CN110110855A - Based on deep-cycle neural network and there is the brain network reconstruction method for supervising dictionary learning - Google Patents
Based on deep-cycle neural network and there is the brain network reconstruction method for supervising dictionary learning Download PDFInfo
- Publication number
- CN110110855A CN110110855A CN201910452205.6A CN201910452205A CN110110855A CN 110110855 A CN110110855 A CN 110110855A CN 201910452205 A CN201910452205 A CN 201910452205A CN 110110855 A CN110110855 A CN 110110855A
- Authority
- CN
- China
- Prior art keywords
- brain
- network
- dictionary
- matrix
- deep
- 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; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Computational Mathematics (AREA)
- Mathematical Analysis (AREA)
- Mathematical Optimization (AREA)
- General Engineering & Computer Science (AREA)
- Software Systems (AREA)
- Pure & Applied Mathematics (AREA)
- Computing Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Databases & Information Systems (AREA)
- Molecular Biology (AREA)
- Algebra (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Operations Research (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Probability & Statistics with Applications (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The present invention relates to a kind of based on deep-cycle neural network and has the brain network reconstruction method of supervision dictionary learning, it is the novel brain network reconfiguration calculation method of a kind of Fusion Model driving and data-driven method advantage, for the functional brain network diversified and concurrent from the reconstruct of task state fMRI data.Specifically, diversification, adaptive regression variable are derived automatically from using deep-cycle neural network;The brain function network for utilizing regression variable reconstruction tasks to cause using there is supervision dictionary learning method.The invention proposes the regression variables that a kind of deep-cycle neural network carrys out automatic learning data driving.Later, the brain network activation figure of these regression variables is reconstructed using effective supervision dictionary learning and sparse representation method.Experiments have shown that this calculation method has superiority in identification diversification and complicated with brain network facet.
Description
Technical field
The invention belongs to field of medical image processing, are related to one kind based on deep-cycle neural network and have supervision dictionary
The brain network reconstruction method of habit can be applied to brain function activity analysis.
Background technique
Task state functional mri (tfMRI) is a kind of to measure the caused blood of neuron activity using the magnetic radiography that shakes
The emerging Brain imaging techniques that hydraulic power changes.Because of its spatial resolution height, Noninvasive, there is no the features such as radioactive exposure,
It is widely applied in terms of brain function activity analysis and clinical diagnosis.To task state fMRI data midbrain network
The detection and reconstruct of ingredient are that one of main research of task state Functional magnetic resonance imaging and cerebral function are living
The basis of dynamic analysis and clinical application.
Although there are many traditional brain network reconstruction method, the primary analysis method of task state fMRI data
Model-driven, that is, general linear model (GLM).The basic thought of this model driven method is using from blood
The regression variable of the hypothesis of kinetic reaction function (HRF) and its derivative carrys out the brain function network of reconstruction tasks initiation.But this
One main problem of class method is regression variable over-simplification, lacks adaptability, does not also account for task state functional MRI
The sequence characteristic of data also has ignored other diversifications, concurrent brain activity network.Therefore, it is necessary to be directed to functional MRI
The characteristics of data and brain activity, carries out the research of new brain network reconstruction method.
Current already present brain network reconfiguration calculation method has the disadvantage in that regression variable over-simplification, lacks pervasive
Property, cause many diversifications, concurrent brain activity network ignored.
Summary of the invention
Technical problems to be solved
In order to avoid the shortcomings of the prior art, the present invention proposes one kind based on deep-cycle neural network and has supervision
The brain network reconstruction method of dictionary learning.
Technical solution
It is a kind of based on deep-cycle neural network and have supervision dictionary learning brain network reconstruction method, it is characterised in that step
It is rapid as follows:
Step 1, tectonic network model: network model is deep-cycle neural network DRNN, including two circulation nervous layers
RNN and full articulamentum FC;Two circulation nervous layer RNN and a full articulamentum FC form cascaded structure;It is described each
RNN layers include 30 LSTM unit, that is, shot and long term memory networks;
The deep-cycle neural network DRNN constructed in step 2, training step 1: with the task matrix in tfMRI data set
(T) input as deep-cycle neural network DRNN, network model are trained to model convergence, and RNN layers at top
In the output of each unit obtain the regression variable of data-driven, the full brain signal matrix S predicted after full articulamentum;
Step 3: using full brain signal matrix S as the input for the dictionary learning for having supervision, the dictionary learning for having supervision is carried out,
When study, by a part of D in dictionary DcIt is not involved in update as constant, only to rest part DlIt optimizes, obtains dictionary D
And coefficient matrices A;
Step 4: every a line of matrix A being mapped in the standard form of brain, generation is led on motor task by DRNN
The brain network activation figure of data-driven out completes the reconstruct of space brain network.
