CN118161166A - Method and device for identifying depressive symptom characteristics based on ERP-fNIRS - Google Patents
Method and device for identifying depressive symptom characteristics based on ERP-fNIRS Download PDFInfo
- Publication number
- CN118161166A CN118161166A CN202410591420.5A CN202410591420A CN118161166A CN 118161166 A CN118161166 A CN 118161166A CN 202410591420 A CN202410591420 A CN 202410591420A CN 118161166 A CN118161166 A CN 118161166A
- Authority
- CN
- China
- Prior art keywords
- depression
- erp
- fnirs
- data
- characteristic
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/254—Fusion techniques of classification results, e.g. of results related to same input data
- G06F18/256—Fusion techniques of classification results, e.g. of results related to same input data of results relating to different input data, e.g. multimodal recognition
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/145—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue
- A61B5/1455—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue using optical sensors, e.g. spectral photometrical oximeters
- A61B5/14551—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue using optical sensors, e.g. spectral photometrical oximeters for measuring blood gases
- A61B5/14553—Measuring characteristics of blood in vivo, e.g. gas concentration or pH-value ; Measuring characteristics of body fluids or tissues, e.g. interstitial fluid or cerebral tissue using optical sensors, e.g. spectral photometrical oximeters for measuring blood gases specially adapted for cerebral tissue
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/16—Devices for psychotechnics; Testing reaction times ; Devices for evaluating the psychological state
- A61B5/165—Evaluating the state of mind, e.g. depression, anxiety
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/24—Detecting, measuring or recording bioelectric or biomagnetic signals of the body or parts thereof
- A61B5/316—Modalities, i.e. specific diagnostic methods
- A61B5/369—Electroencephalography [EEG]
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/72—Signal processing specially adapted for physiological signals or for diagnostic purposes
- A61B5/7235—Details of waveform analysis
- A61B5/7264—Classification of physiological signals or data, e.g. using neural networks, statistical classifiers, expert systems or fuzzy systems
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/10—Pre-processing; Data cleansing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/213—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
- G06F18/253—Fusion techniques of extracted features
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/042—Knowledge-based neural networks; Logical representations of neural networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/02—Preprocessing
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
- Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
- Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Veterinary Medicine (AREA)
- Heart & Thoracic Surgery (AREA)
- Surgery (AREA)
- Animal Behavior & Ethology (AREA)
- Evolutionary Biology (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Psychiatry (AREA)
- Pathology (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Psychology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Signal Processing (AREA)
- Social Psychology (AREA)
- Hospice & Palliative Care (AREA)
- Educational Technology (AREA)
- Developmental Disabilities (AREA)
- Child & Adolescent Psychology (AREA)
- Optics & Photonics (AREA)
- Spectroscopy & Molecular Physics (AREA)
- Neurology (AREA)
Abstract
The invention discloses a depression symptom characteristic identification method and device based on ERP-fNIRS, which are used for redesigning a novel ERP task paradigm according to the requirement of core symptom identification of depression symptoms, synchronously recording blood oxygen activities of brain tissues in combination with fNIRS signals so as to effectively monitor brain function states, extracting depression symptom signal characteristics from highly nonlinear ERP-fNIRS multi-mode data by using a OFSLLES strategy, constructing characteristic data samples, and combining CAG neural networks with double structures to fuse core symptom characteristic information so as to realize accurate resolution of depression symptoms. The invention effectively solves the pain point of depression core disease characteristic detection, greatly improves the disease identification precision, can intelligently assist doctors in judging and evaluating depression, provides objective and quantifiable detection indexes, and opens a new idea for individual accurate diagnosis and treatment of depression.
Description
Technical Field
The invention relates to the field of artificial intelligence technology application, in particular to a depression symptom characteristic identification method and device based on ERP-fNIRS.
Background
Depression is a common affective disorder disease, which seriously affects physical and mental health of patients, clinical diagnosis of depression is mainly based on a scale inquiry mode at the present stage, and how to better assist doctors in identifying, distinguishing and predicting relevant indicators of depression is explored, and construction of objective indicators becomes a long-term and important target for assisting diagnosis and treatment of depression.
The traditional identification method of symptoms and characteristics of depression mainly provides diagnosis by observing and evaluating symptoms and behaviors of patients and combining professional judgment of clinicians. However, this diagnostic method lacks biological markers, is subjectively strong, and may vary from person to person in terms of perception and description of symptoms, which may lead to inconsistent diagnostic results, and takes a long time to observe and evaluate, and does not immediately give a diagnostic result. Meanwhile, symptoms of depression may sometimes overlap with other mental disorders, or be ignored and misinterpreted by patients, resulting in higher rates of omission and misdiagnosis. Accordingly, efforts are underway in the industry to find more accurate and objective methods for diagnosing depression. Among other things, neuroimaging techniques such as event-related potential (ERP), functional magnetic resonance imaging (fMRI), and functional near infrared spectroscopy (fNIRS) are increasingly being demonstrated to provide some useful information.
For example, fNIRS and ERP are combined, which belongs to the combination of electric signals and optical signals, and the electric signals and the optical signals do not interfere with each other, so that the combination of the electric signals and the optical signals is feasible to synchronously collect and analyze.
However, there is a "contradictory conflict treatment defect" for the depressed patient, which is related to the neuronal electrical activity and blood flow change of the depressed patient in the prefrontal cortex, amygdala, hippocampus from an objective point of view, but the existing scheme of fNIRS in combination with ERP is not specific to how the above symptoms of the specific disease are treated, feature identification, etc., which results in the following problems that the ERP-fNIRS neuroimaging technology needs to think and overcome when facing the auxiliary detection of the depressed condition:
1) If the electrode monitors scalp surface electrical activity as the sum of neuronal electrical activity resulting in failure to preserve pain points of depressed prefrontal cortex and deep amygdala, and furthermore, electrical activity has hysteresis problems with brain blood hemoglobin concentration via hypothalamic-pituitary-adrenal axis and sympathetic and parasympathetic nerves, then, if an adapted psychological experimental paradigm is chosen to effectively obtain ERP data in ERP-fNIRS data? How does the psychological experimental paradigm be designed to emphasize the characteristics of typical conditions of depression from acquired electroencephalogram data?
