Attention assessment method, system and medium

Abstract

The present invention provides an attention assessment method, system, and medium, belonging to the field of biomedical engineering technology, comprising: collecting EEG signals and MRI data of the assessment subject respectively by performing a paradigm requiring different cognitive functions; removing electrooculographic artifacts and reconstructing the EEG signal; extracting EEG signals of each frequency band, analyzing the latency and amplitude changes of the EEG components of the EEG signals of each frequency band, judging whether an attention blink has occurred, and obtaining EEG behavioral representations; using a graph theory parameter algorithm to obtain the communication efficiency and network attribute characteristics between brain regions, using a support vector machine (SVM) to train the SVM based on the communication efficiency and network attribute characteristics between brain regions and their corresponding EEG behavioral representations, and obtaining a trained SVM; and obtaining a comprehensive assessment result of the subject's attention level through the trained SVM based on the subject's MRI data. This method quickly and accurately assesses an individual's attention level, significantly improving the speed and timeliness of the assessment.

Description

Technical Field

[0001] The present invention relates to the field of biomedical engineering technology, and in particular to an attention assessment method, system and medium. Background Art

[0002] Current research shows that attention is not only a core topic in psychology and cognitive science but also a direct product of the human brain's structure and function. Attention involves the coordinated work of multiple brain regions, particularly the prefrontal and parietal lobes. Using electroencephalography (EEG), scientists can observe specific attention-related electrical activity, such as event-related potential (ERP) components, which reflect the brain's response to and processing of external stimuli.

[0003] As people's living standards gradually improve, public concern for mental health and cognitive function is also increasing. In this context, a scientific understanding of attention not only promotes academic development but also holds the key to improving education, work efficiency, and the quality of daily life. Therefore, developing methods to accurately assess and enhance attention has become a key direction for scientific research and application.

[0004] Existing methods for assessing attention, including traditional psychological tests and neurophysiological measurements, typically take a long time to complete and struggle to provide immediate feedback. This is particularly inadequate in environments where rapid decision-making is required, such as in education or workplace monitoring. Furthermore, when measuring attention, existing methods are often limited to a single measurement dimension, such as a single physiological or neuroimaging method, which is susceptible to individual state or environmental factors and has limited accuracy and reliability. Summary of the Invention

[0005] To solve the above problems, the present invention provides an attention assessment method, system and medium. This method quickly and accurately assesses an individual's attention level by real-time analysis of EEG ERP components and brain network graph data generated based on nuclear magnetic resonance imaging, significantly improving the speed and timeliness of the assessment.

[0006] To achieve the above objectives, the present invention provides the following technical solutions.

[0007] An attention assessment method comprises the following steps:

[0008] By having the subject perform different cognitive function-requiring paradigms, the subject's electroencephalogram (EEG) signals and magnetic resonance imaging (MRI) data are collected; the MRI data include T1-weighted imaging, T2-weighted imaging, BOLD imaging, and DTI imaging;

[0009] The EEG signal after denoising is processed by empirical mode decomposition and combined with independent component analysis to remove eye artifacts and reconstruct the EEG signal;

[0010] Extract EEG signals of each frequency band, analyze the temporal characteristics of the EEG components P3b and N2 of the EEG signals of each frequency band, obtain the corresponding latency and amplitude changes of the EEG components P3b and N2, determine whether attention blink occurs, and obtain EEG behavioral representation;

[0011] T1-weighted and T2-weighted imaging data were used to analyze brain anatomy, DTI data were used to analyze nerve fiber pathways, BOLD imaging data were used to construct functional connections to construct brain networks, and graph theory parameter algorithms were used to obtain inter-brain communication efficiency and network attribute characteristics, including small-world network properties.

[0012] Using SVM, the SVM is trained based on the communication efficiency and network attribute characteristics between brain regions and their corresponding EEG behavioral representations to obtain the trained SVM;

[0013] The magnetic resonance imaging data of the subject to be evaluated is obtained, and the comprehensive evaluation result of the subject's attention level is obtained through the trained SVM.

[0014] Preferably, the EEG signal after denoising by using empirical mode decomposition and combined with independent component analysis to remove eye artifacts and reconstruct the EEG signal comprises the following steps:

[0015] The EMD is used to perform empirical mode decomposition on the EEG signal after wavelet transform denoising to obtain multiple intrinsic mode function components (IMFs).

