CN109497997A - Based on majority according to the seizure detection and early warning system of acquisition - Google Patents
Based on majority according to the seizure detection and early warning system of acquisition Download PDFInfo
- Publication number
- CN109497997A CN109497997A CN201811504356.3A CN201811504356A CN109497997A CN 109497997 A CN109497997 A CN 109497997A CN 201811504356 A CN201811504356 A CN 201811504356A CN 109497997 A CN109497997 A CN 109497997A
- Authority
- CN
- China
- Prior art keywords
- signal
- acquisition
- multichannel
- parameter
- seizure detection
- 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
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]
-
- 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
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/68—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient
- A61B5/6801—Arrangements of detecting, measuring or recording means, e.g. sensors, in relation to patient specially adapted to be attached to or worn on the body surface
- A61B5/6802—Sensor mounted on worn items
- A61B5/681—Wristwatch-type devices
-
- 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
- A61B5/7253—Details of waveform analysis characterised by using transforms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/74—Details of notification to user or communication with user or patient ; user input means
- A61B5/746—Alarms related to a physiological condition, e.g. details of setting alarm thresholds or avoiding false alarms
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2562/00—Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
- A61B2562/02—Details of sensors specially adapted for in-vivo measurements
- A61B2562/0219—Inertial sensors, e.g. accelerometers, gyroscopes, tilt switches
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B2562/00—Details of sensors; Constructional details of sensor housings or probes; Accessories for sensors
- A61B2562/06—Arrangements of multiple sensors of different types
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Veterinary Medicine (AREA)
- Public Health (AREA)
- General Health & Medical Sciences (AREA)
- Physics & Mathematics (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Surgery (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Signal Processing (AREA)
- Psychiatry (AREA)
- Physiology (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Power Engineering (AREA)
- Psychology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The present invention provides a kind of seizure detection and early warning system based on majority according to acquisition, detection device includes memory and processor, processor executes the computer program in memory and obtains the multiple groups physiological signal after disease stage divides to perform the steps of, multiple groups sample of signal is formed, every group of physiological signal includes multichannel brain electric signal, multichannel skin electrical signal and acceleration signal；The multiple groups sample of signal of acquisition is pre-processed one by one；Extract the characteristic parameter of pretreated every group of sample of signal；Using multiple characteristic parameters of extraction as feature vector, using the corresponding disease stage of every group of characteristic parameter as desired output, multiple decision trees in random forest grader are trained, form Random Forest model；Physiological signal to be analyzed is obtained, obtains after preprocessed and feature extraction to characteristic parameter and is input in Random Forest model, disease stage locating for physiological signal to be analyzed is obtained after Random Forest model is predicted.
Description
Technical field
The present invention relates to computer fields, and it is in particular to a kind of based on majority according to the seizure detection of acquisition and
Early warning system.
Background technique
Epilepsy is to be discharged by cerebral neuron paroxysmal abnormality and brain is caused one kind of of short duration dysfunction occur slowly
Property disease, is commonly called as epilepsy and sheep is insane crazy.Show currently, being investigated according to the World Health Organization (WHO): global about 500,000,000 people suffer from
Epileptic condition.In China, epilepsy has become the second largest common disease that neurology department is only second to headache.Because breaking out position and breaking-out
Inducement is varied, and also complicated multiplicity, epileptic may be whenever and wherever possible because certain be uncertain for the clinical manifestation of epileptic attack
Inducement morbidity, if corresponding treatment means can not be taken in time, patient is likely to traffic accident, burn, gas poisoning etc. occur
It is unexpected, make the life of patient by strong influence, also brings body and psychological injury to oneself and household.
Currently, brain wave (EEG) is still the main tool clinically studied epileptic attack feature, diagnose epileptic condition,
At this stage, the EEG signals of analysis patient are usually removed by human eye by veteran doctor, it is artificial to find epileptic attack point.
But the EEG signals data of patient are huge, while the interference vulnerable to subjective factor, and Artificial Diagnosis False Rate is high.Therefore, one is needed
The equipment that kind can detect automatically epileptic attack according to EEG signals and carry out early warning, to mitigate the workload of medical worker simultaneously
Avoid interference caused by subjective factors, and family numbers of patients allowed to obtain early warning in advance, be ready, reduce patient episode when pain, this
It is of great significance in medical application.At present to there are many analysis methods of EEG signals, but mostly single physiological signal into
Row research, since epileptic attack region is uncertain, the different physiological signal feature performance of brain electricity epileptic attack is also different, therefore with
Analysis method recognition effect based on single physiological signal is not satisfactory.Further, existing analysis method is only single
Analyzed according to the characteristic parameter in time domain or frequency domain, be difficult complete and accurate extracts the hair contained in eeg data
Make information, recognition effect is not satisfactory.
In addition, in current epileptic attack detection and study of warning, mostly using the classification such as SVM, artificial neural network
Device carries out the training identification of characteristic parameter, but SVM classifier is difficult to handle large-scale training sample, will expend a large amount of machine
Memory and operation time, neural network classification algorithm pace of learning is slow, is easy to appear over-fitting.
Summary of the invention
The present invention is in order to overcome asking for existing epileptic attack detection and early warning system status epilepticus recognition effect difference
Topic, provides a kind of recognition accuracy high seizure detection and early warning system based on majority according to acquisition.
