CN116196018B - Dynamic identification method and system for epileptic brain electrical signals based on sharpness and nonlinearity - Google Patents
Dynamic identification method and system for epileptic brain electrical signals based on sharpness and nonlinearity Download PDFInfo
- Publication number
- CN116196018B CN116196018B CN202310484796.1A CN202310484796A CN116196018B CN 116196018 B CN116196018 B CN 116196018B CN 202310484796 A CN202310484796 A CN 202310484796A CN 116196018 B CN116196018 B CN 116196018B
- Authority
- CN
- China
- Prior art keywords
- signal
- electroencephalogram
- multidimensional
- dimension
- epileptic
- 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
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
- A61B5/24—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
- A61B5/24—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]
-
- 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
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
- A61B5/7207—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal of noise induced by motion artifacts
-
- 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
- A61B5/7225—Details of analog processing, e.g. isolation amplifier, gain or sensitivity adjustment, filtering, baseline or drift compensation
-
- 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
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2503/00—Evaluating a particular growth phase or type of persons or animals
- A61B2503/04—Babies, e.g. for SIDS detection
- A61B2503/045—Newborns, e.g. premature baby monitoring
Abstract
The invention discloses a dynamic identification method and a dynamic identification system for epileptic brain electrical signals based on sharpness and nonlinearity, which take complex envelope, sharpness and nonlinearity of an electroencephalogram as key waveform characteristics for exploring epileptic seizures, realize the identification of epileptic seizure events by utilizing more accurate dynamic waveforms, and improve the robustness of the identification of neonatal epileptic seizure events. The invention executes the following scheme: and acquiring multichannel brain electrical signals acquired by the brain electrical equipment. Preprocessing the data of each channel in the acquired electroencephalogram signals one by one, wherein each channel serves as one dimension, and the preprocessed multidimensional electroencephalogram signals are obtained. And taking the corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as key waveform characteristics of the electroencephalogram signal. And the key waveform characteristics of the electroencephalogram signals are used as input machine learning classification models to dynamically identify the epileptic electroencephalogram signals.
Description
Technical Field
The invention relates to the crossing field of neuroscience and information technology, in particular to a dynamic identification method and a system for epileptic brain electrical signals based on sharpness and nonlinearity.
Background
Neonatal seizures have become one of the leading causes of impaired neonatal neurological function and death, whereas term neonatal seizures with stroke, meningitis and hypoxic ischemic encephalopathy may be associated with potential neurological dysfunction, which makes it possible to describe the progress of seizures by neuroelectric features. At present, the judgment of the neonatal epileptic seizure mainly depends on clinical reports, and the screening mode not only needs a great deal of manpower, but also is used as a subjective evaluation, and lacks objective reliability, for example, only 9% of the electroencephalograms of the neonatal epileptic seizure are found to accord with the clinical evaluation in Murray and the like, so that the neonatal epileptic occurrence rate, especially infants with subtle pure clinical manifestations, is seriously underestimated. In addition, the clinical scores are affected by other factors including lack of consistency between the evaluation professionals, different frequencies of seizures/artifacts, etc. at the non-visit time of the neonatal intensive care unit.
In neonatal intensive care units, video electroencephalograms are widely used due to the characteristics of high time resolution, high accuracy and suitability for long-time monitoring. However, the baseThe epileptic screening method in electroencephalogram still requires a great deal of manpower, and also relies on subjective judgment. Amplitude integrated electroencephalogram (aEEG) is a multiwavelocimeter that replaces video electroencephalogram. Traditional analysis is based on the energy and entropy index of aEEG, for example, the power of the band of aEEG changes in infants taking two antiepileptic drugs, i.e. midazolam or lidocaine, both drugs induce energy changes in each frequency bandδ、θThe power is reduced and the power consumption is reduced,βthe power remains unchangedαThe power is unchanged when midazolam is taken, and the power tends to increase when lidocaine is taken. However, there is more evidence that aEEG ignores short-term seizures, particularly brain signal segments with rare, focal, transient and low-amplitude patterns, and that the sensitivity is not high. Greene et al extract the characteristics of the time domain, the frequency domain, the entropy and the like of the brain electrical signals, construct a classifier model for neonatal epileptic seizure detection, and have sensitivity and specificity of less than 85 percent.
The bedside EEG monitor can collect multichannel EEG signals at the same time and support long-time aEEG monitoring. An electroencephalogram may exhibit a remarkably abnormal discharge pattern such as a spike complex, a multi-spike discharge, an epileptiform K complex, etc., and these electroencephalogram waveforms having a specific shape are considered as important factors for distinguishing epileptiform discharges from normal electroencephalograms. In addition, the literature indicates that the waveform characteristics of the brain electrical signals are closely related to sleep modes and pathological limb movement states, so that the analysis of the waveform characteristics of the brain electrical signals has the opportunity to enhance the understanding of the physiological mechanism of brain nerves. For example, in experiments in which rats were induced to develop temporal lobe seizures by pilocarpine injection, individuals with relatively long ascending waveform phases of low frequency brain electrical signals were found to have a higher likelihood of developing epileptic states than those with shorter ascending times than decay times. For absence epilepsy, spike wave oscillation in an electroencephalogram waveform and the burst activity of neurons of a substantia nigra reticula part causing epileptic seizure are balanced with each other, and the difference of waveform characteristics is quantified, so that the progress of absence seizure can be dynamically reflected. Electroencephalogram signalβFrequency oversynchronization is a characteristic feature of parkinson's disease, and such neuronal synchronization can lead to oscillations in the brain electrical signalThe waveform is sharpened. Also, the nonlinear characteristics of the waveform can reflect subthalamic nucleiβThe synchronous change during the burst is made,βthe nonlinearity of the later decreasing segment of the burst will increase and the sharpness will gradually decrease. Analysis of waveform characteristics is useful for monitoring and understanding the onset of sudden, pathological, synchronized neuronal activity. However, in the exploration of neonatal epilepsy, there are currently few quantitative indicators for waveform dynamics, and the artifacts in electroencephalogram are similar to the real epileptic pattern, more importantly, the epileptic seizure pattern of neonatal electroencephalogram has a higher structure, characterized by periodic rhythmic spikes or repeated spikes. Therefore, there is a need to reveal more accurate dynamic waveforms, improving the robustness of neonatal epilepsy recognition.
