Anxiety state evaluation method, system, equipment and medium

Abstract

The invention discloses an anxiety state assessment method, system and device and a medium, and relates to the technical field of electroencephalogram signal assessment. An observed value of the electroencephalogram data sequence at a set moment is extracted, and a time-varying linear coupling relation between the electroencephalogram data sequence and the observed value of the electroencephalogram data sequence is established based on a multivariate adaptive autoregression model; a conversion relation function between states corresponding to the multivariate signals is obtained through a time-varying linear coupling relation between the electroencephalogram data sequence and an observed value of the electroencephalogram data sequence, the conversion relation function is solved to obtain an ATDF value, and a dynamic causal brain network sequence is estimated through the ATDF value; according to the dynamic causal brain network sequence, electroencephalogram network time sequence features are extracted, and then the clinical anxiety state score of the testee is estimated; according to the method, the response to the brain is analyzed from the two dimensions of time and space, and the accuracy of the evaluation result of the clinical anxiety state of the testee is improved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of electroencephalogram signal evaluation, and in particular to an anxiety state evaluation method, system, device and medium. BACKGROUND

[0002] The brain electrical network, as an important indicator of brain neural activity, is closely related to the clinical anxiety state of medical staff. The neural electrical activity of the brain can directly reflect the psychological state of individuals. Medical staff, as a key group in medical work, are in a high-intensity and high-pressure work environment for a long time, which is prone to produce clinical anxiety state. The clinical anxiety state of medical staff not only affects their own physical and mental health, but also may have a negative impact on the quality of nursing work. In clinical work, anxious medical staff may have problems such as lack of concentration and decreased decision-making ability, which affects the judgment of patient's condition and the implementation of nursing measures. Changes in brain electrical networks can provide a unique perspective for early evaluation and prevention of clinical anxiety state of medical staff.

[0003] The brain electrical network is a complex neural electrical signal system that reveals the functional connectivity patterns of the brain by monitoring the electrical activity correlation between different regions of the brain. These connectivity patterns can reflect the working mechanisms of the brain under different cognitive and emotional states. In the study of clinical anxiety state of medical staff, brain electrical network can help us understand the changes in information transmission and cooperative work between different brain regions when anxiety occurs. For example, when medical staff are in an anxious state, the connections between brain regions related to emotional regulation and attention in the brain electrical network may change, which provides a basis for understanding the neural mechanisms of anxiety.

[0004] However, current methods of analyzing brain electrical networks are difficult to characterize information exchange between brain regions when the brain is in an anxious state, and existing methods pay little attention to real-time cognitive processing characteristics of anxiety. For example, eye movement research has found that socially anxious individuals have a "vigilance-avoidance" dynamic attention bias to positive evaluations, but traditional assessments do not integrate such time dimension information, and cannot capture the timing changes of brain networks during the execution of spatial cue tasks, which makes it difficult to explore the network timing dependence characteristics of the brain during cognitive processing and the corresponding dynamic cognitive neural mechanisms. SUMMARY

[0005] In view of the fact that the prior art cannot capture the timing variation characteristics of the brain network during the execution of the spatial cue task, it is difficult to mine the network timing dependence characteristics of the brain during the cognitive processing and the corresponding dynamic cognitive neural mechanism, and the present application provides an anxiety state evaluation method, system, device and medium, which extracts the characteristics of the brain in the anxiety state under external stimulation by using a dynamic brain network, and describes the information exchange between brain regions of the brain in the anxiety state during the execution of the spatial cue task, thereby solving the problems existing in the prior art.

[0006] An anxiety state evaluation method, comprising the following steps: Collecting electroencephalogram data sequences generated by the effective and ineffective clue stimulation of the user to be tested through the spatial cue task test; Extracting the electroencephalogram network timing characteristics with causal relationship of each electroencephalogram data sequence; the extraction process specifically includes: Extracting the observation value of the electroencephalogram data sequence at the set time, establishing the time-varying linear coupling relationship between the electroencephalogram data sequence and its corresponding observation value, obtaining the conversion relationship function between the states corresponding to the electroencephalogram data sequence through the time-varying linear coupling relationship between the electroencephalogram data sequence and its observation value, obtaining the causal influence of the current electroencephalogram sequence on the target electroencephalogram sequence and the adaptive directed transfer function ATDF value of the causal influence at the set frequency point through solving the conversion relationship function, estimating the dynamic causal brain network sequence through the ATDF value, and extracting the electroencephalogram network timing characteristics with causal relationship of the electroencephalogram data sequence according to the dynamic causal brain network sequence; Performing multivariate regression analysis on the electroencephalogram network timing characteristics of the electroencephalogram data sequences generated based on the effective and ineffective clue stimulation, and evaluating the anxiety state of the user to be tested.