To achieve the goals above, the present invention provides a kind of seizure detection based on majority according to acquisition, special
Sign is, including memory and processor, and computer program is stored in memory, and processor executes computer program to realize
Following steps:
The multiple groups physiological signal after disease stage divides is obtained, multiple groups sample of signal is formed, every group of physiological signal wraps
Include multichannel brain electric signal, multichannel skin electrical signal and acceleration signal；
The multiple groups sample of signal of acquisition is pre-processed one by one；
Extract the characteristic parameter of pretreated each sample of signal；
It is that expectation is defeated with the corresponding disease stage of every group of characteristic parameter using multiple characteristic parameters of extraction as feature vector
Out, multiple decision trees in training random forest grader, form Random Forest model；
Physiological signal to be analyzed is obtained, the feature ginseng of physiological signal to be analyzed is obtained after preprocessed and feature extraction
It counts and is input in Random Forest model, hair locating for physiological signal to be analyzed is obtained after Random Forest model is predicted
The sick stage.
An embodiment according to the present invention after the pre-treatment respectively believes sample of signal or physiology to be analyzed from time domain
Number multiple segments with certain time length are divided into, feature extraction is carried out to each segment respectively.
An embodiment according to the present invention extracts the temporal signatures ginseng of multichannel brain electric signal segment in feature extraction
Several and frequency domain character parameter, the time domain charactreristic parameter and frequency domain character parameter and acceleration of multichannel skin electrical signal segment are believed
The displacement acceleration parameter and angular acceleration parameter of number segment.
An embodiment according to the present invention, in the frequency domain character parameter or multichannel skin for extracting multichannel brain electric signal segment
When the frequency domain character parameter of skin electric signal segment, each fragment segmentation at 2 seconds and there is into the small fragment of overlapping in 1 second, is extracted each
Frequency domain character parameter on small fragment.
An embodiment according to the present invention, time domain charactreristic parameter include the related coefficient and its spy in time domain between each channel
Value indicative, frequency domain character parameter include the related coefficient and its characteristic value in frequency domain between each channel.
An embodiment according to the present invention carries out pretreated step to each sample of signal or physiological signal to be analyzed
The removal of eye electricity artefact including multichannel brain electric signal, specific steps are as follows:
It is standardized original multi-channel EEG signals S to obtain signal SC；
Then " db6 " wavelet function is used to carry out seven layers of wavelet transformation, and the wavelet systems that will be obtained after decomposition to signal SC
Number is together in series, and obtains a wavelet coefficient vector matrix X；
The transposition for seeking vector matrix X obtains transposed matrix Y；
Canonical correlation analysis is carried out to vector matrix X and transposed matrix Y, calculates base vector matrix WxAnd Wy, acquire typical case
Canonical variable after constituent analysis identifies eye electricity artefact ingredient using related coefficient, will using canonical correlation analysis inverse transformation
Each representative vectors after removing eye electricity artefact carry out projective transformation, then carry out the inverse transformation of wavelet transformation, and it is pseudo- to obtain removal eye electricity
Physiological signal after mark.
An embodiment according to the present invention carries out pretreated step to each sample of signal or physiological signal to be analyzed
Including removing multichannel brain electric signal respectively, in multichannel skin electrical signal and acceleration signal frequency lower than 0.5 hertz and
Frequency is higher than 50 hertz of signal.
An embodiment according to the present invention, when obtaining physiological signal, eeg signal acquisition device obtains multichannel brain electric
Signal, skin electric inductor and acceleration induction device in Intelligent bracelet obtain multichannel skin electrical signal and acceleration letter respectively
Number.
An embodiment according to the present invention, the multichannel brain electric signal that eeg signal acquisition device obtains are exported to memory
While save to local SD card.
It is corresponding, the present invention also provides it is a kind of based on majority according to acquisition epileptic attack early warning system it include above-mentioned base
In majority according to the seizure detection and output equipment of acquisition.Output equipment is received based on majority according to the epileptic attack of acquisition
The two is simultaneously sent to family numbers of patients or the intelligent terminal of doctor by detection device testing result obtained and early warning result.
To sum up, it is provided by the invention predict according to the seizure detection and early warning system of acquisition based on majority it is insane
Epilepsy combines human body acceleration and multichannel skin electrical signal when breaking out on the basis of EEG signals, acquires epilepsy hair in all directions
Feature when making identifies multimode physiological signal convergence analysis, improves generalization ability, so that Detection accuracy improves, early warning effect
Fruit is more preferable, and multimode signal acquisition can be overcome the problems, such as well because the identification caused by epileptic attack region is uncertain is inaccurate.
And in bio-signal acquisition, EEG signals and skin electrical signal are multi channel signals, and Multi-channel signal analysis detection can be more
Add the complete feature for extracting epileptic attack early period, stage of attack and later period of breaking out comprehensively, further increases recognition effect.
For above and other objects of the present invention, feature and advantage can be clearer and more comprehensible, preferred embodiment is cited below particularly,
And cooperate attached drawing, it is described in detail below.
Detailed description of the invention
Fig. 1 show the principle frame based on majority according to the seizure detection of acquisition of one embodiment of the invention offer
Figure.