Disclosure of Invention
In view of the above, the invention provides a dynamic identification method and a dynamic identification system for epileptic brain electrical signals based on sharpness and nonlinearity, which take complex envelope, sharpness and nonlinearity of an electroencephalogram as key waveform characteristics for exploring epileptic seizures, realize the identification of epileptic brain electrical signals by utilizing more accurate dynamic waveforms, and improve the robustness of the identification of neonatal epileptic brain electrical signals.
In order to achieve the above purpose, the technical scheme of the invention comprises the following steps:
and acquiring multichannel brain electrical signals acquired by the brain electrical equipment.
Preprocessing the data of each channel in the acquired electroencephalogram signals one by one, wherein each channel serves as one dimension, and the preprocessed multidimensional electroencephalogram signals are obtained.
And taking the corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as key waveform characteristics of the electroencephalogram signal.
And (3) utilizing key waveform characteristics of the electroencephalogram signals as input machine learning classification models to dynamically identify the epileptic electroencephalogram signals.
Preferably, preprocessing is performed one by one on the data of each channel in the acquired electroencephalogram signals, specifically:
filtering processing is carried out on data to be processed, and the filtering processing comprises the following steps: removing power frequency interference by using a notch filter; a high pass filter is used to remove motion artifacts.
Performing dimension reduction processing on the filtered data, obtaining a principal component coefficient matrix, a principal component score matrix and a characteristic value vector by adopting a principal component analysis method, accumulating the characteristic values, dividing the accumulated characteristic values by the sum, retaining the characteristic values with accumulated contributions not less than a set value, and recording the number of the characteristic values asKFront of principal component coefficient matrix and principal component score matrixKColumn multiplication to obtain preprocessed multidimensional EEG signalsx i (t),i= 1, 2, …,N,NIs the dimension of the brain electrical signal.
Further, the preprocessed multidimensional computer signal is provided with corresponding complex envelope, sharpness and nonlinearity in each dimension, and the complex envelope is taken as key waveform characteristics of the electroencephalogram signal, wherein the complex envelope is specifically provided with the following steps:
the preprocessed multidimensional electroencephalogram signal is {x i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; signal processing in each dimensionmSecond non-overlapping window segmentation to obtainMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments.
For each multi-dimensional signal segment x i,j ,i= 1, 2, …,N,j=1, 2, …,MPerforming Hilbert transform to obtain analysis signal
The envelope signal of the resolved signal is +.>
。
Calculating envelope signal of each section of analysis signal
Is denoted as x i,j Thereby obtaining a complex envelope for each multi-dimensional signal segment divided by the multi-dimensional computer signal.
Further, the preprocessed multidimensional computer signal is provided with corresponding complex envelope, sharpness and nonlinearity in each dimension, and the complex envelope, sharpness and nonlinearity are taken as key waveform characteristics of the electroencephalogram signal, wherein the sharpness taking process specifically comprises the following steps:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments; one by one adaptive labeling multidimensional signal segment x i,j Is used for the extreme value of (a),i= 1, 2, …,N,j=1, 2, …,Mthe self-adaptive labeling process comprises the following steps: using a sliding Tukey window pair x i,j Performing difference to obtain a difference sequence y i,j Mark y i,j A point smaller than 0 is a falling-stage zero-crossing point, a point larger than 0 is a rising-stage zero-crossing point, a maximum value between the rising-stage zero-crossing point and a next falling-stage zero-crossing point adjacent to the rising-stage zero-crossing point is defined as a wave crest, and a minimum value between the falling-stage zero-crossing point and an immediately next rising-stage zero-crossing point is defined as a wave trough; the collection of peaks and valleys is defined as the extremum.
Taking the extreme point peak of the marked extreme point peak
Two sides of extreme pointtCorresponding to millisecond
And->
And average the absolute voltage difference peak 。
Calculation ofmSecond length x i,j Sharp corresponding to all extreme points in the band peak And find the median, denoted as x i,j Thereby obtaining the sharpness of each multi-dimensional signal segment divided by the multi-dimensional computer signal.
Further, the preprocessed multidimensional computer signal is provided with corresponding complex envelope, sharpness and nonlinearity in each dimension, and the complex envelope, sharpness and nonlinearity are taken as key waveform characteristics of the electroencephalogram signal, wherein the process of taking nonlinearity is specifically as follows:
for the multidimensional electroencephalogram signal { after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j Calculate x i,j Is used for the instantaneous frequency and the zero crossing frequency of the transformer.
By x i,j Extraction of the instantaneous frequency and zero crossing frequency of the signal segment x in relation to the multidimensional signal segment i,j Standard deviation of normalized instantaneous frequency as non-linearityDoN i,j Thus, the nonlinearity of each multidimensional signal segment divided by the multidimensional computer signal is obtained.
Further, the preprocessed multidimensional computer signal takes corresponding complex envelope, sharpness and nonlinearity in each dimension as key waveform characteristics of the electroencephalogram signal, and specifically comprises the following steps:
for EEG signal sequence {x i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j 。
For the complex envelope taken in the ith dimension isE i =[E 1i ,E 2i ,E 3i ,…,E Mi ]The sharpness isS i =[S 1i ,S 2i ,S 3i ,…,S Mi ]The nonlinearity isD i =[D 1i ,D 2i ,D 3i ,…,D Mi ]；E Mi ,S Mi ,D Mi The complex envelope, sharpness and nonlinearity that the mth multidimensional signal segment corresponds to.
And ordering all N multiplied by 3 groups of M-dimensional extraction features by using a minimum redundancy-maximum correlation algorithm, wherein the M-dimensional extraction features are used as key waveform features of the electroencephalogram signals.
The invention also provides an epileptic electroencephalogram dynamic identification system based on sharpness and nonlinearity, which comprises an electroencephalogram acquisition module, an epileptic nonlinear waveform characteristic calculation module and an epileptic electroencephalogram dynamic identification module.
The electroencephalogram signal acquisition module is used for acquiring multichannel electroencephalogram signals acquired by the electroencephalogram equipment, preprocessing the data of each channel in the acquired electroencephalogram signals one by one, taking each channel as one dimension, obtaining preprocessed multidimensional electroencephalogram signals, and sending the preprocessed multidimensional electroencephalogram signals to the epileptic nonlinear waveform characteristic calculation module.
And the epileptic nonlinear waveform characteristic calculation module is used for taking the corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as the key waveform characteristic of the electroencephalogram signal and sending the key waveform characteristic to the epileptic electroencephalogram signal dynamic identification module.
And the epileptic electroencephalogram dynamic identification module utilizes key waveform characteristics of the electroencephalogram as an input machine learning classification model to dynamically identify the epileptic electroencephalogram.
Preferably, the epileptic nonlinear waveform characteristic calculation module adopts a complex envelope process specifically as follows:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; signal processing in each dimensionmSecond non-overlapping window segmentation to obtainMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments.
For each multi-dimensional signal segment x i,j ,i= 1, 2, …,N,j=1, 2, …,MPerforming Hilbert transform to obtain analysis signal
The envelope signal of the resolved signal is +.>
。
Calculating envelope signal of each section of analysis signal
Is denoted as x i,j Thereby obtaining a complex envelope for each multi-dimensional signal segment divided by the multi-dimensional computer signal.
Preferably, the epileptic nonlinear waveform characteristic calculation module takes sharpness as follows:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments; one by one adaptive labeling multidimensional signal segment x i,j Is used for the extreme value of (a),i= 1, 2, …,N,j=1, 2, …,Mthe self-adaptive labeling process comprises the following steps: using a sliding Tukey window pair x i,j Performing difference to obtain a difference sequence y i,j Mark y i,j A point smaller than 0 is a falling-stage zero-crossing point, a point larger than 0 is a rising-stage zero-crossing point, a maximum value between the rising-stage zero-crossing point and a next falling-stage zero-crossing point adjacent to the rising-stage zero-crossing point is defined as a wave crest, and a minimum value between the falling-stage zero-crossing point and an immediately next rising-stage zero-crossing point is defined as a wave trough; the collection of peaks and valleys is defined as the extremum.
Taking the extreme point peak of the marked extreme point peak
Two sides of extreme pointtCorresponding to millisecond
And->
And average the absolute voltage difference peak 。
Calculation ofmSecond length x i,j In the inner partSharp corresponding to all extreme points peak And find the median, denoted as x i,j Thereby obtaining the sharpness of each multi-dimensional signal segment divided by the multi-dimensional computer signal.
Preferably, the epileptic nonlinear waveform characteristic calculation module specifically adopts the process of taking nonlinearity as follows:
for the multidimensional electroencephalogram signal { after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j Calculate x i,j Is used for the instantaneous frequency and the zero crossing frequency of the transformer.
By x i,j Extraction of the instantaneous frequency and zero crossing frequency of the signal segment x in relation to the multidimensional signal segment i,j Standard deviation of normalized instantaneous frequency as non-linearityDoN i,j Thus, the nonlinearity of each multidimensional signal segment divided by the multidimensional computer signal is obtained.
The beneficial effects are that:
1. the invention provides a dynamic identification method of epileptic brain electrical signals based on sharpness and nonlinearity, which takes complex envelope, sharpness and nonlinearity combination of an electroencephalogram as key waveform characteristics, utilizes a plurality of machine learning classification models to identify epileptic brain electrical signals, improves the accuracy of identifying epileptic brain electrical signal waveforms, and the complex envelope, sharpness and nonlinearity are all nonlinear characteristics, which do not independently represent epileptic brain electrical signals and seizure dynamics, thus having complementary characteristics and improving the robustness of identifying epileptic brain electrical signals.
2. The invention also provides an epileptic electroencephalogram dynamic identification system based on sharpness and nonlinearity, and the system result shows that the set of nerve oscillation characteristics, complex envelope, epileptic sharpness and nonlinearity can identify the epileptic electroencephalogram of the neonate, invert the epileptic seizure dynamics of the neonate, and has a certain capability of accurately identifying the epileptic electroencephalogram. The three epileptic nonlinear characteristics extracted by the system are not used for independently representing epileptic brain electrical signals and seizure dynamics, have the characteristic of complementation, and have certain robustness in the aspect of identification of epileptic brain electrical signals.
Drawings
Fig. 1 is a block diagram of an epileptic electroencephalogram signal dynamic identification system based on sharpness and nonlinearity provided by an embodiment of the invention;
FIG. 2 (A) is a schematic diagram of multichannel neonatal electroencephalogram and corresponding waveform characteristics;
FIG. 2 (B) is a schematic diagram of the corresponding trajectories of complex envelope, sharpness and nonlinearity, wherein the seizure phase is represented within the rectangular box;
FIG. 3 (A) is a schematic distribution of three waveform characteristics of seizures versus no seizures;
fig. 3 (B) is a schematic diagram showing the distribution of three waveform characteristics of the beginning and ending phases of the seizure.
Detailed Description
The invention will now be described in detail by way of example with reference to the accompanying drawings.
Example 1
The invention provides a dynamic identification method of epileptic brain electrical signals based on sharpness and nonlinearity, which comprises the following specific steps:
step one: acquiring multichannel electroencephalogram signals acquired by electroencephalogram equipment;
step two: preprocessing the data of each channel in the acquired electroencephalogram signal one by one, wherein each channel is used as a dimension to obtain preprocessed multidimensional electroencephalogram signals; the pretreatment process in the embodiment of the invention specifically comprises the following steps:
filtering processing is carried out on data to be processed, and the filtering processing comprises the following steps: removing power frequency interference by using a notch filter; a high pass filter is used to remove motion artifacts.
The notch filter adopted in the embodiment of the invention is a 50/60-Hz notch filter.
In the embodiment of the invention, a high-pass filter with the cutoff frequency of 0.5/1-Hz is adopted to remove the motion artifact to obtain the signalx f (t) The method comprises the steps of carrying out a first treatment on the surface of the For signalsx f (t) By usingz-score, i.e
Obtainingx z (t) Minimizing the effect of electrode impedance differences, wherein +.>
Andsrespectively isx f (t) Mean and standard deviation of (c).
Performing dimension reduction processing on the filtered data, obtaining a principal component coefficient matrix, a principal component score matrix and a characteristic value vector by adopting a principal component analysis method, accumulating characteristic values, dividing the accumulated characteristic values by a sum, retaining characteristic values with accumulated contribution not less than a set value (set as 95% in the embodiment of the invention), and recording the number of the characteristic values as followsKFront of principal component coefficient matrix and principal component score matrixKColumn multiplication to obtain preprocessed multidimensional EEG signalsx i (t),i= 1, 2, …,N,NIs the dimension of the brain electrical signal.
Step three: and taking the corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as key waveform characteristics of the electroencephalogram signal.
In the embodiment of the invention, the process of taking the complex envelope is specifically as follows:
the preprocessed multidimensional electroencephalogram signal is {x i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; signal processing in each dimensionmSecond non-overlapping window segmentation to obtainMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments;
for each multi-dimensional signal segment x i,j ,i= 1, 2, …,N,j=1, 2, …,MPerforming Hilbert transform to obtain analysis signal
The envelope signal of the resolved signal is +.>
；
Calculating envelope signal of each section of analysis signal
Is denoted as x i,j Thereby obtaining a complex envelope for each multi-dimensional signal segment divided by the multi-dimensional computer signal.
In the embodiment of the invention, the sharpness taking process specifically comprises the following steps:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，NIs the dimension of the signal,MIs the total number of signal segments; one by one adaptive labeling multidimensional signal segment x i,j Is used for the extreme value of (a),i= 1, 2, …,N,j=1, 2, …,Mthe self-adaptive labeling process comprises the following steps: using a sliding Tukey window pair x i,j Performing difference to obtain a difference sequence y i,j Mark y i,j A point smaller than 0 is a falling-stage zero-crossing point, a point larger than 0 is a rising-stage zero-crossing point, a maximum value between the rising-stage zero-crossing point and a next falling-stage zero-crossing point adjacent to the rising-stage zero-crossing point is defined as a wave crest, and a minimum value between the falling-stage zero-crossing point and an immediately next rising-stage zero-crossing point is defined as a wave trough; the collection of peaks and valleys is defined as an extremum;
taking the extreme points of the marked extreme points
Two sides of extreme pointtCorresponding +.>
And
and average the absolute voltage difference peak The formula is as follows:
calculation ofmSecond length x i,j Sharp corresponding to all extreme points in the band peak And find the median, denoted as x i,j Thereby obtaining the sharpness of each multi-dimensional signal segment divided by the multi-dimensional computer signal.
In the embodiment of the invention, the process of taking the nonlinearity is specifically as follows:
for the multidimensional electroencephalogram signal { after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j Calculate x i,j Is used for the instantaneous frequency and the zero crossing frequency of the transformer;
x i,j is of instantaneous frequency ofIF i,j For x i,j Performing Hilbert transform to obtain a phase function
For a pair ofθ i,j (t) Deriving and obtainingIF i,j I.e.IF i,j = dθ i,j (t)/dt；
x i,j Is of zero crossing frequency of
Wherein sgn (‧) is a sign function, i.eIF z The jth signalx i (t) Inner symbol rate of change.
By x i,j Extraction of the instantaneous frequency and zero crossing frequency of the signal segment x in relation to the multidimensional signal segment i,j Standard deviation of normalized instantaneous frequency as non-linearityDoN i,j Thus, the nonlinearity of each multidimensional signal segment divided by the multidimensional computer signal is obtained.
In the embodiment of the invention, the nonlinearity degreeDoN i,j The definition is as follows:
for EEG signal sequence {x i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j 。
For the complex envelope taken in the ith dimension isE i =[E 1i ,E 2i ,E 3i ,…,E Mi ]The sharpness isS i =[S 1i ,S 2i ,S 3i ,…,S Mi ]The nonlinearity isD i =[D 1i ,D 2i ,D 3i ,…,D Mi ]；E Mi ,S Mi ,D Mi The complex envelope, sharpness and nonlinearity that the mth multidimensional signal segment in the ith dimension corresponds to.
And ordering all N multiplied by 3 groups of M-dimensional extraction features by using a minimum redundancy-maximum correlation algorithm, wherein the M-dimensional extraction features are used as key waveform features of the electroencephalogram signals.
Step four: and (3) utilizing key waveform characteristics of the electroencephalogram signals as input machine learning classification models to dynamically identify the epileptic electroencephalogram signals.
In the embodiment of the invention, the applicable machine learning classification model comprises a linear regression model, a support vector machine, a Gaussian process regression model, a classification tree and other models.
Example 2:
the embodiment provides an epileptic electroencephalogram signal dynamic identification system based on sharpness and nonlinearity, and a framework of the epileptic electroencephalogram signal dynamic identification system is shown in fig. 1, and the epileptic electroencephalogram signal dynamic identification system comprises an electroencephalogram signal acquisition module, an epileptic nonlinear waveform characteristic calculation module and an epileptic electroencephalogram signal dynamic identification module.
The electroencephalogram signal acquisition module is used for acquiring multi-channel electroencephalogram signals acquired by electroencephalogram equipment, preprocessing the data of each channel in the acquired electroencephalogram signals one by one, taking each channel as one dimension, acquiring preprocessed multi-dimensional electroencephalogram signals, and sending the preprocessed multi-dimensional electroencephalogram signals to the epileptic nonlinear waveform characteristic calculation module;
and the epileptic nonlinear waveform characteristic calculation module is used for taking the corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as the key waveform characteristic of the electroencephalogram signal and sending the key waveform characteristic to the epileptic electroencephalogram signal dynamic identification module.
And the epileptic electroencephalogram dynamic identification module utilizes key waveform characteristics of the electroencephalogram as an input machine learning classification model to dynamically identify the epileptic electroencephalogram.
The epileptic nonlinear waveform characteristic calculation module adopts the complex envelope process specifically as follows:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; signal processing in each dimensionmSecond non-overlapping window segmentation to obtainMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments.
For each multi-dimensional signal segment x i,j ,i= 1, 2, …,N,j=1, 2, …,MPerforming Hilbert transform to obtain analysis signal
The envelope signal of the resolved signal is +.>
。
Calculating envelope signal of each section of analysis signal
Is denoted as x i,j Thereby obtaining a complex envelope for each multi-dimensional signal segment divided by the multi-dimensional computer signal.
The sharpness taking process comprises the following steps:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signal in each dimension is not duplicatedObtained by splitting laminated windowsMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i= 1, 2, …,N,j=1, 2, …,M，MIs the total number of signal segments; one by one adaptive labeling multidimensional signal segment x i,j Is used for the extreme value of (a),i= 1, 2, …,N,j=1, 2, …,Mthe self-adaptive labeling process comprises the following steps: using a sliding Tukey window pair x i,j Performing difference to obtain a difference sequence y i,j Mark y i,j A point smaller than 0 is a falling-stage zero-crossing point, a point larger than 0 is a rising-stage zero-crossing point, a maximum value between the rising-stage zero-crossing point and a next falling-stage zero-crossing point adjacent to the rising-stage zero-crossing point is defined as a wave crest, and a minimum value between the falling-stage zero-crossing point and an immediately next rising-stage zero-crossing point is defined as a wave trough; the collection of peaks and valleys is defined as the extremum.
Taking the extreme points of the marked extreme points
Two sides of extreme pointtCorresponding +.>
And
and average the absolute voltage difference peak The formula is as follows:
calculation ofmSecond length x i,j Sharp corresponding to all extreme points in the band peak And find the median, denoted as x i,j Thereby obtaining the sharpness of each multi-dimensional signal segment divided by the multi-dimensional computer signal.
The process of taking the nonlinearity is specifically as follows:
for the multidimensional electroencephalogram signal { after pretreatmentx i (t),i= 1, 2, …,N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j Calculate x i,j Is used for the instantaneous frequency and the zero crossing frequency of the transformer.
x i,j Is of instantaneous frequency ofIF i,j For x i,j Performing Hilbert transform to obtain a phase function
For a pair ofθ i,j (t) Deriving and obtainingIF i,j I.e.IF i,j = dθ i,j (t)/dt；
x i,j Is of zero crossing frequency of
Wherein sgn (‧) is a sign function, i.eIF z The jth signalx i (t) Inner symbol rate of change.
By x i,j Extraction of the instantaneous frequency and zero crossing frequency of the signal segment x in relation to the multidimensional signal segment i,j Standard deviation of normalized instantaneous frequency as non-linearityDoN i,j Thus, the nonlinearity of each multidimensional signal segment divided by the multidimensional computer signal is obtained.
In the embodiment of the invention, the nonlinearity degreeDoN i,j The definition is as follows:
the embodiment of the invention is illustrated by taking a group of neonatal epilepsy data sets disclosed by Zenodo as an example, wherein the data sets comprise 79 patients, the ages of the patients after passing are different from 35 to 45 weeks, the electroencephalogram signals of 18 electrodes are recorded, and the average monitoring time is 74 minutes. Meanwhile, the starting time and ending time of each seizure of each patient are known. The sampling frequency of the electroencephalogram signals of each channel of the database is 256Hz, wherein the time sequence of the multi-dimensional electroencephalogram signals 200 seconds data of 18 bipolar leads before and after a certain attack of a patient is shown in fig. 2 (A), wherein the rectangular frame range represents the period of epileptic attack, and fig. 2 (B) represents the corresponding track of complex envelope, sharpness and nonlinearity.
The system result shows that the set of nerve oscillation characteristics, complex envelope, epileptic sharpness and nonlinearity can identify the neonatal epileptic brain electrical signals and invert the neonatal epileptic seizure dynamics. The three epileptic nonlinear characteristics are not used for independently representing epileptic brain electrical signals and seizure dynamics, and have the characteristic of complementation. For example, fig. 3 (a) shows a waveform profile of seizure and non-seizure phases, during which seizures exhibit a higher complex envelope, sharpness, with increasing nonlinearity. To further study the dynamic evolution of epileptic brain electrical signals, fig. 3 (B) shows the feature distribution of the ending and starting phases of epileptic seizures. The end phase nonlinearity is generally high while sharpness is inversely related to nonlinearity. The method for processing the neonatal epileptic brain electrical signal based on sharpness and nonlinearity can be used as an effective system for further monitoring and treating neonatal epileptic.
In summary, the above embodiments are only preferred embodiments of the present invention, and are not intended to limit the scope of the present invention. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (4)
1. The dynamic epileptic electroencephalogram signal identification method based on sharpness and nonlinearity is characterized by comprising the following steps:
acquiring multichannel electroencephalogram signals acquired by electroencephalogram equipment;
preprocessing the data of each channel in the acquired electroencephalogram signal one by one, wherein each channel is used as a dimension to obtain preprocessed multidimensional electroencephalogram signals;
taking the corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as key waveform characteristics of the electroencephalogram signal; the process of taking the complex envelope is specifically as follows:
the preprocessed multidimensional electroencephalogram signal is {x i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; in each dimensionSignal of degreemSecond non-overlapping window segmentation to obtainMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i = 1, 2, …, N, j=1, 2, …, M，MIs the total number of signal segments;
for each multi-dimensional signal segment x i,j , i = 1, 2, …, N, j=1, 2, …, MPerforming Hilbert transform to obtain analysis signal
The envelope signal of the resolved signal is +.>
；
Calculating envelope signal of each section of analysis signal
Is denoted as x i,j Thereby obtaining a complex envelope of each multi-dimensional signal segment divided by the multi-dimensional computer signal;
the sharpness taking process comprises the following steps:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i = 1, 2, …, N, j=1, 2, …, M，NIs the dimension of the signal,MIs the total number of signal segments; one by one adaptive labeling multidimensional signal segment x i,j Is used for the extreme value of (a),i = 1, 2, …, N, j=1, 2, …, Mthe self-adaptive labeling process comprises the following steps: using a sliding Tukey window pair x i,j Performing difference to obtain a difference sequence y i,j Mark y i,j The point smaller than 0 is the falling-stage zero-crossing point, the point larger than 0 is the rising-stage zero-crossing point, the maximum value between the rising-stage zero-crossing point and the next falling-stage zero-crossing point adjacent to the rising-stage zero-crossing point is defined as a peak, and the minimum value between the falling-stage zero-crossing point and the immediately next rising-stage zero-crossing point is defined asA trough; the collection of peaks and valleys is defined as an extremum;
taking the extreme point peak of the marked extreme point peak
Two sides of extreme pointtCorresponding +.>
And
and average the absolute voltage difference peak The method comprises the steps of carrying out a first treatment on the surface of the The formula is as follows:
calculation ofmSecond length x i,j Sharp corresponding to all extreme points in the band peak And find the median, denoted as x i,j To thereby obtain the sharpness of each multi-dimensional signal segment divided by the multi-dimensional computer signal;
the process of taking the nonlinearity is specifically as follows:
for the multidimensional electroencephalogram signal { after pretreatmentx i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j Calculate x i,j Is used for the instantaneous frequency and the zero crossing frequency of the transformer;
by x i,j Extraction of the instantaneous frequency and zero crossing frequency of the signal segment x in relation to the multidimensional signal segment i,j Standard deviation of normalized instantaneous frequency as non-linearityDoN i,j Obtaining the nonlinearity of each multidimensional signal segment divided by the multidimensional computer signal;
and using the key waveform characteristics of the electroencephalogram signals as input machine learning classification models to dynamically identify the epileptic electroencephalogram signals.
2. The dynamic identification method of epileptic electroencephalogram signals based on sharpness and nonlinearity according to claim 1, wherein the preprocessing is performed one by one on the data of each channel in the acquired electroencephalogram signals, specifically:
filtering processing is carried out on data to be processed, and the filtering processing comprises the following steps: removing power frequency interference by using a notch filter; removing motion artifacts using a high pass filter;
performing dimension reduction processing on the filtered data, obtaining a principal component coefficient matrix, a principal component score matrix and a characteristic value vector by adopting a principal component analysis method, accumulating the characteristic values, dividing the accumulated characteristic values by the sum, retaining the characteristic values with accumulated contributions not less than a set value, and recording the number of the characteristic values asKFront of principal component coefficient matrix and principal component score matrixKColumn multiplication to obtain preprocessed multidimensional EEG signalsx i (t), i = 1, 2, …, N, NIs the dimension of the brain electrical signal.
3. The dynamic identification method of epileptic brain electrical signals based on sharpness and nonlinearity according to claim 1, wherein the preprocessed multidimensional computer signals take corresponding complex envelopes, sharpness and nonlinearity in each dimension as key waveform characteristics of brain electrical signals, specifically:
for EEG signal sequence {x i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j ；
For the complex envelope taken in the ith dimension isE i =[E 1i , E 2i , E 3i ,…, E Mi ]The sharpness isS i =[S 1i , S 2i , S 3i ,…, S Mi ]The nonlinearity isD i =[D 1i , D 2i , D 3i ,…, D Mi ]；E Mi , S Mi , D Mi Respectively obtaining complex envelope, sharpness and nonlinearity corresponding to the Mth multidimensional signal segment in the ith dimension;
and ordering all N multiplied by 3 groups of M-dimensional extraction features by using a minimum redundancy-maximum correlation algorithm, wherein the M-dimensional extraction features are used as key waveform features of the electroencephalogram signals.
4. The epileptic electroencephalogram dynamic identification system based on sharpness and nonlinearity is characterized by comprising an electroencephalogram acquisition module, an epileptic nonlinear waveform characteristic calculation module and an epileptic electroencephalogram dynamic identification module;
the electroencephalogram signal acquisition module is used for acquiring multichannel electroencephalogram signals acquired by electroencephalogram equipment, preprocessing data of each channel in the acquired electroencephalogram signals one by one, taking each channel as one dimension, obtaining preprocessed multidimensional electroencephalogram signals, and sending the preprocessed multidimensional electroencephalogram signals to the epileptic nonlinear waveform characteristic calculation module;
the epileptic nonlinear waveform characteristic calculation module is used for taking corresponding complex envelope, sharpness and nonlinearity of the preprocessed multidimensional computer signal in each dimension as key waveform characteristics of the electroencephalogram signals and sending the key waveform characteristics to the epileptic electroencephalogram signal dynamic identification module;
the epileptic electroencephalogram dynamic identification module utilizes key waveform characteristics of the electroencephalogram as an input machine learning classification model to dynamically identify epileptic electroencephalogram;
the epileptic nonlinear waveform characteristic calculation module specifically comprises the following steps of:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; signal processing in each dimensionmSecond non-overlapping window segmentation to obtainMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i = 1, 2, …, N, j=1, 2, …, M，MIs the total number of signal segments;
for each multi-dimensional signal segment x i,j , i = 1, 2, …, N, j=1, 2, …, MPerforming Hilbert transform to obtain analysis signal
The envelope signal of the resolved signal is +.>
；
Calculating envelope signal of each section of analysis signal
Is denoted as x i,j Thereby obtaining a complex envelope of each multi-dimensional signal segment divided by the multi-dimensional computer signal;
the epileptic nonlinear waveform characteristic calculation module is characterized in that the sharpness taking process comprises the following steps:
{ for the multidimensional electroencephalogram signals after pretreatmentx i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j ，i = 1, 2, …, N, j=1, 2, …, M，NIs the dimension of the signal,MIs the total number of signal segments; one by one adaptive labeling multidimensional signal segment x i,j Is used for the extreme value of (a),i = 1, 2, …, N, j=1, 2, …, Mthe self-adaptive labeling process comprises the following steps: using a sliding Tukey window pair x i,j Performing difference to obtain a difference sequence y i,j Mark y i,j A point smaller than 0 is a falling-stage zero-crossing point, a point larger than 0 is a rising-stage zero-crossing point, a maximum value between the rising-stage zero-crossing point and a next falling-stage zero-crossing point adjacent to the rising-stage zero-crossing point is defined as a wave crest, and a minimum value between the falling-stage zero-crossing point and an immediately next rising-stage zero-crossing point is defined as a wave trough; the collection of peaks and valleys is defined as an extremum;
taking the extreme point peak of the marked extreme point peak
Two sides of extreme pointtCorresponding +.>
And
and average the absolute voltage difference peak The method comprises the steps of carrying out a first treatment on the surface of the The formula is as follows:
calculation ofmSecond length x i,j Sharp corresponding to all extreme points in the band peak And find the median, denoted as x i,j To thereby obtain the sharpness of each multi-dimensional signal segment divided by the multi-dimensional computer signal;
the epileptic nonlinear waveform characteristic calculation module specifically comprises the following steps of:
for the multidimensional electroencephalogram signal { after pretreatmentx i (t), i = 1, 2, …, N}，NIs the dimension of the brain electrical signal; the signals in each dimension being obtained by non-overlapping window segmentationMPersonal (S)mSecond-long multidimensional signal segment x i,j Calculate x i,j Is used for the instantaneous frequency and the zero crossing frequency of the transformer;
by x i,j Extraction of the instantaneous frequency and zero crossing frequency of the signal segment x in relation to the multidimensional signal segment i,j Standard deviation of normalized instantaneous frequency as non-linearityDoN i,j Thus, the nonlinearity of each multidimensional signal segment divided by the multidimensional computer signal is obtained.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310484796.1A CN116196018B (en) | 2023-05-04 | 2023-05-04 | Dynamic identification method and system for epileptic brain electrical signals based on sharpness and nonlinearity |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310484796.1A CN116196018B (en) | 2023-05-04 | 2023-05-04 | Dynamic identification method and system for epileptic brain electrical signals based on sharpness and nonlinearity |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116196018A CN116196018A (en) | 2023-06-02 |
| CN116196018B true CN116196018B (en) | 2023-07-07 |
Family
ID=86513302
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310484796.1A Active CN116196018B (en) | 2023-05-04 | 2023-05-04 | Dynamic identification method and system for epileptic brain electrical signals based on sharpness and nonlinearity |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116196018B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110432898A (en) * | 2019-07-04 | 2019-11-12 | 北京大学 | A kind of epileptic attack eeg signal classification system based on Nonlinear Dynamical Characteristics |
| CN111150393A (en) * | 2020-02-19 | 2020-05-15 | 杭州电子科技大学 | Electroencephalogram epilepsy spike discharge joint detection method based on LSTM multichannel |
| CN114469138A (en) * | 2022-01-13 | 2022-05-13 | 博睿康医疗科技(上海)有限公司 | Method, system and medium for detecting electroencephalogram outbreak suppression mode based on time-frequency domain |
| CN115530845A (en) * | 2022-10-17 | 2022-12-30 | 常州瑞神安医疗器械有限公司 | Method for detecting abnormal discharge in epilepsia electroencephalogram signal |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US7630757B2 (en) * | 1997-01-06 | 2009-12-08 | Flint Hills Scientific Llc | System for the prediction, rapid detection, warning, prevention, or control of changes in activity states in the brain of a subject |
| WO2021034351A1 (en) * | 2019-08-22 | 2021-02-25 | Hecox Kurt E | Systems and methods for seizure detection based on changes in electroencephalogram (eeg) non-linearities |
| US20210106247A1 (en) * | 2019-10-15 | 2021-04-15 | Carnegie Mellon University | Methods and Apparatus for Detection and Imaging of Epileptogenicity from Scalp High-Frequency Oscillations |
| DE202022103434U1 (en) * | 2022-06-20 | 2022-06-24 | Khyati Chopra | An encephalogram (EEG)-based device used to detect epileptic seizures |
-
2023
- 2023-05-04 CN CN202310484796.1A patent/CN116196018B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110432898A (en) * | 2019-07-04 | 2019-11-12 | 北京大学 | A kind of epileptic attack eeg signal classification system based on Nonlinear Dynamical Characteristics |
| CN111150393A (en) * | 2020-02-19 | 2020-05-15 | 杭州电子科技大学 | Electroencephalogram epilepsy spike discharge joint detection method based on LSTM multichannel |
| CN114469138A (en) * | 2022-01-13 | 2022-05-13 | 博睿康医疗科技(上海)有限公司 | Method, system and medium for detecting electroencephalogram outbreak suppression mode based on time-frequency domain |
| CN115530845A (en) * | 2022-10-17 | 2022-12-30 | 常州瑞神安医疗器械有限公司 | Method for detecting abnormal discharge in epilepsia electroencephalogram signal |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116196018A (en) | 2023-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US20210000426A1 (en) | Classification system of epileptic eeg signals based on non-linear dynamics features | |
| Yuan et al. | A novel channel-aware attention framework for multi-channel eeg seizure detection via multi-view deep learning | |
| Jiang et al. | An automatic analysis method for detecting and eliminating ECG artifacts in EEG | |
| Hamad et al. | Feature extraction of epilepsy EEG using discrete wavelet transform | |
| CN110811609B (en) | Epileptic spike intelligent detection device based on self-adaptive template matching and machine learning algorithm fusion | |
| Wilson et al. | Seizure detection: evaluation of the Reveal algorithm | |
| Subasi | Application of adaptive neuro-fuzzy inference system for epileptic seizure detection using wavelet feature extraction | |
| Tripathy et al. | Automated detection of congestive heart failure from electrocardiogram signal using Stockwell transform and hybrid classification scheme | |
| Li et al. | FFT-based deep feature learning method for EEG classification | |
| CH716863B1 (en) | Method for building a depression detection classification model based on channel selection of multichannel electroencephalography. | |
| Morabito et al. | Enhanced compressibility of EEG signal in Alzheimer's disease patients | |
| Lai et al. | Automated detection of high frequency oscillations in intracranial EEG using the combination of short-time energy and convolutional neural networks | |
| Jacob et al. | Diagnosis of encephalopathy based on energies of EEG subbands using discrete wavelet transform and support vector machine | |
| Joshi et al. | Computer aided detection of abnormal EMG signals based on tunable-Q wavelet transform | |
| Goshvarpour et al. | Affective visual stimuli: Characterization of the picture sequences impacts by means of nonlinear approaches | |
| Yadav et al. | Morphology-based automatic seizure detector for intracerebral EEG recordings | |
| CN106485208A (en) | The automatic removal method of eye electrical interference in single channel EEG signals | |
| Geng et al. | Epileptic seizure detection based on improved wavelet neural networks in long-term intracranial EEG | |
| CN112861625A (en) | Method for determining stacking denoising autoencoder model | |
| Ranjan et al. | Cardiac artifact noise removal from sleep EEG signals using hybrid denoising model | |
| Orosco et al. | Epileptic seizures detection based on empirical mode decomposition of EEG signals | |
| Jose et al. | Robust classification of intramuscular EMG signals to aid the diagnosis of neuromuscular disorders | |
| CN114532993A (en) | Automatic detection method for electroencephalogram high-frequency oscillation signals of epileptic | |
| CN111887811B (en) | Brain abnormal discharge detection method and system based on electroencephalogram signal characteristics | |
| Fathima et al. | Wavelet based features for classification of normal, ictal and interictal EEG signals |
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 |