CN210277161U - Epileptic seizure early warning equipment based on electroencephalogram signal processing - Google Patents
Epileptic seizure early warning equipment based on electroencephalogram signal processing Download PDFInfo
- Publication number
- CN210277161U CN210277161U CN201821439651.0U CN201821439651U CN210277161U CN 210277161 U CN210277161 U CN 210277161U CN 201821439651 U CN201821439651 U CN 201821439651U CN 210277161 U CN210277161 U CN 210277161U
- Authority
- CN
- China
- Prior art keywords
- electroencephalogram
- upper computer
- signal
- channel signal
- channel
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
Landscapes
- Measurement And Recording Of Electrical Phenomena And Electrical Characteristics Of The Living Body (AREA)
Abstract
The utility model discloses an epileptic seizure early warning equipment based on brain electrical signal handles belongs to data processing technology field, and this equipment includes: the device comprises a signal acquisition device and an upper computer; the signal acquisition equipment acquires an electroencephalogram signal to be detected; the upper computer extracts a channel signal of a preset channel from the electroencephalogram signal to be detected, and divides the channel signal according to a preset time window to obtain a plurality of segments of channel signals; respectively acquiring time domain characteristics and frequency characteristics of each channel signal segment, and combining the time domain characteristics and the frequency characteristics of each channel signal segment according to spatial positions to obtain characteristic vectors; and carrying out epileptic seizure early warning according to the characteristic vector. The utility model discloses when carrying out the epileptic seizure early warning, synthesized the characteristics of two aspects of signal time domain, frequency domain, improved the prediction result degree of accuracy.
Description
Technical Field
The utility model relates to a data processing technology field, in particular to seizure early warning equipment based on brain electrical signal handles.
Background
Epilepsy is a chronic dysfunction disease caused by sudden abnormal discharge of cerebral neurons, and the epilepsy can cause obstacles in aspects of movement, consciousness, sensation, behavior and the like of a patient during the seizure, thereby bringing great pain to the patient. For some incurable epilepsy, the epileptic seizure early warning technology can give an early warning some time before the real epilepsy seizure so as to take necessary treatment measures in advance and reduce the harm to the patient.
At present, the main technology of epilepsy early warning is analysis and processing of electroencephalogram signals. Abnormal features such as spike, sharp wave, spike-slow complex wave and the like can appear in the electroencephalogram signals before the epileptic seizure, and the extraction and classification of the abnormal features are the key points of the epileptic early warning. As a random non-stationary signal, an electroencephalogram signal has the characteristics of strong background noise, huge original data set and the like, so that the extraction of signal characteristics is difficult, and common processing methods comprise time domain analysis, fast Fourier transform, wavelet transform, autoregressive analysis and the like. By using the methods, the characteristic vectors are extracted from the original data, and then a considerable accuracy rate of automatic detection can be obtained by selecting a proper classifier and parameters thereof.
Many relevant studies have been made on the early warning of epileptic seizures. Chinese patent publication No. CN108320800A, provides an electroencephalogram data analysis method for epileptic seizure detection and pre-seizure prediction. The method comprises the steps of obtaining time-frequency domain characteristics through processing scalp electroencephalogram (EEG) signals, then screening a characteristic set through a characteristic selection algorithm, training and classifying the obtained optimal characteristic subset by using a Support Vector Machine (SVM), and determining prediction and detection results according to classification results.
However, since the waveform characteristics of the epileptic seizure and the waveform characteristics of the pre-seizure are different, the contradiction between the detection accuracy and the predicted time length cannot be reconciled by using the same method. In this case, the focus of the technique is still the detection of seizures, which limits the prediction time to only about 5s, resulting in a high detection accuracy of 95% or more. However, in terms of actual condition analysis, the early warning time of about 5s is too short to meet the requirement, and doctors cannot take treatment measures in time.