Fig. 2 show one embodiment of the invention offer based on majority according to processor in the seizure detection of acquisition
Execute step flow chart achieved by computer program.
Fig. 3 show the flow chart of training Random Forest model in Fig. 2.
Fig. 4 show the principle frame based on majority according to the epileptic attack detection system of acquisition of one embodiment of the invention offer
Figure.
Specific embodiment
EEG signals are the main foundations clinically studied epileptic attack feature, diagnose epileptic condition, however due to epilepsy
Breaking out, region is unstable, and according only to EEG signals, this single physiological signal is difficult accurately to detect epileptic attack.
Therefore in view of this, the present embodiment provides a kind of based on majority according to the high seizure detection of the identification accuracy of acquisition and pre-
Alert system.
As depicted in figs. 1 and 2, provided in this embodiment to include according to the seizure detection 100 of acquisition based on majority
Memory 10 and processor 20.Be stored with computer program in memory 10, processor 20 execute computer program with realize with
Lower step: obtaining multiple physiological signals after disease stage divides, and is formed multiple groups sample of signal (step S10), every group of physiology
Signal includes multichannel brain electric signal, multichannel skin electrical signal and acceleration signal.To the multiple groups sample of signal of acquisition
It is pre-processed one by one (step S20).Extract the characteristic parameter (step S40) of pretreated every group of sample of signal.To extract
Multiple characteristic parameters as feature vector, be to be desired for exporting with disease stage corresponding to every group of characteristic parameter, training with
Multiple decision trees in machine forest classified device are formed Random Forest model (step S50).Physiological signal to be analyzed is obtained, is passed through
The characteristic parameter of physiological signal to be analyzed is obtained after pretreatment and feature extraction and is input in Random Forest model, is passed through
Disease stage (step S60) locating for physiological signal to be analyzed is obtained after Random Forest model prediction.Below in conjunction with Fig. 1 and figure
4 are discussed in detail the concrete operating principle provided in this embodiment based on majority according to the seizure detection 100 of acquisition.
It is provided in this embodiment based on majority according to the seizure detection 100 of acquisition obtain first it is multiple through fall ill rank
Physiological signal after section divides forms multiple groups sample of signal, and every group of physiological signal includes multichannel brain electric signal, multichannel skin
Skin electric signal and acceleration signal.However, the present invention is not limited in any way this.In other embodiments, physiological signal can
Other signals are further added by the basis of comprising multichannel brain electric signal.For multichannel brain electric signal, adopted using brain electricity
It leads EEG signals and is put through amplifying circuit in the 19 of collection equipment 30 (including electroencephalograph and Polysomnography etc.) acquisition patient
It is exported in the form of digital signal after big and analog-to-digital conversion.And multichannel skin electrical signal and acceleration signal then pass through patient
Skin electric inductor 41 and acceleration induction device 42 in the Intelligent bracelet 40 of wearing obtain.For convenience of to patient physiological signals
Real-time monitoring, in this present embodiment, brain wave acquisition equipment 30 and Intelligent bracelet 40 for signal acquisition are respectively by wireless
The mode of network is connected to the server including memory 10 and processor 20.Specifically, brain wave acquisition equipment 30 and intelligence
Bracelet 40 is sent on the intelligent terminal 60 of user by the physiological signal that bluetooth will test, and intelligent terminal 60 is by the detection
Signal passes through on wireless network or 4G network transmission to server.However, the present invention is not limited in any way this.In other implementations
In example, the signal that brain wave acquisition equipment and Intelligent bracelet can directly will test is exported to server.Since EEG signals are insane
Epilepsy detects most important foundation, therefore brain wave acquisition equipment 30 is exported by multichannel brain electric signal to intelligent end in this present embodiment
EEG signals are stored to local SD card 50 while holding 60, multichannel brain electric signal is backed up.
Since EEG signals are the basic basis of epileptic attack, therefore in this present embodiment, in the brain telecommunications for obtaining multichannel
Healthcare givers is according to each multichannel brain telecommunications that will acquire before and after onset with the feature of multichannel brain electric signal when morbidity after number
Number carry out divided stages, can specifically be divided into interictal, breaking-out early period, stage of attack and breaking-out later period four-stage, and with
Multichannel brain electric signal after divided stages is to obtain multichannel skin electrical signal and acceleration signal according to corresponding in the time domain
The locating stage.However, the present invention is not limited in any way the number of stages that disease stage divides.In other embodiments, morbidity
The division in stage can be refined more.
