Method, system and medium for measuring attention and memory processing of children and adolescents based on dynamic brain network

Abstract

The application relates to a kind of based on dynamic brain network's children and adolescents attention, memory processing measurement method, belong to brain cognitive processing mechanism research technical field, comprising the following steps: S1: let the measured person according to n-back test requirement and method self-practice certain time, and carry out n-back test after a period of time interval;S2: in the n-back test process, the brain electrical data group of the measured person is collected for pretreatment, and the brain network sequence of the brain electrical data group after pretreatment is obtained based on MS-sPDC;S3: the evaluation feature of the brain electrical data group after pretreatment is obtained from the brain network sequence, and the detection report of the attention, memory processing of the measured person is obtained based on evaluation feature.

Description

Technical Field

[0001] This invention belongs to the field of brain cognitive processing mechanism research technology, and relates to a method, system and medium for measuring attention and memory processing in children and adolescents based on dynamic brain networks. Background Technology

[0002] Attention and memory, as core components of intelligence, are closely linked. Attention is a prerequisite for memory; only information that is noticed can enter the brain for further processing. Research on the relationship between attention and working memory has found that when distractions are present, the selective mechanisms of attention can protect working memory from interference. This is because working memory has a limited capacity, and attention can filter out information irrelevant to the target during the awareness stage, thereby reducing the occupation of working memory capacity by irrelevant information, lowering the memory load, and improving memory efficiency.

[0003] Working memory is an information processing system with limited energy, capable of temporarily processing and storing information. It is a crucial foundation for individuals to handle complex cognitive activities. The concept of working memory was introduced to supplement and refine the concept of short-term memory, revealing the psychological process by which information is retained in the brain for a few seconds or minutes after entering. In daily life, working memory plays a vital role in cognitive activities such as learning and calculation. For example, when adding multiple numbers, we need to remember the sum of the first two numbers and then use this result to add the subsequent numbers; this process involves the use of working memory.

[0004] In summary, attention and working memory are closely linked and interact. Selective and inhibitory mechanisms of attention promote working memory capacity, while the content and breadth of working memory also influence attention. Since working memory is a system with limited capacity, childhood and adolescence are critical periods for attention development. Improving the attention levels of children and adolescents can effectively enhance the utilization rate of working memory, reduce memory load, and thus improve memory efficiency. However, there is currently a lack of research in this area. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a method, system, and medium for measuring attention and memory processing in children and adolescents based on dynamic brain networks.

[0006] To achieve the above objectives, the present invention provides the following technical solution:

[0007] A method for measuring attention and memory processing in children and adolescents based on dynamic brain networks includes the following steps:

[0008] S1: Have the test subject practice the n-back test requirements and methods independently for a certain period of time, and then conduct the n-back test after a certain interval;

[0009] S2: During the n-back test, the subject's EEG data set is collected, preprocessed, and the brain network sequence of the preprocessed EEG data set is obtained based on MS-sPDC.

[0010] S3: Obtain the evaluation features of the preprocessed EEG data set from the brain network sequence, and obtain the test report of the subject's attention and memory processing based on the evaluation features.

[0011] Furthermore, the preprocessing of the EEG data set collected from the test subject in step S2 specifically includes: performing mean removal processing, trend removal processing, and noise removal processing on the EEG data set in sequence.

[0012] Furthermore, step S2, which involves obtaining the brain network sequence of the preprocessed EEG data set based on MS-sPDC, specifically includes the following steps:

[0013] The Laplace distribution-based adaptive dynamic brain network estimation method describes the dynamic connectivity changes between brain regions. Given D joint time series, the dynamic driving relationship at time k is represented by a D-dimensional time-varying multiple autoregressive model (tv-MVAR).

[0014]

[0015] Where, β ij (σ,k),σ=1,2,…,p;i=1,2,…,D;j=1,2…,D。D represents the number of time series; p represents the model order. Formula (1) simplifies to the following form:

[0016] z k =H k x k +ε k (46)

[0017] Where zk=[X1(k),…,X j (k),…,X D (k)] T ∈R D×1 ;x k =[β 11 (1,k),…,β 1D (1,k),…,β 11 (p,k),…,(p,k),…,β 1D (p,k),…,βD1 (1,k),…,β DD (1,k),…,β D1 (p,k),…,β DD (p,k)] T ∈R (D×p×D)×1 ; Observation matrix H k The construction form is represented as:

[0018]

[0019] The tv-MVAR model can be represented as the following state-space model:

[0020] x k =F k-1 x k-1 +w k-1 (48)

[0021] z k =H k x k +ε k (49)

[0022] Wherein, formula (4) is the state equation; formula (5) is the observation equation; k represents the discrete time point; x k ∈R D×p×D Represents the state vector; z k F represents the observation vector; k-1 H represents the state transition matrix; k Represents the observation matrix; w k ∈R D×p×D and ε k ∈R D ×1 These represent state noise and observation noise, respectively; the parameter update process of the state-space model is divided into a single-step prediction process and a parameter update process.

