CN114176609A - Stroke severity classification prediction model establishment method based on electroencephalogram signals - Google Patents
Stroke severity classification prediction model establishment method based on electroencephalogram signals Download PDFInfo
- Publication number
- CN114176609A CN114176609A CN202111624138.5A CN202111624138A CN114176609A CN 114176609 A CN114176609 A CN 114176609A CN 202111624138 A CN202111624138 A CN 202111624138A CN 114176609 A CN114176609 A CN 114176609A
- Authority
- CN
- China
- Prior art keywords
- stroke
- relative power
- prediction model
- classification prediction
- calculating
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 21
- 210000004556 brain Anatomy 0.000 claims abstract description 24
- 238000010801 machine learning Methods 0.000 claims abstract description 10
- 238000004458 analytical method Methods 0.000 claims abstract description 7
- 238000007781 pre-processing Methods 0.000 claims abstract description 4
- 238000004364 calculation method Methods 0.000 claims description 17
- 239000011159 matrix material Substances 0.000 claims description 9
- 238000012545 processing Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 3
- 208000006011 Stroke Diseases 0.000 abstract description 83
- 238000002474 experimental method Methods 0.000 abstract description 8
- 206010008190 Cerebrovascular accident Diseases 0.000 abstract description 6
- 238000000540 analysis of variance Methods 0.000 abstract description 6
- 238000012216 screening Methods 0.000 abstract 2
- 238000010276 construction Methods 0.000 abstract 1
- 238000010586 diagram Methods 0.000 description 5
- 238000003066 decision tree Methods 0.000 description 4
- 238000012706 support-vector machine Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 208000032382 Ischaemic stroke Diseases 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 238000000537 electroencephalography Methods 0.000 description 2
- 238000002610 neuroimaging Methods 0.000 description 2
- 238000004393 prognosis Methods 0.000 description 2
- 230000010410 reperfusion Effects 0.000 description 2
- 238000005070 sampling Methods 0.000 description 2
- 238000002560 therapeutic procedure Methods 0.000 description 2
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000007177 brain activity Effects 0.000 description 1
- 230000005978 brain dysfunction Effects 0.000 description 1
- 210000001627 cerebral artery Anatomy 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000010968 computed tomography angiography Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000012880 independent component analysis Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000007971 neurological deficit Effects 0.000 description 1
- 230000007310 pathophysiology Effects 0.000 description 1
- 230000010412 perfusion Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 230000000284 resting effect Effects 0.000 description 1
- 239000000932 sedative agent Substances 0.000 description 1
- 230000001624 sedative effect Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012549 training Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—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
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
- A61B5/372—Analysis of electroencephalograms
- A61B5/374—Detecting the frequency distribution of signals, e.g. detecting delta, theta, alpha, beta or gamma waves
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4058—Detecting, measuring or recording for evaluating the nervous system for evaluating the central nervous system
- A61B5/4064—Evaluating the brain
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/725—Details of waveform analysis using specific filters therefor, e.g. Kalman or adaptive filters
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
- A61B5/7267—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems involving training the classification device
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7271—Specific aspects of physiological measurement analysis
- A61B5/7275—Determining trends in physiological measurement data; Predicting development of a medical condition based on physiological measurements, e.g. determining a risk factor
Landscapes
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Surgery (AREA)
- Veterinary Medicine (AREA)
- General Health & Medical Sciences (AREA)
- Animal Behavior & Ethology (AREA)
- Biophysics (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Heart & Thoracic Surgery (AREA)
- Medical Informatics (AREA)
- Molecular Biology (AREA)
- Psychiatry (AREA)
- Artificial Intelligence (AREA)
- Physiology (AREA)
- Signal Processing (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Psychology (AREA)
- Neurology (AREA)
- Evolutionary Computation (AREA)
- Fuzzy Systems (AREA)
- Mathematical Physics (AREA)
- Neurosurgery (AREA)
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The invention discloses a stroke severity classification prediction model building method based on electroencephalogram signals. First, the electroencephalogram signal of the subject was recorded during the experiment. And preprocessing the electroencephalogram signals. Then screening characteristic values in delta band relative power, theta band relative power, alpha band relative power, beta band relative power, the ratio of delta band relative power to alpha band relative power, the ratio of slower frequency relative power to faster relative power and brain symmetry index through variance analysis. And finally, a machine learning construction model is used for distinguishing the stroke severity. The invention provides a method for classifying the severity of stroke by screening characteristic values based on analysis of variance and outputting the accuracy by matching machine learning. The model provided by the invention can more effectively distinguish the apoplexy severity degree according to the characteristic values extracted by the electroencephalograms of normal people and patients with different apoplexy severity degrees.
