CN112233086B - fMRI data classification and identification method and device based on brain region functional connection - Google Patents
fMRI data classification and identification method and device based on brain region functional connection Download PDFInfo
- Publication number
- CN112233086B CN112233086B CN202011095774.9A CN202011095774A CN112233086B CN 112233086 B CN112233086 B CN 112233086B CN 202011095774 A CN202011095774 A CN 202011095774A CN 112233086 B CN112233086 B CN 112233086B
- Authority
- CN
- China
- Prior art keywords
- brain
- brain region
- fmri data
- partial
- matrix
- 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.)
- Active
Links
- 210000004556 brain Anatomy 0.000 title claims abstract description 231
- 238000000034 method Methods 0.000 title claims abstract description 68
- 238000002599 functional magnetic resonance imaging Methods 0.000 title claims abstract description 67
- 239000011159 matrix material Substances 0.000 claims abstract description 42
- 239000013598 vector Substances 0.000 claims abstract description 28
- 230000009466 transformation Effects 0.000 claims abstract description 15
- 230000006870 function Effects 0.000 claims abstract description 14
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000007781 pre-processing Methods 0.000 claims abstract description 13
- 230000000284 resting effect Effects 0.000 claims description 14
- 230000003925 brain function Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 230000009467 reduction Effects 0.000 claims description 7
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000012937 correction Methods 0.000 claims description 5
- 238000009499 grossing Methods 0.000 claims description 5
- 230000008030 elimination Effects 0.000 claims description 4
- 238000003379 elimination reaction Methods 0.000 claims description 4
- 238000001914 filtration Methods 0.000 claims description 4
- 210000004884 grey matter Anatomy 0.000 claims description 4
- 238000012417 linear regression Methods 0.000 claims description 4
- 210000004885 white matter Anatomy 0.000 claims description 4
- 210000005013 brain tissue Anatomy 0.000 claims description 3
- 238000006243 chemical reaction Methods 0.000 claims description 3
- 230000004886 head movement Effects 0.000 claims description 3
- 238000003384 imaging method Methods 0.000 claims description 3
- 230000001788 irregular Effects 0.000 claims description 3
- 238000005192 partition Methods 0.000 claims description 3
- 238000005516 engineering process Methods 0.000 description 4
- 238000003012 network analysis Methods 0.000 description 3
- 238000013421 nuclear magnetic resonance imaging Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 206010003805 Autism Diseases 0.000 description 2
- 208000020706 Autistic disease Diseases 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 2
- 230000007177 brain activity Effects 0.000 description 2
- 210000001638 cerebellum Anatomy 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 230000000926 neurological effect Effects 0.000 description 2
- 238000003325 tomography Methods 0.000 description 2
- 238000012549 training Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000012952 Resampling Methods 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000008280 blood Substances 0.000 description 1
- 210000004369 blood Anatomy 0.000 description 1
- 230000008344 brain blood flow Effects 0.000 description 1
- 210000001175 cerebrospinal fluid Anatomy 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013145 classification model Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000007405 data analysis Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000012880 independent component analysis Methods 0.000 description 1
- 238000002595 magnetic resonance imaging Methods 0.000 description 1
- 238000002582 magnetoencephalography Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000002503 metabolic effect Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000002610 neuroimaging Methods 0.000 description 1
- 229910052760 oxygen Inorganic materials 0.000 description 1
- 239000001301 oxygen Substances 0.000 description 1
- 230000002285 radioactive effect Effects 0.000 description 1
- 238000002603 single-photon emission computed tomography Methods 0.000 description 1
- 238000012706 support-vector machine Methods 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10072—Tomographic images
- G06T2207/10088—Magnetic resonance imaging [MRI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
- G06T2207/30016—Brain
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- Radiology & Medical Imaging (AREA)
- Quality & Reliability (AREA)
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The invention designs a fMRI data classification and identification method and device based on brain region functional connection, comprising the following steps: obtaining fMRI data of a subject; preprocessing the obtained fMRI data to obtain a grey brain matter image; dividing a grey brain image into a plurality of brain regions with different functions, and extracting an average voxel time sequence of each brain region; selecting a partial brain region with significant difference from a plurality of functional brain regions based on the fuzzy decision rough set; based on the selected partial brain regions, calculating the Pearson correlation coefficient between different brain regions, and carrying out nonlinear processing on the coefficient by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain regions; sparsifying the correlation coefficient values in the matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into one-dimensional feature vectors; and taking the obtained one-dimensional feature vector as input to a trained SVM recognition model, obtaining an output label of the tested person and judging the fMRI data type of the tested person.
Description
Technical Field
The invention belongs to the technical field of data classification and identification, and particularly relates to an fMRI data classification and identification method based on brain region functional connection.