After obtaining multiple groups sample of signal, step S20 is executed, the multiple groups sample of signal of acquisition is pre-processed, every group
Multichannel brain electric signal, multichannel skin electric signal and acceleration signal are all contained in sample of signal.In this present embodiment, more
The pretreatment of channel EEG signals includes the removal of eye electricity artefact and high fdrequency component, and multichannel skin electric signal and acceleration signal
Pretreatment then be high fdrequency component removal.It filters off specifically, bandpass filter can be used except except frequency is at 50 hertz or more
With 0.5 hertz of signal below.And in multichannel brain electric signal, the removal of eye electricity artefact then as follows: first will be former
The multichannel brain electric signal S of beginning is standardized to obtain signal SC.Then " db6 " wavelet function is used to carry out signal SC
Seven layers of wavelet transformation, and the wavelet coefficient obtained after decomposition is together in series, obtain a wavelet coefficient vector matrix X.Seek square
The transposition of battle array X, obtains transposed matrix Y.Canonical correlation analysis is carried out to vector matrix X and transposed matrix Y, calculates base vector square
Battle array WxAnd Wy, the canonical variable after typical composition is analyzed is acquired, eye electricity artefact ingredient is identified using related coefficient, utilizes typical phase
Each representative vectors after removing eye electricity artefact are carried out projective transformation by the analysis inverse transformation of closing property, then carry out the inversion of wavelet transformation
It changes, the EEG signals after obtaining removal eye electricity artefact.
In this present embodiment, after the pre-treatment, step S30 is executed, by pretreated multichannel brain telecommunications from time domain
Number, multichannel skin electrical signal and acceleration signal be divided into multiple segments with certain time length.Step is executed later
S40 carries out feature extraction to each segment respectively.In this present embodiment, the duration of each segment is 30 seconds.However, this hair
The bright length to segment is not limited in any way.Specifically, extracting multichannel brain electric signal segment respectively in feature extraction
Time domain charactreristic parameter and frequency domain character parameter, the time domain charactreristic parameter of multichannel skin electrical signal segment and frequency domain character parameter with
And the displacement acceleration parameter and angular acceleration parameter of acceleration signal segment.
It will be described in detail the time domain charactreristic parameter of multichannel brain electric signal segment and the extraction step of frequency domain character parameter below
Suddenly, the time domain charactreristic parameter of multichannel skin electrical signal segment and frequency domain character parameter are then adopted and are extracted in a like fashion.
Multichannel brain electric signal time domain charactreristic parameter includes the related coefficient of each interchannel and the feature of correlation matrix
Value, extracting method are as follows: down-sampled processing is carried out to the time series of each multichannel brain electric signal segment, if initial data is Xi
(n), length N, down-sampled multiple be M, it is down-sampled after signal be Yi(n),
I=1,2 ..., 19, yiIt (n) is n-th of down-sampled information number in i-th of channel,For that will be adopted in i-th of channel
With the average value of signal, down-sampled formula is as follows:
Then down-sampled signal Y (n) is normalized, normalization formula is as follows:
To acquire the related coefficient of each interchannel, the characteristic value of correlation matrix is obtained by related coefficient, it is related
Coefficient formula is as follows:
When carrying out the feature extraction of multichannel brain electric signal frequency domain, since each rhythm and pace of moving things wave is no longer than 2 seconds, in order to mention
The feature taken more comprehensively includes the information of seizure of disease, therefore each fragment segmentation at 2 seconds and is had 1 second small pieces being overlapped
Section, multichannel brain electric signal frequency domain characteristic parameter includes the related coefficient of each interchannel and the characteristic value of correlation matrix.It mentions
The step of taking includes: to carry out Fast Fourier Transform (FFT) (FFT) to each 2 seconds small fragments, if small pieces segment signal is xi(n), i=
1,2 ..., 19, Fast Fourier Transform (FFT) formula is as follows:
Wherein N is small pieces segment signal xiLength, k=0,1,2 ..., N-1, WN=e-j2π/N。
Result is normalized, normalization formula is as follows:
Wherein i=1,2 ..., 19, k=0,1 ..., N-1, min (Xi) be Fourier transformation after signal XiMinimum value,
max(Xi) be Fourier transformation after signal XiMaximum value.
δ wave y is extracted using wavelet transformi,1, θ wave yi,2, α wave yi,3, β wave yi,4, seek each rhythm and pace of moving things wave of each interchannel
Related coefficient, formula of correlation coefficient are as follows:
Wherein q=1,2 ..., 171；
Finally, obtaining correlation matrix Q=[rq,1,rq,2,rq,3,rq,4]；
The characteristic value for extracting correlation matrix, seeks the ENERGY E (r of each rhythm and pace of moving things wave related coefficientq,m) and energy ratio ratio
(rq,m) and count their maximum value, minimum value, average value and variance as statistical parameter.E(rq,m) and ratio (rq,m) public
Formula is as follows:
ENERGY E (the r of each rhythm and pace of moving things wave related coefficientq,m) calculation formula are as follows:
Total ENERGY EsCalculation formula are as follows:
Energy ratio ratio (rq,m) calculation formula are as follows:
ratio(rq,m)=E (rq,m)/EsFormula 9
Wherein m=1,2,3,4.
After features described above is extracted, the time domain charactreristic parameter and frequency domain character ginseng of each multichannel brain electric signal segment are obtained
The time domain charactreristic parameter and frequency domain character parameter of several, each multichannel skin electrical signal segment and each acceleration signal segment
Displacement acceleration parameter and angular acceleration parameter.
Step S50 is executed later, using multiple characteristic parameters as feature vector, with the corresponding disease stage of every group of characteristic parameter
For desired output, multiple decision trees in random forest grader are trained, form Random Forest model.In this present embodiment, will
The multiple sample of signal obtained in step S10 are divided into training set and test set, utilize multiple sample of signal training in training set
Multiple decision trees in random forest grader form Random Forest model, utilize multiple sample of signal in test set later
The Random Forest model that test amendment is formed.
