CN112002428A - Whole brain individualized brain function map construction method taking independent component network as reference - Google Patents
Whole brain individualized brain function map construction method taking independent component network as reference Download PDFInfo
- Publication number
- CN112002428A CN112002428A CN202010856401.2A CN202010856401A CN112002428A CN 112002428 A CN112002428 A CN 112002428A CN 202010856401 A CN202010856401 A CN 202010856401A CN 112002428 A CN112002428 A CN 112002428A
- Authority
- CN
- China
- Prior art keywords
- brain
- network
- functional
- sub
- voxel
- 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
- 210000004556 brain Anatomy 0.000 title claims abstract description 156
- 230000003925 brain function Effects 0.000 title claims abstract description 52
- 238000010276 construction Methods 0.000 title claims abstract description 20
- 238000000034 method Methods 0.000 claims abstract description 61
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims abstract description 51
- 238000012880 independent component analysis Methods 0.000 claims abstract description 51
- 230000006870 function Effects 0.000 claims abstract description 33
- 230000000284 resting effect Effects 0.000 claims abstract description 16
- 238000012804 iterative process Methods 0.000 claims abstract description 9
- 238000009826 distribution Methods 0.000 claims description 29
- 239000011159 matrix material Substances 0.000 claims description 20
- 230000008569 process Effects 0.000 claims description 18
- 238000004458 analytical method Methods 0.000 claims description 17
- 238000000513 principal component analysis Methods 0.000 claims description 15
- 238000004364 calculation method Methods 0.000 claims description 14
- 238000007781 pre-processing Methods 0.000 claims description 14
- 238000012937 correction Methods 0.000 claims description 10
- 238000004422 calculation algorithm Methods 0.000 claims description 9
- 230000009467 reduction Effects 0.000 claims description 8
- 230000011218 segmentation Effects 0.000 claims description 8
- 238000012935 Averaging Methods 0.000 claims description 7
- 238000009499 grossing Methods 0.000 claims description 7
- 238000012360 testing method Methods 0.000 claims description 6
- 210000003484 anatomy Anatomy 0.000 claims description 4
- 238000012847 principal component analysis method Methods 0.000 claims description 4
- 210000001638 cerebellum Anatomy 0.000 claims description 2
- 238000007405 data analysis Methods 0.000 claims description 2
- 230000007547 defect Effects 0.000 claims description 2
- 238000009792 diffusion process Methods 0.000 claims description 2
- 238000003709 image segmentation Methods 0.000 claims description 2
- 208000000474 Poliomyelitis Diseases 0.000 claims 1
- 238000010606 normalization Methods 0.000 claims 1
- 210000004940 nucleus Anatomy 0.000 claims 1
- 230000002123 temporal effect Effects 0.000 claims 1
- 201000010099 disease Diseases 0.000 abstract description 4
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 abstract description 4
- 238000003384 imaging method Methods 0.000 abstract description 4
- 208000027418 Wounds and injury Diseases 0.000 abstract description 2
- 230000006378 damage Effects 0.000 abstract description 2
- 208000014674 injury Diseases 0.000 abstract description 2
- 230000007246 mechanism Effects 0.000 abstract description 2
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000001054 cortical effect Effects 0.000 description 3
- 230000036039 immunity Effects 0.000 description 3
- 238000002595 magnetic resonance imaging Methods 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 230000002490 cerebral effect Effects 0.000 description 2
- 230000003920 cognitive function Effects 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 210000004884 grey matter Anatomy 0.000 description 2
- 230000008035 nerve activity Effects 0.000 description 2
- 230000001537 neural effect Effects 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 238000000926 separation method Methods 0.000 description 2
- 238000007619 statistical method Methods 0.000 description 2
- 102000001690 Factor VIII Human genes 0.000 description 1
- 108010054218 Factor VIII Proteins 0.000 description 1
- 238000005481 NMR spectroscopy Methods 0.000 description 1
- 238000010220 Pearson correlation analysis Methods 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 230000007177 brain activity Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 230000019771 cognition Effects 0.000 description 1
- 230000001149 cognitive effect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000004886 head movement Effects 0.000 description 1
- 239000013315 hypercross-linked polymer Substances 0.000 description 1
- 230000005415 magnetization Effects 0.000 description 1
- 210000000653 nervous system Anatomy 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 238000005192 partition Methods 0.000 description 1
- 230000001737 promoting effect Effects 0.000 description 1
- 238000004445 quantitative analysis Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 208000024891 symptom Diseases 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H50/00—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics
- G16H50/70—ICT specially adapted for medical diagnosis, medical simulation or medical data mining; ICT specially adapted for detecting, monitoring or modelling epidemics or pandemics for mining of medical data, e.g. analysing previous cases of other patients
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/05—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves
- A61B5/055—Detecting, measuring or recording for diagnosis by means of electric currents or magnetic fields; Measuring using microwaves or radio waves involving electronic [EMR] or nuclear [NMR] magnetic resonance, e.g. magnetic resonance imaging
-
- A—HUMAN NECESSITIES
- A61—MEDICAL OR VETERINARY SCIENCE; HYGIENE
- A61B—DIAGNOSIS; SURGERY; IDENTIFICATION
- A61B5/00—Measuring for diagnostic purposes; Identification of persons
- A61B5/40—Detecting, measuring or recording for evaluating the nervous system
- A61B5/4058—Detecting, measuring or recording for evaluating the nervous system for evaluating the central nervous system
- A61B5/4064—Evaluating the brain
-
- 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
-
- 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
- A61B5/7203—Signal processing specially adapted for physiological signals or for diagnostic purposes for noise prevention, reduction or removal
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F18/2134—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on separation criteria, e.g. independent component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR 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
- G06F18/2135—Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods based on approximation criteria, e.g. principal component analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Life Sciences & Earth Sciences (AREA)
- Physics & Mathematics (AREA)
- Public Health (AREA)
- Medical Informatics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Pathology (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Animal Behavior & Ethology (AREA)
- Surgery (AREA)
- Molecular Biology (AREA)
- Veterinary Medicine (AREA)
- Heart & Thoracic Surgery (AREA)
- Biophysics (AREA)
- General Physics & Mathematics (AREA)
- Physiology (AREA)
- Signal Processing (AREA)
- Neurology (AREA)
- Psychiatry (AREA)
- Bioinformatics & Cheminformatics (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Multimedia (AREA)
- Primary Health Care (AREA)
- Epidemiology (AREA)
- Psychology (AREA)
- Neurosurgery (AREA)
- High Energy & Nuclear Physics (AREA)
- Databases & Information Systems (AREA)
- Radiology & Medical Imaging (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention is a whole brain individualized brain function map construction method taking an independent component network as a reference, the method utilizes individual tested brain resting state fMRI data, firstly introduces an independent component analysis method to construct a group level brain function sub-network, then reversely reconstructs each tested brain function sub-network and a characteristic time sequence corresponding to the function sub-network by utilizing space-time regression, and takes the characteristic time sequence corresponding to the function sub-network as a reference signal; and introducing an inverse distance weighting coefficient, sub-network inverse variation coefficient weighting, a correlation factor and an iterative process, and further obtaining the whole brain individualized function map with the independent component network as a reference. The method comprises the following steps: pure data driving, complete correspondence of brain areas, full brain coverage, more flexible subdivision of functional brain areas and the like, and provides a more accurate objective imaging tool for researching normal brain operation mechanism and disease-related brain function injury.