[0016] The obtained multiple intrinsic mode function components IMFs are used as input and processed using independent component analysis to obtain the corresponding independent components ICs;

[0017] Calculate the cross-correlation coefficient between the independent components ICs and the EEG signal, and determine the proportion of the electrooculogram component in the independent components according to the preset noise and signal sorting criteria. If the proportion of the electrooculogram component exceeds the preset threshold, the independent component is set to zero.

[0018] The determined independent components are combined and the signal is reconstructed, that is, the sum of each independent component ICs is obtained to obtain the reconstructed EEG signal.

[0019] Preferably, the method of performing empirical mode decomposition on the EEG signal after wavelet transform denoising by using EMD comprises the following steps:

[0020] Calculate all the maximum and minimum points of the original EEG signal x(t), fit the upper and lower envelopes of all extreme points using the cubic spline function, and calculate the mean of the upper and lower envelopes m1(t);

[0021] Extract the first component h1(t), which is the difference between the original EEG signal x(t) and the envelope mean m1(t):

[0022] h1(t)=x(t)-m1(t)

[0023] Check the new component h1(t) to see if it meets the two conditions of the intrinsic mode function: the difference between the number of extreme points and the number of zero crossings of the IMF function over the entire time series of the signal is less than or equal to 1; the mean of the upper and lower envelopes composed of local maxima and minima over the entire time series of the signal is 0;

[0024] If it is satisfied, h1(t) is taken as the first-order intrinsic mode function IMF1; if it is not satisfied, h1(t) is taken as the original EEG signal, and new components are extracted according to the above steps until the extracted components meet the two conditions of the intrinsic mode function;

[0025] The first first-order intrinsic mode function that meets the eigenmode function condition is denoted as c1(t). The difference between the original EEG signal x(t) and the first-order intrinsic mode function c1(t) is defined as the residual r1(t), which is expressed as follows:

[0026] r1(t)=x(t)-c1(t)

[0027] Take the residual r1(t) as the original signal and repeat the above steps until the final residual r n When (t) is a monotonic function, the EMD signal decomposition is completed;

[0028] At this time, the last eigenmode function is c n (t), then the final remaining variable r n The expression of (t) is:

[0029]

[0030] Preferably, the paradigms include an individual memory capacity paradigm, a visual short-term memory paradigm, and a rapid serial visual presentation paradigm.

[0031] Preferably, the determination of attention blink comprises the following steps:

[0032] Collect EEG signals under the rapid serial visual presentation paradigm; extract EEG signals in each frequency band;

[0033] Analyze the EEG components of EEG signals in each frequency band:

[0034] When the peak of the EEG component P3b after the first stimulation reaches its peak 200-250ms after the second stimulation, attentional blink occurs;

[0035] Among them, when the time interval between two stimulations is 700ms, the probability of attentional blink occurs is reduced.

[0036] Preferably, the EEG signals of each frequency band are extracted by bandpass filtering.

[0037] Preferably, the use of a graph theory parameter algorithm to evaluate the communication efficiency and network properties between brain regions comprises the following steps:

[0038] Determine the nodes of the brain network structure based on T1-weighted imaging, T2-weighted imaging, BOLD imaging and DTI imaging;

[0039] Calculate the clustering coefficient of each node:

[0040]

[0041] Where: Ei represents the actual number of connections between a node and its surrounding nodes in a subgraph, ki represents the number of nodes in the subgraph, and ki(ki-1) / 2 represents the maximum number of connections in the subgraph;

[0042] The clustering coefficient of the entire network is the average of the clustering coefficients Ci of each node:

[0043]

[0044] Where: Cp represents the clustering coefficient of the network, n represents the number of all nodes in the network;

[0045] Calculate the characteristic path length:

[0046] The average shortest path of a node i is defined as:

[0047]

[0048] Where min{Lij} represents the shortest absolute path length between two nodes i and j, and the average shortest path length refers to the minimum number of edges connecting two nodes. The characteristic path length is defined as:

[0049]

[0050] Calculate the standard clustering coefficient:

[0051]

[0052] in, is the clustering coefficient of a random network under the same conditions;

[0053] Calculate the standard characteristic path length:

[0054]

[0055] in, It represents the characteristic path length of a random network under the same conditions;

[0056] Then the small-world property is expressed as:

[0057]

[0058] An attention assessment system, comprising:

[0059] processor;

[0060] a memory having stored thereon a computer program executable on the processor;

[0061] Wherein, when the computer program is executed by the processor, the steps of the attention evaluation method are implemented.

[0062] A computer-readable storage medium stores a data processing program, which implements the steps of the attention assessment method when executed by a processor.

