CN112690775B - Bayes-based imaging system for focal zone with abnormal brain activity of children - Google Patents

Bayes-based imaging system for focal zone with abnormal brain activity of children Download PDF

Info

Publication number
CN112690775B
CN112690775B CN202011553134.8A CN202011553134A CN112690775B CN 112690775 B CN112690775 B CN 112690775B CN 202011553134 A CN202011553134 A CN 202011553134A CN 112690775 B CN112690775 B CN 112690775B
Authority
CN
China
Prior art keywords
abnormal
brain
activity
bayesian
sampling data
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Active
Application number
CN202011553134.8A
Other languages
Chinese (zh)
Other versions
CN112690775A (en
Inventor
蔡畅
陈靓影
杨宗凯
徐如意
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Central China Normal University
Original Assignee
Central China Normal University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Central China Normal University filed Critical Central China Normal University
Priority to CN202011553134.8A priority Critical patent/CN112690775B/en
Publication of CN112690775A publication Critical patent/CN112690775A/en
Application granted granted Critical
Publication of CN112690775B publication Critical patent/CN112690775B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves 
    • A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves  involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/0033—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room
    • A61B5/004—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part
    • A61B5/0042—Features or image-related aspects of imaging apparatus classified in A61B5/00, e.g. for MRI, optical tomography or impedance tomography apparatus; arrangements of imaging apparatus in a room adapted for image acquisition of a particular organ or body part for the brain
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
    • A61B5/7235—Details of waveform analysis
    • A61B5/7246—Details of waveform analysis using correlation, e.g. template matching or determination of similarity

Abstract

The invention discloses a Bayes-based imaging system for a focal zone with abnormal brain activity of children. The system comprises: the signal processing module is used for acquiring electroencephalogram magnetic sampling data and preprocessing the electromagnetic sampling data; the brain activity abnormity identification module is used for identifying abnormal waves and non-abnormal waves from the preprocessed electroencephalogram and magnetic sampling data; a noise estimation module for estimating background interference noise from the non-abnormal waves; the brain source activity reconstruction module is used for combining the abnormal wave and the estimated background interference noise and acquiring the focal region position of the brain abnormal activity and the brain source activity time sequence of the focal region by adopting a Bayesian estimation method; and the imaging module is used for imaging according to the brain source activity. The method can greatly improve the accuracy of localization and reconstruction of the focal zone, and is particularly suitable for imaging the focal zone with abnormal brain electrical activity of children with autism.