Background
In recent years, the rapid development of medical images provides a very important clinical reference value for brain image data analysis and brain activity state observation, so that the research of human brains enters a new stage. The current approaches for brain activity research are mainly: functional magnetic resonance imaging (fMRI), electroencephalogram (EEG), magnetoencephalography (MEG), electron emission tomography (PET), single photon emission tomography (SPECT), and the like. The fMRI technology is a non-radioactive and noninvasive means for detecting brain function dynamic activity, has higher space-time resolution, and is the most commonly used and effective measuring means at present.
fMRI studies include both task-state fMRI (T-fMRI) and resting-state fMRI (R-fMRI). The former is poor in clinical practicality due to the diversity and complexity of experiments, and compared with the resting state functional magnetic resonance imaging technology, the resting state functional magnetic resonance imaging technology can image the neural activity of a subject under the resting state without external stimulus by measuring blood oxygen level dependent signals, can effectively reflect the functional changes of the brain blood flow, metabolic activity and the like of the subject under the resting state, and is an effective means for researching abnormal brain functional connection, so that the resting state functional magnetic resonance imaging technology is widely applied to clinical and brain science researches.
In the analysis of R-fMRI data, the previous brain function network analysis method generally adopts an independent component analysis or a method matched with a brain template to divide the whole brain into a plurality of regions of interest, and then constructs a brain function network according to the time sequence of each region of interest. According to the method, a brain function connection matrix is constructed according to all brain regions, however, the signal change of the R-fMRI image data only exists in part of brain regions, and a great amount of redundant information exists when a functional network is constructed according to all brain regions, so that the problem of dimension disaster exists.
Disclosure of Invention
The invention aims to overcome the defects in the prior art, provides a fMRI data classification and identification method based on brain region functional connection, and aims to overcome the problem of dimension disaster in fMRI data processing.
In order to achieve the above purpose, the invention is realized by adopting the following technical scheme:
in a first aspect, the present invention provides a fMRI data classification and identification method based on brain region functional connection, including the following steps: obtaining fMRI data of a subject; preprocessing the obtained fMRI data to obtain a grey brain matter image; dividing a grey brain image into a plurality of brain regions with different functions, and extracting an average voxel time sequence of each brain region; selecting a partial brain region with significant difference from a plurality of functional brain regions based on the fuzzy decision rough set; based on the selected partial brain regions, calculating the Pearson correlation coefficient between different brain regions, and carrying out nonlinear processing on the coefficient by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain regions; sparsifying the correlation coefficient values in the matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into one-dimensional feature vectors; and taking the obtained one-dimensional feature vector as input to a trained SVM recognition model, obtaining an output label of the tested person and judging the fMRI data type of the tested person.
Further, preprocessing the obtained fMRI data to obtain a grey brain matter image, including the following specific steps: performing time-layer correction on brain slice data obtained at different time points to adjust the brain slice data to be at the same time point; performing spatial position matching on the fMRI data, and adjusting an irregular posture caused by head movement of a tested person when the resting fMRI data are acquired; uniformly standardizing individual resting fMRI brain images into a standard space by an EPI (Echo planar Imaging) method and adjusting to the same spatial resolution; convolving the resting fMRI data with a 3D gaussian kernel and performing spatial smoothing; linear trend elimination is carried out on the data signals, and band-pass filtering of 0.01-0.08Hz is adopted to remove low-frequency drift and high-frequency noise; and removing ventricles, white matter and global irrelevant signals according to linear regression to obtain gray image signals.
Further, the method for segmenting the grey brain image into a plurality of brain regions with different functions comprises the following specific steps: the whole brain tissue (excluding cerebellum) was segmented into 90 different functional brain regions against a physiological AAL template for grey matter images of the brain.
Further, the fuzzy decision rough set-based selection of a partial brain region with significant differences from a plurality of functional brain regions comprises the following specific steps: the whole brain time sequence in the sample data and the data type label determined according to the fMRI data classification recognition subject form a classification decision table S= (U, A, V, f), wherein the domain U is the acquired all sample sets, A=C U-D, C n-D=phi, wherein C= { a 1 ,a 2 ,…,a 90 The } is a brain region set, the D is the sample category attribute of the tested person, and the information function f is U multiplied by A to V, whereinV a Is a epsilon A value range; and removing redundant brain regions which are irrelevant to the classification and identification results in the whole brain region set by using a heuristic backward search strategy.
Further, the heuristic backward search strategy includes: initializing a partial brain region set B=C; step two: let i=1; step three: calculating the ith brain region a according to the formula (1) i Alpha attribute importance relative to partial brain region set BWherein->An alpha-positive domain for decision attribute D relative to B; step four: if it isB=b-a i The method comprises the steps of carrying out a first treatment on the surface of the Step five: if i= |c|, then the reduction set B is output, otherwise i=i+1 and jumps to step three loop.
