Target crowd recognition method and device based on multi-level fusion of neural blood flow features

Abstract

The application relates to the technical field of brain function imaging, in particular to a target population recognition method and device based on multi-level fusion of neural blood flow characteristics, wherein the method comprises the following steps: acquiring brain neural blood flow signals of normal and target populations by using a synchronous EEG-fNIRS device, preprocessing and extracting single-mode features, calculating neural blood vessel coupling strength features by using a multi-frequency local neural blood vessel coupling analysis method, and unifying neural blood flow mode space information by using a spatial co-location method, so as to improve position accuracy, realize data layer fusion, integrating and analyzing spatial relationships of different mode features by using a Yeo7 brain template, realizing feature layer fusion, and finally training various machine learning classifiers by using significant difference features, so as to realize target population recognition. Therefore, the method solves the problem that related technologies pay more attention to macro signal correlation, leading to a lack of analysis on the synergistic effect of cross-modal spatial information in specific brain regions, and restricting in-depth understanding of neural blood vessel function mechanisms and revealing of brain region roles.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of brain function imaging, in particular to a target population recognition method and device based on multi-level fusion of neural blood flow characteristics. BACKGROUND

[0002] In the related art, EEG (Electroencephalogram) and fNIRS (Functional Near-Infrared Spectroscopy) are often used in combination to study the brain function of substance use disorder subjects and observe the neural blood flow information of the subjects.

[0003] However, in the related art, the synchronous EEG-fNIRS signal acquisition device has the problems of large and complex structure, increased difficulty of acquisition due to mutual interference between the electrodes and the optical fiber probe, and difficulty in signal integration. When the signal is used for research, although the two technologies can provide high-resolution brain activity information, there is a lack of analysis of the relationship between multi-modal characteristics and the specific location of neural vascular coupling, and only single-modal characteristics are analyzed without discussing the interaction mechanism, which leads to a lack of understanding of brain function and the potential of information fusion is not fully explored.

[0004] In summary, since the related art focuses more on the macro-level signal correlation, the analysis of how spatial information across different modalities cooperates in a specific brain region is still relatively scarce, which limits the understanding of the higher-level mechanism of neural vascular function and fails to fully reveal the unique role of different brain regions in information processing and functional regulation, which needs to be improved. SUMMARY

[0005] The present application provides a target population recognition method and device based on multi-level fusion of neural blood flow characteristics to solve the problem that the related art focuses more on macro signal correlation, leading to a lack of analysis of the cooperative effect of cross-modal spatial information in a specific brain region, which restricts the in-depth understanding of the mechanism of neural vascular function and the revelation of the role of brain regions.

[0006] The first aspect embodiment of the present application provides a target population recognition method based on multi-level fusion of neural blood flow characteristics, comprising the following steps: based on a synchronous EEG-fNIRS acquisition device, acquiring brain neural blood flow signal data containing electroencephalogram signals and near-infrared signals of normal population and target population meeting preset conditions; aligning and preprocessing the brain neural blood flow signal data, and dividing the electroencephalogram signals into multiple electroencephalogram components; based on the preprocessed brain neural blood flow signal data, analyzing and extracting at least one single-modality homologous feature parameter, mapping the single-modality homologous feature parameter on a specific channel position to obtain a specific channel feature; based on the preprocessed brain neural blood flow signal data, analyzing and extracting at least one mixed-modality feature parameter, mapping the mixed-modality feature parameter on a specific channel position to obtain a specific channel feature, and realizing feature fusion at the data layer based on the at least one mixed-modality feature parameter; integrating the single-modality feature and the mixed-modality feature in position by using a Yeo 7 network, realizing fusion at the feature layer, and then analyzing the differences and connections of the single-modality feature and the mixed-modality feature on the same network, taking the homologous feature parameters of the target population and the normal population as independent samples, and discovering brain features and brain networks sensitive to the target population by using mean value hypothesis testing; training multiple machine learning classifiers by using the brain features sensitive to the target population to realize recognition of the target population.

[0007] Through the above technical solutions, the embodiments of the present application can obtain brain neural blood flow data by synchronously acquiring EEG and fNIRS signals, provide rich information for comprehensive analysis of brain function. Data alignment and preprocessing and electroencephalogram component division guarantee data quality and effectiveness of analysis. Extraction and corresponding fusion of single-modality and mixed-modality feature parameters not only deeply mine information of each modality itself, but also realize cross-modality information integration, which is helpful for more accurately describing brain activity. Integrating features by using a Yeo 7 network can analyze feature differences and connections at the network level, and discover target population sensitive features and networks. Finally, machine learning classifiers are used for target population recognition, which improves the accuracy and efficiency of recognition, and provides strong support for neuroscience research, disease diagnosis, etc.

[0008] Optionally, in an embodiment of the present application, the at least one single-modality feature parameter is extracted based on the preprocessed brain nerve blood flow signal data, comprising: segmenting the EGG time domain data according to a preset time, calculating the correlation between each EGG channel signal in each segment by linear time delay coherence, obtaining a functional connectivity matrix of a nerve modality, averaging the functional connectivity matrix of the nerve modality according to the EGG channel to obtain the functional connectivity strength on the specific EGG channel; calculating the correlation matrix between each fNIRS channel time domain data by Pearson correlation, and then nonlinearly correcting the correlation matrix to obtain a functional connectivity matrix of a blood flow modality, averaging the functional connectivity matrix of the blood flow modality according to the fNIRS channel to obtain the functional connectivity strength on the specific fNIRS channel; based on the functional connectivity matrix of the nerve modality and the functional connectivity matrix of the blood flow modality, setting sparsity by a preset extensive threshold, constructing a binary network according to the sparsity, calculating a plurality of graph theory parameters by a graph theory method, and calculating AUC values of the plurality of graph theory parameters to obtain stable feature estimation values.

[0009] Through the above technical solutions, the embodiment of the present application can accurately reflect the nerve activity at a specific position by segmenting the EGG time domain data according to a preset time and obtaining the functional connectivity matrix of the nerve modality by calculating the correlation using linear time delay coherence, and then averaging the functional connectivity strength on the specific channel; the blood flow activity at the specific position can be accurately grasped by calculating the functional connectivity matrix of the blood flow modality by using Pearson correlation and nonlinear correction processing of the fNIRS channel data and then averaging; the binary network is constructed based on the functional connectivity matrix of the two modalities, a plurality of graph theory parameters are calculated by using a certain threshold and a graph theory method, the brain function features can be comprehensively and deeply analyzed, and stable feature estimation values are obtained.

[0010] Optionally, in an embodiment of the present application, the at least one mixed-modality feature parameter is extracted based on the preprocessed brain nerve blood flow signal data, comprising: performing multi-modality EEG-fNIRS source estimation based on spatial positioning on the plurality of brain electrical components to obtain a cortical current source density; reconstructing the cortical current source density at a near-infrared channel position to convolve the current source density at the near-infrared channel position with a preset standard blood flow dynamics function to obtain an expected blood flow change; obtaining a real blood flow response according to the expected blood flow change and the near-infrared signal, calculating a standardized slope by using a general linear model to determine the multi-frequency band local neural vascular coupling strength.

[0011] By the technical solution, the embodiment of the application can perform multi-modal source estimation based on spatial positioning on multiple electroencephalogram components, can accurately obtain the cortical current source density, and provides an accurate basis for subsequent analysis. The current source density is reconstructed at the near-infrared channel position, is convolved with a standard hemodynamic function to obtain an expected blood flow change, and is compared with the near-infrared signal to obtain a real blood flow response, and a general linear model is used to calculate a normalized slope to determine the local neural vascular coupling strength, thereby realizing effective correlation and analysis of electroencephalogram signals and blood flow signals at a local position, and helping to deeply understand the relationship between brain neural activity and blood flow change.

[0012] Optionally, in an embodiment of the application, the calculation formula of the multi-band local neural vascular coupling strength is as follows:

[0013]

[0014] wherein S i(i＝1,2,3,4,5) represents the source activity of multiple (five) electroencephalogram waves, V EEG is the potential recorded by an EEG electrode, W is a weight matrix constructed according to an fNIRS activated region, L is a lead matrix (indicating the relationship between an electrode and a current source) of EEG, H is the current source density selected at the near-infrared channel position, is the convolution of the current source density and a standard hemodynamic response function, Y is the real hemodynamic change, β0 is the intercept of the regression model, and β1 is the regression coefficient (indicating the multi-band local neural vascular coupling strength).

[0015] By the technical solution, the embodiment of the application can comprehensively and systematically construct a quantitative relationship model between neural activity and hemodynamic change by explicitly defining multiple key variables. This accurate mathematical expression can accurately calculate the multi-band local neural vascular coupling strength, provides a solid theoretical basis and effective calculation tool for in-depth research on the brain neural vascular coupling mechanism, and helps to more carefully understand the influence of neural activity on blood flow change under different electroencephalogram wave states.

[0016] Optionally, in an embodiment of the present application, the integration of different modal features in position by using Yeo 7 network comprises: determining the specific spatial positions of the neural modal and blood flow modal channels by using a preset standard system, mapping the specific spatial positions to a unified MNI standard space by using a spatial registration method to obtain the MNI coordinates of each channel of each modality; dividing the MNI standard space into 7 brain networks by using Yeo 7 network, determining the brain network to which each electroencephalogram electrode and the near-infrared channel belongs according to the MNI coordinates of the electroencephalogram electrode and the near-infrared channel, and realizing the integration of the single modal features and the mixed modal features in spatial position; and obtaining the brain features with significant differences as brain features sensitive to the target population by using mean hypothesis testing on the normal population and the target population, and taking the brain network in which the modal brain features of the neural modal, the blood flow modal and the mixed modal are different as the brain network sensitive to the target population.

[0017] Through the above technical solutions, the embodiments of the present application can determine and unify the spatial positions of the modal channels by using a standard system and a spatial registration method, obtain MNI coordinates, and lay a foundation for accurate analysis. The brain networks are divided by using Yeo 7 network and the network to which the channels belong is determined, the spatial integration of the multi-modal features is realized, and the relationship and synergistic effect between different modalities are facilitated for comprehensive research. The brain features and brain networks sensitive to the target population are obtained by using mean hypothesis testing, which can effectively distinguish the normal population from the target population and help to deeply understand the brain function changes of the target population.

[0018] Optionally, in an embodiment of the present application, the training of multiple machine learning classifiers by using the brain features sensitive to the target population to realize the recognition of the target population comprises: training multiple stable binary classifiers by using the brain features sensitive to the target population by using multiple machine learning classification models and a k-fold cross-validation method, the binary classifiers being used for distinguishing the normal population from the target population; classifying and recognizing the target population by using the multiple stable binary classifiers to obtain multiple recognition results; and comprehensively judging the multiple recognition results to realize the recognition of the target population.

[0019] Through the above technical solutions, the embodiments of the present application can give full play to the advantages of different models by using multiple machine learning classification models, improve the accuracy and generalization ability of classification, effectively utilize limited data by using a k-fold cross-validation method, divide the training set and the validation set multiple times, test the parameter effect, ensure that the trained model is stable and reliable, and avoid overfitting or underfitting. The multiple stable binary classifiers are trained by using the brain features sensitive to the target population, the normal population and the target population can be distinguished from multiple angles, and the accuracy of recognition is enhanced. Finally, the recognition results of the multiple classifiers are combined for comprehensive judgment, which further improves the accuracy and reliability of the recognition of the target population, and provides strong support for the accurate identification of the target population.

[0020] The second aspect embodiment of the application provides a target population recognition device based on multi-level fusion of neural blood flow characteristics, comprising: a collection module configured to collect brain neural blood flow signal data of normal population and target population meeting a preset condition based on a synchronous EEG-fNIRS collection device, the brain neural blood flow signal data comprising electroencephalogram signals and near-infrared signals; a preprocessing module configured to align and preprocess the brain neural blood flow signal data, and divide the electroencephalogram signals into a plurality of electroencephalogram components; a single-modal feature extraction module configured to analyze and extract at least one single-modal similar feature parameter based on the preprocessed brain neural blood flow signal data, map the single-modal similar feature parameter on a specific channel position to obtain a specific channel feature; a mixed-modal feature extraction module configured to analyze and extract at least one mixed-modal feature parameter based on the preprocessed brain neural blood flow signal data, map the mixed-modal feature parameter on a specific channel position to obtain a specific channel feature, and realize feature fusion at a data layer based on the at least one mixed-modal feature parameter; a multi-modal feature integration module configured to integrate the single-modal features and the mixed-modal features in position by using a Yeo 7 network based on the at least one single-modal feature parameter and the at least one mixed-modal feature parameter, realize fusion at a feature layer, and then analyze the differences and connections of the single-modal features and the mixed-modal features on the same network, take similar feature parameters of the target population and the normal population as independent samples, and find brain features and brain networks sensitive to the target population by using mean value hypothesis testing; and a recognition module configured to train a plurality of machine learning classifiers by using the brain features sensitive to the target population to realize recognition of the target population.

[0021] Through the above technical solution, the embodiments of the application can obtain brain neural blood flow data by synchronously collecting EEG and fNIRS signals, provide rich information for comprehensive analysis of brain function. Data alignment and preprocessing and division of electroencephalogram components guarantee data quality and effectiveness of analysis. Extraction and corresponding fusion of single-modal and mixed-modal feature parameters not only deeply mine information of each modality itself, but also realize cross-modal information integration, which is helpful for more accurately describing brain activity. Integration of features by using a Yeo 7 network can analyze feature differences and connections at a network level, and find brain features and networks sensitive to the target population. Finally, the target population is recognized by using a machine learning classifier, which improves accuracy and efficiency of recognition, and provides strong support for neuroscientific research, disease diagnosis, and the like.

[0022] Optionally, in an embodiment of the present application, the single modality feature extraction module comprises: a first calculation unit configured to segment EGG time domain data according to a preset time segment, calculate the correlation between each EGG channel signal in each segment by linear time delay coherence, obtain a functional connectivity matrix of a neural modality, average the functional connectivity matrix of the neural modality according to the EGG channel to obtain the functional connectivity strength on a specific EGG channel; a second calculation unit configured to calculate the correlation matrix between each fNIRS channel time domain data by Pearson correlation, and then correct the correlation matrix nonlinearly to obtain a functional connectivity matrix of a blood flow modality, average the functional connectivity matrix of the blood flow modality according to the fNIRS channel to obtain the functional connectivity strength on a specific fNIRS channel; and a third calculation unit configured to set sparsity based on a preset extensive threshold value based on the functional connectivity matrix of the neural modality and the functional connectivity matrix of the blood flow modality, construct a binary network according to the sparsity, calculate a plurality of graph theory parameters by a graph theory method, and calculate AUC values of the plurality of graph theory parameters to obtain stable feature estimation values.

[0023] Through the above technical solutions, the embodiment of the present application can accurately reflect the neural activity at a specific position by segmenting EGG time domain data according to a preset time segment and calculating the correlation by linear time delay coherence to obtain a neural modality functional connectivity matrix, and then averaging the functional connectivity strength on a specific channel. The blood flow activity at a specific position can be accurately grasped by calculating the correlation matrix between fNIRS channel data by Pearson correlation and nonlinear correction. The brain function features can be comprehensively and deeply analyzed by constructing a binary network based on the functional connectivity matrix of the two modalities, calculating a plurality of graph theory parameters by a certain threshold and a graph theory method, and obtaining stable feature estimation values.

[0024] Optionally, in an embodiment of the present application, the mixed modality feature extraction module comprises: a source estimation unit configured to perform multi-modality EEG-fNIRS source estimation based on spatial positioning on the plurality of brain electrical components to obtain a cortical current source density; a reconstruction unit configured to reconstruct the cortical current source density at a near-infrared channel position to convolve the current source density at the near-infrared channel position with a preset standard blood flow dynamics function to obtain an expected blood flow change; and a fourth calculation unit configured to obtain a true blood flow response according to the expected blood flow change and the near-infrared signal, calculate a standardized slope by a general linear model to determine the multi-band local neurovascular coupling strength.