[0007] Further, the time-varying linear coupling relationship between the electroencephalogram data sequence and its corresponding observation value is established by using a multivariate adaptive autoregressive model, which specifically includes the following steps: Definition For the electroencephalogram sequence collected from electrode leads within a set time period, the observation value of the electroencephalogram sequence at the time is , wherein , the time-varying linear coupling relationship between the electroencephalogram data sequence at this time and its observation value is established by using a multivariate adaptive autoregressive model, and is represented as: , wherein represents the th time-varying coefficient vector of the th equation in the model; represents the order of the model; residuals of the th equation, th equation, M th lead at time t th lead at time M th lead at time t th lead at time

[0008] Further, the conversion relationship function between the states corresponding to the electroencephalogram data sequence is obtained by a time-varying linear coupling relationship between the electroencephalogram data sequence and its observation values, and specifically includes the following steps: The conversion relationship function between the states corresponding to the electroencephalogram data sequence is obtained by converting the th equation in the time-varying linear coupling relationship between the electroencephalogram data sequence and its observation values, and specifically includes the following steps: ; Wherein, consists of the first past observation values of each electroencephalogram sequence at time th equation in the multivariate adaptive autoregressive model corresponds to a coefficient vector; The conversion relationship function between the states corresponding to the electroencephalogram data sequence is expressed as: ; Wherein, denotes a state transition matrix; denotes state noise, t at time t at time

[0009] Further, the conversion relationship function is solved to obtain the causal influence of the current electroencephalogram sequence on the target electroencephalogram sequence and the adaptive directed transfer function ATDF value of the causal influence at a set frequency point, and specifically includes the following steps: The conversion relationship function is solved using the least squares method , expressed as: ; Wherein, , denotes R is the covariance matrix of the observation noise at time t According to , the ATDF value corresponding to the set frequency at time is estimated: ;​​​​​ in, For the The time point corresponds to the The model coefficient of order is expressed as: ; And there is , for The inverse of the EEG sequence EEG sequence The normalized ATDF value of the causal effect at the frequency point is: ; in, For the moment , EEG sequence Target EEG sequence The causal effect of frequency ATDF value at .

[0010] Furthermore, the multiple regression analysis of the EEG network temporal characteristics of the EEG data sequences generated based on the effective cue stimulation and the invalid cue stimulation to evaluate the anxiety state of the user to be tested specifically includes the following steps: Using the evaluation features and the set anxiety score, a multiple regression analysis was performed on the EEG network temporal features of the EEG data sequences generated by effective cue stimulation and ineffective cue stimulation to train a multiple regression model; wherein the evaluation features include behavioral indicators, event-related potentials, state proportion, state interval time, and number of state switches; Use hidden Markov models to decode dynamic causal brain network sequences and obtain state time series; The state time series is calculated to obtain the state proportion, state interval time and state switching number; the state proportion is the ratio of the time each state appears to the total time; the state interval time is the interval time when the state appears continuously; the state switching number is the number of times the state sequence changes within a period of time; According to the state proportion, state interval time and state switching number, the anxiety state of the user to be tested is evaluated by the trained multivariate regression model.