Further, the calculation process of the decision attribute D relative to the alpha-positive domain of the brain region set B is as follows: step A1 x i For a sample in the domain U, x i Is the kth brain region a of (2) k Average voxel time series of (a)Where N is the time point length, |B| is the number of partial brain region sets, |F>Representing the kth brain region a k The gray value of the average voxel at time point N; a2, calculating any two tested samples x in the discourse domain U i And x j A similar distance d (x i ,x j ) By measuring sample x i And x j Similarity between the same brain regions to calculate the similarity distance between them +.>Wherein->For sample x i And x j A pearson correlation coefficient at the kth brain region reflecting the correlation between the two time series; />For time series +.>And->Is a covariance of (2); sigma is the standard deviation of the time series; step A3: calculating sample x i And x j Fuzzy similarity relation between->Gaussian kernel parameter δ=0.7; step A4: calculating sample x i Is the fuzzy equivalence class->Step A5: calculating sample x i Fuzzy conditional probability correctly classified into decision class Y>Wherein Y is a data type label set, Y is U/D, U/D is the division of the domain U based on the decision attribute D, and the decision class Y is contained in the decision class D: i.e. one of the class labels; such as: two types of 0/1 tags; class a and class b tags; step A6: for brain region set B, the lower approximation set of YB α (Y)＝{x∈U|P(Y|[x i ] R ) Gtoreq α }, wherein the lower approximation threshold α = 0.9; step A7: calculating the positive field of the decision attribute D relative to B>
Further, the method for obtaining the functional connection matrix of the partial brain region based on the selected partial brain region, calculating the pearson correlation coefficient between different brain regions, and performing nonlinear processing on the coefficient by adopting Fisher-z transformation comprises the following steps: extracting partial brain region after rough set reduction of fuzzy decisionBy the formula->Calculation of brain region a t And a s The pearson correlation coefficient between them, construct brain function connection matrix +.>
Further, the method for thinning the correlation coefficient value in the matrix, reserving the correlation coefficient value above the threshold value, and expanding the matrix into a one-dimensional feature vector comprises the following steps: setting a correlation coefficient threshold ES (es=0.15) and preserving the values of z > ES in the matrix, the coefficient values of the rest being set to zero; taking the feature vector of the upper triangle (without diagonal) of the matrix flattened into one dimension
In a second aspect, the present invention further provides an fMRI data classification and identification device based on brain region functional connection, the device comprising: brain image acquisition module: for acquiring fMRI data of a subject; and a data preprocessing module: the method comprises the steps of preprocessing acquired fMRI data to obtain a brain gray matter image; brain partition module: for segmenting the grey brain image into a plurality of brain regions of different functions and extracting an average voxel time sequence of each brain region; brain region selection module: for selecting a partial brain region having a significant difference from a plurality of functional brain regions based on the fuzzy decision rough set; a transformation matrix module: the method comprises the steps of calculating pearson correlation coefficients among different brain areas based on selected partial brain areas, and carrying out nonlinear processing on the coefficients by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain areas; vector conversion module: the method comprises the steps of performing sparsification on correlation coefficient values in a matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into a one-dimensional feature vector; SVM recognition model module: and the obtained one-dimensional feature vector is used as input to a trained SVM recognition model to obtain an output label of the tested person and judge the category of fMRI data of the tested person.
In a third aspect, the present invention further provides an fMRI data classification and identification device based on brain region functional connection, including a processor and a storage medium; the storage medium is used for storing instructions; the processor is operative according to the instructions to perform steps according to the method described above.
Compared with the prior art, the invention has the beneficial effects that:
1. the method adopts a fuzzy decision rough set to find out partial brain regions with obvious differences among R-fMRI data of a tested person from a plurality of brain regions of the whole brain, has a certain interpretability, and can effectively reduce the dimension of the R-fMRI data characteristic space in the process of processing the R-fMRI data;
2. the method belongs to a pure data driving method, and can automatically and directly find out brain areas closely related to classification results from R-fMRI brain images without manual experience.
Drawings
FIG. 1 is a block flow diagram of an apparatus provided by the present invention;
FIG. 2 is a flow chart of partial brain region selection based on a fuzzy decision rough set.
Detailed Description
The invention is further described below with reference to the accompanying drawings. The following examples are only for more clearly illustrating the technical aspects of the present invention, and are not intended to limit the scope of the present invention.
Embodiment one:
the embodiment provides a fMRI data classification and identification method based on brain region functional connection, which classifies the fMRI data of a tested person into two corresponding classes according to the determined fMRI data classification and identification subject: class A and class B, as shown in FIG. 1, the data classification and identification method mainly comprises the following steps:
step 1: fMRI data acquisition: dividing brain images of a tested person into a class A and a class B according to a subject to be studied, and acquiring resting state functional nuclear magnetic resonance imaging data of the class A and class B tested person;
step 2: preprocessing the acquired R-fMRI brain image data by DPARSF software to realize data standardization and eliminate external interference signals, wherein the data standardization comprises slice time layer correction, head motion correction, spatial standardization, spatial smoothing, linear trend elimination, time band-pass filtering and covariate regression;
step 3: dividing the whole brain tissue (excluding cerebellum) into 90 brain regions with different functions by comparing the preprocessed fMRI brain image data with a physiological AAL template, and extracting an average voxel time sequence of each brain region;
step 4: selecting partial brain regions with significant differences between the class A and class B testees from 90 functional brain regions based on the fuzzy decision rough set;
step 5: based on the selected partial brain regions, calculating the Pearson correlation coefficient between different brain regions of each sample, and carrying out nonlinear processing on the coefficients by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain regions;
step 6: sparsifying the correlation coefficient values in the matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into one-dimensional feature vectors;
step 7: training the one-dimensional feature vector as the input of a support vector machine, and establishing a classification recognition model;
step 8: and 2-6 steps of processing the R-fMRI brain image of the tested person to obtain a one-dimensional feature vector, and sending the one-dimensional feature vector as input into a trained classification and identification model to obtain an output label of the tested person so as to judge which data feature the tested person belongs to.
The fMRI data preprocessing in the step 2 comprises the following specific steps:
(1) Performing time-layer correction on brain slice data obtained at different time points to adjust the brain slice data to be at the same time point;
(2) Performing spatial position matching on the head of the brain, and adjusting an irregular posture caused by head movement of a tested person when R-fMRI data are acquired;
(3) The method EPI (Echo planar Imaging) is adopted to uniformly normalize the individual R-fMRI brain images to a standard space established by Montreal neurological institute (Montreal neurological institute, MNI) and adjust to the same spatial resolution;
(4) Convolving the R-fMRI data with a 3D gaussian kernel and performing spatial smoothing;
(5) Removing the linear trend of the data signal, and removing low-frequency drift and high-frequency noise by adopting band-pass filtering of 0.01-0.08 Hz;
(6) And removing ventricle, white matter and global irrelevant signals by using linear regression, and retaining gray image signals.
The step 4 partial brain region selection comprises the following specific steps:
as shown in fig. 2, the process of selecting a partial brain region with a significant difference between class a and class b from 90 brain regions of the whole brain based on the fuzzy decision rough set is as follows:
(1) The whole brain time sequence in the sample data and the corresponding class labels of the first class data and the second class data form a class decision table S= (U, A, V and f), wherein the domain U is an acquired sample set of all testees, A=C U-D, C n-D=phi, and C= { a 1 ,a 2 ,…,a 90 The information function f is U multiplied by A to V, wherein the value range is setV a Is a epsilon A value range;
(2) And removing redundant brain regions irrelevant to classification and identification results in the whole brain region set by using a heuristic backward search strategy, wherein the heuristic backward search strategy is as follows:
c1: initializing a partial brain region set b=c;
c2: let i=1;
and C3: calculating the ith brain region a according to the formula (1) i Relative toAlpha attribute importance of partial brain region set B
Wherein the method comprises the steps ofAn alpha-positive domain of D relative to B;
and C4: if it isB=b-a i ；
C5: if i= |c|, then the reduction set B is output, otherwise i=i+1 and jumps to the C3 loop.
The alpha-positive domain calculation process of the decision attribute D relative to the brain region set B is as follows:
A1:x i for a sample of a tested person in the domain U, x i Is the kth brain region a of (2) k Average voxel time series of (a)Wherein N is the length of the time point, B is the number of partial brain region sets,representing the kth brain region a k The gray value of the average voxel at time point N.
A2, calculating any two tested person samples x in the domain U i And x j A similar distance d (x i ,x j ) By measuring sample x according to formula (2) i And x j Similarity between identical brain regions to calculate a similarity distance d (x i ,x j )，
Wherein the method comprises the steps ofFor sample x i And x j In the kth brain region a k Is a pearson correlation coefficient reflecting the correlation between two time series, equation (3) is as follows:
wherein the method comprises the steps ofFor time series +.>And->σ is the standard deviation of the time series.
A3: calculating sample x i And x j Fuzzy similarity relation mu between R (x i ,x j ) Equation (4) is shown below:
where δ is the gaussian kernel parameter.
A4: calculating sample x i Is of the fuzzy equivalence class of (2)Equation (5) is shown below:
a5: calculating sample x i Fuzzy conditional probability correctly classified into decision class Y, publicFormula (6) is shown below:
wherein Y is a data type label set, Y epsilon U/D is the division of the domain U based on the decision attribute D.
A6: for brain region set B, the lower approximation set of YB α (Y)＝{x∈U|P(Y|[x i ] R ) ∈α }, where α is the lower approximate threshold.
A7: calculating the positive domain of decision class D relative to B according to equation (7)
The resting state functional network analysis comprises the following steps:
based on selected partial brain regionsWherein n represents the number of partial brain regions and n is less than or equal to 90, and the brain region a is calculated by the formula (8) t And a s Pearson correlation coefficient between->Equation (8) is as follows:
construction of brain function connection matrixAnd performing nonlinear mapping by using Fisher-z transformation so that the correlation coefficient value accords with normal distribution.
In addition, it should be noted that the method of the present embodiment may be applied to not only a plurality of classifications of two, three, four, and so on.
The method adopts the fuzzy decision rough set to find out partial brain regions with obvious difference between class A and class B tested data from 90 brain regions of the whole brain, has certain interpretability, and can effectively reduce the dimension of the R-fMRI data characteristic space. The method belongs to a pure data driving method, and can automatically and directly find out brain areas closely related to classification results from R-fMRI brain images without manual experience. The method can be further popularized to fMRI data classification and identification in the fields of bioinformatics, material science and the like, and has good universality and application prospect.
Embodiment two:
the embodiment provides a specific method for extracting and identifying fMRI features based on partial brain region functional connection, for example, determining autism as a subject of fMRI data classification identification, dividing fMRI data into class a and class b, and respectively representing brain connection abnormal testees and brain connection normal testees, wherein the method comprises the following steps:
step one: 184 tested R-fMRI brain images provided by NYU Langone Medical Center are selected from an autism brain imaging data exchange database ABIDE (http:// fcon_1000. Subjects. Nitrc. Org/indi/ABIDE/abide_I. Html), and the age of the tested person is distributed below 18 years old. Of these, 79 were tested, with a ratio of 68/11 for men and women, an average age of 14.52 years, and a PIQ (Performance Intelligence Quotient) mean of 104; class B was tested in 105 cases, with a ratio of 79/26 for men and women, an average age of 9.46 years, and PIQ mean 113.
Step two: the DPARSF software was used to pre-process the R-fMRI data. The temporal differences of the different slices acquired within the scan are first corrected, and then the brain is readjusted to an intermediate position to correct for inter-scan head motion. The functional image is then spatially normalized to the MNI152 standard template, resampling the size is 3X 3mm 3 . Then using a FWHM (full width at half maximum) Gaussian kernel of 6mm to perform space smoothing, and performing linear trend elimination on each voxel BOLD signal, taking a band-pass filter of 0.01-0.08Hz, and removing low-frequency driftShift and high frequency noise. And eliminating interference co-variables such as head motion parameters, average signals, white matter signals, cerebrospinal fluid signals and the like through linear regression. Finally, the brain was segmented into 90 regions of interest using an automatic anatomical landmark AAL brain atlas.
Step three: the process of selecting partial brain regions with significant differences between class A and class B from 90 brain regions of the whole brain based on the fuzzy decision rough set is as follows:
(1) The whole brain time sequence in the sample data and the corresponding classification label form a classification decision table S= (U, A, V, f), wherein the universe U is the acquired collection of all tested samples, A=C.u.D, C.u.D=phi, wherein C.u. { a = 1 ,a 2 ,…,a 90 The value range is set in brain region, the value range is set in U x A-V, and the value range is set in the information function fV a Is the value range of a epsilon A.
(2) And removing redundant brain areas irrelevant to the class A and class B tested in the whole brain area set by using a heuristic backward search strategy, wherein the heuristic backward search strategy is as follows:
c1: initializing a partial brain region set b=c;
c2: let i=1;
and C3: calculating the ith brain region a according to the formula (9) i Alpha attribute importance relative to partial brain region set B
Wherein the method comprises the steps ofAn alpha-positive domain of D relative to B;
and C4: if it isB=b-a i ；
C5: if i= |c|, then the reduction set B is output, otherwise i=i+1 and jumps to the C3 loop.
The alpha-positive domain calculation process of the decision class D relative to the brain region set B is as follows:
A1:x i for a sample in the domain U, x i Is the kth brain region a of (2) k Average voxel time series of (a)Where N is the time point length, |B| is the number of partial brain region sets, |F>Representing the kth brain region a k The gray value of the average voxel at time point N.
A2, calculating any two tested samples x in the discourse domain U i And x j A similar distance d (x i ,x j ) By measuring sample x according to formula (10) i And x j Similarity between identical brain regions to calculate a similarity distance d (x i ,x j )，
Wherein the method comprises the steps ofFor sample x i And x j In the kth brain region a k Is a pearson correlation coefficient reflecting the correlation between two time series, formula (11) is as follows:
wherein the method comprises the steps ofFor time series +.>And->σ is the standard deviation of the time series.
A3: calculating sample x i And x j Fuzzy similarity relation mu between R (x i ,x j ) Equation (12) is shown below:
where δ is the gaussian kernel parameter.
A4: calculating sample x i Is of the fuzzy equivalence class of (2)Equation (13) is shown below:
a5: calculating sample x i The probability of a fuzzy condition being correctly classified into decision class Y, equation (14) is as follows:
wherein Y is a data type label set, Y epsilon U/D is the division of the domain U based on the decision attribute D.
A6: for brain region set B, the lower approximation set of YB α (Y)＝{x∈U|P(Y|[x i ] R ) ∈α }, where α is the lower approximate threshold.
A7: calculating the positive domain of decision class D relative to B according to equation (15)
The resting state functional network analysis comprises the following steps:
based on selected partial brain regionsWherein n represents the number of partial brain regions and n is not more than 90, and the brain region a is calculated by the formula (16) t And a s The pearson correlation coefficient therebetween, equation (16) is as follows:
construction of brain function connection matrixAnd performing nonlinear processing on the characteristics by using a Fisher-z transformation formula, wherein the Fisher-z transformation formula is as follows:
so that the correlation coefficient value accords with the normal distribution to obtain a matrix
Step five: let the correlation coefficient threshold (es=0.15) and preserve the values of z > ES in the matrix, the coefficient values of the rest being set to zero. Taking the feature vector of the upper triangle (without diagonal) of the matrix flattened into one dimensionWherein the number of extracted feature vectors +.>And finally, taking the feature vector F and a corresponding sample label as input of the SVM, and training a classification model by ten times of cross validation. And obtaining a one-dimensional vector after processing the new sample data, inputting the one-dimensional vector into a trained SVM classification recognition model, and confirming which data type the one-dimensional vector belongs to, thereby determining whether the tested person of the new sample belongs to class A or class B.
Embodiment III:
the present embodiment provides a fMRI data classification and identification device based on brain region functional connection, which can implement the method described in the first embodiment, and the device includes:
brain image acquisition module: for acquiring brain magnetic resonance imaging (fMRI) data of a subject;
and a data preprocessing module: the method comprises the steps of preprocessing acquired brain nuclear magnetic resonance imaging data to obtain a brain gray image;
brain partition module: for segmenting the grey brain image into a plurality of brain regions of different functions and extracting an average voxel time sequence of each brain region;
brain region selection module: for selecting a partial brain region having a significant difference from a plurality of functional brain regions based on the fuzzy decision rough set;
a transformation matrix module: the method comprises the steps of calculating pearson correlation coefficients among different brain areas based on selected partial brain areas, and carrying out nonlinear processing on the coefficients by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain areas;
vector conversion module: the method comprises the steps of performing sparsification on correlation coefficient values in a matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into a one-dimensional feature vector;
SVM recognition model module: the method is used for inputting the obtained one-dimensional feature vector into a trained SVM recognition model to obtain an output label of a tested person and judging the category of brain nuclear magnetic resonance imaging data of the tested person.
Embodiment four:
the fMRI data classification and identification device based on brain region functional connection is characterized by comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor is configured to operate in accordance with the instructions to perform the steps of the method of embodiment one.
The foregoing is merely a preferred embodiment of the present invention, and it should be noted that modifications and variations could be made by those skilled in the art without departing from the technical principles of the present invention, and such modifications and variations should also be regarded as being within the scope of the invention.
Claims (8)
1. The fMRI data classification and identification method based on brain region functional connection is characterized by comprising the following steps of:
obtaining fMRI data of a subject;
preprocessing the obtained fMRI data to obtain a grey brain matter image;
dividing a grey brain image into a plurality of brain regions with different functions, and extracting an average voxel time sequence of each brain region;
selecting a partial brain region with significant difference from a plurality of functional brain regions based on the fuzzy decision rough set;
based on the selected partial brain regions, calculating the Pearson correlation coefficient between different brain regions, and carrying out nonlinear processing on the coefficient by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain regions;
sparsifying the correlation coefficient values in the matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into one-dimensional feature vectors;
the obtained one-dimensional feature vector is used as input to a trained SVM recognition model, so as to obtain an output label of a tested person and judge the fMRI data type of the tested person;
the fuzzy decision rough set-based selection of partial brain regions with significant differences from a plurality of functional brain regions comprises the following specific steps:
the whole brain time sequence in the sample data and the data type label determined according to the fMRI data classification recognition subject form a classification decision tableS= (U, a, V, f), where the universe U is the set of all samples taken, a=c U D, C d=phi, where c= { a 1 ,a 2 ,…,a 90 A brain region set, D is a sample class attribute set of the tested person, f is an information function from U x A to V, which assigns an information value to each attribute of each object, whereinV is the set of value ranges, V a Is a epsilon A value range;
removing redundant brain regions irrelevant to classification and identification results in the whole brain region set by utilizing a heuristic backward search strategy;
the heuristic backward search strategy comprises the following steps:
c1: initializing a partial brain region set b=c;
c2: let i=1;
and C3: calculating the ith brain region a according to the formula (9) i Alpha attribute importance relative to partial brain region set B
Wherein the method comprises the steps ofAn alpha-positive domain of D relative to B;
and C4: if it isB=b-a i ；
C5: if i= |c|, then the reduction set B is output, otherwise i=i+1 and jumps to the C3 loop.
2. The fMRI data classification and identification method based on brain region functional connection according to claim 1, wherein the preprocessing of the obtained fMRI data to obtain a grey brain matter image comprises the following specific steps:
performing time-layer correction on brain slice data obtained at different time points to adjust the brain slice data to be at the same time point;
performing spatial position matching on the fMRI data, and adjusting an irregular posture caused by head movement of a tested person when the resting fMRI data are acquired;
uniformly standardizing an individual resting state fMRI brain image into a standard space by a background plane echo imaging method, and adjusting to the same spatial resolution;
convolving the resting fMRI data with gaussian kernels and performing spatial smoothing;
linear trend elimination is carried out on the data signals, and band-pass filtering of 0.01-0.08Hz is adopted to remove low-frequency drift and high-frequency noise;
and removing ventricles, white matter and global irrelevant signals according to linear regression to obtain gray image signals.
3. The fMRI data classification and identification method based on brain region functional connection according to claim 1, wherein the method of dividing a brain gray matter image into a plurality of brain regions with different functions comprises the following specific steps:
the whole brain tissue was segmented into 90 different functional brain regions against a physiological AAL template.