[0023] The single-step prediction process is as follows:

[0024]

[0025]

[0026] The parameter update process is as follows:

[0027]

[0028] P k|k =(IK k H k )P k|k-1 (54)

[0029] When the model coefficients x at time k are estimatedk Then, the time-frequency representation of directed connectivity is calculated using sPDC. The corresponding formula for sPDC estimation is as follows:

[0030]

[0031] Among them, Ξ f,k State x at time k and frequency f k The frequency domain representation of is obtained through the Z-transform:

[0032]

[0033] Where τ represents the imaginary unit; based on the calculation results of sPDC, the causal relationship between node j and node i in the frequency band [f1,f2] at time k is expressed as follows:

[0034]

[0035] Introducing a t-distribution to calculate the threshold achieves a sparsity effect in the network:

[0036] thre=T(1-p,n·(n-1)) (58)

[0037] Where T(1-p,v) represents the cumulative t-distribution probability density function with v degrees of freedom; 1-p is the confidence level; for L k Standardize the network by ignoring the diagonal elements at time k:

[0038]

[0039] Where △ represents any combination between points [1, n], but the diagonal elements in the network matrix are ignored; if Lsk(i,j) is less than the threshold, then L k The corresponding position (i,j) is 0; the causal network at time k is sparse based on threshold judgment;

[0040] Introducing Bayesian inference into the state-space model, the probability density functions of the state values ​​and observations under Bayesian inference are expressed as follows:

[0041]

[0042] P k|k-1 and R k The prior distribution function p(P) k|k-1 |z 1:k-1 ) and p(R k |z 1:k-1 This can be represented as an inverse-Wishart (IW) distribution.

[0043] p(P k|k-1 |z1:k-1 ) = IW(P k|k-1 ;t k|k-1 ,T k|k-1 ), p(R k |z 1:k-1 ) = IW(R k ;u k|k-1 U k|k-1 (61)

[0044] Among them, t k|k-1 and T k|k-1 They represent p(P) k|k-1 |z 1:k-1 The degrees of freedom parameters and inverse scaling matrix of u; k|k-1 and U k|k-1 They represent p(R) k |z 1:k-1 The degrees of freedom parameters and inverse scaling matrix of p(P); based on the characteristics of the inverse-Vichtht distribution, p(P k|k-1 |z 1:k-1 The expectation of ) is expressed as:

[0045] E[P k|k-1 ] = T k|k-1 / (t k|k-1 -m-1) (62)

[0046] Let t k|k-1 =m+τ+1, where τ≥0 is the tuning parameter; therefore, formula (18) can be further expressed as T k|k-1 =τP k|k-1 Introducing Bayesian inference into the covariance matrix R of observation noise k The expression format is:

[0047] p(R k |z 1:k-1 )=∫p(R k |R k-1 )p(R k-1 |z 1:k-1 )dR k-1 (63)

[0048] Wherein, p(R) k-1 |z 1:k-1 ) follows an inverse-Vichter distribution, i.e., p(R) k-1 |z 1:k-1 ) = IW(R k-1 ;u k-1|k-1 U k-1|k-1 Prior parameter u k|k-1 and U k|k-1 The substitution is represented as u k|k-1 =ρ(u k-1|k-1 -n-1)+n+1,Uk|k-1 =ρU k-1|k-1 ρ∈[0,1] is the forgetting factor;

[0049] By introducing a multivariate Laplace distribution, a Bayesian robust representation is constructed under heavy-tailed measurement noise (HTMN). Specifically, HTMN is modeled as a multivariate Laplace distribution in the following form:

[0050]

[0051] Among them, ML(z) k H k x k ,,R k ) represents z k The multidimensional Laplace distribution function, K·(·) denotes the second-order Bessel function of order n / 2-1; p(z k |x k Rewrite it in the following form:

[0052] p(z k |x k )=∫N(z k H k x k ,π k R k )Exp(π k ;λ k )dπ k (65)

[0053] Where Exp(·) represents the parameter λ k The exponential distribution of p(z); k |x k The marginal likelihood function of is rewritten in Gaussian mixture form:

[0054] p(z k |x k ,π k )=N(z k H k x k ,π k R k ),p(π k ) = Exp(π k ;λ k (66)

[0055] Combining variational Bayes, the joint parameter posterior distribution p(x) k ,π k ,P k|k-1 ,R k |z1:k Further calculations are as follows:

[0056] p(x k ,π k ,P k|k-1 ,R k |z 1:k )≈q(x k )q(π k )q(P k|k-1 )q(R k (67)

[0057] Where q(·) is the approximate posterior probability density function; based on variational Bayesian approximation, the approximate probability density function is achieved by minimizing the Kullback-Leiblerdivergence (KLD) between the approximate probability density function and the true probability density function, that is:

[0058]

[0059] in, The solution for each parameter in the parameter set is expressed as:

[0060]

[0061] in, It is a set of parameters; Φ is Θ k Any element in; It is the parameter set ignoring element Φ; c is the constant term; E[·] denotes the expectation calculation symbol; based on the joint probability density function p(Θ,z) 1:k The logarithmic objective function represented by ) is expressed as:

[0062]

[0063] For parameter updates, first assume Φ = π k Considering the posterior distribution function q (i+1) (π k It exhibits a form similar to the generalized inverse Gaussian distribution (GIG), that is:

[0064]

[0065] The update of the relevant parameters is represented as follows:

[0066]

[0067] Based on the properties of the generalized inverse Gaussian distribution The expected value is expressed as:

[0068]

[0069] Define Φ = R k ,q( i+1 (R) k The posterior probability density function of is expressed as:

[0070]

[0071] in,

[0072]

[0073] Define Φ = P k|k-1 P k|k-1 The posterior probability density function is expressed as:

[0074]

[0075] Therefore:

[0076]

[0077] Let Φ = x k State estimator and the corresponding estimation error covariance matrix It is obtained by using the standard Kalman filter update process formulas (9) and (10); according to the probability density function characteristics of IW, where, and Represented as:

[0078]

[0079] Updated measurement noise covariance matrix and the one-step prediction error covariance matrix Further expressed as:

[0080]

[0081] Posterior probability density function q( i+1 (x) k ) is represented as:

[0082]

[0083] To suppress non-Gaussian noise through an L1 smoothing dynamic process, equation (7) is rewritten as follows:

[0084]

[0085] Among them are:

[0086]

[0087] Q k and R k The initial values ​​are all defined as 10. -4 • I, where I is an identity matrix with relevant dimensions; when the time-varying model coefficients x are estimated k Subsequently, formulas (11) to (15) were used to further estimate the dynamic causal brain network.