Description
The technical field is as follows:
the invention relates to the technical field of brain function map construction, in particular to a whole brain individualized brain function map construction method which is given by using resting-state functional magnetic resonance imaging (rs-fMRI) data and takes an independent component network as a reference.
Background art:
the human brain is one of the most mysterious, complex nervous systems, itself containing approximately a hundred million neurons. Moreover, the neurons are interconnected to form a huge neural network. However, researchers still have limited knowledge of this. Therefore, if it is desired to know the complex network, it is most important to know the nodes constituting the network first, which requires to construct a more detailed, more accurate and more adaptive human brain map for brain science research.
Functional magnetic resonance imaging (fMRI) based on Blood Oxygen Level Dependent (BOLD) signals is increasingly widely used in brain science because it can monitor nerve activity of each brain region of the human brain non-invasively. The task state fMRI can induce the neural activity of the corresponding brain region through designing a specific task, and then positions the brain region dominated by the cognitive function, and is widely applied to the functional subregion segmentation aspects such as vision, motion, language and the like. However, task-state fMRI requires very elaborate tasks to make it possible to accurately segment a functional region; in addition, most areas of the brain are not functionally known, and therefore task-state fMRI cannot define all potential functional subregions of the whole brain. In addition to task-state fMRI, studies have shown that fluctuations in BOLD signals in the human brain at rest can also characterize spontaneous brain activity. When a human brain processes a specific task, a plurality of brain areas need to cooperate together to form a functional network. The synchronism of the BOLD signal between brain regions can characterize the synchronism of nerve activity of each functional brain region of human brain, which is called Functional Connectivity (FC). Similar to anatomical connection, the functional subregions can be identified and segmented according to the similarity of functional connection modes among the brain regions, so that a whole brain function map based on functional connection is constructed.
The human brain, as the neural basis for individual cognition and behavior, has significant individual differences in structure and function. Numerous studies have found that brain functional connectivity in both adult populations and infants is significantly subject to individual differences and brain region specificity, i.e., low individual differences between primary cortical individuals and high individual differences between combined cortical brain regions. The traditional brain atlas is mainly used for analyzing the brain atlas based on group information, functional data of an individual human subject is projected to an average brain template according to the structure of the brain, and image characteristics of different human subjects are compared on the basis. The data registration of the group 'template' based on the anatomical structure or the functional information does not fully consider the distribution difference of brain functional subregions among individuals, so that the data comparison among different testees lacks the functional correspondence, statistical efficiency and accuracy in the true sense. Drawing a functional brain map on an individual level provides not only information on the connection strength between brain regions, but also information on the topology of the brain regions (e.g., brain region shape, size, etc.). These will facilitate the study of dynamic changes in the network of individualized functions, which are also essential for personalized medicine. Therefore, it is necessary to construct an individualized brain function spectrogram to more accurately characterize individual brain function activities and further mine the relation between the individual brain function activities and cognitive behaviors/disease symptoms, which will probably fundamentally improve the evaluation and diagnosis of brain cognitive functions of healthy subjects and disease patients, and has great significance for promoting individualized medical services.
Therefore, the invention provides a whole brain individualized function map construction method taking an independent component network as a reference for the first time. The method firstly uses the independent component network at the group level as a reference and carries out the inverse reconstruction of each network component at the individual level to obtain an individualized independent component network. And then, combining the inverse distance weighting, the independent component network variation coefficient and the function connection similarity of the independent component characteristic vector and the voxel to construct an individualized brain function map.
The invention content is as follows:
the invention innovatively provides a whole-brain individualized brain function map construction method taking an independent component network as a reference. The method comprises the steps of utilizing resting state fMRI data of a brain of an individual to be tested, firstly introducing a pure data-driven independent component analysis method to construct a group-level brain function sub-network, and determining spatial component characteristics (namely spatial distribution weight) of the individual-level brain function sub-network by taking the group-level brain function sub-network as reference; then, in order to avoid the situation that the boundaries of adjacent sub-networks at the individual level are possibly overlapped and mistakenly divided, the invention firstly develops an inverse distance weighting algorithm, a sub-network variation weighting algorithm and a voxel function connection-based iterative algorithm, and combines the spatial component characteristics of the sub-networks at the individual level brain function to determine the function boundary of each sub-network at the individual level; and finally, constructing a brain function map covering the whole brain at an individual level according to the function boundary. The method comprises the following steps: pure data driving, complete correspondence of brain areas, full brain coverage, more flexible subdivision of functional brain areas and the like, and provides a more accurate objective imaging tool for researching normal brain operation mechanism and disease-related brain function injury.
In order to achieve the purpose, the invention specifically adopts the technical scheme that:
a whole brain individualized brain function map construction method taking an independent component network as a reference is characterized in that the method utilizes individual brain resting state fMRI data to be tested, firstly introduces an independent component analysis method to construct a group level brain function sub-network, then reversely reconstructs each tested brain function sub-network and a characteristic time sequence corresponding to the function sub-network by utilizing space-time regression, and takes the characteristic time sequence corresponding to the function sub-network as a reference signal;
and introducing an inverse distance weighting coefficient, sub-network inverse variation coefficient weighting, a correlation factor and an iterative process, and further obtaining the whole brain individualized function map with the independent component network as a reference.
