CN108937922A - A kind of diagnostic model method for building up, memory module and the processing equipment of ADHD - Google Patents
A kind of diagnostic model method for building up, memory module and the processing equipment of ADHD Download PDFInfo
- Publication number
- CN108937922A CN108937922A CN201810333017.7A CN201810333017A CN108937922A CN 108937922 A CN108937922 A CN 108937922A CN 201810333017 A CN201810333017 A CN 201810333017A CN 108937922 A CN108937922 A CN 108937922A
- Authority
- CN
- China
- Prior art keywords
- diagnostic model
- adhd
- parameter
- scale
- building
- 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
- 208000006096 Attention Deficit Disorder with Hyperactivity Diseases 0.000 title claims abstract description 50
- 208000036864 Attention deficit/hyperactivity disease Diseases 0.000 title claims abstract description 50
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000012545 processing Methods 0.000 title claims abstract description 9
- 239000013598 vector Substances 0.000 claims abstract description 41
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 28
- 230000006870 function Effects 0.000 claims abstract description 26
- 210000004556 brain Anatomy 0.000 claims abstract description 21
- 239000011159 matrix material Substances 0.000 claims abstract description 19
- 238000003909 pattern recognition Methods 0.000 claims abstract description 9
- 238000012549 training Methods 0.000 claims abstract description 9
- 238000010606 normalization Methods 0.000 claims abstract description 7
- 230000005611 electricity Effects 0.000 claims abstract description 6
- 238000005457 optimization Methods 0.000 claims description 7
- 239000000284 extract Substances 0.000 claims description 4
- 238000012706 support-vector machine Methods 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 8
- 238000000605 extraction Methods 0.000 description 8
- 238000003745 diagnosis Methods 0.000 description 7
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 7
- 201000010099 disease Diseases 0.000 description 6
- 230000033764 rhythmic process Effects 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000005516 engineering process Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 5
- 208000015802 attention deficit-hyperactivity disease Diseases 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 208000013403 hyperactivity Diseases 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000010183 spectrum analysis Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 208000018152 Cerebral disease Diseases 0.000 description 1
- 208000022306 Cerebral injury Diseases 0.000 description 1
- 241001269238 Data Species 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000003925 brain function Effects 0.000 description 1
- 238000004422 calculation algorithm Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000007812 deficiency Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 206010014599 encephalitis Diseases 0.000 description 1
- 206010015037 epilepsy Diseases 0.000 description 1
- 230000005802 health problem Effects 0.000 description 1
- 238000007689 inspection Methods 0.000 description 1
- 230000001575 pathological effect Effects 0.000 description 1
- 208000020016 psychiatric disease Diseases 0.000 description 1
- 230000003304 psychophysiological effect Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
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
- A61B5/72—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
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Psychiatry (AREA)
- Artificial Intelligence (AREA)
- Signal Processing (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Public Health (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Physiology (AREA)
- Power Engineering (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Psychology (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses diagnostic model method for building up, memory module and the processing equipments of a kind of ADHD, the diagnostic model of the DHD is for identifying ADHD, the present invention obtains discrete brain electricity sample signal x (n) first, then 5 layers of wavelet decomposition of db5 small echo are carried out to x (n), the coefficient of wavelet decomposition matrix of a5, d5, d4, d3 and d2 scale is read respectively, and calculate the energy value of coefficient of wavelet decomposition, the feature parameter vectors T is reconstructed, and to EEG signals energy value EjIt is normalized, the feature vector after being normalized is E'jWith energy feature matrix after being normalized, finally using energy feature matrix after normalization as the input of SVM classifier, Radial basis kernel function carries out the pattern-recognition of brain wave as kernel function, and using the training pattern after progress pattern-recognition as the diagnostic model of ADHD.The identification of ADHD is carried out using diagnostic model established by the present invention, speed is fast, and accuracy rate is high.
Description
Technical field
The present invention relates to medical diagnosis on disease fields, diagnostic model method for building up, storage more specifically to a kind of ADHD
Module and processing equipment.
Background technique
Medically, brain wave inspection plays important work in the supervision of the medical assessment of doctor, diagnosis and medical procedures
With application range includes all kinds of cerebral diseases such as Nao Zhong Liu ﹑ encephalitis, brain function decline, epilepsy and cerebral injury.In addition, the technology
Also the research just applied to other health problems of the mankind extends the range of EEG signals detection and analytical technology.