2) How does the ERP-fNIRS signal have highly nonlinear characteristics from which to mine data structure features that can be effectively used for the identification of typical symptoms of depression?
3) How do ERP-fNIRS combine with artificial intelligence and machine learning techniques? The traditional CNN can realize efficient and accurate feature extraction by utilizing local translation invariance to scan different areas for convolution operation, however, the general CNN model ignores the connection between brain areas and limits the construction of brain structures, and the analysis proves that the invention is not suitable for auxiliary identification of typical characteristics of depression symptoms.
Disclosure of Invention
In view of the foregoing, the present invention aims to provide a method and a device for identifying characteristics of depression symptoms based on ERP-fNIRS, so as to solve the above-mentioned specific problems.
The technical scheme adopted by the invention is as follows:
In a first aspect, the present invention provides a method for identifying depression symptom characteristics based on ERP-fNIRS, comprising:
a Flanker model for identifying depression symptoms is preset;
collecting synchronous multimodal data based on the Flanker paradigm and a change in blood flow rate in a prefrontal cortex region;
performing characteristic dimension reduction processing on the synchronous multi-mode data by utilizing OFSLLE algorithm;
and carrying out feature extraction on the synchronous multi-modal data subjected to dimension reduction under Euclidean structure and graph structure, and outputting classification results of depression symptoms after fusing the double-structure feature information.
In at least one possible implementation manner, the Flanker paradigm preset for identifying a depressive disorder includes:
gray is adopted as a display background;
setting a blue fixation point at the center of the gray background;
after a preset interval time, presenting a central target arrow and left and right flank arrows for a preset duration;
Recording a feedback signal of the tested arrow to the center target during the set time period;
And similarly, continuously performing the test according to the preset presentation times of the central target arrows in different directions so as to complete the whole test times.
In at least one possible implementation manner, the collecting synchronous multi-modal data includes:
Performing brain electrical stimulation based on the Flanker paradigm, and synchronously collecting a tested multichannel brain electrical signal and a multichannel original near infrared light intensity signal;
Preprocessing and analyzing the electroencephalogram signals to obtain space-time sequence data of N200 waveforms; converting the original near infrared light intensity signal into spatiotemporal sequence data for characterizing hemoglobin concentration variation of deoxyhemoglobin concentration variation using a modified beer-lambert law;
And carrying out time sequence alignment on the time-space sequence data of the N200 waveform and the time-space sequence data of the hemoglobin concentration change, and matching space information by using a space index to obtain the synchronous multi-modal data.
In at least one possible implementation manner, the performing feature dimension reduction processing on the synchronous multi-mode data by using OFSLLE algorithm includes: the class information of the synchronous multimodal data for depression disorder identification is integrated into the LLE algorithm and an optimal projection matrix is found therefrom.
In at least one possible implementation manner, the feature extraction of the dimension-reduced synchronous multi-modal data under the euclidean structure and the graph structure, and the merging of the double-structure feature information, the outputting of the classification result of the depression symptoms includes:
Pre-constructing a double-structure model comprising a convolutional neural network and a graph convolution neural network;
Taking the synchronous multi-modal data after dimension reduction as the input of the double-structure model, and taking the classification result of the depression symptoms as the output of the double-structure model.
In at least one possible implementation manner, the identification method further comprises: the convolutional neural network in the double-structure model introduces an attention mechanism and sequentially generates corresponding attention force diagrams from two dimensions of a channel and a space; attention is drawn to the neural pathway mechanisms and brain network connection characteristics of depressed patients from the synchronized multimodal data, in combination with the input of the model.
In at least one possible implementation thereof: generating a channel attention pattern using the inter-channel relationship of the neuron activity features that change instantaneously under the conflicting stimulus;
generating a spatial attention pattern by utilizing the spatial relationship between neurons and blood flow change characteristics of the depressed forehead cortex, amygdala and hippocampus and brain network connection characteristics;
multiplying the channel attention pattern by the model input according to the bit to obtain the channel attention output characteristic;
And multiplying the spatial attention pattern with the channel attention output characteristic by bits to obtain the spatial attention output characteristic.
In a second aspect, the present invention provides a device for identifying characteristics of depressive symptoms based on ERP-fNIRS, wherein a Flanker paradigm for identifying depressive disorders is preset, the identifying device includes:
The synchronous multi-mode data acquisition module is used for collecting synchronous multi-mode data based on the Flanker paradigm and the change of the blood flow rate of the forehead cortex region;
the feature dimension reduction module is used for performing feature dimension reduction processing on the synchronous multi-mode data by utilizing OFSLLE algorithm;
And the depression symptom identification module is used for extracting the characteristics of the synchronous multi-mode data subjected to dimension reduction under the Euclidean structure and the graph structure, fusing the double-structure characteristic information and outputting the classification result of depression symptoms.
In at least one possible implementation thereof, the depression symptom identification module includes: the model building unit is used for pre-building a double-structure model comprising a convolutional neural network and a graph convolution neural network; and the synchronous multi-modal data is used as the input of the double-structure model, and the classification result of the depression symptoms is used as the output of the double-structure model.
In a third aspect, the present invention provides an ERP-fNIRS-based depression symptom characteristic identification device, comprising:
One or more processors, a memory, and one or more computer programs, the memory may employ a non-volatile storage medium, wherein the one or more computer programs are stored in the memory, the one or more computer programs comprising instructions which, when executed by the device, cause the device to perform the method as in the first aspect or any of the possible implementations of the first aspect.