SUMMERY OF THE UTILITY MODEL
For solving foretell whole or partial technical problem, the utility model provides a seizure early warning equipment based on EEG signal processing, seizure early warning equipment based on EEG signal processing includes: the device comprises a signal acquisition device and an upper computer;
the signal acquisition equipment is used for acquiring a current electroencephalogram of a user at the current moment, acquiring a historical electroencephalogram of the user within a preset time length from the current moment, and taking the current electroencephalogram and the historical electroencephalogram as electroencephalograms to be detected;
the upper computer is used for extracting channel signals of a preset channel from the electroencephalogram signals to be detected, and dividing the channel signals according to a preset time window to obtain a plurality of segments of channel signals;
the upper computer is further used for respectively obtaining the time domain characteristics and the frequency characteristics of each channel signal segment, and combining the time domain characteristics and the frequency characteristics of each channel signal segment according to the spatial position to obtain a characteristic vector;
and the upper computer is also used for carrying out epileptic seizure early warning according to the characteristic vector.
Preferably, the upper computer is further configured to obtain a mean value, a variance, a skewness and a kurtosis of each segment of channel signal, and use the mean value, the variance, the skewness and the kurtosis of each segment of channel signal as a time domain feature of each segment of channel signal.
Preferably, the upper computer is further configured to perform wavelet transformation on each segment of channel signal respectively to obtain decomposition coefficients of each segment of channel signal in different frequency components;
and the upper computer is also used for calculating the wavelet energy of each section of channel signal according to the decomposition coefficients of each section of channel signal in different frequency components, and taking the wavelet energy of each section of channel signal as the frequency characteristic of each section of channel signal.
Preferably, the upper computer is further configured to perform db5 wavelet transform on each segment of channel signal to obtain decomposition coefficients of frequency components of each segment of channel signal respectively at D1, D2, D3, D4, D5, D6, D7, and D8.
Preferably, the upper computer is further configured to combine the decomposition coefficients of the channel signals of each segment belonging to the same sub-band, so as to obtain the decomposition coefficients of the channel signals of each segment in different sub-bands;
the upper computer is also used for calculating the wavelet energy of each channel signal by the following formula,
wherein E isiIs the wavelet energy of the i-th channel signal, Di(k) Is the decomposition coefficient of the k sub-band of the i channel signal.
Preferably, the preset channels are 6 channels of the left posterior half brain.
Preferably, the upper computer is further configured to perform seizure early warning through a preset classifier according to the feature vector.
Preferably, the preset classifier is a support vector machine classifier or a logistic regression classifier.
Preferably, the upper computer is further configured to apply FIR filtering to the electroencephalogram signal to be detected to remove component noise above a preset frequency.
Preferably, the upper computer is further configured to remove the baseline drift of the electroencephalogram signal to be detected.
The utility model discloses when carrying out the epileptic seizure early warning, synthesized the characteristics of two aspects of signal time domain, frequency domain, improved the prediction result degree of accuracy.
Drawings
Fig. 1 is a block diagram of an embodiment of the utility model, which is based on electroencephalogram signal processing.
Detailed Description
The following detailed description of the embodiments of the present invention is provided with reference to the accompanying drawings and examples. The following examples are intended to illustrate the invention, but are not intended to limit the scope of the invention.
Fig. 1 is a block diagram of an embodiment of an epileptic seizure early warning device based on electroencephalogram signal processing; referring to fig. 1, the seizure early warning apparatus based on electroencephalogram signal processing includes: the signal acquisition equipment 100 and the upper computer 200;
the signal acquisition equipment 100 is configured to acquire a current electroencephalogram of a user at a current time, acquire a historical electroencephalogram of the user within a preset time length from the current time, and use the current electroencephalogram and the historical electroencephalogram as electroencephalograms to be detected;
it should be noted that, because a certain reaction exists in the electroencephalogram signal before the epileptic seizure occurs, the current electroencephalogram signal and the historical electroencephalogram signal can be used as the electroencephalogram signal to be detected.
Since the frequency variation range of the human electroencephalogram is between 1 to 30Hz, the human electroencephalogram can be divided into several basic waveforms, such as θ, δ, α, β, and the like, according to the frequency, different waveforms dominate in different mental states, and in order to improve the accuracy of subsequent processing, in this embodiment, after the electroencephalogram signal to be detected is obtained, the signal acquisition device 100 may filter the electroencephalogram signal to be detected by using Finite Impulse Response (FIR) to remove component noise above a preset frequency, which may be set to 30Hz, of course.
The upper computer 200 is used for extracting channel signals of preset channels from the electroencephalogram signals to be detected, and dividing the channel signals according to a preset time window to obtain a plurality of segments of channel signals;
in order to improve the accuracy of the subsequent prediction result, in this embodiment, the upper computer 200 is further configured to remove the baseline drift of the electroencephalogram signal to be detected.
In particular implementations, the data partitioning between pre-episode and inter-episode needs to be determined for the same patient. Taking 6min before onset as prophase and at least one hour before and after onset as inter-onset. In order to reduce the contingency of experimental results, a data file before 4 epileptic seizures can be used as pre-seizure data, and data with the same length is intercepted in a data file without epileptic seizures as an inter-seizure period according to symmetry. For the stitched data set, 80% was used for model training and 20% was used as test data. The data is segmented in a 6s non-overlapping time window and the data for each segment is processed below.
It should be noted that each piece of data contains information of all 23 channels, but since the epileptic seizure characteristics are only shown in a part of the channels, too many electrodes will also burden the patient and reduce the algorithm operation speed, so that it is necessary to screen the channels. In order to ensure the accuracy of prediction, the number of electrodes cannot be too small, and about 6 electrodes are generally adopted. Usually, the lesion is located at different positions in different patients, and therefore, the channel should be selected according to actual conditions. In the embodiment, by looking at the channel positions, the information of 6 channels of the left posterior half brain is extracted, which represents the regional local epileptic seizure and the whole-body epileptic seizure whole-brain propagation.
The upper computer 200 is further configured to obtain time domain features and frequency features of each channel signal segment, and combine the time domain features and the frequency features of each channel signal segment according to spatial positions to obtain feature vectors;
it should be noted that, the time domain features and the frequency features of each channel signal are combined together, normalized to form feature vectors, the vectors of 6 channels of the left and right hemispheres are combined according to spatial positions, and a label is added to complete feature extraction.
It can be understood that, from the time domain perspective, the time domain characteristics of each channel signal can be calculated, and the first to fourth moments of the data can be used as the time domain characteristics, that is, the time domain characteristics can include: the mean value, the variance, the skewness and the kurtosis are measures of amplitude distribution symmetry and relatively steep and smooth degree, and from experimental results, the decrease of the variance and the increase of the kurtosis are significant features of an early stage of onset, that is, in this implementation, the upper computer 200 is further configured to obtain the mean value, the variance, the skewness and the kurtosis of each channel signal segment respectively, and use the mean value, the variance, the skewness and the kurtosis of each channel signal segment as time domain features of each channel signal segment;
in a specific implementation, from the angle of a frequency domain, wavelet transform is performed on an electroencephalogram signal, and the signal is decomposed to different levels, so as to obtain sub-signals with different frequency components, that is, the upper computer 200 is further configured to perform wavelet transform on each segment of channel signal, so as to obtain decomposition coefficients of each segment of channel signal in different frequency components;
in this embodiment, the time domain signal adopts statistics (i.e., a mean, a variance, a skewness, and a kurtosis), and the calculation amount of the wavelet transform is small, so that the dimensionality of the generated feature vector is not high, and the prediction time overhead can be reduced on the premise of ensuring the accuracy.
The upper computer 200 is further configured to calculate wavelet energy of each segment of channel signal according to the decomposition coefficients of each segment of channel signal in different frequency components, and use the wavelet energy of each segment of channel signal as the frequency characteristic of each segment of channel signal.
In order to make the obtained sub-bands coincide with the rhythms of the frequency bands, such as θ, δ, α, β, etc., in the electroencephalogram (EEG) signal, in this embodiment, the upper computer 200 is further configured to perform db5 wavelet transformation on each segment of the channel signal, respectively, to obtain the decomposition coefficients of the frequency components of each segment of the channel signal at D1, D2, D3, D4, D5, D6, D7, and D8, respectively.
The upper computer 200 is further configured to merge the decomposition coefficients of the channel signals belonging to the same sub-band to obtain the decomposition coefficients of the channel signals in different sub-bands, and when merging the decomposition coefficients, the following table 1 may be referred to;
TABLE 1
The upper computer 200 is further configured to calculate wavelet energy of each channel signal according to the following formula,
wherein E isiIs the wavelet energy of the i-th channel signal, Di(k) Is the decomposition coefficient of the k sub-band of the i channel signal.
The upper computer 200 is further used for carrying out epileptic seizure early warning according to the characteristic vectors.
In order to facilitate the early warning of the epileptic seizure, in this embodiment, the upper computer 200 is further configured to perform the early warning of the epileptic seizure through a preset classifier according to the feature vector, that is, for the obtained feature vector, in this embodiment, a plurality of classifiers such as a support vector machine and a logistic regression may be used for classification, and since accidental errors possibly occurring in a classification result may cause phenomena such as no alarm and false alarm, the judgment of the early warning start time point and the early warning end time point may be defined. The transition from a stable pre-episode to a stable pre-episode is considered an early warning onset, the transition from a stable pre-episode to a stable pre-episode is considered an early warning resolution point, and a period of time between the onset and resolution points is a risk period. If more than 80% of the classification results are in the inter-attack period or the early-attack period within a certain time, the stable inter-attack period or the early-attack period can be judged to be entered.
When the epileptic seizure early warning is carried out, the characteristics of the signal in the time domain and the frequency domain are integrated, and the accuracy of a prediction result is improved.
The present invention is described below with reference to specific examples, but the scope of the present invention is not limited thereto.
The prediction device uses a Boston children hospital public data set for testing, a database contains data of 24 patients, a sampling rate of 256Hz and 23 channels, file information contains information of epilepsy occurrence time, ending time and the like of each acquired patient, the total time is 844 hours, the number of epileptic seizures is 182, and data files are stored in edf format.
The detection rate and the prediction rate are analyzed by the following methods respectively:
and (3) analyzing the detection rate: a five-fold cross validation method was performed on the data set. The system adopts different classifiers to obtain the detection accuracy of classification results as shown in the following table 2:
TABLE 2
And (3) analyzing the prediction rate: the data are predicted by using the model obtained by training, and the early warning time for the three epilepsies is shown in the following table 3:
| onset of disease | Attack one | Attack two | Three attacks |
| Time of early warning | 52.5min | 35.1min | 5.6min |
TABLE 3
Therefore, on the basis of ensuring the prediction accuracy, longer early warning time is obtained.
Compared with the prior art, the utility model provides a technical scheme is better to epilepsy real-time early warning effect. On one hand, only wavelet transform is used as the features, the dimensionality of the extracted features is low, the computational complexity is reduced, and the operating efficiency of the computer is improved to some extent. On the other hand, the early warning time is prolonged by non-restrictive prediction time and shifting the time window, and sufficient time is reserved for taking measures. The two aspects feature that the prediction result meets the requirements of the application.
The technical solution of the present invention is explained in detail from several aspects of system composition, structure, coefficient calculation principle, upper computer display interface, use flow, etc. through the description of the drawings and the detailed embodiments. The above-described embodiments are only preferred embodiments of the present invention, and it is obvious to those skilled in the art that the present invention can be applied to various medical instrument systems by modifying or replacing the same, and not limited to the system structure described in the embodiments of the present invention, so that the above-described embodiments are only preferred and not restrictive.