[0025] By the technical solution, the embodiment of the application can perform multi-modal source estimation based on spatial positioning on multiple electroencephalogram components, can accurately obtain the cortical current source density, and provides an accurate basis for subsequent analysis. The current source density is reconstructed at the near-infrared channel position, is convolved with a standard hemodynamic function to obtain an expected blood flow change, and is compared with the near-infrared signal to obtain a real blood flow response, and a general linear model is used to calculate a normalized slope to determine the local neural vascular coupling strength, thereby realizing effective correlation and analysis of electroencephalogram signals and blood flow signals at a local position, and helping to deeply understand the relationship between brain neural activity and blood flow change.

[0026] Optionally, in an embodiment of the application, the calculation formula of the multi-band local neural vascular coupling strength is as follows:

[0027]

[0028] wherein S i(i＝1,2,3,4,5) represents the source activity of multiple (five) electroencephalogram waves, V EEG is the potential recorded by the EEG electrode, W is a weight matrix constructed according to the fNIRS activated region, L is a lead matrix (indicating the relationship between the electrode and the current source) of the EEG, H is the current source density selected at the near-infrared channel position, is the convolution of the current source density and the standard hemodynamic response function, Y is the real hemodynamic change, β0 is the intercept of the regression model, and β1 is the regression coefficient (indicating the multi-band local neural vascular coupling strength).

[0029] By the technical solution, the embodiment of the application can comprehensively and systematically construct a quantitative relationship model between neural activity and hemodynamic change by explicitly defining multiple key variables. This accurate mathematical expression can accurately calculate the multi-band local neural vascular coupling strength, provides a solid theoretical basis and effective calculation tool for in-depth research on the brain neural vascular coupling mechanism, and helps to more carefully understand the influence of neural activity on blood flow change under different electroencephalogram wave states.

[0030] Optionally, in an embodiment of the present application, the multi-modal feature integration module comprises: a mapping unit configured to determine specific spatial positions of the neural modality and the blood flow modality channels by using a preset standard system, and map the specific spatial positions to a unified MNI standard space by using a spatial registration method to obtain MNI coordinates of each channel of each modality; an integration unit configured to divide the MNI standard space into 7 brain networks by using a Yeo 7 network, determine the brain network to which each electroencephalogram electrode and the near-infrared channel belongs according to the MNI coordinates of the electroencephalogram electrode and the near-infrared channel, and realize integration of the single-modality features and the mixed-modality features in the spatial positions; and a test unit configured to perform mean hypothesis testing on the normal population and the target population to obtain brain features with significant differences as brain features sensitive to the target population, and obtain brain networks in which the brain features of the neural modality, the blood flow modality and the mixed modality are different as brain networks sensitive to the target population.

[0031] Through the above technical solutions, the embodiments of the present application can determine and unify the spatial positions of the modality channels by using a standard system and a spatial registration method, obtain MNI coordinates, and lay a foundation for accurate analysis. The brain networks are divided by using a Yeo 7 network and the network to which each channel belongs is determined, the multi-modal feature spatial integration is realized, and comprehensive research on the relationship and synergistic effect between different modalities is facilitated. The mean hypothesis testing is used to obtain brain features and brain networks sensitive to the target population, which can effectively distinguish the normal population from the target population and help to deeply understand the brain function changes of the target population.

[0032] Optionally, in an embodiment of the present application, the recognition module comprises: a plurality of stable binary classifiers are trained by using the brain features sensitive to the target population by using a plurality of machine learning classification models and a k-fold cross-validation method, and are used to distinguish the normal population from the target population; the target population is classified and recognized by using the plurality of stable binary classifiers to obtain a plurality of recognition results; and the target population is recognized by comprehensively integrating the plurality of recognition results.

[0033] Through the above technical solutions, the embodiments of the present application can give full play to the advantages of different models by using a plurality of machine learning classification models, improve the accuracy and generalization ability of classification, effectively use limited data by using a k-fold cross-validation method, divide the training set and the validation set multiple times, test the parameter effect, ensure that the trained model is stable and reliable, and avoid overfitting or underfitting. The plurality of stable binary classifiers are trained by using the brain features sensitive to the target population, the normal population and the target population can be distinguished from multiple angles, and the accuracy of recognition is enhanced. Finally, the recognition results of the plurality of classifiers are combined and comprehensively judged, which further improves the accuracy and reliability of the recognition of the target population, and provides strong support for accurate identification of the target population.

[0034] The third aspect of the application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the target population identification method based on multi-level fusion of neural blood flow features as described in the above embodiments.

[0035] The fourth aspect of the application provides a computer readable storage medium storing a computer program, which is executed by a processor to implement the target population identification method based on multi-level fusion of neural blood flow features as described above.

[0036] The fifth aspect of the application provides a computer program product comprising a computer program, which is executed to implement the target population identification method based on multi-level fusion of neural blood flow features as described above.

[0037] The embodiments of the application can obtain rich brain neural blood flow data by synchronously collecting EEG and fNIRS signals, guarantee data quality and analysis effectiveness through data alignment, preprocessing and EEG component division; obtain functional connection matrix and calculate related parameters by using linear time delay coherence, Pearson correlation and non-linear correction to process two kinds of modal data respectively, construct binary network calculation graph to calculate AUC value, and comprehensively and deeply analyze brain function characteristics; perform multi-modal source estimation based on spatial positioning on EEG components, reconstruct current source density and standard function convolution to obtain expected blood flow change, calculate neural vascular coupling strength by comparing real blood flow response, and deeply understand the relationship between neural and blood flow; construct a quantitative relationship model by determining variables to accurately calculate coupling strength; use multiple machine learning classification models combined with k-fold cross-validation, train multiple stable binary classifiers using sensitive brain features, jointly determine the results, and improve the accuracy and reliability of target population identification, thereby providing strong support for neural science research, disease diagnosis and the like.

[0038] Additional aspects and advantages of the application will be in part apparent and in part pointed out hereinafter. BRIEF DESCRIPTION OF DRAWINGS

[0039] The above and / or additional aspects and advantages of the application will become apparent and be readily appreciated from the following description, including the appended drawings, wherein:

[0040] Figure 1 A flowchart of a target population identification method based on multi-level fusion of neural blood flow features according to an embodiment of the application is shown in FIG. 1;

[0041] Figure 2 A schematic diagram of a multi-band local neural vascular coupling analysis method according to an embodiment of the application is shown in FIG. 2;

[0042] Figure 3 A schematic diagram of a method for classifying target people and normal people using machine learning according to an embodiment of the present application;

[0043] Figure 4 A structural schematic diagram of a target people recognition device based on multi-level fusion of neural blood flow features according to an embodiment of the present application.

[0044] Figure 5 A structural example diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0045] Embodiments of the present application are described in detail below with reference to the accompanying drawings, in which the same or similar components or components having the same or similar functions are denoted by the same or similar reference numerals throughout. The embodiments described below by reference to the accompanying drawings are exemplary and are intended to explain the present application, and cannot be understood as limiting the present application.

[0046] A target people recognition method and device based on multi-level fusion of neural blood flow features according to an embodiment of the present application are described below with reference to the accompanying drawings. In view of the problem that the related art mentioned in the background art focuses on macro signal correlation, leading to lack of analysis of cross-modal spatial information in specific brain regions, which restricts in-depth understanding of neural vascular function mechanism and revealing of brain region roles, the present application provides a target people recognition method based on multi-level fusion of neural blood flow features. In the method, brain neural blood flow data can be obtained by synchronously collecting EEG and fNIRS signals, providing rich information for comprehensive analysis of brain function. Data alignment and preprocessing and electroencephalogram component division ensure data quality and effectiveness of analysis. Extraction of single-modal and mixed-modal feature parameters and corresponding fusion not only deeply mines information of each modality, but also realizes integration of cross-modal information, which helps to more accurately depict brain activity. Integration of features using Yeo 7 network can analyze feature differences and relationships from the network level, and discover sensitive features and networks of target people. Finally, a machine learning classifier is used for target people recognition, improving the accuracy and efficiency of recognition, and providing strong support for neuroscience research, disease diagnosis, etc. Thus, the problem that the related art focuses on macro signal correlation, leading to lack of analysis of cross-modal spatial information in specific brain regions, which restricts in-depth understanding of neural vascular function mechanism and revealing of brain region roles, is solved.

[0047] Specifically, Figure 1 A flowchart of a target people recognition method based on multi-level fusion of neural blood flow features according to an embodiment of the present application.

[0048] As Figure 1As shown, the target population recognition method based on the multi-level fusion of neural blood flow characteristics includes the following steps:

[0049] In step S101, based on a synchronous EEG-fNIRS acquisition device, brain neural blood flow signal data containing electroencephalogram signals and near-infrared signals of normal people and target people meeting preset conditions are collected.

[0050] It can be understood that EEG and fNIRS are non-invasive, flexible, compact, user-friendly, easy to access, EEG can provide high time resolution neural electrical activity information, and is particularly suitable for studying brain transient response and neural dynamic process; and fNIRS can provide higher spatial resolution hemodynamic information, and has greater operation flexibility and lower equipment cost compared with functional magnetic resonance imaging and other methods.

[0051] Through the assembly of EEG electrodes and fNIRS light sources and receivers in hardware and the design of multi-modal synchronous signal acquisition software, a synchronous EEG-fNIRS signal acquisition platform is built, wherein the acquisition software supports synchronous marking of two signals, and controls the double-modal signal acquisition error within 100ms, based on the synchronous EEG-fNIRS signal acquisition platform, resting state whole brain neural blood flow signal data is collected, and the brain neural blood flow signal data includes electroencephalogram signals and near-infrared signals.

[0052] The embodiments of the present application can realize the synchronous acquisition of two signals by carefully designing the hardware assembly and multi-modal synchronous signal acquisition software, and control the acquisition error within a very small range, effectively guarantee the accuracy and synchronism of the data, and lay a solid and reliable foundation for subsequent analysis and research based on the resting state whole brain neural blood flow signal data.

[0053] In step S102, the brain neural blood flow signal data is aligned and preprocessed, and the electroencephalogram signals are divided into multiple electroencephalogram components.

[0054] Specifically, the cerebral nerve blood flow signal data collected in step S101 is aligned and preprocessed, and the signals are aligned in time according to the synchronization labels obtained by the synchronization marking in step S101. Further, the electroencephalogram signal is preprocessed, and the preprocessing operation includes but is not limited to: downsampling, re-reference, band-pass filtering and notch filtering, ICA (Independent Component Analysis), frequency band analysis of the electroencephalogram signal, including the characteristics of five commonly used electroencephalogram signal components: delta (1-3 Hz), theta (3-8 Hz), alpha (8-13 Hz), beta (13-30 Hz), and gamma (30-40 Hz); the near-infrared signal is preprocessed, and the preprocessing operation includes: filtering, improved Beer-Lambert variation, TDDR (temporal Derivative Distribution Repair) head motion correction, and using the relatively obvious change of hemoglobin HbO in the near-infrared signal as the basis for analyzing the hemodynamic changes.

[0055] The embodiment of the present application can align and preprocess the cerebral nerve blood flow signal data, realize time alignment according to the synchronization label intercepted signal, ensure the accuracy and consistency of the data in the time dimension, and facilitate subsequent analysis. A series of preprocessing operations are carried out on the electroencephalogram signal, such as downsampling, re-reference, etc., and the electroencephalogram signal is divided into five components according to frequency, which optimizes the signal quality and helps to more carefully mine the information contained in the electroencephalogram signal. And the preprocessing means such as filtering, head motion correction and selecting hemoglobin HbO as the basis for analysis for the near-infrared signal can effectively improve the signal quality and accurately capture the hemodynamic changes, and overall provide a high-quality and reliable data basis for subsequent in-depth research on brain function and analysis of brain activity.

[0056] In step S103, based on the preprocessed cerebral nerve blood flow signal data, at least one single-modality same-class feature parameter is analyzed and extracted, and the single-modality same-class feature parameter is mapped to a specific channel position to obtain a specific channel feature.

[0057] It can be understood that the single-modality same-class feature parameter includes but is not limited to the functional connectivity and graph theory parameters of the analyzed and extracted nerve and blood flow signals.

[0058] Optionally, in an embodiment of the present application, based on the pre-processed brain nerve blood flow signal data, at least one single modality homologous feature parameter is analyzed and extracted, including: segmenting the EGG time domain data according to a preset time segmentation, calculating the correlation between each EGG channel signal in each segment by linear time delay coherence, obtaining the functional connection matrix of the nerve modality, averaging the functional connection matrix of the nerve modality according to the EGG channel, obtaining the functional connection strength on the specific EGG channel; calculating the correlation matrix between each fNIRS channel time domain data by Pearson correlation, and then nonlinearly correcting the correlation matrix to obtain the functional connection matrix of the blood flow modality, and averaging the functional connection matrix of the blood flow modality according to the fNIRS channel to obtain the functional connection strength on the specific fNIRS channel; based on the functional connection matrix of the nerve modality and the functional connection matrix of the blood flow modality, a preset extensive threshold is used to set the sparsity, a binary network is constructed according to the sparsity, and a plurality of graph theory parameters are calculated by a graph theory method, and the AUC value of the plurality of graph theory parameters is calculated to obtain a stable feature estimate value.

[0059] Specifically, the EEG time domain data is segmented according to ten seconds per segment, the correlation between each EEG channel signal in each segment is calculated by linear time delay coherence, and the influence of brain electrical signal volume conduction is reduced; the Pearson correlation between each fNIRS channel time domain data is calculated, and then nonlinearly corrected by Fisher's Z transformation to obtain the functional connection matrix of the two modalities respectively. Then, the functional connection matrix of the calculated nerve blood flow signal is averaged according to the channel to obtain the functional connection strength on the specific channel, so as to achieve the purpose of analyzing each specific brain region.

[0060] Further, the functional connection matrix calculated by using the time domain information of EEG and fNIRS is used to construct a binary network respectively, the nodes of the network are each electroencephalogram or near-infrared channel, and the edges of the network do not reach a consensus on the threshold, so an extensive threshold (0.05≤T≤0.5, interval is 0.05) is used to set the sparsity, each functional connection matrix obtains 11 binary networks, a plurality of graph theory parameters are calculated by a graph theory method, including but not limited to small world property, local efficiency, global efficiency, node efficiency, etc., and then the AUC (Area Under the Curve) value of each graph theory feature of each matrix is calculated to obtain a stable feature estimate value.

[0061] The embodiment of the application can comprehensively analyze the functional connection, graph theory parameters and other aspects of the neural and blood flow signals when extracting single-modality homologous feature parameters based on the preprocessed brain neural blood flow signal data. Specifically, the EEG time domain data is reasonably segmented and the correlation is calculated by linear time delay coherence, which can reduce the volume conduction effect of the electroencephalogram signal; the functional connection matrix is obtained by Pearson correlation and nonlinear correction of the fNIRS channel data, which is scientific and rigorous in operation. The specific channel intensity is obtained by averaging the functional connection matrix by channel, which is beneficial to the analysis of each brain region. The binary network is constructed and the sparsity is set by widely setting the threshold, and the various graph theory parameters are calculated by combining the graph theory method, and then the stable feature estimation value is obtained by calculating the AUC value, which can accurately describe the brain function characteristics from multiple dimensions, and provide reliable and valuable data support for subsequent in-depth research on brain activity, disease diagnosis and the like.

[0062] In step S104, at least one mixed modality feature parameter is analyzed and extracted based on the preprocessed brain neural blood flow signal data, the mixed modality feature parameter is mapped to a specific channel position to obtain a specific channel feature, and feature fusion at the data layer is realized based on the at least one mixed modality feature parameter.

[0063] Optionally, in an embodiment of the application, based on the preprocessed brain neural blood flow signal data, analyzing and extracting at least one mixed modality feature parameter comprises: performing multi-modality EEG-fNIRS source estimation based on spatial positioning on a plurality of electroencephalogram components to obtain cortical current source density; reconstructing the cortical current source density on the near-infrared channel position to convolve the current source density on the near-infrared channel position with a pre-set standard blood flow dynamics function to obtain expected blood flow changes; obtaining real blood flow response according to the expected blood flow changes and the near-infrared signal, and using a general linear model to calculate a normalized slope to determine multi-frequency local neural vascular coupling strength.

[0064] It should be noted that, using the preprocessed brain neural blood flow signal data, the time series needs to be further processed to calculate the multi-frequency local neural vascular coupling degree feature, as shown in Figure 2 .

[0065] In actual execution process, since the device is difficult to simultaneously collect neural blood flow signals at the same position, the cortical current source density information is obtained by a multi-modality EEG-fNIRS source estimation method based on spatial co-positioning of a plurality of brain waves. Specifically, the activated area provided by the fNIRS signal is used as the prior spatial information H of the EEG to reduce the solution space of the EEG source estimation. Then, the EEG electric signal source positioning is combined with the blood flow signal of the fNIRS, and the two modal information is jointly estimated by using the Bayesian model to enhance the spatial accuracy of the source estimation. Among them, the spatial distribution of the fNIRS signal can be expressed as:

[0066] H = H1, H2,..., H n (1)

[0067] where H i represents the hemodynamic response corresponding to the i-th channel, and n is the number of fNIRS channels.

[0068] The EEG source estimation problem is usually modeled as a linear inverse problem as follows:

[0069] V EEG = LS (2)

[0070] where V EEG is the five brain waves (observation data), L is the lead matrix, and S is the source activity.

[0071] A weighted source estimation constraint matrix W is defined, where the active region determined by fNIRS will be given a higher weight, and the weight of other regions will be reduced:

[0072] W = diag(w1, w2,..., wn) (3) p

[0073] where if H i is greater than a threshold, w1 = 1; if H i is less than a threshold, w1 = 0.

[0074] In order to combine the blood flow signal of fNIRS and the electrical signal of EEG, Bayesian inference is used for multi-modal fusion. In the Bayesian framework, EEG source estimation is jointly estimated by prior information (fNIRS signal) and observation data (EEG signal). The posterior probability of source activity S can be expressed as:

[0075] P(S|V EEG ,H) = P(V EEG |S)P(S|H) (4)

[0076] where P(V EEG |S) is the likelihood function of EEG observation data, which is assumed to be Gaussian distributed:

[0077]

[0078] where, is the reconstruction error of source estimation, which represents the difference between EEG observation data and source estimation model. P(S|H) is the prior distribution provided by fNIRS, which represents the constraint of fNIRS activated region on EEG source activity. It can be modeled as:

[0079]

[0080] where,​ is the constraint provided by fNIRS, limiting the EEG source localization to the active region determined by fNIRS. Z1, Z2 are normalization constants, σ 2 and τ 2 are the observation error of EEG and the uncertainty of fNIRS prior, respectively. By maximizing the posterior probability P(S|V EEG , H), the optimal estimate of source activity S * is obtained, which is expressed as:

[0081]

[0082] Further, the scalp EEG component signals are mapped to multiple voxels in the cerebral cortex to obtain the current source density results on each voxel. Then, the current source density of the single voxel closest to the location of the near-infrared channel is selected as the reconstructed current source density to reduce the influence of the volume conduction of the EEG signal as much as possible. Secondly, the expected blood flow change is obtained by convolution with the standard blood flow dynamics function, and then the expected blood flow change signal and the real blood flow dynamics change obtained by analyzing the near-infrared signal are aligned in length and amplitude through downsampling and z-score normalization operations.

[0083] After the above processing of the data is completed, the five EEG component predicted blood flow signals and the real blood flow signals of each subject are respectively taken as the independent variable and the dependent variable into the general linear model (formula 1), and the standardized slope is the degree of influence of the dependent variable real blood flow change on the independent variable neural activity, that is, the degree of multi-frequency local neurovascular coupling. The calculation formula of the multi-frequency local neurovascular coupling strength is:

[0084]

[0085] where S i(i＝1,2,3,4,5) represents the source activity of multiple (five) EEG waves, V EEG is the potential recorded by the EEG electrode, W is the weight matrix constructed according to the fNIRS active region, L is the lead matrix of the EEG (indicating the relationship between the electrode and the current source), H is the current source density selected at the location of the near-infrared channel, is the convolution of the current source density and the standard blood flow dynamics response function, Y is the real blood flow dynamics change, β0 is the intercept of the regression model, and β1 is the regression coefficient (representing the multi-frequency local neurovascular coupling strength).

[0086] The embodiment of the application can extract mixed modality feature parameters based on the pre-processed brain nerve blood flow signal data and realize data layer feature fusion. Through the multi-modality EEG-fNIRS source estimation method based on spatial co-location of multiple brain waves, the fNIRS signal is ingeniously used to provide prior spatial information for EEG, the EEG source estimation solution space is reduced, and then the Bayesian model is used to jointly estimate the two modal information, thereby enhancing the spatial accuracy of source estimation. From calculating the cortical current source density to reconstructing, convolving and obtaining the expected blood flow change in the near-infrared channel position, and then comparing with the real blood flow response, the standardized slope is calculated by using the general linear model to determine the local neural vascular coupling strength in multiple frequency bands, thereby further quantifying the coupling strength, effectively fusing the EEG and blood flow signals in depth, and accurately depicting the correlation between the two.

[0087] In step S105, the single modality features and the mixed modality features are integrated in position by using the Yeo 7 network, the feature layer fusion is realized, and then the differences and the relationship between the single modality features and the mixed modality features on the same network are analyzed. The same type of feature parameters of the target population and the normal population are taken as independent samples, and the mean hypothesis test is used to find the sensitive brain features and brain networks of the target population.

[0088] After all the above steps are completed, that is, after all the features are extracted, the brain function of the target population that has changed significantly needs to be found. According to the MNI coordinates of each EEG and near-infrared channel, the corresponding Yeo 7 network is found, the features of the same network in different modalities are integrated by using the Yeo 7 network, and the brain features and brain networks that have changed significantly in the target population compared with the normal population, that is, the sensitive brain features of the target population, are determined.

[0089] Optionally, in an embodiment of the application, the Yeo 7 network is used to integrate different modality features in position, including: using a preset standard system to determine the specific spatial positions of the neural modality and the blood flow modality channels, mapping the specific spatial positions to a unified MNI standard space by using a spatial registration method to obtain the MNI coordinates of each channel of each modality; using the Yeo 7 network to divide the MNI standard space into 7 brain networks, determining the brain network to which each EEG electrode and near-infrared channel belongs according to the MNI coordinates of the EEG electrode and the near-infrared channel, and realizing the integration of the single modality features and the mixed modality features in spatial position; using the mean hypothesis test on the normal population and the target population to obtain the brain features with significant differences as the sensitive brain features of the target population, and taking the brain networks in which the brain features of the neural modality, the blood flow modality and the mixed modality appear differences as the sensitive brain networks of the target population.

[0090] Specifically, the standard system set is the 10-20 international standard lead system.

[0091] The following takes the etomidate use disorder as an example to identify the target, and the step S105 is described in detail.

[0092] Specifically, there are significant differences in age and education level between etomidate use disorderers and normal people. Linear regression is used to eliminate the differences in age and education level. After ensuring that it is not affected by other factors except the use of etomidate, independent sample T test is performed on each feature of the two groups. For multiple similar features, multiple correction is used to avoid false positive results. After completing the hypothesis test, because the features of different modalities cannot be corresponded in position (electroencephalogram single modality features: features on electroencephalogram channel position; near-infrared single modality features and neurovascular coupling features: features on near-infrared channel position), the two device channels are mapped into the same network according to the position by means of Yeo 7 network template, and the channels that appear significant changes and gather in the network are searched to find the Yeo 7 brain network that appears significant brain function changes. When the brain network that appears in the significant change feature set is found, it can be considered that the network is the sensitive network of the target population. For example, in this embodiment, the features of low-frequency electroencephalogram functional connectivity, graph theory and neurovascular coupling that appear significant changes are concentrated in the somatomotor network and the dorsal attention network, that is, the two Yeo 7 networks are etomidate sensitive networks, and etomidate has a significant impact on the functions of the two brain network regions.

[0093] The embodiments of the present application can integrate different modality features in position and fuse features based on single modality and mixed modality feature parameters by using Yeo 7 network, which can effectively break the analysis limitations caused by the difficulty of corresponding different modality features in position. By taking the similar feature parameters of the target population and the normal population as independent samples, using mean hypothesis test, and combining rigorous methods such as linear regression to eliminate the influence of other factors and multiple correction to avoid false positive results, the sensitive brain features and brain networks of the target population can be accurately found.

[0094] In step S106, a plurality of machine learning classifiers are trained by using the brain features sensitive to the target population to realize the identification of the target population.

[0095] It can be understood that after finding the brain features and brain networks sensitive to the target population, all significant difference features can be selected as machine learning parameters for training, or the difference features on the sensitive network can be selected as parameters for training, which can reduce the cost of identification. For example, in the embodiment of using etomidate in step S105 as the identification target, only the significant difference features on the somatomotor network and the dorsal attention network can be used as parameters to train the machine learning model, that is, in the subsequent targeted signal acquisition process, the device can be simplified, and only a few leads can be used to achieve the identification effect of analyzing all lead information.

[0096] Optionally, in an embodiment of the present application, a plurality of machine learning classifiers are trained using brain features sensitive to the target population to identify the target population, including: using a plurality of machine learning classification models and a k-fold cross-validation method, training a plurality of stable binary classifiers using brain features sensitive to the target population to distinguish between normal people and target people, combining the plurality of stable binary classifiers to classify and identify the target population to obtain a plurality of identification results, and integrating the plurality of identification results to identify the target population.

[0097] In the training process of the machine learning model, three machine learning classification models are used, including SVM (Support Vector Machine), RF (Bootstrap Aggregating), and XGBoost (Extreme Gradient Boosting). The target population sensitive brain features are used to adjust the parameters, train three stable binary classifiers of normal people and target people, find the optimal parameters under the specified feature quantity, and achieve the optimal classification effect.

[0098] Optionally, if the data set is limited and the data for training and verification is less, the k-fold cross-validation method can be used to divide the training set of the data into k subsets, and the three classifiers are repeatedly trained k times. Each time, one of the k subsets is used as the validation set, and the remaining k-1 subsets are used as the training set. The classification effect of 10 times of training is averaged as the classification effect of the current classifier and parameters. By dividing the training set and the validation set multiple times, the parameter effect is tested to obtain a stable model parameter.

[0099] After obtaining the stable classifier, the sample features to be identified are put into the three classifiers, and the classification results of the multiple classifiers are integrated to identify the category of the analysis sample. The machine learning classifier classification and identification process is as shown in Figure 3

[0100] ​The embodiments of the present application can train multiple machine learning classifiers using brain features sensitive to the target population to identify the target population. Both all significant difference features and difference features on sensitive networks can be selected as parameters to effectively reduce the identification cost, such as simplifying device information collection in related embodiments. Using multiple machine learning classification models such as SVM, RF, and XGBoost combined with k-fold cross-validation method can fully utilize the advantages of different models. In the case of limited data, the parameter effect can be tested by dividing the training set and the validation set multiple times to train a stable binary classifier. The classification results of multiple classifiers are combined for identification, which further improves the accuracy, stability and reliability of classification, and provides a strong guarantee for accurate identification of the target population.

[0101] The target population identification method based on multi-level fusion of neural blood flow features according to the embodiments of the present application can obtain brain neural blood flow data by synchronously collecting EEG and fNIRS signals to provide rich information for comprehensive analysis of brain function. Data alignment and preprocessing and EEG component division ensure data quality and effectiveness of analysis. Extraction of single-mode and mixed-mode feature parameters and corresponding fusion not only deeply excavates information of each mode itself, but also realizes integration of cross-modal information, which helps to more accurately depict brain activity. Yeo 7 network is used to integrate features to analyze feature differences and relationships at the network level and find sensitive features and networks of the target population. Finally, machine learning classifiers are used for target population identification to improve the accuracy and efficiency of identification and provide strong support for neuroscience research, disease diagnosis, etc.

[0102] Secondly, the target population identification device based on multi-level fusion of neural blood flow features according to the embodiments of the present application is described with reference to the accompanying drawings.

[0103] Figure 4 is a block schematic diagram of the target population identification device based on multi-level fusion of neural blood flow features according to the embodiments of the present application.

[0104] As shown in Figure 4 , the target population identification device based on multi-level fusion of neural blood flow features 10 includes an acquisition module 100, a preprocessing module 200, a single-mode feature extraction module 300, a mixed-mode feature extraction module 400, a multi-modal feature integration module 500, and an identification module 600.

[0105] Specifically, the acquisition module 100 is configured to acquire brain neural blood flow signal data containing EEG signals and near-infrared signals of normal population and target population satisfying a preset condition based on a synchronous EEG-fNIRS acquisition device.

[0106] The preprocessing module 200 is configured to align and preprocess the brain neural blood flow signal data, and divide the EEG signals into multiple EEG components.

[0107] The single-modal feature extraction module 300 is configured to analyze and extract at least one single-modal homogenous feature parameter based on the preprocessed cerebral neural blood flow signal data, and map the single-modal homogenous feature parameter to a specific channel position to obtain a specific channel feature.

[0108] The mixed-modal feature extraction module 400 is configured to analyze and extract at least one mixed-modal feature parameter based on the preprocessed cerebral neural blood flow signal data, map the mixed-modal feature parameter to a specific channel position to obtain a specific channel feature, and realize feature fusion at a data layer based on the at least one mixed-modal feature parameter.

[0109] The multi-modal feature integration module 500 integrates the single-modal feature and the mixed-modal feature in position by using a Yeo 7 network, realizes fusion at a feature layer, and further analyzes the differences and connections between the single-modal feature and the mixed-modal feature on the same network, takes the homogenous feature parameters of the target population and the normal population as independent samples, and finds the brain features and brain networks sensitive to the target population by using mean value hypothesis testing.

[0110] The recognition module 600 is configured to train a plurality of machine learning classifiers by using the brain features sensitive to the target population to realize recognition of the target population.

[0111] Optionally, in an embodiment of the present application, the single-modal feature extraction module 300 comprises a first calculation unit, a second calculation unit and a third calculation unit.

[0112] The first calculation unit is configured to segment the EGG time domain data according to a preset time, calculate the correlation between each EGG channel signal in each segment by linear time-lag coherence, obtain a functional connection matrix of a neural modality, average the functional connection matrix of the neural modality according to the EGG channel, and obtain the functional connection strength on a specific EGG channel.

[0113] The second calculation unit is configured to calculate the correlation matrix between each fNIRS channel time domain data by Pearson correlation, further correct the correlation matrix by nonlinearity, obtain a functional connection matrix of a blood flow modality, average the functional connection matrix of the blood flow modality according to the fNIRS channel, and obtain the functional connection strength on a specific fNIRS channel.

[0114] The third calculation unit is configured to set the sparsity by a preset extensive threshold based on the functional connection matrix of the neural modality and the functional connection matrix of the blood flow modality, construct a binary network according to the sparsity, calculate a plurality of graph theory parameters by a graph theory method, calculate the AUC value of the plurality of graph theory parameters, and obtain a stable feature estimation value.

[0115] Optionally, in an embodiment of the present application, the mixed modality feature extraction module 400 comprises a source estimation unit, a reconstruction unit and a fourth calculation unit.

[0116] The source estimation unit is configured to perform spatially located multi-modality EEG-fNIRS source estimation on the plurality of brain electrical components to obtain a cortical current source density.

[0117] The reconstruction unit is configured to reconstruct the cortical current source density at the near-infrared channel position, to convolve the current source density at the near-infrared channel position with a preset standard blood flow dynamics function, and to obtain an expected blood flow change.

[0118] The fourth calculation unit is configured to obtain a true blood flow response according to the expected blood flow change and the near-infrared signal, to calculate a normalized slope by using a general linear model, and to determine a multi-band local neurovascular coupling strength.

[0119] Optionally, in an embodiment of the present application, the calculation formula of the multi-band local neurovascular coupling strength is as follows:

[0120]

[0121] wherein S i(i＝1,2,3,4,5) represents source activity of a plurality of (five) brain waves, V EEG is a potential recorded by an EEG electrode, W is a weight matrix constructed according to an fNIRS activated region, L is a lead matrix (indicating the relationship between an electrode and a current source) of the EEG, H is a current source density selected at the near-infrared channel position, is a convolution of the current source density with a standard blood flow dynamics response function, Y is a true blood flow dynamics change, β0 is an intercept of a regression model, and β1 is a regression coefficient (indicating the multi-band local neurovascular coupling strength).

[0122] Optionally, in an embodiment of the present application, the multi-modality feature integration module 500 comprises a mapping unit, an integration unit and a verification unit.

[0123] The mapping unit is configured to determine specific spatial positions of a neural modality and a blood flow modality channel by using a preset standard system, to map the specific spatial positions to a unified MNI standard space by using a spatial registration method, and to obtain MNI coordinates of each channel of each modality.

[0124] The integration unit is configured to divide the MNI standard space into 7 brain networks by using a Yeo 7 network, to determine a brain network to which each EEG electrode and near-infrared channel belongs according to the MNI coordinates of the EEG electrode and the near-infrared channel, and to realize integration of single-modality features and mixed-modality features in spatial positions.

[0125] The test unit is used to test the normal population and the target population by using the mean hypothesis, obtain the brain features with significant differences as the brain features sensitive to the target population, and obtain the brain networks with differences in the brain features of the neural modalities, the blood flow modalities and the mixed modalities as the brain networks sensitive to the target population.

[0126] Optionally, in an embodiment of the present application, the identification module 600 comprises: using multiple machine learning classification models and a k-fold cross-validation method, training multiple stable binary classifiers using the brain features sensitive to the target population for distinguishing the normal population and the target population, jointly classifying the target population using the multiple stable binary classifiers to obtain multiple identification results, and comprehensively integrating the multiple identification results to realize identification of the target population.

[0127] It should be noted that the foregoing explanation and description of the embodiment of the target population identification method based on multi-level fusion of neural blood flow features is also applicable to the embodiment of the target population identification device based on multi-level fusion of neural blood flow features, which will not be described here.

[0128] The target population identification device based on multi-level fusion of neural blood flow features according to the embodiment of the present application can obtain brain neural blood flow data by synchronously collecting EEG and fNIRS signals, and provide rich information for comprehensive analysis of brain function. Data alignment and preprocessing and EEG component division ensure data quality and effectiveness of analysis. Extraction of single-modality and mixed-modality feature parameters and corresponding fusion not only deeply excavate information of each modality itself, but also realize integration of cross-modality information, which is helpful for more accurately describing brain activity. Integration of features using Yeo 7 network can analyze feature differences and connections from the network level, and find target population sensitive features and networks. Finally, the target population is identified by means of a machine learning classifier, which improves the accuracy and efficiency of identification, and provides strong support for neuroscience research, disease diagnosis and the like.

[0129] Figure 5 The electronic device provided in the embodiment of the present application has the structure shown in the structural schematic diagram of the electronic device. The electronic device can comprise:

[0130] The memory 501, the processor 502, and the computer program stored in the memory 501 and executable on the processor 502.

[0131] The processor 502 implements the target population identification method based on multi-level fusion of neural blood flow features provided in the above embodiments when executing the program.

[0132] Further, the electronic device further comprises:

[0133] The communication interface 503 is used for communication between the memory 501 and the processor 502.

[0134] The memory 501 is configured to store a computer program capable of being executed on the processor 502.

[0135] The memory 501 can include a high-speed RAM memory, and can further include a non-volatile memory, for example, at least one disk memory.

[0136] If the memory 501, the processor 502 and the communication interface 503 are independently implemented, the communication interface 503, the memory 501 and the processor 502 can be connected to each other through a bus and complete communication between each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For convenience of representation, Figure 5 Only one thick line is used in the figure to represent the bus, but it does not mean that there is only one bus or only one type of bus.

[0137] Optionally, in a specific implementation, if the memory 501, the processor 502 and the communication interface 503 are integrated on a chip, the memory 501, the processor 502 and the communication interface 503 can complete communication between each other through an internal interface.

[0138] The processor 502 can be a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of the present application.

[0139] The embodiments of the present application further provide a computer readable storage medium, which has a computer program stored thereon, and the program is executed by a processor to implement the target population recognition method based on multi-level fusion of neural blood flow features as described above.

[0140] The embodiments of the present application further provide a computer program product, which includes a computer program, and the computer program is executed to implement the target population recognition method based on multi-level fusion of neural blood flow features as described above.

[0141] In the description of the application, reference to "one embodiment", "some embodiments", "an example", "a specific example", or "some examples" means that a particular feature, structure, material, or characteristic being described is included in at least one embodiment or example of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment or example. Furthermore, the described specific features, structures, materials, or characteristics can be combined in any suitable manner in one or more embodiments or examples. In addition, the usage of "N" means at least two, for example, two, three or the like, unless explicitly stated otherwise.

[0142] Furthermore, the terms "first", "second", or the like, are used merely as a designation of certain elements or features, and do not imply or connote relative importance or a specific order of categorization thereof. Accordingly, features described as "first" or "second" can be explicitly or implicitly included in at least one of the features. In the description of the application, the meaning of "N" is at least two, for example, two, three, etc., unless explicitly specified otherwise.

[0143] Any process or method descriptions or blocks in flow charts or otherwise described herein represent embodiments which can be managed as one or more modules, segments, or portions of code which include one or more executable instructions for implementing specific logic functions or steps, and alternate implementations are possible. In some embodiments, the processes and methods described can be executed by one or more apparatuses or devices, either directly or after conversion to another language. Alternate implementations are possible.

[0144] The logic and / or steps represented in the flowcharts and / or described herein, for example, can be considered as a sequence of executable instructions stored in a computer readable medium, which can be executed by an instruction execution system, apparatus or device, such as a computer-based system, a processor-based system, or other system that can fetch the instructions from the instruction execution system, apparatus or device and execute the instructions, or a combination of the above. For the purposes of this specification, a "computer readable medium" can be any apparatus that can contain, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus or device. The computer readable medium can be a computer readable storage medium or a computer readable signal medium. The computer readable storage medium can include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or a propagation medium. The computer readable signal medium can include, but is not limited to, a computer readable medium that facilitates transfer of the program from one place to another. A specific example of a computer readable medium is a non-transitory computer-readable storage medium. A specific example of a computer readable signal medium is a source or destination of the computer readable medium. Another specific example of a computer readable signal medium is a computer readable signal travelling through space. Thus, a computer readable medium can take many forms of hardware to carry out the program for use by or in connection with the instruction execution system, apparatus or device.

[0145] It should be understood that aspects of the application can be implemented in hardware, software, firmware or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in a memory and executed by a suitable instruction execution system. If implemented in hardware and in another embodiment, the hardware can be implemented using any or a combination of the following technologies, which are each well known in the art: a discrete logic circuit(s) having logic gates for implementing logic functions upon an application of data signals, an application specific integrated circuit having appropriate combinational logic gates, a programmable gate array(s) (PGA), a field programmable gate array (FPGA), etc.

[0146] Those of skill in the art would understand that the steps of the methods carried out above can be carried out wholly or partly by a program instructing relevant hardware, and the program can be stored in a computer readable storage medium, and when executed, includes one or a combination of the steps of the method embodiments.

[0147] In addition, each of the functional units in the various embodiments of the present application can be integrated in one processing module, or each of the units can be physically present separately, or two or more units can be integrated in one module. The integrated module can be implemented in the form of hardware or in the form of a software functional module. When the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer readable storage medium.

[0148] The storage medium mentioned above can be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it should be understood that the above embodiments are exemplary and should not be construed as limiting the present application, and those skilled in the art can make changes, modifications, replacements and variations to the above embodiments within the scope of the present application.

Claims

1. A target population identification method based on multi-level fusion of neural blood flow features, characterized in that, The method comprises the following steps: Based on a synchronous EEG-fNIRS acquisition device, brain neural blood flow signal data containing EEG signals and near-infrared signals of normal people and target people meeting preset conditions are collected; The brain neural blood flow signal data are aligned and preprocessed, and the EEG signals are divided into multiple EEG components; Based on the preprocessed brain neural blood flow signal data, at least one single-modality same-class feature parameter is analyzed and extracted, and the single-modality same-class feature parameter is mapped to a specific channel position to obtain a specific channel feature; Based on the preprocessed brain neural blood flow signal data, at least one mixed-modality feature parameter is analyzed and extracted, the mixed-modality feature parameter is mapped to a specific channel position to obtain a specific channel feature, and feature fusion at a data layer is realized based on the at least one mixed-modality feature parameter, wherein the analysis and extraction of the at least one mixed-modality feature parameter based on the preprocessed brain neural blood flow signal data comprises: Multi-modality EEG-fNIRS source estimation based on spatial co-location is performed on the multiple EEG components to obtain a cortical current source density; The cortical current source density is reconstructed on a near-infrared channel position to convolve the current source density on the near-infrared channel position with a preset standard blood flow dynamics function to obtain an expected blood flow change; A real blood flow response is obtained according to the expected blood flow change and the near-infrared signal, a standardized slope is calculated by using a general linear model to determine a multi-band local neural vascular coupling strength, wherein the calculation formula of the multi-band local neural vascular coupling strength is: ， wherein, is a source activity representing a variety of brain waves, is a potential recorded by an EEG electrode, is a weight matrix constructed from fNIRS activated regions, is a lead matrix of the EEG, is a current source density selected on near-infrared channel positions, is a convolution of the current source density with a standard hemodynamic response function, is a real hemodynamic change, is an intercept of a regression model, is a regression coefficient; Yeo 7 network is used to integrate the single-modality same-class feature parameters and the mixed-modality feature parameters in position to realize fusion at a feature layer, and then differences and connections of the single-modality same-class feature parameters and the mixed-modality feature parameters on the same network are analyzed, same-class feature parameters of the target people and the normal people are taken as independent samples, and mean hypothesis testing is used to find brain features and brain networks sensitive to the target people; A plurality of machine learning classifiers are trained by using the brain features sensitive to the target people to realize recognition of the target people.

2. The method of claim 1, wherein, The analysis and extraction of the at least one single-modality same-class feature parameter based on the preprocessed brain neural blood flow signal data comprises: EGG time domain data are segmented according to a preset time, the correlation between each EGG channel signal in each segment is calculated by linear time delay coherence to obtain a functional connection matrix of a neural modality, and the functional connection matrix of the neural modality is averaged according to the EGG channel to obtain a functional connection strength on a specific EGG channel; A correlation matrix between each fNIRS channel time domain data is calculated by Pearson correlation, and then the correlation matrix is nonlinearly corrected to obtain a functional connection matrix of a blood flow modality, and the functional connection matrix of the blood flow modality is averaged according to the fNIRS channel to obtain a functional connection strength on a specific fNIRS channel; Based on the functional connectivity matrix of the neural modality and the functional connectivity matrix of the blood flow modality, a preset extensive threshold is set to set sparsity, a binary network is constructed according to the sparsity, a plurality of graph theory parameters are calculated by a graph theory method, and AUC values of the plurality of graph theory parameters are calculated to obtain stable feature estimation values.

3. The method of claim 1, wherein, The integration of different modal features in position by using the Yeo 7 network includes: A preset standard system is used to determine the specific spatial position of the neural modality and the blood flow modality channel, and the specific spatial position is mapped to a unified MNI standard space by using a spatial registration method to obtain the MNI coordinates of each modality and each channel; The MNI standard space is divided into 7 brain networks by using the Yeo 7 network, and the brain network to which each brain electrical electrode and near-infrared channel belongs is determined according to the MNI coordinates of the brain electrical electrode and the near-infrared channel, so as to realize the integration of the single-modality similar feature parameters and the mixed-modality feature parameters in the spatial position; The normal population and the target population are tested by using mean hypothesis testing, and the brain features with significant differences are obtained as brain features sensitive to the target population, and the brain networks in which the differences in the plurality of modal brain features of the neural modality, the blood flow modality and the mixed modality appear are taken as the brain networks sensitive to the target population.

4. The method of claim 1, wherein, The use of the brain features sensitive to the target population to train a plurality of machine learning classifiers to realize the identification of the target population includes: A plurality of machine learning classification models and a k-fold cross-validation method are used, and the brain features sensitive to the target population or the features on the brain networks sensitive to the target population are used to train a plurality of stable binary classifiers for distinguishing the normal population and the target population. The plurality of stable binary classifiers are combined to classify and identify the target population to obtain a plurality of identification results, and the plurality of identification results are integrated to realize the identification of the target population.

5. A target crowd recognition device based on multi-level fusion of cerebral blood flow characteristics, characterized in that, It includes: A collection module is configured to collect brain neural blood flow signal data of a normal population and a target population satisfying a preset condition based on a synchronous EEG-fNIRS collection device, the brain neural blood flow signal data including brain electrical signals and near-infrared signals; A preprocessing module is configured to align and preprocess the brain neural blood flow signal data, and divide the brain electrical signals into a plurality of brain electrical components; A single-modality similar feature parameter extraction module is configured to analyze and extract at least one single-modality similar feature parameter based on the preprocessed brain neural blood flow signal data, and map the single-modality similar feature parameter to a specific channel position to obtain a specific channel feature; A mixed-modality feature parameter extraction module is configured to analyze and extract at least one mixed-modality feature parameter based on the preprocessed brain neural blood flow signal data, map the mixed-modality feature parameter to a specific channel position to obtain a specific channel feature, and realize feature fusion at a data layer based on the at least one mixed-modality feature parameter, wherein the analysis and extraction of the at least one mixed-modality feature parameter based on the preprocessed brain neural blood flow signal data include: Multi-modality EEG-fNIRS source estimation based on spatial co-localization is performed on the plurality of brain electrical components to obtain a cortical current source density; Reconstructing the cortical current source density at near-infrared channel locations to convolve the current source density at the near-infrared channel locations with a predetermined standard hemodynamic function to obtain an expected blood flow change; According to the expected blood flow change and the near-infrared signal, a real blood flow response is obtained, a general linear model is used to calculate a normalized slope to determine a multi-band local neurovascular coupling strength, and a calculation formula of the multi-band local neurovascular coupling strength is: ， wherein, is a source activity representing a plurality of brain waves, is a potential recorded by an EEG electrode, is a weight matrix constructed from fNIRS activated regions, is a lead matrix of the EEG, is a current source density selected on near-infrared channel locations, is a current source density convolved with a standard hemodynamic response function, is a true hemodynamic change, is an intercept of a regression model, is a regression coefficient; A multi-modal feature integration module is configured to integrate the single-modal same-type feature parameters and the mixed-modal feature parameters in position by using a Yeo 7 network, to realize fusion of feature layers, and to analyze differences and connections between the single-modal same-type feature parameters and the mixed-modal feature parameters on the same network, to take the same-type feature parameters of the target population and the normal population as independent samples, and to find brain features and brain networks sensitive to the target population by using mean value hypothesis testing; A recognition module is configured to train a plurality of machine learning classifiers by using the brain features sensitive to the target population to realize recognition of the target population.

6. An electronic device, comprising: Comprise: A memory, a processor, and a computer program stored on the memory and executable on the processor, the processor executing the program to implement the target population recognition method based on multi-level fusion of neurovascular flow features according to any one of claims 1-4.

7. A computer readable storage medium having stored thereon a computer program, characterized in that, The program is executed by the processor to implement the target population recognition method based on multi-level fusion of neurovascular flow features according to any one of claims 1-4.

8. A computer program product comprising a computer program, characterized in that, The computer program is executed to implement the target population recognition method based on multi-level fusion of neurovascular flow features according to any one of claims 1-4.

Images