The specific process of the whole brain individualized brain function map construction method taking the independent component network as the reference comprises the following steps:
(1) data acquisition and preprocessing: acquiring brain fMRI data of each tested resting state, preprocessing the acquired brain fMRI data, ensuring that different tested brain fMRI data are aligned to the same standard space, ensuring that the anatomical structures of all tested voxels are the same, and acquiring the fMRI data of each tested and the time sequence of the tested voxels;
(2) group level and individual independent component network reconstruction: after all the preprocessed fMRI data are connected in series according to the time direction, a group of horizontal space ICA analyses are carried out, which mainly comprises the following steps: (i) carrying out dimensionality reduction on each preprocessed fMRI data to be tested in a time point direction by adopting principal component analysis; (ii) connecting a plurality of tested dimension-reduced functional data in series, and performing PCA dimension reduction on the connected data again; (iii) carrying out spatial ICA analysis on the reduced PCA data to obtain horizontal ICA spatial components; (iv) performing space-time regression on the individual fMRI data processed by the independent component analysis algorithm based on the group horizontal ICA space components to obtain the space component weight of each individual functional subnetwork, the whole brain voxel value and the characteristic time sequence of each individual functional subnetwork;
(3) constructing an individual functional brain map: introducing inverse distance weighting coefficients, sub-network inverse variation coefficient weighting, correlation factors and an iterative process to further obtain a whole brain individualized function map with an independent component network as a reference, and mainly comprising the following steps of:
(i) setting the number of each individual function sub-network, the weight value range and the block size range of the core block mass in each function sub-network, searching and determining the position and the number of the core block mass in each function sub-network according to the set weight value and the block size based on the spatial distribution weight of each individual function sub-network, and calculating the distance from other voxels of the whole brain to the nearest voxel of the core block mass by taking the core block mass as the center;
the degree of adjacency of a voxel to the functional sub-network is measured by an inverse distance weighting factor, i.e. a neighborhood factor NF, whose formula is given by equation (1):
wherein D is the Euclidean distance from a certain voxel of the whole brain to the core cluster of the functional subnetwork; vox is a certain voxel currently calculated; hub _ cluster is a core cluster;
if the number of the core clusters in the functional sub-network is multiple, calculating a current certain voxel and multiple neighbor factors of the functional sub-network according to a formula (1), and averaging all the neighbor factors to obtain a final neighbor factor;
(ii) the coefficient of variation at the voxel level of each functional subnetwork tested was calculated, and the expression of coefficient of variation CV is formula (2):
wherein, σ is the standard deviation of the spatial distribution weight of a certain voxel in a certain functional subnetwork among the individual testees, and μ is the average value of the spatial distribution weight of a certain voxel in a certain functional subnetwork among the individual testees; obtaining an anti-noise factor ANF according to a formula (3) by using the coefficient of variation:
ANF=e-cv (3)
(iii) taking the characteristic time sequence of the functional sub-networks obtained in the step (2) as a reference signal, performing Pearson correlation calculation on the characteristic time sequence of each functional sub-network obtained in the step (2) and the voxel-by-voxel time sequence, and dividing a correlation coefficient serving as a correlation factor CF for dividing the voxel into the functional sub-networks;
(iv) multiplying the functional sub-network spatial component weight, the final neighbor factor, the anti-noise factor and the correlation factor; taking the product as the weight of the voxel to be divided into all functional sub-networks, sorting the weights of the voxel to be divided into all functional sub-networks, and assigning the voxel to the functional sub-network with the highest weight value;
(v) repeating the step (iv) until all voxels are assigned, obtaining a roughly divided whole brain function map, presenting a plurality of brain areas, and recording the average value of all voxel time sequences in each brain area as the average time sequence of the brain area;
(vi) extracting an average time sequence of the brain region obtained by the previous segmentation, averaging the average time sequence with a previous reference signal to obtain a new reference signal, and performing Pearson correlation calculation on each voxel and the new reference signal to obtain a new correlation factor;
(vii) repeating the steps (iv) - (v) to obtain a new generation of whole brain function map;
(viii) and (5) calculating the overlapping rate of the current brain atlas and the previous brain atlas, if the overlapping rate is lower than a set value, repeating the steps (vi) - (vii), and otherwise, terminating the iteration to obtain the final whole brain individualized brain function atlas.
The invention has the following beneficial effects:
1. brain area subdivision is more flexible: the number of functional sub-networks (namely independent components) can be artificially defined, and further individualized brain function maps with different fineness degrees can be constructed to meet different requirements.
2. The brain regions correspond completely: each subregion between individuals has a definite corresponding relation, so that a single subregion (namely, a brain map is obtained through an independent component network) can be obtained through calculation according to the invention, and the statistical analysis of the single subregion across the testees is carried out.
3. Pure data driving: the brain network constructed by the blind source analysis method ICA is used as reference, the existing prior template is not relied on, the data of a specific population is selected when the data set is obtained, the constructed individualized brain atlas is more suitable for the population corresponding to the data set, and a special atlas suitable for the population can be constructed for different populations, such as the brain atlas specially aiming at infants and old people.
4. Whole brain coverage: the individual functional brain atlas not only comprises cortex subregions, but also comprises a lower gray matter nucleus and cerebellum subregions, and can more comprehensively depict the functional boundary of the human gray matter subregions.
The core innovation points of the invention comprise:
(1) the invention firstly utilizes the independent component network as a reference to carry out individualized functional subregion segmentation on the whole brain. Independent Component Analysis (ICA) is one of the blind source signal separation algorithms that are widely used. The invention firstly separates spatially relatively independent group level brain functional sub-networks by utilizing an ICA (independent component analysis) technology based on a large sample fMRI (magnetic resonance imaging) data set, and the functional connection mode in each sub-network is similar and is different from the connection modes of other functional sub-networks. Because the ICA method is a blind source analysis method driven by pure data, users can set different numbers of functional sub-networks according to the requirements of the users, the flexibility of brain atlas subdivision is increased, and the method is more beneficial to and meets scenes with different purposes. Then, a spatial spatiotemporal regression technique is used to inversely reconstruct the functional sub-network of each individual. The space-time regression method can ensure that each functional sub-network at the group level completely corresponds to the individual level, the spatial distribution of the individual sub-networks is different from person to person, and the difference of the functional connection modes among individuals can be well described. The invention is a new idea of applying the method to the construction of the brain atlas.
(2) The individual functional subnetworks obtained by the ICA method described above are not absolutely independent, resulting in a situation where the boundaries of neighboring subnetworks may overlap and diverge. In order to solve the problem, the invention adopts the introduction of inverse distance weighting, sub-network variation weighting and iteration (including correlation factors and an iteration process) based on voxel function connection to determine stable and accurate boundaries among sub-networks and construct the whole brain individualized function brain atlas with different fineness degrees.
Description of the drawings:
FIG. 1: the invention discloses a flow schematic diagram of a whole brain individualized brain function map construction method taking an independent component network as a reference.
FIG. 2: ICA-based group level some independent component spatial distribution.
FIG. 3: each weight factor schematic diagram in the whole brain individualized function map construction method.
FIG. 4: based on the method disclosed by the invention, two tested whole brain individualized function maps are obtained.
The specific implementation mode is as follows:
the invention is further explained below with reference to examples and figures.
The invention designs a whole brain individualized function map construction method taking an independent component network as a reference. The flow of the construction method is shown in fig. 1, and the method mainly comprises data acquisition and brain fMRI pretreatment, spatial ICA (independent component analysis) based on group information, and individualized brain function map construction. Firstly, a three-dimensional T1 weighted sequence of a magnetic resonance device is utilized to obtain a brain structure image and a brain fMRI image with high spatial resolution; secondly, preprocessing collected fMRI data, and registering each brain function image to a standard space template by methods of time layer correction, motion correction and space registration, image segmentation, space standardization, space smoothing and the like so as to reduce the influence of individual anatomical position difference on the result; then, obtaining group horizontal space independent components by a group horizontal space ICA method, and then reversely reconstructing the space distribution (brain function sub-network) of each tested object and a characteristic time sequence corresponding to the components, namely reference signals, by utilizing space-time regression; finally, the individual brain function sub-network combines the inverse distance weighting, the sub-network variation weighting, the correlation factor and the iterative process to obtain an individual brain function map; wherein the correlation factors and the iterative process implement a voxel-based iteration of the functional connection. The iterative process means that the current correlation factor is continuously recalculated, and the repeated calculation determines that each voxel of the whole brain is assigned to each functional sub-network according to the four parameters of the functional sub-network spatial component weight, the final neighbor factor, the anti-noise factor and the correlation factor, so as to complete brain region segmentation until the overlapping rate of the brain atlas reaches a set value.
