CN110811557A - Anesthesia depth monitoring system and method based on micro-state power spectrum analysis - Google Patents

Anesthesia depth monitoring system and method based on micro-state power spectrum analysis Download PDF

Info

Publication number
CN110811557A
CN110811557A CN201911118047.7A CN201911118047A CN110811557A CN 110811557 A CN110811557 A CN 110811557A CN 201911118047 A CN201911118047 A CN 201911118047A CN 110811557 A CN110811557 A CN 110811557A
Authority
CN
China
Prior art keywords
micro
state
electroencephalogram
anesthesia
time
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
Application number
CN201911118047.7A
Other languages
Chinese (zh)
Inventor
王刚
刘治安
施文
闫相国
李雅敏
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Xian Jiaotong University
Original Assignee
Xian Jiaotong University
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Xian Jiaotong University filed Critical Xian Jiaotong University
Priority to CN201911118047.7A priority Critical patent/CN110811557A/en
Publication of CN110811557A publication Critical patent/CN110811557A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/48—Other medical applications
    • A61B5/4821—Determining level or depth of anaesthesia
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue
    • A61B5/14542—Measuring characteristics of blood in vivo, e.g. gas concentration, pH value; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid, cerebral tissue for measuring blood gases
    • 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
    • A61B5/7235—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
    • A61B5/7253—Details of waveform analysis characterised by using transforms
    • A—HUMAN NECESSITIES
    • A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
    • A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
    • A61B5/00—Measuring for diagnostic purposes; Identification of persons
    • A61B5/72—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

Abstract

A anesthesia depth monitoring system and method based on microstate power spectral analysis, the system includes electroencephalogram signal acquisition module A, microstate time sequence construction module B, electroencephalogram time-frequency information calculation module C, microstate power spectrum construction module D and classification identification module E five modules, firstly, high-density electroencephalogram is used for acquiring electroencephalogram signals of a tested whole brain, and then a corresponding microstate time sequence is constructed according to a microstate algorithm; meanwhile, decomposing the electroencephalogram signals through the MEMD to obtain the instantaneous frequency bands and the power of the electroencephalogram signals at different time points, and further obtaining the Hilbert spectrum of each channel; combining the micro-state time sequence structure with brain electrical state time-frequency information to obtain micro-state power spectrums with different frequencies; then, inputting the micro-state power spectrums of different frequency bands into the SVM for pattern recognition and classification respectively; the invention utilizes the brain electrical micro-state and the time-frequency information thereof to monitor the anesthesia depth, and can effectively and accurately monitor the anesthesia depth of the patient.