[0088] Furthermore, the evaluation features described in step S3 include working memory brain region temporal variability, working memory processing ground-state link fluctuations, short-term memory network structural variability, and working memory encoding efficiency.

[0089] The temporal variability of working memory brain regions is extracted from brain region lateralization, and the calculation formula is as follows:

[0090]

[0091] in, γ represents the out-degree and in-degree strengths of network node i at time t, respectively; N and M are the node sets of the specified brain region and the reference brain region, respectively; T is the number of time points in the brain network; t It reflects the dominance of a specific brain region throughout the working memory process; Γ is a cognitive processing indicator of the temporal variability of working memory brain regions.

[0092] The ground-state link fluctuations in working memory processing are extracted by calculating the average distance of the brain network, as shown in the following formula:

[0093]

[0094] Among them, A t Let be the brain network matrix at time t, and D(A,B) represent the distance between matrices A and B;

[0095] The structural variability of the short-term memory network is extracted by calculating the overall spatial diversity, using the following formula:

[0096]

[0097]

[0098] in: It is accumulated t After the probability density function is sparse t The binary network i at time i, j Node connections;

[0099] The working memory encoding efficiency is extracted by calculating network time efficiency, and the calculation formula is as follows:

[0100]

[0101] in, express t At time i, network node j The average shortest path length between them.

[0102] Furthermore, step S3, which involves obtaining a test report on the subject's attention and memory processing based on evaluation features, specifically includes:

[0103] The working memory capacity of children and adolescents is measured based on the working memory brain region time variability index. The higher the working memory brain region time variability index value, the stronger the subject's working memory capacity. The ground state link fluctuation of the working memory processing reflects the brain's "readiness" or adaptability in working memory tasks. The short-term memory network structure variability reflects the subject's cognitive resource allocation in working memory tasks. The working memory encoding efficiency reflects the mental workload that children and adolescents can withstand.

[0104] On the other hand, the present invention provides a measurement system for attention and memory processing in children and adolescents based on dynamic brain networks, including an attention and memory induction module, an EEG preprocessing module, a brain network sequence acquisition module, and an attention and memory processing feedback module;

[0105] The attention and memory elicitation module is used to explain the requirements and methods of the n-back test to the test subjects, and let them practice independently for a certain period of time according to the requirements.

[0106] The EEG preprocessing module is used to collect and preprocess the EEG data set of the test subject.

[0107] The brain network sequence acquisition module obtains the dynamic network sequence of the preprocessed EEG data set based on MS-sPDC.

[0108] The attention and memory processing feedback module obtains the brain network sequence obtained by MS-sPDC, acquires the subject's attention and memory processing measurement index, and obtains the subject's attention and memory processing detection results based on the measurement index.

[0109] Furthermore, it also includes a human-computer interaction module, which is used to output the detection results of attention and memory processing and form an evaluation report.

[0110] Thirdly, the present invention provides a computer storage medium storing a computer program that, when run on a computer, can execute the above-described method for measuring attention and memory processing in children and adolescents based on dynamic brain networks.

[0111] The beneficial effects of this invention are as follows:

[0112] 1. Standardize the integration of EEG signal acquisition, processing, analysis, and feedback into a single system;

[0113] 2. By leveraging the advantages of the n-back experiment in inducing attention and memory, and the attention and memory processing mechanisms of children and adolescents, a feasible method for measuring attention and memory processing in children and adolescents was developed.

[0114] 3. By utilizing advanced dynamic brain network construction algorithms, robustness to outlier noise is achieved, and the dynamic connectivity changes between brain regions can be accurately described.

[0115] 4. The online real-time testing and training system is easy to operate and requires no pre-training.

[0116] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0117] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein:

[0118] Figure 1 This is a flowchart of the attention and memory processing measurement method of the present invention;

[0119] Figure 2 This is a framework diagram of the attention and memory processing measurement method of the present invention;

[0120] Figure 3 This is a schematic diagram of the attention and memory test task of the present invention;

[0121] Figure 4 Flowchart of the EEG preprocessing module;

[0122] Figure 5 Flowchart for generating attention and memory test reports. Detailed Implementation

[0123] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0124] It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Therefore, the drawings only show the components related to the present invention and are not drawn according to the actual number, shape and size of the components in the actual implementation. In the actual implementation, the form, quantity and proportion of each component can be arbitrarily changed, and the layout of the components may also be more complex.

[0125] In the following description, numerous details are explored to provide a more thorough explanation of embodiments of the invention. However, it will be apparent to those skilled in the art that embodiments of the invention may be practiced without these specific details. In other embodiments, well-known structures and devices are shown in block diagram form rather than in detail to avoid obscuring embodiments of the invention.

[0126] like Figure 1-5 As shown, the present invention provides a method for measuring attention and memory processing in children and adolescents based on dynamic brain networks, comprising the following steps:

[0127] S101. Before the test, the children and adolescents being tested are explained the requirements and methods of the n-back test, and they are asked to practice independently for 15 minutes according to the requirements. The formal test is then conducted after the first interval.

[0128] S102, in the formal test, the EEG data set of children and adolescents was collected and preprocessed, and a network model of the preprocessed EEG data set was obtained based on MS-sPDC.