The specific process is as follows:
(1) data acquisition and preprocessing: the method comprises the steps of acquiring brain fMRI data in each tested resting state by using a nuclear magnetic resonance device, and then performing preprocessing such as time layer correction, motion correction, spatial registration, spatial smoothing and the like on the acquired brain fMRI data by using the conventional FSL (FMRIB Software library) open source Software, wherein the preprocessing ensures that different tested brain fMRI data are aligned to the same standard space, so as to solve the problems of brain morphology difference and inconsistent spatial positions during scanning between different tested subjects, ensure that the anatomical structures of all tested voxels are the same, and obtain the fMRI data of each tested subject and the voxel time sequence of the tested subject.
(2) Group level and individual independent component network reconstruction: after all the preprocessed fMRI data are connected in series according to the time direction, a group of horizontal space ICA analyses are carried out, which mainly comprises the following steps: (i) firstly, performing Principal Component Analysis (PCA) on preprocessed fMRI data of each tested object in a time point direction to perform dimensionality reduction (individual PCA); (ii) connecting a plurality of tested reduced-dimension functional data in series, and performing PCA (principal component analysis) dimension reduction on the connected data again (group PCA); (iii) performing spatial ICA analysis on the reduced PCA data set by using independent component analysis algorithms such as Informatx or FastICA and the like to obtain horizontal ICA spatial components of the set; (iv) and performing space-time regression on the individual fMRI data processed by the independent component analysis algorithm based on the group horizontal ICA space components to obtain the space component weight and the whole brain voxel value of each individual functional subnetwork and the characteristic time sequence of each individual functional subnetwork.
(3) Constructing an individual functional brain map: in order to ensure that each individual functional sub-network is absolutely independent (no overlap) in spatial distribution and has whole brain coverage (no interval), the invention introduces inverse distance weighting coefficients, sub-network inverse dissimilarity coefficient weighting, correlation factors and an iterative process, thereby obtaining a whole brain individualized functional map with an independent component network as a reference, and the main steps comprise:
(i) setting the number of each individual functional sub-network (the number of different tested functional sub-networks is the same), the weight value range and the block size range of the core block in each functional sub-network, searching and determining the position and the number of the core block in each functional sub-network according to the set weight value, the block size and other factors based on the spatial distribution weight of each individual functional sub-network, and calculating the distance from other voxels (all voxels except the voxel corresponding to the core block) of the whole brain to the nearest voxel of the core block by taking the core block as the center (the distance is marked as 0 if the voxels fall into the core block). The closer to the core blob, the greater the likelihood that the voxel is classified into the functional subnetwork. The present invention measures the proximity of a voxel to the functional subnetwork, i.e. the Neighbor Factor (NF), with an inverse distance weighting factor.
The formula expression of the neighbor factor, namely the inverse distance weighting coefficient NF, is formula (1):
wherein D is the Euclidean distance from a certain voxel of the whole brain to the core cluster of the functional subnetwork; vox is a certain voxel currently calculated; hub _ cluster is a core cluster;
if the number of the core clusters in the functional sub-network is multiple, calculating a current certain voxel and multiple neighbor factors of the functional sub-network according to a formula (1), and averaging all the neighbor factors to obtain a final neighbor factor;
(ii) in order to reduce the influence of noise on the functional sub-region boundary division, the invention calculates the variation coefficient of each tested functional sub-network at the voxel level. For a certain functional sub-network, the more varied voxels (generally at the boundary of the functional sub-network), the more affected the noise, the more likely the voxels are wrongly classified as the functional sub-network; conversely, the smaller the coefficient of variation, the more likely the sub-network is to be divided. Therefore, sub-network inverse difference coefficients are used to measure the noise immunity level, i.e. noise immunity factor (ANF), of the corresponding voxels of each sub-network.
Coefficient of Variation (CV) is a concept in probability theory and statistics, and is mainly used to measure the magnitude of the dispersion degree of a certain distribution, and its value is the ratio of the standard deviation of the distribution to the average value of the distribution, and its expression is as follows:
wherein, σ is the standard deviation of the distribution of the spatial distribution weight of a certain voxel in a certain functional subnetwork among the individual testees, and μ is the average value of the distribution of the spatial distribution weight of a certain voxel in a certain functional subnetwork among the individual testees. The variation coefficient can be used for carrying out quantitative analysis on the distribution variation of each voxel in an independent component among individual testees, and the larger the value of the variation coefficient is, the more remarkable the variation of the voxel among the testees is, so that the current voxel is closer to noise.
Thus, the anti-noise factor (ANF) is expressed by the following equation:
ANF=e-cv (3)
(iii) and (3) taking the characteristic time sequence of the functional sub-networks obtained in the step (2) as a reference signal, performing Pearson correlation calculation on the voxel-by-voxel time sequence (each voxel time sequence can be obtained after the acquisition of the fMRI data in the resting state) and the characteristic time sequence of each functional sub-network obtained in the step (2), and dividing a correlation coefficient into Correlation Factors (CF) of each functional sub-network as the voxel.
The Correlation Factor (CF) characterizes the synchronicity of a certain voxel time series with a functional subnetwork reference signal, measured by the pearson correlation coefficient. If we assume that the rs-fMRI data time series of all voxels of the whole brain can be represented by TC ═ TC1,tc2,…,tcv]And the functional subnetwork reference signal may be represented by TC _ ref ═ TC _ ref1,tc_ref2,…,tc_refj]Is represented by, wherein tcvIs the time series of the v-th voxel (v is 1, 2, …, Q is the number of voxels), and tc _ refjFor the jth functional subnetwork reference signal (J ═ 1, 2, …, J is the number of functional subnetworks), then the Correlation Factor (CF) of each voxel with the functional subnetwork reference signal is expressed by the following equation:
CF(v,j)=corr(tcv,tc_refj) (4)
wherein corr represents Pearson correlation analysis.
(iv) Multiplying the functional sub-network spatial component weight, the final neighbor factor, the anti-noise factor and the correlation factor; the product is used as the weight of dividing the voxel into functional sub-networks, the weights of different functional sub-networks are sorted, and the voxel is assigned to the functional sub-network with the highest weight value.
(v) Repeating the step (iv) until all voxels are assigned, obtaining a roughly divided whole brain function map, presenting a plurality of brain areas, and recording the average value of all voxel time sequences in each brain area as the average time sequence of the brain area;
(vi) extracting an average time sequence of the brain region obtained by the previous segmentation, averaging the average time sequence with a previous reference signal to obtain a new reference signal, and performing Pearson correlation calculation on each voxel and the new reference signal to obtain a new correlation factor;
(vii) repeating the steps (iv) - (v) to obtain a new generation of whole brain function map;
(viii) and (3) calculating the overlapping rate of the current brain atlas and the previous brain atlas, if the overlapping rate is lower than a set value (such as 99%), repeating the steps (vi) - (vii), and otherwise, terminating the iteration to obtain the final whole brain individualized brain function atlas.