The above embodiments are only examples of the present invention, and the above embodiments are only used for illustrating the technical solutions and concepts of the present invention and not for limiting the scope of the claims of the present invention. Other technical solutions that may be obtained by a person skilled in the art through logic analysis, reasoning or limited experiments based on the concepts of the present patent in combination with the prior art should also be considered to fall within the scope of the present invention.
The above embodiments are only used for illustrating the present invention, and not for limiting the present invention, and those skilled in the relevant technical field can make various changes and modifications without departing from the spirit and scope of the present invention, so that all equivalent technical solutions also belong to the scope of the present invention, and the protection scope of the present invention should be defined by the claims.
Claims (10)
1. An epileptic seizure early warning device based on electroencephalogram signal processing, characterized in that the epileptic seizure early warning device based on electroencephalogram signal processing comprises: the device comprises a signal acquisition device and an upper computer;
the signal acquisition equipment is used for acquiring a current electroencephalogram of a user at the current moment, acquiring a historical electroencephalogram of the user within a preset time length from the current moment, and taking the current electroencephalogram and the historical electroencephalogram as electroencephalograms to be detected;
the upper computer is used for extracting channel signals of a preset channel from the electroencephalogram signals to be detected, and dividing the channel signals according to a preset time window to obtain a plurality of segments of channel signals;
the upper computer is further used for respectively obtaining the time domain characteristics and the frequency characteristics of each channel signal segment, and combining the time domain characteristics and the frequency characteristics of each channel signal segment according to the spatial position to obtain a characteristic vector;
and the upper computer is also used for carrying out epileptic seizure early warning according to the characteristic vector.
2. The electroencephalogram signal processing-based epileptic seizure early-warning device of claim 1, wherein the upper computer is further configured to obtain a mean value, a variance, a skewness and a kurtosis of each segment of channel signal respectively, and use the mean value, the variance, the skewness and the kurtosis of each segment of channel signal as time-domain features of each segment of channel signal.
3. The EEG processing-based epileptic seizure early warning device as recited in claim 1, wherein said upper computer is further configured to perform wavelet transformation on each channel signal segment to obtain decomposition coefficients of each channel signal segment in different frequency components;
and the upper computer is also used for calculating the wavelet energy of each section of channel signal according to the decomposition coefficients of each section of channel signal in different frequency components, and taking the wavelet energy of each section of channel signal as the frequency characteristic of each section of channel signal.
4. The electroencephalogram signal processing-based epileptic seizure pre-warning device of claim 3, wherein the upper computer is further configured to perform db5 wavelet transformation on each channel signal respectively to obtain decomposition coefficients of frequency components of each channel signal respectively at D1, D2, D3, D4, D5, D6, D7 and D8.
5. The electroencephalogram signal processing-based epileptic seizure early-warning device as claimed in claim 4, wherein the upper computer is further configured to combine the decomposition coefficients of the channel signals belonging to the same sub-band to obtain the decomposition coefficients of the channel signals in different sub-bands;
the upper computer is also used for calculating the wavelet energy of each channel signal by the following formula,
wherein E isiIs the wavelet energy of the i-th channel signal, Di(k) Is the decomposition coefficient of the k sub-band of the i channel signal.
6. The electroencephalogram signal processing-based epileptic seizure early-warning device as claimed in any one of claims 1 to 5, wherein the preset channels are 6 channels of the left and back half brains.
7. The electroencephalogram signal processing-based epileptic seizure early warning device as claimed in any one of claims 1 to 5, wherein the upper computer is further used for carrying out epileptic seizure early warning through a preset classifier according to the feature vectors.
8. The electroencephalogram signal processing-based seizure pre-warning device of claim 7, wherein the preset classifier is a support vector machine classifier or a logistic regression classifier.