The healthy growth and development of children are constantly subjected to the great attention of parent and society.But children are growing up and are developing
Some problems that stage shows such as self-closing disease, hyperactivity, retarded etc. are then sent out unlike other diseases are so obvious in time
It is existing.Attention deficit hyperactivity disorder (Attention Deficit Hyperactivity Disorder, ADHD) is commonly known as
Hyperactivity is the Childhood most common mental disease, and world wide is interior, and there are about 5% children to be influenced by ADHD, performance
Predominantly symptoms, these illnesss such as poor, mostly dynamic, impulsion of control force can be always existed with child growth.Have approximately half of
The above problem of ADHD children continue to teenager even adulthood, and with for example break laws and commit crime, social function it is bad, study
The generation of the problems such as under achievement.It was verified that early find and timely intervene, can effectively correct children hyperactivity,
The diseases such as retarded or self-closing disease, currently, the diagnosis of these diseases can be completed by the detection and analysis to brain wave.
Summary of the invention
The technical problem to be solved in the present invention is that for it is existing to ADHD diagnosis there are accuracy rate it is lower and diagnosis when
Between slow technological deficiency, provide diagnostic model method for building up, memory module and the processing equipment of a kind of ADHD.
The invention discloses diagnostic model method for building up, memory module and the processing equipments of a kind of ADHD, and the DHD's examines
For disconnected model for identifying to ADHD, the present invention obtains discrete brain electricity sample signal x (n) first, then carries out to x (n)
5 layers of wavelet decomposition of db5 small echo, read the coefficient of wavelet decomposition matrix of a5, d5, d4, d3 and d2 scale respectively, and calculate small
The energy value of Wave Decomposition coefficient reproduces the feature parameter vectors T, and to EEG signals energy value EjIt is normalized, obtains normalizing
Feature vector after change is E'jIt, finally will energy feature matrix conduct after normalization with energy feature matrix after being normalized
The input of SVM classifier, Radial basis kernel function carry out the pattern-recognition of brain wave as kernel function, and will be after progress pattern-recognition
Diagnostic model of the training pattern as ADHD.The identification of ADHD is carried out using diagnostic model established by the present invention, speed is fast,
Accuracy rate is high.
Detailed description of the invention
Present invention will be further explained below with reference to the attached drawings and examples, in attached drawing:
Fig. 1 is the wavelet decomposition tree schematic diagram that scale is 5;
Fig. 2 is each frequency band reconstruction signal of the EEG signal based on wavelet decomposition;
Fig. 3 is the eeg signal classification recognition principle block diagram based on WT and SVM;
Fig. 4 is change curve of the classification accuracy rate with parameter d;
Fig. 5 is change curve of the classification accuracy rate with parameter (g, C);
Fig. 6 is grid data service optimizing result figure;
Fig. 7 is the flow chart of the diagnostic model method for building up of ADHD.
Specific embodiment
For a clearer understanding of the technical characteristics, objects and effects of the present invention, now control attached drawing is described in detail
A specific embodiment of the invention.
The characteristics extraction of ADHD EEG signals:
EEG signals are made of multiple frequency contents, containing a large amount of useful informations, select method appropriate to brain wave patterns
It is analyzed and to extract pathological characters most important.Normal EEG signals and the distribution of ADHD EEG signals energy space are different, energy
The change of amount also means that the change of EEG signals feature, and wavelet transformation (Wavelet Transform, WT) analysis method
EEG signals can be decomposited according to different frequency and carry out independent analysis, therefore multiresolution wavelet joint time frequency analysis can be used
Method carries out multiple dimensioned power feature extraction to typical case's non-stationary signal as EEG signals.
The wavelet decomposition of EEG signals:
Multiresolution wavelet analysis is actually the detail signal approximation signal and high frequency that signal decomposition is low frequency,
Second of decomposition is done to the signal of low frequency part again, is decomposed into two parts of approximation signal and detail signal again, however to height
The signal of frequency is further processed, and is successively decomposed and has been obtained the decomposition coefficient of different frequency sections.Wavelet decomposition tree (scale 5
When) as shown in Figure 1, wherein A represents approximation signal, D represents detail signal, and serial number immediately after is then the number of plies decomposed.So
Signal is reconstructed using coefficient of wavelet decomposition afterwards, then the EEG signals of the available different rhythm and pace of moving things.