The specific calculation process of the subnetwork spatial component weight of the functional subnetwork is as follows:
ICA is a blind source signal separation algorithm that estimates the original signal by using the observed signal and through the unmixing matrix under the condition that the source signals are assumed to be independent from each other. In ICA analysis, S ═ is assumed (S)1,S2,…,SN)TRepresenting N independent source signals, the source signals being passed through an unknown mixing matrix AM×NLinear mixing to generate observation signal X ═ X1,X2,…,Xt,…,XM)T. Wherein, XtRepresents a random variable in the observed signal, and M represents the number of observed signals. Therefore, under the condition of the noise-free ICA model, the formula expression is as follows:
X=AS (5)
as known from the above formula, each independent source signal and mixing matrix in the ICA method are unknown, so the uncertainty of the sequence of ICA output components and the unpredictability of the number of components make it a great challenge for the multi-tested rs-fMRI analysis. Therefore, ICA analysis between multiple subjects requires establishing correspondence between the functional sub-networks of multiple subjects to more effectively obtain the specificity of individual functional networks.
Considering the defects of the traditional ICA analysis in the aspects of uncertainty of output component sequence and unpredictability of component number, the invention utilizes a spatial ICA method based on group information to obtain the brain function network specificity and correspondence of an individual to be tested. The process of the spatial ICA method based on the group information is as follows: ICA analysis is firstly carried out on all the groups of data to obtain independent components on the group level, and then each tested component is obtained through reconstruction based on the obtained independent components on the group level. The method has the advantages that the method can establish the correspondence between different tested components, and is favorable for directly carrying out statistical analysis based on a single brain subregion.
All brain rs-fMRI data tested can be expressed as X ═ X1,x2,…,xi]Is represented by the formula (I) in which xiThe data is the preprocessed whole brain rs-fMRI data of the ith test (i is 1, 2, …, n, n is the number of the test); second, individual PCA: using principal component analysis method to xiPerform data dimension reduction (i.e. reduce the dimension of the dataBroadly refers to a calculation process for performing principal component analysis), as shown in equation 5, X is reduced to KiA matrix of xl (K is the number of principal components, L is the number of voxels); then, the matrixes after different tested dimensionalities are connected in series into one matrix according to the time point directionA matrix of (a); and then PCA: using principal component analysis method pairIs reduced in dimension (i.e., G)-1The calculation process of principal component analysis is performed again) to obtain an N × L matrix (N represents the number of independent source signals (i.e., the number of set functional subnets); finally, using formula 5, the obtained N × L matrix is subjected to data analysis by the ICA method, and a series of spatial distributions S of brain functional sub-networks based on the group level are obtained.
From the above equation, S is the spatial distribution of the group horizontal brain function sub-network obtained by the ICA method, i.e., the independent signal source in the above equation (5).
Subsequently, as shown in formula (6), time regression is performed on the rs-fMRI data of the whole brain of the individual based on the obtained independent components at the group level, and the unmixing matrix a of the individual to be tested is obtainedi(characteristic time series of individual functional subnetworks). Then, based on the individual unmixing matrix, the individual fMRI data is subjected to spatial regression to obtain the spatial distribution characteristics s of the individual functional sub-networksiAs shown in equations 7 and 8.
The ith tested time regression expression is:
Ai=xiS-1 (7)
the spatial regression expression for the ith test was:
remember of siIs the spatial distribution of individual horizontal brain functional subnetworks, i.e. the spatial component weights of the functional subnetworks, obtained by a two-step regression method.
Example 1
Magnetic resonance data acquisition: the present invention currently uses the Human brain connection Project (HCP) common database to derive the brain image data. The database collected MRI data using a custom version of the HCP scanner with a magnetic field strength of 3.0Tesla WU-Minn-Ox. Among them, the three-dimensional high-resolution T1 weighted structure: preparing a rapid acquisition gradient echo (MPRAGE) sequence using magnetization, with a repetition Time (TR)/echo Time (TE)/inversion Time (TI) of 2400/2.14/1000ms, a flip angle of 8 °, a field of view (FOV) of 224mm × 224mm, an imaging matrix of 320 × 320, a layer thickness of 0.7mm, 0.7-mm3The isotropic voxels of (a); resting state fMRI: using a multiband single gradient echo EPI (MB-SS-GRE-EPI) sequence with TR/TE 720/33.1ms, flip angle 52 °, slice acceleration factor 8, FOV 208mm × 180mm, imaging matrix 104 × 90, layer thickness2mm, 72 layers, 2-mm3Isotropic voxels, the resting-state scanning process adds an exchange phase encoding direction scan (left-fold) in order to reduce the deformation due to susceptibility artifacts>Right/right->Left phase encoding) to obtain 4 sets of resting state fMRI data, each set of fMRI data including 1200 time points. The remaining brain image MRI scan parameters can be consulted for the HCP common database official network (http:// protocols. humanconnected. org/HCP/3T/imaging-protocols. html). In addition, the invention is also applicable to other public databases or collected resting state fMRI data, and collected data parameters can be referred to a HCP public database official network.
(1) Preprocessing of the fMRI data in the resting state: the invention adopts a HCP common database standard preprocessing process which mainly comprises gradient distortion correction, head movement correction, space standardization and high-pass filtering[1,2]And denoising based on independent component analysis (ICA-FIX)[3]And the like. Therefore, the resting state fMRI data used in the present invention is denoised fMRI data of "ICA-FIX" disclosed by the HCP. The data preprocessing process is integrated into the FSL (FMRIB Software library) open source toolkit, and the detailed processes of the rest of data preprocessing can be referred to (https:// www.humanconnectome.org/storage/app/media/documentation/S1200/HCP _ S1200_ Relea _ Reference _ Manual. In addition, the present invention performs spatial smoothing on the data, and the process takes into account the influence of a certain voxel gray level in the space on the gray levels of other surrounding voxels, and diffuses to other surrounding voxels through a Gaussian kernel function (Gaussian kernel function). An indicator characterizing this spatially smooth diffusion range is full width at half maximum (FWHM). Smoothing can reduce registration errors and enhance the signal-to-noise ratio of the image; the normality of the data is increased in order to statistically improve the statistical power, and the FWHM is usually set to 2-3 times the voxel size. The FWHM is 6mm3The process may also be implemented in the FSL (FMRIB Software library) open source toolkit. In addition, the invention can also process the resting state fMRI data by using the standard preprocessing method (time layer correction, motion correction, spatial registration, spatial smoothing and the like) provided in the FSL open source toolkit, and the detailed method is shown in FSL (FMRIB Software library)Source kit instruction manual.
(2) Functional subnetwork spatial component weights: the present invention utilizes ICA-based group-level independent components (http:// www.humanconnectome.org/study/HCP-yong-adult) provided by the public database of healthy adults with HCPs. As shown in fig. 2 (L is the left side, Z is the weight value) (only the sub-network spatial component weight of one component is shown in fig. 2), the present embodiment selects 100 group level independent component results of the whole brain partition provided by the public database as reference. Then, the tested independent components, namely the spatial component weights of the functional sub-networks are obtained based on the group of horizontal independent component reconstruction. Also, the characteristic time series corresponding to the component (i.e. the characteristic time series of the functional sub-network, denoted as reference signal, as the start signal of the first iteration). In addition, individual components and corresponding time series of each subject of the present invention can also be obtained by providing group-level-based ICA analysis methods in the resting state fMRI data and the FSL (FMRIB Software library) open source kit.