[0011] The present invention also includes an anxiety state assessment system, comprising: An acquisition module is used to acquire EEG data sequences generated by the user under test after effective cue stimulation and invalid cue stimulation through a spatial cue task test; The EEG network temporal feature extraction module is used to extract the EEG network temporal features with causal relationships for each EEG data sequence. The extraction process specifically includes: Extract the observation value of the EEG data sequence at a set time, establish a time-varying linear coupling relationship between the EEG data sequence and its corresponding observation value; obtain the conversion relationship function between the states corresponding to the EEG data sequence through the time-varying linear coupling relationship between the EEG data sequence and its observation value; obtain the causal influence of the current EEG sequence on the target EEG sequence and the adaptive directed transfer function ATDF value of the causal influence at the set frequency point by solving the conversion relationship function; estimate the dynamic causal brain network sequence through the ATDF value; and extract the EEG network time series features with causal relationship of the EEG data sequence based on the dynamic causal brain network sequence; The evaluation module is used to perform multiple regression analysis on the EEG network temporal characteristics of the EEG data sequences generated based on effective clue stimulation and invalid clue stimulation to evaluate the anxiety state of the user to be tested.

[0012] The present invention also includes an anxiety state assessment computer device, comprising: a memory, a processor, and a computer program stored in the memory, wherein the processor implements the steps of the anxiety state assessment method when executing the computer program.

[0013] The present invention also includes a readable storage medium, wherein the readable storage medium stores a computer program, wherein the computer program includes program instructions, and when the program instructions are executed by a processor, the steps of the anxiety state assessment method are executed.

[0014] The present invention provides a method for assessing anxiety state, which has the following beneficial effects: The present invention constructs a dynamic brain network by using the adaptive directed transfer function ATDF, captures the changes in brain functional connections from two dimensions of time and space, and further derives the state transition matrix through the time-varying linear coupling relationship to describe the conversion law of EEG sequences between different states; through the state transition relationship, the causal influence of the current EEG sequence on the target EEG sequence and the ATDF value at a specific frequency point are solved. The ATDF value changes with time, which can capture the dynamic reorganization process of the brain network, distinguish the direction of information flow, and reveal the regulatory abnormalities of key brain areas under anxiety conditions; by integrating the ATDF values ​​at each moment and frequency point, a dynamic causal brain network sequence is constructed to reflect the time-varying characteristics of brain functional connections. The dynamic network can capture millisecond-level changes, more accurately reflect the real-time response of the brain, and improve the accuracy of the assessment results of the clinical anxiety state of the subjects. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 This is a flow chart of the anxiety state assessment method in an embodiment of the present invention; Figure 2 This is a framework diagram of the anxiety state assessment system in an embodiment of the present invention; Figure 3A space cue task diagram for a spatial cue task in an embodiment of the present application; Figure 4 A flow chart of an EEG preprocessing module and a clinical anxiety state score generation in an embodiment of the present application. DETAILED DESCRIPTION

[0016] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, but not all the embodiments.

[0017] The present application provides an anxiety state evaluation method based on an EEG network, as shown in Figure 1 , Figure 4 The method specifically comprises the following steps: S1, making the subject complete a status-trait anxiety inventory (STAI) to evaluate the anxiety state and trait anxiety of the subject.

[0018] The subject needs to complete the status-trait anxiety inventory (STAI), which covers two sub-inventories, a total of 40 questions, to evaluate the state anxiety and trait anxiety of the user. Among them, the state anxiety inventory has 20 questions, and each question requires the user to answer the anxiety feeling at a specific moment (anxiety feeling from completely not anxious to very anxious according to four-level scoring); the trait anxiety has 20 questions to evaluate the frequency of feeling anxious (anxiety frequency from almost not to almost always according to four-level scoring).

[0019] S2, introduce the test requirements and methods of the space cue task (SC) to the subject, and let the subject practice for a period of time and then test.

[0020] Introduce the test requirements and methods of the space cue task (SC) to the subject, and let the subject practice autonomously for 15 minutes according to the requirements, and then perform formal testing after a period of time. The 15-minute autonomous practice is used for the subject to fully understand the test task of SC. In the task execution stage, the user needs to press the key according to the orientation of the target, wherein the key "f" means that the target appears on the left, and the key "j" means that the target appears on the right. Each experiment totals 43 trials, each of which contains two stimuli, i.e., a cue stimulus and a target stimulus. Among them, the cue stimulus appears before the target stimulus, and the cue stimulus is divided into effective stimulus (the cue appears on the opposite side of the target, a total of 288 trials) and ineffective stimulus (the cue appears on the same side of the target, a total of 144 trials).

