A brain microstate analysis method and device based on hidden Markov

Abstract

The present invention discloses a brain microstate analysis method and device based on hidden Markov. The original time-series EEG signal data of autism groups and healthy development groups are selected as observation sequence data, and the analysis of hidden brain microstates is completed through the construction and decoding process of the hidden Markov model. For the acquired brain microstate hidden process sequence, a series of statistical indicators in the time domain dimension and the frequency domain dimension are calculated. Finally, the change process and dynamic activity pattern of the brain microstate between different pathological groups are quantitatively described and compared, forming a neural marker conclusion for analysis reference. The present invention proposes an EEG data modeling method that converts the observation perspective from the static average brain structure to the dynamic time-varying brain microstate structure. It can provide a millisecond-scale brain state analysis model from an individual perspective. Through accurate analysis results and reliable descriptive indicators, the brain activity patterns of different pathological states can be effectively interpreted and compared.

Description

Technical Field

[0001] The present invention relates to the field of computer technology and is a method for analyzing brain microstates based on resting-state electroencephalogram (EEG). The present invention specifically relates to a method for analyzing latent brain microstates based on a hidden Markov model and a method for constructing neural markers for reference and comparative analysis between different pathological groups. The present invention particularly relates to a method and device for analyzing brain microstates based on a hidden Markov model. Background Art

[0002] Autism is a spectrum disorder with unknown pathological mechanisms, and is heterogeneous and complex. Therefore, in the clinical medical evaluation process, it is necessary to find neuromarker indicators that can be used for reference and comparative analysis from the perspective of brain development to evaluate the brain state at the individual level. Electroencephalogram (EEG) data is a potential curve that records brain activity over time using electrodes on the surface of the brain. It is a window for observing cerebral cortical activity and is key to the coordinated work between brain regions. A full understanding of EEG signals can comprehensively and systematically explore the mysteries of the brain from the perspective of neurons and electrophysiological brain signals, and then conduct in-depth analysis of brain function and the patterns of brain activity.

[0003] Traditional index extraction and calculation methods based on EEG feature analysis focus on the average functional localization of brain oscillation patterns in each brain region or frequency band, and are unable to explore and discover the dynamic activity patterns of large-scale brain network functions. Brain microstate analysis methods, on the other hand, can shift the perspective from static average brain structure to dynamic, time-varying brain microstate structure. They can capture recurring microscopic dynamic activity patterns in the brain and capture the dynamic temporal changes of EEG that integrate spatial information. This provides a millisecond-scale brain state analysis method for individuals with illnesses and is currently one of the effective and reliable indicators for studying brain dysfunction and neural function patterns.

[0004] Existing research methods based on resting-state brain microstates focus on single-subject microstate identification using clustering methods. These methods attempt to decompose EEG time series data into a set of brain microstate evolutions with similar topological structures. However, this clustering method based on the topological structure of the inter-channel voltage standard deviation destroys the spatial discreteness of the brain. Furthermore, this similarity-based clustering method leads to a certain degree of randomness in the results, and requires a large number of iterative experiments to select relatively representative results. Summary of the Invention

[0005] In order to adaptively analyze the multivariate dynamic EEG characteristics of the resting-state brain of autistic and normal subjects that change over time, capture brain microstructural patterns at the individual and group levels, link them to corresponding brain dysfunctions, and establish rich and reliable neural marker indicators, the present invention proposes a brain microstate time series process analysis method based on a hidden Markov model. Brain microstates are modeled from the perspective of a multivariate Gaussian distribution observation model, and instantaneous and recurring brain microstate structures are adaptively extracted in a completely data-driven manner. Descriptive indicators in the time domain, frequency domain, and spatial domain are obtained on a millisecond time scale. The brain activity patterns of different pathological states are interpreted and compared through the analysis results and descriptive indicators, thereby improving the analysis effect.

[0006] In order to achieve the above object, the technical solution of the present invention is:

[0007] The first aspect provides a brain microstate analysis method based on Hidden Markov Model, comprising:

[0008] S1: Obtain the original EEG data of healthy and diseased groups, and obtain clean time-series EEG signal data through a series of preprocessing operations;

[0009] S2: Based on the individual level, the clean time-series EEG signal data is used as the observation sequence in the hidden Markov modeling process to complete the construction of the hidden Markov model;

[0010] S3: Define the hidden brain microstate categories as four resting-state brain microstate categories and train the constructed hidden Markov model;

[0011] S4: Decoding the trained hidden Markov model to obtain the state time course sequences of four different resting-state brain microstate categories;

[0012] S5: Calculate statistical indicators of the time domain dimension based on the group level for the state time course sequences of the four different resting-state brain microstate categories obtained by the decoding operation;

[0013] S6: Calculate frequency domain statistical indicators based on the population level for the time course sequences of the four different resting-state brain microstate categories obtained through decoding.