(3) A neighbor factor: a core blob (in this embodiment, the core blob is set to have a default weight value >2.2 and the blob size >70voxels) is searched in each independent component spatial range (functional subnetwork), the core blob is the spatial position of the maximum weight value in the functional subnetwork, the distance from other voxels in the whole brain to the voxel point is calculated by taking the spatial position as the center, and a neighbor factor, i.e., a neighbor coefficient, is obtained according to the distance, and the obtained result is shown in fig. 3.
(4) Coefficient of variation and noise immunity factor: considering the influence of noise on resting state functional data, the distribution change of each voxel in each independent component among individual subjects, namely the coefficient of variation, is calculated respectively, and an anti-noise factor, namely an anti-noise coefficient, is obtained according to the formula (3) according to the coefficient of variation, and the obtained result is shown in fig. 3.
(5) Correlation factor: and (3) taking the characteristic time sequence of the functional sub-network obtained in the step (2) as a reference signal, measuring the synchronism of a certain voxel time sequence and the reference signal of the functional sub-network through a Pearson correlation coefficient, and obtaining a correlation factor through correlation calculation of the voxel time sequence and the reference signal of the functional sub-network.
(6) The product of the spatial component weight, the neighborhood factor, the anti-noise factor and the correlation factor of the functional sub-network is used as the weight of dividing the voxel into each component (functional sub-network), the weights of dividing the voxel into different components are ranked, and the voxel is assigned to the functional sub-network with the highest weight. Assigning values of all voxels according to the standard to obtain a roughly divided whole brain function map, presenting a plurality of brain regions, and recording the average value of all voxel time sequences in each brain region as the average time sequence of the brain region
(7) Then, continuously optimizing a whole brain function map, extracting an average time sequence of a brain region obtained by previous segmentation, averaging the average time sequence with a previous reference signal to obtain a new reference signal, and performing Pearson correlation calculation on each voxel and the new reference signal to obtain a new correlation factor; repeating the step (6) to obtain a new generation of whole brain function map;
until the overlapping rate of the current brain atlas and the previous brain atlas reaches a set value, if the overlapping rate is not lower than the set value (for example, 99%), the iteration is terminated to obtain a final individual brain function atlas of the whole brain, and the final segmentation result is shown in fig. 4, it can be seen from fig. 4 that two tested subjects can construct respective whole brain function atlases through the method of the present application, and the atlases, brains and cerebellar subregions can be distinguished obviously in the atlas.
The method of the invention has the following characteristics: 1) functional brain area subdivision is more flexible: the number of independent components, i.e. functional sub-networks, can be defined according to the needs of the user; 2) pure data driving: does not rely on existing prior templates; 3) the brain regions correspond completely: each subregion between individuals has a definite corresponding relation, and the brain subregions obtained by the method are beneficial to subsequent horizontal analysis of the region of interest of the single subregion; 4) and (3) full coverage: the functional brain atlas includes cortical subregions, as well as subclinical nuclei and cerebellar subregions.
Nothing in this specification is said to apply to the prior art.
Reference documents:
[1]Glasser M F,Sotiropoulos S N,Wilson J A,et al.The minimal preprocessing pipelines for the Human Connectome Project[J].Neuroimage,2013,80:105-124.
[2]Van Essen D C,Smith S M,Barch D M,et al.The WU-Minn Human Connectome Project:an overview[J].Neuroimage,2013,80:62-79.
[3]Griffanti L,Salimi-Khorshidi G,Beckmann C F,et al.ICA-based artefact removal and accelerated fMRI acquisition for improved resting state network imaging[J].Neuroimage,2014,95:232-247.
Claims (7)
1. a whole brain individualized brain function map construction method taking an independent component network as a reference is characterized in that the method utilizes individual brain resting state fMRI data to be tested, firstly introduces an independent component analysis method to construct a group level brain function sub-network, then reversely reconstructs each tested brain function sub-network and a characteristic time sequence corresponding to the function sub-network by utilizing space-time regression, and takes the characteristic time sequence corresponding to the function sub-network as a reference signal;
and introducing an inverse distance weighting coefficient, sub-network inverse variation coefficient weighting, a correlation factor and an iterative process, and further obtaining the whole brain individualized function map with the independent component network as a reference.
2. The construction method according to claim 1, characterized in that the specific process of the method is as follows:
(1) data acquisition and preprocessing: acquiring brain fMRI data of each tested resting state, preprocessing the acquired brain fMRI data, ensuring that different tested brain fMRI data are aligned to the same standard space, ensuring that the anatomical structures of all tested voxels are the same, and acquiring the fMRI data of each tested and the time sequence of the tested voxels;
(2) group level and individual independent component network reconstruction: after all the preprocessed fMRI data are connected in series according to the time direction, a group of horizontal space ICA analyses are carried out, which mainly comprises the following steps: (i) carrying out dimensionality reduction on each preprocessed fMRI data to be tested in a time point direction by adopting principal component analysis; (ii) connecting a plurality of tested dimension-reduced functional data in series, and performing PCA dimension reduction on the connected data again; (iii) carrying out spatial ICA analysis on the reduced PCA data to obtain horizontal ICA spatial components; (iv) performing space-time regression on the individual fMRI data processed by the independent component analysis algorithm based on the group horizontal ICA space components to obtain the space component weight of each individual functional subnetwork, the whole brain voxel value and the characteristic time sequence of each individual functional subnetwork;
(3) constructing an individual functional brain map: introducing inverse distance weighting coefficients, sub-network inverse variation coefficient weighting, correlation factors and an iterative process to further obtain a whole brain individualized function map with an independent component network as a reference, and mainly comprising the following steps of:
(i) setting the number of each individual function sub-network, the weight value range and the block size range of the core block mass in each function sub-network, searching and determining the position and the number of the core block mass in each function sub-network according to the set weight value and the block size based on the spatial distribution weight of each individual function sub-network, and calculating the distance from other voxels of the whole brain to the nearest voxel of the core block mass by taking the core block mass as the center;
the degree of adjacency of a voxel to the functional sub-network is measured by an inverse distance weighting factor, i.e. a neighborhood factor NF, whose formula is given by equation (1):
wherein D is the Euclidean distance from a certain voxel of the whole brain to the core cluster of the functional subnetwork; vox is a certain voxel currently calculated; hub _ cluster is a core cluster;
if the number of the core clusters in the functional sub-network is multiple, calculating a current certain voxel and multiple neighbor factors of the functional sub-network according to a formula (1), and averaging all the neighbor factors to obtain a final neighbor factor;
(ii) the coefficient of variation at the voxel level of each functional subnetwork tested was calculated, and the expression of coefficient of variation CV is formula (2):
wherein, σ is the standard deviation of the spatial distribution weight of a certain voxel in a certain functional subnetwork among the individual testees, and μ is the average value of the spatial distribution weight of a certain voxel in a certain functional subnetwork among the individual testees; obtaining an anti-noise factor ANF according to a formula (3) by using the coefficient of variation:
ANF=e-cv (3)
(iii) taking the characteristic time sequence of the functional sub-networks obtained in the step (2) as a reference signal, performing Pearson correlation calculation on the characteristic time sequence of each functional sub-network obtained in the step (2) and the voxel-by-voxel time sequence, and dividing a correlation coefficient serving as a correlation factor CF for dividing the voxel into the functional sub-networks;
(iv) multiplying the functional sub-network spatial component weight, the final neighbor factor, the anti-noise factor and the correlation factor; taking the product as the weight of the voxel to be divided into all functional sub-networks, sorting the weights of the voxel to be divided into all functional sub-networks, and assigning the voxel to the functional sub-network with the highest weight value;
(v) repeating the step (iv) until all voxels are assigned, obtaining a roughly divided whole brain function map, presenting a plurality of brain areas, and recording the average value of all voxel time sequences in each brain area as the average time sequence of the brain area;
(vi) extracting an average time sequence of the brain region obtained by the previous segmentation, averaging the average time sequence with a previous reference signal to obtain a new reference signal, and performing Pearson correlation calculation on each voxel and the new reference signal to obtain a new correlation factor;
(vii) repeating the steps (iv) - (v) to obtain a new generation of whole brain function map;
(viii) and (5) calculating the overlapping rate of the current brain atlas and the previous brain atlas, if the overlapping rate is lower than a set value, repeating the steps (vi) - (vii), and otherwise, terminating the iteration to obtain the final whole brain individualized brain function atlas.
3. The construction method according to claim 2, wherein the specific calculation process of the functional sub-network spatial component weight is:
considering the defects of the traditional ICA analysis in the aspects of uncertainty of output component sequence and unpredictability of component number, utilizing a spatial ICA method based on group information to obtain the brain function network specificity and correspondence of an individual to be tested; the process of the spatial ICA method based on the group information is as follows: firstly, ICA analysis is carried out on all group data once to obtain independent components on a group level, then each tested component is obtained based on the obtained independent components on the group level, and the correspondence among different tested components can be established;
all brain rs-fMRI data tested can be expressed as X ═ X1,x2,…,xi]Is represented by the formula (I) in which xiThe data is the preprocessed whole brain rs-fMRI data of the ith test (i is 1, 2, …, n, n is the number of the test); second, individual PCA: using principal component analysis method to xiReducing dimension of data to KiA matrix of xl (K is the number of principal components, L is the number of voxels); then, the matrixes after different tested dimensionalities are connected in series into one matrix according to the time point directionA matrix of (a); and then PCA: using principal component analysis method pairPerforming dimensionality reduction on the matrix to obtain an NxL matrix, wherein N represents the number of independent source signals; finally, carrying out data analysis on the obtained NxL matrix by an ICA method so as to obtain a series of brain function sub-network spatial distribution S based on group level;
then, based on the obtained independent components on the group level, time regression is carried out on the rs-fMRI data of the whole brain of the individual, and the unmixing matrix A of the individual to be tested is obtainedi(ii) a Then, based on the individual unmixing matrix, the individual fMRI data is subjected to spatial regression to obtain the spatial distribution characteristics s of the individual functional sub-networksiSuch as formula7 and equation 8;
the ith tested time regression expression is:
Ai=xiS-1 (7)
the spatial regression expression for the ith test was:
remember of siIs the spatial distribution of individual horizontal brain functional subnetworks, i.e. the spatial component weights of the functional subnetworks, obtained by a two-step regression method.
4. Construction method according to claim 2, characterized in that the data preprocessing comprises temporal layer correction, motion correction and spatial registration, image segmentation, spatial normalization, spatial smoothing, the size of the full width at half maximum FWHM of the indicator of the spatial smooth diffusion in the spatial smoothing is set to 2-3 times the voxel size, preferably 6mm FWHM3。
5. The method of construction according to claim 2, wherein the weight value of the core blob is >2.2 and the blob size is >70 voxels.
6. The method according to claim 2, wherein the core cluster is a spatial position of a maximum weight value in the functional subnetwork, and the distances from other voxels of the whole brain to the voxel point are calculated with the position as a center, and the neighborhood factor is obtained from the distances.
7. The method of any one of claims 1-6, wherein the map is capable of distinguishing between sub-regions of polio nuclei, brain and cerebellum.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010856401.2A CN112002428B (en) | 2020-08-24 | 2020-08-24 | Whole brain individualized brain function map construction method taking independent component network as reference |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010856401.2A CN112002428B (en) | 2020-08-24 | 2020-08-24 | Whole brain individualized brain function map construction method taking independent component network as reference |
Publications (2)
Publication Number | Publication Date |
---|---|
CN112002428A true CN112002428A (en) | 2020-11-27 |
CN112002428B CN112002428B (en) | 2022-03-08 |
Family
ID=73471716
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010856401.2A Active CN112002428B (en) | 2020-08-24 | 2020-08-24 | Whole brain individualized brain function map construction method taking independent component network as reference |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112002428B (en) |
Cited By (6)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113066113A (en) * | 2021-04-27 | 2021-07-02 | 河北医科大学第二医院 | Method and equipment for constructing spontaneous hypertensive rat brain template and map set |
CN113345572A (en) * | 2021-05-13 | 2021-09-03 | 东南大学 | Brain disease judging system and judging device based on meta-analysis |
CN114376558A (en) * | 2022-03-24 | 2022-04-22 | 之江实验室 | Brain atlas individuation method and system based on magnetic resonance and twin map neural network |
CN115546124A (en) * | 2022-09-21 | 2022-12-30 | 电子科技大学 | Integration method of cerebellum-brain motor function combining ICA and functional gradient |
EP4156200A1 (en) * | 2021-09-27 | 2023-03-29 | Universitätsmedizin der Johannes Gutenberg-Universität Mainz | Method for reconstructing a brain network |
CN116597994A (en) * | 2023-05-16 | 2023-08-15 | 天津大学 | Mental disease brain function activity assessment device based on brain activation clustering algorithm |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
US20150272468A1 (en) * | 2014-03-31 | 2015-10-01 | Hesheng Liu | System And Method For Functional Brain Organization Mapping |
CN105955929A (en) * | 2016-04-26 | 2016-09-21 | 上海大学 | Data scientific visualization-oriented inverse distance weighting mixed interpolation method |
CN106709244A (en) * | 2016-12-12 | 2017-05-24 | 西北工业大学 | Brain function network modeling method for resting state synchronization EEG-fMRI |
CN109496338A (en) * | 2017-12-05 | 2019-03-19 | 北京师范大学 | Based on personal feature through cranium brain map generation method, air navigation aid and its system |
CN110598793A (en) * | 2019-09-12 | 2019-12-20 | 常州大学 | Brain function network feature classification method |
KR20200005303A (en) * | 2018-07-06 | 2020-01-15 | 고려대학교 산학협력단 | Device and method for estimating task-related spatial patterns using task-negative default-mode networks and recording medium for performing the method |
CN110838173A (en) * | 2019-11-15 | 2020-02-25 | 天津医科大学 | Three-dimensional texture feature-based individual brain covariant network construction method |
CN111081351A (en) * | 2019-12-02 | 2020-04-28 | 北京优脑银河科技有限公司 | Method and system for drawing brain function map |
-
2020
- 2020-08-24 CN CN202010856401.2A patent/CN112002428B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