The main design concept of the invention is to redesign a novel ERP task paradigm according to the requirements of the core symptom recognition of the depression symptoms, synchronously record the blood oxygen activity of brain tissues in combination with fNIRS signals to effectively monitor the brain function state, extract the highly nonlinear ERP-fNIRS multi-modal data from the highly nonlinear ERP-fNIRS multi-modal data after dimension reduction by utilizing OFSLLES strategies, simplify and fully reserve comprehensive multi-modal data information at the same time, and then combine CAG neural networks with double structures to fuse the core symptom characteristic information so as to realize accurate resolution of the depression symptoms. According to the invention, based on the Flanker paradigm acquisition of the electroencephalogram waveform and the blood flow rate change of the prefrontal cortex region, the ERP-fNIRS multi-modal data is synchronously collected and predicted under the model framework of CNN combined with GCN, so that pain points of depression core disorder characteristic detection are effectively solved, disorder recognition precision is greatly improved, a doctor can be intelligently assisted in judging and evaluating depression, objective and quantifiable detection indexes are provided, and a new idea is opened for individual accurate diagnosis and treatment of depression.
Further, the mechanism of classification of the two-structure GAC neural network disorder can be specifically explained that, on one hand, in the euclidean structure, CNN is used and a global shared convolution kernel is defined, and the same structure and high-dimensional characteristics in electrode channel information are extracted by utilizing the local translational invariance of the CNN, so that the neuron activity characteristics of a depressed patient, which change instantaneously under the stimulation of Flanker range, are captured; on the other hand, in a non-Euclidean structure, ERP-fNIRS multi-mode signals are represented by utilizing a graph structure, so that the problem that translational invariance is not applicable any more is solved, the spatial characteristics of the brain are effectively represented, the characteristics of neurons and blood flow changes of the prefrontal cortex, amygdalin and hippocampus of depression and the characteristics of brain network connection are fully and comprehensively extracted through the combination of the two, and the adaptability of the model is improved by preferably introducing an attention mechanism and multi-layer convolution.
Drawings
For the purpose of making the objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described with reference to the accompanying drawings, in which:
FIG. 1 is a frame diagram of a method for identifying depression symptom characteristics based on ERP-fNIRS according to an embodiment of the present invention;
Fig. 2 is a schematic diagram of a device for identifying characteristics of depression symptoms based on ERP-fNIRS according to an embodiment of the present invention.
Detailed Description
Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein like or similar reference numerals refer to like or similar elements or elements having like or similar functions throughout. The embodiments described below by referring to the drawings are illustrative only and are not to be construed as limiting the invention.
The invention provides an embodiment of a depression symptom characteristic identification method based on ERP-fNIRS, which can be specifically combined with an overall conception frame diagram shown in fig. 1, wherein the method mainly comprises the following key steps:
Step S1, a Flanker paradigm for identifying depression symptoms is preset;
The Flanker paradigm is a psychological experimental paradigm usually used for researching attention control and cognitive control, but the stimulation effect of the paradigm is not good in the past, so that the difference of the obtained ERP waveform characteristics for the requirements of the invention is small, and therefore, in order that ERP data recorded based on the Flanker paradigm can be used as an auxiliary detection scene of the characteristics of the depression symptoms of the invention, the invention redesigns a novel Flanker paradigm: through repeated training optimization, 160 test times are designed, the interval time is 1500ms, the sufficient response time is given, and the sufficient stimulation quantity is ensured, so that the pain points with poor stimulation effect and small waveform characteristic difference in the past can be solved. Further, based on color psychology and mood regulation disorder of a depressed patient, gray background and blue stimulation symbols are adopted in the novel Flanker paradigm, so that the mood disorder and inhibition control disorder of the depressed patient can be further amplified by utilizing the conflict between peace and calm released by blue and the heart negative mood of the depressed patient, and it is found through verification that ERP data obtained from the novel Flanker paradigm can highlight the characteristics of representing depression symptoms, namely, the novel Flanker paradigm can fully mobilize the cortex conflict inhibition function of the depressed patient, and effectively amplify depression characteristic signals.
For development, based on Matlab and E-prime 2.0 software, with RGB (127,127,127) gray as background, stimulus is RGB (72,116,203) blue, each test run starts to present 1500ms with a center gaze point (blue cross "+"), then presents the target arrow and both left and right flank arrows at the same time 1500ms. In practice, the stimulus may be presented in the display (blue stimulus symbol is set in the gray background center of the display), the display resolution may be 1280×720 pixels, the refresh rate may be 60Hz, and the viewing distance to be tested may be set to 60cm from the display screen center.
For example, during the course of the experiment it may be required that the test ignore the two-sided arrow and react to the center arrow, the arrow being to the left, the left index finger pressing the "F" key; the arrow points to the right, the index finger of the right hand presses the "J" key, and reacts as soon as possible on the basis of ensuring the accuracy. A total of 160 trials are set, the directions of the arrows are consistent and inconsistent, 80 trials are respectively set, each trial lasts for 1500ms, and the total duration is about 8 min.
Step S2, based on the Flanker model and the change of the blood flow rate of the prefrontal cortex area, collecting synchronous multi-mode data (namely ERP-fNIRS signals);
Specifically, an electrode is placed on the scalp layer of a tested person by adopting a 10-20 international standard lead system electrode placement method, and an electroencephalogram stimulation experiment is executed based on the Flanker experimental paradigm to synchronously acquire the 18-channel electroencephalogram signals and 48-channel original near-infrared light intensity signals of the tested person.
Preprocessing and analyzing the electroencephalogram signals to obtain multichannel N200 space-time sequence data; the original collected fNIRS signals are light intensity signals, and the original light intensity signals are converted into the change quantity of the deoxyhemoglobin concentration by using the modified beer-lambert law, namely the time-space sequence data for representing the change of the multichannel hemoglobin concentration.
ERP-fNIRS multi-modal data (set) is obtained by aligning the time series based on supersampling and matching spatial information with spatial index by combining ERP time-space series data containing N200 component with time-space series data of hemoglobin concentration variation.