As shown in figure 3, specific training process is as follows:
Step S501, sampling that is random and having playback, forming quantity training sample identical with training set in training set
Collection.Step S502 obtains the characteristic parameter of each training set sample in training sample set, forms set of eigenvectors.Step S503,
Random and there is playback to sample, the forming quantity set of eigenvectors to be selected identical with set of eigenvectors in set of eigenvectors.Step
S504 randomly selects part feature to be selected in feature vector to be selected concentration, optimal spy to be selected is selected in the feature to be selected of part
Sign, and with the optimal feature to be selected from root node divided.Step S505 judges whether that leaf section can be become
Point.If reaching termination condition on present node, it is leaf node that present node, which is arranged, and the prediction output of the leaf node is
That most one kind of quantity in present node sample set；If present node does not reach termination condition, continue to repeat to walk
Rapid S504 and step S505, until current decision tree stops growing.Step S501 is repeated to step S505 to train next to determine
Plan tree completes the training of Random Forest model after the quantity of the decision tree of formation meets sets requirement.The termination condition
For the stage that can obtain morbidity on present node.
After the training and amendment for completing Random Forest model, step S60 is executed.Step S60 includes that acquisition is to be analyzed
Physiological signal pre-processes physiological signal to be analyzed using the algorithm in step S20, later using the algorithm in S30
The algorithm carried out in fragment segmentation and step S40 carries out characteristic parameter extraction, is included to obtain EEG signals to be analyzed
Multiple characteristic parameters, disease stage locating for each segment is obtained after analyzing, physiology letter to be analyzed is obtained after counting
Period of disease locating for number realizes the prediction of epileptics disease stage.
It is corresponding, as shown in figure 4, the present embodiment also provides a kind of epileptic attack detection system based on majority according to acquisition
System, which includes the seizure detection 100 and output equipment 200 provided in this embodiment based on majority according to acquisition.
Output equipment 200 is received based on majority according to the testing result obtained of seizure detection 100 and early warning result of acquisition
And the two is sent to family numbers of patients or the intelligent terminal 300 of doctor, family numbers of patients or the current disease of doctor patient are reminded in time
Reason state.In this present embodiment, the intelligent terminal is smart phone.However, the present invention is not limited in any way this.
To sum up, it is provided by the invention predict according to the seizure detection and early warning system of acquisition based on majority it is insane
Epilepsy combines human body acceleration and skin electrical signal when breaking out on the basis of EEG signals, acquires epileptic attack in all directions
Feature identifies multimode physiological signal convergence analysis, improves generalization ability, so that Detection accuracy improves, early warning effect is more preferable,
Multimode signal acquisition can be overcome the problems, such as well because the identification caused by epileptic attack region is uncertain is inaccurate.And in physiology
When signal detection, EEG signals and skin electrical signal are multi channel signals, and Multi-channel signal analysis detection can be more comprehensively complete
The whole feature for extracting epileptic attack early period, stage of attack and later period of breaking out, further increases recognition effect.
Although the present invention is disclosed above by preferred embodiment, however, it is not intended to limit the invention, this any known skill
Skill person can make some changes and embellishment without departing from the spirit and scope of the present invention, therefore protection scope of the present invention is worked as
Subject to claims range claimed.
Claims (10)
1. it is a kind of based on majority according to the seizure detection of acquisition, which is characterized in that it is described including memory and processor
Computer program is stored in memory, the processor executes the computer program to perform the steps of
The multiple groups physiological signal after disease stage divides is obtained, forms multiple sample of signal, every group of physiological signal includes more
Channel EEG signals, multichannel skin electrical signal and acceleration signal；
The multiple groups sample of signal of acquisition is pre-processed one by one；
Extract the characteristic parameter of pretreated every group of sample of signal；
Using multiple characteristic parameters of extraction as feature vector, using the corresponding disease stage of every group of characteristic parameter as desired output,
Multiple decision trees in training random forest grader, form Random Forest model；
Physiological signal to be analyzed is obtained, obtains the characteristic parameter of physiological signal to be analyzed simultaneously after preprocessed and feature extraction
It is input in Random Forest model, morbidity rank locating for physiological signal to be analyzed is obtained after Random Forest model is predicted
Section.
2. it is according to claim 1 based on majority according to the seizure detection of acquisition, which is characterized in that pre-processing
Sample of signal or physiological signal to be analyzed are divided into multiple segments with certain time length from time domain respectively afterwards, it is right respectively
Each segment carries out feature extraction.
3. it is according to claim 2 based on majority according to the seizure detection of acquisition, which is characterized in that mentioned in feature
When taking, the time domain charactreristic parameter and frequency domain character parameter of multichannel brain electric signal segment, multichannel skin electrical signal segment are extracted
Time domain charactreristic parameter and frequency domain character parameter and acceleration signal segment displacement acceleration parameter and angular acceleration ginseng
Number.
4. it is according to claim 3 based on majority according to the seizure detection of acquisition, which is characterized in that extract it is more
When the frequency domain character parameter of the frequency domain character parameter of channel EEG signals segment or multichannel skin electrical signal segment, by each
Section is divided into 2 seconds and has the small fragment of overlapping in 1 second, extracts the frequency domain character parameter on each small fragment.