9. The electroencephalogram signal processing-based epileptic seizure early-warning device as claimed in any one of claims 1 to 5, wherein the upper computer is further configured to apply FIR filtering to the electroencephalogram signal to be detected so as to remove component noise above a preset frequency.
10. The electroencephalogram signal processing-based epileptic seizure early-warning device as claimed in any one of claims 1 to 5, wherein the upper computer is further used for removing baseline drift of the electroencephalogram signal to be detected.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201821439651.0U CN210277161U (en) | 2018-09-04 | 2018-09-04 | Epileptic seizure early warning equipment based on electroencephalogram signal processing |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201821439651.0U CN210277161U (en) | 2018-09-04 | 2018-09-04 | Epileptic seizure early warning equipment based on electroencephalogram signal processing |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN210277161U true CN210277161U (en) | 2020-04-10 |
Family
ID=70058651
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201821439651.0U Active CN210277161U (en) | 2018-09-04 | 2018-09-04 | Epileptic seizure early warning equipment based on electroencephalogram signal processing |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN210277161U (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109106365A (en) * | 2018-09-04 | 2019-01-01 | 杭州航弈生物科技有限责任公司 | Epileptic attack source of early warning based on EEG Processing |
-
2018
- 2018-09-04 CN CN201821439651.0U patent/CN210277161U/en active Active
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109106365A (en) * | 2018-09-04 | 2019-01-01 | 杭州航弈生物科技有限责任公司 | Epileptic attack source of early warning based on EEG Processing |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Hamad et al. | Feature extraction of epilepsy EEG using discrete wavelet transform | |
| Zhang et al. | Low-complexity seizure prediction from iEEG/sEEG using spectral power and ratios of spectral power | |
| Riaz et al. | EMD-based temporal and spectral features for the classification of EEG signals using supervised learning | |
| Kaya et al. | 1D-local binary pattern based feature extraction for classification of epileptic EEG signals | |
| Iriarte et al. | Independent component analysis as a tool to eliminate artifacts in EEG: a quantitative study | |
| Fathima et al. | Wavelet based features for epileptic seizure detection | |
| Kumari et al. | Seizure detection in EEG using time frequency analysis and SVM | |
| Gruszczyńska et al. | Epilepsy identification based on EEG signal using RQA method | |
| Hosseini et al. | Emotional stress recognition system for affective computing based on bio-signals | |
| Clark et al. | Multiresolution decomposition of non-stationary EEG signals: a preliminary study | |
| Boubchir et al. | A review of feature extraction for EEG epileptic seizure detection and classification | |
| CN109106365A (en) | Epileptic attack source of early warning based on EEG Processing | |
| Xia et al. | Seizure detection approach using S-transform and singular value decomposition | |
| Schönwald et al. | Quantifying chirp in sleep spindles | |
| Chaibi et al. | A reliable approach to distinguish between transient with and without HFOs using TQWT and MCA | |
| CN106485208A (en) | The automatic removal method of eye electrical interference in single channel EEG signals | |
| Zhang et al. | Automatic detection of interictal epileptiform discharges based on time-series sequence merging method | |
| Ramos-Aguilar et al. | Analysis of EEG signal processing techniques based on spectrograms | |
| WO2015066791A1 (en) | Method and apparatus for processing eeg signals | |
| Rashed-Al-Mahfuz et al. | Artifact suppression and analysis of brain activities with electroencephalography signals | |
| Hanrahan | Noise reduction in EEG signals using convolutional autoencoding techniques | |
| CN210277161U (en) | Epileptic seizure early warning equipment based on electroencephalogram signal processing | |
| Fathima et al. | Wavelet based features for classification of normal, ictal and interictal EEG signals | |
| Antonakakis et al. | Synchronization coupling investigation using ICA cluster analysis in resting MEG signals in reading difficulties | |
| Shahnaz et al. | Seizure detection exploiting EMD-wavelet analysis of EEG signals |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| GR01 | Patent grant | ||
| GR01 | Patent grant |