[0063] Beneficial effects of the present invention:

[0064] The present invention proposes an attention assessment method, system and medium. This method significantly shortens the time of attention assessment by real-time analysis of ERP and MRI data, and adapts to modern application environments that require rapid decision-making and response. By combining the time sensitivity of ERP and the spatial detail of MRI, the present invention can more comprehensively and accurately assess the level of attention, especially when considering the complex interactions of brain network structure and function. During the evaluation process, this method only collects MRI data for judgment, optimizes the data processing and analysis process, reduces dependence on high-end equipment, and makes the present invention easier to deploy and use in non-professional environments. The method provides a multi-dimensional attention assessment including the functional status of brain networks, helping users to fully understand the individual's performance under different cognitive tasks, which is very valuable for applications such as personalized education plans and work efficiency optimization. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 is a flow chart of an attention assessment method according to an embodiment of the present invention;

[0066] Figure 2 is a schematic diagram of an individual memory capacity paradigm according to an embodiment of the present invention;

[0067] Figure 3 is a schematic diagram of a visual short-term memory paradigm according to an embodiment of the present invention;

[0068] Figure 4is a schematic diagram of a rapid serial visual presentation paradigm according to an embodiment of the present invention;

[0069] Figure 5 4 is a flow chart of the EMD-ICA algorithm according to an embodiment of the present invention;

[0070] Figure 6 A is a time series characteristic diagram of the EEG response component of attentional blink in an embodiment of the present invention;

[0071] Figure 6 B is a time series characteristic diagram of the EEG response component when no attention blink occurs in the embodiment of the present invention;

[0072] Figure 6 C is the recognition accuracy of the embodiment of the present invention at T1 and T2. DETAILED DESCRIPTION

[0073] In order to make the purpose, technical solutions and advantages of the present invention more clearly understood, the present invention will be further described in detail below with reference to 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.

[0074] Example 1

[0075] Recent advances in medical imaging technology, particularly magnetic resonance imaging (MRI), have provided crucial tools for understanding brain structure and function. T1- and T2-weighted MRI images clearly depict brain anatomy, while functional MRI (fMRI), which reflects the BOLD (Blood Oxygen Level Dependent) signal, is used to observe dynamic changes in brain activity. Furthermore, diffusion tensor imaging (DTI) reveals the orientation and connectivity of neural fibers within the brain, providing a unique perspective for studying brain network structure.

[0076] Event-related potential (ERP) technology uses electroencephalography (EEG) to record the brain's response to specific events, providing precise temporal information about the brain's electrical activity. Combined with ERP technology, researchers can analyze in detail an individual's cognitive processes during specific tasks. Meanwhile, brain network graph theory, applied to the analysis of the brain's structural and functional networks, reveals the complex interactions between brain regions. By evaluating properties such as small-world networks, scientists can better understand the brain's ability to efficiently process information, which is particularly important for assessing and understanding attention.

[0077] By comprehensively utilizing ERP technology and brain network graph analysis of nuclear magnetic resonance, the present invention aims to provide a new, multi-dimensional attention assessment method that can not only quickly and accurately assess an individual's attention level, but also provide dynamic brain function data under different physiological and cognitive states. The development of this method is applied to multiple fields such as education, occupational health, and personal cognitive function improvement, providing a comprehensive, fast and accurate attention assessment method to meet the needs of modern society for individual cognitive function assessment and promote the optimization of individual performance in various environments. Figure 1 As shown, Figure 1 This is a flowchart of the attention assessment method, which specifically includes the following steps:

[0078] S1: The subject is assessed by performing a paradigm requiring different cognitive functions, and the EEG signals and MRI data of the subject are collected respectively; the MRI data include T1-weighted imaging, T2-weighted imaging, BOLD imaging, and DTI imaging.

[0079] S2: The EEG signal after denoising is processed using empirical mode decomposition, and the independent component analysis method is combined to remove eye artifacts and reconstruct the EEG signal.

[0080] S3: Extract EEG signals of each frequency band, analyze the temporal characteristics of the EEG components P3b and N2 of the EEG signals of each frequency band, obtain the latency and amplitude changes corresponding to the EEG components P3b and N2, determine whether attention blink occurs, and obtain EEG behavioral representation.

[0081] S4: Use T1-weighted imaging and T2-weighted imaging data to analyze the anatomical structure of the brain, use DTI data to analyze the paths of nerve fibers, apply BOLD imaging data to functionally connect and construct brain networks, and use graph theory parameter algorithms to obtain communication efficiency and network attribute characteristics between brain regions, including small-world network attributes.