Description

Bayes-based imaging system for focal zone with abnormal brain activity of children
Technical Field
The invention belongs to the technical field of reconstruction of brain focal areas of autistic children, and particularly relates to a Bayes-based imaging system for brain activity abnormal focal areas of children.
Background
The brain is the most important part of the central nervous system and controls higher thinking of people, such as cognition, learning, social interaction and the like, and the cognitive learning process and brain function disorder brain operation mechanism of autistic children need to be clarified firstly in the education and brain function disorder intervention treatment of autistic children. Electroencephalogram source imaging is an important means for exploring brain activity mechanism, understanding various thinking and sensing dynamic change process of brain activity, and has important significance for disclosing the brain operation mechanism of autistic children.
The brain electrical source imaging inverse problem is a process of inverting neuron activity information in the brain by sampling brain electromagnetic data, and is the core of brain electrical source imaging. In recent years, the technology can be used for cognitive learning of autistic children and study of dysfunction electroencephalogram source imaging technology at home and abroad, but the existing brain source imaging technology is mostly based on a minimum norm method, the degree of freedom is limited, and the characteristics of high complexity and high resolution requirement of human brain source activities lead to low cognitive learning and dysfunction electroencephalogram source imaging accuracy of autistic children. How to realize the automatic identification of the electroencephalogram and magnetic anomaly sampling data, and how to quickly and accurately reconstruct the brain focus area activities and the positions of the brain focus area activities of the autistic children are still a challenge subject.
Disclosure of Invention
Aiming at least one defect or improvement requirement in the prior art, the invention provides a Bayes-based imaging system for the brain activity abnormal focal zone of the child, which can greatly improve the accuracy of localization and reconstruction of the focal zone and is particularly suitable for imaging the brain activity abnormal focal zone of the child with the autism.
In order to achieve the above object, the present invention provides a bayesian-based imaging system for a focal zone of abnormal brain activity of a child, comprising:
the signal processing module is used for acquiring electroencephalogram sampling data and preprocessing the electroencephalogram sampling data;
the brain activity abnormity identification module is used for identifying abnormal waves and non-abnormal waves from the preprocessed electroencephalogram and magnetic sampling data;
the noise estimation module is used for constructing a first sampling data generation model of the non-abnormal wave, solving the first sampling data generation model by adopting a variational Bayesian method and acquiring the distribution information of background interference noise of data acquisition;
the brain electrical source activity reconstruction module is used for constructing a second sampling data generation model of the abnormal wave, the distribution information of the background interference noise is a parameter in the second sampling data generation model, and the second sampling data generation model is solved by adopting an empirical Bayesian method to obtain a brain electrical source activity time sequence;
and the imaging module is used for imaging according to the brain electrical source activity time sequence.
Preferably, the signal processing module includes:
the sampling offset removing module is used for removing sampling offset in the electroencephalogram magnetic sampling data;
the interference removing module is used for removing noise of the signal from which the sampling offset is removed by adopting an independent component analysis method and a space projection method;
and the filtering module is used for filtering the signal subjected to denoising processing.
Preferably, the first sampling data generation model is:
YconFU + E, wherein YconNon-abnormal waves, F is an unknown M x K weighting matrix, and U is an unknown K x tKFactor parameter of, tKIs the number of samples in the non-anomalous wave, U ═ U1,u2,...,uK]TE is a residual signal after the decomposition of the non-abnormal wave, M is the number of channels of the acquisition point, and K is less than M.
Preferably, the solving of the first sampling data generation model by using the variational bayes method includes the steps of:
let the prior distribution of the factor parameter U be a standard gaussian distribution:
Figure BDA0002857602630000021
the distribution of the background interference noise E for data acquisition is:
Figure BDA0002857602630000022
wherein Λ ═ diag (λ)1,λ2,...,λM) Is a diagonal matrix, Λ is the prior variance of E,
the weighting matrix F is set to gaussian:
Figure BDA0002857602630000031
wherein, Fm,kAs a value in the weighting matrix F, alphakTo addThe prior variance of the weight matrix F, K is more than or equal to 1 and less than or equal to K, the first sampling data generation model is evolved based on variational Bayesian estimation, and the updating mode of the obtained weight matrix F and the factor parameter U is as follows:
Figure BDA0002857602630000032
Figure BDA0002857602630000033
Ψ＝Ruu+α
Figure BDA0002857602630000034
Figure BDA0002857602630000035
wherein R isyyFor the covariance matrix of the sampled data, RuuTo estimate the factor covariance matrix, RuyFor multiplying the estimation factor by the sample data, α ═ diag (α)1,α2...,αk)；
Performing iterative cycle according to the updating mode of the weighting matrix F and the factor parameter U until the values of the weighting matrix F and the factor parameter U are not changed any more, and acquiring a reduced-dimension and background interference noise covariance matrix sigma according to the values of the weighting matrix F and the factor parameter U at the momentEComprises the following steps:
Figure BDA0002857602630000036
preferably, the second sampling data generation model is:
Yact＝LS+ε，
wherein, YactIs an abnormal wave, Yact＝[y(t1),y(t2),...,y(tK)]，tKSampling points for abnormal wavesThe number of the first and second groups is,
Figure BDA0002857602630000037
obtaining the extracerebral sampling intensity generated by known unit source activity, wherein N is the number of sources to be solved, S is the time sequence of the electroencephalogram source activity, epsilon is background interference noise, and the extracerebral sampling intensity generated by the unit source activity at the position N is Ln＝[l1,l2,...,lM]N is less than N, M is the channel number of the acquisition point, M is less than N, and S is [ S ═ S [1,S2,...,SN]T。
Preferably, the solving of the second sampling data generation model by using the empirical bayesian method includes the steps of:
setting the brain electrical source activity time sequence S to be mutually independent in space and time and obey Gaussian distribution, and converting the second sampling data generation model into a second sampling data probability generation model by adopting an empirical Bayes method;
and constructing a cost function based on the convex function plane boundary, and solving the second sampling data probability generation model by optimizing the cost function.
Preferably, the probability generation model of the second sampled data is:
the brain electrical activity time series S are spatially and temporally independent of each other and obey a gaussian distribution:
Figure BDA0002857602630000041
Figure BDA0002857602630000042
being the variance of the prior gaussian distribution for position n, the conditional probability distribution p (Y | S) is,
Figure BDA0002857602630000043
ΣEis a covariance matrix, basis, of background interference noise obtained by a noise estimation blockThe empirical Bayesian model can be used to obtain:
Figure BDA0002857602630000044
posterior probability distribution p (S | Y)act) Comprises the following steps:
Figure BDA0002857602630000045
Γ-1in order to be a hyper-parameter,
Figure BDA0002857602630000046
and Γ-1The posterior probability mean and variance of the source activity are sampled at time t, respectively.
Preferably, the hyperparameter Γ-1By maximizing the edge distribution p (Y)act| γ), edge distribution logarithmic model logp (Y)actγ) is:
Figure BDA0002857602630000047
Figure BDA0002857602630000048
is an abnormal wave covariance matrix model,
Figure BDA0002857602630000049
is the n-th source activity point
Figure BDA00028576026300000410
Construction based on gamma-1Log | Σ ofyThe | boundary function is such that,
Figure BDA0002857602630000051
tr (×) is the trace of the computation matrix;
the cost function based on the convex function plane boundary is:
Figure BDA0002857602630000052
Λnand Λ0Is an auxiliary variable;
the posterior distribution p (S | Y) can be obtained from the probability generation model of the second sampling dataact) The mean and variance of (a) are as follows:
Figure BDA0002857602630000053
Figure BDA0002857602630000054
for cost function
Figure BDA0002857602630000055
Solving for the over parameter gammanAnd auxiliary parameter ΛnThe reciprocal, one can obtain:
Figure BDA0002857602630000056
Figure BDA0002857602630000057
make it
Figure BDA0002857602630000058
Is zero, over-parameter gammanAnd auxiliary parameter ΛnThe formula of (1) is as follows:
Figure BDA0002857602630000059
Figure BDA00028576026300000510
by iteration of loops
Figure BDA00028576026300000511
Γ-1And gammanUntil the edge distribution logarithm model converges, at which time
Figure BDA00028576026300000512
I.e. to solve the brain power activity time series.
Preferably, the identifying of the abnormal wave and the non-abnormal wave comprises the steps of:
identifying a first candidate abnormal wave according to the amplitude of the preprocessed electromagnetic sampling signal;
and reading the standard abnormal wave stored in advance, matching the local part of the first candidate abnormal wave with the standard abnormal wave, and identifying the abnormal wave according to the matching result.
Preferably, the imaging according to the brain electrical source activity time sequence comprises the following steps:
and obtaining pre-stored model parameters, and imaging according to the model parameters and the electroencephalogram source activity time sequence.
In general, compared with the prior art, the invention has the following beneficial effects: the brain activity background interference noise extraction method based on variational Bayes is adopted in the imaging system of the lesion areas with abnormal brain activity of the children, and the lesion areas with abnormal brain activity of the children are reconstructed and positioned based on the empirical Bayes model by combining with the sampled abnormal brain electromagnetic data. Is particularly suitable for imaging focal zones of autistic children with abnormal brain activities.
Drawings
FIG. 1 is a schematic block diagram of an imaging system according to an embodiment of the invention;
fig. 2 is a functional diagram of an imaging system according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention. In addition, the technical features involved in the embodiments of the present invention described below may be combined with each other as long as they do not conflict with each other.
As shown in fig. 1, the bayesian-based imaging system for a focal zone of abnormal brain activity of a child according to an embodiment of the present invention includes a signal acquisition and processing module, a brain activity abnormality recognition module, a noise estimation module, an electroencephalogram source activity reconstruction module, and an imaging module. The system can be applied to various scenes for identifying diseases through imaging of the focal region with the abnormal brain electromagnetic activity, and is particularly suitable for the infantile autism.
And the signal processing module is used for acquiring the electromagnetic sampling data and preprocessing the electromagnetic sampling data.
A preferred implementation of the signal processing module is described below.
Signal acquisition:
inputting brain electromagnetic sampling data, Yraw＝[yraw(1),yraw(2),...,yraw(T)]T is the number of original signal sampling points, yraw(t)＝[y1(t),y2(t),…,yM(t)]TWhere M is the number of channels of acquisition points and T represents the transpose matrix. The input data needs to be preliminarily screened to ensure the validity and integrity of the sampled data, such as data format conversion and parameter confirmation, including information of the sampled data.
Signal preprocessing:
the signal acquisition processing module comprises a sampling offset removing module, an interference removing module and a filtering module, and is used for preprocessing input data, including preprocessing such as sampling offset removing, drying removing and filtering.
The formula for removing the sampling offset is as follows:
Figure BDA0002857602630000071
the interference removal needs to be performed based on the sampled data characteristics.
Preferably, based on Independent Component Analysis (ICA) or spatial projection, such as interference elimination based on ICA (including electro-oculogram), the following formula is:
Figure BDA0002857602630000072
wherein A ═ a1,a2,...,aN]To decompose the weighting matrix, X ═ X1,x2,...,xN]TIs an independent component, let xmContaining significant interference or noise, the de-dried sampled data YnewThe following were used:
Figure BDA0002857602630000073
preferably, the signal after the ICA-based interference removal processing is further processed by using a projection-based interference removal method. The interference elimination method based on projection needs to find out a projection matrix, which can be based on features such as brain structure space information, etc., assuming that the projection matrix is W, the interference elimination formula is:
Yclean＝WTYnew (4)
filtering, namely filtering the sampled data to remove the sampled data of the non-brain activity frequency band, mainly applying time-frequency filtering based on Fourier or wavelet transform and reserving the interested frequency band data. The accuracy of subsequent abnormal recognition is guaranteed through the preprocessed data.
And the brain activity abnormity identification module is used for identifying abnormal waves and non-abnormal waves from the preprocessed electroencephalogram and magnetic sampling data.
Preferably, brain activity abnormalities are identified based on the amplitude and local waveform of the sampled brain electrical data.
Primarily selecting according to the amplitude: carrying out statistical analysis on the amplitude of the sampled electroencephalogram data, and setting a threshold value to screen potential abnormal sampling waveforms; the threshold is set to be roughly selected, and preferably, a plurality of sampling data segments are satisfied as a first candidate abnormal wave to enter the identification of the next stage.
And (3) waveform matching, namely further selecting the primarily selected segments, matching the primarily selected waveforms based on abnormal wave forms pre-stored in a waveform library, realizing window size switching by adjusting sampling frequency, and further selecting abnormal waves between 200ms and 100ms for identification. In one embodiment, the brain activity sample data with a matching degree greater than 80% is preferably used as the threshold of abnormal activity. The step is used for marking the sampling abnormal wave at the same time, and marking the non-abnormal wave with the same time scale before the abnormal wave appears secondarily as the input of the noise estimation module.
A noise estimation module: the method comprises the steps of constructing a first sampling data generation model of the non-abnormal wave, solving the first sampling data generation model by adopting a variational Bayes method, and acquiring background interference noise of data acquisition, namely estimating background interference noise statistical distribution information by adopting a variational Bayes-based method to analyze the non-abnormal electroencephalogram.
Non-anomalous waves are also called stationary waves. Probability distribution information is solved by carrying out variational Bayesian estimation on the stationary wave to be used as a covariance matrix of background interference noise. The specific process is as follows, assuming that the non-abnormal wave is YconVariational bayesian factor analysis assumes that the sample data is represented as:
Ycon＝FU+E (5)
where F is an unknown M K weighting matrix, and U is [ U ]1,u2,...,uK]TIs unknown KxtKFactor parameter of, tKAnd setting K < M for the number of sampling points in the non-abnormal wave, namely K is a preset arbitrary positive integer less than M, and E is a residual signal after the decomposition of the non-abnormal wave. The prior distribution of the factor parameters is set as the standard gaussian distribution as follows:
Figure BDA0002857602630000081
while the distribution of sensor noise is as follows:
Figure BDA0002857602630000082
Λ＝diag(λ1,λ2,...,λM) Is a diagonal matrix, λmIs the variance of the mth channel E, and Λ is the prior variance of E. The weighting matrix also satisfies the following gaussian distribution:
Figure BDA0002857602630000091
αkk is more than or equal to 1 and less than or equal to K and is the prior variance of the weighting matrix F.
The update rule of unknown parameters based on variational bayes estimation is as follows:
Figure BDA0002857602630000092
Figure BDA0002857602630000093
Ψ＝Ruu+α (11)
Figure BDA0002857602630000094
Figure BDA0002857602630000095
wherein the content of the first and second substances,
Figure BDA0002857602630000096
is the average of all the elements of the M x K weighting matrix,
Figure BDA0002857602630000097
is KxtKThe mean value of the posterior probability distribution, R, of all elements in the factor parameter matrix of (1)yyFor the covariance matrix of the sampled data, RuuTo estimate the factor covariance matrix, RuyFor multiplying the estimation factor by the sample data, α ═ diag (α)1,α2...,αk). Equations 9 through 13 are iteratively cycled through
Figure BDA0002857602630000098
And
Figure BDA0002857602630000099
is not changing, according to which time
Figure BDA00028576026300000912
And
Figure BDA00028576026300000913
the value of (A) can obtain the covariance matrix information sigma of the interference and the background noise with reduced dimensionEThe description is as follows:
Figure BDA00028576026300000910
equation 14 is the background interference information estimated based on the non-abnormal brain activity sampling data, which is used for the focal zone input of the brain source activity reconstruction module brain abnormal activity.
The brain source activity reconstruction module is used for constructing a second sampling data generation model of the abnormal wave, background interference noise is a parameter in the second sampling data generation model, the second sampling data generation model is solved by adopting an empirical Bayesian method, and a brain source activity time sequence is obtained.
Preferably, the solving of the second sampling data generation model by using the empirical bayesian method comprises the steps of: setting the brain electrical source activity time sequence S to be mutually independent in space and time and obey Gaussian distribution, and converting the second sampling data generation model into a second sampling data probability generation model by adopting an empirical Bayes method; and constructing a cost function based on the convex function plane boundary, and solving the second sampling data probability generation model by optimizing the cost function.
The following describes in detail a preferred implementation of the brain power activity reconstruction module.
Modeling the second sampling data generation model of the abnormal wave:
is provided with Yact＝[y(t1),y(t2),...,y(tK)]For identified abnormal electroencephalogram sampling data, tKThe number of sampling points of the abnormal wave is the same as that of the non-abnormal wave. y (t)k)∈RM*1Is tkAnd (3) sampling data of the electroencephalogram at a moment, wherein M is the number of channels. The brain source activity space comprises N source activity points, and the unit source activity of the position i generates the extracerebral sampling intensity of Ln＝[l1,l2,...,lM]，LnThe steering matrix of the nth source activity is a known parameter, can be solved through a source reconstruction positive problem, and can also be generated through open source software fieldtrip, nutmeg, spm and the like. The second sampling data generation model of the abnormal wave is as follows:
Yact＝LS+ε (15)
S＝[S1,S2,...,SN]Tbeing a time sequence of source activities, SnThe source activity time series for position n, ε is the noise and interference data, whose statistical signature information is solved by the noise estimation module equation 14. Let the brain source activities be spatially and temporally independent and obey a gaussian distribution as follows:
Figure BDA0002857602630000101
Figure BDA0002857602630000102
variance γ ═ diag (γ) of a priori gaussian distribution for position n1,…,γn) The conditional probability distribution p (Y | S) is,
Figure BDA0002857602630000103
ΣEis the covariance matrix of the noise and interference estimated in S4. Based on an empirical Bayesian model, the following results can be obtained:
Figure BDA0002857602630000104
sample data p (Y)act) The method is difficult to calculate, only has a reduction effect in the empirical Bayesian estimation process, and does not influence the estimation result. Due to the prior distribution p (S | gamma) and the conditional distribution p (Y)actS) are all Gaussian distributions, posterior probability distribution p (S | Y)act) Is composed of
Figure BDA0002857602630000111
Wherein the content of the first and second substances,
Figure BDA0002857602630000112
is the mean, Γ, of the posterior probability distribution of the source activity-1The posterior distribution equal variance.
Equations 16, 17, 18 and 19 constitute a second sample data probability generating model of the second sample data generating model, and model parameters are solved below.
S5.2: cost function optimization based on convex function plane boundary
Hyper-parameter gamma-1By maximizing the edge distribution p (Y)act| γ), edge distribution logarithmic model logp (Y)actγ) is:
Figure BDA0002857602630000113
Figure BDA0002857602630000114
for the abnormal wave sample data covariance matrix model, equation 21 can be optimized by the expectation-maximization algorithm (EM algorithm)But log | ∑ oyThe convergence rate of | is extremely slow due to log | ∑ syIs gamma-1Constructing a gamma-based convex function of-1Log | ∑ ofyThe function of the | boundary(s),
Figure BDA0002857602630000115
optimized solution-log | sigmayThe maximum of i, i.e. the minimum of the boundary cost function, is solved, as shown below,
Figure BDA0002857602630000116
Λnand Λ0For auxiliary variables, tr (×) is the trace of the computation matrix, based on the cost function of the convex function boundary
Figure BDA0002857602630000117
Satisfy always:
Figure BDA0002857602630000121
logp(Yactthe maximum value of the [ gamma ]) is optimized to be the cost function
Figure BDA0002857602630000122
The minimum value of (2) is solved.
S5.3 Complex focal zone reconstruction
From the above analysis, it can be seen that the solution of brain power first assumes that the source activity obeys mutually independent gaussian distributions, and by empirical bayesian estimation, the values of the source activity are solved by substituting equations 16, 17 and 19 into equation 18 to solve posterior distribution p (sy)act) The mean and variance of (a) are as follows:
Figure BDA0002857602630000123
Figure BDA0002857602630000124
Figure BDA0002857602630000125
is a diagonal matrix. Mean value in equation 20
Figure BDA0002857602630000126
Namely a source activity time sequence to be solved, and the solution of the source activity time sequence needs to be known by the hyper-parameter gamma. Solving the maximum value of edge distribution to obtain the hyper-parameter gamma, and constructing a cost function based on a convex function boundary
Figure BDA0002857602630000127
Model, conversion to cost function by hyper-parametric gamma solution
Figure BDA0002857602630000128
And (4) optimizing. To function
Figure BDA0002857602630000129
Solving for the over parameter gammanAnd auxiliary parameter ΛnThe reciprocal, one can obtain:
Figure BDA00028576026300001210
let equation 25 be zero, over-parameter γnAnd auxiliary parameter ΛnThe formula of (1) is as follows:
Figure BDA00028576026300001211
by iteratively iterating equations 24 and 26 in a loop until cost function equation 20 converges,
Figure BDA00028576026300001212
i.e. to solve the activity time series of the brain power.
And the imaging module is used for outputting and imaging the electroencephalogram source activity according to the electroencephalogram source activity time sequence acquired by the electroencephalogram source activity reconstruction module.
Obtaining brain power activity time series
Figure BDA00028576026300001213
And then, displaying the result, and further, displaying the result based on the existing model parameters, such as the source position after brain space segmentation and brain structure information, wherein the MRI template is adopted in the patent, the source activity and the operation result on the general brain model are rendered, and the operation result is shown in fig. 2.
It must be noted that in any of the above embodiments, the methods are not necessarily executed in order of sequence number, and as long as it cannot be assumed from the execution logic that they are necessarily executed in a certain order, it means that they can be executed in any other possible order.
It will be understood by those skilled in the art that the foregoing is only a preferred embodiment of the present invention, and is not intended to limit the invention, and that any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the scope of the present invention.

Claims (10)

1. A Bayesian-based imaging system for a focal zone of abnormal brain activity in a child, comprising:
the signal processing module is used for acquiring electroencephalogram sampling data and preprocessing the electroencephalogram sampling data;
the brain activity abnormity identification module is used for identifying abnormal waves and non-abnormal waves from the preprocessed electroencephalogram and magnetic sampling data;
the noise estimation module is used for constructing a first sampling data generation model of the non-abnormal wave, solving the first sampling data generation model by adopting a variational Bayesian method and acquiring the distribution information of background interference noise of data acquisition;
the brain electrical source activity reconstruction module is used for constructing a second sampling data generation model of the abnormal wave, the distribution information of the background interference noise is a parameter in the second sampling data generation model, and the second sampling data generation model is solved by adopting an empirical Bayesian method to obtain a brain electrical source activity time sequence;
and the imaging module is used for imaging according to the brain electrical source activity time sequence.
2. The bayesian-based imaging system of a focal zone of brain activity abnormality of a child according to claim 1, wherein said signal processing module comprises:
the sampling offset removing module is used for removing sampling offset in the electroencephalogram magnetic sampling data;
the interference removing module is used for removing noise of the signal from which the sampling offset is removed by adopting an independent component analysis method and a space projection method;
and the filtering module is used for filtering the signal subjected to denoising processing.
3. The bayesian-based imaging system for a focal zone of brain activity abnormality in a child according to claim 1, wherein said first sampled data generating model is:
YconFU + E, wherein YconNon-abnormal waves, F is an unknown M x K weighting matrix, and U is an unknown K x tKFactor parameter of, tKIs the number of samples in the non-anomalous wave, U ═ U1,u2,...,uK]TE is a residual signal after the decomposition of the non-abnormal wave, M is the number of channels of the acquisition point, and K is less than M.
4. The bayesian-based imaging system for a focal zone of abnormal brain activity in a child according to claim 3, wherein said solving the first sampled data generating model using the variational bayesian method comprises the steps of:
let the prior distribution of the factor parameter U be a standard gaussian distribution:
Figure FDA0002857602620000021
the distribution of the background interference noise E for data acquisition is:
Figure FDA0002857602620000022
wherein Λ ═ diag (λ)1,λ2,...,λM) Is a diagonal matrix, Λ is the prior variance of E,
the weighting matrix F is set to gaussian:
Figure FDA0002857602620000023
wherein, Fm,kAs a value in the weighting matrix F, alphakThe prior variance of the weighting matrix F is more than or equal to 1 and less than or equal to K, the first sampling data generation model is evolved based on variational Bayesian estimation, and the updating mode of the weighting matrix F and the factor parameter U is as follows:
Figure FDA0002857602620000024
Figure FDA0002857602620000025
Ψ＝Ruu+α
Figure FDA0002857602620000026
Figure FDA0002857602620000027
wherein R isyyFor the covariance matrix of the sampled data, RuuTo estimate the factor covariance matrix, RuyFor multiplying the estimation factor by the sample data, α ═ diag (α)1,α2...,αk)；
Performing iterative cycle according to the updating mode of the weighting matrix F and the factor parameter U until the values of the weighting matrix F and the factor parameter U are not changed any more, and acquiring a reduced-dimension and background interference noise covariance matrix sigma according to the values of the weighting matrix F and the factor parameter U at the momentEComprises the following steps:
Figure FDA0002857602620000031
5. the bayesian-based imaging system for a focal zone of brain activity abnormality in a child according to claim 1, wherein said second sampled data generating model is:
Yact＝LS+ε，
wherein, YactIs an abnormal wave, Yact＝[y(t1),y(t2),...,y(tK)]，tKIs the number of sampling points of the abnormal wave,
Figure FDA0002857602620000032
obtaining the extracerebral sampling intensity generated by known unit source activity, wherein N is the number of sources to be solved, S is the time sequence of the electroencephalogram source activity, epsilon is background interference noise, and the extracerebral sampling intensity generated by the unit source activity at the position N is Ln＝[l1,l2,...,lM]N is less than N, M is the channel number of the acquisition point, M is less than N, and S is [ S ═ S [1,S2,...,SN]T。
6. The bayesian-based imaging system for a focal zone of abnormal brain activity in a child according to claim 5, wherein said solving the second sampled data generating model using an empirical bayesian approach comprises the steps of:
setting the brain electrical source activity time sequence S to be mutually independent in space and time and obey Gaussian distribution, and converting the second sampling data generation model into a second sampling data probability generation model by adopting an empirical Bayes method;
and constructing a cost function based on the convex function plane boundary, and solving the second sampling data probability generation model by optimizing the cost function.
7. The Bayesian-based imaging system for brain activity abnormality focal zone of children as recited in claim 6, wherein said second sampled data probabilistic generative model is:
the brain electrical activity time series S are spatially and temporally independent of each other and obey a gaussian distribution:
Figure FDA0002857602620000033
Figure FDA0002857602620000034
being the variance of the prior gaussian distribution for position n, the conditional probability distribution p (Y | S) is,
Figure FDA0002857602620000035
ΣEthe covariance matrix of the background interference noise acquired by the noise estimation module is obtained based on an empirical Bayesian model:
Figure FDA0002857602620000041
posterior probability distribution p (S | Y)act) Comprises the following steps:
Figure FDA0002857602620000042
Γ-1in order to be a hyper-parameter,
Figure FDA0002857602620000043
Figure FDA0002857602620000044
and Γ-1The posterior probability mean and variance of the source activity are sampled at time t, respectively.
8. A Bayesian-based imaging system of focal zones of abnormal brain activity in children as recited in claim 7,
hyper-parameter gamma-1By maximizing the edge distribution p (Y)act| γ), edge distribution logarithmic model logp (Y)actγ) is:
Figure FDA0002857602620000045
Figure FDA0002857602620000046
is an abnormal wave covariance matrix model,
Figure FDA0002857602620000047
is the n-th source activity point
Figure FDA0002857602620000048
Construction based on gamma-1Log | Σ ofyThe | boundary function is such that,
Figure FDA0002857602620000049
tr (×) is the trace of the computation matrix;
the cost function based on the convex function plane boundary is:
Figure FDA0002857602620000051
Λnand Λ0Is an auxiliary variable;
the posterior distribution p (S | Y) can be obtained from the probability generation model of the second sampling dataact) The mean and variance of (a) are as follows:
Figure FDA0002857602620000052
Figure FDA0002857602620000053
for cost function
Figure FDA0002857602620000054
Solving for the over parameter gammanAnd auxiliary parameter ΛnThe reciprocal, one can obtain:
Figure FDA0002857602620000055
Figure FDA0002857602620000056
make it
Figure FDA0002857602620000057
Is zero, over-parameter gammanAnd auxiliary parameter ΛnThe formula of (1) is as follows:
Figure FDA0002857602620000058
Figure FDA0002857602620000059
by iteration of loops
Figure FDA00028576026200000510
Γ-1And gammanUntil the edge distribution logarithm model converges, at which time
Figure FDA00028576026200000511
I.e. to solve the brain power activity time series.
9. The bayesian-based imaging system of a focal zone of abnormal brain activity in a child according to claim 1, wherein said identifying of said abnormal waves and said non-abnormal waves comprises the steps of:
identifying a first candidate abnormal wave according to the amplitude of the preprocessed electromagnetic sampling signal;
and reading the standard abnormal wave stored in advance, matching the local part of the first candidate abnormal wave with the standard abnormal wave, and identifying the abnormal wave according to the matching result.
10. The bayesian-based imaging system of a focal zone of abnormal brain activity in a child according to claim 1, wherein said imaging based on a time series of brain electrical source activity comprises the steps of:
and obtaining pre-stored model parameters, and imaging according to the model parameters and the electroencephalogram source activity time sequence.
CN202011553134.8A 2020-12-24 2020-12-24 Bayes-based imaging system for focal zone with abnormal brain activity of children Active CN112690775B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011553134.8A CN112690775B (en) 2020-12-24 2020-12-24 Bayes-based imaging system for focal zone with abnormal brain activity of children

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011553134.8A CN112690775B (en) 2020-12-24 2020-12-24 Bayes-based imaging system for focal zone with abnormal brain activity of children

Publications (2)

Publication Number Publication Date
CN112690775A CN112690775A (en) 2021-04-23
CN112690775B true CN112690775B (en) 2022-04-29

Family

ID=75509934

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011553134.8A Active CN112690775B (en) 2020-12-24 2020-12-24 Bayes-based imaging system for focal zone with abnormal brain activity of children

Country Status (1)

Country Link
CN (1) CN112690775B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113951899B (en) * 2021-10-28 2024-04-19 华中师范大学 Brain source activity high-resolution reconstruction system and method

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107095684A (en) * 2017-03-23 2017-08-29 兰州大学 A kind of autism of children risk evaluating system based on brain electricity
CN108416822A (en) * 2018-03-22 2018-08-17 武汉大学 A kind of multi-level multiple dimensioned chromatography imaging method based on Bayesian Estimation
CN110251083A (en) * 2019-06-20 2019-09-20 深圳大学 A kind of processing method, system and the storage medium of the location data of epileptic focus
CN111035383A (en) * 2019-12-20 2020-04-21 东南大学 Analysis method for electroencephalogram signals of children facing autism spectrum disorder
CN111543949A (en) * 2020-05-13 2020-08-18 北京航空航天大学 Child ASD diagnosis device based on magnetoencephalogram and electroencephalogram

Family Cites Families (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP5531237B2 (en) * 2009-02-24 2014-06-25 本田技研工業株式会社 Brain information output device, robot, and brain information output method
ES2919141T3 (en) * 2016-07-18 2022-07-22 Univ Daix Marseille Amu Modulation method of epileptogenicity in the brain of a patient
US10945622B2 (en) * 2016-08-23 2021-03-16 Regents Of The University Of Minnesota System and method for assessing electrical activity using an iterative sparse technique

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN107095684A (en) * 2017-03-23 2017-08-29 兰州大学 A kind of autism of children risk evaluating system based on brain electricity
CN108416822A (en) * 2018-03-22 2018-08-17 武汉大学 A kind of multi-level multiple dimensioned chromatography imaging method based on Bayesian Estimation
CN110251083A (en) * 2019-06-20 2019-09-20 深圳大学 A kind of processing method, system and the storage medium of the location data of epileptic focus
CN111035383A (en) * 2019-12-20 2020-04-21 东南大学 Analysis method for electroencephalogram signals of children facing autism spectrum disorder
CN111543949A (en) * 2020-05-13 2020-08-18 北京航空航天大学 Child ASD diagnosis device based on magnetoencephalogram and electroencephalogram

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
Robust Empirical Bayesian Reconstruction of Distributed Sources for Electromagnetic Brain Imaging;Chang Cai et al;《IEEE TRANSACTIONS ON MEDICAL IMAGING》;20200331;全文 *
基于贝叶斯理论和时空约束的脑电源成像方法研究;刘柯;《中国博士学位论文全文数据库 (医药卫生科技辑)》;20170515;第567-577页 *

Also Published As

Publication number Publication date
CN112690775A (en) 2021-04-23

Similar Documents

Publication Publication Date Title
Colclough et al. A symmetric multivariate leakage correction for MEG connectomes
Sanei et al. A new adaptive line enhancer based on singular spectrum analysis
Wu et al. A hierarchical Bayesian approach for learning sparse spatio-temporal decompositions of multichannel EEG
Bolstad et al. Space–time event sparse penalization for magneto-/electroencephalography
CN116503507B (en) Magnetic particle image reconstruction method based on pre-training model
CN115326783B (en) Raman spectrum preprocessing model generation method, system, terminal and storage medium
Hansen et al. Unmixing oscillatory brain activity by EEG source localization and empirical mode decomposition
CN112690775B (en) Bayes-based imaging system for focal zone with abnormal brain activity of children
Ge et al. Spike sorting by stochastic simulation
CN115414051A (en) Emotion classification and recognition method of electroencephalogram signal self-adaptive window
CN108470335B (en) Multi-correlation source scanning imaging method based on brain source space segmentation
CN113378650B (en) Emotion recognition method based on electroencephalogram source imaging and regularization common space mode
Deng et al. Impulse feature extraction method for machinery fault detection using fusion sparse coding and online dictionary learning
Gui et al. Classification of task-state fMRI data based on circle-EMD and machine learning
Washizawa et al. Blind extraction of global signal from multi-channel noisy observations
Lu et al. A method combining fractal analysis and single channel ICA for vibration noise reduction
Murad et al. Radial undersampling-based interpolation scheme for multislice CSMRI reconstruction techniques
CN115132349A (en) Event-related brain activity time-space-frequency dynamic positioning system and method
Jurdana et al. Application of the 2D Local Entropy Information in Sparse TFD Reconstruction
Truong et al. Assessing learned features of Deep Learning applied to EEG
CN113995422A (en) Transient electroencephalogram source positioning method and system based on nonnegative block sparse Bayesian learning
Deng et al. Approximate sparsity and nonlocal total variation based compressive MR image reconstruction
Ramzan et al. Analyzing and identifying oscillatory activities of emotions based on Electroencephalography signals using power spectral density and spatial filtering techniques
Liu et al. One to multiple mapping dual learning: Learning multiple sources from one mixed signal
Liu et al. One to multiple mapping dual learning: Learning multiple signals from one mixture

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant