Alzheimer disease early screening method

Abstract

The invention relates to the technical field of cognitive impairment image processing, in particular to an Alzheimer's disease early screening method which comprises the following steps: extracting time and space coordinates of a fixation point from a visual glancing task participated by an Alzheimer's disease patient, recording direction change of a jumping behavior in the time and space coordinates, and calculating the direction change of the jumping behavior in the time and space coordinates; and generating jump behavior distribution data. According to the method, by extracting the time and space coordinates of the fixation point and combining the direction change of the jump behavior, comprehensive description of the dynamic law of the fixation behavior is realized, and high-dimensional support is provided for fine analysis of the fixation behavior of a patient in the task execution process. By constructing jump event distribution, track dynamic change and complexity quantification of the gazing path, the offset characteristic, the jump rule and the path fractal dimension in the gazing behavior are accurately revealed, and quantitative description of the abnormal behavior is provided for screening.

Description

Technical Field

[0001] The present invention relates to the technical field of cognitive impairment image processing, and in particular to an early screening method for Alzheimer's disease. Background Art

[0002] Cognitive impairment image processing technology studies and solves problems related to cognitive dysfunction by analyzing brain images, behavioral videos or other related visual data. This technology mainly relies on artificial intelligence, deep learning and pattern recognition technology, combined with multimodal imaging (such as MRI, CT, PET, etc.) and neuroscience knowledge to identify the early characteristics of cognitive dysfunction and provide a scientific basis for medical diagnosis, rehabilitation program design and health monitoring. It is widely used in early detection and intervention of neurological diseases such as Alzheimer's disease, Parkinson's disease, and autism.

[0003] Among them, the early screening method for Alzheimer's disease is an important application direction in the field of cognitive impairment image processing, which aims to discover early biomarkers or pathological characteristics of Alzheimer's disease by analyzing brain images or other related data. Its main purpose is to help clinicians diagnose the disease early through rapid and accurate screening methods, so as to formulate personalized intervention measures.

[0004] When analyzing gaze behavior, the existing technology has difficulty in dynamically recording the direction, amplitude and time series characteristics of gaze point jump behavior, and cannot deeply identify the changing patterns of gaze paths within and outside the task stage. For example, in a memory search task, the existing method can only roughly count the distribution of gaze points, but it is difficult to capture the microscopic characteristics of the distribution of return points, bifurcation points and hot spots, resulting in the behavioral characteristics of abnormal task execution of patients being ignored. For the analysis of the complexity and changing patterns of gaze paths, the existing technology has difficulty in quantifying the complexity of gaze behavior through means such as path fractal dimension, making it difficult to identify cognitive abnormalities hidden in the patient's jump path. In the comparative analysis of hot spots and healthy groups, the existing method has difficulty in counting the hot spot distribution of gaze behavior based on jump frequency and direction changes, and cannot reveal significant differences between patients and the control group during the task switching stage. The above-mentioned deficiencies lead to insufficient coverage of dynamic behavioral characteristics in disease screening by the existing technology, which may lead to a decrease in the sensitivity of screening results to abnormal cognitive functions. Summary of the invention

[0005] The purpose of the present invention is to solve the shortcomings of the prior art and to propose an early screening method for Alzheimer's disease.

[0006] In order to achieve the above object, the present invention adopts the following technical scheme: an early screening method for Alzheimer's disease, comprising the following steps:

[0007] S1: Extract the temporal and spatial coordinates of the fixation point from the visual saccade task of Alzheimer's disease patients, record the directional changes of the jumping behavior in the temporal and spatial coordinates, and generate the jumping behavior distribution data;

[0008] S2: based on the jump behavior distribution data, the direction, amplitude and fixation duration of the jump events are grouped and arranged, and the position changes are recorded, and the density of the fixation point distribution position is combined to generate a fluctuation characteristic diagram of the visual scanning task;

[0009] S3: Based on the fluctuation characteristic diagram of the visual scanning task, the distribution of target points and non-target points in the fixation path is recorded, the continuity and jump changes of the fixation behavior are referred to, the return points and bifurcation points are marked, the number and length of path switching are counted, and the path characteristic data of the patient's memory search task is generated;

[0010] S4: Based on the path characteristic data of the patient's memory search task, according to the relative position change between the target point and the fixation point, the direction and amplitude of the deviation event are recorded and the deviation category is distinguished, so as to generate the path deviation characteristic data in the patient's search task;

[0011] S5: Collect the data of the memory search link of the healthy population, compare it with the path deviation characteristic data in the patient search task, mark the hotspot areas and fixation points, analyze the difference between the hotspot distribution of the patient jump path and the control group, and generate hotspot fixation feature data;

[0012] S6: Based on the hotspot gaze feature data, the number and position of gazes on the target object and the distractor are recorded, the gaze path during the task switching phase is analyzed, the behavioral characteristics of the target gaze path are extracted, and early screening data for Alzheimer's disease is generated for early screening of Alzheimer's disease.

[0013] The present invention has the following improvements: the jump behavior distribution data includes time records of fixation points, multi-dimensional position distribution of jump directions, and fixation time duration data; the fluctuation characteristic diagram of the visual scanning task includes jump frequency change distribution in each stage of the fixation trajectory, dynamic fluctuation range of jump direction consistency, and trajectory distribution density information of the fixation time period; the path characteristic data of the patient memory search task includes fixation coverage distribution of target points, jump behavior statistics of non-target points, and the number and position relationship of return points in the path; the path offset characteristic data in the patient search task includes the distance change range between the target point and the fixation point, the direction angle distribution of the offset event, and the frequency statistics of the offset amplitude; the hotspot fixation feature data includes the coverage difference of the fixation point hotspot area between patients and healthy people, the distribution density of multi-frequency fixation points, and the priority grouping of the fixation hotspot area; the early screening data for Alzheimer's disease includes the stage change distribution of the number of target object fixations in dynamic task switching, the jump frequency distribution of the distractor fixation path, and the classification result of the fixation priority.

[0014] The present invention has been improved in that the specific steps of extracting the time and space coordinates of the fixation point from the visual scanning task participated by the Alzheimer's disease patient, recording the direction change of the jumping behavior in the time and space coordinates, and generating the jumping behavior distribution data are as follows:

[0015] S101: Based on the time and space coordinates of the gaze points of Alzheimer's patients recorded by the eye movement device in the visual scanning task, identify the spatial position relationship between the gaze points and the task target, distractors, and blank areas in the visual scanning task, classify the association records between the gaze points and various types of areas, and generate the spatial area distribution data of the gaze points;

[0016] S102: Based on the spatial distribution data of the fixation points, the number and duration of fixations on the target object, the distractors and the blank area are counted, the time proportion of the fixation types is classified, the distribution characteristics of the fixation types are marked, and the fixation type time statistics results are generated;

[0017] S103: Based on the time statistics of the gaze type, record the spatial jump changes between gaze points, calculate the angle characteristics of the jump direction, mark the change rules of the jump direction according to the time series of the jump events, and generate jump behavior distribution data.