Description

Anesthesia depth monitoring system and method based on micro-state power spectrum analysis
Technical Field
The invention relates to the technical field of biomedical signal processing, in particular to an anesthesia depth monitoring system and method based on micro-state power spectrum analysis.
Background
Anesthesia, especially general anesthesia, is a common means of clinical treatment. Generally speaking, in the operation process, the central nerve of a patient is inhibited through inhalation of anesthetic or intravenous injection, so that the patient shows states of unconsciousness, motor function reduction, pain stimulation reaction disappearance and the like, the patient loses the memory of pain sense in the operation, the safety of the operation is improved, and the operation is convenient. The monitoring of the depth of anesthesia is an important method for guaranteeing the quality of anesthesia in clinical operations. If the depth of anesthesia is too heavy, the medication cost is increased, the recovery time of the patient is prolonged, and even anesthesia sequelae are caused to the nervous system. On the other hand, if the degree of anesthesia is shallow, the patient may be "known during the operation", which not only affects the normal operation, but also causes great physical and mental trauma to the patient.
In clinical practice, there is no universal "gold standard" for monitoring the depth of anesthesia, and in actual clinical practice, the application is relatively wide, and there are mainly monitoring methods based on clinical signs of patients and monitoring methods based on electroencephalogram signals. The former is widely used, mainly for Minimum Alveolar Concentration (MAC) monitoring, and is defined as the concentration of inhalation anesthetic in alveolar gas when 50% of subjects do not respond to traumatic stimulation, but has a disadvantage that it can only be used for evaluating the efficacy of inhalation anesthetic, and cannot be used for evaluating the depth of intravenous anesthesia and mixed anesthesia. The latter is mainly used for monitoring the electroencephalogram Bispectrum Index (BIS) of spontaneous brain electricity and monitoring the Auditory Evoked Potential (AEP) of evoked brain electricity. BIS is a dimensionless parameter, defined in the range 0-100, and is evaluated as 100 when the subject is absolutely awake; and evaluated as 0 at the most deeply anesthetized. However, BIS is strongly drug dependent, e.g. with isoflurane and N2O has no correlation. Secondly, BIS is also large for different racesAnd (4) difference. Furthermore, BIS sometimes fails to predict the patient's time to wake and recovery procedures. AEP monitors the depth of anesthesia using the patient's auditory evoked potentials generated by a repetitive sound stimulus, which reflects neuronal activity in the thalamus and primary auditory cortex and is not affected by opioids and inducers. However, AEP monitors are susceptible to the surrounding environment and AEP is dependent on human hearing, making this approach difficult for patients with hearing problems.
Disclosure of Invention
In order to overcome the problems of the above methods, the present invention aims to provide an anesthesia depth monitoring system and method based on micro-state power spectrum analysis, wherein the brain electrical micro-state is considered to represent the global scalp electric field activity, the time information and the space information of the brain electrical can be combined, the characteristics of the brain electrical topology on the time domain can be well reflected, the anesthesia depth change of the brain can be more comprehensively monitored from the time-frequency domain by combining the power spectrum of the micro-state sequence on the frequency domain, the power spectrums of different frequency bands at any time of different micro-states can be obtained by constructing the time sequence of the brain electrical micro-state tested in different states and then combining Multivariate Empirical Mode Decomposition (MEMD), and then the obtained micro-state power spectrums are input into a Support Vector Machine (SVM) as the characteristics for anesthesia state monitoring, distinguishing between awake and anesthetized states of the subject; the invention takes the brain as an organic whole, monitors the anesthesia depth by utilizing the electroencephalogram micro-state and the time-frequency information thereof, and has higher accuracy and sensitivity by combining the SVM classifier.
In order to achieve the purpose, the specific technical scheme of the invention is as follows:
an anesthesia depth monitoring system based on micro-state power spectrum analysis comprises five modules, namely an electroencephalogram signal acquisition module (A), a micro-state time sequence construction module (B), an electroencephalogram time-frequency information calculation module (C), a micro-state power spectrum construction module (D) and a classification identification module (E);
the electroencephalogram signal acquisition module (A) is used for acquiring electroencephalogram signals of samples which are subjected to general anesthesia in different anesthesia states;
the micro-state time sequence construction module (B) analyzes the electroencephalogram signals in different anesthesia states from the signals acquired by the electroencephalogram signal acquisition module (A) through an electroencephalogram micro-state algorithm to construct corresponding micro-state time sequences;
the electroencephalogram time-frequency information calculation module (C) decomposes the signals obtained by the electroencephalogram signal acquisition module (A) through the MEMD to obtain the instantaneous frequency ranges and the powers of the electroencephalogram signals at different time points, and further obtains the Hilbert spectrums of all channels;
the micro-state power spectrum construction module (D) combines the micro-state time sequence obtained by the micro-state time sequence construction module (B) with the electroencephalogram instantaneous frequency and power information obtained by the electroencephalogram time-frequency information calculation module (C) to obtain micro-state power spectrums with different frequencies;
and the classification and identification module (E) respectively inputs the micro-state power spectrums of different frequency bands obtained by the micro-state power spectrum construction module (D) as features into the SVM for pattern recognition and classification, and the two classification results of the SVM are the monitoring results of the anesthesia depth.
A monitoring method of an anesthesia depth monitoring system based on micro-state power spectrum analysis comprises the following steps:
(1) the brain blood oxygen signal acquisition module (A) is used for acquiring electroencephalogram signals of samples subjected to general anesthesia in different anesthesia states;
(2) analyzing the acquired signals by an electroencephalogram micro-state algorithm through a micro-state time sequence construction module (B) to construct corresponding micro-state time sequences;
(3) decomposing the electroencephalogram signals by using an electroencephalogram time-frequency information calculation module (C) through the MEMD to obtain the instantaneous frequency and power of the electroencephalogram signals at different time points, and further obtaining the Hilbert spectrum of each channel;
(4) combining the obtained micro-state time sequence with the instantaneous frequency and power information of the electroencephalogram signal by using a micro-state power spectrum construction module (D) to obtain micro-state power spectrums of different frequency bands;
(5) and (3) using a classification recognition module (E) to obtain the micro-state power spectrums of different frequency bands as characteristic values, inputting the characteristic values into an SVM (support vector machine) for pattern recognition and classification, wherein the two classification results of the SVM are the detection results of the anesthesia depth.
The step (2) specifically comprises:
(2.1) determining the ideal number of micro-states for the measured brain electrical signals through cross-validation (CV) and Krzanowski-LaiCriterion (KL) criteria;
(2.2) extracting peak points of a Global energy spectrum (GFP) of the electroencephalogram signal, wherein the specific formula of the GFP is as follows:
Figure BDA0002274627190000041
in the formula, K represents the total number of conductive connections, ViA potential representing the ith lead;
(2.3) clustering the EEG signals at the corresponding moment of GFP by using a Modified k-means algorithm to obtain a micro-state topology, calculating Global interpretable Variance (GEV) while clustering, and calculating 100 times of GEV to ensure that the finally obtained micro-state topology can overcome the randomness of the Modified k-means algorithm to the maximum extent;
(2.4) carrying out spatial correlation pairing on the clustered micro-states and the electroencephalogram of the GFP function peak point of the original electroencephalogram signal, taking the micro-state topology with the highest spatial correlation as the current micro-state, namely marking the electroencephalogram at the moment as the serial number of the micro-state;
the formula for spatial correlation is as follows:
Figure BDA0002274627190000051
in the formula, C is spatial correlation, n is the number of leads, u is the brain electrical mapping of Map u, v is the brain electrical mapping of Map v, and i is the ith lead;
and (2.5) marking the rest of the electroencephalogram signals according to the micro-state corresponding to the nearest GFP peak value to obtain the micro-state sequence of the corresponding anesthesia state.
The step (3) specifically comprises:
(3.1) obtaining Intrinsic Mode Functions (IMFs) of the original signal through MEMD, wherein the IMFs are expressed as follows:
Figure BDA0002274627190000052
where s (t) is the EEG signal of n channels, ci(t) is an intrinsic mode function; r (t) is a margin function;
(3.2) calculating instantaneous frequency and energy through Hilbert transform to perform short-time Fourier analysis; for the jth channel, the ith IMF (c)ij(t)) the hilbert transform is as follows:
Figure BDA0002274627190000061
wherein P represents a cauchy principal value;
(3.3) further deriving the corresponding instantaneous amplitude and frequency as follows:
aij(t)＝|cij(t)+jH(cij(t))|
Figure BDA0002274627190000062
wherein a isij(t) represents the instantaneous amplitude, θij(t) denotes the instantaneous phase, ωij(t) represents the instantaneous frequency and hence the Hilbert spectrum for each channel, e.g. for channel j at ω frequency at time t, it can be represented as Hj(ω,t)。
The step (4) specifically comprises:
(4.1) according to the determined time sequence of the micro-state, marking the obtained power spectrum of the brain electricity signal with different power spectrums, and further obtaining different micro-statesRespective Hilbert spectrum, which corresponds to the Hilbert power spectrum h for the micro-state kk(ω) is:
Figure BDA0002274627190000064
wherein the total duration is Ls，TsThe time length corresponding to the micro state k is n is the total number of channels;
(4.2) aiming at the selected frequency range, the power spectrums of different frequency bands corresponding to the micro-states can be obtained:
in the formula EkPower spectrum, omega, corresponding to the micro-state k0Is the selected frequency range.
The invention has the advantages that: the invention provides a method for continuously monitoring the anesthesia depth of a patient by combining an electroencephalogram micro state with an SVM. On the basis of taking the brain as an organic whole, the time information and the space information of the brain electricity can be combined, the anesthesia depth can be monitored by utilizing the brain electricity micro-state and the time-frequency information thereof, and meanwhile, the anesthesia depth of a patient can be effectively and accurately monitored. But also can provide a certain solution to the problem of the specificity of the anesthesia of different groups to patients.
Drawings
FIG. 1 is a flow chart of the present invention.
Fig. 2 is a diagram showing the selection result of the ideal number of micro-states.
FIG. 3 is a diagram of the results of the micro-topological clustering of 20 cases of anesthesia sessions tested.
FIG. 4 is a global interpretable variance boxplot of 20 tested micro-states at different stages of anesthesia.
Fig. 5 is a graph of the results of the micro-state power spectra of 20 tested samples.
FIG. 6 is a ROC graph of SVM classification results of BS-ML anesthesia phase in 5 sub-bands.
FIG. 7 is a ROC graph of classification results of SVM on BS-MD anesthesia phase in 5 sub-bands.
FIG. 8 is a ROC graph of classification results of SVM on BS-ML and BS-MD anesthesia phases in the full frequency band.
Detailed Description
The present invention will be described in detail with reference to the accompanying drawings. The following examples are intended to illustrate the invention but are not intended to limit the scope of the invention.
Referring to fig. 1, the anesthesia depth monitoring system based on the electroencephalogram microstate comprises an electroencephalogram signal acquisition module (A), a microstate time sequence construction module (B), an electroencephalogram time-frequency information calculation module (C), a microstate power spectrum construction module (D) and a classification identification module (E),
the electroencephalogram signal acquisition module (A) is used for acquiring electroencephalogram signals of samples which are subjected to general anesthesia in different anesthesia states;
the micro-state time sequence construction module (B) firstly analyzes the electroencephalogram signals in different anesthesia states from the signals acquired by the electroencephalogram signal acquisition module (A) through an electroencephalogram micro-state algorithm to construct corresponding micro-state time sequences;
the micro-state parameter calculation module (C) decomposes the electroencephalogram signals through the MEMD to obtain the instantaneous frequency and power of the electroencephalogram signals at different time points, and further obtains the Hilbert spectrum of each channel;
the micro-state power spectrum construction module (D) combines the obtained micro-state time sequence with the instantaneous frequency and power information of the electroencephalogram signal to obtain micro-state power spectrums of different frequency bands;
and the classification and identification module (E) takes the obtained micro-state power spectrums of different frequency bands as characteristic values, inputs the characteristic values into the SVM to perform pattern recognition and classification, and the two classification results of the SVM are the detection results of the anesthesia depth.
The embodiment is a detection method based on the monitoring system, and the detection method comprises the following steps:
(1) an electroencephalogram signal acquisition module (A) is utilized to acquire the whole brain electroencephalogram signals of 20 cases of subjects who are subjected to general anesthesia.
The step (1) specifically comprises:
(1.1) 20 collected subjects were tested for cardiovascular and cerebrovascular disease and were subjected to non-head related surgery;
(1.2) the experiment is divided into four stages, namely, resting state of eyes closed, light anesthesia, Moderate anesthesia and recovery state, which are respectively marked as Baseline (BS), Mild session (ML), Moderate session (MD) and Recovery (RC), and propofol is injected into the plasma to be tested for 10 minutes in each stage under the control of a digital injection pump (the concentration is respectively ML:0.6 mug/ML and MD:1.2 mug/ML);
(1.3) sample length of about 7 minutes, acquired using 128-lead high density EEG (in μ V; sampling frequency of 250 Hz);
(1.4) the EEG signal is not continuous for about 7 minutes per person, but is divided by EEGLAB into data segments (epochs) of 10 seconds in length. The mean (standard deviation) of the number of effectively analyzable epochs corresponding to the four anesthesia phases are: 38(5), 39(4), 38(4) and 40 (2).
(1.5) preprocessing the acquired signals, including correcting a base line, removing artifacts and noises, and filtering in a frequency range of 0.5-45 Hz;
(1.6) the finally obtained data for subsequent research is 91-lead electroencephalogram data.
(2) Analyzing the electroencephalogram signals in different anesthesia states through an electroencephalogram micro-state algorithm on the signals acquired by the electroencephalogram signal acquisition module (A), and constructing corresponding micro-state time sequences;
the step (2) specifically comprises:
(2.1) determining the ideal number of micro-states (the number corresponding to the first maximum value after 3) by two criteria of cross-evaluation (CV) and Krzanowski-LaiCriterion (KL) for the measured brain electrical signals, wherein in the present example, as shown in FIG. 2, although the ideal number of states in the BS and RC periods is 4, the ideal number of states in the ML and MD periods is 5, so that the optimal ideal number of states in the micro-states is 5 for the sake of the uniformity and rationality of the analysis;
(2.2) extracting peak points of a Global energy spectrum (GFP) of the electroencephalogram signal, wherein the specific formula of the GFP is as follows:
Figure BDA0002274627190000101
in the formula, K represents the total number of conductive connections, ViRepresenting the potential of the ith lead.
Then 91 lead brain electrical data of the moment corresponding to the GFP local maximum value is extracted;
and (2.3) clustering the EEG signals at the corresponding moment of the GFP by using a Modified k-means algorithm to obtain a micro-state topology. In this example, clustering is performed in two passes. Clustering for the first time, and clustering 10 topologies for each signal segment of each person; the second clustering, namely clustering the first clustering topology result under each anesthesia state, and clustering 5 final topologies in each state, as shown in fig. 3; because Global Extended Variance (GEV) can represent the interpretation degree of a given micro-state on total Variance, the GEV is calculated for 100 times to ensure that the finally obtained micro-state topology can overcome the randomness of the Modified k-means algorithm to the maximum extent during clustering;
and (2.4) carrying out spatial correlation pairing on the clustered micro-states and the electroencephalogram of the GFP function peak point of the original electroencephalogram signal, taking the micro-state topology with the highest spatial correlation as the current micro-state, namely marking the electroencephalogram at the moment as the serial number (namely A, B, C, D and F) of the micro-state. The microstate A, B, C, D is a well-known classical brain electrical microstate, and the microstate F mainly presents a topology with a peak value located at the center of the back of the scalp, which is close to the microstate F defined by Custo et al in the 7 brain electrical microstate defined in 2017 by combining with functional nuclear magnetic resonance, so the fifth microstate in this example is defined as microstate F.
The formula for spatial correlation is as follows:
Figure BDA0002274627190000111
in the formula, C is spatial correlation, n is the number of leads, u is the brain electrical mapping of Map u, v is the brain electrical mapping of Map v, and i is the ith lead;
(2.5) marking the rest brain electrical signals according to the micro state corresponding to the nearest GFP peak value, wherein the marking principle is as follows: the brain electrical micro-state at a certain moment is consistent with the brain electrical micro-state corresponding to the marked GFP peak value nearest to the brain electrical micro-state. Then the micro-state sequence of the corresponding anesthesia state can be obtained.
The global interpretable variance boxplot for each microstate at each anesthesia stage is shown in fig. 4, and these five microstates together account for over 60% of the peak variance of GFP in the electroencephalographic data for different tested, different anesthesia states, where the interpretability (i.e., GEV) for microstate C is always highest for all four states. Meanwhile, the spatial correlation between the micro-state topology of the BS and the micro-state topologies of the other three states is calculated to obtain the graph 5, and the difference between the states of the similar topologies is very small, and the topology of the BS and the corresponding topologies of the other three states have very high spatial correlation (mean: 97.1%, standard deviation: 3.6%), and the five micro-states exist stably in the consciousness transition process, and none of the micro-states appears and disappears suddenly.
(3) And decomposing the electroencephalogram signals by using an electroencephalogram time-frequency information calculation module (C) through the MEMD to obtain the instantaneous frequency and power of the electroencephalogram signals at different time points, and further obtaining the Hilbert spectrum of each channel.
The step (3) specifically comprises:
(3.1) for the electroencephalogram signals of n channels, obtaining Intrinsic Mode Functions (IMFs) of original signals through the MEMD, wherein the IMFs are expressed as follows:
Figure BDA0002274627190000121
where s (t) is the EEG signal of n channels, ci(t) is an intrinsic mode function; r (t) is a margin function;
(3.2) Hilbert transform to calculate instantaneous frequency and energy for short-time Fourier analysis. For the jth channel, the ith IMF (c)ij(t)) the hilbert transform is as follows:
Figure BDA0002274627190000122
wherein P represents a cauchy principal value;
(3.3) further deriving the corresponding instantaneous amplitude and frequency as follows:
aij(t)＝|cij(t)+jH(cij(t))|
Figure BDA0002274627190000124
wherein a isij(t) represents the instantaneous amplitude, θij(t) denotes the instantaneous phase, ωij(t) represents the instantaneous frequency and hence the Hilbert spectrum for each channel, e.g. for channel j at ω frequency at time t, it can be represented as Hj(ω,t)；
(4) And combining the obtained micro-state time sequence with the instantaneous frequency and power information of the electroencephalogram signal by using a micro-state power spectrum construction module (D) to obtain micro-state power spectrums of different frequency bands.
The step (4) specifically comprises:
(4.1) according to the determined time sequence of the micro-state, marking the obtained power spectrum of the brain electricity signal with different power spectrums, and further obtaining respective Hilbert spectrums of different micro-states, for example, for the micro-state k, the corresponding Hilbert power spectrum h thereofk(ω) is:
Figure BDA0002274627190000131
wherein the total duration is Ls，TsThe time length corresponding to the micro state k is n is the total number of channels;
(4.2) obtaining power spectrums of different frequency bands corresponding to the micro-states according to the selected frequency range:
Figure BDA0002274627190000132
in the formula EkPower spectrum, omega, corresponding to the micro-state k0Is the selected frequency range.
In this example, five bands are divided, delta (0.5-4Hz), theta (4-8Hz), alpha (8-15Hz), beta (15-25Hz), and low gamma (25-45 Hz). The obtained power spectral lines are summed in the same frequency band, so as to obtain the distribution of the power values of each frequency band, as shown in fig. 5.
From the general trend of spectral lines, it can be seen that all the micro-states are relatively similar in distribution in the three stages of BS, ML and RC, the power is higher in the delta and low alpha bands, and there is a small peak in the low alpha band. However, in the MD phase, the power of the five micro-states makes a large difference. First, there is a more or less reduction in power for all five micro-states in the low alpha band, however the C and D reductions are most pronounced; second, significant increases in power occur in both beta and low gamma (25-30Hz) and delta bands, with the largest change being the micro-regime F.
From the power level of the micro-states, it can be seen that the power of the micro-state C is maximum in all frequency bands. The micro state F is at the position where the power is lowest in all frequency bands from the rest state at the beginning, and the power is obviously increased in all frequency bands compared with other micro states as the anesthesia degree is increased to the MD anesthesia stage. This change disappears again as the subject recovers from the anesthetic state, and the power spectra of all the micro-states recover again similar to the resting state.
(4) And (3) using a classification recognition module (E) to obtain the micro-state power spectrums of different frequency bands as characteristic values, inputting the characteristic values into an SVM (support vector machine) for pattern recognition and classification, wherein the two classification results of the SVM are the detection results of the anesthesia depth.
The step (5) specifically comprises:
(5.1) selecting a basic microscopic state power spectrum of each subject in BS, ML and MD states as a feature, inputting the feature into an SVM (linear kernel, c is 1) for training and testing, wherein power spectrums of 5 sub-bands divided by a full band are respectively tested, and then the power spectrums of the sub-bands which are not divided by the full band are tested;
(5.2) the results were verified by Leave-One-Out Cross Validation (LOOCV). Since each test was run once during LOOCV, the results are extracted to generate a Receiver Operating Characteristic (ROC) curve. In the process of drawing the AUC, the data is divided into a positive class and a negative class according to a two-classification mode, and a calculation formula reflecting the accuracy (accuracyy) of the accuracy standard in all data and the sensitivity (sensitivity) of the accuracy standard in the positive class data is as follows:
Figure BDA0002274627190000151
Figure BDA0002274627190000152
in the formula: the TP judges the data number of the positive type actually; the FN judges the number of data of the negative class, actually the positive class. The TN judges as the data number of the negative type, actually the negative type; FP judges the data number of the positive class and actually the negative class.
In the present invention, the data in the anesthesia phase is defined as positive class, the data in the awake phase is defined as negative class, and the plotting results are shown in fig. 6-8, wherein fig. 6 is the ROC graph of classification results of SVM on BS-ML anesthesia phase in 5 sub-bands. FIG. 7 is a ROC graph of classification results of SVM on BS-MD anesthesia phase in 5 sub-bands. FIG. 8 is a ROC graph of classification results of SVM on BS-ML and BS-MD anesthesia phases in the full frequency band.
In the case of molecular frequency bands, Area Under the Curve (AUC) achieves that the classification of SVM to BS and MD is 0.920 of beta segment at the most and 0.625 of alpha segment to BS and ML. Under the condition of no sub-band division, the classification of the SVM to the BS-MD and the BS-ML is respectively 0.887 and 0.607 by the area under the curve. The method shows good distinguishing capability for distinguishing the waking state from the anesthesia state, and has high operability and application value.

Claims (5)

1. An anesthesia depth monitoring system based on an electroencephalogram micro-state is characterized by comprising an electroencephalogram signal acquisition module (A), a micro-state time sequence construction module (B), an electroencephalogram time-frequency information calculation module (C), a micro-state power spectrum construction module (D) and a classification identification module (E);
the electroencephalogram signal acquisition module (A) is used for acquiring electroencephalogram signals of samples which are subjected to general anesthesia in different anesthesia states;
the micro-state time sequence construction module (B) firstly analyzes the electroencephalogram signals in different anesthesia states from the signals acquired by the electroencephalogram signal acquisition module (A) through an electroencephalogram micro-state algorithm to construct corresponding micro-state time sequences;
the electroencephalogram time-frequency information calculation module (C) decomposes the signals obtained by the electroencephalogram signal acquisition module (A) through the MEMD to obtain the instantaneous frequency ranges and the powers of the electroencephalogram signals at different time points, and further obtains the Hilbert spectrums of all channels;
the micro-state power spectrum construction module (D) combines the micro-state time sequence obtained by the micro-state time sequence construction module (B) with the electroencephalogram instantaneous frequency and power information obtained by the electroencephalogram time-frequency information calculation module (C) to obtain micro-state power spectrums with different frequencies;
and the classification and identification module (E) respectively inputs the micro-state power spectrums of different frequency bands obtained by the micro-state power spectrum construction module (D) as features into the SVM for pattern recognition and classification, and the two classification results of the SVM are the monitoring results of the anesthesia depth.
2. The monitoring method of the anesthesia depth monitoring system based on the electroencephalogram micro-state is characterized by comprising the following steps of:
(1) the brain blood oxygen signal acquisition module (A) is used for acquiring electroencephalogram signals of samples subjected to general anesthesia in different anesthesia states;
(2) analyzing the acquired signals by an electroencephalogram micro-state algorithm through a micro-state time sequence construction module (B) to construct corresponding micro-state time sequences;
(3) decomposing the electroencephalogram signals by using an electroencephalogram time-frequency information calculation module (C) through the MEMD to obtain the instantaneous frequency and power of the electroencephalogram signals at different time points, and further obtaining the Hilbert spectrum of each channel;
(4) combining the obtained micro-state time sequence with the instantaneous frequency and power information of the electroencephalogram signal by using a micro-state power spectrum construction module (D) to obtain micro-state power spectrums of different frequency bands;
(5) and (3) using a classification recognition module (E) to obtain the micro-state power spectrums of different frequency bands as characteristic values, inputting the characteristic values into an SVM (support vector machine) for pattern recognition and classification, wherein the two classification results of the SVM are the detection results of the anesthesia depth.
3. The monitoring method according to claim 2, wherein the step (2) specifically comprises:
(2.1) determining the ideal number of the micro-states by two criteria of cross-validation and Krzanowski-Lai Criterion for the measured brain electrical signals;
(2.2) extracting peak points of a Global energy spectrum (GFP) of the electroencephalogram signal, wherein the specific formula of the GFP is as follows:
Figure FDA0002274627180000021
in the formula, K represents the total number of conductive connections, ViA potential representing the ith lead;
(2.3) clustering the electroencephalogram signals at the corresponding moment of GFP by using a Modified k-means algorithm to obtain a micro-state topology, calculating a global interpretable variance (GEV) while clustering, and calculating 100 GEVs to maximize the GEV so as to ensure that the finally obtained micro-state topology can overcome the randomness of the Modified k-means algorithm to the maximum extent;
(2.4) carrying out spatial correlation pairing on the clustered micro-states and the electroencephalogram of the GFP function peak point of the original electroencephalogram signal, taking the micro-state topology with the highest spatial correlation as the current micro-state, namely marking the electroencephalogram at the moment as the serial number of the micro-state;
the formula for spatial correlation is as follows:
Figure FDA0002274627180000031
wherein C is a spatial correlation, NeThe number of leads is U, the brain electrical mapping of Map u is u, v is the brain electrical mapping of Map v, and i is the ith lead;
and (2.5) marking the rest of the electroencephalogram signals according to the micro-state corresponding to the nearest GFP peak value to obtain the micro-state sequence of the corresponding anesthesia state.
4. The monitoring method according to claim 2, wherein the step (3) specifically comprises:
(3.1) obtaining the intrinsic mode functions IMFs of the original signal through MEMD, and expressing the IMFs in the following forms:
Figure FDA0002274627180000032
where s (t) is the EEG signal of n channels, ci(t) is an intrinsic mode function; r (t) is a margin function;
(3.2) using Hilbert transform to calculate instantaneous frequency and energy for short-time Fourier analysis; for the jth channel, the ith IMF (c)ij(t)) the hilbert transform is as follows:
Figure FDA0002274627180000041
wherein P represents a cauchy principal value;
(3.3) further deriving the corresponding instantaneous amplitude and frequency as follows:
aij(t)＝|cij(t)+jH(cij(t))|
Figure FDA0002274627180000042
wherein a isij(t) represents the instantaneous amplitude, θij(t) denotes the instantaneous phase, ωij(t) represents the instantaneous frequency and hence the Hilbert spectrum for each channel, e.g. for channel j at ω frequency at time t, it can be represented as Hj(ω,t)。
5. The monitoring method according to claim 2, wherein the step (4) specifically comprises:
(4.1) according to the determined time sequence of the micro-state, marking the obtained power spectrum of the brain electricity signal with different power spectrums, further obtaining respective Hilbert spectrums of different micro-states, and regarding the micro-state k, corresponding Hilbert power spectrum hk(ω) is:
Figure FDA0002274627180000044
wherein the total duration is Ls，TsThe time length corresponding to the micro state k is n is the total number of channels;
(4.2) obtaining power spectrums of different frequency bands corresponding to the micro-states according to the selected frequency range:
Figure FDA0002274627180000045
in the formula EkPower spectrum, omega, corresponding to the micro-state k0Is the selected frequency range.
CN201911118047.7A 2019-11-15 2019-11-15 Anesthesia depth monitoring system and method based on micro-state power spectrum analysis Pending CN110811557A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201911118047.7A CN110811557A (en) 2019-11-15 2019-11-15 Anesthesia depth monitoring system and method based on micro-state power spectrum analysis

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201911118047.7A CN110811557A (en) 2019-11-15 2019-11-15 Anesthesia depth monitoring system and method based on micro-state power spectrum analysis

Publications (1)

Publication Number Publication Date
CN110811557A true CN110811557A (en) 2020-02-21

Family

ID=69555547

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201911118047.7A Pending CN110811557A (en) 2019-11-15 2019-11-15 Anesthesia depth monitoring system and method based on micro-state power spectrum analysis

Country Status (1)

Country Link
CN (1) CN110811557A (en)

Cited By (4)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111407231A (en) * 2020-03-30 2020-07-14 河北省科学院应用数学研究所 Method and device for detecting risk of Alzheimer's disease and terminal equipment
CN111671420A (en) * 2020-06-17 2020-09-18 河北省科学院应用数学研究所 Method for extracting features from resting electroencephalogram data and terminal equipment
CN113367705A (en) * 2021-04-07 2021-09-10 西北工业大学 Motor imagery electroencephalogram signal classification method based on improved micro-state analysis
CN113974557A (en) * 2021-10-28 2022-01-28 中国人民解放军陆军军医大学第二附属医院 Deep neural network anesthesia depth analysis method based on electroencephalogram singular spectrum analysis

Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2003086188A2 (en) * 2002-04-15 2003-10-23 Instrumentarium Corporation Method and apparatus for determining the cerebral state of a patient with fast response
EP1495715A1 (en) * 2003-07-07 2005-01-12 Instrumentarium Corporation A method and apparatus based on combination of three phsysiological parameters for assessment of analgesia during anesthesia or sedation
CN109645989A (en) * 2018-12-10 2019-04-19 燕山大学 A kind of depth of anesthesia estimation method and system

Patent Citations (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2003086188A2 (en) * 2002-04-15 2003-10-23 Instrumentarium Corporation Method and apparatus for determining the cerebral state of a patient with fast response
EP1495715A1 (en) * 2003-07-07 2005-01-12 Instrumentarium Corporation A method and apparatus based on combination of three phsysiological parameters for assessment of analgesia during anesthesia or sedation
CN109645989A (en) * 2018-12-10 2019-04-19 燕山大学 A kind of depth of anesthesia estimation method and system

Non-Patent Citations (4)

* Cited by examiner, † Cited by third party
Title
ALLARD-SHI: "update version 1", 《GITHUB》 *
MAILLARD JULIEN ET AL.: "Alteration of temporal organization of EEG microstate sequences during propofol-induced loss of consciousness", 《SWISS MEDICAL WEEKLY》 *
MERCEDEH JAHANSEIR ET AL: "Automatic anesthesia depth staging using entropy measures and relative power of electroencephalogram frequency bands", 《AUSTRALASIAN PHYSICAL & ENGINEERING SCIENCES IN MEDICINE》 *
NORDEN E. HUANG ET AL: "A REVIEW ON HILBERT-HUANG TRANSFORM:METHOD AND ITS APPLICATIONS TO GEOPHYSICAL STUDIES", 《REVIEWS OF GEOPHYSICS》 *

Cited By (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN111407231A (en) * 2020-03-30 2020-07-14 河北省科学院应用数学研究所 Method and device for detecting risk of Alzheimer's disease and terminal equipment
CN111671420A (en) * 2020-06-17 2020-09-18 河北省科学院应用数学研究所 Method for extracting features from resting electroencephalogram data and terminal equipment
CN111671420B (en) * 2020-06-17 2023-07-18 河北省科学院应用数学研究所 Method for extracting features from resting state electroencephalogram data and terminal equipment
CN113367705A (en) * 2021-04-07 2021-09-10 西北工业大学 Motor imagery electroencephalogram signal classification method based on improved micro-state analysis
CN113974557A (en) * 2021-10-28 2022-01-28 中国人民解放军陆军军医大学第二附属医院 Deep neural network anesthesia depth analysis method based on electroencephalogram singular spectrum analysis

Similar Documents

Publication Publication Date Title
Durongbhan et al. A dementia classification framework using frequency and time-frequency features based on EEG signals
Cao et al. Extraction of SSVEPs-based inherent fuzzy entropy using a wearable headband EEG in migraine patients
Vaid et al. EEG signal analysis for BCI interface: A review
CN110811557A (en) Anesthesia depth monitoring system and method based on micro-state power spectrum analysis
US7373198B2 (en) Method and apparatus for the estimation of anesthetic depth using wavelet analysis of the electroencephalogram
Blanco et al. Unsupervised classification of high-frequency oscillations in human neocortical epilepsy and control patients
EP2498676B1 (en) Brain activity as a marker of disease
CN110811556A (en) Anesthesia depth monitoring system and method based on electroencephalogram micro-state analysis
Ong et al. Power spectral density analysis for human EEG-based biometric identification
US20160029946A1 (en) Wavelet analysis in neuro diagnostics
Hosseini et al. Emotional stress recognition system for affective computing based on bio-signals
US20060135880A1 (en) Identification of a dominant signal component in a biosignal
Podgorelec Analyzing EEG signals with machine learning for diagnosing Alzheimer’s disease
Tong et al. Emotion recognition based on photoplethysmogram and electroencephalogram
Kim et al. Wedea: A new eeg-based framework for emotion recognition
Li et al. Single-channel selection for EEG-based emotion recognition using brain rhythm sequencing
Choi et al. Determination of the concentrated state using multiple EEG channels
Wen et al. Hybrid approach of eeg stress level classification using k-means clustering and support vector machine
Fernández-Soto et al. Neural correlates of phrase quadrature perception in harmonic rhythm: an EEG study using a brain–computer interface
Zhang et al. Decoding olfactory EEG signals for different odor stimuli identification using wavelet-spatial domain feature
Zainuddin et al. Performance of support vector machine in classifying EEG signal of dyslexic children using RBF kernel
Mahmud et al. Multivariate models for decoding hearing impairment using EEG gamma-band power spectral density
Suhail et al. Distinguishing cognitive states using electroencephalography local activation and functional connectivity patterns
Suhail et al. Electroencephalography based detection of cognitive state during learning tasks: An extensive approach.
Sushkova et al. An investigation of the specificity of features of early stages of Parkinson’s disease obtained using the method of cortex electrical activity analysis based on wave trains

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
WD01 Invention patent application deemed withdrawn after publication

Application publication date: 20200221

WD01 Invention patent application deemed withdrawn after publication