Description
Technical Field
The invention belongs to the field of bioelectricity signal processing, and relates to a stroke severity classification prediction model establishing method based on a brain electric signal.
Background
Stroke is a major cause of adult disability worldwide, and electroencephalography, one of the most important techniques for diagnosing stroke, is considered as a standard tool for assessing changes in brain activity, with a temporal resolution on the order of milliseconds. The severity of stroke and the speed at which stroke treatment is found to have a large impact on disability and mortality. However, despite the demographic, clinical and imaging factors associated with stroke prognosis, early prediction of stroke short-term and long-term prognosis remains challenging due to the large individual variation. Therefore, it is extremely important to find a biological signal that can distinguish different stroke severity levels from the control group as a characteristic value.
Neuroimaging and clinical evaluation have been shown to play a key role in the diagnosis of ischemic stroke at present. Treatment eligibility was determined primarily on neuroimaging after admission, but it was important to distinguish between eligible and ineligible patients as quickly as possible. If this can be done before admission, the enthusiasm of the team can be mobilized, the clinical management can be optimized, the best clinical management and effect can be realized for the patient, and the medical service resources can be more effectively utilized.
The most commonly used assessment methods for stroke-related neurological deficits are: stroke scales, mri or ct perfusion techniques, which are time-consuming and costly, are well established in the national institutes of health. Electroencephalograms are a non-invasive, inexpensive, and high-time-resolution diagnostic method, which are helpful for rapidly recording the electrical activity of the brain, and have been used as a diagnostic and prognostic tool for stroke. Electroencephalograms are of great value in assessing brain dysfunction. Some studies have shown that continuous monitoring of the electroencephalographic index prior to any clinical change can provide information for immediate assessment of the effectiveness of ischemic stroke reperfusion therapy. In addition, electroencephalography can provide continuous monitoring of changing brain pathophysiology, whereas techniques such as electron computed tomography angiography of the cerebral arteries cannot. There are many valuable electroencephalogram indicators in stroke monitoring, such as total power, relative delta and alpha power, ratio of slower and faster frequencies, and brain symmetry indicators, which are commonly used as tools for assessing whether stroke is a concern.
Stroke patients of varying severity receive treatment that varies greatly. Patients with severe stroke need to be assured of being quickly transported to a hospital that provides emergency reperfusion therapy. The treatment effect of patients is seriously influenced by the problems that mild stroke is misjudged into severe stroke, up to 30 percent of severe stroke patients are missed in the pre-hospital examination, and the like. If the characteristic value for judging the severity of the stroke is successfully found, the time for the patient to receive the treatment is greatly saved, which is beneficial to the treatment recovery of the patient. Therefore, it is imperative to quickly distinguish the severity of stroke.
Disclosure of Invention
In order to overcome the defects in the prior art, the invention provides a stroke severity classification prediction model building method based on electroencephalogram signals.
The method firstly collects the electroencephalogram signals of a testee in a resting fist-making state, and preprocesses the electroencephalogram signals through an eeglab toolbox in Matlab. Then, the characteristic value for judging the severity of the stroke is screened by using single-factor variance analysis. And finally, performing accuracy fitting on the extracted characteristic values by using a machine learning classifier so as to complete modeling.
In order to achieve the above object, the method of the present invention mainly comprises the following steps:
step 1: collecting electroencephalograms of normal people, patients with mild stroke, patients with moderate stroke and patients with severe stroke.
Step 2: and (4) preprocessing the electroencephalogram signals.
And step 3: and (4) calculating an electroencephalogram signal.
And 4, step 4: the eigenvalues were extracted by analysis of variance.
And 5: and inputting the characteristic value into a machine learning classifier for accuracy calculation.
Step 5-1: and inputting the characteristic value into a support vector machine and carrying out accuracy calculation.
Step 5-2: inputting the characteristic values into a decision tree and carrying out accuracy calculation;
step 5-3: and inputting the characteristic value into K nearest neighbor and carrying out accuracy calculation.
Step 5-4: and inputting the characteristic value into naive Bayes and carrying out accuracy calculation.
Step 6: and establishing a stroke severity classification prediction model by using the characteristic values.
Further, in the step 1, according to the international 10-20 standard, the electroencephalogram signal of the subject is collected. Further, the electroencephalogram signal processing step in the step 2 is as follows:
(1) a channel is located. The coordinates are located according to the international 10-20 standard.
(2) The useless electrodes and the re-reference are removed.
(3) And (4) selecting a channel.
(4) And (4) band-pass filtering. The lower limit of the frequency band is selected to be 1Hz, and the upper limit of the frequency band is selected to be 30 Hz.
(5) And (6) correcting the baseline. The data of the rest part at the time of data acquisition was used as a baseline for correction.
(6) Independent component analysis removes artifacts.
(7) Look at the independent source components.
(8) And (5) eliminating noise.
Further, the calculation of the electroencephalogram signal in step 3 includes:
(1) the delta band relative power is calculated. The expression is as follows:
where rPSD is the relative power, f1、f2Respectively corresponding to frequency bandStart-stop frequency band. f. ofL、fH1Hz and 30Hz respectively, PSD refers to the power spectral density of the corresponding frequency.
(2) The relative power of the theta band is calculated. The expression is as follows:
(3) the alpha band relative power is calculated. The expression is the same as above.
(4) The beta band relative power is calculated. The expression is the same as above.
(5) The ratio of the delta band relative power to the alpha band relative power is calculated.
(6) The ratio of the slower frequency relative power to the faster relative power is calculated. The expression is as follows:
where R is the ratio of the slower frequency relative power to the faster relative power, rPSDdeltaIs delta band relative power, rPSDthetaIs the theta band relative power, rPSDalphaIs the alpha band relative power, rPSDbetaIs the beta band relative power.
(7) And calculating the brain symmetry index. The expression is as follows:
where pdBSI represents the brain symmetry index, i represents each channel, and j represents each sample point. R represents the right brain, L represents the left brain, M is the number of channels, and N is the number of sampling points.
(8) And (5) calculating the Shannon entropy. The expression is as follows:
where H (x) represents Shannon entropy, n is the number of classes, xiIs of the i-th class, P (x)i) Is xiThe probability of (c).
Further, the machine learning classifier in step 5 includes:
(1) and a support vector machine.
(2) And (4) a decision tree.
(3) And K is nearest neighbor.
(4) Naive bayes.
Further, the accuracy calculation in step 5 includes:
and (3) taking the sample with the labeled known category as the input of the classifier, and comparing the overall accuracy of each classifier by selecting different classifiers. The accuracy of each severity of stroke is compared by computing the confusion matrix for the different classifiers.
Further, the step 6 of establishing a stroke severity prediction model by using the feature values includes:
and selecting three groups with most significant differences after the analysis of variance in pairs as characteristic values to be input into a machine learning classifier. And selecting the classifier with high overall accuracy or high accuracy of each severity degree according to the requirement.
The invention has the beneficial effects that: previous models were not able to efficiently distinguish between strokes of different severity or only between stroke patients and normal persons. The classification model method provided by the invention can more effectively distinguish the apoplexy severity degree according to the characteristic values extracted by the electroencephalograms of normal people and patients with different apoplexy severity degrees.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a diagram of an electroencephalogram signal experimental data acquisition process, wherein a hand-clenching state to an opening state lasts for 15 seconds, and the hand-clenching state is an experimental recording state and is marked as record; the hand is opened and the fist is clenched for 15 seconds, and the state is a rest state and is recorded as rest; three experiments were recorded as one group and three groups were recorded per subject.
FIG. 3 is a graph showing the results of a relative power ANOVA, wherein (a) the relative delta power, (b) the relative theta power, (c) the relative alpha power, and (b) the relative beta power.
FIG. 4 is a graph of power ratio ANOVA results, wherein (a) the ratio of delta band relative power to alpha band relative power, and (b) the ratio of slower frequency relative power to faster frequency relative power.
FIG. 5 is a diagram of Shannon entropy analysis results, wherein (a) delta band Shannon entropy, (b) theta band Shannon entropy, (c) alpha band Shannon entropy, and (b) beta band Shannon entropy.
FIG. 6 is a diagram showing the results of a brain symmetry-forming exponential ANOVA.
FIG. 7 is a schematic diagram of various classifier confusion matrices, wherein (a) a support vector machine confusion matrix, (b) a K nearest neighbor confusion matrix, (c) a decision tree confusion matrix, and (d) a naive Bayes confusion matrix.
FIG. 8 is a diagram illustrating prediction accuracy of different classifiers.
Detailed Description
As shown in fig. 1, the present embodiment includes the following steps:
step 1: and (6) data acquisition. Collecting the electroencephalogram data of a subject. The data acquisition is divided into an experimental record part and a rest part. The experimental record part lasted 15 seconds from the fist-closed state to the open-hand state. Rest was from open hand to fist for 15 seconds. At the beginning of the experiment, the subject was asked to stare at a white cross on the black screen of the notebook for 30 seconds at rest. When the subject hears the "clenches" the subject clenches the fist and medical personnel record the data and apply a label. When the subject hears the "loose hand", the subject opens the hand and medical personnel end the data recording and labeling. 3 consecutive recordings were made for one experiment, 3 experiments were made per subject. The experiment required subjects to take sedative drugs prior to the experiment, see figure 2.
Step 2: and preprocessing the collected electroencephalogram data through Matlab. The processing flow is to introduce data into an eeglab tool box of matlab, sequentially perform channel positioning, re-reference, channel selection, band-pass filtering and baseline correction, analyze independent components to remove artifacts, view independent source components and remove noise.
And step 3: the electroencephalogram signal calculation comprises relative power calculation, brain symmetry index calculation and Shannon entropy calculation. The relative power calculation method is as follows:
and step two, eliminating the noise electroencephalogram data in the step two.
The brain formation symmetry index calculation method comprises the following steps:
where pdBSI represents the brain symmetry index, i represents each channel, and j represents each sample point. R represents the right brain, L represents the left brain, M is the number of channels, and N is the number of sampling points.
The Shannon entropy calculation method comprises the following steps:
where H (x) represents Shannon entropy, n is the number of classes, xiIs of the i-th class, P (x)i) Is xiThe probability of (c).
And 4, step 4: and (4) carrying out variance analysis on the result of the step (3) by using SPSS software, and judging the significance. The significant difference, i.e., the value of P in SPSS less than 0.01, was chosen as the eigenvalue.
And 5: the feature values are input to a machine learning classifier. The data set is divided into 10 equal subsets. The ratio of training set to test set was 7: 3. The experiment was repeated one thousand times and the average was taken as the final result.
Step 6: and selecting the classifier with the highest accuracy as a stroke severity prediction model.
The method respectively calculates the relative delta power, the relative theta power, the relative alpha power, the relative beta power, the ratio of the delta frequency band relative power to the alpha frequency band relative power, the ratio of the slower frequency relative power to the faster relative power, the delta frequency band Shannon entropy, the theta frequency band Shannon entropy, the alpha frequency band Shannon entropy, the beta frequency band Shannon entropy and the brain symmetry index of the stroke patient with different severity degrees.
