A human state evaluation method based on network connection analysis

Abstract

A human state evaluation method based on the method comprises the steps of obtaining an electroencephalogram signal; electroencephalography and reconstruction, calculating the activity distribution of each brain area according to Brodmann brain partition to generate source space electroencephalogram signal; in all time domain, the unbiased weighted phase delay index connection matrix of different samples of all people and all channel pairs is calculated; the permutation test between the two groups of people is used to find out the frequency band with significant difference and the corresponding unbiased weighted phase delay index connection matrix; the brain network parameters of the sensor space electroencephalogram signal and the source space electroencephalogram signal are calculated under different sparseness, and the area under the curve of the corresponding parameters is calculated as a feature according to the sparseness; the feature is input into a support vector machine model to complete the distinction of easy anesthesia population and non-easy anesthesia population; the present application has high calculation efficiency, improves the precision, and can realize prediction only by using the data in the awake state during testing.

Description

Technical Field

[0001] This invention belongs to the field of machine learning and pattern recognition technology, and specifically relates to a human state assessment method based on network connectivity analysis for accurate brain monitoring during anesthesia. Background Technology

[0002] Electroencephalogram (EEG) is a method of recording brain activity using electrophysiological indicators. It is formed by summing the postsynaptic potentials that occur synchronously among a large number of neurons during brain activity. It records the changes in electrical waves during brain activity and is a comprehensive reflection of the electrophysiological activity of brain nerve cells on the surface of the cerebral cortex or scalp.

[0003] Brain state pattern recognition algorithms based on EEG mainly involve feature extraction and classification of EEG signals. Currently, algorithms for EEG signal feature extraction mainly include time-domain, frequency-domain, and time-frequency analysis methods; while classifiers include linear discriminant analysis and support vector machines. In recent years, a prominent highlight in machine learning has been the introduction of Convolutional Neural Networks (CNNs), especially in computer vision tasks, where they have significantly improved recognition accuracy compared to traditional machine learning methods.

[0004] The existing patent application, "A Method for Analyzing Brain Functional Connectivity Networks in Mathematically Gifted Adolescents Based on Resting-State EEG Signals" (patent application number: CN201711385303.X), describes a method for analyzing brain functional connectivity networks in mathematically gifted adolescents based on resting-state EEG signals. This method mainly includes: constructing a connectivity network based on resting-state EEG signals using a phase lag algorithm based on the Wilcoxon signed-rank test (PLWT); and then using graph theory metrics to measure the connectivity network and characterize the connectivity network features of mathematically gifted adolescents, including active brain regions and graph theory metrics. However, this method has some drawbacks. First, it fails to extract significantly different EEG signal frequency bands. The long frequency bands of EEG signals mean that full-band input analysis may result in redundancy of effective information and an overabundance of ineffective information. Second, when using the phase lag algorithm based on the Wilcoxon signed-rank test for brain network analysis, it does not perform brain power imaging, failing to consider the mutual influence between channel pairs across all channels, resulting in a discrepancy between the constructed brain network connectivity and the actual spatial reality. Summary of the Invention

[0005] To overcome the shortcomings of the prior art, the present invention aims to provide a human condition assessment method based on [the present invention name]. By constructing brain networks with different sparsity, a machine learning model is established based on multi-feature data for classification, which realizes the classification of two groups of people: those who are easily anesthetized and those who are not. The method has high computational efficiency and improved accuracy. At the same time, the test only uses data in the conscious state, and the prediction can be achieved by observing the classification effect of samples in the conscious state.

[0006] To achieve the above objectives, the technical solution of the present invention is as follows:

[0007] A method for assessing human condition based on network connectivity analysis includes the following steps:

[0008] Step 1: Acquire multi-channel EEG signals from the sensor space and preprocess the acquired EEG signals in the conscious state;

[0009] Step 2: Brain power source imaging and reconstruction, which involves calculating the brain electrical signals in the sensor space to obtain their mapping in the brain power source space; and calculating the activity distribution of each brain region according to the Brodmann brain partitions to perform brain power source reconstruction and generate source space brain electrical signals.

[0010] Step 3: Calculate the unbiased weighted phase delay exponential connectivity matrix for all channels of all people with respect to different samples in all time domains; use the permutation test between the two groups of people to find the frequency bands with significant differences and the corresponding unbiased weighted phase delay exponential connectivity matrices.

[0011] Step 4: Based on the unbiased weighted phase delay exponential connection matrix of the frequency bands with significant differences, calculate the brain network parameters of the sensor space EEG signal and the source space EEG signal under different sparsities, and calculate the area under the curve (AUC) of the corresponding parameters according to the sparsity as the feature.

[0012] Step 5: Establish a support vector machine model. Use the features calculated in step 4 as the feature input of the support vector machine model. The support vector machine model uses a Gaussian kernel to distinguish between people who are easily anesthetized and those who are not, thus realizing the preoperative assessment of human condition in clinical practice.

[0013] In step one, during the acquisition of multi-channel EEG signals in the sensor space, the acquired multi-channel EEG signals include two segments: one from the conscious state and one from the moderately anesthetized state. The sensor space EEG signal from the moderately anesthetized state is only used to determine whether the subject is susceptible to anesthesia, while the EEG signal from the conscious state is used as data for subsequent analysis. Therefore, the acquired sensor space conscious state multi-channel EEG signal is represented as follows: Where n represents whether the subject is susceptible to anesthesia, and m is the number of the Noc-guided EEG signal collected from the m-th subject during the experiment. Represents a real matrix. There is a Noc row and a t column, where t is the number of sampling points, and the sampling time is assumed to be t. t If the sampling frequency is F, then t = t t ×F.

[0014] Step one involves preprocessing the EEG signals in the conscious state, including bandpass filtering, period segmentation, baseline correction, and removal of eye movement or muscle artifacts. During bandpass filtering, the retained sensor spatial EEG signals are processed... All EEG signals from all channels were filtered using a bandpass filter of 0.5–45 Hz. In the epoch division, all EEG signals from all channels were divided into multiple equal-length epochs, each numbered as an epoch. e 'e' represents the period number, and the maximum value of 'e' depends on the number of periods divided, with a maximum value of 'e'. max Each resulting epoch is numbered accordingly. e The EEG signal relative to the entire EEG signal Baseline correction was performed using the average voltage; eye movement or muscle artifacts were removed by calculating the EEG signal. Normalized variance was used to identify channels and periods of anomalous noise, which were then manually rejected or retained through visual inspection. The preprocessed sensor spatial EEG data were then obtained.

[0015] The specific process of brain electrical activity imaging and reconstruction in step two is as follows:

[0016] (1) Solve the problem of EEG, that is, to establish a head model in the brain that can explain how the activity of neurons in the cerebral cortex is projected to the scalp through the brain volume effect and the conduction of the skull. Freesurferaverage based on magnetic resonance imaging is used as the anatomical brain template, and the default parameters of Brainstorms in the boundary element method in OpenMEEG are used to generate the head model of each participant. The mathematical model of scalp EEG signal generation is established as follows:

[0017]

[0018] The sensor EEG signal obtained after step one is represented; A represents the lead field, which is a zero-lead matrix that simulates the distribution of neurons in the cerebral cortex and the changes in their postsynaptic potentials; X represents the dipole moment of the cortical current dipole, which simulates the distribution of neurons in the cerebral cortex and the changes in their postsynaptic potentials; noise represents the measurement noise.

[0019] (2) The inverse problem is solved by using the minimum norm estimation method;

[0020]

[0021] L2 regularization is used to further constrain the solution in order to find the solution X that minimizes both the residual and energy of the source space. This additional constraint is represented by the second term in equation (2). The regularization parameter λ is obtained using the L-curve method to balance the influence of these two terms. In the case of L2 norm, the analytical solution of equation (2) yields the optimal estimate of the source distribution X, as shown in (3).

[0022]

[0023] When W is only the identity matrix, this solution Known as Minimum-norm Estimation (MNE), it is the mapping of EEG signals in the sensor space to the brain power space. T This represents the flip matrix of the leading field.

[0024] In step two, the activity distribution of each brain region is calculated based on the Brodmann brain regions, and brain power source reconstruction is performed to generate source spatial EEG signals. Specifically, the number of dipoles is classified according to the Brodmann brain regions, and the average value of the measurements of dipoles in the same region at the same time is taken as the output of the entire brain region, thereby obtaining the source distribution matrix of a single subject. That is, the EEG signal mapped from the sensor space to the source space, N c Represents EEG signals from sensors The number of channels mapped into the source space, i.e. the number of regions of interest after being divided according to the Brodmann partition; T represents the number of sampling points.

[0025] The unbiased weighted phase delay exponential connection matrix in step three is calculated as follows:

[0026] x and y are The two different rows of EEG signals represent different channels of the source space EEG signals. For different epochs of the two different channels of EEG signals x and y... e The sample segments are represented as signals respectively. and It then performs time-frequency analysis, sets the time bin length, and generates a time bin number. ft Set the frequency bin length, generate the frequency bin, and number it bin. fi Through Fourier transform signal and Multiplication yields the cross spectrum in express The complex conjugate of the cross spectrum is calculated using the same epoch of signals x and y. e The same bin ft The values ​​are fragments that correspond completely on the time scale. The phase lag index (PLI) of signals x and y from different channels is defined by the following formula:

[0027]

[0028] Represents cross spectrum The imaginary part of , sign is the sign function, and Ε represents solving all bins. ft The mean, Indicates to Take the absolute value;

[0029] Since PLI has a positive bias, the formula for calculating the unbiased phase delay index is:

[0030]

[0031] Index e j and e k Representing different bins ft The time period number; the weighted phase lag index (WPLI) is less sensitive to noise than the phase lag index and is defined as:

[0032]

[0033] Similar to PLI, DWPLI also exhibits a positive deviation. DWPLI is calculated as follows:

[0034]

[0035] In the formula, j and k represent bin ft The values ​​are accumulated in the time bins. When calculating DWPLI, it is required that each frequency bin be calculated within the same frequency bin. fi corresponding Different channels Combined, to generate matrix Elements in the matrix express The EEG signals of the i-th channel and the j-th channel in a certain epoch e bin fi This frequency is calculated within the chamber. n indicates whether the subject is susceptible to anesthesia, and m is the number of the m-th subject.

[0036] In step three, a permutation test between the two groups of people is used to identify frequency bands with significant differences. Specifically, this involves first solving the bin for each frequency band. fi of The mean of the matrix, binning the values ​​of the same type. fi Below Each element in Add and divide by N c ×(N c -1), to obtain any frequency bin of any subject segment sample. Each epoch e Get multiple Value; will use the same bin fi Value Form a sequence

[0037] For each sequence Perform a permutation test, with the number of permutations being times. Within each permutation, ... Randomly shuffle the sequence, but do not shuffle the labels. Calculate the mean value of the susceptible population for anesthesia based on the indices of the randomly shuffled sequence. forward Simultaneously, the mean value (nean) is calculated by taking the values ​​of the less easily anesthetized population from the randomly shuffled sequence. backpro minus timei timei represents the minus obtained from the i-th permutation. timei , all minus timei The MINUS sequence is obtained by sorting the elements in ascending order from smallest to largest. The MINUS sequence has a total of TIME elements.

[0038] Set the confidence coefficient `trust` and calculate the upper limit of the threshold. and threshold lower limit The calculation formula is as follows:

[0039]

[0040]

[0041] In the above formula, MINUS{floor[TIME×(trust / 2)]} means taking the element with index floor[TIME×(trust / 2)] in the MINUS sequence, where floor represents the floor function.

[0042] All warehouses belonging to the same category and frequency Take the average to get pass Obtain the Dec sequence, its elements Will and and When a comparison is performed, Greater than or less At that time, it was assumed that the number was bin fi The frequency bands corresponding to the frequency bins are those with significant differences. These significantly different frequency bands are extracted, and the frequency bins belonging to the frequency bins with significant differences are selected. The total number of selected frequency bins is [number missing].

[0043] Step four, which involves calculating the brain network parameters of the sensor spatial EEG signal and the source spatial EEG signal under different sparsities based on the unbiased weighted phase delay exponential connection matrix with significantly different frequency bands, specifically includes:

[0044] Select frequency bands whose corresponding frequency zones are significantly different. Selected multiple Corresponding element Find all epochs e The mean of the values ​​is calculated using the following formula:

[0045]

[0046] In equation (10), only the mean values ​​of frequency bands with significant differences are calculated. Combining to obtain For each subject, the corresponding adjacency matrix W is calculated using different brain connectivity sparse coefficients (0 ≤ sparse ≤ 1). sparse , by the adjacency matrix W sparse Obtain the degree matrix D sparse ;

[0047] Adjacency matrix W sparse In the calculation process, the brain connectivity sparsity coefficient is first determined, and then... All elements Extract the results sorted in ascending order from smallest to largest. The sequence is then used to determine the threshold, which is calculated using the following formula:

[0048]

[0049] In the above formula, Indicates taking The element index is floor[(1-sparse)×N] c ×N c ] elements, calculate Afterwards, All elements and In comparison, if the value is greater than W, it is considered that a connection exists. sparse middle The weight of the element at the corresponding position remains unchanged, and is If it is less than W, it is considered not connected. sparse middle The element at the corresponding position is set to 0, via W. sparse The sum of the elements in each row is used as D. sparse The values ​​of the diagonal elements in the corresponding row;

[0050] Solving for sensor spatial EEG signals adjacency matrix W sparse Sum degree matrix D sparse At this time, it is necessary to first divide the Fourier transform signal into different frequency bins, select the frequency bands with significant differences, and then, following the above steps, obtain the source spatial EEG data. The corresponding adjacency matrix W sparse Sum degree matrix D sparse Obtained using the same method;

[0051] For the obtained adjacency matrix W sparse Sum degree matrix D sparse Under different sparsity conditions, four different features are calculated: clustering coefficient, feature path length, modularity, and standard deviation of participation coefficient. These features are denoted as CC. sparse CP sparse Q sparse P sparse Different brain connectivity sparse coefficients were used to calculate different CC values. sparse CP sparse Q sparse P sparse .

[0052] In step four, during the calculation of the area under the curve (AUC) of the corresponding parameters based on sparsity, the brain connectivity sparsity coefficients are used as the horizontal axis. sparse CP sparse Q sparse P sparse Using the value of as the vertical axis, we treat the four different features as one-dimensional functions and solve for the area under the curve of the one-dimensional function as the feature.

[0053] Compared with the prior art, the present invention has the following advantages:

[0054] (1) In step one, the present invention divides the EEG signals of each subject into different samples, and then calculates and analyzes the samples, performs data augmentation, improves the robustness of the model, and uses the EEG signals in the waking state as subsequent processing.

[0055] (2) In step two, the present invention performs brain power spatial imaging based on the theoretical basis of neurophysiology and establishes source space EEG signals, which is beneficial to reduce the volume conduction effect during EEG signal transmission and noise interference during the acquisition process.

[0056] (3) In step three, DWPLI takes into account the influence of different frequencies on phase synchronization, corrects statistical bias, is robust to volume conduction, has a fast calculation speed, and is less sensitive to noise, which is advantageous in analyzing EEG signals.

[0057] (4) In step three, the permutation test is used to select the frequency bands with significant differences. Subsequent analysis is also based on the frequency bands with significant differences, which makes the model have more effective inputs and the analysis results more accurate.

[0058] (5) In step four, the features of network connections under different sparsities are integrated by calculating the area under the curve. Different features are combined as input to the model, which improves the comprehensive representation of network connections in the model and makes the model more generalizable.

[0059] (6) In step five, the present invention uses only the sensor spatial multi-channel EEG signal in the waking state to assess the human body state, without the need for any EEG signal in the anesthesia state, and only the normal resting EEG signal of the human body is needed to complete the assessment of the human body state.

[0060] In summary, this invention improves accuracy by constructing brain networks with different sparsity and establishing machine learning models based on multi-feature data to classify individuals into two groups: those who are easily anesthetized and those who are not. At the same time, it achieves preoperative clinical assessment of the human condition using only EEG signals in a conscious state. Attached Figure Description

[0061] Figure 1 This is a flowchart of the present invention.

