CN113768464A - Epileptic seizure time period detection system and method based on empirical mode decomposition - Google Patents
Epileptic seizure time period detection system and method based on empirical mode decomposition Download PDFInfo
- Publication number
- CN113768464A CN113768464A CN202110961989.2A CN202110961989A CN113768464A CN 113768464 A CN113768464 A CN 113768464A CN 202110961989 A CN202110961989 A CN 202110961989A CN 113768464 A CN113768464 A CN 113768464A
- Authority
- CN
- China
- Prior art keywords
- electroencephalogram
- information
- module
- dtf
- empirical mode
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 63
- 206010010904 Convulsion Diseases 0.000 title claims abstract description 62
- 238000000354 decomposition reaction Methods 0.000 title claims abstract description 47
- 206010015037 epilepsy Diseases 0.000 title claims abstract description 31
- 208000028329 epileptic seizure Diseases 0.000 title claims abstract description 29
- 238000001514 detection method Methods 0.000 title claims abstract description 23
- 210000004556 brain Anatomy 0.000 claims abstract description 23
- 230000001037 epileptic effect Effects 0.000 claims abstract description 23
- 238000012706 support-vector machine Methods 0.000 claims abstract description 12
- 238000012546 transfer Methods 0.000 claims abstract description 12
- 238000010276 construction Methods 0.000 claims abstract description 10
- 238000000605 extraction Methods 0.000 claims abstract description 10
- 239000011159 matrix material Substances 0.000 claims description 24
- 238000002790 cross-validation Methods 0.000 claims description 17
- 238000012360 testing method Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 238000005457 optimization Methods 0.000 claims description 7
- 210000002569 neuron Anatomy 0.000 claims description 4
- 238000003909 pattern recognition Methods 0.000 claims description 4
- 230000008569 process Effects 0.000 claims description 4
- 230000008859 change Effects 0.000 claims description 3
- 230000003993 interaction Effects 0.000 claims description 2
- 238000010827 pathological analysis Methods 0.000 claims description 2
- 230000002093 peripheral effect Effects 0.000 claims description 2
- 238000003672 processing method Methods 0.000 abstract description 2
- 238000000537 electroencephalography Methods 0.000 description 5
- 230000035945 sensitivity Effects 0.000 description 4
- 238000012549 training Methods 0.000 description 4
- 239000013598 vector Substances 0.000 description 4
- 230000002159 abnormal effect Effects 0.000 description 2
- 238000003066 decision tree Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000001575 pathological effect Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 208000012902 Nervous system disease Diseases 0.000 description 1
- 208000025966 Neurological disease Diseases 0.000 description 1
- 238000012274 Preoperative evaluation Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000003542 behavioural effect Effects 0.000 description 1
- 230000002490 cerebral effect Effects 0.000 description 1
- 230000001684 chronic effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000013100 final test Methods 0.000 description 1
- 230000001314 paroxysmal effect Effects 0.000 description 1
- 230000035790 physiological processes and functions Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000000306 recurrent effect Effects 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 208000011580 syndromic disease Diseases 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
- 230000007704 transition Effects 0.000 description 1
- 238000010200 validation analysis Methods 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4076—Diagnosing or monitoring particular conditions of the nervous system
- A61B5/4094—Diagnosing or monitoring seizure diseases, e.g. epilepsy
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
-
- 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
-
- 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
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- 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
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/20—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for computer-aided diagnosis, e.g. based on medical expert systems
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Pathology (AREA)
- Molecular Biology (AREA)
- Animal Behavior & Ethology (AREA)
- Veterinary Medicine (AREA)
- Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Biophysics (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Neurology (AREA)
- Signal Processing (AREA)
- Fuzzy Systems (AREA)
- Neurosurgery (AREA)
- Mathematical Physics (AREA)
- Theoretical Computer Science (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Epidemiology (AREA)
- Primary Health Care (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Physics & Mathematics (AREA)
- Psychology (AREA)
- General Engineering & Computer Science (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
A epileptic seizure time period detection system and method based on empirical mode decomposition comprises an electroencephalogram signal acquisition module A, a self-adaptive decomposition module B, an information flow network construction module C, an outflow information extraction module D and a classification identification module E, wherein multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptic patients are acquired firstly based on the five modules, the electroencephalogram signals are decomposed into a plurality of IMF components and margins by using a multi-element empirical mode decomposition (MEMD) method, information flow intensity among different brain areas is calculated through a Directed Transfer Function (DTF), the IMF components and the margins are used as electroencephalogram characteristic values to be input into a cost-sensitive support vector machine (CSVM) to be used for distinguishing signals of the seizure periods and the non-seizure periods, the multi-element empirical mode decomposition (MEMD) is a self-adaptive signal time-frequency processing method, can carry out self-adaptive decomposition on multichannel data at the same time, and is suitable for analyzing electroencephalogram signals with high correlation and non-stationarity, the accuracy of the frequency information of the positioning electroencephalogram signal can be enhanced.