5. it is according to claim 3 based on majority according to the seizure detection of acquisition, which is characterized in that temporal signatures
Parameter includes the related coefficient and its characteristic value in time domain between each channel, and frequency domain character parameter includes in frequency domain between each channel
Related coefficient and its characteristic value.
6. it is according to claim 1 based on majority according to the seizure detection of acquisition, which is characterized in that every group of letter
Number sample or physiological signal to be analyzed carry out the removal that pretreated step includes the eye electricity artefact of multichannel brain electric signal, have
Body step are as follows:
It is standardized original multi-channel EEG signals S to obtain signal SC；
Then " db6 " wavelet function is used to carry out seven layers of wavelet transformation, and the wavelet coefficient string that will be obtained after decomposition to signal SC
Connection gets up, and obtains a wavelet coefficient vector matrix X；
The transposition for seeking vector matrix X obtains transposed matrix Y；
Canonical correlation analysis is carried out to vector matrix X and transposed matrix Y, calculates base vector matrix WxAnd Wy, acquire typical composition
Canonical variable after analysis identifies eye electricity artefact ingredient using related coefficient, will be removed using canonical correlation analysis inverse transformation
Each representative vectors after the electric artefact of eye carry out projective transformation, then carry out the inverse transformation of wavelet transformation, after obtaining removal eye electricity artefact
Physiological signal.
7. it is according to claim 1 based on majority according to the seizure detection of acquisition, which is characterized in that every group of letter
It includes removing multichannel brain electric signal, multichannel skin respectively that number sample or physiological signal to be analyzed, which carry out pretreated step,
Frequency is lower than 0.5 hertz in electric signal and acceleration signal and frequency is higher than 50 hertz of signal.
8. it is according to claim 1 based on majority according to the seizure detection of acquisition, which is characterized in that obtain give birth to
When managing signal, eeg signal acquisition device obtains multichannel brain electric signal, skin electric inductor and acceleration in Intelligent bracelet
Inductor obtains multichannel skin electrical signal and acceleration signal respectively.
9. it is according to claim 8 based on majority according to the seizure detection of acquisition, which is characterized in that EEG signals
The multichannel brain electric signal that acquisition device obtains is saved while output to memory to local SD card.
10. it is a kind of based on majority according to the epileptic attack early warning system of acquisition characterized by comprising
It is according to any one of claims 1 to 9 based on majority according to the seizure detection of acquisition；
Output equipment, receive based on majority according to acquisition seizure detection testing result obtained and early warning result simultaneously
The two is sent to family numbers of patients or the intelligent terminal of doctor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811504356.3A CN109497997A (en) | 2018-12-10 | 2018-12-10 | Based on majority according to the seizure detection and early warning system of acquisition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811504356.3A CN109497997A (en) | 2018-12-10 | 2018-12-10 | Based on majority according to the seizure detection and early warning system of acquisition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109497997A true CN109497997A (en) | 2019-03-22 |
Family
ID=65752076
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811504356.3A Pending CN109497997A (en) | 2018-12-10 | 2018-12-10 | Based on majority according to the seizure detection and early warning system of acquisition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109497997A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109998530A (en) * | 2019-04-15 | 2019-07-12 | 杭州妞诺科技有限公司 | Portable brain pyroelectric monitor system based on VR glasses |
| CN110123342A (en) * | 2019-04-17 | 2019-08-16 | 西北大学 | A kind of network addiction detection method and system based on brain wave |
| CN110353672A (en) * | 2019-07-15 | 2019-10-22 | 西安邮电大学 | Eye artefact removal system and minimizing technology in a kind of EEG signals |
| CN111462887A (en) * | 2020-03-31 | 2020-07-28 | 首都医科大学宣武医院 | Wearable epileptic digital assistant system |
| CN112336354A (en) * | 2020-11-06 | 2021-02-09 | 山西三友和智慧信息技术股份有限公司 | Epilepsy monitoring method based on EEG signal |
| CN112438733A (en) * | 2020-11-06 | 2021-03-05 | 南京大学 | Portable neonatal convulsion electroencephalogram monitoring system |
| CN112842358A (en) * | 2019-11-26 | 2021-05-28 | 阿里健康信息技术有限公司 | Brain physiological data processing system, method, device and storage medium |
| CN113598792A (en) * | 2021-08-04 | 2021-11-05 | 杭州电子科技大学 | Epilepsy electroencephalogram classification method based on supervised feature fusion algorithm |
| CN114431829A (en) * | 2020-11-05 | 2022-05-06 | 四川大学华西医院 | Modeling method of epileptic seizure judgment model, and epileptic monitoring method and device |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102058413A (en) * | 2010-12-03 | 2011-05-18 | 上海交通大学 | Method for detecting EEG (electroencephalogram) alertness based on continuous wavelet transform |
| WO2011072684A1 (en) * | 2009-12-16 | 2011-06-23 | Ictalcare A/S | A system for the prediction of epileptic seizures |
| US8024032B1 (en) * | 2005-11-28 | 2011-09-20 | Flint Hills Scientific Llc | Method and system for the prediction, rapid detection, warning, prevention, or control of changes in the brain states of a subject using hurst parameter estimation |
| CA2812959A1 (en) * | 2010-10-01 | 2012-04-05 | Ivan Osorio | Detecting, quantifying, and/or classifying seizures using multimodal data |
| CN103258215A (en) * | 2013-05-10 | 2013-08-21 | 杭州电子科技大学 | Multi-lead correlation analysis electroencephalo-graph (EEG) feature extraction method |
| CN103476326A (en) * | 2011-01-28 | 2013-12-25 | 纽罗斯凯公司 | Dry sensor EEG/EMG and motion sensing system for seizure detection and monitoring |
| CN104173045A (en) * | 2014-08-15 | 2014-12-03 | 浙江大学医学院附属第二医院 | Epileptic seizure prewarning system |
| CN104720797A (en) * | 2015-03-20 | 2015-06-24 | 合肥工业大学 | Method for eliminating myoelectricity noise in electroencephalogram signal based on single channel |
| CN204889968U (en) * | 2015-08-05 | 2015-12-23 | 深圳市易特科信息技术有限公司 | A wearable equipment for epileptics is monitored in early days |
| CN105266818A (en) * | 2015-11-09 | 2016-01-27 | 四川大学华西医院 | Wearable monitoring instrument for monitoring partial psychomotor seizure of epileptic |
| CN106137185A (en) * | 2016-06-21 | 2016-11-23 | 华南理工大学 | A kind of epileptic chracter wave detecting method based on structure of transvers plate small echo |
| CN106913351A (en) * | 2015-12-28 | 2017-07-04 | 西南交通大学 | A kind of recognition methods of Mental Workload level |
| CN107260166A (en) * | 2017-05-26 | 2017-10-20 | 昆明理工大学 | A kind of electric artefact elimination method of practical online brain |
| CN107408144A (en) * | 2014-11-14 | 2017-11-28 | Zoll医疗公司 | Medical precursor event estimation |
| CN107693014A (en) * | 2017-11-22 | 2018-02-16 | 北京工业大学 | A kind of epileptic attack method for early warning based on bag of words and extreme learning machine |
| CN108836319A (en) * | 2018-03-08 | 2018-11-20 | 义乌市杰联电子科技有限公司 | A kind of nervous feedback system of fusion individuation encephalomere rule ratio and forehead myoelectricity energy |
| CN108962391A (en) * | 2018-05-02 | 2018-12-07 | 杭州电子科技大学 | Epileptics prediction technique early period based on wavelet packet character and random forest |
-
2018
- 2018-12-10 CN CN201811504356.3A patent/CN109497997A/en active Pending
Patent Citations (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8024032B1 (en) * | 2005-11-28 | 2011-09-20 | Flint Hills Scientific Llc | Method and system for the prediction, rapid detection, warning, prevention, or control of changes in the brain states of a subject using hurst parameter estimation |
| WO2011072684A1 (en) * | 2009-12-16 | 2011-06-23 | Ictalcare A/S | A system for the prediction of epileptic seizures |
| EP2512331A1 (en) * | 2009-12-16 | 2012-10-24 | IctalCare A/S | A system for the prediction of epileptic seizures |
| CA2812959A1 (en) * | 2010-10-01 | 2012-04-05 | Ivan Osorio | Detecting, quantifying, and/or classifying seizures using multimodal data |
| CN102058413A (en) * | 2010-12-03 | 2011-05-18 | 上海交通大学 | Method for detecting EEG (electroencephalogram) alertness based on continuous wavelet transform |
| CN103476326A (en) * | 2011-01-28 | 2013-12-25 | 纽罗斯凯公司 | Dry sensor EEG/EMG and motion sensing system for seizure detection and monitoring |
| CN103258215A (en) * | 2013-05-10 | 2013-08-21 | 杭州电子科技大学 | Multi-lead correlation analysis electroencephalo-graph (EEG) feature extraction method |
| CN104173045A (en) * | 2014-08-15 | 2014-12-03 | 浙江大学医学院附属第二医院 | Epileptic seizure prewarning system |
| CN107408144A (en) * | 2014-11-14 | 2017-11-28 | Zoll医疗公司 | Medical precursor event estimation |
| CN104720797A (en) * | 2015-03-20 | 2015-06-24 | 合肥工业大学 | Method for eliminating myoelectricity noise in electroencephalogram signal based on single channel |
| CN204889968U (en) * | 2015-08-05 | 2015-12-23 | 深圳市易特科信息技术有限公司 | A wearable equipment for epileptics is monitored in early days |
| CN105266818A (en) * | 2015-11-09 | 2016-01-27 | 四川大学华西医院 | Wearable monitoring instrument for monitoring partial psychomotor seizure of epileptic |
| CN106913351A (en) * | 2015-12-28 | 2017-07-04 | 西南交通大学 | A kind of recognition methods of Mental Workload level |
| CN106137185A (en) * | 2016-06-21 | 2016-11-23 | 华南理工大学 | A kind of epileptic chracter wave detecting method based on structure of transvers plate small echo |
| CN107260166A (en) * | 2017-05-26 | 2017-10-20 | 昆明理工大学 | A kind of electric artefact elimination method of practical online brain |