[0021] As shown in Figure 3The flow of one trial of SC is shown, first a crosshair is presented for 500 ms, the user needs to focus on the crosshair; after the crosshair is presented, a blank background appears for 750 ms; then a cue stimulus appears for 300 ms, followed by a target stimulus for 200 ms. The user needs to press the "f" or "j" key according to whether the target stimulus appears on the left or right side of the screen.

[0022] S3, in the SC test process, the brain electrical data set of the tested person is collected for preprocessing, and the brain network sequence of the preprocessed brain electrical data set is obtained based on an adaptive transfer directed function (ATDF).

[0023] During the task performed by the tested person, the relevant brain electrical signals are synchronously collected and reference corrected using a reference electrode standardization technique. For each trial, the brain electrical data from 2 seconds before the appearance of the cue to 2 seconds after the appearance of the cue are analyzed. The brain electrical cap used by the system, i.e., the combination of a non-invasive brain electrical electrode, an electrode wire and a fixing cap, conforms to the international standard 10-20 system, and all 64 channels are selected as data acquisition channels.

[0024] The preprocessing includes: the collected brain electrical data are sequentially subjected to mean removal processing, trend removal processing and noise removal processing, so as to improve the signal quality and analyzability. Among them, the brain electrical signals subjected to the cue stimulus are subjected to band-pass filtering processing of 4~8Hz respectively, so as to remove useless signals, thereby eliminating and distinguishing other useless noise interference signals into two groups of brain electrical data generated by effective cue stimulus and brain electrical data generated by ineffective cue stimulus.

[0025] The two groups of brain electrical data after preprocessing are respectively extracted using ATDF to obtain the corresponding dynamic brain network state sequence, which includes the following steps: Definition For the brain electrical sequence collected from electrode leads in time, the observation value at the moment is , wherein , the time-varying linear coupling relationship between the observation values at this moment can be described by the following multivariate adaptive autoregressive model: , wherein, represents the th time-varying coefficient vector of the th equation in the model; represents the order of the model; represents the residuals of the equations correspond to time-varying covariances, is the th M t is the EEG data recorded at time i from the th M is the noise at time i from the th t The th where, is composed of the first p past observations at time i from each sequence; is the coefficient vector corresponding to the th The state changes from the th M

[0026] where, is the state transition matrix; is the state noise. The parameter coefficients of the model at time i can be estimated using the least squares method by:

[0027] where, , , .

[0028] When the parameter coefficients of the model at time i are estimated, the ATDF values corresponding to each frequency point at this time can be estimated by: where, is the th is the model coefficient of the th order at the th and , is the inverse of , then the normalized ATDF value of the causal influence of any EEG sequence on the EEG sequence at each frequency point can be calculated by:​​​​​​​ wherein, is the time point , the electroencephalogram sequence The causal influence of the electroencephalogram sequence The ATDF value at the frequency point , the ATDF information of the time point in the frequency band The integration of the ATDF information can be characterized by the average ATDF, and the calculation formula is as follows: .

[0029] S4, obtain the electroencephalogram network time sequence characteristics from the dynamic network sequence, and perform multiple regression analysis based on the time sequence characteristics and the anxiety score to estimate the clinical anxiety state score of the subject.

[0030] The evaluation characteristics include behavioral indicators, event-related potentials (ERP), state proportion, state interval time and state switching number. The behavioral indicators include accuracy and reaction time. The accuracy is the proportion of the correct judgment of the subject in the SC test to the total number of trials. The reaction time is the time consumed from the appearance of the target to the pressing of the key by the subject, including the average reaction time of accepting valid clues , the average reaction time of accepting invalid clues , and the average reaction time of all trials . For subjects in an anxious state, the reaction time will be longer, and more reaction time will be needed for invalid clues. The ERP is the P2 amplitude, including the average amplitude of accepting valid clues and the average amplitude of accepting invalid clues The amplitude of subjects in an anxious state will be slightly larger. The state proportion, state interval time and state switching number are obtained by decoding the dynamic electroencephalogram network sequence obtained in step S103 using a hidden Markov model (HMM) to obtain the corresponding state time sequence, and then calculating the following indicators from the hidden state sequence:

[0031] State proportion (FO), i.e. the ratio of the time of each state appearing to the total time. High proportion indicates that the state appears for a long time, and vice versa.