Description
Technical Field
The invention relates to the technical field of biomedical signal processing, in particular to a seizure time period detection system and method based on empirical mode decomposition.
Background
Epilepsy is a chronic neurological disease or syndrome characterized by sudden, transient, recurrent epileptic seizures. Epileptic seizures are the clinical manifestations of paroxysmal abnormal hypersynchronous electrical activity of the nerve cell population in the brain. Repeated and sudden epileptic seizures are very dangerous, threaten the life of a patient and also add great burden to the family. In general, clinically, doctors use long-range electroencephalography (EEG) to monitor epileptic seizures, however, this work is tedious, time-consuming, and largely dependent on the clinician's own experience and subjective judgment, resulting in low accuracy and repeatability of manual detection results. In order to enable epileptic seizures to be detected in a short time, the development of epileptic brain electrical automatic detection and identification technology is particularly critical.
Because the electroencephalogram signals have non-stationarity and non-linear characteristics, the electroencephalogram signals are often distinguished by calculating various non-linear characteristic values by mainly using a traditional time domain, frequency domain or time frequency combination method for electroencephalogram signal analysis. Most of data preprocessing methods are to decompose signals by wavelet transformation, but the selection of layering number and basis function has great influence on the result and does not have the decomposition capability of signal self-adaption. At present, methods such as a Support Vector Machine (SVM), a Decision Tree (DT), a Random Forest (Random Forest) and the like are mostly used for carrying out electroencephalogram classification, but in the actual long-range electroencephalogram detection, the time of an attack period is far shorter than the time of a non-attack period, the decision boundary of a classifier is deviated due to unbalanced data sets, and the classification effect of the model is finally influenced. And a Cost Sensitive Support Vector Machine (CSVM) takes the misclassification costs of different types of samples into consideration during modeling, and embeds the misclassification costs into a standard SVM algorithm.
The features currently used for detecting epileptic signals are mainly the amplitude, dominant frequency, coefficient of variation, entropy, etc. of EEG, and do not consider the information exchange and flow that the brain as an organic whole will inevitably have between parts when behavioural or physiological function changes occur. Pathological studies have shown that the source of epileptic production (known as the focal source) is not completely coincident with the massive seizure brain region, so there is a flow of information from epileptic production to seizure brain region. And because epilepsy is caused by abnormal discharge of cerebral neurons, information flow between different brain areas in the period of epileptic seizure is different from that of normal state, and the difference has both strength difference and direction difference. By using the difference, the status of the epileptic seizure period and the status of the non-seizure period can be effectively distinguished, and the signal detection of the epileptic seizure period can be carried out.
Disclosure of Invention
In order to overcome the defects of the prior art, the present invention provides a system and a method for detecting a seizure time period based on empirical mode decomposition, which decompose an electroencephalogram signal into a plurality of IMF components and residuals by using a Multivariate Empirical Mode Decomposition (MEMD) method, calculate information flow intensity between different brain areas through a Directed Transfer Function (DTF), and input the information flow intensity as an electroencephalogram characteristic value into a cost sensitive support vector machine (CSVM) for distinguishing signals of a seizure period from signals of a non-seizure period.
In order to achieve the purpose, the technical scheme of the invention is as follows:
the utility model provides an epileptic seizure time quantum detecting system based on empirical mode decomposition, includes five modules of EEG signal acquisition module A, self-adaptation decomposition module B, information flow network construction module C, outflow information extraction module D and categorised identification module E:
the electroencephalogram signal acquisition module A: the method is used for collecting multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptics, and numbering electrodes in the multichannel electroencephalogram signals from 1 to N, wherein N is the total number of the electrodes;
the self-adaptive decomposition module B: the method is used for dividing a moving window for the electroencephalogram signal according to the time length and carrying out multi-element empirical mode decomposition (MEMD): dividing the multi-channel electroencephalogram signal data acquired by the electroencephalogram signal acquisition module A into moving window segments according to time length; performing multi-element empirical mode decomposition on each moving window to obtain multi-channel electroencephalogram data IMF components and allowance;
the information flow network construction module C: the method is used for establishing a multivariate autoregressive model MVAR for each IMF component and the residual in the self-adaptive decomposition module B and the original electroencephalogram signal and solving the strength and the direction of a directed transfer function DTF: for each mobile window, determining the order of the MVAR model according to a Bayesian criterion (SBC) criterion, and establishing the MVAR model; the DTF strength and direction among all channels are calculated according to the MVAR model coefficient of the moving window;
the outflow information extraction module D: calculating the outflow information intensity of each channel according to the DTF value obtained by the information flow network construction module C: summing the squares of the DTF values under each frequency between the same pair of channels to obtain the information flow intensity between the pair of channels; calculating the outflow information intensity of each channel according to the DTF matrix;
the classification identification module E: combining cross validation, taking the outflow information intensity in the outflow information extraction module D as a characteristic value, using a cost sensitive support vector machine (CSVM) to perform identification and classification, and inputting the outflow information intensities of all channels as the characteristic value into the CSVM to perform pattern identification and classification; and performing in-sample optimization and out-sample testing simultaneously by using k-fold cross validation, wherein the two classification results of the CSVM are the detection results of the epileptic seizure signals.
The detection method based on the system for detecting the epileptic seizure time period based on empirical mode decomposition comprises the following steps:
(1): the electroencephalogram signal acquisition module A acquires multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptics, numbers electrodes in the multichannel electroencephalogram signals from 1 to N, wherein N is the total number of the electrodes, and the electroencephalogram signals are expressed as X (t) ═ X1(t)，X2(t)，...，XN(T) }, where T denotes the transpose of the matrix, XN(t) (N ═ 1, 2.., N) represents the electroencephalogram signal of the nth channel;
(2): the self-adaptive decomposition module B divides a moving window for the electroencephalogram signal according to the time length and carries out Multivariate Empirical Mode Decomposition (MEMD) to obtain IMF (intrinsic mode function) components and allowance of the multichannel electroencephalogram data;
(3): the information flow network construction module C establishes a multivariate autoregressive Model (MVAR) for each IMF component, the residual quantity and the original electroencephalogram signal and calculates the strength and the direction of a directional transfer function DTF;
(4): the outflow information extraction module D calculates the outflow information intensity of each channel according to the obtained DTF value;
(5): and the classification identification module E combines cross validation to use a cost sensitive support vector machine CSVM to perform identification and classification by taking the outflow information intensity as a characteristic value: inputting the outflow information intensities of all channels into the CSVM as characteristic values to perform pattern recognition classification; and performing in-sample optimization and out-sample testing simultaneously by using k-fold cross validation, wherein the two classification results of the CSVM are the detection results of the epileptic seizure signals.
The step (2) specifically comprises:
(B1) the method comprises the following steps Dividing the collected multi-channel electroencephalogram data into moving window segments according to time length;
(B2) the method comprises the following steps Performing multivariate empirical mode decomposition on each moving window: the brain electrical signal is expressed as follows
Where s denotes the number of IMF components, di(t)＝{di，1(t)，di，2(t)，...di，N(t) denotes the i-th IMF component of N-channel electroencephalogram data, and r (t) { r }1(t)，r2(t)，...rN(t) } denotes an N-channel electroencephalogram data margin.
The step (3) specifically comprises:
(C1) the method comprises the following steps For each moving window, q (t) { q ═ q is defined1(t)，q2(t)，...，qs+2(t) }, in which q isi(t) (i ═ 1, 2.., s +2) is described below
(C2) The method comprises the following steps Determining the order of the MVAR model according to a Bayesian criterion (SBC) criterion, and establishing the MVAR model: the sequence is represented as follows
Where p denotes the order of the MVAR model, ArThe coefficient matrix N × N, r 1, 2.., p, e (t) indicates an estimation error, and the estimation of the coefficient matrix can be obtained by an arfit algorithm.
(C3) The method comprises the following steps And (3) solving the DTF strength and direction among channels according to the MVAR model coefficient of the moving window: for MVAR model coefficient matrix ArFourier transform to obtain
Wherein f represents a discrete frequency variable, and the transfer matrix is defined as H (f) ═ (1-A (f))-1In turn, obtaining a stream of information from lead j to lead i at frequency f
Wherein Hij(f) Element of i-th row and j-th column of H (f), hi(f) The ith column, DTF, of the matrix H (f)ij(f) Representing the strength and direction of information flow from lead j to lead i at frequency f.
The step (4) specifically comprises:
(D1) the method comprises the following steps Summing the squares of the DTF values at each frequency between the same pair of channels to obtain the information flow strength between the channels;
(D2) the method comprises the following steps And calculating the outflow information intensity of each channel according to the DTF matrix: the outflow information for lead j can be given by
The value of the outgoing information represents lead j as the source, leading to the otherThe information sum transmitted in a joint way can represent the information interaction process between a specific brain region and other different brain regions, and considering that the pathological diagnosis of the epileptic attack mainly comprises that nerve cells in the epileptic origin brain region are over-discharged firstly and then transition to the peripheral brain region, so that more information flows out from the epileptic origin brain region in the epileptic attack process, and the information flow change before and after the epileptic attack can be better represented by the outflow information.
The invention has the advantages that:
the multi-element empirical mode decomposition (MEMD) is a self-adaptive signal time-frequency processing method, can perform self-adaptive decomposition on multi-channel data at the same time, is suitable for analyzing electroencephalogram signals with high correlation and non-stationarity, and can enhance the accuracy of frequency information of the positioning electroencephalogram signals. Based on the estimation of a multi-channel autoregressive Model (MVAR), a directed network can be constructed by a Directed Transfer Function (DTF) method, the connection strength between different leads can be reflected, and the direction of information flow can be represented.
The method utilizes the multivariate empirical mode decomposition to carry out self-adaptive decomposition on the original electroencephalogram signal, and enhances the accuracy of the frequency information of the positioning electroencephalogram signal, thereby effectively improving the identification capability of the electroencephalogram signal. On the basis, a new method for detecting the epileptic seizure time slot by utilizing the DTF method and CSVM classification is provided from the perspective of the direction and the strength of information flow of different brain areas of the brain, the method not only can effectively and accurately detect the epileptic seizure time slot, but also can well accord with the pathological characteristics of the epilepsia, and the physiological explanation of the method can be clinically accepted. The method has more accurate detection on the epileptic seizure time slot, can be applied to reducing the workload of reading epileptic EEG clinically, helps to confirm the epileptic seizure time slot better, and can be used as an auxiliary means for the preoperative evaluation of the epileptic surgery. Compared with the algorithm without using MEMD (DTF-CSVM), the result shows that the method has higher accuracy and sensitivity.
Drawings
Fig. 1 is a diagram of an automatic epileptic seizure detection system based on multivariate empirical mode decomposition according to the present invention.
Fig. 2 is a flowchart of an automatic epileptic seizure period detection method based on multivariate empirical mode decomposition according to the present invention.
Fig. 3(a) is a color graph of information flow between channels in a moving window during a patient's episode (considering the information flow from the channel to the channel, the diagonal i-j is not set to zero); fig. 3(b) is a color graph of information flow between channels in a moving window during a non-attack period of a patient (considering information flow from the patient to the channel, the diagonal i-j is not set to zero).
Detailed Description
The present invention will be described in further detail with reference to the accompanying drawings and examples. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
Referring to fig. 1, an automatic epileptic seizure detection system based on multivariate empirical mode decomposition comprises five templates,
the electroencephalogram signal acquisition module A: the method is used for collecting multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptics, and numbering electrodes in the multichannel electroencephalogram signals from 1 to N, wherein N is the total number of the electrodes;
the self-adaptive decomposition module B: the method is used for dividing a moving window of an electroencephalogram signal according to time length and performing multi-element empirical mode decomposition (MEMD): dividing the collected multi-channel electroencephalogram data into moving window segments according to time length; performing multi-element empirical mode decomposition on each moving window to obtain multi-channel electroencephalogram data IMF components and allowance;
the information flow network construction module C: the method is used for establishing a multivariate autoregressive Model (MVAR) for each IMF component, the residual quantity and the original electroencephalogram signal and solving the strength and the direction of a Directional Transfer Function (DTF): for each mobile window, determining the order of the MVAR model according to a Bayesian criterion (SBC) criterion, and establishing the MVAR model; the DTF strength and direction among all channels are calculated according to the MVAR model coefficient of the moving window;
the outflow information extraction module D: and calculating the outflow information intensity of each channel according to the obtained DTF value: summing the squares of the DTF values under each frequency between the same pair of channels to obtain the information flow intensity between the pair of channels; calculating the outflow information intensity of each channel according to the DTF matrix;
a classification identification module E: for identifying and classifying by using a cost sensitive support vector machine (CSVM) with cross validation using outflow information intensity as a characteristic value. Inputting the outflow information intensities of all channels into the CSVM as characteristic values to perform pattern recognition classification; and performing in-sample optimization and out-sample testing simultaneously by using k-fold cross validation, wherein the two classification results of the CSVM are the detection results of the epileptic seizure signals.
Referring to fig. 2, a method for automatically detecting a seizure time period based on multivariate empirical mode decomposition includes the following steps:
(A) the method comprises the following steps Acquiring multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptics, and numbering electrodes in the multichannel electroencephalogram signals from 1 to 19, wherein 19 is the total number of the electrodes: in the embodiment, the electroencephalogram 19 is acquired, the electrode placement position of the electroencephalogram follows the international 10/20 standard, and the electrode on the top of the head CZ is taken as a reference electrode. The sampling frequency is 200Hz, and the passband cut-off frequency of the data is 0.5 Hz-60 Hz.
(B) The method comprises the following steps Dividing a moving window for the electroencephalogram signal according to the time length and performing Multivariate Empirical Mode Decomposition (MEMD) to obtain IMF (intrinsic mode function) components and allowance of the multichannel electroencephalogram data;
the step B specifically comprises the following steps:
(B1) the method comprises the following steps Dividing the acquired multi-channel electroencephalogram data into moving window segments according to time length: respectively dividing 19-channel electroencephalogram data in a non-attack period and an attack period into non-overlapped 2s moving window segments by considering that the minimum frequency of the acquired electroencephalogram signal is 0.5 Hz;
(B2) the method comprises the following steps Performing multivariate empirical mode decomposition on each moving window: by adopting a Hammersley sequence sampling method,
representing corresponding angles on an 18-dimensional sphere
A set of direction vectors, where k is 1, 2. Calculating original EEG signal
In the k-th direction vector
Is projected as
Where T represents the number of samples of the moving window.
Finding projection signals of direction vectors
Instantaneous moment corresponding to extreme value
Where i represents the position of the extreme point, i ∈ [1, T ]]。
Interpolation function difference extreme point by multi-element spline
Resulting in 19 multivariate envelopes
For the 19 directional vectors in spherical space, the local mean is,
extracting a natural mode function d (t) by d (t) ═ X (t) -m (t), if d (t) meets the criteria of judging the multivariate mode function IMF, taking X (t) -m (t) as an input signal to extract new multivariate IMF components again, otherwise, taking m (t) as the input signal to continue screening. Finally, 7 IMF components and a margin are obtained.
(C) The method comprises the following steps Establishing a multivariate autoregressive Model (MVAR) for each IMF component, the residual quantity and the original electroencephalogram signal, and solving the strength and the direction of a Directed Transfer Function (DTF);
the step C specifically comprises the following steps:
(C1) for each moving window, q (t) { q ═ q is defined1(t)，q2(t)，...，q9(t) }, in which q isi(t), i ═ 1, 2.., 9, described below
Wherein X (t) represents 19 channels of original electroencephalogram data, di(t) denotes the i-th IMF component of X (t), and r (t) denotes the residual amount of X (t).
(C2) The method comprises the following steps Determining the order of the MVAR model according to a Bayesian criterion (SBC criterion):
wherein
Representing the white noise variance, p representing the model order, taking 0-20, and N representing the number of samples. According to the SBC criterion, the k value corresponding to the first minimum point in SBC (p) is the model order.
Establishing an MVAR model: by building a multi-channel autoregressive model, qiThe (t) sequence may be expressed as,
where p denotes the order of the MVAR model, ArA coefficient matrix representing nxn, r 1, 2.. p, e (t) representing an estimation error; ideally white uncorrelated noise with a mean value of zero. Coefficient matrix ArThe estimate of (c) can be found using the arfit algorithm.
(C3) The method comprises the following steps And (3) solving the DTF strength and direction among channels according to the MVAR model coefficient of the moving window: for MVAR model coefficient matrix ArFourier transform to obtain
Where f represents a discrete frequency variable.
The transfer matrix is defined as the matrix of,
H(f)＝(1-A(f))-1 (4)
in turn obtaining the strength and direction of information flow from lead j to lead i at frequency f,
wherein Hij(f) Element of i-th row and j-th column of H (f), hi(f) Indicating the ith column, DTF, of the matrix Hij(f) Representing the strength and direction of information flow from lead j to lead i at frequency f.
(D) The method comprises the following steps And calculating the outflow information intensity of each channel according to the obtained DTF value.
The step D specifically comprises the following steps:
(D1) the method comprises the following steps The sum of squares of DTF values under each frequency between the same pair of channels is the information flow strength between the channels, and the total energy of DTF of the information flow from the lead j to the lead i is set as gammaijThen, then
Fig. 3 shows a color graph of information flow between channels of an attack period and a non-attack period of a certain patient (considering information flow from the channel to the channel, the diagonal i-j is not set to zero);
(D2) the method comprises the following steps And calculating the outflow information intensity of each channel according to the DTF matrix: the outflow information for lead j can be given by
The value of the outflow information represents lead j as a source, and the sum of information transmitted to other leads can represent the information flow change before and after the epileptic seizure and is used as a characteristic value for classification.
(E) The method comprises the following steps And performing identification classification by using a cost sensitive support vector machine (CSVM) by taking the strength of the outflow information as a characteristic value in combination with cross validation.
The step E specifically comprises the following steps:
(E1) the method comprises the following steps And inputting the outflow information intensity of all channels into the CSVM as a characteristic value for pattern recognition and classification: to calculate 9 qi(t) the outgoing information strengths of 19 channels of the sequence are used as eigenvalues, which are input into the CSVM for training and testing.
(E2) The method comprises the following steps Performing in-sample optimization and out-sample testing simultaneously by using k-fold cross validation, wherein the two classification results of the CSVM are the detection results of the epileptic seizure time period: in order to ensure that the classification accuracy of the CSVM on the data really represents the actual accuracy, the embodiment adopts 5-fold cross validation, and performs double cross validation to achieve in-sample optimization and out-sample detection. In the outer layer cross validation, the data set was equally divided into 5 groups, each time trained with 4 groups of data, one group of data was used for testing. In the inner-layer cross validation, 80% of the training set is randomly selected, a CSVM model is established, and F is calculated by using the validation results of the model on the rest 20% of the training set2Thereby, the optimal model is evaluated and selected. And after the CSVM model training is finished, testing the model through a reserved test set, and evaluating the model. The results of 5 times were averaged as the final test result.
Reference is made to the following table: the results of the detection of the seizure time period are the CSVM output of 10 tested patients which are subjected to 5-fold cross validation.
The classification result can be evaluated by using the following indexes, wherein True Positive (TP) is the epileptic seizure time period segment judged correctly by the algorithm, False Positive (FP) is the epileptic seizure time period segment judged incorrectly by the algorithm, True Negative (TN) is the non-epileptic seizure time period segment judged correctly by the algorithm, and False Negative (FN) is the non-epileptic seizure time period segment judged incorrectly by the algorithm.
Accuracy (Accuracy)
Selectivity (Selectivity)
Sensitivity (Sensitivity)
F2Value of
The classification results of CSVM on the seizure time period in this embodiment are shown in table 1, and it can be seen that the classification accuracy, selectivity, sensitivity, and F2 value are all relatively high, which are quite consistent with the real detection results, and have relatively high operability and application value.
Claims (5)
1. The utility model provides an epileptic seizure time quantum detecting system based on empirical mode decomposition which characterized in that, includes five modules of EEG signal acquisition module A, self-adaptation decomposition module B, information flow network construction module C, outflow information extraction module D and categorised identification module E:
the electroencephalogram signal acquisition module A: the method is used for collecting multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptics, and numbering electrodes in the multichannel electroencephalogram signals from 1 to N, wherein N is the total number of the electrodes;
the self-adaptive decomposition module B: the method is used for dividing a moving window for the electroencephalogram signal according to the time length and carrying out multi-element empirical mode decomposition (MEMD): dividing the multi-channel electroencephalogram signal data acquired by the electroencephalogram signal acquisition module A into moving window segments according to time length; performing multi-element empirical mode decomposition on each moving window to obtain multi-channel electroencephalogram data IMF components and allowance;
the information flow network construction module C: the method is used for establishing a multivariate autoregressive model MVAR for each IMF component and the residual in the self-adaptive decomposition module B and the original electroencephalogram signal and solving the strength and the direction of a directed transfer function DTF: for each mobile window, determining the order of the MVAR model according to a Bayesian criterion (SBC) criterion, and establishing the MVAR model; the DTF strength and direction among all channels are calculated according to the MVAR model coefficient of the moving window;
the outflow information extraction module D: calculating the outflow information intensity of each channel according to the DTF value obtained by the information flow network construction module C: summing the squares of the DTF values under each frequency between the same pair of channels to obtain the information flow intensity between the pair of channels; calculating the outflow information intensity of each channel according to the DTF matrix;
the classification identification module E: combining cross validation, taking the outflow information intensity in the outflow information extraction module D as a characteristic value, using a cost sensitive support vector machine (CSVM) to perform identification and classification, and inputting the outflow information intensities of all channels as the characteristic value into the CSVM to perform pattern identification and classification; and performing in-sample optimization and out-sample testing simultaneously by using k-fold cross validation, wherein the two classification results of the CSVM are the detection results of the epileptic seizure signals.
2. The detection method of the system for detecting epileptic seizure periods based on Empirical Mode Decomposition (EMD) according to claim 1, characterized by comprising the following steps:
(1): the electroencephalogram signal acquisition module A acquires multichannel electroencephalogram signals of seizure periods and non-seizure periods of epileptics, numbers electrodes in the multichannel electroencephalogram signals from 1 to N, wherein N is the total number of the electrodes, and the electroencephalogram signals are expressed as X (t) ═ X1(t)，X2(t)，...，XN(T) }, where T denotes the transpose of the matrix, XN(t) (N ═ 1, 2.., N) denotes the nth fluxA channel electrical brain signal;
(2): the self-adaptive decomposition module B divides a moving window for the electroencephalogram signal according to the time length and carries out Multivariate Empirical Mode Decomposition (MEMD) to obtain IMF (intrinsic mode function) components and allowance of the multichannel electroencephalogram data;
(3): the information flow network construction module C establishes a multivariate autoregressive Model (MVAR) for each IMF component, the residual quantity and the original electroencephalogram signal and calculates the strength and the direction of a directional transfer function DTF;
(4): the outflow information extraction module D calculates the outflow information intensity of each channel according to the obtained DTF value;
(5): and the classification identification module E combines cross validation to use a cost sensitive support vector machine CSVM to perform identification and classification by taking the outflow information intensity as a characteristic value: inputting the outflow information intensities of all channels into the CSVM as characteristic values to perform pattern recognition classification; and performing in-sample optimization and out-sample testing simultaneously by using k-fold cross validation, wherein the two classification results of the CSVM are the detection results of the epileptic seizure signals.
3. The method for detecting epileptic seizure time period based on Empirical Mode Decomposition (EMD) according to claim 2, wherein the step (2) specifically comprises:
(B1) the method comprises the following steps Dividing the collected multi-channel electroencephalogram data into moving window segments according to time length;
(B2) the method comprises the following steps Performing multivariate empirical mode decomposition on each moving window: the brain electrical signal is expressed as follows
Where s denotes the number of IMF components, di(t)＝{di，1(t)，di，2(t)，...di，N(t) denotes the i-th IMF component of N-channel electroencephalogram data, and r (t) { r }1(t)，r2(t)，...rN(t) } denotes an N-channel electroencephalogram data margin.
4. The method for detecting epileptic seizure time period based on Empirical Mode Decomposition (EMD) according to claim 2, wherein the step (3) specifically comprises:
(C1) the method comprises the following steps For each moving window, q (t) { q ═ q is defined1(t)，q2(t)，...，qs+2(t) }, in which q isi(t) (i ═ 1, 2.., s +2) is described below
(C2) The method comprises the following steps Determining the order of the MVAR model according to a Bayesian criterion (SBC) criterion, and establishing the MVAR model: the sequence is represented as follows
Where p denotes the order of the MVAR model, ArThe coefficient matrix N × N, r 1, 2.., p, e (t) indicates an estimation error, and the estimation of the coefficient matrix can be obtained by an arfit algorithm.
(C3) The method comprises the following steps And (3) solving the DTF strength and direction among channels according to the MVAR model coefficient of the moving window: for MVAR model coefficient matrix ArFourier transform to obtain
Wherein f represents a discrete frequency variable, and the transfer matrix is defined as H (f) ═ (1-A (f))-1, and thus the information flow from lead j to lead i in frequency f
Wherein Hij(f) Element of i-th row and j-th column of H (f), hi(f) The ith column, DTF, of the matrix H (f)ij(f) Representing the strength and direction of information flow from lead j to lead i at frequency f.
5. The method for detecting epileptic seizure time period based on Empirical Mode Decomposition (EMD) according to claim 2, wherein the step (4) specifically comprises:
(D1) the method comprises the following steps Summing the squares of the DTF values at each frequency between the same pair of channels to obtain the information flow strength between the channels;
(D2) the method comprises the following steps And calculating the outflow information intensity of each channel according to the DTF matrix: the outflow information for lead j can be given by
The value of the outflow information represents that lead j is used as a source, the sum of information transmitted to other leads can represent the information interaction process between a specific brain region and other different brain regions, and the pathological diagnosis of epileptic attack is considered that nerve cells in the brain region of the epileptic origin are over-discharged firstly and then transited to the peripheral brain region, so that more information flows out of the brain region of the epileptic origin in the epileptic attack process, and the information flow change before and after the epileptic attack can be better represented by the outflow information.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110961989.2A CN113768464A (en) | 2021-08-20 | 2021-08-20 | Epileptic seizure time period detection system and method based on empirical mode decomposition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110961989.2A CN113768464A (en) | 2021-08-20 | 2021-08-20 | Epileptic seizure time period detection system and method based on empirical mode decomposition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113768464A true CN113768464A (en) | 2021-12-10 |
Family
ID=78838383
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110961989.2A Pending CN113768464A (en) | 2021-08-20 | 2021-08-20 | Epileptic seizure time period detection system and method based on empirical mode decomposition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113768464A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114052668A (en) * | 2022-01-17 | 2022-02-18 | 北京航空航天大学杭州创新研究院 | Brain function analysis method based on magnetoencephalogram data |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104720796A (en) * | 2015-02-12 | 2015-06-24 | 西安交通大学 | Automatic detecting system and method for epileptic attack time period |
| CN110367980A (en) * | 2019-07-10 | 2019-10-25 | 南京邮电大学 | EEG signals Emotion identification method based on polynary empirical mode decomposition |
-
2021
- 2021-08-20 CN CN202110961989.2A patent/CN113768464A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104720796A (en) * | 2015-02-12 | 2015-06-24 | 西安交通大学 | Automatic detecting system and method for epileptic attack time period |
| CN110367980A (en) * | 2019-07-10 | 2019-10-25 | 南京邮电大学 | EEG signals Emotion identification method based on polynary empirical mode decomposition |
Non-Patent Citations (3)
| Title |
|---|
| SUMAN DUTTA 等: "Automated classification of non-motor mental task inelectroencephalogram based brain-computer interface using multivariate autoregressive model in the intrinsic mode function domain", 《BIOMEDICAL SIGNAL PROCESSING AND CONTROL》 * |
| 李冬梅 等: "经验模式分解与代价敏感支持向量机在癫痫脑电信号分类中的应用", 《生物医学工程研究》 * |
| 王栋 等: "利用高频脑电的局灶性癫痫患者癫痫发作检测", 《西安交通大学学报》 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114052668A (en) * | 2022-01-17 | 2022-02-18 | 北京航空航天大学杭州创新研究院 | Brain function analysis method based on magnetoencephalogram data |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Aslan et al. | Automatic Detection of Schizophrenia by Applying Deep Learning over Spectrogram Images of EEG Signals. | |
| CN111340142B (en) | Epilepsia magnetoencephalogram spike automatic detection method and tracing positioning system | |
| CN104720796B (en) | A kind of automatic checkout system and method for epileptic attack time section | |
| CN109645990B (en) | Computer mode identification method for electroencephalogram signals of epileptics | |
| CN104586387B (en) | Method for extracting and fusing time, frequency and space domain multi-parameter electroencephalogram characters | |
| CN109480833A (en) | The pretreatment and recognition methods of epileptic's EEG signals based on artificial intelligence | |
| Boashash et al. | A methodology for time-frequency image processing applied to the classification of non-stationary multichannel signals using instantaneous frequency descriptors with application to newborn EEG signals | |
| CN111091074A (en) | Motor imagery electroencephalogram signal classification method based on optimal region common space mode | |
| CN113052113A (en) | Depression identification method and system based on compact convolutional neural network | |
| CN113662560B (en) | Method for detecting seizure-like discharge between attacks, storage medium and device | |
| CN108511055A (en) | Ventricular premature beat identifying system and method based on Multiple Classifier Fusion and diagnostic rule | |
| CN115081486B (en) | System and method for positioning epileptic focus by using intracranial brain electrical network in early stage of epileptic seizure | |
| CN112057087B (en) | Autonomic nerve function data processing method and device for high-risk schizophrenic people | |
| CN114970599A (en) | Identification method and identification device for attention defect associated electroencephalogram signals and storage medium | |
| CN111067513B (en) | Sleep quality detection key brain area judgment method based on characteristic weight self-learning | |
| CN114532993A (en) | Automatic detection method for electroencephalogram high-frequency oscillation signals of epileptic | |
| CN113768519A (en) | Method for analyzing consciousness level of patient based on deep learning and resting state electroencephalogram data | |
| TWI628625B (en) | Method of producing imaging biomarker for brain disease | |
| CN113768464A (en) | Epileptic seizure time period detection system and method based on empirical mode decomposition | |
| JP2003310564A (en) | Automatic brain wave analyzing apparatus and method | |
| CN113128585B (en) | Deep neural network based multi-size convolution kernel method for realizing electrocardiographic abnormality detection and classification | |
| Zhou et al. | A hybrid expert system for individualized quantification of electrical status epilepticus during sleep using biogeography-based optimization | |
| CN111643076A (en) | BECT spike intelligent detection method based on multi-channel electroencephalogram signals | |
| CN116484290A (en) | Depression recognition model construction method based on Stacking integration | |
| CN110200624A (en) | Based on convolutional neural networks-Recognition with Recurrent Neural Network-support vector machines mixed model disease identification algorithm |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20211210 |