[0082] S5: Using SVM, the SVM is trained according to the communication efficiency and network attribute characteristics between brain regions and their corresponding EEG behavioral representations to obtain a trained SVM.

[0083] S6: Obtain the magnetic resonance imaging data of the subject to be evaluated, and obtain a comprehensive evaluation result of the attention level of the subject to be evaluated through the trained SVM.

[0084] 1. Data collection:

[0085] The data collection phase of this invention primarily involves two types of data: electroencephalogram (EEG) data and magnetic resonance imaging (MRI) data. This data is collected through paradigms with varying cognitive function requirements to comprehensively assess the subject's attention and cognitive processing abilities. Each paradigm is described in detail below, including the individual memory capacity paradigm, the visual short-term memory paradigm, and the rapid serial visual presentation paradigm.

[0086] (1) Individual Memory Capacity Paradigm:

[0087] Purpose: To assess the subject's ability to process and retain information.

[0088] Methods: Individual memory capacity (N-back) paradigm ( Figure 2 ) is often used in studies related to cognitive function. It requires participants to compare a stimulus that has just appeared with the nth stimulus that preceded it. Memory load is manipulated by controlling the number of stimuli between the current stimulus and the target stimulus. When n = 1, participants are asked to compare the current stimulus with the stimulus immediately preceding it; when n = 2, they are asked to compare the current stimulus with the stimulus one position ahead of it; when n = 3, they are asked to compare the current stimulus with the stimulus two positions ahead of it, and so on, to achieve different levels of task difficulty. The advantage of this paradigm is that tasks can be designed to impose a continuous, variable-parameter memory load on working memory, while other tasks need to remain constant.

[0089] This example used the classic N-back paradigm for data collection. The letter stimuli used in this experiment were b, B, d, D, g, G, p, P, t, T, v, and V. These letters are closed syllables, and the interference of syllables and uppercase letters is eliminated, which is more conducive to exploring patients' memory capacity in this example. During data collection, the 1-back and 2-back tasks were performed alternately to avoid a certain degree of practice effect.

[0090] (2) Visual short-term memory paradigm:

[0091] Purpose: To measure the subject's ability to process and recall visual information in a short period of time.

[0092] Methods: Visual short-term memory (VSTM) is a short-term storage of visual information after the visual stimulus disappears. It lasts only a few seconds and has a very limited capacity, generally believed to be only 3-4 units. This experiment uses a dual-task VSTM paradigm ( Figure 3) assesses the static storage capacity of the subjects' working memory (i.e., working memory capacity). The VSTM paradigm first presents the subjects with four horizontally arranged Arabic numerals (1, 2, 3, 4; randomly ordered in the experiment) for 500ms. After a 500ms fixation cross sign, five colored circles of different colors are displayed for 800ms, followed by a 1500ms fixation cross sign. Afterwards, a picture with only one colored circle appears until the subject determines whether the circle is the same color as the circle in the corresponding position of the five different colored circles, or it continues to be displayed for 3000ms. Finally, four horizontally arranged Arabic numerals (1, 2, 3, 4; randomly ordered in the experiment) are presented and remain displayed for 3000ms, or until the subject determines whether it is the same as the initial Arabic numeral.

[0093] (3) Rapid serial visual presentation paradigm ( Figure 4 ):

[0094] Purpose: To examine the subjects' ability to quickly identify and process information in high-speed information flow.

[0095] Methods: In the experimental stimulus stream, each stimulus image containing numbers and letters was displayed for 50ms, and a blank image was inserted between each stimulus image for 50ms to prevent the visual persistence effect. At the same time, 7-14 letter stimuli were randomly presented before the T1 stimulus to prevent the subjects from learning the pattern of T1's appearance. Given that the human attentional blink phenomenon often occurs when the task of recognizing two stimuli separated by less than 500ms, this example collected behavioral data from subjects in two situations where the interval between T1 and T2 was 3 letters (lg4) or 6 letters (lg7), that is, 300ms and 600ms, to test the attentional blink phenomenon. This experiment included a total of 40 trials.

[0096] T2 recognition accuracy (hit rate) = N2 / (N1+N2)

[0097] False alarm rate = 100% - hit rate

[0098] Wherein, N1 represents the number of times T1 is correctly identified, and N2 represents the number of times T2 is correctly identified.