[0062] Figure 2(a) is a diagram of the scalp EEG acquisition device, and Figure 2(b) is a diagram of the EEG channel distribution.

[0063] Figure 3 This is a schematic diagram of an experimental scheme for acquiring spatial EEG signals using sensors. Detailed Implementation

[0064] The process and advantages of the present invention will now be described in detail with reference to the accompanying drawings.

[0065] The specific experiments of this invention were conducted under the Windows 10 (64-bit) operating system. The overall algorithm flow was based on Matlab R2019b, and the brain power source localization was implemented using the brainst orm toolbox based on Matlab 2019b.

[0066] Figure 1 This is a brief overview of the entire algorithm, which consists of five modules: acquisition and extraction of spatial EEG signals from the sensor, brain power imaging and reconstruction, extraction of significantly different frequency bands, feature extraction, and SVM classifier. Figure 2(a) shows the scalp EEG acquisition device, and Figure 2(b) shows the distribution of EEG channels.

[0067] Experimental protocol for sensor-based spatial EEG signal acquisition as follows Figure 3 As shown, 22 neurologically healthy adults participated in data collection. Due to technical issues, data from two participants were unavailable, and data from the remaining 20 participants (9 males, 11 females, mean age = 30.85; standard deviation = 10.98) were analyzed. Participants were successively injected with 0.6 μg / ml (mild anesthesia) and 1.2 μg / ml (moderate anesthesia). EEG signals were collected simultaneously during the participants' conscious state and after anesthesia recovery. Behavioral tests were used to assess the participants' condition, and analysis was used to differentiate between those who were easily and difficult to anesthetize.

[0068] Based on the aforementioned sensor spatial EEG signal dataset, the following was adopted: Figure 1 The algorithm flow shown in this invention, a human body state assessment method based on network connectivity analysis, includes the following steps:

[0069] Step 1: Acquire multi-channel EEG signals from the sensor space and preprocess the acquired EEG signals in the conscious state;

[0070] In step one, during the acquisition of multi-channel EEG signals in the sensor space, the acquired multi-channel EEG signals include two segments: one from the conscious state and one from the moderately anesthetized state. The EEG signal from the moderately anesthetized state is used to distinguish whether the subject is susceptible to anesthesia. This invention uses only the conscious multi-channel EEG signal for subsequent analysis and classification; that is, the acquired conscious multi-channel EEG signal in the sensor space is represented as follows: Where m represents the subject number, and each subject's Noc-guided EEG signal was collected (128 channels of high-density EEG data were collected in the experiment, and data from 91 channels on the scalp surface were retained for further analysis; therefore, the maximum Noc value was 91). The EEG measurement unit is microvolt (μV), and the sampling frequency is F (unit: Hz). n indicates whether the subject is in a group susceptible to anesthesia (in the experiment, n=0 indicates that the subject is not easily anesthetized, and n=1 indicates that the subject is easily anesthetized). m is the Noc-guided EEG signal number collected from the m-th subject in the experiment. Represents a real matrix. There is a Noc row and a t column, where t is the number of sampling points, and the sampling time is assumed to be t. t If the sampling frequency is F, then t = t t ×F.

[0071] Step one involves preprocessing the EEG signals acquired in the conscious state, including bandpass filtering, sample segmentation, baseline correction, and removal of eye movement or muscle artifacts. In bandpass filtering, the retained sensor spatial EEG signals are processed... All EEG signals from all channels were filtered using a bandpass filter of 0.5–45 Hz. During sample partitioning, all EEG signals from all channels were divided into 10-second samples, each numbered as an epoch. e 'e' represents the period number, and the maximum value of 'e' depends on the number of periods divided, with a maximum value of 'e'. max Each resulting number is an epoch. e The EEG signal relative to the entire EEG signal The average voltage was baseline-corrected to remove eye movement or muscle artifacts. Abnormal noise channels and periods were identified by calculating their normalized variance. These were then manually rejected or retained through visual inspection. The preprocessed sensor spatial EEG data were then... The size of T is 10×F, meaning the period is divided into 10 seconds.

[0072] Step 2: Brain power source imaging and reconstruction, which involves calculating the brain electrical signals in the sensor space to obtain their mapping in the brain power source space; and calculating the activity distribution of each brain region according to the Brodmann brain partitions to perform brain power source reconstruction and generate source space brain electrical signals.

[0073] The specific process of brain electrical activity imaging and reconstruction in step two is as follows:

[0074] (1) Solving the problem of EEG, namely, establishing a head model that illustrates how the activity of neurons in the cerebral cortex is projected onto the scalp through the brain volume effect and the conduction through the skull. The main parts of the head model include the geometry of the brain and the conductivity characteristics of the different tissues (gray matter, white matter, cerebrospinal fluid, skull, and skin) through which neural electrical activity propagates. This invention uses the Freesurfer average based on magnetic resonance imaging (MRI) as the anatomical brain template and uses the Brainstorms default parameters of the Boundary Element Method (BEM) in OpenMEEG to generate a head model for each participant. The mathematical model for the generation of scalp EEG signals is established as follows:

[0075]

[0076] The sensor EEG signal obtained after step one is represented; A represents the lead field, which is a zero-lead matrix that simulates the distribution of neurons in the cerebral cortex and the changes in their postsynaptic potentials; X represents the dipole moment of the cortical current dipole, which simulates the distribution of neurons in the cerebral cortex and the changes in their postsynaptic potentials (source activity); noise represents the measurement noise.

[0077] (2) The inverse problem is solved by using the minimum norm estimation method.

[0078]

[0079] Theoretically, the optimal solution for X is the one that minimizes the residual between the estimate and the recorded scalp EEG in the L2 norm least squares sense. Considering that the number of dipoles is often greater than the number of electrodes, and many types of dipole distributions can produce similar results on the scalp, L2 regularization is used to further constrain the solution in order to find the solution X that minimizes both the residual and energy in the source space. This additional constraint is represented by the second term in (2); the regularization parameter λ is obtained using the L-curve method to balance the influence of these two terms. In the L2 norm case, the analytical solution of (2) yields the optimal estimate of the source distribution X, as shown in (3).

[0080]

[0081] When W is only the identity matrix, this solution This is known as the minimum-norm estimation (MNE). T This represents the flip matrix of the leading field.

[0082] In step two, the activity distribution of each brain region is calculated based on the Brodmann brain regions, and brain power source reconstruction is performed to generate source spatial EEG signals. Specifically, the number of dipoles is classified according to the Brodmann brain regions, and the average value of the measurements of dipoles in the same region at the same time is taken as the output of the entire brain region, thereby obtaining the source distribution matrix of a single subject. (i.e., the EEG signal mapped from the sensor space to the source space), N c Represents EEG signals from sensors The number of channels mapped into the source space (i.e., the number of regions of interest after being divided according to the Brodmann partition); T represents the number of sampling points, and the size of T is 10×F, that is, the length of the divided period is 10 seconds.

[0083] Step 3: Calculate the unbiased weighted phase delay exponential connectivity matrix for all channels of all people with respect to different samples in all time domains; use the permutation test between the two groups of people to find the frequency bands with significant differences and the corresponding unbiased weighted phase delay exponential connectivity matrices.

[0084] The debiased weighted phase lag index (DWPLI) connection matrix in step three is calculated as follows:

[0085] To perform segmented analysis of frequencies and to increase the sample size, two different channels of EEG signals x and y (x and y are...) were analyzed. The two different rows of EEG signals represent different epochs in different channels of the source space EEG signal. e Signal segmented by the sample and To perform time-frequency analysis, the first step is to... and Time-frequency decomposition was performed, with the time bin length set to 0.04 seconds. For a 10-second time bin... and 250 time bins were generated, numbered bin ft (The maximum value of ft in the experiment was 20×250×e) max Each subject's segmented sample can be divided into 250 time bins, and the number of frequency bins is set to 512, numbered bin. fi (In the experiment, there were 512 frequency bins, meaning the maximum fi value was 512). The Fourier transform signal was then analyzed. and Multiplication yields the cross spectrum in express The complex conjugate of the cross spectrum is used; note that the cross spectrum calculation uses the same epoch of signals x and y. eThe same bin ft The values ​​are fragments that correspond completely on the time scale. The phase lag index (PLI) of signals x and y from different channels is defined by the following formula:

[0086]

[0087] Represents cross spectrum The imaginary part of , sign is the sign function, and Ε represents solving all bins. ft The mean, Indicates to Take the absolute value.

[0088] PLI has a positive bias, meaning that the bin used in the calculation has a positive bias. ft When the ft value is small, the PLI is usually high. The formula for calculating the debiased phase lag index (dbPLI) is:

[0089]

[0090] Index e j and e k Representing different bins ft The time period number. The weighted phase lag index (WPLI) is less sensitive to noise than the phase lag index, and it is defined as:

[0091]

[0092] Similar to PLI, WPLI exhibits a positive deviation. DWPLI can be calculated as:

[0093]

[0094] Note that when calculating DWPLI, it is required that each frequency bin be calculated within the same frequency bin and the same time bin. fi bin at a certain time segment ft corresponding Different channels Combined, to generate matrix Elements in the matrix express The EEG signals of the i-th channel and the j-th channel in bin fi This frequency is within the warehouse and bin ft Calculated within this time segment (each in the experiment) Contains N c ×N c (element), each subject's epoch e Time segmentation bin under the sample ft Each has its own corresponding n represents whether the subject is in the easily anesthetized population (in the experiment, n=0 indicates that the subject belongs to the easily anesthetized population, and n=1 indicates that the subject belongs to the difficult-to-anesthetize population), and m is the number of the m-th subject.

[0095] In step three, a permutation test between the two groups is used to identify frequency bands with significant differences. Specifically, in the subsequent analysis using these specific frequency bands, it is first necessary to solve the bin of all time periods. ft Each frequency bin fi of The mean of a matrix needs to be calculated by binning matrices with the same values. fi Below Each element in Add and divide by N c ×N c Obtain any frequency bin of any subject segment sample Each time bin ft You can get 512. Value. (The same bin) fi Value Form a sequence (The order of elements in the sequence is not required; they are randomly generated. In this experiment, a total of 512 sequences were generated, each containing 20 × e^(-1 / 2). max The number of elements is because a total of 20 subjects were divided into e. max (samples)

[0096] For each Perform a permutation test, with the number of permutations being times. Within each permutation, ... Randomly shuffle the sequence and select the first N elements from the shuffled sequence based on their indices. SET Calculate the mean value forward At the same time, take the last N indices of the randomly shuffled sequence. SET Calculate the mean of the values. backpro minus timei timei represents the minus obtained from the i-th permutation. timei , all minus timei The MINUS sequence is obtained by sorting the elements in ascending order (from smallest to largest). The MINUS sequence has a total of TIME elements (in this experiment, the MINUS sequence has a total of 512 elements).