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 |
US20150272468A1 (en) * | 2014-03-31 | 2015-10-01 | Hesheng Liu | System And Method For Functional Brain Organization Mapping |
CN105955929A (en) * | 2016-04-26 | 2016-09-21 | 上海大学 | Data scientific visualization-oriented inverse distance weighting mixed interpolation method |
CN106709244A (en) * | 2016-12-12 | 2017-05-24 | 西北工业大学 | Brain function network modeling method for resting state synchronization EEG-fMRI |
CN109496338A (en) * | 2017-12-05 | 2019-03-19 | 北京师范大学 | Based on personal feature through cranium brain map generation method, air navigation aid and its system |
KR20200005303A (en) * | 2018-07-06 | 2020-01-15 | 고려대학교 산학협력단 | Device and method for estimating task-related spatial patterns using task-negative default-mode networks and recording medium for performing the method |
CN110598793A (en) * | 2019-09-12 | 2019-12-20 | 常州大学 | Brain function network feature classification method |
CN110838173A (en) * | 2019-11-15 | 2020-02-25 | 天津医科大学 | Three-dimensional texture feature-based individual brain covariant network construction method |
CN111081351A (en) * | 2019-12-02 | 2020-04-28 | 北京优脑银河科技有限公司 | Method and system for drawing brain function map |
Non-Patent Citations (2)
Title |
---|
ZHANGYANG ETC.: ""Altered brain activation and functional connectivity in working memory related networks in patients with type 2 diabetes: An ICA-based analysis"", 《SCIENTIFIC REPORTS》 * |
王振松: ""CT图像头颈部组织器官分割方法研究"", 《CNKI学位论文》 * |
Cited By (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN113066113A (en) * | 2021-04-27 | 2021-07-02 | 河北医科大学第二医院 | Method and equipment for constructing spontaneous hypertensive rat brain template and map set |
CN113066113B (en) * | 2021-04-27 | 2022-01-11 | 河北医科大学第二医院 | Method and equipment for constructing spontaneous hypertensive rat brain template and map set |
CN113345572A (en) * | 2021-05-13 | 2021-09-03 | 东南大学 | Brain disease judging system and judging device based on meta-analysis |
CN113345572B (en) * | 2021-05-13 | 2024-03-15 | 东南大学 | Brain disease judging system and judging device based on meta-analysis |
EP4156200A1 (en) * | 2021-09-27 | 2023-03-29 | Universitätsmedizin der Johannes Gutenberg-Universität Mainz | Method for reconstructing a brain network |
WO2023046960A1 (en) * | 2021-09-27 | 2023-03-30 | Universitätsmedizin Der Johannes Gutenberg-Universität Mainz | Method for reconstructing a brain network |
CN114376558A (en) * | 2022-03-24 | 2022-04-22 | 之江实验室 | Brain atlas individuation method and system based on magnetic resonance and twin map neural network |
WO2023178916A1 (en) * | 2022-03-24 | 2023-09-28 | 之江实验室 | Brain atlas individualized method and system based on magnetic resonance and twin graph neural network |
CN115546124A (en) * | 2022-09-21 | 2022-12-30 | 电子科技大学 | Integration method of cerebellum-brain motor function combining ICA and functional gradient |
CN116597994A (en) * | 2023-05-16 | 2023-08-15 | 天津大学 | Mental disease brain function activity assessment device based on brain activation clustering algorithm |
CN116597994B (en) * | 2023-05-16 | 2024-05-14 | 天津大学 | Mental disease brain function activity assessment device based on brain activation clustering algorithm |
Also Published As
Publication number | Publication date |
---|---|
CN112002428B (en) | 2022-03-08 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN112002428B (en) | Whole brain individualized brain function map construction method taking independent component network as reference | |
CN109376751B (en) | Human brain function network classification method based on convolutional neural network | |
CN109522894B (en) | Method for detecting dynamic covariation of fMRI brain network | |
Thirion et al. | Detection of signal synchronizations in resting-state fMRI datasets | |
WO2020252805A1 (en) | Dynamic function mode learning method inspired by fmri brain network mechanism | |
Zhang et al. | Deep learning models unveiled functional difference between cortical gyri and sulci | |
CN117172294B (en) | Method, system, equipment and storage medium for constructing sparse brain network | |
CN113947157B (en) | Dynamic brain effect connection network generation method based on hierarchical clustering and structural equation model | |
CN112233086B (en) | fMRI data classification and identification method and device based on brain region functional connection | |
Byeon et al. | Artificial neural network inspired by neuroimaging connectivity: application in autism spectrum disorder | |
CN115005798A (en) | Brain image feature extraction method based on edge connection function connection | |
Rathi et al. | Detection and characterization of brain tumor using segmentation based on HSOM, wavelet packet feature spaces and ANN | |
CN115272295A (en) | Dynamic brain function network analysis method and system based on time domain-space domain combined state | |
CN111931578A (en) | Electroencephalogram identification method based on minimum spanning tree and regional double-layer network | |
CN109994204A (en) | A kind of cerebral function network key node searching method | |
Sen | Generalized Prediction Model for Detection of Psychiatric Disorders | |
Donnelly-Kehoe et al. | Consistent local dynamics in the brain across sessions are revealed by whole brain modeling of resting state activity | |
CN114176518B (en) | Complex fMRI data space component phase anti-correction method for improving CNN classification performance | |
Shankaran | Multi-site Diagnostic Classification Of Schizophrenia Using 3D CNN On Aggregated Task-based fMRI Data | |
Li et al. | Functional Brain Network Estimation with Human-Guided Modularity Representation | |
Marui et al. | MD-LMS algorithm based brain functional connectivity analysis in resting state fMRI | |
Komorowski et al. | MULTI-LEVEL EXPLANATIONS IN NEUROSCIENCE II: EEG SPECTRAL FINGERPRINTS AND TENSOR DECOMPOSITIONS FOR UNDERSTANDING BRAIN ACTIVITY--INITIAL RESULTS. | |
CN116090225A (en) | Brain connection network construction method based on homogeneity constraint multi-set typical correlation analysis | |
Gomez-Laberge et al. | Selection criteria for the analysis of data-driven clusters in cerebral fMRI | |
Zhao et al. | Exploring brain structure-function relationship using constant-block PLSC via MRI data |
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 | ||
CP02 | Change in the address of a patent holder | ||
CP02 | Change in the address of a patent holder |
Address after: No. 1, Rongyuan Road, Huayuan Industrial Zone, High-tech Industrial Development Zone, Binhai New Area, Tianjin 300384 Patentee after: Tianjin Medical University Address before: 300070 No. 22 Meteorological Observatory Road, Heping District, Tianjin Patentee before: Tianjin Medical University |
|
OL01 | Intention to license declared | ||
OL01 | Intention to license declared |