4. The fMRI data classification and identification method based on brain region functional connection according to claim 1, wherein the calculation process of the decision attribute D with respect to the α -positive domain of the brain region set B is as follows:
step A1 x i For a sample in the domain U, x i Is the kth brain region a of (2) k Average voxel time series of (a)Wherein N is the length of the time point, B is the number of partial brain region sets,representing the kth brain region a k The gray value of the average voxel at time point N;
a2, calculating any two tested samples x in the discourse domain U i And x j A similar distance d (x i ,x j ) By measuring sample x i And x j Similarity between identical brain regions to calculate a similarity distance between themWherein the method comprises the steps ofFor sample x i And x j A pearson correlation coefficient at the kth brain region reflecting the correlation between the two time series; />For time series +.>And->Is a covariance of (2); />For time series +.>And->Standard deviation of (2);
step A3: calculating sample x i And x j Fuzzy similarity relation betweenδ is a gaussian kernel parameter, here δ=0.7;
step A4: calculating sample x i Is of the fuzzy equivalence class of (2)
Step A5: calculating sample x i Fuzzy conditional probability of correctly classified into decision class YWherein Y is a data category label set, and Y epsilon U/D is the division of the domain U based on the decision attribute D,wherein mu Y (x j )＝1,x j ∈Y；/>
Step A6: for brain region set B, the lower approximation set of YB α (Y)＝{x∈U|P(Y|[x i ] R ) Gtoreq α }, wherein the lower approximation threshold α = 0.9;
step A7: calculating the positive field of decision attribute D relative to B
5. The fMRI data classification and identification method based on brain region functional connection of claim 4, wherein the method for calculating pearson correlation coefficients between different brain regions based on the selected partial brain regions and performing nonlinear processing on the coefficients by using Fisher-z transformation to obtain the functional connection matrix of the partial brain regions comprises the following steps:
extracting partial brain region after rough set reduction of fuzzy decision
By the formulaCalculation of brain region a t And a s Peel in betweenForest correlation coefficient, a t ,a s E, B, constructing a brain function connection matrix ∈B> Respectively, the samples are in brain region a t And a s Time series on>Time series>And->Standard deviation of (2).
6. The method for classifying and identifying fMRI data based on brain region functional connections according to claim 5, wherein the method for thinning the correlation coefficient values in the matrix, retaining the correlation coefficient values above the threshold, and expanding the matrix into one-dimensional eigenvectors comprises the steps of:
setting the value of a correlation coefficient threshold ES and reserving the value of z & gtES in the matrix, wherein the coefficient value of the rest is set to be zero;
taking the upper triangle of the matrix, without diagonal lines, flattening and folding the upper triangle into one-dimensional characteristic vector
7. An fMRI data classification and identification device based on brain region functional connection, characterized in that the device comprises:
brain image acquisition module: for acquiring fMRI data of a subject;
and a data preprocessing module: the method comprises the steps of preprocessing acquired fMRI data to obtain a brain gray matter image;
brain partition module: for segmenting the grey brain image into a plurality of brain regions of different functions and extracting an average voxel time sequence of each brain region;
brain region selection module: for selecting a partial brain region having a significant difference from a plurality of functional brain regions based on the fuzzy decision rough set;
a transformation matrix module: the method comprises the steps of calculating pearson correlation coefficients among different brain areas based on selected partial brain areas, and carrying out nonlinear processing on the coefficients by adopting Fisher-z transformation to obtain a functional connection matrix of the partial brain areas;
vector conversion module: the method comprises the steps of performing sparsification on correlation coefficient values in a matrix, reserving the correlation coefficient values above a threshold value, and expanding the matrix into a one-dimensional feature vector;
SVM recognition model module: the method comprises the steps of using the obtained one-dimensional feature vector as input to a trained SVM recognition model, obtaining an output label of a tested person and judging the category of fMRI data of the tested person;
the fuzzy decision rough set-based selection of partial brain regions with significant differences from a plurality of functional brain regions comprises the following specific steps:
the whole brain time sequence in the sample data and the data type label determined according to the fMRI data classification recognition subject form a classification decision table S= (U, A, V, f), wherein the domain U is the acquired all sample sets, A=C U-D, C n-D=phi, wherein C= { a 1 ,a 2 ,…,a 90 A brain region set, D is a sample class attribute set of the tested person, f is an information function from U x A to V, which assigns an information value to each attribute of each object, whereinV is the set of value ranges, V a Is a epsilon A value range;
removing redundant brain regions irrelevant to classification and identification results in the whole brain region set by utilizing a heuristic backward search strategy;
the heuristic backward search strategy comprises the following steps:
c1: initializing a partial brain region set b=c;
c2: let i=1;
and C3: calculating the ith brain region a according to the formula (9) i Alpha attribute importance relative to partial brain region set B
Wherein the method comprises the steps ofAn alpha-positive domain of D relative to B;
and C4: if it isB=b-a i ；
C5: if i= |c|, then the reduction set B is output, otherwise i=i+1 and jumps to the C3 loop.