[0129] S103, based on the dynamic network sequence calculated by MS-sPDC, obtains the evaluation features of the preprocessed EEG data set, and obtains the detection report of attention and memory processing of children and adolescents based on the evaluation features.

[0130] In step S101, the n-back test requirements and methods are explained to the students before the test, and they practice independently for 15 minutes. The formal test is then conducted after the first interval. The 15 minutes of independent practice allows children and adolescents to fully understand the n-back test task, which is used to stimulate their attention and memory processing. The first interval is a break period during which no visual stimulation is provided. Practical application is as follows:

[0131] Step 1) Provide children and teenagers with a 5-second start prompt to remind them that the test has begun;

[0132] Step 2) Provide children and adolescents with 2 seconds of digital stimulation to induce their attention and memory processing;

[0133] Step 3) Provide children and adolescents with a 0.3s reaction time to test their stimulus response;

[0134] Step 3) Provide children and adolescents with 0.5 seconds of feedback on the results to indicate whether the results are correct;

[0135] Step 4) Provide children and adolescents with a 5-second end prompt to remind them that the test is over.

[0136] like Figure 2 Before the test, the requirements and methods of the n-back test are explained to the children and adolescents. They practice independently for 15 minutes, followed by a short rest, before the actual test. By collecting EEG data from the children and adolescents, their attention and memory processing patterns are characterized. In this invention, the children and adolescents' response involves recalling the nth preceding digit stimulus based on the current digit stimulus and adding it to the current digit stimulus (addition within 10). This response is obtained as EEG data. For example, in the 3-back test, the children and adolescents need to recall the third preceding digit and add it to the current digit. Note that these tasks must be completed before the current digit stimulus ends. After the current digit stimulus ends, there is a 0.3-second reaction time during which the children and adolescents need to immediately input the addition result. Feedback is then provided based on the input result. A correct result will display a smiley face during the feedback phase; an incorrect result or exceeding the time limit will display a blank face.

[0137] In step S102, the EEG data set of children and adolescents is acquired and preprocessed. The dynamic network sequence of the preprocessed EEG data set is obtained based on MS-sPDC. The system uses an EEG cap, which integrates non-invasive EEG electrodes, electrode wires, and a cap for fixation, conforming to the international standard 10-20 system, and all 64 channels are selected as data acquisition channels. Preprocessing includes: performing mean reduction, detrending, and noise reduction on the acquired EEG data sequentially to improve signal quality and analyzability. Specifically, the EEG signals after digital stimulation are subjected to 4–8 Hz bandpass filtering to remove unwanted signals and eliminate other useless noise interference signals.

[0138] Brain network attributes are extracted from preprocessed EEG data, and these attributes can characterize the causal driving relationships between brain regions. This invention employs an adaptive dynamic brain network estimation method based on the Laplace distribution. This method is robust to artifact noise in EEG and can accurately describe the dynamic connectivity changes between brain regions. The algorithm principle is as follows:

[0139] Given D joint time series, the dynamic driving relationship at time k can be represented by the following D-dimensional time-varying multiple autoregressive model (tv-MVAR):

[0140]

[0141] Where, β ij (σ,k), σ=1,2,…,p;i=1,2,…,D;j=1,2…,D. D represents the number of time series; p represents the model order. Therefore, formula (1) can be further simplified to the following form:

[0142] z k =H k x k +ε k (90)

[0143] Among them, z k =[X1(k),…,X j (k),…,X D (k)] T ∈R D×1 ;x k =[β 11 (1,k),…,β 1D (1,k),…,β 11 (p,k),…,(p,k),…,β 1D (p,k),…,β D1(1,k),…,β DD (1,k),…,β D1 (p,k),…,β DD (p,k)] T ∈R (D ×p×D)×1 Meanwhile, the observation matrix H k The construction form can be represented as:

[0144]

[0145] In summary, the tv-MVAR model can be represented as the following state-space model, then:

[0146] x k =F k-1 x k-1 +w k-1 (92)

[0147] z k =H k x k +ε k (93)

[0148] Wherein, formula (4) is the state equation; formula (5) is the observation equation; k represents the discrete time point; x k ∈R D×p×D Represents the state vector; z k F represents the observation vector; k-1 H represents the state transition matrix; k Represents the observation matrix; w k ∈R D×p×D and ε k ∈R D ×1 Let these represent state noise and observation noise, respectively. Theoretically, the parameter update process of a state-space model can be divided into two parts: the single-step prediction process and the parameter update process. Therefore:

[0149] Single-step prediction:

[0150]

[0151]

[0152] Parameter update:

[0153]

[0154] P k|k =(IK k H k )P k|k -1 (98)

[0155] When the model coefficients x at time k are estimated k Then, sPDC is used to calculate the time-frequency representation of directed connectivity. Compared with traditional PDC-based dynamic causal network construction methods, sPDC can more accurately and stably measure the directed causal relationship from time series j to i. The corresponding formula for sPDC estimation is as follows:

[0156]

[0157] Among them, Ξ f,k It is the frequency domain representation of state xk at time k and frequency f. Where Ξ f,k It can be obtained through the Z-transform, then we have:

[0158]

[0159] Where τ represents the imaginary unit. Therefore, based on the sPDC calculation results, the causal relationship between node j and node i in the frequency band [f1, f2] at time k can be expressed as:

[0160]

[0161] Next, this paper introduces the t-distribution to calculate the corresponding threshold, thereby achieving the sparsity effect of the network. Then:

[0162] thre=T(1-p,n·(n-1)) (102)

[0163] Where T(1-p,v) represents the cumulative t-distribution probability density function with v degrees of freedom; 1-p is the confidence level. Then, for L... k The standardization process ignores the diagonal elements of the network at time k, i.e.:

[0164]

[0165] Here, △ represents any combination of points between [1, n], but the diagonal elements of the network matrix are ignored. This paper defines p as 0.001. If Lsk(i,j) is less than the threshold, then L... k The corresponding position (i,j) is 0. Based on the threshold judgment, the causal network at time k is thus made sparse.