[0097] Set the confidence coefficient `trust` (0.05 in the experiment) and solve for the upper limit of the threshold. and threshold lower limit The calculation formula is as follows:

[0098]

[0099]

[0100] In the above formula, MINUS{floor[TIME×(trust / 2)]} means taking the element with index floor[TIME×(trust / 2)] in the MINUS sequence, where floor represents the floor function.

[0101] Bins belonging to the same category (with the same n value) and of the same frequency fi All (with the same value) Adding them together gives (A total of 512 × 2 were obtained in the experiment) Divided into two groups of people, and 512 frequency warehouses, through... Obtain the Dec sequence, its elements Will and and When a comparison is performed, Greater than or less At that time, it was assumed that the number was bin fi The frequency bands corresponding to the frequency bins are those with significant differences. This invention extracts these significantly different frequency bands and selects those frequency bins that belong to the frequency bins with significant differences. The total number of selected frequency bins is [number missing]. The frequency band with significant differences extracted from the source space EEG signal was 8.79-10.25 Hz, while the frequency band with significant differences extracted from the sensor space EEG signal was 8.30-10.74 Hz.

[0102] (The frequency bands with significant differences selected in the experiment are 8.30-10.74 Hz for EEG data in the conscious state, and 8.79-10.25 Hz for the source spatial EEG signals.)

[0103] Step 4: Based on the unbiased weighted phase delay exponential connectivity matrix of significantly different frequency bands, calculate the brain network parameters of the sensor spatial EEG signal and the source spatial EEG signal under different sparsities, and use the area under the curve (AUC) of the corresponding parameters as features according to the sparsity; the specific process is as follows:

[0104] Based on the unbiased weighted phase delay exponential connectivity matrix of significantly different frequency bands, brain network parameters of sensor spatial EEG signals and source spatial EEG signals are calculated under different sparsities. In the process of calculating the area under the curve (AUC) of the corresponding parameters according to sparsity as features, it is first necessary to select frequency bins whose corresponding frequency bands completely belong to the significantly different frequency bands. Selected multiple Corresponding element Find all epochs e The mean of the values ​​is calculated using the following formula:

[0105]

[0106] In the above formula, only the mean of the frequency bands with significant differences is calculated. Combining to obtain For each subject, the corresponding adjacency matrix W can be calculated for different brain connectivity sparse coefficients (0 ≤ sparse ≤ 1). sparse , by the adjacency matrix W sparse The degree matrix D can be obtained. sparse .

[0107] Adjacency matrix W sparse In the calculation process, the sparse coefficient of brain connectivity must first be determined, and then... All elements Extract the results sorted in ascending order (from smallest to largest). The sequence is then used to determine the threshold, which is calculated using the following formula:

[0108]

[0109] In the above formula, Indicates taking The element index is floor[(1-sparse)×N] c ×N c ] elements, calculate Afterwards, All elements and In comparison, if the value is greater than W, it is considered that a connection exists. sparse middle The weight of the element at the corresponding position remains unchanged, and is If it is less than W, it is considered not connected. sparse middle The element at the corresponding position is set to 0, via W. sparse The sum of the elements in each row is used as D. sparseThe value of the diagonal element of the corresponding row.

[0110] Solving for sensor spatial EEG signals adjacency matrix W sparse Sum degree matrix D sparse At this time, the Fourier transform signal needs to be divided into different frequency bands (see step three for details), and the frequency bands with significant differences need to be selected. Following the above steps, the sensor spatial EEG data... The corresponding adjacency matrix W sparse Sum degree matrix D sparse It can be obtained using the same method, except that the sensor spatial EEG data is... There are Noc channels, and there is a difference in the calculation formula (11).

[0111] For the obtained adjacency matrix W sparse Sum degree matrix D sparse Under different sparsity conditions, four different features are calculated: clustering coefficient, feature path length, modularity, and standard deviation of participation coefficient. These features are denoted as CC. sparse CP sparse Q sparse P sparse Different brain connectivity sparse coefficients were used to calculate different CC values. sparse CP sparse Q sparse P sparse .

[0112] In step four, during the process of calculating the area under the curve (AUC) of the corresponding parameters according to sparsity, the brain connectivity sparse coefficients are used as the horizontal axis. sparse CP sparse Q sparse P sparse Using the value as the vertical axis, the four different features can be regarded as one-dimensional functions. The area under the curve (AUC) of the one-dimensional function is calculated as the feature. A total of 8 features can be obtained from the sensor spatial EEG signal and the source spatial EEG signal.

[0113] Step 5: Establish a support vector machine (SVM) model to distinguish between patients who are easily anesthetized and those who are not, and to achieve preoperative clinical assessment of the patient's condition.

[0114] The specific process includes: using the features calculated in step four as feature input to the support vector machine (SVM) model. The SVM model uses a Gaussian kernel to calculate the human state of the sample, classifying the samples into easily anesthetized and difficult-to-anesthetize groups, thus achieving preoperative clinical human state assessment. The features calculated in step four are based on the EEG signal processing in the conscious state in step one; therefore, the training data only uses EEG signals in the conscious state to achieve human state assessment.

[0115] Accuracy evaluation for the support vector machine model:

[0116] This invention employs M-fold cross-validation to evaluate a proposed human state assessment method based on network connectivity analysis. Each test involves using all sample data from two subjects, ensuring that the two subjects belong to different population groups (different n values) as much as possible. This process is repeated M times, and the average recognition accuracy is calculated to evaluate the performance of the proposed algorithm.

[0117] The table below shows the accuracy obtained by ten-fold cross-validation on a test set of twenty subjects using the specific implementation method of this invention. The final average accuracy was 78.46%.

[0118] Accuracy of 10-fold cross-validation on a test set of 20 subjects

[0119]

Claims

1. A method for assessing human condition based on network connectivity analysis, characterized in that, Includes the following steps: Step 1: Acquire multi-channel EEG signals from the sensor space and preprocess the acquired EEG signals in the conscious state; Step 2: Brain power source imaging and reconstruction, which involves calculating the brain electrical signals in the sensor space to obtain their mapping in the brain power source space; and calculating the activity distribution of each brain region according to the Brodmann brain partitions to perform brain power source reconstruction and generate source space brain electrical signals. Step 3: Calculate the unbiased weighted phase delay index connectivity matrix for different samples of all people and all channels in all time domains; use the permutation test between the two groups of people to find the frequency bands with significant differences and the corresponding unbiased weighted phase delay index connectivity matrix. Step 4: Based on the unbiased weighted phase delay exponential connection matrix of the frequency bands with significant differences, calculate the brain network parameters of the sensor space EEG signal and the source space EEG signal under different sparsities, and calculate the area under the curve (AUC) of the corresponding parameters according to the sparsity as the feature. Step 5: Establish a support vector machine model. Use the features calculated in step 4 as the feature input of the support vector machine model. The support vector machine model uses a Gaussian kernel to distinguish between people who are easily anesthetized and those who are not, thus realizing the preoperative assessment of human condition in clinical practice.

2. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, In step one, during the acquisition of multi-channel EEG signals in the sensor space, the acquired multi-channel EEG signals include two segments: one from the conscious state and one from the moderately anesthetized state. The sensor space EEG signal from the moderately anesthetized state is only used to determine whether the subject is susceptible to anesthesia, while the EEG signal from the conscious state is used as data for subsequent analysis. Therefore, the acquired sensor space conscious state multi-channel EEG signal is represented as follows: Where n represents whether the subject is susceptible to anesthesia, and m is the number of the Noc-guided EEG signal collected from the m-th subject during the experiment. Represents a real matrix. There is a Noc row and a t column, where t is the number of sampling points, and the sampling time is assumed to be t. t If the sampling frequency is F, then t = t t ×F.

3. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, Step one involves preprocessing the EEG signals in the conscious state, including bandpass filtering, period segmentation, baseline correction, and removal of eye movement or muscle artifacts. During bandpass filtering, the retained sensor spatial EEG signals are processed... All EEG signals from all channels were filtered using a bandpass filter of 0.5–45 Hz. In the epoch division, all EEG signals from all channels were divided into multiple equal-length epochs, each numbered as an epoch. e 'e' represents the period number, and the maximum value of 'e' depends on the number of periods divided, with a maximum value of 'e'. max Each resulting epoch is numbered accordingly. e The EEG signal relative to the entire EEG signal Baseline correction was performed using the average voltage; eye movement or muscle artifacts were removed by calculating the EEG signal. Normalized variance was used to identify channels and periods of anomalous noise, which were then manually rejected or retained through visual inspection. The preprocessed sensor spatial EEG data were then obtained.

4. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, The specific process of brain electrical activity imaging and reconstruction in step two is as follows: (1) Solve the problem of EEG, that is, to establish a head model in the brain that can explain how the activity of neurons in the cerebral cortex is projected to the scalp through the brain volume effect and the conduction of the skull. Freesurferaverage based on magnetic resonance imaging is used as the anatomical brain template, and the default parameters of Brainstorms in the boundary element method in OpenMEEG are used to generate the head model of each participant. The mathematical model of scalp EEG signal generation is established as follows: The sensor EEG signal obtained after step one is represented; A represents the lead field, which is a zero-lead matrix that simulates the distribution of neurons in the cerebral cortex and the changes in their postsynaptic potentials; X represents the dipole moment of the cortical current dipole, which simulates the distribution of neurons in the cerebral cortex and the changes in their postsynaptic potentials; noise represents the measurement noise. (2) The inverse problem is solved by using the minimum norm estimation method; L2 regularization is used to further constrain the solution in order to find the solution X that minimizes both the residual and energy of the source space. This additional constraint is represented by the second term in equation (2). The regularization parameter λ is obtained using the L-curve method to balance the influence of these two terms. In the case of L2 norm, the analytical solution of equation (2) yields the optimal estimate of the source distribution X, as shown in (3). When W is only the identity matrix, this solution Known as Minimum-norm Estimation (MNE), it is the mapping of EEG signals in the sensor space to the brain power space. T This represents the flip matrix of the leading field.

5. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, In step two, the activity distribution of each brain region is calculated based on the Brodmann brain regions, and brain power source reconstruction is performed to generate source spatial EEG signals. Specifically, the number of dipoles is classified according to the Brodmann brain regions, and the average value of the measurements of dipoles in the same region at the same time is taken as the output of the entire brain region, thereby obtaining the source distribution matrix of a single subject. That is, the EEG signal mapped from the sensor space to the source space, N c Represents EEG signals from sensors The number of channels mapped into the source space, i.e. the number of regions of interest after being divided according to the Brodmann partition; T represents the number of sampling points.

6. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, The unbiased weighted phase delay exponential connection matrix in step three is calculated as follows: x and y are The two different rows of EEG signals represent different channels of the source space EEG signals. For different epochs of the two different channels of EEG signals x and y... e The sample segments are represented as signals respectively. and It then performs time-frequency analysis, sets the time bin length, and generates a time bin number. ft Set the frequency bin length, generate the frequency bin, and number it bin. fi Through Fourier transform signal and Multiplication yields the cross spectrum in express The complex conjugate of the cross spectrum is calculated using the same epoch of signals x and y. e The same bin ft The values ​​are fragments that correspond completely on the time scale. The phase lag index (PLI) of signals x and y from different channels is defined by the following formula: Represents cross spectrum The imaginary part of , sign is the sign function, and Ε represents solving all bins. ft The mean, Indicates to Take the absolute value; Since PLI has a positive bias, the formula for calculating the unbiased phase delay index is: Index e j and e k Representing different bins ft The time period number; the weighted phase lag index (WPLI) is less sensitive to noise than the phase lag index and is defined as: Similar to PLI, WPLI also exhibits a positive deviation. DWPLI is calculated as follows: In the formula, j and h represent bin ft The values ​​are accumulated in the time bins. When calculating DWPLI, it is required that each frequency bin be calculated within the same frequency bin. fi corresponding Different channels Combined, to generate matrix Elements in the matrix express The EEG signals of the i-th channel and the j-th channel in a certain epoch e bin fi This frequency is calculated within the chamber. n indicates whether the subject is susceptible to anesthesia, and m is the number of the m-th subject.

7. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, In step three, a permutation test between the two groups of people is used to identify frequency bands with significant differences. Specifically, this involves first solving the bin for each frequency band. fi of The mean of the matrix, binning the values ​​of the same type. fi Below Each element in Add and divide by N c ×(N c -1), to obtain any frequency bin of any subject segment sample. Each epoch e Get multiple Value; will use the same bin fi Value Form a sequence For each sequence Perform a permutation test, with the number of permutations being times. Within each permutation, ... Randomly shuffle the sequence, but do not shuffle the labels. Calculate the mean value of the susceptible population for anesthesia based on the indices of the randomly shuffled sequence. forward Simultaneously, the mean value is calculated by taking the values ​​of the less easily anesthetized population from the randomly shuffled sequence. backpro minus timei timei represents the minus obtained from the i-th permutation. timei , all minus timei The MINUS sequence is obtained by sorting the elements in ascending order from smallest to largest. The MINUS sequence has a total of TIME elements. Set the confidence coefficient `trust` and calculate the upper limit of the threshold. and threshold lower limit The calculation formula is as follows: In the above formula, MINUS{floor[TIME×(trust / 2)]} means taking the element with index floor[TIME×(trust / 2)] in the MINUS sequence, where floor represents the floor function. All warehouses belonging to the same category and frequency Take the average to get pass Obtain the Dec sequence, its elements Will and and When a comparison is performed, Greater than or less At that time, it was assumed that the number was bin fi The frequency bands corresponding to the frequency bins are those with significant differences. These significantly different frequency bands are extracted, and the frequency bins belonging to the frequency bins with significant differences are selected. The total number of selected frequency bins is [number missing].

8. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, In step four, the brain network parameters of the sensor spatial EEG signal and the source spatial EEG signal are calculated based on the unbiased weighted phase delay exponential connection matrix with significantly different frequency bands, under different sparsities. Specifically, the process is as follows: Select frequency bands whose corresponding frequency zones are significantly different. Selected multiple Corresponding element Find all epochs e The mean of the values ​​is calculated using the following formula: In equation (10), only the mean values ​​of frequency bands with significant differences are calculated. Combining to obtain For each subject, the corresponding adjacency matrix W is calculated using different brain connectivity sparse coefficients (0 ≤ sparse ≤ 1). sparse , by the adjacency matrix W sparse Obtain the degree matrix D sparse ; Adjacency matrix W sparse In the calculation process, the brain connectivity sparsity coefficient is first determined, and then... All elements Extract the results sorted in ascending order from smallest to largest. The sequence is then used to determine the threshold, which is calculated using the following formula: In the above formula, Indicates taking The element index is floor[(1-sparse)×N] c ×N c ] elements, calculate Afterwards, All elements and In comparison, if the value is greater than W, it is considered that a connection exists. sparse middle The weight of the element at the corresponding position remains unchanged, and is If it is less than W, it is considered not connected. sparse middle The element at the corresponding position is set to 0, via W. sparse The sum of the elements in each row is used as D. sparse The values ​​of the diagonal elements in the corresponding row; Solving for sensor spatial EEG signals adjacency matrix W sparse Sum degree matrix D sparse At this time, it is necessary to first divide the Fourier transform signal into different frequency bins, select the frequency bands with significant differences, and then, following the above steps, obtain the source spatial EEG data. The corresponding adjacency matrix W sparse Sum degree matrix D sparse Obtained using the same method; For the obtained adjacency matrix W sparse Sum degree matrix D sparse Under different sparsity conditions, four different features are calculated: clustering coefficient, feature path length, modularity, and standard deviation of participation coefficient. These features are denoted as CC. sparse CP sparse Q sparse P sparse， Different brain connectivity sparse coefficients are used to calculate different CC values. sparse CP sparse Q sparse P sparse .

9. The method for assessing human condition based on network connectivity analysis according to claim 1, characterized in that, In step four, during the process of calculating the area under the curve (AUC) of the corresponding parameters according to sparsity, the brain connectivity sparsity coefficient is used as the horizontal axis, and CC... sparse CP sparse Q sparse P sparse Using the value of as the vertical axis, we treat the four different features as one-dimensional functions and solve for the area under the curve of the one-dimensional function as the feature.

Images