[0014] S7: Interpret and compare statistical indicators between healthy and diseased groups, and construct neurobiological markers to analyze the differences in brain activity patterns under different pathological conditions.

[0015] In one embodiment, step S2 includes:

[0016] S2.1: Define the brain microstate discrimination hidden Markov model λ = (N, M, A, B, π), where N is the number of hidden states, M is the number of possible observations for each state, A is the time-independent state transition probability matrix, B is the probability distribution matrix of observations in a given state, π is the probability distribution of the initial state space, and the set of all hidden state spaces is The set of all observable sequence spaces is Right now The hidden state sequence I = {i1, i2, ..., i T The corresponding observation sequence is O = {o1, o2, ..., o T};

[0017] S2.2: Define the assumptions of the hidden Markov model, which include the homogeneous Markov assumption and the observation independence assumption. The homogeneous Markov assumption means that in a series of events, the probability of a given event occurring depends only on the event that occurred at the previous moment, that is, the state i at any time t t Only depends on the state i at the previous time t-1 t-1 , which is independent of the state and observation at other times. The observation independence hypothesis means that the observation value o at any time t t Only depends on the Markov chain state i at the current time t t , which is independent of other states and observations.

[0018] In one embodiment, step S3 includes:

[0019] The short-term recurring brain state is defined as a brain microstate, which is a finite set of hidden states that cannot be directly observed and need to be inferred. The time-series EEG signal data is used as the observation sequence of the hidden Markov model, which is a result that can be directly observed and serves as the observation space set for decoding the hidden state sequence.

[0020] Define the data of time series EEG signal data at different time points The brain microstate corresponding to time t is x t ∈{1, 2, ..., K};

[0021] The observed sequence EEG time series signal data y t Input the hidden Markov model and infer the most likely brain microstate hidden state sequence x t , at each time point t each state x t There is a corresponding active probability p t .

[0022] In one embodiment, step S4 includes:

[0023] The brain microstate time process sequence is extracted through the hidden Markov decoding process, and the data probability observation pattern of the brain microstate is represented by a multivariate Gaussian distribution:

[0024]

[0025] where μ k is the mean matrix of the data, ∑ k It is a covariance matrix that encodes the variance and covariance between channels. The multivariate Gaussian distribution is used to complete the modeling process of a single brain microstate based on the characteristics between multiple channels. Each brain microstate is characterized by the parameters of the multivariate Gaussian distribution.

[0026] In one embodiment, step S5 includes:

[0027] Calculate the total number of times each brain microstate appears during the entire recording time, that is, the number of times a single brain microstate category appears;

[0028] Calculate the total proportion of coverage when each microstate dominates the entire recording time, that is, the proportion of state coverage that occurs in a single brain microstate category;

[0029] Calculate the average duration of a brain microstate remaining stably active after it appears before switching to another microstate, that is, the average duration of a single brain microstate category;

[0030] Calculate the average time interval between the continuous active appearance of a specific brain microstate, that is, the average interval time between the appearance of a single brain microstate category.

[0031] In one embodiment, step S6 includes:

[0032] S6.1: Map the time course sequences corresponding to different resting-state brain microstate categories obtained in step S4 to different time periods in the original time-series EEG data, and calculate parameter indicators corresponding to the frequency domain dimension for different time period windows;

[0033] S6.2: Calculate the statistical information of the original EEG signal data in the frequency domain within the mapped time window, obtain the spectrum estimate of the EEG signal data using a multi-window spectrum analysis method, generate a series of windowed data using a discrete long spherical sequence composed of a series of orthogonal conical windows, and calculate the average value of the periodogram of these windowed data as the spectrum estimate of the signal;

[0034] S6.3: Based on the spectral estimation values ​​based on the multi-window spectrum, the coherence calculation method is used to measure the degree of linear correlation between different brain regions in the frequency domain to complete the calculation of brain functional connectivity indicators, which serves as a representation model of collaborative communication between different brain regions.

[0035] In one embodiment, step S7 includes:

[0036] S7.1: Based on the calculation results of steps S5 and S6, perform summary and average statistics for individuals within the healthy and diseased groups, respectively, to obtain statistical information and parameter indicators at the corresponding group level, including the time course series of the four brain microstate categories, the time dimension parameters of individual brain microstates, and the frequency domain dimension parameters of individual brain microstates;

[0037] S7.2: Compare and analyze different groups of parameters between different groups to obtain differences in brain cognitive function status related to pathological conditions and form reference neural marker conclusions.