Select suitable small echo and reasonable Decomposition order extremely important for the extraction of characteristic value, it is small using db5 herein
Wave carries out 5 rank wavelet decompositions as morther wavelet, by collected original eeg data, then reconstructs to every layer of wavelet coefficient,
Corresponding brain wave rhythm signal can be obtained.
According to the method described above, EEG signals decomposition, reconstruct are carried out by programming in Matlab, obtains result of spectrum analysis
As shown in Figure 2.
Original EEG signals are decomposed using wavelet transformation technique to obtain the wavelet coefficient of each layer, by being reconstructed
Different brain wave rhythms are obtained, spectrum analysis then is carried out to different brain wave rhythms again, analysis result is as shown in Figure 2.
Each detail signal of EEG signals with wavelet decomposition scale from small to large, according to high frequency to low frequency successively by
It decomposites and, myoelectricity and Hz noise are mainly manifested on D2 scale (frequency is 31.25~62.5Hz), and brain electricity β wave is main
Respectively on D3 scale (frequency is 15.6~31.25Hz), α wave is mainly distributed on D4 scale (frequency is 7.8~15.6Hz),
θ wave is mainly distributed on D5 scale (frequency be 3.91~7.8Hz), δ wave be mainly distributed on A5 scale (frequency is 0.97~
3.91Hz).By wavelet multi_resolution analysis, original EEG signals are broken down on different frequency band, to obtained different brains
The electric rhythm and pace of moving things is further to be studied.
Using normalized frequency, frequency range representated by this 5 frequency bands of extraction, as shown in table 1.
Table 1: the frequency range of each frequency range
The characteristics extraction of EEG signals:
Obtained coefficient matrix dimension is very high after carrying out wavelet multi_resolution analysis due to discrete EEG signals, if directly
It connects and carries out Classification and Identification using these data, when operation will be inconvenient, and classifying quality is also bad.And signal energy can be preferable
Indicate the feature of signal, therefore the present invention constructs the feature parameter vectors of EEG signals, this method can substantially reduce the dimension of matrix
ADHD children accurately can be identified and be classified by number.
The energy for the different brain wave rhythms that the present invention is decomposed using upper section is classified as feature vector, to discrete
The range value of EEG signals carries out the cumulative energy that can be obtained by EEG signals of quadratic sum.
In formula | xij| indicate i-th point in j-th of frequency band of range value.EEG signals energy value EjCharacterize original brain
For electric signal in j-th of frequency band, the number of the energy value of certain time period has reacted original EEG signals in the frequency band
Interior feature.
Time-frequency energy value is defined as signal characteristic, then the extraction step of signal characteristic is as follows:
Step l: discrete EEG signals x (n) is obtained;
Step 2: carrying out 5 layers of wavelet decomposition of db5 small echo to x (n), read the coefficient of wavelet decomposition matrix of each scale, press
The energy value of formula (1) calculating coefficient of wavelet decomposition;
Step 3: construction the feature parameter vectors.
The feature parameter vectors T constructed according to above step is as follows
T=[Ea5, Ed5, Ed4, Ed3, Ed2] (2)
Signal energy EjA usually biggish numerical value can make troubles to subsequent Classification and Identification, therefore can be to brain
Electrical signal energy EjIt is normalized, the feature vector after definition normalization is E'j, E'jThe calculation of one of following three formula can be used
:
The present invention is normalized signal energy using formula (5).
The energy value of EEG signals different frequency sections thus can be obtained.According to the algorithm above, to ADHD children and just
Normal children's EEG signals are handled, and construct the feature parameter vectors.
Characteristics extraction result:
Experimental data A group (normal child) and each 80 of B group (ADHD children), successively handle this 160 by upper section method
EEG signals, the energy feature value matrix of available 160*4, these matrixes characterize the feature of ADHD EEG signals.From institute
It is random respectively in each 80 eigenvalue matrix of obtained normal child and ADHD children to extract 5, energy feature such as 2 institute of table
Show.
Table 2: each frequency band energy of EEG signals
Each energy eigenvalue characterizes one group of wavelet coefficient, can reflect out the information in two domain of EEG signals time-frequency.
As seen from Table 2, in the electrical energy of brain characteristic value of ADHD children, the energy eigenvalue of θ wave is apparently higher than normal child, β wave, α wave
Energy eigenvalue be significantly lower than normal child.After obtaining the feature vector of EEG signals, the design of classifier can be carried out.
In fact the extraction of characteristic value and the design of classifier, which generally are intended to constantly test repeatedly just, can be obtained preferable classifying quality.
EEG signals mode identification technology:
EEG signals pattern-recognition refers to for after collected original EEG signals, using suitable recognizer, knows
Not Chu subject's acquisition when locating psychophysiological state or thinking mistake area.Eeg signal classification identification in most critical a step just
It is signal mode identification, can directly influence correctly Classification and Identification go out spirit locating for tester or thinking mistake area.It is right
Original EEG signals carry out Classification and Identification, and classification results are a series of classification numbers, and inhomogeneity alias corresponds to the different state of mind.
Method of EEG signals classification mainly uses at present: linear classification, artificial neural network classification and support to
The methods of amount machine method (Support Vector Machine, SVM).Wherein, support vector machines is the machine that recent two decades just propose
Device learning method.Support vector machines is simpler than other classification methods and generalization ability is strong.
Eeg signal classification identification based on WT and SVM:
From above-mentioned 160 experimental datas, chooses 100 and be used as training sample set (normal child and ADHD child dataset
Each 50), in addition 60 are used as test sample collection, train classifier using training sample set, reuse test sample collection
The classifying quality of inspection-classification device.During using support vector cassification, the Selection of kernel function of support vector machines, parameter
The accuracy rate that Bu Tong all will affect classifier of optimum choice and training sample and test sample quantity.Experimental program block diagram is such as
Shown in Fig. 3.
The classification performance of classifier can be measured by calculating classification accuracy (Accuracy).
Classification accuracy is defined as:
Wherein, true positives TP represents practical ADHD and is predicted as ADHD, and false positive FP represents practical non-ad HD and is predicted as ADHD,
True negative TN represents practical non-ad HD and is predicted as non-ad HD, and false negative FN represents practical ADHD and is predicted as non-ad HD.
Using after normalization, energy feature matrix is as the input vector of SVM classifier, after determining input vector, next
It is exactly the selection of SVM kernel function and the optimization of kernel functional parameter.
The selection of kernel function:
The common kernel function of support vector machines is linear, multinomial, radial basis function (RBF), Sigmoid etc..Using more
Item formula kernel function (K (x, xi)=[(x, xi)+1]d) carry out ADHD eeg signal classification experiment when, different polynomial kernel letters are set
Number parameter d (d=l, 2 ..., 15), and the accuracy tested every time is recorded, Fig. 4 is classified to obtain using different parameters d
Accuracy.With the raising of d, the dimension of feature space can become larger, polynomial order d it is excessive or it is too small all will be to classifying quality
It has an impact, when parameter d increases, overfitting phenomenon can occur for SVM, to reduce the generalization ability of SVM.
Use Radial basis kernel functionWhen being tested, the group of nuclear parameter g and punishment parameter C
The change curve for closing corresponding classification accuracy rate is as shown in Figure 5.When punishment parameter C is smaller, classification accuracy rate is lower, increases parameter C
The generalization ability of classifier significantly increases when value.
From the point of view of the generalization ability of above experimental result and classifier, Radial basis kernel function is a kind of common opposite
Stable kernel function has better generalization ability for EEG signals.The present invention chooses Radial basis kernel function as kernel function.
The optimization of parameter:
In the case where the kernel function of support vector machines has been determined, the quality for analyzing result is exactly the parameter decision of kernel function
's.Since choose is Radial basis kernel function to the present invention, the critical issue next to be solved is exactly Radial basis kernel function
Parameter g and punishment parameter C selection.
Very perfect method not yet, currently used supporting vector are optimized to the parameter of support vector machines now
Machine parameter optimization method mainly has: experimental method note and grid search (Grid Search) method etc..Grid data service is used herein
Optimizing result, the search range for being provided with C is [2-8,28], the search range that g is arranged is [2-8,28], setting search step pitch
It is 0.5, as a result as shown in Figure 6.
If grid data service is in the case where infinitely great search space range and infinitesimal search step pitch, system is
Optimized parameter can be searched out.But the classifying quality of most parameter group is all excessively poor in search space, only one
The classifying quality of parameter group in a smaller search space is relatively good, therefore all parameter groups meetings ten in traversal search spatial dimension
Divide and expends the time.
To overcome the defects of above parameter optimisation procedure, a kind of improved grid data service is used herein.Firstly, using
Biggish search step pitch carries out rough search within the scope of big search space, and extract so that classification accuracy it is highest that
The value of group parameter.Then, after searching out local optimum parameter, a lesser search space is chosen near the parameter group,
It carries out second using traditional grid data service to search for, to obtain optimized parameter.
Traditional grid data service and improved grid data service carry out contrast and experiment such as 3 institute of table of parameter optimization
Show.Experiment is concentrated from training sample extracts data training classifier progress optimizing test, has done 5 groups of experiments altogether.
Table 3: the result that different grid data services carry out parameter optimization compares
From table 3 it is observed that obtain classification accuracy relatively high for traditional grid data service, but take a long time, and this hair
Though the classification accuracy of bright proposed improved method is more slightly lower than traditional grid data service, time of its optimizing well below
Traditional grid data service.
Classification results:
The classification of (1) one dimensional input vector
Divided respectively using feature vector corresponding to α, β, θ, δ and θ/β, θ/α as a dimensional input vector of SVM
Class.Table 4 is the classification results of a dimensional input vector classifier.
4: one dimensional input vector classifier classification results of table
(2) multidimensional input vector is classified
Classify feature vector corresponding to α, β, θ, δ and θ/β, θ/α as SVM multidimensional input vector.Table 5 is
The classification results of multidimensional input vector classifier.
Table 5: multidimensional input vector classifier classification results
By table 4 and table 5 for ADHD children's EEG signals, with β, θ energy value and θ/β, θ/α energy ratio are according to tool
There is preferable classifying quality, and the classification accuracy of the SVM of multidimensional input vector is higher than the svm classifier standard of a dimensional input vector
True rate.Therefore the present invention uses β, θ energy value and θ/β, the input vector that θ/α energy ratio is support vector machines.
After determining kernel function and input vector, optimized parameter is found, ADHD children are examined using the SVM model
It is disconnected.60 (normal child and each 30 of ADHD child dataset) test samples in 160 eigenvalue matrix that last chapter is extracted
Collect the input vector as diagnosis.Its diagnostic result is as shown in table 6.
Table 6:ADHD diagnosis output result
From diagnostic result it can be calculated that using wavelet decomposition, β, θ energy value and θ/β, θ/α energy ratio are extracted
It is inputted as feature, the accuracy rate of svm classifier method is 91.7%, and classification results are more satisfactory.This method, which can be used as, quickly examines
A kind of approach of disconnected ADHD children.
The present invention has studied emphatically eeg signal acquisition using ADHD children EEG signals as research object
The relevant technologies and ADHD EEG signals classification and recognition methods.By carrying out feature to the children's EEG signals collected
Value is extracted, classification and identification, to diagnose whether patient suffers from attention deficit hyperactivity disorder disease.The experimental results showed that small echo is become
It changes and is combined with support vector machines technology, can effectively analyze children's brain wave and diagnose ADHD.
Data are analyzed based on the above principles, and the invention proposes the diagnostic model method for building up of ADHD a kind of, and the DHD's examines
Disconnected model is for identifying ADHD.With reference to Fig. 7, above-mentioned diagnostic model method for building up includes the following steps:
S1, discrete brain electricity sample signal x (n) is obtained;
S2,5 layers of wavelet decomposition that db5 small echo is carried out to x (n), read the small echo of a5, d5, d4, d3 and d2 scale respectively
Decomposition coefficient matrix is calculated the energy value of coefficient of wavelet decomposition by formula (1);
In formula, | xij| indicate i-th point in j-th of frequency band of range value, EjFor EEG signals energy value;D2 scale generation
Table myoelectricity and Hz noise, d3 scale represent β wave, and d4 scale represents α wave, and d5 scale represents θ wave, and a5 scale represents δ wave;
S3, the feature parameter vectors T is constructed according to formula (2), and to EEG signals energy value EjIt is normalized, obtains normalizing
Feature vector after change is E'jWith energy feature matrix after being normalized;
T=[Ea5, Ed5, Ed4, Ed3, Ed2] (8)
S4, using energy feature matrix after normalization as the input of SVM classifier, Radial basis kernel function as kernel function into
The pattern-recognition of row brain wave, and using the training pattern after progress pattern-recognition as the diagnostic model of ADHD.
Wherein, the corresponding frequency of d2 scale is 31.25~62.5Hz, and the corresponding frequency of d3 scale is 15.6~31.25Hz,
The corresponding frequency of d4 scale is 7.8~15.6Hz, and the corresponding frequency of d5 scale is 3.91~7.8Hz, the corresponding frequency of a5 scale
For 0.97~3.91Hz.
To EEG signals energy value E described in step S2jIt is normalized, the feature vector after being normalized is E'jTool
Body refers to using any one formula in following formula (9)~(11) to EjIt is normalized;
Further include the steps that optimizing the parameter g and punishment parameter C of Radial basis kernel function in step S4, when optimization adopts
With improved grid data service: firstly, rough search is carried out within the scope of big search space using biggish search step pitch, and
Extract the value so that highest that group of parameter of classification accuracy;Then, after searching out local optimum parameter, in the parameter group
A lesser search space is nearby chosen, second is carried out using traditional grid data service and searches for, to obtain optimized parameter.
When the improved grid data service used scans for there is parameter g and punishment parameter C to be respectively provided with preset search
Rope range, and there is preset search step pitch.
The input vector of SVM classifier be 5 dimensional input vectors, respectively the corresponding feature of α, β, θ, δ, θ/β and θ/α to
Amount.
The present invention also provides a kind of memory modules, are stored with program instruction, and described program instructs for realizing above-mentioned
Diagnostic model method for building up.
The present invention also provides a kind of processing equipment, what the above-mentioned diagnostic model method for building up of the processing equipment was established is examined
Disconnected model carries out the identification of ADHD.
The embodiment of the present invention is described with above attached drawing, but the invention is not limited to above-mentioned specific
Embodiment, the above mentioned embodiment is only schematical, rather than restrictive, those skilled in the art
Under the inspiration of the present invention, without breaking away from the scope protected by the purposes and claims of the present invention, it can also make very much
Form, all of these belong to the protection of the present invention.
Claims (8)
1. the diagnostic model method for building up of ADHD a kind of, for identifying to ADHD, feature exists the diagnostic model of the DHD
In the diagnostic model method for building up includes the following steps:
S1, discrete brain electricity sample signal x (n) is obtained;
S2,5 layers of wavelet decomposition that db5 small echo is carried out to x (n), read the wavelet decomposition of a5, d5, d4, d3 and d2 scale respectively
Coefficient matrix is calculated the energy value of coefficient of wavelet decomposition by formula (1);
In formula, | xij| indicate i-th point in j-th of frequency band of range value, EjFor EEG signals energy value;D2 scale represents flesh
Electricity and Hz noise, d3 scale represent β wave, and d4 scale represents α wave, and d5 scale represents θ wave, and a5 scale represents δ wave;
S3, the feature parameter vectors T is constructed according to formula (2), and to EEG signals energy value EjIt is normalized, after obtaining normalization
Feature vector be E'jWith energy feature matrix after being normalized;
T=[Ea5, Ed5, Ed4, Ed3, Ed2] (2)
S4, using energy feature matrix after normalization as the input of SVM classifier, Radial basis kernel function carries out brain as kernel function
The pattern-recognition of electric wave, and using the training pattern after progress pattern-recognition as the diagnostic model of ADHD.
2. diagnostic model method for building up according to claim 1, which is characterized in that the corresponding frequency of d2 scale is 31.25
The corresponding frequency of~62.5Hz, d3 scale is 15.6~31.25Hz, and the corresponding frequency of d4 scale is 7.8~15.6Hz, d5 scale
Corresponding frequency is 3.91~7.8Hz, and the corresponding frequency of a5 scale is 0.97~3.91Hz.
3. diagnostic model method for building up according to claim 1, which is characterized in that EEG signals energy described in step S2
Magnitude EjIt is normalized, the feature vector after being normalized is E'jIt specifically refers to using appointing in following formula (3)~(5)
A formula anticipate to EjIt is normalized;
。
4. diagnostic model method for building up according to claim 1, which is characterized in that further include to radial direction in the step S4
The step of parameter g and punishment parameter C of base kernel function are optimized uses improved grid data service when optimization: firstly, using
Biggish search step pitch carries out rough search within the scope of big search space, and extract so that classification accuracy it is highest that
The value of group parameter;Then, after searching out local optimum parameter, a lesser search space is chosen near the parameter group,
It carries out second using traditional grid data service to search for, to obtain optimized parameter.
5. diagnostic model method for building up according to claim 4, which is characterized in that the improved grid data service of use into
There is parameter g and punishment parameter C to be respectively provided with preset search range when row search, and there is preset search step pitch.
6. diagnostic model method for building up according to claim 1, which is characterized in that the input vector of the SVM classifier
For 5 dimensional input vectors, the respectively corresponding feature vector of α, β, θ, δ, θ/β and θ/α.
7. a kind of memory module, which is characterized in that be stored with program instruction, described program instructs for realizing such as claim 1-
Diagnostic model method for building up described in 6 any one.
8. a kind of processing equipment, which is characterized in that use diagnostic model method for building up as claimed in any one of claims 1 to 6
The diagnostic model established carries out the identification of ADHD.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810333017.7A CN108937922A (en) | 2018-04-13 | 2018-04-13 | A kind of diagnostic model method for building up, memory module and the processing equipment of ADHD |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201810333017.7A CN108937922A (en) | 2018-04-13 | 2018-04-13 | A kind of diagnostic model method for building up, memory module and the processing equipment of ADHD |
Publications (1)
Publication Number | Publication Date |
---|---|
CN108937922A true CN108937922A (en) | 2018-12-07 |
Family
ID=64498589
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201810333017.7A Pending CN108937922A (en) | 2018-04-13 | 2018-04-13 | A kind of diagnostic model method for building up, memory module and the processing equipment of ADHD |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108937922A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110335685A (en) * | 2019-07-05 | 2019-10-15 | 河海大学常州校区 | A kind of ADHD classification diagnosis method based on brain connection dual subspace study |
CN110384507A (en) * | 2019-07-16 | 2019-10-29 | 西安石油大学 | A kind of detection method based on lip optics woundless measurement of blood sugar concentration |
CN111603135A (en) * | 2020-05-11 | 2020-09-01 | 江南大学 | Low-power-consumption epilepsy detection circuit based on master-slave support vector machine |
Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102149320A (en) * | 2008-04-28 | 2011-08-10 | 科技研究局(A*Star) | Method and system for concentration detection |
US20110245633A1 (en) * | 2010-03-04 | 2011-10-06 | Neumitra LLC | Devices and methods for treating psychological disorders |
CN103720456A (en) * | 2012-10-12 | 2014-04-16 | 昆山冠强软件科技有限公司 | Attention defect disorder feedback type game treatment device based on bioelectricity |
CN103845050A (en) * | 2012-12-06 | 2014-06-11 | 上海市浦东新区知识产权保护协会 | Attention deficiency disorder feedback type play therapy device based on bioelectricity |
US9095266B1 (en) * | 2010-08-02 | 2015-08-04 | Chi Yung Fu | Method for treating a patient |
CN105342569A (en) * | 2015-11-25 | 2016-02-24 | 新乡医学院 | Mental state detection system based on electroencephalogram analysis |
CN105677035A (en) * | 2016-01-26 | 2016-06-15 | 杭州电子科技大学 | EEMD (Ensemble Empirical Mode Decomposition) and wavelet threshold based motor imagery electroencephalogram signal denoising method |
WO2018060878A1 (en) * | 2016-09-27 | 2018-04-05 | Mor Research Applications Ltd. | Eeg microstates for controlling neurological treatment |
-
2018
- 2018-04-13 CN CN201810333017.7A patent/CN108937922A/en active Pending
Patent Citations (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102149320A (en) * | 2008-04-28 | 2011-08-10 | 科技研究局(A*Star) | Method and system for concentration detection |
US20110245633A1 (en) * | 2010-03-04 | 2011-10-06 | Neumitra LLC | Devices and methods for treating psychological disorders |
US9095266B1 (en) * | 2010-08-02 | 2015-08-04 | Chi Yung Fu | Method for treating a patient |
CN103720456A (en) * | 2012-10-12 | 2014-04-16 | 昆山冠强软件科技有限公司 | Attention defect disorder feedback type game treatment device based on bioelectricity |
CN103845050A (en) * | 2012-12-06 | 2014-06-11 | 上海市浦东新区知识产权保护协会 | Attention deficiency disorder feedback type play therapy device based on bioelectricity |
CN105342569A (en) * | 2015-11-25 | 2016-02-24 | 新乡医学院 | Mental state detection system based on electroencephalogram analysis |
CN105677035A (en) * | 2016-01-26 | 2016-06-15 | 杭州电子科技大学 | EEMD (Ensemble Empirical Mode Decomposition) and wavelet threshold based motor imagery electroencephalogram signal denoising method |
WO2018060878A1 (en) * | 2016-09-27 | 2018-04-05 | Mor Research Applications Ltd. | Eeg microstates for controlling neurological treatment |
Non-Patent Citations (7)
Title |
---|
付强等: "表面波和Retinex结合的水声图像增强方法", 《应用科学学报》 * |
姚涛等: "ADHD患儿脑电生物反馈治疗前、后脑电图变化情况分析", 《海南医学院学报》 * |
李钢等: "支持向量机在脑电信号分类中的应用", 《计算机应用》 * |
梁静坤著: "《基于想象驾驶行为的脑机接口控制》", 31 December 2015, 北京国防工业出版社 * |
牛洁: "注意力相关脑电的特征提取及分类方法研究", 《中国优秀硕士学位论文全文数据库信息科技》 * |
罗甜等: "测风激光雷达信号载噪比和有效性算法研究", 《软件与算法》 * |
郭士琦等: "电压骤降暂态扰动点检测方法研究", 《农业科技与装备》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110335685A (en) * | 2019-07-05 | 2019-10-15 | 河海大学常州校区 | A kind of ADHD classification diagnosis method based on brain connection dual subspace study |
CN110384507A (en) * | 2019-07-16 | 2019-10-29 | 西安石油大学 | A kind of detection method based on lip optics woundless measurement of blood sugar concentration |
CN111603135A (en) * | 2020-05-11 | 2020-09-01 | 江南大学 | Low-power-consumption epilepsy detection circuit based on master-slave support vector machine |
CN111603135B (en) * | 2020-05-11 | 2021-09-28 | 江南大学 | Low-power-consumption epilepsy detection circuit based on master-slave support vector machine |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Aslan et al. | Automatic Detection of Schizophrenia by Applying Deep Learning over Spectrogram Images of EEG Signals. | |
Rubin et al. | Recognizing abnormal heart sounds using deep learning | |
CN109934089B (en) | Automatic multi-stage epilepsia electroencephalogram signal identification method based on supervised gradient raiser | |
CN109770932B (en) | Processing method of multi-modal brain nerve image features | |
Subasi | Epileptic seizure detection using dynamic wavelet network | |
US8577451B2 (en) | System and methods for neurologic monitoring and improving classification and treatment of neurologic states | |
CN111202517B (en) | Sleep automatic staging method, system, medium and electronic equipment | |
CN111134664B (en) | Epileptic discharge identification method and system based on capsule network and storage medium | |
CN108937922A (en) | A kind of diagnostic model method for building up, memory module and the processing equipment of ADHD | |
CN110598793A (en) | Brain function network feature classification method | |
CN114366124A (en) | Epilepsia electroencephalogram identification method based on semi-supervised deep convolution channel attention single classification network | |
CN112932498A (en) | T wave morphology classification system with strong generalization capability based on deep learning | |
Jadhav et al. | Automated sleep stage scoring using time-frequency spectra convolution neural network | |
Zeng et al. | Ensemble learning on heartbeat type classification | |
Liu et al. | Using deep learning and machine learning to detect epileptic seizure with electroencephalography (EEG) data | |
CN113796873B (en) | Wearable dynamic electrocardiosignal classification method and system | |
Qin et al. | Deep multi-scale feature fusion convolutional neural network for automatic epilepsy detection using EEG signals | |
Skarga-Bandurova et al. | A Multilayer LSTM Auto-Encoder for Fetal ECG Anomaly Detection | |
Chen et al. | Identification of atrial fibrillation from electrocardiogram signals based on deep neural network | |
Lin et al. | Discrete Wavelet Transform based ECG classification using gcForest: A deep ensemble method | |
Wu et al. | Automatic diagnostics of EEG pathology via capsule network with multi-level feature fusion | |
Vu et al. | A Comparison of Machine Learning Algorithms for Parkinson’s Disease Detection | |
Sakli et al. | Lightweight CNN Towards Skin Lesions Automated Diagnosis In Dermoscopic Images | |
Kumar et al. | Early Prediction of Parkinson’s Disease using Multiple SVM Classifiers | |
Shi et al. | Lung sound recognition method based on multi-resolution interleaved net and time-frequency feature enhancement |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181207 |
|
RJ01 | Rejection of invention patent application after publication |