[0166] Considering that the above methods are not applicable to dynamic network estimation under unknown noise, we introduce Bayesian inference into the state-space model, taking into account its adaptability in the dynamic causal estimation process. Then, the probability density functions of the state values ​​and observations under Bayesian inference can be expressed as follows:

[0167]

[0168] Considering P k|k-1 and R k These are positive definite matrices, and their prior distribution function p(P) k|k-1 |z 1:k-1 ) and p(R k |z 1:k-1 If it can be represented as an inverse-Wishart (IW) distribution, then:

[0169] p(P k|k-1 |z 1:k-1 ) = IW(P k|k-1 ;t k|k-1 ,T k|k-1 ), p(R k |z 1:k-1 ) = IW(R k ;u k|k-1 U k|k-1 (105)

[0170] Among them, t k|k-1 and T k|k-1 They represent p(P) k|k-1 |z 1:k-1 The degrees of freedom parameters and inverse scaling matrix of u; k|k-1 and U k|k-1 They represent p(R) k |z 1:k-1 The degrees of freedom parameters and inverse scaling matrix of p(P). Based on the properties of the inverse-Vichter distribution, p(P k|k-1 |z 1:k-1 The expectation of ) can be expressed as:

[0171] E[P k|k-1 ] = T k|k-1 / (t k|k-1 -m-1) (106)

[0172] Next, let t k|k -1=m+τ+1, where τ≥0 is the tuning parameter; therefore, formula (18) can be further expressed as T k|k -1=τP k|k-1 Furthermore, Bayesian inference is further introduced into the covariance matrix R of the observation noise. k Therefore, its expression forms are:

[0173] p(R k |z 1:k-1 )=∫p(R k |R k-1 )p(R k-1 |z 1:k-1 )dR k-1 (107)

[0174] Wherein, p(R) k-1 |z 1:k-1 ) follows an inverse-Vichter distribution, i.e., p(R) k-1 |z 1:k-1 ) = IW(R k-1 ;u k-1|k-1 U k-1|k-1 Therefore, the prior parameter u k|k-1 and U k|k-1 It can be replaced by u k|k-1 =ρ(u k-1|k-1 -n-1)+n+1,U k|k-1 =ρU k-1|k-1 ρ∈[0,1] is the forgetting factor. In summary, equations (16) to (19) construct an adaptive parameter estimation process within a Bayesian Gaussian state space. Based on the estimated time-varying model coefficients, equations (11) to (15) are used to further estimate the dynamic causal brain network.

[0175] While Bayesian dynamic causal networks achieve Bayesian modeling, their observation noise is still essentially assumed to be subject to L2 norm constraints. Therefore, to achieve robust estimation of dynamic causal networks, this invention introduces a multivariate Laplace distribution to construct a Bayesian robust representation under heavy-tailed measurement noise (HTMN). Specifically, HTMN can be modeled as a multivariate Laplace distribution in the following form:

[0176]

[0177] Among them, ML(z) k H k x k ,,R k ) represents z k The multidimensional Laplace distribution function, K·(·), represents the Bessel function of the second kind with order n / 2-1. Theoretically, a specific distribution in a Monte Carlo experiment can be represented as a proportional mixture of normal distributions. Therefore, p(z) k |x k It can be rewritten in the following form:

[0178] p(z k |x k )=∫N(z k H k x k ,π k R k )Exp(π k ;λ k )dπ k (109)

[0179] Where Exp(·) represents the parameter λ k The exponential distribution. Next, p(z) k |x k The marginal likelihood function of can be rewritten in Gaussian mixture form, then:

[0180] p(z k |x k ,π k )=N(z k H k x k ,π k R k ),p(π k ) = Exp(π k ;λ k (110)

[0181] In summary, equations (16) to (22) constitute an adaptive dynamic causal brain network estimation model based on Bayesian-multidimensional Laplace constraints. This model exhibits adaptability and robustness under noise. Therefore, in order to accurately estimate the state value x... k Combining variational Bayes, the joint parameter posterior distribution p(x) k ,π k ,P k|k-1 ,R k |z 1:k This can be further calculated as follows:

[0182] p(x k ,π k ,P k|k-1 ,R k |z 1:k )≈q(x k )q(π k )q(P k|k-1 )q(R k (111)

[0183] Here, q(·) is the approximate posterior probability density function. Based on variational Bayesian approximation, the approximate probability density function can be achieved by minimizing the Kullback-Leiblerdivergence (KLD) between the approximate and true probability density functions, i.e.:

[0184]

[0185] in, Therefore, the solution for each parameter in the parameter set can be expressed as:

[0186]

[0187] in, It is a set of parameters; Φ is Θ k Any element in; It is the parameter set ignoring element Φ; c is the constant term; E[·] denotes the expectation calculation symbol. In summary, based on the joint probability density function p(Θ,z) 1:k The logarithmic objective function represented by ) can be expressed as:

[0188]

[0189] For parameter updates, first assume Φ = π k Considering the posterior distribution function q( i+1 (π) k It exhibits a form similar to the generalized inverse Gaussian distribution (GIG), that is:

[0190]

[0191] The update of relevant parameters can be expressed as:

[0192]

[0193] Based on the properties of the generalized inverse Gaussian distribution The expected value can be expressed as:

[0194]

[0195] Next, define Φ = R k , then q (i+1) (R k The posterior probability density function of () can be expressed as:

[0196]

[0197] in,

[0198]

[0199] Define Φ = P k|k-1 P k|k-1 The posterior probability density function can be expressed as:

[0200]

[0201] Therefore:

[0202]

[0203] Let Φ = xk Considering that the state-space equations still conform to the Gaussian property, the state estimator and the corresponding estimation error covariance matrix This can be obtained by using the standard Kalman filter update process formulas (9) and (10). Furthermore, based on the probability density function characteristics of IW, where, It can be represented as:

[0204]

[0205] Updated measurement noise covariance matrix and the one-step prediction error covariance matrix This can be further expressed as:

[0206]

[0207] Then, the posterior probability density function q( i+1 (x) k ) is represented as:

[0208]

[0209] It is worth noting that this patent employs an advanced method to initially generate a suitable Q. k Value. That is, non-Gaussian noise is suppressed through an L1 smoothing dynamic process. This method has better performance in systems with impulse and Laplace noise, while improving the effect of coefficient sparsity. Therefore, equation (7) can be rewritten as:

[0210]

[0211] Among them are:

[0212]

[0213] In addition, Q k and R k The initial values ​​are all defined as 10. -4 Let I be an identity matrix with relevant dimensions. Then, when the time-varying model coefficients x are estimated... k Subsequently, formulas (11) to (15) were used to further estimate the dynamic causal brain network.

[0214] This invention names this dynamic brain network estimation method MS-sPDC to distinguish it from other time-varying PDC methods. The calculated network model is used to draw a brain network topology diagram, visualizing the brain's connection patterns and information transmission paths. Furthermore, this time-varying network is used to extract network attributes and characterize brain cognitive processing patterns.

[0215] Using the calculated brain network sequence, this invention extracts working memory cognitive processing indicators through the following steps:

[0216] (1) Temporal variability of working memory brain regions

[0217] Different brain regions exhibit varying degrees of activation during working memory cognitive processing. This invention employs brain region lateralization to extract this indicator, and the calculation formula is as follows:

[0218]

[0219] in, They represent in t The out-degree and in-degree strengths of network node i at time N. M These are the node sets for the specified brain region and the reference brain region, respectively. T represents the number of time points in the brain network. γ t It reflects the dominance of a specific brain region throughout the entire working memory process. Γ This refers to the cognitive processing index of working memory brain region time variability. In practice, the designated regions in this invention are the prefrontal cortex and parietal cortex, and the reference brain regions are other brain regions besides these two. The higher the value of the working memory brain region time variability index, the stronger the subject's working memory ability.

[0220] (2) Ground-state link fluctuations in working memory processing

[0221] Moderate fluctuations in brain network connectivity may be associated with improved cognitive abilities, especially in situations requiring high flexibility, such as working memory. The brain needs to be able to flexibly adjust network connectivity according to task demands to effectively store, retain, and retrieve information. This invention extracts this indicator by calculating the average distance of brain networks, using the following formula:

[0222]

[0223] Among them, A t It is the brain network matrix at time t. D( (A, B) represents the distance between matrices A and B. The ground-state link variability index of working memory processing reflects the brain's "readiness" or adaptability in working memory tasks.

[0224] (3) Structural variability of short-term memory networks

[0225] The overall spatial diversity index of dynamic brain networks reflects the complexity and breadth of connectivity patterns, which is fundamental to efficient cognitive processing. This invention extracts this index by calculating the degree of overall spatial diversity:

[0226]

[0227] in: It is accumulated t After the probability density function is sparse t The binary network i at time i, j Node connections. Short-term memory network structural variability indicators characterize the allocation of cognitive resources by subjects in working memory tasks.

[0228] (4) Working memory encoding efficiency

[0229] Performing increasingly complex cognitive tasks requires more brain resources, leading to a higher mental workload. In this case, the brain needs to process the task with greater efficiency. This invention extracts this indicator by calculating network time efficiency:

[0230]

[0231] in, express t At time i, network node j The average shortest path length between objects. Working memory coding efficiency index characterizes the level of mental workload that a subject can withstand.

[0232] This application also provides a system for measuring attention and memory processing in children and adolescents based on dynamic brain networks. The system includes at least one software functional module stored in a storage module or embedded in an operating system (OS) in the form of software or firmware. An attention and memory induction module and a brain network sequence acquisition module are used to execute executable modules stored in the storage module, such as the software functional modules and computer program modules included in the system for measuring attention and memory processing in children and adolescents based on dynamic brain networks.

[0233] The system also includes an attention and memory induction module, an EEG preprocessing module, a brain network sequence acquisition module, and an attention and memory processing feedback module.

[0234] The attention and memory stimulation module explains the n-back test requirements and methods to the children and adolescents before the test, allowing them to practice independently for 15 minutes, and then conduct the formal test after the first interval.

[0235] The EEG preprocessing module and the brain network sequence acquisition module collect and preprocess EEG data sets from children and adolescents during formal testing, and obtain network models of the preprocessed EEG data sets based on MS-sPDC.

[0236] The attention and memory processing feedback module, based on the dynamic network sequence calculated by MS-sPDC, obtains the evaluation features of the preprocessed EEG data set, and generates a test report on the attention and memory processing of children and adolescents based on the evaluation features.

[0237] The attention and memory processing feedback module is coupled with a human-computer interaction module to output an evaluation report on the attention and memory processing effect.

[0238] In this embodiment, the storage module can be, but is not limited to, random access memory, read-only memory, programmable read-only memory, erasable programmable read-only memory, electrically erasable programmable read-only memory, etc. In this embodiment, the storage module can be used to store algorithms such as the MS-sPDC algorithm in the brain network sequence acquisition module. Of course, the storage module can also be used to store programs, which the processing module executes after receiving an execution instruction.

[0239] Based on the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by hardware or by using software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application can be embodied in the form of a software product. This software product can be stored in a non-volatile storage medium (such as CD-ROM, USB flash drive, mobile hard drive, etc.) and includes several instructions to cause a computer device (such as a personal computer, electronic device, or network device, etc.) to execute the methods described in the various implementation scenarios of this application.

[0240] In summary, this application provides a rehabilitation training method, system, and medium based on a multimodal brain-computer interface. The rehabilitation training method based on a multimodal brain-computer interface includes the following steps: S101, explaining the n-back test requirements and methods to the children and adolescents before the test, allowing them to practice independently for 15 minutes, and conducting the formal test after a first time interval; S102, during the formal test, collecting and preprocessing the children's and adolescents' EEG data sets, and obtaining the network model of the preprocessed EEG data sets based on MS-sPDC; S103, obtaining the evaluation features of the preprocessed EEG data sets based on the dynamic network sequence calculated by MS-sPDC, and obtaining a detection report on the children's and adolescents' attention and memory processing based on the evaluation features. This invention has high accuracy, high information transmission rate, convenient operation, requires no training, and has high practical value.

[0241] In the above embodiments, the reference to "this embodiment" in the specification indicates that a specific feature, structure, or characteristic described in connection with the embodiment is included in at least some embodiments, but not necessarily all embodiments. Multiple appearances of "this embodiment" do not necessarily refer to the same embodiment.

[0242] In the above embodiments, although the invention has been described in conjunction with specific embodiments thereof, many substitutions, modifications, and variations of these embodiments will be apparent to those skilled in the art from the foregoing description. For example, other memory structures (e.g., dynamic RAM (DRAM)) may be used with the embodiments discussed. The embodiments of the invention are intended to cover all such substitutions, modifications, and variations falling within the broad scope of the appended claims.

[0243] This embodiment also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements any of the methods in this embodiment.

[0244] This embodiment also provides an electronic terminal, including: a processor and a memory;

[0245] The memory is used to store computer programs, and the processor is used to execute the computer programs stored in the memory to cause the terminal to perform any of the methods in this embodiment.

[0246] As will be understood by those skilled in the art, the computer-readable storage medium described in this embodiment allows for the implementation of all or part of the steps in the above method embodiments by computer program-related hardware. The aforementioned computer program can be stored in a computer-readable storage medium. When executed, the program performs the steps of the above method embodiments; and the aforementioned storage medium includes various media capable of storing program code, such as ROM, RAM, magnetic disks, or optical disks.

[0247] The electronic terminal provided in this embodiment includes a processor, a memory, a transceiver, and a communication interface. The memory and the communication interface are connected to the processor and the transceiver and complete communication between them. The memory is used to store computer programs, the communication interface is used to perform communication, and the processor and the transceiver are used to run the computer programs, so that the electronic terminal performs the steps of the above method.

[0248] In this embodiment, the memory may include random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device.

[0249] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0250] This invention can be used in a wide range of general-purpose or special-purpose computing system environments or configurations. Examples include: personal computers, server computers, handheld or portable devices, tablet devices, multiprocessor systems, microprocessor-based systems, set-top boxes, programmable consumer electronics, network PCs, minicomputers, mainframe computers, and distributed computing environments including any of the above systems or devices, etc.

[0251] This invention can be described in the general context of computer-executable instructions, such as program modules, that are executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, etc., that perform a specific task or implement a specific abstract data type. This invention can also be practiced in distributed computing environments where tasks are performed by remote processing devices connected via a communication network. In distributed computing environments, program modules can reside in local and remote computer storage media, including storage devices.

[0252] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for measuring attention and memory processing in children and adolescents based on dynamic brain networks, characterized in that: Includes the following steps: S1: Have the test subject practice the n-back test requirements and methods independently for a certain period of time, and then conduct the n-back test after a certain interval; S2: During the n-back test, the subject's EEG data set is preprocessed, and the brain network sequence of the preprocessed EEG data set is obtained based on the dynamic brain network estimation model MS-sPDC. The specific steps include: An adaptive dynamic brain network estimation method based on the Laplace distribution describes the dynamic connectivity changes between brain regions. Given D joint time series, time is represented by a D-dimensional time-varying multivariate autoregressive model tv-MVAR. k The dynamic driving relationship; The tv-MVAR model is represented as a state-space model, including state equations and observation equations; For the state-space model, the parameter estimation process includes two stages: single-step prediction and recursive update. The estimated model coefficients at each time step are obtained iteratively. k After obtaining the model coefficients, the sPDC method is used to perform frequency domain analysis on the directed connectivity between brain regions. By performing frequency domain transformation on the time-varying model coefficients, the connectivity strength at each frequency is obtained, and the causal relationship between brain regions is characterized within a given frequency band. A threshold strategy based on the t-distribution is introduced to standardize the connection matrix and remove diagonal elements. Weak connections are then screened out by threshold determination, thereby achieving a sparse representation of dynamic brain networks. By introducing a multivariate Laplace distribution, a Bayesian-based robust representation is constructed for heavy-tailed observation noise (HTMN). Specifically, HTMN is modeled as a multivariate Laplace distribution: in, z k Indicates time k The observed vector below, For state variables, For the observation matrix, To observe the noise covariance matrix, ML ( z k; H k x k , , R k )express z k The multidimensional Laplace distribution function, K · (·) indicates that the order is n / 2-1 The second kind of Bessel function; p ( z k | x k Rewrite it in the following form: in, Exp (·) indicates that the parameter is λ k The exponential distribution; p ( z k | x k The marginal likelihood function of is rewritten in Gaussian mixture form: Combining variational Bayesian methods with joint parameter posterior distributions Further calculations are as follows: Among them, error covariance , Here is the state transition matrix. The covariance of the state equation. q (·) is the approximate posterior probability density function; based on variational Bayesian approximation, the approximate probability density function is achieved by minimizing the Kullback-Leibler divergence between the approximate probability density function and the true probability density function, that is: in, The solution for each parameter in the parameter set is expressed as follows: in, It is a set of parameters; yes Any element in; Ignore element The parameter set afterwards; c It is a constant term; E [·] indicates the expected calculation symbol; P k|k-1 and R k The prior is based on the inverse-Vichter distribution: in, t k|k-1 and u k|k-1 For the degree of freedom parameter, T k|k-1 and U k|k-1 The inverse scaling matrix is ​​used; based on this, the joint probability density function is... The logarithmic objective function is expressed as: For parameter updates, first press the settings... Considering the posterior distribution function It exhibits a form similar to the generalized inverse Gaussian distribution GIG, namely: The update of the relevant parameters is represented as follows: Based on the properties of the generalized inverse Gaussian distribution The expected value is expressed as: definition , The posterior probability density function is expressed as: in, definition , P k|k-1 The posterior probability density function is expressed as: Therefore: make State estimator and the corresponding estimation error covariance matrix It is obtained by updating using standard Kalman filtering; based on the probability density function characteristics of IW, where, and Represented as: Updated measurement noise covariance matrix and the one-step prediction error covariance matrix Further expressed as: Posterior probability density function Represented as: Non-Gaussian noise is suppressed through an L1 smoothing dynamic process, and the error covariance is rewritten as: Among them are: Q k and R k The initial values ​​are all defined as 10. -4 · I , I It is an identity matrix with relevant dimensions; when the time-varying model coefficients are estimated... x k Subsequently, the dynamic causal brain network was further estimated using sPDC; S3: Obtain the evaluation features of the preprocessed EEG data set from the brain network sequence, and obtain the test report of the subject's attention and memory processing based on the evaluation features.

2. The method for measuring attention and memory processing in children and adolescents based on dynamic brain networks according to claim 1, characterized in that: The preprocessing of the EEG data set collected from the test subject in step S2 specifically includes: performing mean removal processing, trend removal processing, and noise removal processing on the EEG data set in sequence.

3. The method for measuring attention and memory processing in children and adolescents based on dynamic brain networks according to claim 1, characterized in that: The evaluation features described in step S3 include working memory brain region temporal variability, ground-state link fluctuations in working memory processing, short-term memory network structural variability, and working memory encoding efficiency. The temporal variability of working memory brain regions is extracted from brain region lateralization, and the calculation formula is as follows: in, , They represent in Time network nodes The out-of-range intensity and the in-range intensity; , These are the node sets for the specified brain region and the reference brain region, respectively. The number of time points in the brain network; It reflects the dominance of a specific brain region throughout the entire working memory process; This is an indicator of cognitive processing based on the time variability of brain regions used for working memory; The ground-state link fluctuations in working memory processing are extracted by calculating the average distance of the brain network, as shown in the following formula: in, It is the brain network matrix at time t. Representation matrix ， The distance; The structural variability of the short-term memory network is extracted by calculating the overall spatial diversity, using the following formula: in: It is accumulated After the probability density function is sparse Binary networks at different times , Node connections; The working memory encoding efficiency is extracted by calculating network time efficiency, and the calculation formula is as follows: in, express Current network nodes , The average shortest path length between them.

4. The method for measuring attention and memory processing in children and adolescents based on dynamic brain networks according to claim 3, characterized in that: Step S3, which involves obtaining a test report on the subject's attention and memory processing based on evaluation features, specifically includes: The working memory capacity of children and adolescents is measured based on the working memory brain region time variability index. The higher the working memory brain region time variability index value, the stronger the working memory capacity of the subject. The ground state link fluctuation of the working memory processing reflects the brain's "readiness" or adaptability in working memory tasks. The short-term memory network structure variability reflects the subject's cognitive resource allocation in working memory tasks. The working memory encoding efficiency reflects the mental load that children and adolescents can withstand.

5. A measurement system for attention and memory processing in children and adolescents based on dynamic brain networks, characterized in that: It includes modules for eliciting attention and memory, EEG preprocessing, brain network sequence acquisition, and attention and memory processing feedback. The attention and memory induction module is used to explain the n-back test requirements and methods to the test subjects and let them practice independently for a certain period of time according to the requirements. The EEG preprocessing module is used to collect and preprocess the EEG data set of the test subject. The brain network sequence acquisition module obtains the dynamic network sequence of the preprocessed EEG data group based on the dynamic brain network estimation model MS-sPDC. The attention and memory processing feedback module obtains the brain network sequence obtained by MS-sPDC, acquires the subject's attention and memory processing measurement index, and obtains the subject's attention and memory processing detection results based on the measurement index; The above system implements the method for measuring attention and memory processing in children and adolescents based on dynamic brain networks as described in any one of claims 1-4.

6. The measurement system for attention and memory processing in children and adolescents based on dynamic brain networks according to claim 5, characterized in that: It also includes a human-computer interaction module, which is used to output the detection results of attention and memory processing and form an evaluation report.

7. A computer storage medium storing a computer program, characterized in that: When the computer program is run on a computer, it is able to perform the method for measuring attention and memory processing in children and adolescents based on dynamic brain networks as described in any one of claims 1-4.

Images