S3, performing characteristic dimension reduction processing on the synchronous multi-mode data by utilizing OFSLLE algorithm;
The local linear embedding (Locally Linear Embedding, LLE) is a nonlinear dimension-reducing algorithm for mapping high-dimensional data into a low-dimensional space, and the LLE algorithm is already used for the aspects of gas emission prediction, seismic data, elevator fault data and the like at present, but is still blank in the analysis and research of one-dimensional nonlinear electroencephalogram signals; furthermore, redundant information exists in the low-dimensional feature space, and the category information of the sample is not embedded in the low-dimensional feature space, which is important for successfully mining the essential information of the depression disorder characteristics from the ERP-fNIRS multi-mode data. In addition, depression patients have a bias towards forehead area dysfunction in brain network connection, so that to extract this feature needs to fully preserve the positional relationship between the electrode channels, from which vectors can be constructed.
Based on the LLE of the local linear matrix algorithm, the invention proposes to adopt the local linear embedded maximum dispersion matrix algorithm (OFSLLE) with uncorrelated statistics, and perform data dimension reduction processing through the OFSLLE algorithm, so that the class information of the synchronous multi-mode data for identifying depression symptoms can be integrated into the LLE algorithm and an optimal projection matrix can be found out from the class information, thereby not only fully keeping the position relation, but also eliminating redundant information in the multi-mode data. That is, OFSLLE algorithm solves the problem that the traditional local linear embedding algorithm has redundant information in the low-dimensional feature space and does not embed the class information of the sample in the low dimension, and the complete nonlinear structural feature is mined in the ERP-fNIRS signal to obtain the remarkable cluster distribution characteristic of the ERP-fNIRS signal in the low-dimensional space.
The specific processing procedure of OFSLLE algorithm is described as follows:
1. the OFSLLE algorithm integrates class information into the LLE algorithm based on a local linear matrix algorithm to find an optimal projection matrix to satisfy the following objective function:
（1）
In the formula (i) the formula (ii), Representing the considered graph divergence matrix,Representing a similar divergence matrix.
2. A neighborhood is selected. For ERP-fNIRS datasetsFinding each sample point based on Euclidean distance between data pointsA kind of electronic deviceNearest neighbors of eachWhereinAs a dimension of the original data,In order to be the number of samples of the data,Representing sample pointsA kind of electronic deviceA set of nearest neighbor subscripts.
3. Structure of the deviceAndIs a neighborhood graph of (1). If it isIs thatK nearest neighbor of (F)Is shown in the neighborhood graph. At the position ofAndThere is an edge between them. If it isAndIs a sample of the same class, inIn the neighborhood map of (c),AndThere is an edge between them. A thermonuclear estimation algorithm is employed to determine the weights. At the position ofIf in the neighborhood graph of (1)AndAn edge exists between them, then:
otherwise it is . At the position ofIf in the neighborhood graph of (1)AndAn edge exists between them, thenFor the weight, otherwise. Similar matrix is. Also, consider a graph matrixFrom the following componentsAndComplement map construction of (c). Neighborhood in the figureIf in complement diagram of (2)AndIf K neighbor of (1), there is no edge, otherwise, rebuilding weight matrix. In the neighborhood of the graphIf in complement diagram of (2)AndSamples of the same class do not have one edge; otherwise, reconstructing the weight matrix. Thus, the first and second substrates are bonded together,It is considered that the graph matrix,. Solving for similar matricesAnd consider a graph matrix。
（3）
（4）
The problem of converting the formula of (1) into generalized eigenvalues is:
（5）
And solving the eigenvalue and the eigenvector by a formula. If the eigenvalue of the corresponding eigenvector Is thatThenIs a projection matrix.
4. In order to ensure that the reconstruction weight keeps the position information unchanged before and after the dimension reduction relative to the multi-mode data of the ERP-fNIRS,The following constraints need to be satisfied:
（6）
Using a matrix of weights Local geometric information of ERP-fNIRS multi-modal data in a high-dimensional space may be used for distribution of data in a low-dimensional space.
5. Searching for realizable neighborhood matrix in low-dimensional spaceIs of the construction point of (a)，Representing the embedding dimension. And (3) making:
（7）
to ensure low-dimensional embedded coordinates as a cost function The adjacent point can reflect the position information of the corresponding sample point in the high-dimensional observation space, and avoid generating a refund solution, and the following constraint conditions are required to be satisfied simultaneously:
（8）
（9）
writing formula (7) in matrix form, namely:
（10）
In the method, in the process of the invention, Is an n x n matrix. Calculating the characteristic values, and arranging the characteristic values in order from small to large, and recording asThe corresponding feature vector is. Because ofTo ensure that the value of the cost function is minimum, a matrix is takenThe feature vectors corresponding to the d+1 minimum feature values form embedded coordinates of multi-mode data of the high-dimensional space ERP-fNIRS in the low-dimensional space, namely:
（11）
And S4, extracting features of the synchronous multi-mode data subjected to dimension reduction under Euclidean structures and graph structures, and outputting classification results of depression symptoms after fusing the double-structure feature information.
In actual operation, a GAC (gas-computer) double-structure neural network model of CNN combined with GCN can be constructed, so that more sufficient and comprehensive depression characteristic information can be captured, a depression symptom identification scene is fully adapted, a deep analysis can be performed on space-time signals of ERP-fNIRS, and the characteristic extraction and construction of a brain region topological structure are reserved. In other words, the dual structure processing mechanism can not only preserve the high efficiency and accuracy of the CNN processing euclidean data, but also display brain connections through the GCN, so as to more fully and comprehensively preserve the feature extraction of brain structures.
First preferred aspect
For the CNN network, an attention mechanism can be introduced, attention force diagrams are generated from two different dimensions of a channel and a space in sequence, the attention force diagrams are multiplied by the input characteristic diagrams in a phase manner, and adaptive characteristic optimization is carried out, so that the neural pathway mechanism and the brain network connection characteristics of the depressed patient are fully extracted from the synchronous multi-mode data.