[0038] Based on the same inventive concept, the second aspect of the present invention provides a brain microstate analysis device based on Hidden Markov Model, comprising:

[0039] The data acquisition and preprocessing module is used to obtain the original EEG data of healthy and diseased groups, and obtain clean time-series EEG signal data through a series of preprocessing operations;

[0040] The model building module is used to use the clean time-series EEG signal data obtained at the individual level as the observation sequence in the hidden Markov modeling process to complete the construction of the hidden Markov model;

[0041] A model training module is used to define hidden brain microstate categories as four resting-state brain microstate categories and train the constructed hidden Markov model;

[0042] A decoding module is used to decode the trained hidden Markov model to obtain state time process sequences of four different resting-state brain microstate categories;

[0043] The time-domain dimension statistical indicator calculation module is used to calculate the time-domain dimension statistical indicators based on the group level for the state time process sequences of four different resting-state brain microstate categories obtained by the decoding operation;

[0044] The frequency domain dimension statistical index calculation module is used to calculate the frequency domain dimension statistical index based on the group level for the state time process series of four different resting-state brain microstate categories obtained by the decoding operation;

[0045] The comparative analysis module is used to interpret and compare statistical indicators between healthy and diseased groups, and to construct neurobiological markers to analyze the differences in brain activity patterns under different pathological conditions.

[0046] Based on the same inventive concept, the third aspect of the present invention provides a computer-readable storage medium having a computer program stored thereon, which implements the method described in the first aspect when the program is executed.

[0047] Based on the same inventive concept, the fourth aspect of the present invention provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method described in the first aspect when executing the program.

[0048] Compared with the prior art, the advantages and beneficial technical effects of the present invention are as follows:

[0049] 1) This invention models EEG data through brain microstate analysis. This analysis method shifts the perspective from static average brain structure to dynamic, time-varying brain microstate structure. This allows for capturing recurring microscopic patterns of dynamic activity in the brain, providing a millisecond-scale approach for analyzing brain states in individuals. By analyzing the differences in brain microstates between healthy and diseased groups, a rich and reliable set of neuromarker indicators can be established.

[0050] 2) This invention analyzes brain microstate time series processes using a hidden Markov model, modeling brain microstates from the perspective of a multivariate Gaussian distribution observation model. This method adaptively extracts transient and recurring brain microstate structures in a completely data-driven manner, acquiring a variety of descriptive indicators in the time, frequency, and spatial domains on a millisecond timescale. These indicators include the average duration of each brain microstate category, state coverage percentage, average interval time, spectral estimation, and coherence analysis between different channels. Through precise analysis results and reliable descriptive indicators, it can effectively interpret and compare brain activity patterns in different pathological states. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.

[0052] Figure 1 This is a flowchart of a brain microstate analysis method based on Hidden Markov Model disclosed in the implementation of the present invention;

[0053] Figure 2 This is a structural diagram of the model for brain microstate analysis based on resting-state EEG provided in the implementation of the present invention. DETAILED DESCRIPTION

[0054] This invention discloses a method for latent brain microstate analysis and neural marker construction based on a hidden Markov model (HMM) for comparison and reference between different pathological groups. Raw time-series EEG signal data from autistic and healthy individuals are selected as observation sequence data. Hidden brain microstate analysis is achieved through the construction and decoding of a HMM. Based on the acquired latent brain microstate process sequence, a series of statistical indicators in the time and frequency domains are calculated, including the duration, interval, and coverage of individual brain microstate categories, as well as the transition process and transition probability between different brain microstates, and the frequency domain activity of the corresponding individual brain microstate categories. Finally, the changing processes and dynamic activity patterns of brain microstates across different pathological groups are quantitatively described and compared, generating neural marker conclusions for reference analysis. This invention proposes an EEG data modeling method that shifts the observation perspective from static average brain structure to dynamic, time-varying brain microstate structure. This method provides a millisecond-scale brain state analysis model from an individual perspective. Through accurate analysis results and reliable descriptive indicators, it can effectively interpret and compare brain activity patterns across different pathological states.

[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention more clear, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the present invention.

[0056] Example 1

[0057] This embodiment provides a brain microstate analysis method based on Hidden Markov. Figure 1 , the method comprising:

[0058] S1: Obtain the original EEG data of healthy and diseased groups, and obtain clean time-series EEG signal data through a series of preprocessing operations;

[0059] S2: Based on the individual level, the clean time-series EEG signal data is used as the observation sequence in the hidden Markov modeling process to complete the construction of the hidden Markov model;

[0060] S3: Define the hidden brain microstate categories as four resting-state brain microstate categories and train the constructed hidden Markov model;

[0061] S4: Decoding the trained hidden Markov model to obtain the state time course sequences of four different resting-state brain microstate categories;

[0062] S5: Calculate statistical indicators of the time domain dimension based on the group level for the state time course sequences of the four different resting-state brain microstate categories obtained by the decoding operation;

[0063] S6: Calculate frequency domain statistical indicators based on the population level for the time course sequences of the four different resting-state brain microstate categories obtained through decoding.

[0064] S7: Interpret and compare statistical indicators between healthy and diseased groups, and construct neurobiological markers to analyze the differences in brain activity patterns under different pathological conditions.

[0065] Specifically, step S5 is to quantitatively describe the changes in brain microstates by calculating a series of statistical indicators in the time domain dimension for the acquired latent process sequence of brain microstates, so as to facilitate subsequent comparative analysis between different pathological groups. Step S6 is to quantitatively describe the changes in brain microstates by calculating a series of statistical indicators in the frequency domain dimension for the acquired latent process sequence of brain microstates, so as to facilitate subsequent comparative analysis between different pathological groups.

[0066] In the specific implementation process, the preprocessing of the raw EEG data of the healthy group and the diseased group can be achieved by the following methods:

[0067] Step S1.1: Referring to the 128-channel electrode distribution diagram in the international standard 10 / 10 system, the 8-channel electrodes corresponding to the original 8-channel EEG data obtained by the present invention are mapped to different cerebral hemispheres and brain regions. The corresponding mapping relationship is as follows: F3 - left frontal lobe, F4 - right frontal lobe, T3 - left temporal lobe, C3 - central region, C4 - central region, T4 - right temporal lobe, O1 - left occipital lobe, O2 - right occipital lobe.

[0068] Step S1.2: Preprocess the EEG data using a standardized preprocessing process for large-scale EEG analysis. This removes the influence of the acquisition instrument and environment, facilitating the construction of a large EEG database. Divide the EEG into one-second segments, remove bad channels in each segment based on a threshold, and replace the bad channel signals with whole-brain signal fitting, thus becoming a global reference.

[0069] Step S1.3: Use the EEGLAB toolbox in MATLAB to observe the forming data, manually screen and check, and remove some interference signals.

[0070] In one embodiment, step S2 includes:

[0071] S2.1: Define the brain microstate discrimination hidden Markov model λ = (N, M, A, B, π), where N is the number of hidden states, M is the number of possible observations for each state (i.e., the length of the observation sequence), A is the time-independent state transition probability matrix, B is the probability distribution matrix of the observations in a given state, π is the probability distribution of the initial state space, and the set of all hidden state spaces is The set of all observable sequence spaces is Right now where q1, q2, ..., q N denote the first, second and Nth hidden states, {v1, v2, ..., v M} represents an observable sequence of length M, and a hidden state sequence of length T I = {i1, i2, ..., i T The corresponding observation sequence is O = {o1, o2, ..., o T};

[0072] S2.2: Define the assumptions of the hidden Markov model, which include the homogeneous Markov assumption and the observation independence assumption. The homogeneous Markov assumption means that in a series of events, the probability of a given event occurring depends only on the event that occurred at the previous moment, that is, the state i at any time t t Only depends on the state i at the previous time t-1 t-1 , which is independent of the state and observation at other times. The observation independence hypothesis means that the observation value o at any time t t Only depends on the Markov chain state i at the current time t t , which is independent of other states and observations.

[0073] See Figure 2 , is a structural diagram of the model for brain microstate analysis based on resting-state EEG provided in the implementation of the present invention.

[0074] Specifically, the hidden Markov model is determined by the initial state probability distribution, the state transition probability distribution, and the observation probability distribution.

[0075] Among them, the state transition probability matrix is:

[0076]

[0077] Where a ij Indicates that at the current time t, it is in state q i Under the condition that at the next moment t+1, it transfers to state q j Probability of:

[0078] a ij =P(i t+1 =q j|i t =q i )i=1, 2, ..., N; j=1, 2, ..., N (2)

[0079] The observation probability matrix is:

[0080]

[0081] Where b i (k) means that at the current time t, i When the state is , the observation value v is generated k The probability that given i t =q i Under the conditions of t =v k Probability of:

[0082] b i (k)=P(o t =v k |i t =q i )t=1, 2,...; i=1, 2,..., N; k=1, 2,..., M (4)

[0083] The initial state probability distribution is:

[0084] π=(π1,π2,...,π N ) T (5)

[0085] Where, π i Indicates that at time t it is in state q i Probability: π i =P(i t =q i ), and the conditions are met

[0086] Regarding the assumptions of the hidden Markov model, one is the homogeneous Markov assumption, which states that the probability of a given event occurring in a series of events depends only on the event that occurred at the previous moment, that is, the state i at any time t t Only depends on the state i at the previous time t-1 t-1 , which is independent of the state and observation at other times:

[0087] P(i t |i t-1 ,i t-2 ,...,i1,o t ,...,o1)=P(i t |i t-1), t=2,3,...,T (6)

[0088] The second is the observation independence assumption, that is, the observation value o at any time t t Only depends on the Markov chain state i at the current time t t , independent of other states and observations:

[0089] P(o t |i T , o T ,...,i t , o t ,i t-1 , o t-1 ,...,i1,o1)=P(o t |i t ) (7)

[0090] In one embodiment, step S3 includes:

[0091] The short-term recurring brain state is defined as a brain microstate, which is a finite set of hidden states that cannot be directly observed and need to be inferred. The time-series EEG signal data is used as the observation sequence of the hidden Markov model, which is a result that can be directly observed and serves as the observation space set for decoding the hidden state sequence.

[0092] Define the data of time series EEG signal data at different time points The brain microstate corresponding to time t is x t ∈{1, 2, ..., K};

[0093] The observed sequence EEG time series signal data y t Input the hidden Markov model and infer the most likely brain microstate hidden state sequence x t , at each time point t each state x t There is a corresponding active probability p t .

[0094] Specifically, based on the construction of the hidden Markov model, the model is trained, and the training process is the process of modeling the EEG time series signal data.

[0095] In one embodiment, step S4 includes:

[0096] The brain microstate time process sequence is extracted through the hidden Markov decoding process, and the data probability observation pattern of the brain microstate is represented by a multivariate Gaussian distribution:

[0097]

[0098] where μ kis the mean matrix of the data, ∑ k It is a covariance matrix that encodes the variance and covariance between channels. The multivariate Gaussian distribution is used to complete the modeling process of a single brain microstate based on the characteristics between multiple channels. Each brain microstate is characterized by the parameters of the multivariate Gaussian distribution.

[0099] In one embodiment, step S5 includes:

[0100] Calculate the total number of times each brain microstate appears during the entire recording time, that is, the number of times a single brain microstate category appears;

[0101] Calculate the total proportion of coverage when each microstate dominates the entire recording time, that is, the proportion of state coverage that occurs in a single brain microstate category;

[0102] Calculate the average duration of a brain microstate remaining stably active after it appears before switching to another microstate, that is, the average duration of a single brain microstate category;

[0103] Calculate the average time interval between the continuous active appearance of a specific brain microstate, that is, the average interval time between the appearance of a single brain microstate category.

[0104] In the specific implementation process, the specific steps for calculating statistical indicators in the time domain dimension include:

[0105] Step S3.1: Calculate the number of state occurrences. That is, calculate the total number of times each brain microstate appears during the entire recording time:

[0106] OC(k)=∑ t (((x t = = k)-(x t-1 = = k)) = = 1) (9)

[0107] where x t Indicates the brain microstate category corresponding to the EEG signal data at the current moment, x t-1 Indicates the brain microstate category corresponding to the EEG signal data at the previous moment. Calculates the frequency of occurrence of a single microstate in the overall microstate sequence.

[0108] Step S3.2: Calculate the state coverage ratio. That is, calculate the total coverage ratio of each microstate when it dominates the overall recording time:

[0109]

[0110] Where T is the total length of the brain microstate time series data, and the main active state of the brain at the current time t is microstate k when x tThe k value is 1. The ratio of the duration of a single brain microstate to the duration of the overall microstate time series is calculated.

[0111] Step S3.3: Calculate the average duration. This means calculating the average duration that a brain microstate remains stable and active after it appears before switching to another microstate:

[0112]

[0113] where ∑ t (x t = = k) represents the sum of all durations of the brain in microstate category k, and OC(k) is the result calculated in step 3.1. Calculate the ratio of the duration of a single brain microstate to the total number of occurrences of the corresponding microstate analogy.

[0114] Step 3.4: Calculate the average interval time. That is, calculate the average time interval between consecutive active occurrences of a specific brain microstate:

[0115]

[0116] Where T is the total length of the brain microstate time series data, and OC(k) is the result calculated in step 3.1. The difference between the coverage time of a single brain microstate category and the overall time is averaged.

[0117] In one embodiment, step S6 includes:

[0118] S6.1: Map the time course sequences corresponding to different resting-state brain microstate categories obtained in step S4 to different time periods in the original time-series EEG data, and calculate parameter indicators corresponding to the frequency domain dimension for different time period windows;

[0119] S6.2: Calculate the statistical information of the original EEG signal data in the frequency domain within the mapped time window, obtain the spectrum estimate of the EEG signal data using a multi-window spectrum analysis method, generate a series of windowed data using a discrete long spherical sequence composed of a series of orthogonal conical windows, and calculate the average value of the periodogram of these windowed data as the spectrum estimate of the signal;

[0120] S6.3: Based on the spectral estimation values ​​based on the multi-window spectrum, the coherence calculation method is used to measure the degree of linear correlation between different brain regions in the frequency domain to complete the calculation of brain functional connectivity indicators, which serves as a representation model of collaborative communication between different brain regions.

[0121] Regarding the calculation process of statistical indicators in the frequency domain dimension, the specific steps include:

[0122] S6.1: Based on the time process sequences corresponding to different brain microstate categories obtained above, they are mapped to different time periods in the original time series EEG data, and parameter indicators corresponding to the frequency domain dimension are calculated for different time period windows.

[0123] S6.2: Calculate the frequency domain representation of the EEG signal data, namely, the distribution of signal power along frequency in the time series signal. Estimate the spectral density of a random process from a series of time samples of the random process to obtain a spectrum estimate. The present invention utilizes a multi-window spectrum analysis method to obtain a spectrum estimate of the EEG signal data. A discrete prolate spherical sequence composed of a series of orthogonal conical windows is used to generate a series of windowed data. The average value of the periodogram of these windowed data is calculated as the spectrum estimate of the signal.

[0124] The multi-window spectrum is defined as:

[0125]

[0126] Where C is the number of data windows, m is the mth sequence of the signal, t is time, ω is frequency, is the spectrum of the kth data window, where k=0...C-1.

[0127] is calculated as follows:

[0128]

[0129] In the formula, x(n) is the data sequence, N is the data length, e -jnω is the complex exponential form of the nth segment signal, a k (n) is the kth data window, and any number of data windows are orthogonal to each other, that is,

[0130]

[0131] where a i (n) and a j (n) represents an unrelated data window. If i≠j, then the two different data windows are orthogonal.

[0132] Therefore, the series of tapered windows used in the multi-window spectrum calculation method of the present invention have the characteristics of orthogonality and also have the best time-frequency concentration characteristics, and the obtained spectrum estimation has small variance and high frequency resolution.

[0133] Step S6.3: Based on the spectrum estimation value based on the multi-window spectrum, the present invention uses the coherence calculation method to measure the linear correlation degree between different brain regions in the frequency domain to complete the calculation of brain functional connectivity index. First, the cross-correlation between different channels is calculated:

[0134]

[0135] Where H represents the number of EEG signal data channels, and are the means corresponding to different channel sequences x(i) and y(i), σ x and σ y is the corresponding variance, o xy ∈[0,1]. The cross-correlation calculation obtains the linear synchronization between the sequence data of different channels, where a value of 1 indicates that the synchronization effect between the two channels is the strongest.

[0136] Calculate the coherence functional connectivity index:

[0137]

[0138] Where, O xy (f), O xx (f) and O yy (f) are the spectra after Fourier transform of the cross-correlation calculation results.

[0139] In one embodiment, step S7 includes:

[0140] S7.1: Based on the calculation results of steps S5 and S6, perform summary and average statistics for individuals within the healthy and diseased groups, respectively, to obtain statistical information and parameter indicators at the corresponding group level, including the time course series of the four brain microstate categories, the time dimension parameters of individual brain microstates, and the frequency domain dimension parameters of individual brain microstates;

[0141] S7.2: Compare and analyze different groups of parameters between different groups to obtain differences in brain cognitive function status related to pathological conditions and form reference neural marker conclusions.

[0142] The present invention has the following positive effects and advantages:

[0143] 1) This invention models EEG data through brain microstate analysis. This analysis method shifts the perspective from static average brain structure to dynamic, time-varying brain microstate structure. This allows for capturing recurring microscopic patterns of dynamic activity in the brain, providing a millisecond-scale approach for analyzing brain states. By analyzing the differences in brain microstates between healthy and diseased groups, a rich and reliable set of neuromarker indicators can be established.

[0144] 2) This invention analyzes brain microstate time series processes using a hidden Markov model, modeling brain microstates from the perspective of a multivariate Gaussian distribution observation model. This method adaptively extracts transient and recurring brain microstate structures in a completely data-driven manner, acquiring a variety of descriptive indicators in the time, frequency, and spatial domains on a millisecond timescale. These indicators include the average duration of each brain microstate category, state coverage percentage, average interval time, spectral estimation, and coherence analysis between different channels. Through precise analysis results and reliable descriptive indicators, it can effectively interpret and compare brain activity patterns in different pathological states.

[0145] Example 2

[0146] Based on the same inventive concept, the present invention discloses a brain microstate analysis device based on Hidden Markov Model, comprising:

[0147] The data acquisition and preprocessing module is used to obtain the original EEG data of healthy and diseased groups, and obtain clean time-series EEG signal data through a series of preprocessing operations;

[0148] The model building module is used to use the clean time-series EEG signal data obtained at the individual level as the observation sequence in the hidden Markov modeling process to complete the construction of the hidden Markov model;

[0149] A model training module is used to define hidden brain microstate categories as four resting-state brain microstate categories and train the constructed hidden Markov model;

[0150] A decoding module is used to decode the trained hidden Markov model to obtain state time process sequences of four different resting-state brain microstate categories;

[0151] The time-domain dimension statistical indicator calculation module is used to calculate the time-domain dimension statistical indicators based on the group level for the state time process sequences of four different resting-state brain microstate categories obtained by the decoding operation;

[0152] The frequency domain dimension statistical index calculation module is used to calculate the frequency domain dimension statistical index based on the group level for the state time process series of four different resting-state brain microstate categories obtained by the decoding operation;

[0153] The comparative analysis module is used to interpret and compare statistical indicators between healthy and diseased groups, and to construct neurobiological markers to analyze the differences in brain activity patterns under different pathological conditions.

[0154] Since the device described in Example 2 of the present invention is used to implement the brain microstate analysis method based on Hidden Markov Model in Example 1 of the present invention, the specific structure and variations of the device are well understood by those skilled in the art based on the method described in Example 1 of the present invention, and therefore will not be described in detail here. All devices used in the method of Example 1 of the present invention fall within the scope of protection of the present invention.

[0155] Example 3

[0156] Based on the same inventive concept, the present invention further provides a computer-readable storage medium on which a computer program is stored. When the program is executed, the method described in the first embodiment is implemented.

[0157] Since the computer-readable storage medium described in the third embodiment of the present invention is the computer-readable storage medium used to implement the brain microstate analysis method based on the hidden Markov model described in the first embodiment of the present invention, the specific structure and variations of the computer-readable storage medium are understood by those skilled in the art based on the method described in the first embodiment of the present invention, and therefore are not further described here. All computer-readable storage media used in the method of the first embodiment of the present invention fall within the scope of protection of the present invention.

[0158] Example 4

[0159] Based on the same inventive concept, the present application also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the method in the first embodiment is implemented.

[0160] Since the computer device described in Example 4 of the present invention is the computer device used to implement the brain microstate analysis method based on Hidden Markov Model in Example 1 of the present invention, the specific structure and variations of the computer device are readily understood by those skilled in the art based on the method described in Example 1 of the present invention, and therefore will not be described in detail here. All computer devices used in the method of Example 1 of the present invention fall within the scope of protection of the present invention.

[0161] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0162] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0163] Although preferred embodiments of the present invention have been described, those skilled in the art may make additional changes and modifications to these embodiments once they are aware of the basic inventive concepts. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present invention. Obviously, those skilled in the art may make various changes and modifications to the embodiments of the present invention without departing from the spirit and scope of the embodiments of the present invention. Thus, the present invention is intended to include such changes and modifications to the embodiments of the present invention if they fall within the scope of the claims and their equivalents.

Claims

1. A brain microstate analysis method based on Hidden Markov Model, characterized in that: include: S1: Obtain the original EEG data of healthy and diseased groups, and obtain clean time-series EEG signal data through a series of preprocessing operations; S2: Based on the individual level, the clean time-series EEG signal data is used as the observation sequence in the hidden Markov modeling process to complete the construction of the hidden Markov model; S3: Define the hidden brain microstate categories as four resting-state brain microstate categories and train the constructed hidden Markov model; S4: Decoding the trained hidden Markov model to obtain the state time course sequences of four different resting-state brain microstate categories; S5: Calculate statistical indicators of the time domain dimension based on the group level for the state time course sequences of the four different resting-state brain microstate categories obtained by the decoding operation; S6: Calculate frequency domain statistical indicators based on the population level for the time course sequences of the four different resting-state brain microstate categories obtained through decoding. S7: Interpret and compare statistical indicators between healthy and diseased groups, and construct neurobiological markers to analyze the differences in brain activity patterns under different pathological conditions.

2. The brain microstate analysis method based on Hidden Markov Model according to claim 1, characterized in that: Step S2 includes: S2.1: Defining a Hidden Markov Model for Brain Microstate Discrimination ,in is the number of hidden states, is the number of possible observations for each state, is the time-independent state transition probability matrix, is the probability distribution matrix of observations at a given state, is the probability distribution of the initial state space, and the set of all hidden state spaces is , the set of all observable sequence spaces is ,Right now , , the length is The hidden state sequence of The corresponding observation sequence is ; S2.2: Define the assumptions of the hidden Markov model, which include the homogeneous Markov assumption and the observation independence assumption. The homogeneous Markov assumption means that in a series of events, the probability of a given event occurring depends only on the event that occurred at the previous moment, that is, at any moment Status Depends only on the previous moment Status , which is independent of the state and observation at other times. The observation independence hypothesis means that at any time Observed values Depends only on the current moment The Markov chain state , which is independent of other states and observations.

3. The brain microstate analysis method based on Hidden Markov Model according to claim 1, characterized in that: Step S3 includes: The short-term recurring brain state is defined as a brain microstate, which is a finite set of hidden states that cannot be directly observed and need to be inferred. The time-series EEG signal data is used as the observation sequence of the hidden Markov model, which is a result that can be directly observed and serves as the observation space set for decoding the hidden state sequence. Define the data of time series EEG signal data at different time points , corresponding to time The brain microstate on ; Observe sequential EEG signal data Input the hidden Markov model to infer the most likely hidden state sequence of brain microstates , at each time point Each state There is a corresponding active probability .

4. The brain microstate analysis method based on Hidden Markov Model according to claim 3, characterized in that: Step S4 includes: The brain microstate time process sequence is extracted through the hidden Markov decoding process, and the data probability observation pattern of the brain microstate is represented by a multivariate Gaussian distribution: in is the mean matrix of the data, is the covariance matrix that encodes the variance and covariance between channels, Represents the brain microstate category, and completes the modeling process of a single brain microstate based on the characteristics between multiple channels through multivariate Gaussian distribution. Each brain microstate is characterized by the parameters of the multivariate Gaussian distribution.

5. The brain microstate analysis method based on Hidden Markov Model according to claim 1, characterized in that: Step S5 includes: Calculate the total number of times each brain microstate appears during the entire recording time, that is, the number of times a single brain microstate category appears; Calculate the total proportion of coverage when each microstate dominates the entire recording time, that is, the proportion of state coverage that occurs in a single brain microstate category; Calculate the average duration of a brain microstate remaining stably active after it appears before switching to another microstate, that is, the average duration of a single brain microstate category; Calculate the average time interval between the continuous active appearance of a specific brain microstate, that is, the average interval time between the appearance of a single brain microstate category.

6. The brain microstate analysis method based on Hidden Markov Model according to claim 1, characterized in that: Step S6 includes: S6.1: Map the time course sequences corresponding to different resting-state brain microstate categories obtained in step S4 to different time periods in the original time-series EEG data, and calculate parameter indicators corresponding to the frequency domain dimension for different time period windows; S6.2: Calculate the statistical information of the original EEG signal data in the frequency domain within the mapped time window, obtain the spectrum estimate of the EEG signal data using a multi-window spectrum analysis method, generate a series of windowed data using a discrete long spherical sequence composed of a series of orthogonal conical windows, and calculate the average value of the periodogram of these windowed data as the spectrum estimate of the signal; S6.3: Based on the spectral estimation values ​​based on the multi-window spectrum, the coherence calculation method is used to measure the degree of linear correlation between different brain regions in the frequency domain to complete the calculation of brain functional connectivity indicators, which serves as a representation model of collaborative communication between different brain regions.

7. The brain microstate analysis method based on Hidden Markov Model according to claim 1, characterized in that: Step S7 includes: S7.1: Based on the calculation results of steps S5 and S6, perform summary and average statistics for individuals within the healthy and diseased groups, respectively, to obtain statistical information and parameter indicators at the corresponding group level, including the time course series of the four brain microstate categories, the time dimension parameters of individual brain microstates, and the frequency domain dimension parameters of individual brain microstates; S7.2: Compare and analyze different groups of parameters between different groups to obtain differences in brain cognitive function status related to pathological conditions and form reference neural marker conclusions.

8. A brain microstate analysis device based on Hidden Markov, characterized in that: include: The data acquisition and preprocessing module is used to obtain the original EEG data of healthy and diseased groups, and obtain clean time-series EEG signal data through a series of preprocessing operations; The model building module is used to use the clean time-series EEG signal data obtained at the individual level as the observation sequence in the hidden Markov modeling process to complete the construction of the hidden Markov model; A model training module is used to define hidden brain microstate categories as four resting-state brain microstate categories and train the constructed hidden Markov model; A decoding module is used to decode the trained hidden Markov model to obtain state time process sequences of four different resting-state brain microstate categories; The time-domain dimension statistical indicator calculation module is used to calculate the time-domain dimension statistical indicators based on the group level for the state time process sequences of four different resting-state brain microstate categories obtained by the decoding operation; The frequency domain dimension statistical index calculation module is used to calculate the frequency domain dimension statistical index based on the group level for the state time process series of four different resting-state brain microstate categories obtained by the decoding operation; The comparative analysis module is used to interpret and compare statistical indicators between healthy and diseased groups, and to construct neurobiological markers to analyze the differences in brain activity patterns under different pathological conditions.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed, the method according to any one of claims 1 to 7 is implemented.

10. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the program, the method according to any one of claims 1 to 7 is implemented.

Images