[0099] By combining these paradigms, the present invention can assess a subject's attention and cognitive abilities across multiple dimensions, including memory, information processing speed, and visual processing. This multifaceted cognitive function assessment provides a solid scientific foundation for the accurate measurement of attention.

[0100] (4) EEG data collection

[0101] While using the above paradigm, EEG data were collected using a 64-channel EEG device. This allows for precise monitoring and recording of EEG activity during various cognitive tasks, particularly ERP components closely related to cognition and attention.

[0102] (5) NMR data acquisition:

[0103] At the same time, MRI technology was used to collect structural and functional imaging data. These data included T1- and T2-weighted imaging for observing brain anatomical structure, and BOLD and DTI imaging for analyzing brain functional activity and nerve fiber pathways, providing a basis for brain network graph analysis.

[0104] 2. ERP Component Analysis

[0105] After EEG signal acquisition is complete, data preprocessing is crucial for feature extraction and classification. The multi-channel, continuous raw EEG data obtained by the EEG acquisition system is excessively redundant, so extracting valid data segments becomes the primary task of preprocessing. Based on the experimental process, the EEG data of all subjects was collated and analyzed, and the EEG data for the motor imagery task segment was extracted based on the experimental markers.

[0106] EEG signals are weak, nonlinear, non-stationary, and random. They are susceptible to external interference during signal acquisition. This interference primarily includes noise from human physiological activities, such as electrocardiogram (ECG), electrooculogram (EOG), and electromyography (EMG), as well as electromagnetic and power frequency interference from electrical devices and power supplies, such as mobile phones and computers. Interference and noise complicate signal feature extraction and affect classification and recognition accuracy.

[0107] Specifically, such as Figure 5 As shown in the figure, according to the principles of EMD and ICA algorithms, the main steps of the EMD-ICA algorithm to remove electrooculogram artifacts are as follows:

[0108] Step 1: Use EMD to perform empirical mode decomposition on the EEG signal after wavelet transform denoising to obtain multiple intrinsic mode function components (IMFs).

[0109] Step 2: Take the obtained multiple intrinsic mode function components IMFs as input and process them using ICA to obtain the corresponding independent components ICs.

[0110] Step 3: Calculate the cross-correlation coefficient between the independent components ICs and the source signal. According to the noise and signal separation criteria selected in this paper, determine the proportion of the electrooculogram component in the independent components. If the proportion of the electrooculogram component is high, set the independent component to zero.

[0111] Step 4: Combine the determined independent components and reconstruct the signal, which is the sum of the independent components ICs.

[0112] Among them, using EMD to perform empirical mode decomposition on the EEG signal after wavelet transform denoising, includes the following steps:

[0113] Calculate all the maximum and minimum points of the original EEG signal x(t), fit the upper and lower envelopes of all extreme points using the cubic spline function, and calculate the mean of the upper and lower envelopes m1(t);

[0114] Extract the first component h1(t), which is the difference between the original EEG signal x(t) and the envelope mean m1(t):

[0115] h1(t)=x(t)-m1(t)

[0116] Check the new component h1(t) to see if it meets the two conditions of the intrinsic mode function: the difference between the number of extreme points and the number of zero crossings of the IMF function over the entire time series of the signal is less than or equal to 1; the mean of the upper and lower envelopes composed of local maxima and minima over the entire time series of the signal is 0;

[0117] If it is satisfied, h1(t) is taken as the first-order intrinsic mode function IMF1; if it is not satisfied, h1(t) is taken as the original EEG signal, and new components are extracted according to the above steps until the extracted components meet the two conditions of the intrinsic mode function;

[0118] The first first-order intrinsic mode function that meets the eigenmode function condition is denoted as c1(t). The difference between the original EEG signal x(t) and the first-order intrinsic mode function c1(t) is defined as the residual r1(t), which is expressed as follows:

[0119] r1(t)=x(t)-c1(t)

[0120] Take the residual r1(t) as the original signal and repeat the above steps until the final residual r n When (t) is a monotonic function, the EMD signal decomposition is completed;

[0121] At this time, the last eigenmode function is c n (t), then the final remaining variable r n The expression of (t) is:

[0122]

[0123] The main principle of the ICA algorithm is to separate each independent component from the blind source signal by finding a separation matrix. It is a major method for blind source separation and feature extraction of biological signals. Its model can be expressed as:

[0124] X=A*S

[0125] Where X=[x1,x2,…,x n ] T and S=[s1,s2,…,s m ] T are the observation signal and source signal collected on m channels respectively. A represents the unknown constant linear mixing matrix. The main purpose of the ICA algorithm is to find a separation matrix W to separate the source signal S from the observation signal X. Its expression is:

[0126] y=Wx

[0127] Where y=[y1,y2,…,y m ] represents the approximation signal of the source signal S.

[0128] The EEG data were preprocessed, including filtering and artifact removal, to extract clear ERP signals.

[0129] Electroencephalogram (EEG) has a high temporal resolution and can be used to explore the temporal evolution mechanism of attentional blink. The EEG signals recorded when the subjects performed the attentional blink behavioral paradigm task were preprocessed by noise reduction, artifact removal, etc., and the EEG signals of various frequency bands such as δ (1-4Hz), θ (4-8Hz), α (8-13Hz), β (13-30Hz) and γ (30-45Hz) were extracted by bandpass filtering. By analyzing the behavioral data of 150 people combined with the EEG component timing results, the key neural response timing characteristics of attentional blink and the key parameters of three types of timing feature windows were revealed: (1) When the N2 EEG component caused by T2 overlaps with the P3b component generated by the late processing of T1, that is, P3b reaches its peak 200-250ms after T2, T2 cannot be recognized, and attentional blink occurs ( Figure 6 A); (2) When the N2 component triggered by T2 does not overlap with the P3b component processed later in T1, that is, P3b reaches its peak 100-200ms after T2, T2 can be identified and no attentional blink occurs ( Figure 6B). (3) The recognition accuracy of T2 can be improved by extending the time interval between T1 and T2. When the interval between T1 and T2 is 400ms, the probability of attentional blink is higher, the recognition accuracy of T2 is lower, and it is difficult to be recognized. When the interval between T1 and T2 is 700ms, the probability of attentional blink is lower, the recognition accuracy of T2 is higher, and T2 is easier to be recognized. Figure 6 C). By analyzing the neural response signals of attentional blink using EEG, the present invention proposes and discovers the key timing characteristics of attentional blink, namely, whether an attentional blink occurs is primarily determined by the peak occurrence time of the P3b EEG component responsible for T1 target processing. If P3b reaches its peak 200-250ms after T2, it indicates that the EEG component responsible for T1 processing occupies more resources and competes for resources with the N2 EEG component responsible for T2 target processing. The resources required for T2 target processing are competed for, making it difficult to identify. Conversely, if the P3b responsible for T1 target processing completes T1 processing in a shorter time (reaching its peak 100-200ms after T2) and does not overlap with the N2 EEG component responsible for T2 target processing, then sufficient resources are allocated to the processing of the T2 target, and the T2 target can be identified. This example also explored the crucial regulatory node between T1 and T2, namely, 400ms, where resource competition between them is likely to occur, preventing T2 from being recognized. However, when the interval is greater than 700ms, resource competition ceases, allowing T2 target recognition. This allows for millisecond-level analysis of the perceptual-cognitive information flow. The analysis focused on the P3b and N2 waveforms, two ERP components believed to be closely related to attention regulation. Attention levels were assessed by analyzing their latency and amplitude changes.

[0130] 3. Brain network graph analysis of MRI data:

[0131] T1 and T2 imaging data are used to analyze brain anatomy, and DTI data are used to study the pathways of nerve fibers. Functional connectivity analysis using BOLD imaging data assesses inter-brain communication efficiency and network properties, particularly the presence of small-world properties, which contributes to understanding the role of brain networks in efficient information processing.

[0132] (1) Data preprocessing

[0133] Structural data preprocessing

[0134] T1 and T2 structural data preprocessing was performed using the Freesurfer software platform, developed jointly by the MIT Institute of Health Sciences and Technology and Massachusetts General Hospital. This software integrates a series of algorithms for quantifying brain function, structural connectivity, and properties. First, DICOM data were converted to NIFTI format and oriented to align with the MNI152 standard space. Using the fully automated cortical reconstruction command in Freesurfer, multiple steps were performed, including nonuniform intensity normalization, skull removal, and white matter segmentation, to obtain detailed anatomical parameters such as cortical thickness, area, and gray matter volume.

[0135] Functional data preprocessing

[0136] BOLD functional data preprocessing was performed on the FSL platform. The processing workflow included multiple steps, including excluding images at the starting time point, slice timing correction, and head motion correction. Specifically, all functional MRI data were coregistered to the MNI152 standard template to standardize the analysis process. Furthermore, coregistering, tissue segmentation, and filtering procedures enhanced data quality, ensuring the accuracy and reliability of the analysis results.