The deep-cycle neural network DRNN constructed in the training step 1 of the step 2 is: by the stimulation square at each moment
Battle array is straightened as vector, and the stimulus vector of t time is spliced into matrix T, and size is n × t, as the defeated of DRNN network
Enter;It exports to obtain the regression variable for the brain activity for representing specific time in each unit of the RNN layer at top, connect entirely
Layer, obtains the tfMRI signal matrix S of full brain;The vector is dimension n × 1;The tfMRI signal matrix S dimension of the full brain is
M × t, wherein m is voxel number, and t is the time.
1. the step 3: the tfMRI signal S of full brain is that the sparse linear of the atom of basic dictionary D combines, every in dictionary D
The signal of a dictionary atom represents the functional activity of specific brain regions network, and corresponding weight vectors represent the brain network in matrix A
Spatial distribution;Dictionary atom is divided into two parts:
Wherein DCAs the dictionary atom of predefined model-driven, DlFor by the dictionary atom of tfMRI data-driven, and
Only to DlIt optimizes, steps are as follows:
Step a: input signalWherein
D0For initial dictionary matrix, DCFor predefined dictionary atom, Dl0Random initializtion, T are cycle-index;
Step b: circulation starts, cycle-index iter=1:T;I=iter%n wherein T > n;
Step c: s is extracted from signal Si;
Step d: sparse coding:
Step e: D is updatedl(t), but DCIt remains unchanged
Step f: end loop returns to D and A matrix.
Beneficial effect
It is proposed by the present invention it is a kind of based on deep-cycle neural network and have supervision dictionary learning brain network reconstruction method,
It is the novel brain network reconfiguration calculation method of a kind of Fusion Model driving and data-driven method advantage, is used for from task state function
Magnetic resonance imaging data reconstruct diversification and concurrent functional brain network.Specifically, certainly using deep-cycle neural network
Dynamic export diversification, adaptive regression variable;Regression variable reconstruction tasks are utilized to draw using there is supervision dictionary learning method
The brain function network of hair.
The invention proposes the regression variables that a kind of deep-cycle neural network carrys out automatic learning data driving.Later, it adopts
The brain network activation figure of these regression variables is reconstructed with effective supervision dictionary learning and sparse representation method.Experiments have shown that this
Calculation method has superiority in identification diversification and complicated with brain network facet.
Firstly, compared with model-driven is as general linear model (GLM) this conventional method of regression variable, the present invention
The method of proposition uses data-driven and exports regression variable.Secondly, in the part using supervision dictionary learning reconstruct brain network,
It, partially as the constant in dictionary atom, will only optimize other dictionary atoms as derived from DRNN, sufficiently show brain network and exist
Specific activities under task stimulation, and can also identify many other concurrent cerebral function networks, such as different time
Delay network illustrates that this model has robustness and superiority.
Detailed description of the invention
Fig. 1: it show the flow chart of the method for the present invention.The present invention is first using task stimulation matrix T as input, by two
A RNN layers and a full articulamentum, export full brain tfMRI signal matrix S.Dictionary learning by there is supervision obtain dictionary D and
Every a line of A matrix is reconstructed the space brain network under task stimulation by coefficient matrices A.
Data-driven regression variable derived from the DRNN of part under Fig. 2: Motor task
Fig. 3: the individual of the regression variable reconstruct of the data-driven as derived from DRNN and the example of group roomage response
Fig. 4: the brain cyberspace for the regression variable reconstruct that different time postpones under Motor task design
Specific embodiment
Now in conjunction with embodiment, attached drawing, the invention will be further described:
Step 1: choosing data set.HCP data set is considered as one of most system and most comprehensive neuroimaging data set.
Therefore by taking HCP data set as an example, brain network reconfiguration of the calculation method under task stimulation in the verifying present invention has exploitativeness.
Step 2: construction depth Recognition with Recurrent Neural Network model (DRNN).Network model DRNN in this experiment is by two RNN
Layer and a full articulamentum composition, are wherein made of for each DRNN layers 30 LSTM units.Two RNN layers and one full connections
Layer forms cascaded structure.
Step 3: deep-cycle neural network (DRNN) being instructed using task stimulation matrix (T) in HCP data set
Practice.
The input of step 3a:DRNN network is that task stimulates matrix Task Design (T).By the stimulation square at each moment
Battle array is straightened as vector (dimension n × 1), and the stimulus vector of a period of time is spliced into matrix T, and size is n × t, as
The input of DRNN network.