As can be seen from FIG. 3, there is a significant difference in relative delta power between the patients with severe stroke and the healthy control group (p 0.044). There were significant differences in relative theta power between the severe and moderate stroke groups (p 0.008), the control and severe stroke groups (p 0.008), the mild and severe stroke groups (p 0.009). There were significant differences in relative alpha efficacy between the control and severe stroke (p 0.037), mild and severe stroke (p 0.029). There were significant differences in relative beta efficacy between the control and moderate stroke groups (p 0.028), mild stroke and moderate stroke group (p 0.021).
As can be seen from fig. 4. There were significant differences between the severe and moderate stroke groups (p 0.042), the control and severe stroke groups (p 0.048), and the mild and severe stroke groups (p 0.024) in the ratio of delta band relative power to alpha band relative power. There were significant differences between the control group and the severe stroke group (p 0.041), the mild stroke group and the moderate stroke group (p 0.039), and the mild stroke group and the severe stroke group (p 0.026) with respect to the ratio of the slower frequency relative power to the faster relative power.
Fig. 5 shows a comparison of shannon entropy for patients with different degrees of stroke severity. There was a significant difference in shannon entropy between the control and mild stroke (p 0.005), the control and severe stroke (p 0.006), the mild and moderate stroke (p 0.009), the mild and severe stroke (p 0.005), and the moderate and severe stroke (p 0.009) in the delta band. In the theta band, the Shannon entropy of only the mild stroke group and the severe stroke group (p is 0.0385) is significantly different. There was no significant difference in shannon entropy in the alpha band. In the beta frequency band, there is a significant difference in Shannon entropy between the control group and the moderate stroke group (p is 0.006), and between the control group and the severe stroke group (p is 0.008).
Figure 6 shows a comparison of brain symmetry indices for patients of different severity over different frequency bands. There was no significant difference between brain symmetry indices in the delta band. In the theta band, there was a significant difference in brain symmetry indices between the mild stroke group and the severe stroke group (p ═ 0.242). In the alpha frequency band, there were significant differences in brain symmetry indices between the control group and the mild stroke group (p 0.367) and between the control group and the severe stroke group (p 0.393). There was a significant difference in brain symmetry indices in the beta band between the moderate stroke group and the control group (p 0.251). However, in general, the change in the brain symmetry index depends on the frequency, and the significance is not obvious at low frequencies. With increasing frequency, the brain symmetry index of the healthy control group increased significantly.
The confusion matrix is used to determine how well the classifier predicts correctly. It is a specific matrix for the performance of the visual supervised learning algorithm. The diagonal elements represent the number of predicted object classes equal to the actual object class. The off-diagonal elements correspond to the target class of misclassification or prediction error. Fig. 7 shows the confusion matrix for four classifiers. The results show that: for the support vector machine, the accuracy rate of judgment of patients with mild stroke is lowest, and the accuracy rate of judgment of patients with moderate stroke is highest; for K nearest neighbor, the judgment accuracy rate of patients with mild apoplexy is lowest, and the judgment accuracy rate of patients with severe apoplexy is highest; for the decision tree, the judgment accuracy rate of patients with mild stroke is lowest, and the judgment accuracy rate of patients with severe stroke is highest; for naive Bayes, the accuracy of judgment of patients with severe stroke is lowest, and the accuracy of judgment of patients with mild stroke is highest. In the figure, 1 represents patients with mild stroke, 2 represents patients with moderate stroke, 3 represents patients with severe stroke, and 4 represents a control group.
Three groups with most significant differences are selected after variance analysis in pairs as characteristic values to be input into a machine learning classifier, so that Shannon entropy, the ratio of delta frequency band relative power to alpha frequency band relative power and the ratio of slower frequency relative power to faster relative power are selected as the characteristic values. Fig. 8 shows that the accuracy of the overall output result of the K nearest neighbor classifier is the highest.
Claims (6)
1. The stroke severity classification prediction model building method based on the electroencephalogram signals is characterized by comprising the following steps:
step 1: collecting electroencephalograms of normal people, patients with mild stroke, patients with moderate stroke and patients with severe stroke;
step 2: preprocessing an electroencephalogram signal;
and step 3: calculating an electroencephalogram signal;
and 4, step 4: extracting characteristic values through variance analysis;
and 5: inputting the characteristic value into a machine learning classifier, and carrying out accuracy calculation;
step 6: and establishing a stroke severity classification prediction model by using the characteristic values.
2. The stroke severity classification prediction model building method based on electroencephalogram signals as claimed in claim 1, characterized in that: in the step 1, according to the international 10-20 standard, electroencephalogram signals of a subject are collected; the experimental movements include fist making and hand opening.
3. The stroke severity classification prediction model building method based on electroencephalogram signals as claimed in claim 1, characterized in that: the electroencephalogram signal processing step in the step 2 is as follows:
(1) a positioning channel;
(2) removing the useless electrodes and the re-reference;
(3) selecting a channel;
(4) band-pass filtering;
(5) correcting a baseline;
(6) analyzing independent components to remove artifacts;
(7) viewing the independent source components;
(8) and (5) eliminating noise.
4. The stroke severity classification prediction model building method based on electroencephalogram signals as claimed in claim 1, characterized in that: step 3, electroencephalogram signal calculation comprises the following steps:
(1) calculating delta band relative power;
(2) calculating the relative power of the theta frequency band;
(3) calculating alpha frequency band relative power;
(4) calculating beta frequency band relative power;
(5) calculating the ratio of the delta band relative power to the alpha band relative power;
(6) calculating a ratio of the slower frequency relative power to the faster relative power;
(7) calculating a brain symmetry index;
(8) and (5) calculating the Shannon entropy.
5. The stroke severity classification prediction model building method based on electroencephalogram signals as claimed in claim 1, characterized in that: step 5, the accuracy calculation comprises the following steps:
(1) selecting different classifiers, and comparing the overall accuracy of each classifier;
(2) the accuracy of each severity of stroke is compared by computing the confusion matrix for the different classifiers.
6. The stroke severity classification prediction model building method based on electroencephalogram signals as claimed in claim 1, characterized in that: step 6, establishing a stroke severity classification prediction model by using the characteristic values comprises the following steps:
(1) selecting three groups with most significant differences after variance analysis in pairs as characteristic values to be input into a machine learning classifier;
(3) and selecting the classifier with high overall accuracy or high accuracy of each severity degree according to the requirement.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111624138.5A CN114176609A (en) | 2021-12-27 | 2021-12-27 | Stroke severity classification prediction model establishment method based on electroencephalogram signals |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111624138.5A CN114176609A (en) | 2021-12-27 | 2021-12-27 | Stroke severity classification prediction model establishment method based on electroencephalogram signals |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114176609A true CN114176609A (en) | 2022-03-15 |
Family
ID=80545115
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111624138.5A Pending CN114176609A (en) | 2021-12-27 | 2021-12-27 | Stroke severity classification prediction model establishment method based on electroencephalogram signals |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114176609A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102715903A (en) * | 2012-07-09 | 2012-10-10 | 天津市人民医院 | Method for extracting electroencephalogram characteristic based on quantitative electroencephalogram |
| CN109934089A (en) * | 2018-10-31 | 2019-06-25 | 北京航空航天大学 | Multistage epileptic EEG Signal automatic identifying method based on supervision gradient lifter |
| CN111028944A (en) * | 2019-12-16 | 2020-04-17 | 南昌大学第二附属医院 | Cerebrovascular disease nerve function damage degree prediction model based on kernel principal component analysis and polynomial characteristics |
| CN111067514A (en) * | 2020-01-08 | 2020-04-28 | 燕山大学 | Multi-channel electroencephalogram coupling analysis method based on multi-scale multivariable transfer entropy |
-
2021
- 2021-12-27 CN CN202111624138.5A patent/CN114176609A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102715903A (en) * | 2012-07-09 | 2012-10-10 | 天津市人民医院 | Method for extracting electroencephalogram characteristic based on quantitative electroencephalogram |
| CN109934089A (en) * | 2018-10-31 | 2019-06-25 | 北京航空航天大学 | Multistage epileptic EEG Signal automatic identifying method based on supervision gradient lifter |
| CN111028944A (en) * | 2019-12-16 | 2020-04-17 | 南昌大学第二附属医院 | Cerebrovascular disease nerve function damage degree prediction model based on kernel principal component analysis and polynomial characteristics |
| CN111067514A (en) * | 2020-01-08 | 2020-04-28 | 燕山大学 | Multi-channel electroencephalogram coupling analysis method based on multi-scale multivariable transfer entropy |
Non-Patent Citations (1)
| Title |
|---|
| CASSANDRA M. WILKINSON 等: "Application of the Muse portable EEG system to aid in rapid diagnosis of stroke", 《MEDRXIV》, 5 June 2020 (2020-06-05), pages 1 - 30 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Naz et al. | From ECG signals to images: a transformation based approach for deep learning | |
| CN111340142B (en) | Epilepsia magnetoencephalogram spike automatic detection method and tracing positioning system | |
| CN104427932B (en) | For the method and apparatus of multi-mode nerve assessment | |
| AlSharabi et al. | EEG signal processing for Alzheimer’s disorders using discrete wavelet transform and machine learning approaches | |
| CN110063732B (en) | System for early detection and risk prediction of schizophrenia | |
| Bairagi | EEG signal analysis for early diagnosis of Alzheimer disease using spectral and wavelet based features | |
| CN111568446A (en) | Portable electroencephalogram depression detection system combined with demographic attention mechanism | |
| US20100217144A1 (en) | Diagnostic and predictive system and methodology using multiple parameter electrocardiography superscores | |
| Ertuğrul et al. | Automatic diagnosis of cardiovascular disorders by sub images of the ECG signal using multi-feature extraction methods and randomized neural network | |
| CN112220485A (en) | Epilepsia electroencephalogram/magnetoencephalogram abnormal wave automatic detection method and positioning system | |
| Wang et al. | Multi-class arrhythmia detection based on neural network with multi-stage features fusion | |
| Shaban | Automated screening of parkinson's disease using deep learning based electroencephalography | |
| Webb et al. | Automated detection of artefacts in neonatal EEG with residual neural networks | |
| Rafiei et al. | Automated detection of major depressive disorder with EEG signals: a time series classification using deep learning | |
| CN115281688A (en) | Cardiac hypertrophy multi-label detection system based on multi-mode deep learning | |
| CN115363531A (en) | Epilepsy detection system based on bimodal electroencephalogram signal information bottleneck | |
| CN115590529A (en) | Epilepsia electroencephalogram signal monitoring method and system based on space-time attention mechanism | |
| CN112704503A (en) | Electrocardiosignal noise processing method | |
| Magesh et al. | Ensemble feature extraction-based prediction of fetal arrhythmia using cardiotocographic signals | |
| CN117017310A (en) | Acoustic-electric dual-mode congenital heart disease prediction device based on knowledge distillation | |
| CN114176609A (en) | Stroke severity classification prediction model establishment method based on electroencephalogram signals | |
| CN115736920A (en) | Depression state identification method and system based on bimodal fusion | |
| CN115017996A (en) | Mental load prediction method and system based on multiple physiological parameters | |
| Anchana et al. | Comparative study of detection of ADHD using EEG signals | |
| Mitra et al. | Analyzing Clinical 12-Lead ECG Images Using Deep Learning Algorithms for Objective Detection of Cardiac Diseases |
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 |