[0032] State interval time (IT), i.e. the interval time when the state appears continuously.

[0033] ​State switching number (N-index), represented as the number of state sequence changes in a period of time.

[0034] Based on the evaluation characteristics and the anxiety score, multiple regression analysis is performed to estimate the clinical anxiety state of the testee, specifically including: performing multiple regression analysis on the above behavior indicators, ERPs, state proportion, state interval time and state switching number evaluation characteristics and the anxiety score obtained in step S1 to train a multiple regression model, which can be used to measure the anxiety state of the testee. Figure 2 As shown in the figure, the framework of the anxiety state evaluation system of the present application is shown.

[0035] Based on the same inventive concept, the present application also provides an anxiety state evaluation system, comprising: The acquisition module is used to collect the EEG data sequence generated by the user to be tested under effective and ineffective cue stimulation through the spatial cue task test.

[0036] The EEG network time series feature extraction module is used to extract the EEG network time series features with causal relationship of each EEG data sequence respectively; the extraction process specifically includes: The observation value of the EEG data sequence at a set time is extracted, and the time-varying linear coupling relationship between the EEG data sequence and its corresponding observation value is established; the conversion relationship function between the states corresponding to the EEG data sequence is obtained through the time-varying linear coupling relationship between the EEG data sequence and its observation value; the causal influence of the current EEG sequence on the target EEG sequence and the adaptive directed transfer function (ATDF) value of the causal influence at a set frequency point are obtained by solving the conversion relationship function; the dynamic causal brain network sequence is estimated by the ATDF value; and the EEG network time series features with causal relationship of the EEG data sequence are extracted according to the dynamic causal brain network sequence.

[0037] The evaluation module is used to perform multiple regression analysis on the EEG network time series features of the EEG data sequence generated based on effective and ineffective cue stimulation to evaluate the anxiety state of the user to be tested.

[0038] The present application also provides an anxiety state evaluation computer device, comprising: a memory, a processor and a computer program stored in the memory, and the processor executes the computer program to realize the steps of the anxiety state evaluation method.

[0039] The present application also provides a readable storage medium, which stores a computer program, and the computer program comprises program instructions, which are executed by the processor to perform the steps of the anxiety state evaluation method.

[0040] The above merely describes preferred specific embodiments of the present application, but the protection scope of the present application is not limited thereto, and any person skilled in the art, according to the technical solution and inventive concept of the present application, makes equivalent replacement or change within the technical range disclosed by the present application, which should be covered within the protection scope of the present application.

Claims

1. A method for assessing anxiety state, characterized in that: The following steps are involved: Through the spatial cue task test, the EEG data sequences generated by the user under test after effective cue stimulation and invalid cue stimulation are collected; Extract the EEG network temporal features with causal relationship for each EEG data sequence respectively; The extraction process specifically includes: Extract the observation value of the EEG data sequence at a set time, establish a time-varying linear coupling relationship between the EEG data sequence and its corresponding observation value; obtain the conversion relationship function between the states corresponding to the EEG data sequence through the time-varying linear coupling relationship between the EEG data sequence and its observation value; obtain the causal influence of the current EEG sequence on the target EEG sequence and the adaptive directed transfer function ATDF value of the causal influence at the set frequency point by solving the conversion relationship function; estimate the dynamic causal brain network sequence through the ATDF value; and extract the EEG network time series features with causal relationship of the EEG data sequence based on the dynamic causal brain network sequence; A multiple regression analysis was performed on the EEG network temporal characteristics of the EEG data sequences generated based on effective cue stimulation and invalid cue stimulation to evaluate the anxiety state of the user to be tested.

2. The anxiety state assessment method according to claim 1, characterized in that: The multivariate adaptive autoregressive model is used to establish a time-varying linear coupling relationship between the EEG data sequence and its corresponding observations, which specifically includes the following steps: definition For those from The EEG sequence collected by the electrode leads within the set time period is The observation value at time is ,in , then a multivariate adaptive autoregressive model is used to establish the time-varying linear coupling relationship between the EEG data sequence and its observation value at that moment, which can be expressed as: in, Indicates the model The first A time-varying coefficient vector; represents the model order; Expressed as The time-varying covariance corresponding to the residuals of the equations is, For the M Leads in t -Electroencephalogram data recorded at every moment, Indicates the M Leads in t The noise of the moment.

3. The anxiety state assessment method according to claim 2, characterized in that: The method of obtaining the conversion relationship function between the states corresponding to the EEG data sequence through the time-varying linear coupling relationship between the EEG data sequence and its observation value specifically includes the following steps: By using the time-varying linear coupling relationship between the EEG data series and its observations Transform the equation to get: ； in, Each EEG sequence Before the moment past observations; Represented as the first The coefficient vector corresponding to the equation; Then the conversion relationship function between the states corresponding to the EEG data sequence is expressed as: ； in, represents the state transition matrix; represents the state noise, for t -1 moment condition t Estimates of the state coefficients at each moment.

4. The anxiety state assessment method according to claim 3, characterized in that: The method of solving the conversion relationship function to obtain the causal influence of the current EEG sequence on the target EEG sequence and the adaptive directed transfer function ATDF value of the causal influence at the set frequency point specifically includes the following steps: Solve the transformation relationship function using the least squares method , expressed as: ； in, , express R for t The covariance matrix of the moment-by-moment observation noise; according to , estimate the corresponding set frequency at the current moment ATDF value at: ； in, For the The time point corresponds to the The model coefficient of order is expressed as: ； And there is , for The inverse of the EEG sequence EEG sequence The normalized ATDF value of the causal effect at the frequency point is: ； in, For the moment , EEG sequence Target EEG sequence The causal effect of frequency ATDF value at .

5. The anxiety state assessment method according to claim 1, characterized in that: The method of performing a multiple regression analysis on the EEG network temporal characteristics of the EEG data sequences generated based on effective cue stimulation and invalid cue stimulation to evaluate the anxiety state of the user to be tested specifically includes the following steps: Using the evaluation features and the set anxiety score, a multiple regression analysis was performed on the EEG network temporal features of the EEG data sequences generated by effective cue stimulation and ineffective cue stimulation to train a multiple regression model; wherein the evaluation features include behavioral indicators, event-related potentials, state proportion, state interval time, and number of state switches; Use hidden Markov models to decode dynamic causal brain network sequences and obtain state time series; The state time series is calculated to obtain the state proportion, state interval time and state switching number; the state proportion is the ratio of the time each state appears to the total time; the state interval time is the interval time when the state appears continuously; the state switching number is the number of times the state sequence changes within a period of time; According to the state proportion, state interval time and state switching number, the anxiety state of the user to be tested is evaluated by the trained multivariate regression model.

6. An anxiety state assessment system, characterized in that: include: An acquisition module is used to acquire EEG data sequences generated by the user under test after effective cue stimulation and invalid cue stimulation through a spatial cue task test; The EEG network temporal feature extraction module is used to extract the EEG network temporal features with causal relationship for each EEG data sequence; The extraction process specifically includes: Extract the observation value of the EEG data sequence at a set time, establish a time-varying linear coupling relationship between the EEG data sequence and its corresponding observation value; obtain the conversion relationship function between the states corresponding to the EEG data sequence through the time-varying linear coupling relationship between the EEG data sequence and its observation value; obtain the causal influence of the current EEG sequence on the target EEG sequence and the adaptive directed transfer function ATDF value of the causal influence at the set frequency point by solving the conversion relationship function; estimate the dynamic causal brain network sequence through the ATDF value; and extract the EEG network time series features with causal relationship of the EEG data sequence based on the dynamic causal brain network sequence; The evaluation module is used to perform multiple regression analysis on the EEG network temporal characteristics of the EEG data sequences generated based on effective clue stimulation and invalid clue stimulation to evaluate the anxiety state of the user to be tested.

7. An anxiety state assessment computer device, characterized in that include: A memory, a processor, and a computer program stored in the memory, wherein when the processor executes the computer program, the steps of the anxiety state assessment method according to any one of claims 1 to 5 are implemented.

8. A readable storage medium, characterized in that: The readable storage medium stores a computer program, which includes program instructions. When the program instructions are executed by a processor, they are used to execute the steps of the anxiety state assessment method according to any one of claims 1 to 5.

Images