The output of step 3b:DRNN network is full brain signal Whole Brain Signals (S).The RNN layer at top it is every
A unit output represents the brain activity of specific time, provides input as regression variable for next step analysis.It is complete by one
Articulamentum obtains the tfMRI signal matrix S of full brain (matrix dimensionality is m × t, and wherein m is voxel number, and t is the time).
Step 4: using full brain signal matrix S obtained in step 3 as input, carrying out the dictionary learning for having supervision.
Step 4a: full brain tfMRI signal matrix S can sparsely be indicated by dictionary D.In dictionary learning, the tfMRI of full brain
Signal S can be counted as the sparse linear combination of the atom of basic dictionary D.For example, si=D × Ai, S=D × A, wherein A is dilute
Dredge the correlation coefficient matrix indicated.Particularly, the function that the signal of each dictionary atom represents specific brain regions network in dictionary D is lived
Dynamic, corresponding weight vectors represent the spatial distribution of the brain network in matrix A.
Step 4b: in the present invention, dictionary atom is particularly divided into two parts, wherein DCAs predefined model-driven
Dictionary atom, DlFor by the dictionary atom of tfMRI data-driven.In this calculation method, only to DlIt optimizes.
For signal S, cost function is defined as:
Wherein λ is regularization parameter.In short, the problem of supervision dictionary learning, can be rewritten as the matrix point in (3-4)
Solution problem:
Calculation method circulation step is as follows:
Step a: input signalWherein
D0For initial dictionary matrix, DCFor predefined dictionary atom, Dl0Random initializtion, T are cycle-index.
Step b: circulation starts, cycle-index iter=1:T;I=iter%n wherein T > n;
Step c: s is extracted from signal Si;
Step d: sparse coding:
Step e: D is updatedl(t), but DCIt remains unchanged
Step f: end loop returns to D and A matrix.
Identical with the partitioning scheme of dictionary D, coefficient matrices A is equally resorted to two parts, returns and becomes as derived from DRNN
Brain cyberspace A under the task stimulation of amount reconstructCAnd the complicated with brain cyberspace of data-driven is distributed Al。
Step 5: being gone out on missions using the reconstruct of coefficient matrices A obtained in step 4 stimulates the brain network caused.By coefficient square
Every a line of matrix A is mapped in the standard form of brain by battle array A as input, and generation is exported on motor task by DRNN
Data-driven brain network activation figure, to realize the reconstruct work of space brain network.
Claims (3)
1. it is a kind of based on deep-cycle neural network and have supervision dictionary learning brain network reconstruction method, it is characterised in that step
It is as follows:
Step 1, tectonic network model: network model is deep-cycle neural network DRNN, including two circulation nervous layer RNN with
One full articulamentum FC;Two circulation nervous layer RNN and a full articulamentum FC form cascaded structure;Described each RNN layers
Including 30 LSTM unit, that is, shot and long term memory networks;
The deep-cycle neural network DRNN constructed in step 2, training step 1: with the task matrix (T) in tfMRI data set
As the input of deep-cycle neural network DRNN, network model is trained to model convergence, in top RNN layers
The output of each unit obtains the regression variable of data-driven, the full brain signal matrix S predicted after full articulamentum;
Step 3: using full brain signal matrix S as the input for the dictionary learning for having supervision, carrying out the dictionary learning for having supervision, learn
When, by a part of D in dictionary DcIt is not involved in update as constant, only to rest part DlIt optimizes, obtain dictionary D and is
Matrix number A;
Step 4: every a line of matrix A being mapped in the standard form of brain, is generated on motor task as derived from DRNN
The brain network activation figure of data-driven completes the reconstruct of space brain network.
2. based on deep-cycle neural network and there is the brain network reconstruction method for supervising dictionary learning according to claim 1,
It is characterized by: the deep-cycle neural network DRNN constructed in the training step 1 of the step 2 is: by the thorn at each moment
Sharp matrix is straightened as vector, and the stimulus vector of t time is spliced into matrix T, and size is n × t, as DRNN network
Input;It exports to obtain the regression variable for the brain activity for representing specific time in each unit of the RNN layer at top, connect entirely
Layer, obtains the tfMRI signal matrix S of full brain;The vector is dimension n × 1;The tfMRI signal matrix S dimension of the full brain is
M × t, wherein m is voxel number, and t is the time.
3. based on deep-cycle neural network and there is the brain network reconstruction method for supervising dictionary learning according to claim 1,
It is characterized by: the step 3: the tfMRI signal S of full brain is that the sparse linear of the atom of basic dictionary D combines, in dictionary D
The signal of each dictionary atom represents the functional activity of specific brain regions network, and corresponding weight vectors represent the brain network in matrix A
Spatial distribution;Dictionary atom is divided into two parts:
Wherein DCAs the dictionary atom of predefined model-driven, DlFor by the dictionary atom of tfMRI data-driven, and it is only right
DlIt optimizes, steps are as follows:
Step a: input signalWherein D0For
Initial dictionary matrix, DCFor predefined dictionary atom, Dl0Random initializtion, T are cycle-index;
Step b: circulation starts, cycle-index iter=1:T;I=iter%n wherein T > n;
Step c: s is extracted from signal Si;
Step d: sparse coding:
Step e: D is updatedl(t), but DCIt remains unchanged
Step f: end loop returns to D and A matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910452205.6A CN110110855B (en) | 2019-05-28 | 2019-05-28 | Brain network reconstruction method based on deep cycle neural network and supervised dictionary learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910452205.6A CN110110855B (en) | 2019-05-28 | 2019-05-28 | Brain network reconstruction method based on deep cycle neural network and supervised dictionary learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110110855A true CN110110855A (en) | 2019-08-09 |
CN110110855B CN110110855B (en) | 2022-04-29 |
Family
ID=67492686
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910452205.6A Active CN110110855B (en) | 2019-05-28 | 2019-05-28 | Brain network reconstruction method based on deep cycle neural network and supervised dictionary learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110110855B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112419437A (en) * | 2019-11-29 | 2021-02-26 | 上海联影智能医疗科技有限公司 | System and method for reconstructing magnetic resonance images |
Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102138782A (en) * | 2011-03-10 | 2011-08-03 | 电子科技大学 | Brain functional effective connection analyzing method |
CN102366323A (en) * | 2011-09-30 | 2012-03-07 | 中国科学院自动化研究所 | Detection method for causal connection strength of magnetic resonance brain imaging based on PCA (Principal component analysis) and GCA (Granger causality analysis) |
CN103034778A (en) * | 2012-09-28 | 2013-04-10 | 中国科学院自动化研究所 | Method for extracting brain function network of individual based on analysis of multiple tested brain function data |
CN103325119A (en) * | 2013-06-27 | 2013-09-25 | 中国科学院自动化研究所 | Default state brain network center node detecting method based on modality fusion |
CN103345749A (en) * | 2013-06-27 | 2013-10-09 | 中国科学院自动化研究所 | Method for detecting brain network function connectivity lateralization based on modality fusion |
CN105596004A (en) * | 2015-12-28 | 2016-05-25 | 中国人民解放军国防科学技术大学 | Brain functional magnetic resonance imaging blind source separation method based on group canonical correlation analysis |
CN106997581A (en) * | 2017-03-01 | 2017-08-01 | 杭州电子科技大学 | A kind of method that utilization deep learning rebuilds high spectrum image |
CN107658018A (en) * | 2017-10-12 | 2018-02-02 | 太原理工大学 | A kind of fusion brain network establishing method based on structure connection and function connects |
CN108577835A (en) * | 2018-05-17 | 2018-09-28 | 太原理工大学 | A kind of brain function network establishing method based on micro- state |
CN108921286A (en) * | 2018-06-29 | 2018-11-30 | 杭州电子科技大学 | A kind of tranquillization state function brain network establishing method for exempting from threshold value setting |
CN109065128A (en) * | 2018-09-28 | 2018-12-21 | 郑州大学 | A kind of sparse brain network establishing method of weighted graph regularization |
CN109741806A (en) * | 2019-01-07 | 2019-05-10 | 北京推想科技有限公司 | A kind of Medical imaging diagnostic reports auxiliary generating method and its device |
-
2019
- 2019-05-28 CN CN201910452205.6A patent/CN110110855B/en active Active
Patent Citations (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102138782A (en) * | 2011-03-10 | 2011-08-03 | 电子科技大学 | Brain functional effective connection analyzing method |
CN102366323A (en) * | 2011-09-30 | 2012-03-07 | 中国科学院自动化研究所 | Detection method for causal connection strength of magnetic resonance brain imaging based on PCA (Principal component analysis) and GCA (Granger causality analysis) |
CN103034778A (en) * | 2012-09-28 | 2013-04-10 | 中国科学院自动化研究所 | Method for extracting brain function network of individual based on analysis of multiple tested brain function data |
CN103325119A (en) * | 2013-06-27 | 2013-09-25 | 中国科学院自动化研究所 | Default state brain network center node detecting method based on modality fusion |
CN103345749A (en) * | 2013-06-27 | 2013-10-09 | 中国科学院自动化研究所 | Method for detecting brain network function connectivity lateralization based on modality fusion |
CN105596004A (en) * | 2015-12-28 | 2016-05-25 | 中国人民解放军国防科学技术大学 | Brain functional magnetic resonance imaging blind source separation method based on group canonical correlation analysis |
CN106997581A (en) * | 2017-03-01 | 2017-08-01 | 杭州电子科技大学 | A kind of method that utilization deep learning rebuilds high spectrum image |
CN107658018A (en) * | 2017-10-12 | 2018-02-02 | 太原理工大学 | A kind of fusion brain network establishing method based on structure connection and function connects |
CN108577835A (en) * | 2018-05-17 | 2018-09-28 | 太原理工大学 | A kind of brain function network establishing method based on micro- state |
CN108921286A (en) * | 2018-06-29 | 2018-11-30 | 杭州电子科技大学 | A kind of tranquillization state function brain network establishing method for exempting from threshold value setting |
CN109065128A (en) * | 2018-09-28 | 2018-12-21 | 郑州大学 | A kind of sparse brain network establishing method of weighted graph regularization |
CN109741806A (en) * | 2019-01-07 | 2019-05-10 | 北京推想科技有限公司 | A kind of Medical imaging diagnostic reports auxiliary generating method and its device |
Non-Patent Citations (3)
Title |
---|
TAO HU等: "《Electron Microscopy Reconstruction of Brain Structure Using Sparse Representations Over Learned Dictionaries》", 《IEEE TRANSACTIONS ON MEDICAL IMAGING》 * |
YIMIN ZHOU等: "《Distributed Dimensionality Reconstruction Algorithm for High Dimensional Data in Internet of Brain Things》", 《IEEE ACCESS》 * |
胡颖等: "《静息态功能磁共振成像的脑功能分区综述》", 《中国图象图形学报》 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112419437A (en) * | 2019-11-29 | 2021-02-26 | 上海联影智能医疗科技有限公司 | System and method for reconstructing magnetic resonance images |
Also Published As
Publication number | Publication date |
---|---|
CN110110855B (en) | 2022-04-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Li et al. | A multi-scale fusion convolutional neural network based on attention mechanism for the visualization analysis of EEG signals decoding | |
D’Angelo et al. | Realistic modeling of neurons and networks: towards brain simulation | |
CA2880758C (en) | Method and computing system for modelling a primate brain | |
Kasabov et al. | New algorithms for encoding, learning and classification of fMRI data in a spiking neural network architecture: A case on modeling and understanding of dynamic cognitive processes | |
CN110619322A (en) | Multi-lead electrocardio abnormal signal identification method and system based on multi-flow convolution cyclic neural network | |
CN108135520B (en) | Generating natural language representations of psychological content from functional brain images | |
US10726546B2 (en) | Tissue-to-flow image generation in medical imaging | |
CN110090017B (en) | Electroencephalogram signal source positioning method based on LSTM | |
CN108542390A (en) | Vascular plaque ingredient recognition methods based on more contrast nuclear magnetic resonance images | |
Bzdok et al. | Semi-supervised factored logistic regression for high-dimensional neuroimaging data | |
Rejer | EEG feature selection for BCI based on motor imaginary task | |
CN109816630A (en) | FMRI visual coding model building method based on transfer learning | |
He et al. | Alzheimer's disease diagnosis model based on three-dimensional full convolutional DenseNet | |
CN107248180A (en) | A kind of fMRI natural image coding/decoding methods based on hidden state model | |
CN110321827A (en) | A kind of pain level appraisal procedure based on face pain expression video | |
Qiang et al. | Learning brain representation using recurrent Wasserstein generative adversarial net | |
CN113180695B (en) | Brain-computer interface signal classification method, system, equipment and storage medium | |
CN110110855A (en) | Based on deep-cycle neural network and there is the brain network reconstruction method for supervising dictionary learning | |
CN110392549A (en) | Determine the method and apparatus for causing the best big brain stimulation of expected behavior | |
Qiao et al. | Cheart: A conditional spatio-temporal generative model for cardiac anatomy | |
DE102019213931A1 (en) | Method and computer program product for identifying a vehicle user and control device for automated driving functions | |
Shpigelman et al. | Spikernels: Embedding spiking neurons in inner-product spaces | |
Li et al. | Latent source mining of fMRI data via deep belief network | |
Zhang et al. | Deep Linear Modeling of Hierarchical Functional Connectivity in the Human Brain | |
Ma | Neurophysiological analysis of the genesis mechanism of EEG during the interictal and ictal periods using a multiple neural masses model |
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 |