[0018] The present invention is improved in that, based on the jump behavior distribution data, the direction, amplitude and fixation duration of the jump events are grouped and arranged and the position changes are recorded, and the specific steps of generating the fluctuation characteristic diagram of the visual scanning task in combination with the density of the fixation point distribution position are as follows:

[0019] S201: Based on the jumping behavior distribution data, the jumping events are segmented in time according to the time interval of the gaze point and the change of the coordinates, the fractal dimension of the jumping path is calculated by dividing the coordinate interval into equally spaced grids, the complexity and change law of the jumping behavior in the gaze path are quantified, and the segmented trajectory record of the jumping event is established;

[0020] S202: based on the jump event segmented trajectory record, extract the continuous change record of the gaze trajectory of the visual scanning task in chronological order, record the direction, amplitude and gaze time interval of each jump event in the gaze trajectory, and generate a gaze trajectory sequence feature record;

[0021] S203: Based on the gaze trajectory sequence feature record, the dynamic changes of the gaze trajectory within the task stage are analyzed, and according to the time change trend of the jump frequency, the transfer pattern of the gaze point in the gaze concentration area of ​​the visual scanning task, and the difference in the distribution of the gaze time within and outside the task stage, a fluctuation characteristic diagram of the visual scanning task is established.

[0022] The present invention is improved in that the fractal dimension of the jump path is calculated by using the formula:

[0023]

[0024] Get the fractal dimension D f , represents the self-similarity of the path at multiple scales;

[0025] Where ∈ is the grid size, N(∈) is the number of grids covered by the jump path, T represents the temporal intensity of the fixation point in the jump path, log is used to map the number of grids covered by the jump path N(∈) and the change of the grid scale ∈ to the logarithmic space, and log(1 / ∈) is the logarithm of the inverse of the grid scale, reflecting the change of the discretization scale of the path.

[0026] The present invention is improved in that, based on the fluctuation characteristic diagram of the visual scanning task, the distribution of target points and non-target points in the fixation path is recorded, the continuity and jump changes of the fixation behavior are referred to, the return points and bifurcation points are marked, the number and length of path switching are counted, and the specific steps of generating the path characteristic data of the patient's memory search task are as follows:

[0027] S301: Based on the fluctuation characteristic diagram of the visual scanning task, extract the data source of the memory search link in the visual scanning task by random clustering, record the starting point and the ending point of the fixation path in the memory search link, mark the spatial distribution positions of the target points and non-target points in the fixation path, and generate the target and non-target distribution data of the memory search link;

[0028] S302: Based on the target and non-target distribution data of the memory search phase, extract the position change of the gaze point in the gaze path according to the time series of the gaze path, record the direction change and frequency of the gaze point jump event, mark the return point and bifurcation point in the gaze path, and generate the memory search phase gaze path jump characteristic data;

[0029] S303: Based on the gaze path jump characteristic data of the memory search link, the number of gaze point switches and the path length in the gaze path are grouped and counted, the gaze behavior is classified according to the density of the intersection points of the gaze path, the stage characteristics of the gaze path are integrated into the overall distribution of the task path, and the patient's memory search task path characteristic data is generated.

[0030] The present invention is improved in that, based on the path characteristic data of the patient's memory search task, according to the relative position change between the target point and the fixation point, the direction and amplitude of the deviation event are recorded and the deviation category is distinguished, and the specific steps of generating the path deviation characteristic data in the patient's search task are as follows:

[0031] S401: Based on the patient's memory search task path characteristic data, the relative spatial position of the target point and the gaze point in the gaze path in the memory search phase is recorded, the direction angle and amplitude of each offset event in the gaze path are marked, and the spatial relationship data between the target point and the gaze point in the gaze path are generated;

[0032] S402: Based on the spatial relationship data between the target point and the gaze point in the gaze path, classify the offset events according to the direction consistency and amplitude of the gaze point offset events in the gaze path, annotate multiple distance offsets and direction distribution characteristics of the distance offset events, and generate classification data of the offset events in the gaze path;

[0033] S403: Based on the classification data of the deviation events in the gaze path, the frequency distribution of each category of deviation events is counted, the general direction and amplitude range of the deviation events in the gaze path are marked, the statistical characteristics of the deviation categories in the gaze path are summarized, and the path deviation characteristic data in the patient search task is generated.

[0034] The present invention has the following improvements: collecting data from the memory search link of healthy people, comparing it with the path deviation characteristic data in the patient search task, marking hotspot areas and fixation points, analyzing the difference between the hotspot distribution of the patient jump path and the control group, and generating hotspot fixation characteristic data in the following specific steps:

[0035] S501: Based on the collected data of the memory search phase of the healthy population, the data of the path deviation characteristics in the patient search task are compared, the distribution of the gaze points in the gaze paths of the patients and the healthy population is marked, the number of gazes and path changes of the target points and non-target points are recorded by category, and the comparison results of the gaze paths of the patients and the healthy population are generated;

[0036] S502: Based on the comparison results of the gaze paths of the patient and the healthy population, the hot spot distribution of the patient and the healthy population is counted, the spatial positions and gaze times of the multi-frequency gaze points are marked, the difference in the hot spot distribution in the jump path of the patient and the healthy population is analyzed, and the difference in the gaze hot spot distribution between the patient and the healthy population is generated;

[0037] S503: Based on the difference data of the gaze hotspot distribution between the patients and the healthy population, the distribution center of the patient's gaze hotspot area is marked, and the coverage of the patient's hotspot area and the target point is counted to generate hotspot gaze feature data.

[0038] The present invention has been improved in that, based on the hotspot gaze feature data, the number and position of gazes on the target object and the distractor are recorded, the gaze path in the task switching phase is analyzed, the behavioral characteristics of the target gaze path are extracted, and early screening data for Alzheimer's disease is generated. The specific steps for applying the data to early screening for Alzheimer's disease are as follows:

[0039] S601: Based on the hotspot gaze feature data, the gaze point types of the gaze path in the hotspot area are marked by hierarchical residual analysis, the gaze times and spatial distribution of the target object and the distractor are recorded, and the gaze point distribution characteristic data of the hotspot area is generated according to the distribution characteristics of the gaze points in the hotspot area;

[0040] S602: Based on the hotspot area gaze point distribution characteristic data, record the gaze path in the task dynamic switching stage, and according to the switching order and frequency of the target object and the interference object in the jump path, count the path priority area and jump frequency changes in the task switching stage to generate the gaze path characteristic data in the task dynamic switching stage;

[0041] S603: Based on the gaze path characteristic data in the dynamic switching stage of the task, the gaze priority deviation in the patient's gaze hotspot is marked, and the target point focusing frequency, jump direction and amplitude stability, and hotspot area coverage are recorded by task stage. The gaze behavior characteristics in the dynamic switching stage are extracted, and the patient's gaze behavior tendency in subsequent tasks is predicted to generate early screening data for Alzheimer's disease.

[0042] The present invention is improved in that, for predicting the patient's gaze behavior tendency in subsequent tasks, the formula is adopted:

[0043]

[0044] Calculate the gaze behavior tendency value P t , which measures whether the patient is more inclined to fixate on the target point, distractor, or hot spot in the subsequent task, and usually ranges from 0 to 1;

[0045] Among them, A i Represents the weight factor of the fixation type in the current task stage, D′ i is the corrected duration of the fixation point, F i is the jump frequency of the fixation point, and n is the total number of fixations recorded in the current task.

[0046] Compared with the prior art, the advantages and positive effects of the present invention are:

[0047] In the present invention, by extracting the time and space coordinates of the gaze point and combining the change in the direction of the jumping behavior, a comprehensive characterization of the dynamic laws of the gaze behavior is achieved, providing high-dimensional support for the detailed analysis of the patient's gaze behavior during the task execution process. By constructing the jump event distribution, trajectory dynamic changes and complexity quantification of the gaze path, the offset characteristics, jump laws and path fractal dimensions in the gaze behavior are accurately revealed, providing a quantitative description of abnormal behavior for screening. Through the distribution statistics of target points and non-target points and the marking of return points and bifurcation points, the granularity of the path analysis of the memory search task is significantly improved, making the behavioral differences inside and outside the task stage clearer. Hot spot area analysis combines the priority areas of the gaze path with the law of task switching, and provides systematic data support for the comparative analysis of the patient's gaze behavior characteristics in complex tasks with healthy groups. By extracting behavioral feature data and dynamic change trends, the patient's gaze hot spot distribution and the abnormalities in the task switching stage are fully captured, providing data support for early clinical intervention measures. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 A flow chart of an early screening method for Alzheimer's disease is proposed for the present invention;

[0049] Figure 2 This is a detailed flow chart of step S1 of the present invention;

[0050] Figure 3 This is a detailed flow chart of step S2 of the present invention;

[0051] Figure 4 This is a detailed flow chart of step S3 of the present invention;

[0052] Figure 5 This is a detailed flow chart of step S4 of the present invention;

[0053] Figure 6 This is a detailed flow chart of step S5 of the present invention;

[0054] Figure 7 This is a detailed flow chart of step S6 of the present invention. DETAILED DESCRIPTION

[0055] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0056] In the description of the present invention, it should be understood that the terms "length", "width", "up", "down", "front", "back", "left", "right", "vertical", "horizontal", "top", "bottom", "inside", "outside", etc., indicating positions or positional relationships, are based on the positions or positional relationships shown in the accompanying drawings, and are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as limiting the present invention. In addition, in the description of the present invention, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.

[0057] See also Figure 1 The present invention provides a technical solution: an early screening method for Alzheimer's disease, comprising the following steps:

[0058] S1: Extract the temporal and spatial coordinates of the fixation point from the visual saccade task of Alzheimer's disease patients, record the directional changes of the jumping behavior in the temporal and spatial coordinates, and generate the jumping behavior distribution data;

[0059] S2: Based on the jump behavior distribution data, the direction, amplitude and fixation duration of the jump events are grouped and arranged, and the position changes are recorded. Combined with the density of the fixation point distribution position, a fluctuation characteristic diagram of the visual saccade task is generated;

[0060] S3: Based on the fluctuation characteristic diagram of the visual scanning task, the distribution of target points and non-target points in the gaze path is recorded, the continuity and jump changes of the gaze behavior are referred to, the return points and bifurcation points are marked, the number and length of path switching are counted, and the path characteristic data of the patient's memory search task is generated;

[0061] S4: Based on the path characteristic data of the patient's memory search task, according to the relative position change between the target point and the fixation point, the direction and amplitude of the deviation event are recorded and the deviation category is distinguished to generate the path deviation characteristic data in the patient's search task;

[0062] S5: Collect data on the memory search process of healthy people, compare it with the path deviation characteristic data in the patient search task, mark the hotspot areas and fixation points, analyze the differences in the hotspot distribution of the patient jump path and the control group, and generate hotspot fixation feature data;

[0063] S6: Based on the hotspot gaze feature data, record the number and position of gazes on the target and distractors, analyze the gaze path during the task switching phase, extract the behavioral characteristics of the target gaze path, generate early screening data for Alzheimer's disease, and apply it to early screening for Alzheimer's disease.

[0064] The jump behavior distribution data include the time record of the fixation point, the multi-dimensional position distribution of the jump direction and the fixation time duration data. The fluctuation characteristic diagram of the visual scanning task includes the jump frequency change distribution in each stage of the fixation trajectory, the dynamic fluctuation range of the jump direction consistency and the trajectory distribution density information of the fixation time period. The path characteristic data of the patient's memory search task include the fixation coverage distribution of the target point, the jump behavior statistics of the non-target point and the number and position relationship of the return points in the path. The path deviation characteristic data in the patient's search task include the distance variation range between the target point and the fixation point, the direction angle distribution of the deviation event and the frequency statistics of the deviation amplitude. The hotspot fixation feature data include the coverage difference of the fixation point hotspot area between patients and healthy people, the distribution density of multi-frequency fixation points and the priority grouping of the fixation hotspot area. The early screening data for Alzheimer's disease include the stage change distribution of the number of fixations on the target object in dynamic task switching, the jump frequency distribution of the distractor fixation path and the classification results of the fixation priority.

[0065] See also Figure 2 ,From the visual saccade task participated in by Alzheimer’s patients, the temporal and spatial coordinates of the fixation points are extracted, the ,directional changes of the jumping behavior in the temporal and spatial coordinates are recorded, and the specific steps of generating the ,jumping behavior distribution data are as follows:

[0066] S101: Based on the time and space coordinates of the gaze points of Alzheimer's patients recorded by the eye movement device in the visual scanning task, identify the spatial position relationship between the gaze points and the task target, distractors, and blank areas in the visual scanning task, classify the association records between the gaze points and various types of areas, and generate the spatial area distribution data of the gaze points;

[0067] First, the original eye movement data is cleaned up, and incomplete data records and abnormal coordinate values ​​that deviate from the main task area are deleted in the time series. For example, if the distance between a fixation point and the center of the main task area exceeds a certain threshold, the fixation point record can be removed. Then the retained data is sorted in chronological order to construct a complete time and space trajectory. The fixation points are divided into target objects, distractors, and blank areas using clustering methods. In the specific operation, the Euclidean distance from the fixation point to the center of the area is used as the main classification basis. At the same time, the physical size and boundary data of the task area are introduced for boundary adjustment. For example, when analyzing a task containing three target areas and two distractor areas, the density of the fixation point distribution in each area can be evaluated, the classification effect can be checked by visualization, and the fixation points in the fuzzy area within the boundary can be reallocated. Finally, the classification and labeling of the fixation points and area types are completed, and the spatial area distribution data of the fixation points are generated. The K-Means method in Scikit-learn is used through Python tools to perform cluster analysis, and Matplotlib is called to generate a visual distribution map. The classification of data points within the fuzzy boundary is manually adjusted according to the results.

[0068] S102: Based on the spatial distribution data of the fixation points, the number and duration of fixations on the target object, the distractors and the blank area are counted, the time proportion of the fixation types is classified, the distribution characteristics of the fixation types are marked, and the fixation type time statistics are generated;

[0069] By establishing a time series matrix, we can obtain the duration and frequency of each fixation point in different areas. The specific operations include extracting the start and end time of fixation from the timestamp, calculating the duration of each fixation, accumulating the total time in the area, and counting the frequency. For example, for the fixation points in a target area, we record the duration of each fixation point in the area and calculate the total number of stays. Then, we generate a fixation time distribution table based on the area type, analyze the time distribution characteristics of each area, and present the fixation time and frequency in a time series graph. Finally, we mark the differences in the time distribution characteristics between the target area and the distractor area. For example, the total fixation time of a patient in the target area may be significantly lower than that of the healthy group. The analysis results of this characteristic data can further reveal the abnormal fixation behavior of patients during task completion. We use Excel to import fixation data to generate a time distribution table, draw a fixation time series graph through ggplot2 in R language, and further present the results in a direct way by combining a bar graph of the regional fixation time ratio.

[0070] S103: Based on the time statistics of the gaze type, record the spatial jump changes between the gaze points, calculate the angle characteristics of the jump direction, mark the change law of the jump direction according to the time series of the jump event, and generate the jump behavior distribution data;

[0071] First, the coordinates of adjacent fixation points are calculated to obtain the spatial change vector of the jump event, such as recording the displacement length and direction from a target fixation point to the next distractor fixation point. Then, the distribution of the directional angle is used as the analysis focus, and the jump direction of the fixation point is divided into multiple intervals according to the angle, such as dividing 360 degrees into 8 45-degree intervals. By counting the frequency distribution in each interval, the main direction trend of the jump behavior is analyzed. Further, according to the jump frequency and direction change distribution, abnormal high-frequency direction areas or disordered jump areas are marked. For example, the jump path of a patient shows frequent jumps to the distractor direction and lacks obvious rules, which is significantly different from the distribution of the target as the main jump direction in the healthy control group. These feature data can provide effective clues for the early identification of Alzheimer's disease. The directional angle of each jump event is calculated by the angle function of MATLAB, and the frequency is counted by matching the directional angle with the interval boundary, and the bar function in MATLAB is called to generate a directional angle distribution bar chart to intuitively present the directional jump characteristics.

[0072] See also Figure 3 Based on the jump behavior distribution data, the direction, amplitude and fixation duration of the jump events are grouped and arranged and the position changes are recorded. Combined with the density of the fixation point distribution position, the specific steps for generating the fluctuation characteristic diagram of the visual scanning task are as follows:

[0073] S201: Based on the jumping behavior distribution data, the jumping events are segmented in time according to the time interval of the fixation points and the coordinate changes, and the fractal dimension of the jumping path is calculated by dividing the coordinate interval into equally spaced grids, so as to quantify the complexity and change law of the jumping behavior in the fixation path and establish a segmented trajectory record of the jumping events;

[0074] The jump events are segmented in time according to the time interval and coordinate change of the fixation points. For example, the time interval of the jump path from point A (2, 3) to point B (5, 6) is 0.5 seconds, its direction is the vector angle from A to B, the amplitude is the Euclidean distance from point A to point B, and the fixation duration is 0.2 seconds for point A and 0.3 seconds for point B. The direction, amplitude and duration of all paths will be converted into discretized time and space trajectory data as the basis for fractal dimension calculation. The grid generation of the jump path is based on the coordinate range of the fixation point. First, the maximum and minimum coordinate values ​​of the task area are extracted, and the coordinate interval is divided into equally spaced grids according to the set grid size ∈. For example, in a 100×100 task area, when ∈=10, the side length of each grid is 10 units, and the task area will be divided into a 10×10 grid matrix, and each grid corresponds to a unique spatial index. The path data is projected onto the grid index to determine the number of grids covered N(∈). For example, point A (15, 25) and point B (45, 65) in the jump path fall on grids (2, 3) and (5, 7) respectively, and the jump path between the two points covers the index ranges of these grids.

[0075] For calculating the fractal dimension of the jump path, the formula is used:

[0076]

[0077] Get the fractal dimension D f , represents the self-similarity of the path at multiple scales;

[0078] Among them, D fIt is used to quantify the complexity of the jump path. ∈ is the grid size, which indicates the size of the grid during the discretization of the path. It is usually generated by setting a fixed ratio of grid intervals. The appropriate ∈ value can be selected according to the task area size and resolution of the experimental scene. N(∈) is the number of grids covered by the jump path. It is used to measure the spatial complexity of the path under a specific grid size ∈. It is obtained by counting the grids containing jump path data in the discretized grid. T is the sum of the fixation durations, which indicates the time intensity of the fixation points in the jump path. It is obtained by summing the time intervals of all fixation points. log is used to map the number of grids covered by the jump path N(∈) and the change of the grid scale ∈ to the logarithmic space, which is convenient for linear processing of data at different orders of magnitude, so as to calculate the fractal dimension of the path. The role of logarithmic operation is to smooth the differences in high-order data, making the fitting calculation more robust. log(1 / ∈) means that the value divided by 1 and the grid size ∈ is the logarithm of the independent variable. Its meaning is to take the logarithm of the inverse of the grid scale, which can reflect the change of the discretization scale of the path.

[0079] For example, if the jump path data is as follows: grid size ∈: 0.1, 0.01, 0.001; number of grids N(∈): 50, 500, 5000; total fixation duration T = 10 seconds; log is the natural logarithm (with base e), calculate:

[0080] log(N(∈)·T)=log(50·10)=2.698, log(500·10)=3.698, log(5000·10)=4.698

[0081] log(1 / ∈)=log(1 / 0.1)=1, log(1 / 0.01)=2, log(1 / 0.001)=3

[0082] Fitting slope:

[0083]

[0084] This result shows that the fractal dimension D f =1, fractal dimension D f The value range of is usually between 1 and 2. The closer the value is to 2, the more complex the jump path is and the wider the coverage is. The closer the value is to 1, the more linear the path is. For example, if the D f =1.8, indicating that this path shows high complexity and dispersion, while D f=1.2 indicates that the path tends to be simple and has strong directionality. According to the local changes in the fractal dimension, the abnormal features in the jump path are extracted. For example, if the fractal dimension of a certain path segment is significantly higher than that of other path segments, it can be marked as a complex jump segment, and the distribution characteristics of the gaze points in this segment can be further analyzed; if the fractal dimension gradually decreases, it can be marked as a concentrated jump segment, pointing to the jump behavior in the target area. The jump path is divided into multiple segments according to the time and space changes. Each segment corresponds to a fractal dimension value. The jump direction, amplitude and fixation duration in the segment are recorded to generate a trajectory record of the jump event. For example, a certain trajectory from time 0 to 2 seconds has a direction of 45°, an amplitude of 10 units, and a fractal dimension of 1.7; the next trajectory from time 2 to 4 seconds has a direction of 90°, an amplitude of 5 units, and a fractal dimension of 1.3. Through these segmentation information, the jump behavior characteristics in the fixation path can be fully described. Combined with the distribution of fractal dimensions, the jump trajectory is divided into high-complexity jumps (such as fractal dimension D f >1.5), low complexity jumps (such as D f ≤1.5) and directional concentrated characteristic jumps, generate segmented trajectory records of jumping events according to classification, and provide basic data support for subsequent behavioral feature analysis.

[0085] S202: based on the jump event segmented trajectory record, extract the continuous change record of the gaze trajectory of the visual saccade task in chronological order, record the direction, amplitude and gaze time interval of each jump event in the gaze trajectory, and generate a gaze trajectory sequence feature record;

[0086] The gaze trajectory is organized into a continuous change record in chronological order. The direction, amplitude and gaze time interval of each jump event are extracted. The timestamp information in the segmented trajectory is sorted and the jump order is marked. Each jump event is connected according to the start and end points of the time axis and organized into a continuous time series record. The direction data is directly obtained from the segmented trajectory, the amplitude data is combined with the spatial position change of the segmented trajectory record, and the gaze time interval is calculated by the difference between the start time and the end time. The record format of each jump event is as follows: time period (such as 0-2 seconds), direction (such as 45°), amplitude (such as 10 units), and gaze time interval (such as 2 seconds). By continuously splicing each record, a complete sequence of the gaze trajectory is established, and the sequence is organized into a feature list segmented by time.

[0087] S203: Based on the gaze trajectory sequence feature record, the dynamic changes of the gaze trajectory in the task stage are analyzed, and according to the time change trend of the jump frequency, the transfer pattern of the gaze point in the gaze concentration area of ​​the visual saccade task, and the difference in the gaze time distribution within and outside the task stage, a fluctuation characteristic diagram of the visual saccade task is established;

[0088] The dynamic change data of gaze trajectory in the task stage were extracted. The stages were divided according to time. The gaze trajectory was segmented by time. The jump frequency, direction distribution and amplitude of each segment were extracted. The regularity of gaze behavior was analyzed by using the relationship between the total amount of segmented gaze time and the jump frequency. The data segmentation and statistics were realized through MATLAB tools. The readtable function was used to import the gaze trajectory sequence feature record file and segment it according to the timestamp. The jump frequency of each stage was extracted and the stage frequency distribution table was generated by combining the histcounts function. The spatial distribution of the gaze points in the gaze trajectory was extracted. The focus area and the transfer path between areas were recorded according to the area number. The area transfer matrix was generated by counting the frequency of the gaze point transferring from one area to another in each stage. The rows in the matrix represent the starting area, the columns represent the target area, and the values ​​are the area transfer frequencies. After the area transfer matrix was generated by the accumarray function of MATLAB, the law of gaze transfer between areas can be intuitively reflected. The distribution characteristics of the gaze time in the task stage were compared with the gaze time outside the stage, and the difference of gaze time inside and outside the task was extracted. Finally, a fluctuation characteristic diagram was generated, including the trend of jump frequency change, the transfer pattern of the focus area, and the difference of gaze time distribution inside and outside the task stage.

[0089] See also Figure 4 ,Based on the fluctuation characteristic diagram of the visual scanning task, the distribution of target points and non-target points in the gaze path is recorded, the continuity and jump changes of gaze behavior are referred to, the return points and bifurcation points are marked, the number and length of path switching are counted, and the specific steps for generating the path characteristic data of the patient's memory search task are as follows:

[0090] S301: Based on the fluctuation characteristic diagram of the visual scanning task, extract the data source of the memory search link in the visual scanning task by random clustering, record the starting point and the ending point of the fixation path in the memory search link, mark the spatial distribution positions of the target points and non-target points in the fixation path, and generate the target and non-target distribution data of the memory search link;

[0091] The data source of the memory search link in the visual scanning task is extracted by random clustering, and the patient's gaze path data in the task is imported into the cluster analysis tool, such as the Scikit-learn library in Python. Clustering is performed by calling the KMeans method, and the number of clusters is set based on the actual number of task targets. The gaze points are divided into multiple clusters according to the Euclidean distance of the spatial position. Each cluster represents a gaze area. The clustering results are manually corrected to exclude non-target areas and are finally marked as target points or non-target points. Combined with the starting and ending point data of the gaze path, a time series record is generated to record the spatial distribution positions of the target points and non-target points in each gaze path. For example, the proportion of the time a patient's gaze path stays in the target point cluster is 70%, and the proportion of the time it stays in the non-target point is 30%. This distribution data can reflect the patient's ability to focus on the target point during the memory search process.

[0092] S302: Based on the target and non-target distribution data of the memory search phase, the position changes of the gaze points in the gaze path are extracted according to the time series of the gaze path, the direction changes and frequencies of the gaze point jump events are recorded, the return points and bifurcation points in the gaze path are marked, and the memory search phase gaze path jump characteristic data are generated;

[0093] The continuous position changes of the gaze point are extracted according to the time series of the gaze path, and the jump events of the gaze point are recorded. The jump direction characteristics are extracted by calculating the direction changes of adjacent gaze points. The atan2 function in MATLAB is used to calculate the direction angle difference. By inputting the coordinates of the two gaze points (x1, y1) and (x2, y2), the vector angle formed by the two points is calculated. The output direction difference is directly stored in the data table. At the same time, the frequency of jump events is counted, and the gaze points whose direction mutation frequency exceeds a certain threshold are marked as return points. For example, the return points of a patient account for 20% of the total number of gaze points, reflecting the instability of his jumping behavior. The bifurcation points are marked by counting the event frequency of a single gaze point jumping in multiple directions, and the bifurcation point dense area is extracted. For example, in a certain section of the path, the bifurcation points of a patient are concentrated in a specific target area. Combined with the bifurcation point distribution, the patient's jumping bias in the task is evaluated.

[0094] S303: Based on the gaze path jump characteristic data of the memory search phase, the number of gaze point switches and the path length in the gaze path are grouped and counted, the gaze behavior is classified according to the density of the intersection points of the gaze path, the stage characteristics of the gaze path are integrated into the overall distribution of the task path, and the patient's memory search task path characteristic data is generated;

[0095] The gaze paths were grouped by stages, and the number of gaze point switches in each stage was counted. The path length was calculated through the gaze point sequence of the time period. The distribution of intersections was counted through the gaze point dense area, and the gaze path crossing behavior was divided into high-density paths and low-density paths. The path complexity was evaluated by the segmented jump frequency and direction changes. For example, the jumping frequency of a patient in the high-density area accounted for 60%, while that of the healthy control group was only 30%, reflecting the excessive dispersion of the patient's jumping behavior in the task. The gaze paths were divided into behavioral patterns of different complexities through the number and distribution position characteristics of the intersections, and the time period and intersection data were integrated to generate the overall distribution of the task path.

[0096] See also Figure 5 ,Based on the path characteristic data of the patient's memory search task, according to the relative position change between the target point and the fixation point, the direction and amplitude of the deviation event are recorded and the deviation category is distinguished. The specific steps for generating the path deviation characteristic data in the patient's search task are as follows:

[0097] S401: Based on the patient's memory search task path characteristic data, the relative spatial position of the target point and the fixation point in the fixation path in the memory search phase is recorded, the direction angle and amplitude of each offset event in the fixation path are marked, and the spatial relationship data between the target point and the fixation point in the fixation path are generated;

[0098] The gaze path data was processed by using the pandas and numpy libraries in Python tools. The difference between the spatial coordinates of the target point and the spatial coordinates of each gaze point was calculated, and the position difference in the x and y directions was extracted respectively and stored as two columns of data. The distance from the gaze point to the target point was then calculated using the Euclidean distance formula and stored in a spatial relationship data file. On this basis, the direction angle calculation formula was used in combination with Python's math.atan2 function to calculate the direction angle of the gaze point offset event. The offset event was annotated based on the amplitude calculation result. For example, the proportion of offset events with a direction angle of 30° in a certain path was 20%, indicating that the patient had a regular offset in a specific direction. Finally, the spatial relationship data between the target point and the gaze point in the gaze path was generated.

[0099] S402: Based on the spatial relationship data between the target point and the gaze point in the gaze path, classify the offset events according to the direction consistency and amplitude of the gaze point offset events in the gaze path, annotate multiple distance offsets and direction distribution characteristics of the distance offset events, and generate classification data of the offset events in the gaze path;

[0100] The deviation events in the gaze path were classified according to their directional consistency and amplitude. The directional deviation consistency ratio of each segment of the gaze path was statistically analyzed by segment. The directional deviation of the gaze path was classified and statistically analyzed using Python's groupby function. The amplitude was divided into short-distance deviation (less than a certain threshold) and long-distance deviation (greater than the threshold). Each deviation event was labeled based on the directional distribution characteristics. For example, in the data of a certain patient, the frequency of long-distance deviation events was higher than that of the healthy control group, reflecting the lack of stability in the task execution process. Finally, the classification data of deviation events in the gaze path was generated.

[0101] S403: Based on the classification data of the deviation events in the gaze path, the frequency distribution of each category of deviation events is counted, the conventional direction and amplitude range of the deviation events in the gaze path are marked, the statistical characteristics of the deviation categories in the gaze path are summarized, and the path deviation characteristic data in the patient search task is generated;

[0102] The frequency distribution of each category of offset events was statistically analyzed, and the frequencies of all offset events in the gaze path were summarized according to their classification results. The frequencies were further subdivided by combining the direction and amplitude range. The frequency bar graph of each offset event was drawn using Python's matplotlib library, and the conventional direction range of the offset events was marked. For example, the offset events of a certain patient were concentrated in the direction of 0° to 45°, and the amplitude range was 75% of the short-distance offset. The data of the healthy control group were more evenly distributed. The offset characteristics of the patients were extracted through comparative analysis. Combined with the statistical characteristics of the frequencies of different categories, the offset category information in the gaze path was summarized as the offset characteristic data of the patient's search task path.

[0103] See also Figure 6 ,Collect the data of memory search link of healthy people, compare it with the path deviation characteristic data in the patient search task, annotate the hotspot areas and fixation points, analyze the difference between the hotspot distribution of the patient jump path and the control group, and generate the hotspot fixation feature data in the following specific steps:

[0104] S501: Based on the collected data of the memory search phase of the healthy population, the data of the path deviation characteristics in the patient search task are compared, the distribution of the gaze points in the gaze paths of the patients and the healthy population is marked, the number of gazes and path changes of the target points and non-target points are recorded in categories, and the comparison results of the gaze paths of the patients and the healthy population are generated;

[0105] The gaze path data of patients and healthy people were extracted, and the data were imported into Python's pandas and numpy libraries for data preprocessing. Pandas was used to filter out the task target point-related gaze points and non-target point gaze points, and the starting and ending points of each gaze path were recorded. The number of gazes on the target point and the number of gazes on the non-target point were counted separately, and the data were integrated into two groups: target gaze video data and non-target gaze video data. Combined with the sequence characteristics of the gaze path, the gaze distribution characteristics of patients and healthy people were annotated, and the distribution diagrams of the two groups of data were drawn using matplotlib. For example, in a certain test stage, the proportion of target point gaze time in healthy people was 75%, while that of patients was only 50%, reflecting the problem of patients' attention deviation.

[0106] S502: Based on the comparison results of the gaze paths of patients and healthy people, the hot spot area distribution of patients and healthy people is counted, the spatial positions and gaze times of multi-frequency gaze points are marked, the difference in the hot spot distribution in the jump path of patients and healthy people is analyzed, and the difference data of the gaze hot spot distribution between patients and healthy people is generated;

[0107] The hotspot distribution of patients and healthy people was statistically analyzed, and the DBSCAN method in Python's scikit-learn library was used for spatial clustering. The coordinates of each fixation point were input, and the hotspot areas were divided according to the Euclidean distance parameter. The areas where high-frequency fixations were located were identified, and the center coordinates and fixation times of each hotspot area were extracted. The hotspot area positions were marked by drawing a two-dimensional spatial distribution map, and the fixation proportion of each hotspot area was statistically analyzed by pandas. For example, the fixation time of a patient in a hotspot area accounted for 65%, while that of healthy people was only 20%. The differences in fixation behavior between patients and healthy people were analyzed in combination with the hotspot distribution characteristics, and the difference data of fixation hotspot distribution between patients and healthy people were generated.

[0108] S503: Based on the difference data of the gaze hotspot distribution between the patients and the healthy population, the distribution center of the patient's gaze hotspot area is marked, and the coverage of the patient's hotspot area and the target point is counted to generate hotspot gaze feature data;

[0109] The distribution center of the patient's gaze hotspot area is marked. Combined with the location data of the target point, a scatter plot is drawn using Python's matplotlib library to indicate the relative position of the hotspot area and the target point. The patient's gaze tendency at the target point is evaluated by counting the number of hotspot gazes within the coverage range of the target point. For example, in the hotspot area of ​​a patient, the gaze ratio of the target point coverage range is 35%, while that of the healthy population is 80%. Combined with the distribution data, the patient's gaze hotspot characteristics are quantitatively evaluated, and the spatial overlap rate of the hotspot and the target point is calculated. The spatial overlap rate of the hotspot and the target point can be obtained by spatially overlapping the coverage range of the hotspot area and the target point. First, based on the coordinates and radius of the center point of the hotspot area, the geometric shape of the hotspot area is created using the shapely library in Python. Then, combined with the geometric range of the target point, the intersection function in shapely is used to obtain the overlapping area of ​​the hotspot area and the coverage range of the target point. The proportion of the overlapping area to the coverage area of ​​the target point is counted. For example, if the overlapping area of ​​the hotspot area and the target point is 15 square units, and the total coverage area of ​​the target point is 50 square units, the overlap rate is 30%. This value reflects the degree to which the patient's gaze hotspot area is concentrated on the target point. The patient's gaze hotspot offset characteristics can be further analyzed by comparing multiple sets of data.

[0110] See also Figure 7 Based on the hotspot gaze feature data, the number and position of gazes on the target and distractors are recorded, the gaze path in the task switching phase is analyzed, the behavioral characteristics of the target gaze path are extracted, and the early screening data for Alzheimer's disease is generated. The specific steps for applying it to the early screening of Alzheimer's disease are as follows:

[0111] S601: Based on the hotspot gaze feature data, the gaze point types of the gaze path in the hotspot area are annotated by hierarchical residual analysis, the gaze times and spatial distribution of the target object and the distractor are recorded, and the gaze point distribution characteristic data of the hotspot area is generated according to the distribution characteristics of the gaze points in the hotspot area;

[0112] First, by collecting the patient's gaze path information in the visual scanning task, the fixation points are superimposed with the boundaries of the hotspot area, and ArcGIS is used for spatial analysis to identify the specific locations of the target and distractors. The distribution of fixation points on the target and distractors in the hotspot area is extracted respectively, and then the fixation points are further classified using hierarchical residual analysis. The specific process of hierarchical residual analysis includes: using SPSS to group the fixation points according to the attributes of the target, distractors, and blank areas, setting the model variables as the spatial position and fixation duration of the fixation points, calculating the difference between each group of fixation points and the theoretical model, and quantifying them as residual values. Through the significance test of the residuals, the fixation point areas that deviate from the theoretical distribution are screened out to mark the abnormal distribution in the hotspot area. After the analysis is completed, the results are converted into classification layers, and the differences in the number of fixations and spatial distribution of the target and distractors are displayed through visualization software such as Tableau.

[0113] S602: Based on the hotspot area gaze point distribution characteristic data, record the gaze path in the task dynamic switching stage, and according to the switching order and frequency of the target object and the interference object in the jump path, count the path priority area and jump frequency changes in the task switching stage to generate the gaze path characteristic data in the task dynamic switching stage;

[0114] Firstly, the gaze paths in the dynamic switching stage were extracted, and the path data were spatially visualized using QGIS. The spatial distribution of gaze points and jump paths was converted into intuitive layers. Then, the distribution range of each gaze path in the dynamic switching stage of tasks was clarified by combining the location annotation of hotspot areas. In the analysis process, the time series of the gaze switching sequence of the target and distractors was decomposed by using Python's pandas and matplotlib modules, and the jump events in the path switching were extracted. In order to analyze the changes in the path priority area and jump frequency in the task switching stage, the Origin software was used to perform time series statistics on the gaze point data, and the frequency of the gaze points in different time intervals was associated with the spatial distribution, and the jump frequency map in the hotspot area was generated. Finally, by integrating the gaze path, the switching sequence and frequency characteristics of the target and distractors, and combining the geospatial analysis tools, the gaze path characteristic data in the dynamic switching stage were output.

[0115] S603: Based on the gaze path characteristic data during the dynamic switching stage of the task, the gaze priority area deviation in the patient's gaze hotspot is marked, and the target point focusing frequency, jump direction and amplitude stability, and hotspot area coverage are recorded by task stage. The gaze behavior characteristics in the dynamic switching stage are extracted, and the patient's gaze behavior tendency in subsequent tasks is predicted to generate early screening data for Alzheimer's disease.

[0116] In the analysis of the dynamic switching stage, the deviation of the priority area in the gaze hotspot is marked by combining the distribution of hotspot areas in the patient's gaze path and the gaze records in the task stage. The specific process includes first defining the priority area as the part of the hotspot area with a higher density of gaze points, and then filtering out the gaze points that deviate from the center of the hotspot area according to the gaze path records in the task stage. Next, the frequency of focus on the target point is recorded by task stage, including the number and duration of the patient's gaze on the target point in each stage. The jump direction and amplitude stability are obtained by analyzing the jump events between the patient's gaze points, specifically the range of jump direction changes and the degree of fluctuation of the amplitude value. For the coverage of the hotspot area, the degree of match between the patient's gaze hotspot and the task target point is evaluated based on the spatial overlap rate between the gaze point and the target point in the hotspot area. Combined with these data, the gaze behavior characteristics in the dynamic switching stage are extracted, and finally a complete description of the changing rules of the patient's gaze behavior in different task stages is formed.

[0117] To predict the patient's gaze behavior tendency in the subsequent task, the formula is used:

[0118]

[0119] Calculate the gaze behavior tendency value P t , which measures whether the patient is more inclined to fixate on the target point, distractor, or hot spot in the subsequent task, and usually ranges from 0 to 1;

[0120] Among them, A i represents the weight factor of the fixation point type in the current task stage. This value is determined through experimental analysis and obtained by comparing the fixation preferences of healthy people and patients in the task. i It is the corrected duration of the fixation point, in seconds. The corrected duration is adjusted by combining the patient's standard fixation duration with the actual fixation duration of the task phase. The correction formula is: D' i =DL i +k·(T′ norm -T′ act ), DL i is the original duration of the fixation point, obtained directly from the eye tracking device. The timestamp of the fixation point, T′ norm is the standard fixation time, referring to the historical average fixation time of healthy people, T′ actis the actual fixation time, which is recorded by the patient’s duration at the fixation point. k is the adjustment coefficient, which depends on the experimental correction results. For example, in a certain experiment, the task data of 50 subjects were fitted and corrected. The initial setting k = 1 was found through step-by-step adjustment that when k = 1.25, the model’s prediction error of the focus frequency of the target point was minimized, and the fitting result of the jump amplitude stability was consistent with the experimental observation. Therefore, k = 1.25 was used as the final adjustment coefficient for subsequent analysis. i It represents the jumping frequency of the fixation point, which is the number of times the fixation point jumps from other fixation points to the current point per unit time. This data is obtained through the event record statistics of the fixation path data during the dynamic switching phase of the task. n represents the total number of fixation points recorded in the current task, and is obtained by counting the number of records of the fixation path.

[0121] For example, if a task contains the following fixation point data: Fixation point 1 (target point): A1 = 1.0, DL1 = 3s, T′ norm =4s, T′ act =2s, F1=5; fixation point 2 (interference object): A2=0.5, DL2=2s, T′ norm =3s, T′ act =2.5s, F2=3; fixation point 3 (hot spot): A3=0.8, DL3=4s, T′ norm =5s, T′ act =3s, F3=4.

[0122] Corrected duration calculation:

[0123] For fixation point 1: D'1 = 3 + k·(4-2); for fixation point 2: D'2 = 2 + k·(3-2.5); for fixation point 3: D'3 = 4 + k·(5-3). Assuming k = 0.5, the calculation results are: D'1 = 3 + 0.5·2 = 4s; D'2 = 2 + 0.5·0.5 = 2.25s; D'3 = 4 + 0.5·2 = 5s.

[0124] Substitute into the formula:

[0125]

[0126] The results show that the patient's gaze behavior tendency index in subsequent tasks is higher, close to the distribution of target points and hot spots. Combined with this tendency value, it can further predict the patient's ability to focus on the target point, thereby providing a reference for the evaluation of patients' cognitive function changes in the early screening of Alzheimer's disease.

[0127] The above are only preferred embodiments of the present invention and are not intended to limit the present invention in other forms. Any technician familiar with the profession may use the technical contents disclosed above to change or modify them into equivalent embodiments with equivalent changes and apply them to other fields. However, any simple modification, equivalent change and modification made to the above embodiments based on the technical essence of the present invention without departing from the technical solution of the present invention still falls within the protection scope of the technical solution of the present invention.

Claims

1. A method for early screening of Alzheimer's disease, characterized in that: The following steps are involved: Extract the temporal and spatial coordinates of the fixation point from the visual saccade task of Alzheimer's disease patients, record the directional changes of the jumping behavior in the temporal and spatial coordinates, and generate the jumping behavior distribution data; Based on the jump behavior distribution data, the direction, amplitude and fixation duration of the jump events are arranged in groups and the position changes are recorded, and the fluctuation characteristic diagram of the visual scanning task is generated in combination with the density of the fixation point distribution position; Based on the fluctuation characteristic diagram of the visual scanning task, the distribution of target points and non-target points in the gaze path is recorded, the continuity and jump changes of the gaze behavior are referred to, the return points and bifurcation points are marked, the number and length of path switching are counted, and the path characteristic data of the patient's memory search task is generated; Based on the path characteristic data of the patient's memory search task, according to the relative position change between the target point and the fixation point, the direction and amplitude of the deviation event are recorded and the deviation category is distinguished, so as to generate the path deviation characteristic data in the patient's search task; Collect the data of memory search in healthy people, compare it with the path deviation characteristic data in the patient search task, mark the hot spots and fixation points, analyze the difference between the hot spots distribution of the patient jump path and the control group, and generate hotspot fixation feature data; Based on the hotspot gaze feature data, the number and position of gazes on the target object and the distractor are recorded, the gaze path during the task switching phase is analyzed, the behavioral characteristics of the target gaze path are extracted, and early screening data for Alzheimer's disease is generated for application in early screening of Alzheimer's disease.

2. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: The jump behavior distribution data includes the time record of the fixation point, the multi-dimensional position distribution of the jump direction and the fixation time duration data; the fluctuation characteristic diagram of the visual scanning task includes the jump frequency change distribution of each stage in the fixation trajectory, the dynamic fluctuation range of the jump direction consistency and the trajectory distribution density information of the fixation time period; the path characteristic data of the patient memory search task includes the fixation coverage distribution of the target point, the jump behavior statistics of the non-target point and the number and position relationship of the return points in the path; the path deviation characteristic data in the patient search task includes the distance change range between the target point and the fixation point, the direction angle distribution of the deviation event and the frequency statistics of the deviation amplitude; the hotspot fixation feature data includes the coverage difference of the fixation point hotspot area between patients and healthy people, the distribution density of multi-frequency fixation points and the priority grouping of the fixation hotspot area; the early screening data for Alzheimer's disease includes the stage change distribution of the number of target object fixations in dynamic task switching, the jump frequency distribution of the distractor fixation path and the classification results of the fixation priority.

3. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: The specific steps for extracting the temporal and spatial coordinates of the fixation point from the visual saccade task in which Alzheimer's disease patients participated, recording the directional changes of the jumping behavior in the temporal and spatial coordinates, and generating the jumping behavior distribution data are as follows: Based on the time and space coordinates of the gaze points of Alzheimer's patients recorded by eye movement equipment in visual scanning tasks, the spatial position relationship between the gaze points and the task targets, distractors, and blank areas in the visual scanning tasks is identified, and the association records between the gaze points and various types of areas are classified to generate the spatial area distribution data of the gaze points; Based on the spatial distribution data of the fixation points, the number and duration of fixations on the target object, the distractors and the blank area are counted, the time proportion of the fixation types is classified, the distribution characteristics of the fixation types are marked, and the statistical results of the fixation type time are generated; Based on the temporal statistics of the gaze types, the spatial jump changes between gaze points are recorded, the angular characteristics of the jump direction are calculated, and the change rules of the jump direction are marked according to the time series of the jump events to generate the jump behavior distribution data.

4. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: Based on the jump behavior distribution data, the direction, amplitude and fixation duration of the jump events are grouped and arranged, and the position changes are recorded. Combined with the density of the fixation point distribution position, the specific steps of generating the fluctuation characteristic diagram of the visual scanning task are as follows: Based on the jumping behavior distribution data, the jumping events are time segmented according to the time interval of the gaze point and the change of the coordinates, the fractal dimension of the jumping path is calculated by dividing the coordinate interval into equally spaced grids, the complexity and change law of the jumping behavior in the gaze path are quantified, and the segmented trajectory record of the jumping event is established; Based on the jump event segmented trajectory record, extract the continuous change record of the gaze trajectory of the visual scanning task in time sequence, record the direction, amplitude and gaze time interval of each jump event in the gaze trajectory, and generate a gaze trajectory sequence feature record; Based on the characteristic records of the gaze trajectory sequence, the dynamic changes of the gaze trajectory within the task stage are analyzed, and according to the time variation trend of the jump frequency, the transfer pattern of the gaze point in the gaze concentration area of ​​the visual scanning task and the difference in the distribution of the gaze time within and outside the task stage, a fluctuation characteristic diagram of the visual scanning task is established.

5. The method for early screening of Alzheimer's disease according to claim 4, characterized in that: For calculating the fractal dimension of the jump path, the formula is used: Get the fractal dimension D f , represents the self-similarity of the path at multiple scales; Where ∈ is the grid size, N(∈) is the number of grids covered by the jump path, T represents the temporal intensity of the fixation point in the jump path, log is used to map the number of grids covered by the jump path N(∈) and the change of the grid scale ∈ to the logarithmic space, and log(1 / ∈) is the logarithm of the inverse of the grid scale, reflecting the change of the discretization scale of the path.

6. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: Based on the fluctuation characteristic diagram of the visual scanning task, the distribution of target points and non-target points in the gaze path is recorded, the continuity and jump changes of the gaze behavior are referred to, the return points and bifurcation points are marked, the number and length of path switching are counted, and the specific steps of generating the path characteristic data of the patient's memory search task are as follows: Based on the fluctuation characteristic diagram of the visual scanning task, extract the data source of the memory search link in the visual scanning task by random clustering, record the starting point and the ending point of the fixation path in the memory search link, mark the spatial distribution positions of the target points and non-target points in the fixation path, and generate the target and non-target distribution data of the memory search link; Based on the target and non-target distribution data of the memory search phase, the position changes of the gaze points in the gaze path are extracted according to the time series of the gaze path, the direction changes and frequencies of the gaze point jump events are recorded, the return points and bifurcation points in the gaze path are marked, and the memory search phase gaze path jump characteristic data are generated; Based on the gaze path jumping characteristic data of the memory search link, the number of gaze point switches and the path length in the gaze path are grouped and counted, the gaze behavior is classified according to the density of the intersections of the gaze path, and the stage characteristics of the gaze path are integrated into the overall distribution of the task path to generate the patient's memory search task path characteristic data.

7. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: Based on the path characteristic data of the patient memory search task, the specific steps of recording the direction and amplitude of the deviation event and distinguishing the deviation category according to the relative position change between the target point and the fixation point, and generating the path deviation characteristic data in the patient search task are as follows: Based on the patient's memory search task path characteristic data, the relative spatial position of the target point and the fixation point in the fixation path in the memory search phase is recorded, the direction angle and amplitude of each offset event in the fixation path are marked, and the spatial relationship data between the target point and the fixation point in the fixation path are generated; Based on the spatial relationship data between the target point and the gaze point in the gaze path, the offset events in the gaze path are classified according to the directional consistency and amplitude of the offset events, and multiple distance offsets and directional distribution characteristics of the distance offset events are marked to generate classification data of the offset events in the gaze path; Based on the classification data of the deviation events in the gaze path, the frequency distribution of each category of deviation events is counted, the conventional direction and amplitude range of the deviation events in the gaze path are marked, the statistical characteristics of the deviation categories in the gaze path are summarized, and the path deviation characteristic data in the patient search task is generated.

8. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: The data of the memory search link of the healthy population was collected, compared with the path deviation characteristic data in the patient search task, the hotspot areas and fixation points were marked, and the difference in the hotspot distribution of the patient jump path and the control group was analyzed. The specific steps for generating the hotspot fixation feature data are as follows: Based on the collected data of the memory search phase of the healthy population, the data are compared with the path deviation characteristic data in the patient search task, the distribution of gaze points in the gaze paths of the patients and healthy population is marked, the number of gazes and path changes of the target points and non-target points are recorded in categories, and the comparison results of the gaze paths of the patients and healthy population are generated; Based on the comparison results of the gaze paths of the patients and healthy people, the hot spot area distribution of the patients and the healthy people is counted, the spatial positions and gaze times of the multi-frequency gaze points are marked, the difference between the hot spot distribution in the patients' jump paths and that of the healthy people is analyzed, and the difference data of the gaze hot spot distribution between the patients and the healthy people is generated; Based on the difference data of the gaze hotspot distribution between the patients and the healthy population, the distribution center of the patient's gaze hotspot area is marked, the coverage of the patient's hotspot area and the target point is counted, and the hotspot gaze feature data is generated.

9. The method for early screening of Alzheimer's disease according to claim 1, characterized in that: Based on the hotspot gaze feature data, the number and position of gazes on the target and distractors are recorded, the gaze path in the task switching phase is analyzed, the behavioral characteristics of the target gaze path are extracted, and early screening data for Alzheimer's disease is generated. The specific steps for applying it to early screening for Alzheimer's disease are as follows: Based on the hotspot gaze feature data, the gaze point types of the gaze paths in the hotspot area are annotated by hierarchical residual analysis, the gaze times and spatial distribution of the target object and the distractor are recorded, and the gaze point distribution characteristic data of the hotspot area is generated according to the distribution characteristics of the gaze points in the hotspot area; Based on the hotspot area gaze point distribution characteristic data, the gaze path in the task dynamic switching stage is recorded, and according to the switching order and frequency of the target object and the interference object in the jump path, the path priority area and the jump frequency changes in the task switching stage are counted to generate the gaze path characteristic data in the task dynamic switching stage; Based on the gaze path characteristic data during the dynamic switching stage of the task, the gaze priority deviation in the patient's gaze hotspot is marked, and the target point focusing frequency, jump direction and amplitude stability, and hotspot area coverage are recorded according to the task stage. The gaze behavior characteristics in the dynamic switching stage are extracted, and the patient's gaze behavior tendency in subsequent tasks is predicted to generate early screening data for Alzheimer's disease.

10. The method for early screening of Alzheimer's disease according to claim 9, characterized in that: To predict the patient's gaze behavior tendency in the subsequent task, the formula is used: Calculate the gaze behavior tendency value P t , which measures whether the patient is more inclined to fixate on the target point, distractor, or hot spot in the subsequent task, and usually ranges from 0 to 1; Among them, A i Represents the weight factor of the fixation type in the current task stage, D′ i is the corrected duration of the fixation point, F i is the jump frequency of the fixation point, and n is the total number of fixations recorded in the current task.

Images