appear after seeing the target fixation point appear.

[0129] After all stimulus sets in the countersaccade task were presented, the test phase of the visual paired comparison task began. The test phase consisted of 10 pairs of test images. Each pair consisted of a learning image presented during the training phase and a new image, again a black-and-white sketch with the required content. Each pair of test images was presented for 6 seconds, with the user explicitly instructed to fixate on the new image.

[0130] Please refer to Figure 3 , Figure 3 A flowchart of a cognitive risk assessment method based on eye movement characteristics provided by the present invention, the method comprising:

[0131] S301, sequentially presenting a learning phase of a visual paired comparison task, a reverse saccade task, and a test phase of a visual paired comparison task in a cognitive risk assessment task on a display screen of a mobile terminal;

[0132] Among them, the learning phase of the visual paired comparison task includes presenting a first number of pairs of learning pictures in sequence, and the display duration of each pair of learning pictures is the first duration; the testing phase of the visual paired comparison task includes presenting a second number of pairs of test pictures in sequence, each pair of test pictures includes a learning picture and a new picture that has not been presented in the learning phase, and the display duration of each pair of test pictures is the second duration; the reverse saccade task includes a third number of groups of stimulation tasks, when each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen, and after the fixed fixation point is displayed for a preset delay time, a target fixation point at a different position from the fixed fixation point is displayed on the display screen, and the display duration of the target fixation point is the third duration. The preset delay time corresponding to each group of stimulation tasks changes according to preset rules.

[0133] S302. Obtain a facial video captured by the front camera of the mobile terminal when the user completes the cognitive risk assessment task in accordance with the task rules, determine raw eye movement data based on the facial video, and determine quantitative eye movement features corresponding to the visual paired comparison task and the reverse saccade task based on the raw eye movement data and the cognitive risk assessment task information.

[0134] S303: Input the quantified eye movement features into a pre-trained machine learning model, and use the machine learning model to determine the cognitive risk assessment results based on the quantified eye movement features.

[0135] For a detailed introduction to the cognitive risk assessment method based on eye movement features provided by the present invention, please refer to the above-mentioned embodiment of the cognitive risk assessment system based on eye movement features, and the present invention will not be elaborated here.

[0136] Please refer to Figure 4 , Figure 4This is a schematic diagram of the structure of a cognitive risk assessment device based on eye movement characteristics provided by the present invention, which includes:

[0137] Memory 401, used for storing computer programs;

[0138] The processor 402 is configured to implement the steps of the above-mentioned cognitive risk assessment method based on eye movement characteristics when executing the computer program.

[0139] For a detailed introduction to the cognitive risk assessment device based on eye movement features provided by the present invention, please refer to the above-mentioned embodiment of the cognitive risk assessment system based on eye movement features, and the present invention will not be elaborated here.

[0140] The present invention also provides a storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-mentioned cognitive risk assessment method based on eye movement features.

[0141] For a detailed introduction to a storage medium provided by the present invention, please refer to the above-mentioned embodiment of the cognitive risk assessment system based on eye movement characteristics, and the present invention will not be elaborated here.

[0142] The various embodiments in this specification are described in a progressive manner, and each embodiment focuses on the differences from other embodiments. It should also be noted that in this specification, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between these entities or operations. Moreover, the terms "comprises", "includes" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, article or device that includes a series of elements includes not only those elements, but also other elements that are not explicitly listed, or also includes elements that are inherent to such process, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a..." does not exclude the presence of other identical elements in the process, article or device that includes the element.

[0143] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A cognitive risk assessment system based on eye movement features, characterized in that: include: a cognitive paradigm presentation module, configured to sequentially present on a display screen of a mobile terminal a learning phase of a visual paired comparison task, a reverse saccade task, and a test phase of the visual paired comparison task in a cognitive risk assessment task; The learning phase of the visual paired comparison task includes sequentially presenting a first number of pairs of learning pictures, and the display duration of each pair of learning pictures is the first duration; the testing phase of the visual paired comparison task includes sequentially presenting a second number of pairs of test pictures, each pair of test pictures includes one of the learning pictures and a new picture that has not been presented in the learning phase, and the display duration of each pair of test pictures is the second duration; the reverse saccade task includes a third number of groups of stimulation tasks, when each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen, and after the fixed fixation point is displayed for a preset delay time, a target fixation point at a different position from the fixed fixation point is displayed on the display screen, and the display duration of the target fixation point is the third duration, and the preset delay time corresponding to each group of stimulation tasks changes according to a preset rule; an eye movement data processing module, configured to obtain a facial video captured by a front-facing camera of the mobile terminal while the user completes the cognitive risk assessment task in accordance with the task rules, determine raw eye movement data based on the facial video, and determine quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task based on the raw eye movement data and cognitive risk assessment task information; Among them, the quantitative eye movement features corresponding to the visual paired comparison task include the dwell time on the new picture, the dwell time on the old picture, the number of gaze blinks, the duration of the gaze area, the proportion of dwell time on the new picture, and the preference ratio between the new and old pictures; the quantitative eye movement features corresponding to the reverse saccade task include the maximum saccade velocity, the number of saccade points before stimulation, the number of saccade points after stimulation, the degree of gaze stability, the number of gaze blinks, the initial saccade gain, the saccade gain ratio, the saccade accuracy rate, the saccade error rate, the correct saccade delay time, the correct correct saccade delay time, the number of error corrections, and the correction delay; A cognitive risk assessment module is used to input the quantified eye movement features into a pre-trained machine learning model, and use the machine learning model to determine a cognitive risk assessment result based on the quantified eye movement features.

2. The cognitive risk assessment system based on eye movement features according to claim 1, characterized in that: Determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes: The saccade gain ratio, the correction delay, the maximum saccade speed, the proportion of stay on the new image, the gaze area duration, and the saccade accuracy are input into the machine learning model, so that the machine learning model determines a cognitive assessment risk value based on an eye movement logistic regression formula, and the cognitive assessment risk value is positively correlated with the probability that the user has Alzheimer's disease; the eye movement logistic regression formula is: ； in, Assessing a risk value for the cognition, is a constant term, 、 to The independent variables 、 to The coefficient of is the saccadic gain ratio, For the correction delay, is the maximum eye speed, is the retention ratio of the new image, is the duration of the gaze area, is the eye saccade accuracy rate.

3. The cognitive risk assessment system based on eye movement features according to claim 1, characterized in that: Determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes: Inputting the percentage of new images the user stays on during the visual paired comparison task into the machine learning model, so that the machine learning model determines the user's initial memory ability value based on the memory ability initial value calculation formula, determines a memory ability score based on the memory ability initial value and the memory score calculation formula, and uses the memory ability score as the cognitive risk assessment result; The calculation formula for the initial value of the memory capacity is: ,in, is the initial value of the memory capacity, is a constant term, is the first coefficient, is the second coefficient, is the retention ratio of the new image, The user's age; The memory score calculation formula is: ,in, is the memory ability score, is a constant term, is the third coefficient.

4. The cognitive risk assessment system based on eye movement features according to claim 1, wherein: Determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes: Inputting the correct saccade delay time and the maximum saccade speed into the machine learning model, so that the machine learning model determines the user's initial execution ability value based on an execution ability initial value calculation formula, calculates an execution ability score based on the initial execution ability value and an execution ability score calculation formula, and uses the execution ability score as the cognitive risk assessment result; The calculation formula for the initial value of the execution capability is: ,in, is the initial value of the execution capability, is a constant term, 、 as well as The fourth coefficient, the fifth coefficient and the sixth coefficient are respectively, is the correct saccade delay time, is the maximum saccade velocity, The user's age; The execution ability score calculation formula is: ,in, is the execution ability score, is a constant term, The seventh coefficient.

5. The cognitive risk assessment system based on eye movement features according to claim 1, wherein: Determining a cognitive risk assessment result based on the quantified eye movement features using the machine learning model includes: The gaze area duration is input into the machine learning model so that the machine learning model determines the user's attention ability score based on the gaze area duration and an attention ability score calculation formula, and uses the attention ability score as the cognitive risk assessment result; the attention ability score calculation formula is: ,in, is the attention ability score, is the duration of the gaze area, is a constant term, The eighth coefficient.

6. The cognitive risk assessment system based on eye movement features according to any one of claims 1 to 5, characterized in that: The cognitive paradigm presentation module is further configured to: Select black and white sketch images from the preset image library, and the images that meet the preset task image requirements as learning images and test images in the visual paired comparison task; The method includes determining the screen features of the current mobile terminal, determining the size and display position of each learning image and test image in the visual paired comparison task based on the screen features, and determining the display positions of the fixed fixation point and the target fixation point in the countersaccade task. After the determination, the method then proceeds to the step of sequentially presenting the learning phase of the visual paired comparison task, the countersaccade task, and the test phase of the visual paired comparison task in the cognitive risk assessment task on the display screen of the mobile terminal.

7. A cognitive risk assessment method based on eye movement characteristics, characterized in that: include: sequentially presenting a learning phase of a visual paired comparison task, a reverse saccade task, and a test phase of the visual paired comparison task in a cognitive risk assessment task on a display screen of a mobile terminal; The learning phase of the visual paired comparison task includes sequentially presenting a first number of pairs of learning pictures, and the display duration of each pair of learning pictures is the first duration; the testing phase of the visual paired comparison task includes sequentially presenting a second number of pairs of test pictures, each pair of test pictures includes one of the learning pictures and a new picture that has not been presented in the learning phase, and the display duration of each pair of test pictures is the second duration; the reverse saccade task includes a third number of groups of stimulation tasks, when each group of stimulation tasks is presented, a fixed fixation point is first displayed on the display screen, and after the fixed fixation point is displayed for a preset delay time, a target fixation point at a different position from the fixed fixation point is displayed on the display screen, and the display duration of the target fixation point is the third duration, and the preset delay time corresponding to each group of stimulation tasks changes according to a preset rule; Obtaining a facial video captured by a front-facing camera of the mobile terminal when the user completes the cognitive risk assessment task in accordance with the task rules, determining raw eye movement data based on the facial video, and determining quantitative eye movement features corresponding to the visual paired comparison task and the countersaccade task based on the raw eye movement data and cognitive risk assessment task information; Among them, the quantitative eye movement features corresponding to the visual paired comparison task include the dwell time on the new picture, the dwell time on the old picture, the number of gaze blinks, the duration of the gaze area, the proportion of dwell time on the new picture, and the preference ratio between the new and old pictures; the quantitative eye movement features corresponding to the reverse saccade task include the maximum saccade velocity, the number of saccade points before stimulation, the number of saccade points after stimulation, the degree of gaze stability, the number of gaze blinks, the initial saccade gain, the saccade gain ratio, the saccade accuracy rate, the saccade error rate, the correct saccade delay time, the correct correct saccade delay time, the number of error corrections, and the correction delay; The quantified eye movement features are input into a pre-trained machine learning model, and the machine learning model is used to determine a cognitive risk assessment result based on the quantified eye movement features.

8. A cognitive risk assessment device based on eye movement characteristics, characterized in that: include: memory for storing computer programs; A processor is configured to implement the steps of the method for cognitive risk assessment based on eye movement features as claimed in claim 7 when executing the computer program.

9. A storage medium, characterized in that: The storage medium stores a computer program, which, when executed by a processor, implements the steps of the cognitive risk assessment method based on eye movement features according to claim 7.

Images