8. The fMRI data classification and identification device based on brain region functional connection is characterized by comprising a processor and a storage medium;
the storage medium is used for storing instructions;
the processor being operative according to the instructions to perform the steps of the method according to any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011095774.9A CN112233086B (en) | 2020-10-14 | 2020-10-14 | fMRI data classification and identification method and device based on brain region functional connection |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011095774.9A CN112233086B (en) | 2020-10-14 | 2020-10-14 | fMRI data classification and identification method and device based on brain region functional connection |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112233086A CN112233086A (en) | 2021-01-15 |
| CN112233086B true CN112233086B (en) | 2023-08-04 |
Family
ID=74112804
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011095774.9A Active CN112233086B (en) | 2020-10-14 | 2020-10-14 | fMRI data classification and identification method and device based on brain region functional connection |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112233086B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113052800B (en) * | 2021-03-09 | 2022-02-22 | 山东大学 | Alzheimer disease image analysis method and device |
| WO2023123380A1 (en) * | 2021-12-31 | 2023-07-06 | 深圳先进技术研究院 | Dynamic addiction neural circuit generation method and system based on weakly supervised contrastive learning |
| CN115909016B (en) * | 2023-03-10 | 2023-06-23 | 同心智医科技(北京)有限公司 | GCN-based fMRI image analysis system, method, electronic equipment and medium |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392907A (en) * | 2017-09-01 | 2017-11-24 | 上海理工大学 | Parahippocampal gyrus function division method based on tranquillization state FMRI |
| CN110414605A (en) * | 2019-07-30 | 2019-11-05 | 西南交通大学 | A kind of feature selection approach based on the steady Fuzzy-rough Set Model of multicore |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2992823B1 (en) * | 2013-05-01 | 2021-06-30 | Advanced Telecommunications Research Institute International | Brain activity training device |
-
2020
- 2020-10-14 CN CN202011095774.9A patent/CN112233086B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107392907A (en) * | 2017-09-01 | 2017-11-24 | 上海理工大学 | Parahippocampal gyrus function division method based on tranquillization state FMRI |
| CN110414605A (en) * | 2019-07-30 | 2019-11-05 | 西南交通大学 | A kind of feature selection approach based on the steady Fuzzy-rough Set Model of multicore |
Non-Patent Citations (1)
| Title |
|---|
| M.D. Alexiuk等.Cluster validation indices for fMRI data: Fuzzy C-Means with feature partitions versus cluster merging strategies. IEEE Annual Meeting of the Fuzzy Information, 2004. Processing NAFIPS '04..2004,全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112233086A (en) | 2021-01-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11049243B2 (en) | Target detection in latent space | |
| CN112233086B (en) | fMRI data classification and identification method and device based on brain region functional connection | |
| Wismüller et al. | Cluster analysis of biomedical image time-series | |
| Kapur et al. | Segmentation of brain tissue from magnetic resonance images | |
| US10235750B2 (en) | Segmentation of cardiac magnetic resonance (CMR) images using a memory persistence approach | |
| Gerig et al. | Automating segmentation of dual-echo MR head data | |
| CN112837274B (en) | Classification recognition method based on multi-mode multi-site data fusion | |
| Zacharaki et al. | Measuring brain lesion progression with a supervised tissue classification system | |
| Mørch et al. | Nonlinear versus linear models in functional neuroimaging: Learning curves and generalization crossover | |
| US9990719B2 (en) | Method and system for generating multiparametric nosological images | |
| CN108596228B (en) | Brain function magnetic resonance image classification method based on unsupervised fuzzy system | |
| Byeon et al. | Artificial neural network inspired by neuroimaging connectivity: application in autism spectrum disorder | |
| Daliri et al. | Skull segmentation in 3D neonatal MRI using hybrid Hopfield Neural Network | |
| CN116433967B (en) | Personalized target spot selection method oriented to noninvasive nerve regulation technology | |
| Elakkiya | Toward improving the accuracy in the diagnosis of schizophrenia using functional magnetic resonance imaging (fMRI) | |
| Cocosco et al. | Automatic generation of training data for brain tissue classification from MRI | |
| Al-Waeli | An automated system for the classification and segmentation of brain tumours in MRI images based on the modified grey level co-occurrence matrix | |
| VIANNEY-KINANI et al. | Computer-aided diagnosis of brain tumors using image enhancement and fuzzy logic | |
| Gritsenko et al. | Automatic Identification of Twin Zygosity in Resting-State Functional MRI | |
| Sen | Generalized Prediction Model for Detection of Psychiatric Disorders | |
| Muldoon et al. | Lightweight and interpretable left ventricular ejection fraction estimation using mobile U-Net | |
| Giacomantone et al. | ROC performance evaluation of RADSPM technique | |
| Lemieux | Automatic, accurate, and reproducible segmentation of the brain and cerebro-spinal fluid in T1-weighted volume MRI scans and its application to serial cerebral and intracranial volumetry | |
| CN116958683A (en) | MRI data classification method based on low-rank multi-mode fusion network | |
| Tripoliti et al. | Recent developments in computer methods for fMRI data processing |
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 |