Specifically, the synchronous multi-modal data after feature dimension reduction optimization may be input into a convolutional neural network CNN, which may include 3 convolutional layers and 2 maximum pooling layers in actual operation. The convolution kernel size of the first convolution layer is selected to be Fs/2 based on the sampling frequency of the ERP-fNIRS signal, considering the portability and generalization capability of the model. Each convolution layer performs three operations in turn: convolution operation, batch normalization and function activation. Wherein the attention mechanism module is arranged on a2 nd pooling layer; the activation layer uses a nonlinear function ReLU function to enhance the nonlinear fitting capability of the network. In order to reduce the overfitting of the network and the time consumption of the network model, a rejection layer is applied after two maximum pooling layers, the rejection function is set to zero the amount of information input to the rejection layer, and the parameter p represents the zero return probability and takes a value of 0.5.
The attention module described above comprises two parts: channel attention and spatial attention, and generating attention force diagrams from two different dimensions of the channel and the space in turn, and carrying out adaptive feature optimization by carrying out bit multiplication on the attention force diagrams and input features of the model.
Specifically, for CNNs in a dual structure, (1) a channel attention module is constructed that generates a channel attention pattern using the inter-channel relationship of the neuron activity features that change instantaneously under conflicting stimuli. By understanding the importance of each feature, attention is paid to the feature distribution channel according to the inputted feature map. Each channel of the feature map acts as a feature detector, focusing on meaningful input information. Generating two different spatial descriptors using spatial information of average pooled and maximum pooled parallel aggregated feature mapsAndThe two descriptors are then sent to a shared network consisting of a multi-layer perceptron (MLP) and a hidden layer. And obtaining two activated results after the ReLU function is activated, summing the two results according to bits to output a feature vector, and normalizing the feature vector by an S-shaped function to obtain the output of the channel attention module. Channel attention patternIs defined as:
(2) The spatial attention module is constructed, and the spatial attention pattern is generated by utilizing the characteristics of neurons and blood flow changes of the depressed forehead cortex, amygdala, hippocampus and the spatial relationship between brain network connection characteristics. Unlike channel attention, spatial attention is focused on the rich part of the information, which is complementary to channel attention. To calculate spatial attention, an average pooling and maximum pooling operation needs to be performed to aggregate the channel information of the feature map, generating two-dimensional maps: And/> For representing the average pool feature and the maximum pool feature, respectively, and combining them to generate a valid feature descriptor, and then performing a 7 x 7 convolution of the positions of the encoded salient or suppressed regions by generating a two-dimensional spatial attention feature map. Finally, the final attention graph/>, is obtained by normalizing the sigmoid functionSpace attention is directed to the intentionThe definition is as follows:
(3) Attention to the channel Multiplying the input signal with the original input signal F by bits to obtain the attention output characteristic，The expression is:
Map the space attention AndMultiplying by bits to obtain spatial attention output characteristicsI.e., the final output characteristics,The expression is:
Second preferred aspect
For the GCN network, it may be stated that the synchronous multi-mode data is sliced in units of 1s, the original data is divided into N channels of data with a duration of 1s, and one channel is 1 node. An undirected full-join graph gi= (V, ei) is defined, representing the interaction relationship between the channels of slice i. Where V represents a set of N (n= |v|) channels, the features of the N nodes form an N X D dimensional feature matrix X, and D represents feature numbers. Ei ⊂ V denotes the set of undirected edges between node sets V. A represents an adjacency matrix of the graph-rolled neural network, which is an n×n-dimensional matrix formed based on the correlations between nodes. X and A are used as initial inputs to the model, each sample feature matrix is defined as X i, X represents the temporal background of X i, and y represents the X i depression label. Since the diagonal elements of the adjacency matrix a are all zero, its multiplication with the feature matrix X can only calculate a weighted sum of the features of each node and other neighboring nodes besides itself, while the features of the node itself cannot participate in the operation. To overcome the defect, a re-parameterization technique is adopted to add a unit matrix with the same dimension to construct an adjacent matrix，The expression of (2) is:
Wherein I is an identity matrix of the same dimension, Is a constant coefficient. Adoption=1, The nature of the default node itself is the same as the importance of its neighbors. At this time, the liquid crystal display device,
Non-normalized matrixMultiplying by the feature matrix X will change the original distribution of features, thus adjacency matrixNormalization is required. Adopting weighted average as normalization strategy, giving lower degree node weight, and the weighted average makes the influence of the lower degree node on the adjacent node larger; while the impact of the height nodes is smaller because their impact is distributed over many neighbors. First construct adjacency matrixDegree matrixAnd matrixThe definition is as follows:
Symmetry and normalization matrix First order proximity information of each node is extracted by using chebyshev approximation.The definition is as follows
H is used for representing the characteristics of each layer in the network structure, and the expression for layer-by-layer propagation of the characteristics of the graph roll-up neural network is obtained as follows:
wherein, W is a weight coefficient and l is the layer number for nonlinear activation function. For the input layer, H is X. Calculating cross entropy loss functions of all marked nodes, wherein the expression is as follows:
In summary, the main design concept of the invention is to redesign a novel ERP task paradigm according to the need of identifying the core symptoms of the depression disorder, synchronously record the blood oxygen activity of brain tissues in combination with fNIRS signals to effectively monitor the brain function state, extract the highly nonlinear ERP-fNIRS multi-mode data from the highly nonlinear ERP-fNIRS multi-mode data after dimension reduction by utilizing OFSLLES strategy, construct a characteristic data sample, simplify and fully reserve the comprehensive multi-mode data information at the same time, and then combine the CAG neural network with a double structure to fuse the core symptom characteristic information so as to realize accurate resolution of the depression disorder. According to the invention, based on the Flanker paradigm acquisition of the electroencephalogram waveform and the blood flow rate change of the prefrontal cortex region, the ERP-fNIRS multi-modal data is synchronously collected and predicted under the model framework of CNN combined with GCN, so that pain points of depression core disorder characteristic detection are effectively solved, disorder recognition precision is greatly improved, a doctor can be intelligently assisted in judging and evaluating depression, objective and quantifiable detection indexes are provided, and a new idea is opened for individual accurate diagnosis and treatment of depression.
Corresponding to the above embodiments and preferred solutions, the present invention further provides an embodiment of a device 200 for identifying characteristics of depressive symptoms based on ERP-fNIRS, wherein a Flanker model for identifying depressive disorders is preset, and the identifying device may specifically include the following components as shown in fig. 2:
A synchronous multi-modal data acquisition module 201, configured to collect synchronous multi-modal data based on the Flanker paradigm and the change in blood flow rate in the prefrontal cortex region;
The feature dimension reduction module 202 is configured to perform feature dimension reduction processing on the synchronous multi-mode data by using OFSLLE algorithm;
The depression symptom identifying module 203 is configured to extract features of the synchronized multimodal data after the dimension reduction under euclidean structure and graph structure, and fuse the dual-structure feature information to output a classification result of depression symptoms.
In at least one possible implementation thereof, the depression symptom identification module includes: the model building unit is used for pre-building a double-structure model comprising a convolutional neural network and a graph convolution neural network; and the synchronous multi-modal data is used as the input of the double-structure model, and the classification result of the depression symptoms is used as the output of the double-structure model.
It should be understood that the division of the components in the ERP-fNIRS-based depression symptom characteristic recognition apparatus shown in fig. 2 is merely a division of logic functions, and may be fully or partially integrated into a physical entity or may be physically separated. And these components may all be implemented in software in the form of a call through a processing element; or can be realized in hardware; it is also possible that part of the components are implemented in the form of software called by the processing element and part of the components are implemented in the form of hardware. For example, some of the above units may be individually set up processing elements, or may be integrated in a chip of the electronic device. The implementation of the other components is similar. In addition, all or part of the components can be integrated together or can be independently realized. In implementation, each step of the above method or each component above may be implemented by an integrated logic circuit of hardware in a processor element or an instruction in the form of software.
For example, the above components may be one or more integrated circuits configured to implement the above methods, such as: one or more Application SPECIFIC INTEGRATED Circuits (ASIC), or one or more microprocessors (DIGITAL SIGNAL Processor (DSP), or one or more field programmable gate arrays (Field Programmable GATE ARRAY; FPGA), etc. For another example, these components may be integrated together and implemented in the form of a System-On-a-Chip (SOC).
In view of the foregoing examples and preferred embodiments thereof, it will be appreciated by those skilled in the art that in actual operation, the technical concepts of the present invention may be applied to various embodiments, and the present invention is schematically illustrated by the following carriers:
(1) An ERP-fNIRS-based depression symptom characteristic identification device. The device may specifically include: one or more processors, memory, and one or more computer programs, wherein the one or more computer programs are stored in the memory, the one or more computer programs comprising instructions, which when executed by the device, cause the device to perform the steps/functions of the foregoing embodiments or equivalent implementations.
The electronic device may be an electronic device related to a computer, such as, but not limited to, various computing terminals, electronic products, and the like.
Specifically, the device/terminal may be a computer device, and the hardware structure of the computer device may further specifically include: at least one processor, at least one communication interface, at least one memory and at least one communication bus; the processor, the communication interface and the memory can all communicate with each other through a communication bus. The processor may be a central Processing unit CPU, DSP, microcontroller or digital signal processor, and may further include a GPU, an embedded neural network processor (Neural-network Process Units; hereinafter referred to as NPU) and an image signal processor (IMAGE SIGNAL Processing; hereinafter referred to as ISP), where the processor may further include an ASIC (application specific integrated circuit) or one or more integrated circuits configured to implement embodiments of the present invention, and the processor may further have a function of operating one or more software programs, where the software programs may be stored in a storage medium such as a memory; and the aforementioned memory/storage medium may include: nonvolatile Memory (Non-Volatile Memory), such as a Non-removable magnetic disk, a USB flash disk, a removable hard disk, an optical disk, and the like, and Read-Only Memory (ROM), random access Memory (Random Access Memory; RAM), and the like.
(2) A computer data storage medium having stored thereon a computer program or the above-mentioned means which, when executed, causes a computer to perform the steps/functions of the foregoing embodiments or equivalent implementations.
In several embodiments provided by the present invention, any of the functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer data storage medium. Based on such understanding, certain aspects of the present invention may be embodied in the form of a software product as described below, in essence, or as a part of, contributing to the prior art.
It is especially pointed out that the storage medium may refer to a server or a similar computer device, in particular, i.e. a storage means in the server or a similar computer device, in which the aforementioned computer program or the aforementioned means are stored.
(3) A computer program product (which may comprise the apparatus described above) which, when run on a terminal device, causes the terminal device to perform the method for identifying depression symptom characteristics based on ERP-fNIRS of the preceding example or equivalent implementation.
From the above description of embodiments, it will be apparent to those skilled in the art that all or part of the steps of the above described methods may be implemented in software plus necessary general purpose hardware platforms. Based on such understanding, the above-described computer program product may include, but is not limited to, an APP.
Claims (10)
1. A depression symptom characteristic identification method based on ERP-fNIRS, comprising:
a Flanker model for identifying depression symptoms is preset;
collecting synchronous multimodal data based on the Flanker paradigm and a change in blood flow rate in a prefrontal cortex region;
performing characteristic dimension reduction processing on the synchronous multi-mode data by utilizing OFSLLE algorithm;
and carrying out feature extraction on the synchronous multi-modal data subjected to dimension reduction under Euclidean structure and graph structure, and outputting classification results of depression symptoms after fusing the double-structure feature information.
2. The ERP-fNIRS-based depression symptom characteristic recognition method according to claim 1, wherein the preset Flanker paradigm for depression disorder recognition includes:
gray is adopted as a display background;
setting a blue fixation point at the center of the gray background;
after a preset interval time, presenting a central target arrow and left and right flank arrows for a preset duration;
Recording a feedback signal of the tested arrow to the center target during the set time period;
And similarly, continuously performing the test according to the preset presentation times of the central target arrows in different directions so as to complete the whole test times.
3. The ERP-fNIRS-based depression symptom characteristic identification method of claim 1, wherein the collecting synchronous multimodal data comprises:
Performing brain electrical stimulation based on the Flanker paradigm, and synchronously collecting a tested multichannel brain electrical signal and a multichannel original near infrared light intensity signal;
Preprocessing and analyzing the electroencephalogram signals to obtain space-time sequence data of N200 waveforms; converting the original near infrared light intensity signal into spatiotemporal sequence data for characterizing hemoglobin concentration variation of deoxyhemoglobin concentration variation using a modified beer-lambert law;
And carrying out time sequence alignment on the time-space sequence data of the N200 waveform and the time-space sequence data of the hemoglobin concentration change, and matching space information by using a space index to obtain the synchronous multi-modal data.
4. The method for identifying depression symptom characteristics based on ERP-fNIRS according to claim 1, wherein the feature dimension reduction processing of the synchronous multi-modal data by using OFSLLE algorithm comprises: the class information of the synchronous multimodal data for depression disorder identification is integrated into the LLE algorithm and an optimal projection matrix is found therefrom.
5. The method for identifying depression symptom characteristics based on ERP-fNIRS according to any one of claims 1 to 4, wherein the feature extraction of the dimension-reduced synchronous multi-modal data under euclidean structure and graph structure, and the fusion of the double-structure feature information, the output of the classification result of depression symptoms comprises:
Pre-constructing a double-structure model comprising a convolutional neural network and a graph convolution neural network;
Taking the synchronous multi-modal data after dimension reduction as the input of the double-structure model, and taking the classification result of the depression symptoms as the output of the double-structure model.
6. The ERP-fNIRS-based depression symptom characteristic identification method of claim 5, wherein the identification method further comprises: the convolutional neural network in the double-structure model introduces an attention mechanism and sequentially generates corresponding attention force diagrams from two dimensions of a channel and a space; attention is drawn to the neural pathway mechanisms and brain network connection characteristics of depressed patients from the synchronized multimodal data, in combination with the input of the model.
7. The ERP-fNIRS-based depression symptom characteristic recognition method according to claim 6, wherein:
Generating a channel attention pattern using the inter-channel relationship of the neuron activity features that change instantaneously under the conflicting stimulus;
generating a spatial attention pattern by utilizing the spatial relationship between neurons and blood flow change characteristics of the depressed forehead cortex, amygdala and hippocampus and brain network connection characteristics;
multiplying the channel attention pattern by the model input according to the bit to obtain the channel attention output characteristic;
And multiplying the spatial attention pattern with the channel attention output characteristic by bits to obtain the spatial attention output characteristic.
8. A depression symptom characteristic recognition device based on ERP-fNIRS, characterized in that a Flanker paradigm for recognition of depression symptoms is preset, the recognition device comprising:
The synchronous multi-mode data acquisition module is used for collecting synchronous multi-mode data based on the Flanker paradigm and the change of the blood flow rate of the forehead cortex region;
the feature dimension reduction module is used for performing feature dimension reduction processing on the synchronous multi-mode data by utilizing OFSLLE algorithm;
And the depression symptom identification module is used for extracting the characteristics of the synchronous multi-mode data subjected to dimension reduction under the Euclidean structure and the graph structure, fusing the double-structure characteristic information and outputting the classification result of depression symptoms.
9. The ERP-fNIRS-based depression symptom characteristic recognition apparatus of claim 8, wherein the depression symptom recognition module comprises: the model building unit is used for pre-building a double-structure model comprising a convolutional neural network and a graph convolution neural network; and the synchronous multi-modal data is used as the input of the double-structure model, and the classification result of the depression symptoms is used as the output of the double-structure model.
10. An ERP-fNIRS-based depression symptom characteristic recognition apparatus, comprising:
One or more processors, memory, and one or more computer programs, wherein the one or more computer programs are stored in the memory, the one or more computer programs comprising instructions, which when executed by the device, cause the device to perform the ERP-fNIRS-based depression symptom characteristic identification method of any of claims 1-7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410591420.5A CN118161166B (en) | 2024-05-14 | 2024-05-14 | Method and device for identifying depressive symptom characteristics based on ERP-fNIRS |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410591420.5A CN118161166B (en) | 2024-05-14 | 2024-05-14 | Method and device for identifying depressive symptom characteristics based on ERP-fNIRS |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN118161166A true CN118161166A (en) | 2024-06-11 |
| CN118161166B CN118161166B (en) | 2024-07-05 |
Family
ID=91357051
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410591420.5A Active CN118161166B (en) | 2024-05-14 | 2024-05-14 | Method and device for identifying depressive symptom characteristics based on ERP-fNIRS |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN118161166B (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119014870A (en) * | 2024-10-24 | 2024-11-26 | 慧创科仪(北京)科技有限公司 | A method for evaluating the executive function status of a subject based on fNIRS blood oxygen data and EEG data |
| CN119548114A (en) * | 2024-12-27 | 2025-03-04 | 北京理工大学 | Target population recognition method and device based on multi-level fusion of neural blood flow characteristics |
| CN119867754A (en) * | 2025-01-23 | 2025-04-25 | 成都大学 | Depression risk assessment method based on functional near infrared spectrum |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110119212A1 (en) * | 2008-02-20 | 2011-05-19 | Hubert De Bruin | Expert system for determining patient treatment response |
| US20200253527A1 (en) * | 2017-02-01 | 2020-08-13 | Conflu3Nce Ltd | Multi-purpose interactive cognitive platform |
| CN112417157A (en) * | 2020-12-15 | 2021-02-26 | 华南师范大学 | Emotion classification method of text attribute words based on deep learning network |
| CN112818892A (en) * | 2021-02-10 | 2021-05-18 | 杭州医典智能科技有限公司 | Multi-modal depression detection method and system based on time convolution neural network |
| CN113920442A (en) * | 2021-09-29 | 2022-01-11 | 中国人民解放军火箭军工程大学 | Hyperspectral classification method combining graph structure and convolutional neural network |
| CN117612710A (en) * | 2024-01-19 | 2024-02-27 | 合肥市第四人民医院 | A medical diagnosis assistance system based on EEG signal and artificial intelligence classification |
-
2024
- 2024-05-14 CN CN202410591420.5A patent/CN118161166B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110119212A1 (en) * | 2008-02-20 | 2011-05-19 | Hubert De Bruin | Expert system for determining patient treatment response |
| US20200253527A1 (en) * | 2017-02-01 | 2020-08-13 | Conflu3Nce Ltd | Multi-purpose interactive cognitive platform |
| CN112417157A (en) * | 2020-12-15 | 2021-02-26 | 华南师范大学 | Emotion classification method of text attribute words based on deep learning network |
| CN112818892A (en) * | 2021-02-10 | 2021-05-18 | 杭州医典智能科技有限公司 | Multi-modal depression detection method and system based on time convolution neural network |
| CN113920442A (en) * | 2021-09-29 | 2022-01-11 | 中国人民解放军火箭军工程大学 | Hyperspectral classification method combining graph structure and convolutional neural network |
| CN117612710A (en) * | 2024-01-19 | 2024-02-27 | 合肥市第四人民医院 | A medical diagnosis assistance system based on EEG signal and artificial intelligence classification |
Non-Patent Citations (2)
| Title |
|---|
| FRANK RICHTER ET AL: "The potential of substance P to initiate and perpetuate cortical spreading depression (CSD) in rat in vivo", 《SCI REP》, 31 December 2018 (2018-12-31) * |
| 于洪丽 等: "基于混合神经网络和注意力机制的卒中后抑郁早期筛查分类方法研究", 《电子测量技术》, vol. 46, no. 12, 30 June 2023 (2023-06-30) * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN119014870A (en) * | 2024-10-24 | 2024-11-26 | 慧创科仪(北京)科技有限公司 | A method for evaluating the executive function status of a subject based on fNIRS blood oxygen data and EEG data |
| CN119548114A (en) * | 2024-12-27 | 2025-03-04 | 北京理工大学 | Target population recognition method and device based on multi-level fusion of neural blood flow characteristics |
| CN119867754A (en) * | 2025-01-23 | 2025-04-25 | 成都大学 | Depression risk assessment method based on functional near infrared spectrum |
| CN119867754B (en) * | 2025-01-23 | 2026-01-06 | 成都大学 | A method for assessing the risk of depression based on functional near-infrared spectroscopy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN118161166B (en) | 2024-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gao et al. | Complex networks and deep learning for EEG signal analysis | |
| CN118161166A (en) | Method and device for identifying depressive symptom characteristics based on ERP-fNIRS | |
| CN114224342A (en) | Multi-channel electroencephalogram emotion recognition method based on space-time fusion feature network | |
| KR20190128978A (en) | Method for estimating human emotions using deep psychological affect network and system therefor | |
| Cui et al. | A multiview sparse dynamic graph convolution-based region-attention feature fusion network for major depressive disorder detection | |
| CN118902458B (en) | Multimodal emotion recognition and consciousness detection method based on electroencephalogram and micro-expression | |
| Guo et al. | A dual transfer learning method based on 3D-CNN and vision transformer for emotion recognition: Z. Guo et al. | |
| Zeng et al. | Adaptive node feature extraction in graph-based neural networks for brain diseases diagnosis using self-supervised learning | |
| Jain et al. | Electroencephalogram (EEG) based fuzzy logic and spiking neural networks (FLSNN) for advanced multiple neurological disorder diagnosis | |
| Zhang et al. | Multi-view graph contrastive learning via adaptive channel optimization for depression detection in EEG signals | |
| Zheng et al. | Deep multidilation temporal and spatial dependence modeling in stereoscopic 3-D EEG for visual discomfort assessment | |
| Deshpande et al. | Computational intelligence in neuroinformatics: Technologies and data analytics | |
| Senthil Kumar et al. | ResDense Fusion: enhancing schizophrenia disorder detection in EEG data through ensemble fusion of deep learning models | |
| Abbas et al. | Brain and Heart Rate Variability Patterns Recognition for Depression Classification of Mental Health Disorder. | |
| Korani et al. | Time frequency images as predictors for depressed patients’ respondent status to SSRI antidepressant | |
| Tolu et al. | Perspective on investigation of neurodegenerative diseases with neurorobotics approaches | |
| Hu et al. | SCANet: An Innovative Multiscale Selective Channel Attention Network for EEG-Based ADHD Recognition | |
| Wang et al. | A hybrid graph neural network for enhanced EEG-Based depression detection | |
| Dia et al. | Multichannel convolutional transformer for detecting mental disorders using electroancephalogrpahy records | |
| Daadaa | EmotionNet: Dissecting stress and anxiety through EEG-based deep learning approaches | |
| Chen et al. | A local-global graph convolutional network for depression recognition using EEG signals | |
| Yuan et al. | Discovery of shared latent nonlinear effective connectivity for EEG-based depression detection | |
| Li et al. | Multi-branch fusion graph neural network based on multi-head attention for childhood seizure detection | |
| Kumar et al. | Integrating physiological signals with dynamical attention networks for personality trait analysis | |
| Jridi et al. | Explainable ai-driven EEG channel selection for accurate and interpretable epilepsy diagnosis |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |