115 results about "Resting state fMRI" patented technology

The invention discloses a myocardial perfusion image classification method and system based on a single resting state, and belongs to intelligent analysis and auxiliary diagnosis of medical images. The method comprises the following steps: acquiring SPECT three-dimensional voxel data in the single resting state, and performing image reconstruction by adopting an OSEM algorithm; segmenting the myocardial region by using a pre-trained U-Net convolutional neural network, and mapping a segmentation result to a two-dimensional polar coordinate graph conforming to the AHA17 segment model; extracting a multi-dimensional feature vector; a single-phase inference model MSR-Net based on biphase labeling is constructed, in the training stage, segment classification labels of resting-load biphase images are used, in combination with segment consistency indexes, label correction is carried out, and in the inference stage, only resting state features are input, and then a pixel-level perfusion defect distribution diagram and ischemia scores of 17 myocardial segments can be output. Quantitative and segmental evaluation of myocardial ischemia can be completed by using single resting state imaging, exercise or drug load examination is avoided, and cardiovascular adverse events and complication risks are reduced.

The invention provides a method for generating a personalized nerve regulation and control stimulation scheme, and aims to generate a precise nerve regulation and control stimulation scheme for an individual through an electroencephalogram evaluation result. The method comprises the following steps: firstly, constructing a scheme library containing a plurality of transcranial electrical stimulation basic schemes; secondly, collecting resting-state electroencephalogram signals of a user, extracting electroencephalogram characteristic indexes related to emotion, cognition and sleep, and comparing the electroencephalogram characteristic indexes with a norm database to judge cognition risks; generating a primary stimulation scheme according to an evaluation result, and if a cognitive risk exists, further detecting that a feature index is abnormal and generating a targeted correction scheme; all the schemes are subjected to priority ranking, and the sequence is determined according to the abnormal severity degree and the clinical weight; and finally, outputting a personalized treatment scheme sequence of one week. According to the method, the accuracy and effectiveness of treatment are improved, high individuation, systematicness, practicability and dynamic optimization potential are achieved, and powerful support is provided for nerve regulation and control treatment.

The invention discloses a dementia identification method based on brain computer network space cross attention fusion. The dementia identification method comprises the steps that resting-state electroencephalogram signals are acquired, preprocessing and brain network construction are carried out, the frequency band power ratio is calculated, and a training set and a test set are divided; the brain network fuses the spatial information, and spatial features are obtained through an isotropic graph neural network and a local feature enhancement module; the frequency band power ratio is coded by a multi-layer perceptron to obtain frequency spectrum statistical characteristics; a bidirectional cross attention module is input, and complementary fusion features are extracted; and performing mixed pooling and splicing on the complementary fusion features, then sending the fused features to a Chebyshev Kolmogorov-Arnold network classifier, and outputting an Alzheimer's disease / frontotemporal dementia / health control (AD / FTD / HC) classification result and a cognitive scale evaluation score (MMSE). According to the method, the multi-feature complementary information is effectively integrated through joint modeling of the spatial diagram features and the global spectrum features, and the accuracy, stability and generalization ability of AD and FTD classification in a complex brain network are improved.

The invention discloses an autism detection method based on multi-mode collaborative embedding, and belongs to the technical field of medical image analysis and artificial intelligence. The method comprises the following steps: firstly, obtaining resting state functional magnetic resonance imaging data and non-imaging data of a subject; a Markov transition field is utilized to encode the time sequence into an image so as to retain dynamic features, and feature extraction is carried out through an efficient multi-scale attention module; then realizing effective fusion and semantic alignment of multi-modal information by adopting a three-level fusion architecture and a joint loss function; and then adaptively constructing a graph structure based on the fusion features, dynamically learning a node relationship by using a graph attention network, and completing a classification decision. According to the method, the defects of a traditional method in the aspects of dynamic feature modeling, multi-modal fusion and heterogeneous graph structure processing are effectively overcome, the autism detection accuracy and robustness are remarkably improved, and a reliable tool is provided for clinical intelligent diagnosis.

The invention discloses a calculation method for integrating task induction and intrinsic spontaneous brain function activity. The calculation method comprises the following steps: calculating a brain activation mode when an individual executes a corresponding cognitive task based on task state functional magnetic resonance imaging data and a general linear model; identifying individual large-scale nerve avalanche with spatial continuity based on resting state functional magnetic resonance imaging data; the method comprises the following steps: performing principal component analysis on resting state functional magnetic resonance data of an individual to construct a low-dimensional state space; a task-induced brain activation mode and intrinsic spontaneous nerve avalanche are projected to an individual low-dimensional state space; calculating the Euclidean distance between the task-induced brain activity and the intrinsic spontaneous nerve avalanche in the low-dimensional state space; and detecting the prediction effect of the geometric distance on the performance of the tested task through the regression model. The method is verified on a real data set, and experimental results show that the method not only can integrate two basic brain function activities, but also can significantly predict individual cognitive performance differences.

The invention relates to the technical field of brain magnetic signal denoising, and provides a brain magnetic background noise suppression method and device based on multi-scale frequency domain subspace projection filtering. The method comprises the following steps: firstly, converting a multi-channel resting-state noise signal and a multi-channel brain magnetic signal into a time-frequency domain through wavelet packet transformation, and then decomposing data after wavelet packet transformation into different frequency bands so as to separate noise components more clearly; carrying out singular value decomposition on the sub-band coefficient matrix, and adaptively selecting a threshold value by combining an energy accumulation method and a second-order difference method so as to eliminate noise related components; and finally, denoising is performed on each frequency band by using a common subspace projection method, and then the denoised data is reconstructed to obtain the denoised brain magnetic signals, so that the method has a good noise suppression effect, and high-quality clean data can be provided for subsequent brain magnetic signal analysis.

The invention relates to a brain anomaly network positioning method and system based on functional connection network mapping in the technical field of neural image data processing. The brain abnormal network positioning method comprises the following steps: calculating a whole brain function network diagram connected with each abnormal site by using resting state function connection data of large-scale health subjects based on a plurality of dispersed abnormal sites reported in previous literatures; then superposing the function network diagrams to obtain a network probability graph; and finally, filtering the probability graph through a threshold value to obtain a final core anomaly network graph. According to the method, the inherent functional connection architecture of the brain is used as a reference system, abnormal sites which seem to be uncorrelated in different researches are successfully traced and unified to a common and stable functional network, and compared with a single brain region marker, the generated network-level biomarker integrates more source evidences, so that the network-level biomarker has the advantages that the network-level biomarker can be widely applied to the field of biomarkers of the brain region. Therefore, the problems of result heterogeneity and inconsistency in brain abnormality discovery are solved.

The invention relates to the technical field of medical health, in particular to an epilepsy auxiliary evaluation method and device based on heart rate variability, equipment and a medium. The method comprises the following steps: acquiring at least two target physiological states in a current acquisition combination, wherein the target physiological states comprise a resting state and one or two stimulation states; based on the target physiological state, electrocardiosignals of the subject are collected; based on a preset signal processing algorithm, heart rate variability analysis is conducted on each group of electrocardiosignals, a corresponding resting state HRV feature set and at least one stimulation state HRV feature set are obtained, and each HRV feature set comprises a plurality of feature items; determining a group of target feature items suitable for the current collection combination according to the independent judgment efficiency, the correlation judgment efficiency and / or the global judgment efficiency of each feature item on epilepsy; and generating a screening report for assisting epilepsy risk assessment based on the target feature item. And the operation efficiency is improved while the evaluation accuracy is guaranteed.

The invention discloses a brain function training device and method, and belongs to the field of brain function training.The method comprises the steps that before training is started, the historical average value of resting state numerical values of target brain function indexes of a user and the current resting state numerical value are determined; determining a dynamic offset according to a difference value between the current resting state numerical value and a historical average value; adjusting and updating the reference threshold sequence through the dynamic offset to obtain a target threshold sequence; after training is started, a target stage corresponding to a real-time interval (the interval where a real-time numerical value of a target brain function index is located in a target threshold sequence) in the multi-stage feedback content can be determined; feedback content of the target stage is fed back to the user; on one hand, the target threshold sequence can be dynamically determined by referring to the current and historical resting state numerical values (of the target brain function index) of the user, the training effect is improved, on the other hand, the interestingness of brain function training is improved through the multi-stage feedback content, and the training enthusiasm of the user is improved.

The application discloses a closed-loop dynamic modeling system for heart, blood vessels and brain, and relates to the field of physiological system modeling.The system comprises a signal processing module, a closed-loop dynamic modeling module, a personalized training module and a state analysis module; the signal processing module extracts a feature sequence from multi-modal physiological signals; the closed-loop dynamic modeling module constructs a dynamic skeleton with physical differential equations, and embeds a neural network subject to physical constraints in nonlinear regulation nodes of brain central control, autonomic nervous regulation, heart blood pumping and blood vessel transmission to generate a dynamic numerical trajectory of system state variables; the personalized training module firstly fixes the neural network on resting state data to determine physical model parameters, and then jointly optimizes the neural network and the physical model parameters based on data fitting terms and physiological constraint terms on load task data; and the state analysis module calculates brain-heart sympathetic drive gain, pressure reflex sensitivity and blood vessel hardening index to determine dominant factors of cardiovascular regulation function.

This invention discloses a novel intelligent rapid screening device for depression / Alzheimer's disease, used for early auxiliary detection of either condition. It comprises: an EEG acquisition module, a human-computer interaction module, a wireless transmission module, and an intelligent analysis module. The EEG acquisition module collects resting-state and task-oriented EEG signals from the subject. The human-computer interaction module selects the screening mode, inputs subject information, performs impedance detection, signal verification, and provides task guidance. The wireless transmission module enables data transmission between the modules. The intelligent analysis module receives the EEG signals, analyzes the data using a machine learning model, and outputs the screening results. This device uses dry electrode technology, eliminating the need for conductive adhesive. The brain-computer interface is made of skin-friendly material for comfortable wear. Wireless transmission and cloud analysis provide high flexibility. AI-based data analysis and pattern recognition improve diagnostic efficiency and accuracy.

The invention relates to a brain induced activity-based electroencephalogram accurate decoding method and a brain-computer interface system. The brain induced activity-based electroencephalogram accurate decoding method comprises an electroencephalogram acquisition module for acquiring original electroencephalogram signals in a resting state and a motion state; an electroencephalogram preprocessing module; the brain dynamics model takes the pure resting state EEG signals and the stimulation state EEG signals processed by the electroencephalogram preprocessing module as input, the internal state of brain spontaneous activity and the internal state of brain spontaneous activity and induced activity mixture are obtained respectively, and then the internal state of brain induced activity is obtained; the internal state-EEG conversion network takes the pure signal processed by the electroencephalogram preprocessing module as input, and obtains a reconstructed EEG signal by using a coding and decoding network; the training module is used for optimizing model parameters and internal state-EEG conversion network parameters at the same time; and the decoding module is used for decoding the motion behavior according to the internal state of the brain induced activity output by the brain dynamics model. The accuracy of model parameter and internal state estimation is improved, and interference generated by spontaneous activity is eliminated.

The embodiment of the invention provides an electroencephalogram emotion recognition method based on a graph neural network and federal learning. The method is applied to the technical field of artificial intelligence. The method comprises the following steps: acquiring a resting state functional magnetic resonance imaging time sequence and non-image personalized data; preprocessing the resting state functional magnetic resonance imaging time sequence, and constructing a dynamic graph sequence for the preprocessed resting state functional magnetic resonance imaging time sequence based on a plurality of preset brain maps by adopting a sliding window technology; inputting the dynamic graph sequence into a shared feature layer for feature extraction to obtain a space-time shared feature vector; inputting the non-image personalized data into an independent personalized layer for feature extraction to obtain a personalized feature vector; performing feature fusion processing on the space-time sharing feature vector and the personalized feature vector to obtain a fused feature; and the fused features are mapped into the electroencephalogram emotion category probability through the classifier, an electroencephalogram emotion recognition result is obtained, and the electroencephalogram emotion recognition accuracy and the generalization ability of the model are improved.

This invention relates to a precise EEG decoding method and brain-computer interface system based on evoked brain activity, comprising: an EEG acquisition module for acquiring raw EEG signals in resting and motor states; an EEG preprocessing module; a brain dynamics model that, using the purified resting-state and stimulated-state EEG signals processed by the EEG preprocessing module as input, obtains the internal states of spontaneous brain activity and the mixed internal states of spontaneous and evoked brain activity, thereby obtaining the internal states of evoked brain activity; an internal state-EEG conversion network that, using the purified signals processed by the EEG preprocessing module as input, obtains reconstructed EEG signals using an encoding-decoding network; a training module that simultaneously optimizes the model parameters and the internal state-EEG conversion network parameters; and a decoding module that decodes motor behavior based on the internal states of evoked brain activity output by the brain dynamics model. This improves the accuracy of model parameters and internal state estimation and eliminates interference from spontaneous activity.

The invention provides a lower limb rehabilitation evaluation method, system and device based on a brain-computer interface and a medium, and relates to the technical field of rehabilitation medicine and medical electronics. According to a lower limb rehabilitation evaluation technology based on a brain-computer interface, a two-way evaluation system of peripheral muscle form-central nervous activity is constructed through synchronous acquisition of wearable A-type ultrasound and electroencephalogram signals; a-type ultrasound monitors the change of the thickness of the rectus femoris in real time, and electroencephalogram signals analyze the coherence of Alpha / Beta frequency bands, so that functional difference evaluation of a stroke patient in a resting state, a passive motion state and an active motion state with different resistances is realized, the limitation of single subjective scoring of a traditional scale is avoided, and a multi-dimensional quantitative index is provided for rehabilitation evaluation.

The invention provides an EEG (electroencephalogram)-based self-sensitivity quantitative evaluation method, and belongs to the field of physiological signal processing and mode recognition. The method comprises the following steps: collecting resting state and task state electroencephalogram signals of a subject, and obtaining a self-efficacy feeling score of the subject in combination with a scale; the electroencephalogram signals are preprocessed, window segmentation is carried out on the preprocessed signals, and multi-dimensional features such as a time domain, a frequency domain, a power spectrum entropy and a micro state are extracted; and inputting the features and the performance feeling score into a deep learning model based on LSTM-Attention for training, and constructing an evaluation model. The model can output a quantitative score according to the electroencephalogram signal, and the evaluation result continuously iterates and optimizes the model through continuous feedback data, so that the evaluation precision is improved. According to the method, objective and accurate quantification of self-effectiveness is achieved by combining the electroencephalogram signal features and the deep learning technology, the internal association between the self-effectiveness and electroencephalogram activity is revealed, and high practicability is achieved.

The invention discloses a dementia identification method based on an electroencephalogram norm network and double-current attention fusion, which comprises the following steps: collecting resting-state electroencephalogram data of a subject, and preprocessing the resting-state electroencephalogram data; constructing a normalized standard brain network mode based on age change; calculating the deviation degree of each brain function connection of the subject relative to the healthy norm; extracting an image feature vector Ie and a numerical feature vector Ve from the deviation matrix by adopting a parallel double-flow architecture; according to the image feature vector Ie and the numerical value feature vector Ve, respectively obtaining an image feature ViTfeed and a numerical value feature zfeed, and fusing the ViTfeed and the zfeed to obtain a final feature vector Ffuse; and sending the final feature vector Ffuse into a downstream classifier, carrying out final classification discrimination, and outputting a final prediction result. The dementia identification method based on the electroencephalogram norm network and the double-current attention fusion has the advantages that the image features and the numerical features can be effectively fused, and the identification accuracy and the early diagnosis sensitivity of the dementia are improved.

The application relates to the technical field of biomedical information processing, in particular to an autism classification system based on resting-state electroencephalogram signals, which comprises an EEG signal acquisition and preprocessing unit and an EEG signal classification unit, the EEG signal classification unit is used for classification by using a trained Rest-HGCN network model; the Rest-HGCN network model comprises a resting-state mixed graph network module, an attention learning module and a classification module; the resting-state mixed graph network module comprises a cognitive graph branch and a data-driven graph branch and is used for extracting corresponding feature mappings; the attention learning module is used for fusing the feature mappings extracted by the resting-state mixed graph network module to obtain a final feature mapping; and the classification module is used for classifying the final feature mapping to obtain a classification result. Through the classification system, the problems of ASD patient EEG feature extraction difficulty and low recognition rate in the prior art can be effectively solved, and only a small amount of features are needed to achieve the purpose of more efficient ASD classification recognition.

The application relates to the cross field of biomedical engineering and artificial intelligence, and discloses a personalized brain power development training method based on electroencephalogram signals. The method comprises the following steps: collecting resting state and task state multi-channel electroencephalogram signals of a subject, constructing a functional connection matrix after pretreatment, identifying individualized weak connection target points through difference operation and cluster analysis; matching a neural feedback training protocol from a preset paradigm library based on the target points, and generating a feedback signal by extracting a target point synchronicity feature in real time during training to guide the subject to actively enhance the weak connection; updating the model after each training and dynamically optimizing subsequent parameters to form a closed-loop regulation. The application improves working memory and attention through individualized targeted training, induces neural plasticity, and realizes efficient and accurate brain power development.

This invention provides a method and system for predicting brain diseases by fusing amplitude and phase information with graph-aware hybrid expert graph neural networks. It relates to the field of neuroimaging analysis technology. The method includes: acquiring raw resting-state functional magnetic resonance imaging (fMRI) data of a subject and corresponding brain disease category labels; preprocessing the raw resting-state fMRI data; extracting the mean oxygenation level dependent signal time series of multiple brain regions of the subject and using the mean oxygenation level dependent signal time series as the signal time series; constructing a functional connectivity matrix and a phase adjacency matrix based on the signal time series; constructing a training dataset; constructing a brain disease prediction model based on a two-branch hybrid expert graph neural network; training the brain disease prediction model using the training dataset; acquiring resting-state fMRI data; inputting the resting-state fMRI data into the trained brain disease prediction model for prediction, and outputting the brain disease prediction result.

Resting-state magnetic resonance imaging (fMRI) technology records brain activity by measuring the global blood oxygen saturation of the subject's brain, providing high spatial resolution data. Although the temporal resolution is lower, in elderly individuals at rest, the introduction of time series data still allows for good observation of the hemodynamic direction of individual pixels, making it one of the important evaluation indicators for intervention and treatment of brain diseases. The neural circuit features constructed based on a single brain region towards the whole brain are also an important evaluation method for fMRI data, namely, the quantification of the physiological interaction of the brain region towards the whole brain in a certain state. This invention constructs a method for quantifying the treatment effect of Alzheimer's disease (AD) based on resting-state magnetic resonance imaging (fMRI) technology. This method mainly addresses: (1) the difficulty in quantifying the rehabilitation effect during the treatment of AD patients; and (2) the quantification of the rehabilitation effect of different treatment modes during the treatment of AD patients.

A biomarker predictive of a survival outcome of a brain tumor patient is disclosed. The biomarker includes a functional connectivity matrix that includes a plurality of matrix elements. Each matrix element includes a correlation of resting-state fMRI activities of a first and second region of interest from a plurality of regions of interest within the patient's brain. Computing device and systems are disclosed to transform a resting-state fMRI dataset obtained from the patient into the biomarker and to transform the biomarker into a predicted survival outcome using a machine learning model.

The application discloses a short-term memory discrimination method based on multi-modal neuroimaging and deep learning, and comprises the following steps: S1, synchronizing an EEG-fMRI device to collect electroencephalogram signals and BOLD signals in a resting state and a memory task; S2, pre-processing the data collected in the step S1; S3, extracting feature data of the EEG and the fMRI through a convolutional neural network; S4, fusing the feature data of the EEG and the fMRI through position coding and attention mechanism weighting to generate a joint feature vector; and S5, classifying the joint feature vector based on an ELM algorithm to output a memory state discrimination result.The application has the beneficial effect that the fMRI and EEG data obtained through collection and fusion are combined with the CNN and ELM algorithms to realize high-precision memory state classification, the correlation analysis is performed with a behavioral result, the short-term memory characteristic signal is discriminated by using neuroimaging data, and the neural mechanism of memory coding and consolidation is analyzed.

The application belongs to the technical field of medical image analysis, and provides a dynamic brain function network analysis method and system based on a time-space joint state. The method comprises the following steps: in the time dimension, according to a preset sliding window, a four-dimensional resting-state functional magnetic resonance image data segment after preprocessing is intercepted, and the Pearson correlation coefficient of the time sequence signals between any two brain regions in the window segment is calculated to obtain a dynamic brain function connection matrix; the independent component analysis method is used to extract the dynamic brain function connection matrix to obtain the individual corresponding independent component and the time sequence corresponding to the independent component; based on the individual corresponding independent component and the time sequence corresponding to the independent component, an efficient forward search strategy and a classifier are used to construct a time-space fusion discrimination model, specific brain dynamic networks related to specific neuropsychiatric diseases are identified, and a decision value is output, so that quantitative measurement of the dynamic brain function network at the individual level is realized.

The application discloses a cognitive disorder screening method and system based on resting state electroencephalogram and multi-task paradigm, wherein the method comprises the following steps: collecting baseline closed-eye resting state electroencephalogram signals of a subject under a task-free stimulation condition to establish an individualized neural steady state reference; presenting a multi-dimensional cognitive task covering multiple cognitive function networks; collecting closed-eye resting state electroencephalogram signals in a preset time threshold after the task ends to represent the steady state regression process after neural disturbance. Based on the baseline and recovery period electroencephalogram signals, the neural recovery amplitude feature, the neural recovery rate feature and the brain network reconstruction feature are calculated, the neural recovery feature set is mapped to the healthy neural recovery norm space, the deviation degree of the neural steady state regulation ability is evaluated, and the cognitive disorder risk assessment result is output. The application overcomes the problem of insufficient sensitivity of the traditional resting state or task state single analysis mode to early cognitive disorders, and is suitable for early screening and clinical auxiliary evaluation of cognitive disorders.

The invention discloses a brain function network classification method and system based on adversarial graph comparative learning, and the method comprises the steps: obtaining resting-state functional magnetic resonance image data, and carrying out the preprocessing of the data, and constructing a functional connection matrix X; and inputting the X into an adversarial graph comparison learning classification model for classification. The model comprises an image augmentor, a feature extraction layer, a projection head and a classifier. A trainable encoder is arranged in the image augmentation device, and an edge deletion probability matrix P and an augmentation image X 'are generated through the encoder; the feature extraction layer extracts X and X 'feature representations; the projection head maps the feature representation to a contrast learning space to obtain an optimal weight parameter; the classifier performs brain function classification based on the feature representation. The method has the advantages that data driving and task-oriented dynamic augmentation are achieved, classification related function connection is reserved, redundant connection is deleted, the model can distinguish different brain region function specificity, and the problems that a traditional model cannot distinguish different brain region function differences due to node replacement invariance, so that classification performance is poor, and explanatory performance loses practical significance are solved.

This invention discloses a hypergraph representation method for brain functional networks. The steps include: preprocessing resting-state functional magnetic resonance imaging (fMRI) to obtain time series data for all brain regions; dividing the entire time series into multiple overlapping sub-sequence segments using a sliding window; constructing a dynamic brain functional network and transforming it into an optimization model; constructing a hypergraph of the dynamic brain functional network using the nearest neighbor algorithm; dynamically modifying the hypergraph structure through convolution operations and extracting features to obtain a new dynamic hypergraph; extracting the Laplacian matrix of the dynamic hypergraph; constructing the manifold regularization term of the Laplacian matrix and simultaneously introducing the manifold regularization term and the L1 norm regularization term into the optimization model to obtain the hypergraph representation of the brain functional network. This invention is used to represent functional interactions and higher-order relationships between multiple brain regions, determine discriminative brain functional network classification features, and effectively improve the classification performance of brain disease features.

The application discloses a system and method for autism risk assessment and early warning, and relates to the field of computer and electroencephalogram signal processing. The system comprises an electroencephalogram data acquisition module, a feature extraction module, a risk assessment module and a risk early warning module. The preset task state covers resting state, local motion state and social motion state. In the resting state, the user observes low-speed target objects in the display picture for a preset time length. In the local motion state, the user observes the first type of target person completing local limb motion. In the social motion state, the user observes the second type of target person waving hands to say hello. By collecting electroencephalogram data in each task state and extracting features, the autism risk probability and the corresponding risk level are determined, and the early warning is triggered according to the preset logic rules. The system does not need frequent language interaction with children, is suitable for young children, significantly reduces the screening manpower and time cost, and is convenient for popularization in areas with scarce medical resources.

The invention relates to a depression brain function image classification method and system based on a dynamic high-order connection multi-scale space-time diagram, and the method comprises the steps: obtaining a resting state functional magnetic resonance imaging signal, calculating a Pearson's correlation coefficient between regions of interest in a time window, and obtaining a dynamic low-order functional connection sequence; based on the dynamic low-order function connection sequence, calculating a Pearson's correlation coefficient between the low-order function distribution sequences of the region of interest to obtain a dynamic high-order function connection sequence; low-order function features and high-order function features are obtained in parallel through time slice graph construction based on multi-modal information fusion, node embedding representation updating, important node screening, multi-level feature extraction and scaling and centralization processing; based on the low-order function features and the high-order function features, obtaining fusion features through multi-modal fusion based on a multi-head self-attention mechanism; based on the fusion features, classification of depression and health control is predicted, and importance visualization of the brain region is realized based on attention weight.

The invention discloses a depression analysis method based on deep learning and resting-state electroencephalogram data, and relates to the technical field of electroencephalogram data analys.The depression analysis method comprises the steps that short-time Fourier transform is conducted on collected environment radio-frequency signals and electroencephalogram data of a patient, and the closeness degree of the environment radio-frequency signals and the electroencephalogram data is quantified; capturing micro-motion data of the head of the patient in real time, marking micro-motion intervals and motion intensity in each micro-motion interval, and generating a bad contact event log by using an impedance detection function of electroencephalogram equipment; and aligning the environment radio frequency signal, the six-axis IMU data and the channel impedance data by adopting a dynamic time warping algorithm, performing Min-Max normalization on the characteristics of each mode, analyzing a generation quality tag of EEG data per second, and outputting a patient resting state EEG data interference thermodynamic diagram. Problem time periods can be quickly positioned, pollution data are eliminated or acquisition parameters are adjusted during analysis, and the result accuracy is improved.