[0137] DTI data preprocessing

[0138] DTI data preprocessing was also performed on the FSL software platform. After conversion from the original DICOM format to the NIFTI format, the data were corrected for eddy current and motion artifacts. Then, the FSL Diffusion Toolkit was used to fit the data to a tensor model, obtaining parameters such as the anisotropy index (FA) and mean diffusivity (MD).

[0139] (2) NMR data analysis

[0140] This patent utilizes structural and functional imaging data through a series of precise data preprocessing and analysis steps to deeply assess the communication efficiency and network characteristics between brain regions, with a particular focus on small-world network properties. Small-worldness σ is a quantitative parameter that compares brain networks with random networks. The normal brain generally exhibits a small-world structure, achieving a stable balance between the needs of localized, proprietary information processing and the needs of integrated processing across the entire network. The specific analysis process is as follows:

[0141] Determine the nodes of the brain network structure based on T1-weighted imaging, T2-weighted imaging, BOLD imaging and DTI imaging;

[0142] Calculate the clustering coefficient of each node:

[0143]

[0144] Where: Ei represents the actual number of connections between a node and its surrounding nodes in a subgraph, ki represents the number of nodes in the subgraph, and ki(ki-1) / 2 represents the maximum number of connections in the subgraph;

[0145] The clustering coefficient of the entire network is the average of the clustering coefficients Ci of each node:

[0146]

[0147] Where: Cp represents the clustering coefficient of the network, n represents the number of all nodes in the network;

[0148] Calculate the characteristic path length:

[0149] The average shortest path of a node i is defined as:

[0150]

[0151] Where min{Lij} represents the shortest absolute path length between two nodes i and j, and the average shortest path length refers to the minimum number of edges connecting two nodes. The characteristic path length is defined as:

[0152]

[0153] Calculate the standard clustering coefficient:

[0154]

[0155] in, is the clustering coefficient of a random network under the same conditions;

[0156] Calculate the standard characteristic path length:

[0157]

[0158] in, It represents the characteristic path length of a random network under the same conditions;

[0159] Then the small-world property is expressed as:

[0160]

[0161] By analyzing ERP and MRI data in real time, the present invention significantly shortens the time for attention assessment and adapts to modern application environments that require rapid decision-making and response. By combining the temporal sensitivity of ERP and the spatial detail of MRI, the present invention can more comprehensively and accurately assess attention levels, especially when considering the complex interactions of brain network structure and function. This method optimizes the data processing and analysis process, reduces dependence on high-end equipment, and makes the present invention easier to deploy and use in non-professional environments. The method provides a multi-dimensional attention assessment, including the functional state of brain networks, to help users fully understand the individual's performance under different cognitive tasks, which is very valuable for applications such as personalized education plans and work efficiency optimization.

[0162] The above are only preferred embodiments of the present invention and are not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for assessing attention, characterized in that: The following steps are involved: By having the subject perform different cognitive function-requiring paradigms, the subject's electroencephalogram (EEG) signals and magnetic resonance imaging (MRI) data are collected; the MRI data include T1-weighted imaging, T2-weighted imaging, BOLD imaging, and DTI imaging; The EEG signal after denoising is processed by empirical mode decomposition and combined with independent component analysis to remove eye artifacts and reconstruct the EEG signal; Extract EEG signals of each frequency band, analyze the temporal characteristics of the EEG components P3b and N2 of the EEG signals of each frequency band, obtain the corresponding latency and amplitude changes of the EEG components P3b and N2, determine whether attention blink occurs, and obtain EEG behavioral representation; T1-weighted and T2-weighted imaging data were used to analyze brain anatomy, DTI data were used to analyze nerve fiber pathways, BOLD imaging data were used to construct functional connections to construct brain networks, and graph theory parameter algorithms were used to obtain inter-brain communication efficiency and network attribute characteristics, including small-world network properties. Using SVM, the SVM is trained based on the communication efficiency and network attribute characteristics between brain regions and their corresponding EEG behavioral representations to obtain the trained SVM; The magnetic resonance imaging data of the subject to be evaluated is obtained, and the comprehensive evaluation result of the subject's attention level is obtained through the trained SVM.

2. The attention assessment method according to claim 1, characterized in that The EEG signal after denoising by using empirical mode decomposition is combined with independent component analysis to remove eye artifacts and reconstruct the EEG signal, which includes the following steps: The EMD is used to perform empirical mode decomposition on the EEG signal after wavelet transform denoising to obtain multiple intrinsic mode function components (IMFs). The obtained multiple intrinsic mode function components IMFs are used as input and processed using independent component analysis to obtain the corresponding independent components ICs; Calculate the cross-correlation coefficient between the independent components ICs and the EEG signal, and determine the proportion of the electrooculogram component in the independent components according to the preset noise and signal sorting criteria. If the proportion of the electrooculogram component exceeds the preset threshold, the independent component is set to zero. The determined independent components are combined and the signal is reconstructed, that is, the sum of each independent component ICs is obtained to obtain the reconstructed EEG signal.

3. The attention assessment method according to claim 2, characterized in that The method of performing empirical mode decomposition on the EEG signal after wavelet transform denoising by using EMD comprises the following steps: Calculate all the maximum and minimum points of the original EEG signal x(t), fit the upper and lower envelopes of all extreme points using the cubic spline function, and calculate the mean of the upper and lower envelopes m1(t); Extract the first component h1(t), which is the difference between the original EEG signal x(t) and the envelope mean m1(t): h1(t)=x(t)-m1(t) Check the new component h1(t) to see if it meets the two conditions of the intrinsic mode function: the difference between the number of extreme points and the number of zero crossings of the IMF function over the entire time series of the signal is less than or equal to 1; the mean of the upper and lower envelopes composed of local maxima and minima over the entire time series of the signal is 0; If it is satisfied, h1(t) is taken as the first-order intrinsic mode function IMF1; if it is not satisfied, h1(t) is taken as the original EEG signal, and new components are extracted according to the above steps until the extracted components meet the two conditions of the intrinsic mode function; The first first-order intrinsic mode function that meets the eigenmode function condition is denoted as c1(t). The difference between the original EEG signal x(t) and the first-order intrinsic mode function c1(t) is defined as the residual r1(t), which is expressed as follows: r1(t)=x(t)-c1(t) Take the residual r1(t) as the original signal and repeat the above steps until the final residual r n When (t) is a monotonic function, the EMD signal decomposition is completed; At this time, the last eigenmode function is c n (t), then the final remaining variable r n The expression of (t) is:

4. The attention assessment method according to claim 1, wherein: The paradigms include an individual memory capacity paradigm, a visual short-term memory paradigm, and a rapid serial visual presentation paradigm.

5. The attention assessment method according to claim 4, characterized in that: The determination of attention blink comprises the following steps: Collect EEG signals under the rapid serial visual presentation paradigm; extract EEG signals in each frequency band; Analyze the EEG components of EEG signals in each frequency band: When the peak of the EEG component P3b after the first stimulation reaches its peak 200-250ms after the second stimulation, attentional blink occurs; Among them, when the time interval between two stimulations is 700ms, the probability of attentional blink occurs is reduced.

6. The attention assessment method according to claim 5, characterized in that: The EEG signals in each frequency band are extracted by bandpass filtering.

7. The attention assessment method according to claim 1, characterized in that: The method of using a graph theory parameter algorithm to evaluate the communication efficiency and network properties between brain regions includes the following steps: Determine the nodes of the brain network structure based on T1-weighted imaging, T2-weighted imaging, BOLD imaging and DTI imaging; Calculate the clustering coefficient of each node: Where: Ei represents the actual number of connections between a node and its surrounding nodes in a subgraph, ki represents the number of nodes in the subgraph, and ki(ki-1) / 2 represents the maximum number of connections in the subgraph; The clustering coefficient of the entire network is the average of the clustering coefficients Ci of each node: Where: Cp represents the clustering coefficient of the network, n represents the number of all nodes in the network; Calculate the characteristic path length: The average shortest path of a node i is defined as: Where min{Lij} represents the shortest absolute path length between two nodes i and j, and the average shortest path length refers to the minimum number of edges connecting two nodes. The characteristic path length is defined as: Calculate the standard clustering coefficient: in, is the clustering coefficient of a random network under the same conditions; Calculate the standard characteristic path length: in, It represents the characteristic path length of a random network under the same conditions; Then the small-world property is expressed as:

8. An attention assessment system, characterized in that: The system comprises: processor; a memory having stored thereon a computer program executable on the processor; When the computer program is executed by the processor, the steps of the attention assessment method according to any one of claims 1 to 7 are implemented.

9. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a data processing program, which, when executed by a processor, implements the steps of the attention evaluation method according to any one of claims 1 to 7.

Images