| CN107693014A (en) * | 2017-11-22 | 2018-02-16 | 北京工业大学 | A kind of epileptic attack method for early warning based on bag of words and extreme learning machine |
| CN108836319A (en) * | 2018-03-08 | 2018-11-20 | 义乌市杰联电子科技有限公司 | A kind of nervous feedback system of fusion individuation encephalomere rule ratio and forehead myoelectricity energy |
| CN108962391A (en) * | 2018-05-02 | 2018-12-07 | 杭州电子科技大学 | Epileptics prediction technique early period based on wavelet packet character and random forest |
Non-Patent Citations (1)
| Title |
|---|
| 李明爱: "眼电伪迹自动去除方法的研究与分析", 《仪器仪表学报》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109998530A (en) * | 2019-04-15 | 2019-07-12 | 杭州妞诺科技有限公司 | Portable brain pyroelectric monitor system based on VR glasses |
| CN110123342A (en) * | 2019-04-17 | 2019-08-16 | 西北大学 | A kind of network addiction detection method and system based on brain wave |
| CN110353672A (en) * | 2019-07-15 | 2019-10-22 | 西安邮电大学 | Eye artefact removal system and minimizing technology in a kind of EEG signals |
| CN110353672B (en) * | 2019-07-15 | 2022-02-22 | 西安邮电大学 | System and method for removing eye artifacts in electroencephalogram signals |
| CN112842358A (en) * | 2019-11-26 | 2021-05-28 | 阿里健康信息技术有限公司 | Brain physiological data processing system, method, device and storage medium |
| CN111462887A (en) * | 2020-03-31 | 2020-07-28 | 首都医科大学宣武医院 | Wearable epileptic digital assistant system |
| CN111462887B (en) * | 2020-03-31 | 2023-08-29 | 首都医科大学宣武医院 | Wearable epileptic digital assistant system |
| CN114431829A (en) * | 2020-11-05 | 2022-05-06 | 四川大学华西医院 | Modeling method of epileptic seizure judgment model, and epileptic monitoring method and device |
| CN112336354A (en) * | 2020-11-06 | 2021-02-09 | 山西三友和智慧信息技术股份有限公司 | Epilepsy monitoring method based on EEG signal |
| CN112438733A (en) * | 2020-11-06 | 2021-03-05 | 南京大学 | Portable neonatal convulsion electroencephalogram monitoring system |
| CN113598792A (en) * | 2021-08-04 | 2021-11-05 | 杭州电子科技大学 | Epilepsy electroencephalogram classification method based on supervised feature fusion algorithm |
| CN113598792B (en) * | 2021-08-04 | 2024-01-30 | 杭州电子科技大学 | Epileptic electroencephalogram classification method based on supervised feature fusion algorithm |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109497997A (en) | Based on majority according to the seizure detection and early warning system of acquisition | |
| CN110338786B (en) | Epileptic discharge identification and classification method, system, device and medium | |
| Tzallas et al. | Automatic seizure detection based on time‐frequency analysis and artificial neural networks | |
| CN113786204A (en) | Epilepsia intracranial electroencephalogram early warning method based on deep convolution attention network | |
| CN109602417A (en) | Sleep stage method and system based on random forest | |
| CN105796096B (en) | A kind of heart rate variance analyzing method, system and terminal | |
| CN103750844B (en) | A kind of based on the phase locked personal identification method of brain electricity | |
| Shedeed | A new method for person identification in a biometric security system based on brain EEG signal processing | |
| Kumari et al. | Seizure detection in EEG using time frequency analysis and SVM | |
| CN110811609A (en) | Intelligent epileptic spike detection method based on fusion of adaptive template matching and machine learning algorithm | |
| WO2015200750A1 (en) | Early detection of hemodynamic decompensation using taut-string transformation | |
| JP2002502270A (en) | System for predicting, early detecting, alerting, preventing or controlling changes in brain activity | |
| Wang et al. | A novel multi-scale dilated 3D CNN for epileptic seizure prediction | |
| CN108256629A (en) | The unsupervised feature learning method of EEG signal based on convolutional network and own coding | |
| Bedeeuzzaman et al. | Automatic seizure detection using higher order moments | |
| Obayya et al. | Automatic classification of sleep stages using EEG records based on Fuzzy c-means (FCM) algorithm | |
| Zhao et al. | An IoT-based wearable system using accelerometers and machine learning for fetal movement monitoring | |
| Matta et al. | Real-time continuous identification system using ECG signals | |
| CN114532993B (en) | Automatic detection method for electroencephalogram high-frequency oscillation signals of epileptic patients | |
| Sabor et al. | Robust arrhythmia classification based on QRS detection and a compact 1D-CNN for wearable ECG devices | |
| CN106175754A (en) | During sleep state is analyzed, waking state detects device | |
| Djamal et al. | Significant variables extraction of post-stroke EEG signal using wavelet and SOM kohonen | |
| Ma et al. | Early prediction of epileptic seizure based on the BNLSTM-CASA model | |
| CN113274037A (en) | Method, system and equipment for generating dynamic brain function network | |
| CN115081486A (en) | Epileptic focus positioning system and method for intracranial electroencephalogram network in early stage of epileptic seizure |
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 |