CN101292871A - Method for specification extraction of magnetic resonance imaging brain active region based on pattern recognition - Google Patents
Method for specification extraction of magnetic resonance imaging brain active region based on pattern recognition Download PDFInfo
- Publication number
- CN101292871A CN101292871A CNA2007100986913A CN200710098691A CN101292871A CN 101292871 A CN101292871 A CN 101292871A CN A2007100986913 A CNA2007100986913 A CN A2007100986913A CN 200710098691 A CN200710098691 A CN 200710098691A CN 101292871 A CN101292871 A CN 101292871A
- Authority
- CN
- China
- Prior art keywords
- voxel
- distance
- different
- active region
- data
- 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
- 238000000034 method Methods 0.000 title claims abstract description 35
- 238000002595 magnetic resonance imaging Methods 0.000 title claims abstract description 9
- 210000004556 brain Anatomy 0.000 title claims description 90
- 238000000605 extraction Methods 0.000 title claims description 12
- 238000003909 pattern recognition Methods 0.000 title claims description 3
- 230000000694 effects Effects 0.000 claims abstract description 39
- 238000002599 functional magnetic resonance imaging Methods 0.000 claims abstract description 39
- 238000012360 testing method Methods 0.000 claims description 42
- 238000004458 analytical method Methods 0.000 claims description 35
- 238000004422 calculation algorithm Methods 0.000 claims description 35
- 230000006870 function Effects 0.000 claims description 27
- 238000000491 multivariate analysis Methods 0.000 claims description 10
- 230000008707 rearrangement Effects 0.000 claims description 6
- 238000002203 pretreatment Methods 0.000 claims description 5
- 230000008569 process Effects 0.000 claims description 4
- 238000007635 classification algorithm Methods 0.000 claims description 3
- 238000010219 correlation analysis Methods 0.000 claims description 3
- 238000009795 derivation Methods 0.000 claims description 3
- 238000002156 mixing Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000012706 support-vector machine Methods 0.000 claims description 3
- 238000013401 experimental design Methods 0.000 claims description 2
- 238000001558 permutation test Methods 0.000 claims 1
- 230000000638 stimulation Effects 0.000 abstract description 8
- 230000002490 cerebral effect Effects 0.000 abstract 13
- 230000033001 locomotion Effects 0.000 abstract 3
- 238000001514 detection method Methods 0.000 description 20
- 230000004913 activation Effects 0.000 description 17
- 238000007405 data analysis Methods 0.000 description 8
- 238000001914 filtration Methods 0.000 description 8
- 238000007473 univariate analysis Methods 0.000 description 8
- 238000002474 experimental method Methods 0.000 description 7
- 238000009499 grossing Methods 0.000 description 7
- 230000003925 brain function Effects 0.000 description 6
- 238000006243 chemical reaction Methods 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 239000000284 extract Substances 0.000 description 4
- 239000011159 matrix material Substances 0.000 description 4
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 description 3
- 239000008280 blood Substances 0.000 description 3
- 210000004369 blood Anatomy 0.000 description 3
- 230000001149 cognitive effect Effects 0.000 description 3
- 230000000004 hemodynamic effect Effects 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 239000002085 irritant Substances 0.000 description 3
- 231100000021 irritant Toxicity 0.000 description 3
- 229910052760 oxygen Inorganic materials 0.000 description 3
- 239000001301 oxygen Substances 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 230000035945 sensitivity Effects 0.000 description 3
- 238000004088 simulation Methods 0.000 description 3
- 230000004936 stimulating effect Effects 0.000 description 3
- 238000005481 NMR spectroscopy Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 238000012937 correction Methods 0.000 description 2
- 230000004044 response Effects 0.000 description 2
- 101100009524 Anas platyrhynchos HSD17B12 gene Proteins 0.000 description 1
- 206010020751 Hypersensitivity Diseases 0.000 description 1
- 239000006002 Pepper Substances 0.000 description 1
- 101100256975 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) SIP2 gene Proteins 0.000 description 1
- 101100203850 Saccharomyces cerevisiae (strain ATCC 204508 / S288c) SPT2 gene Proteins 0.000 description 1
- 101100291461 Schizosaccharomyces pombe (strain 972 / ATCC 24843) spm2 gene Proteins 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 208000026935 allergic disease Diseases 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000003727 cerebral blood flow Effects 0.000 description 1
- 230000036992 cognitive tasks 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
- 210000002451 diencephalon Anatomy 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000009610 hypersensitivity Effects 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000002610 neuroimaging Methods 0.000 description 1
- 238000012797 qualification Methods 0.000 description 1
- 238000009877 rendering Methods 0.000 description 1
- 238000005316 response function Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 230000003335 steric effect Effects 0.000 description 1
- 230000002123 temporal effect Effects 0.000 description 1
Images
Landscapes
- Magnetic Resonance Imaging Apparatus (AREA)
Abstract
The present invention discloses an arithmetic of picking up magnetic resonance imaging cerebral active regions by sorting based on mode identification, which comprises the steps that cerebral active regions are extracted based on the multi-element mode distance between fine activity modes in partial cerebral regions for the pretreatment of an fMRI image; a partial consistent cerebral region is obtained by clustering; the combined activities of a plurality of tissues inside the partial consistent cerebral region are used for constructing the multi-element mode; the multi-element distance function is constructed by a mode sorting method to measure the separable characters of the partial cerebral region motion under different stimulation conditions, so as to judge whether the cerebral region is activated or not. The present invention indicates the cerebral motions under different stimulation conditions by multi-element mode information formed by multiple tissues inside the partial cerebral region directly, the multi-element mode can reflect the partial cerebral motion state over all, and multi-element statistical distance can be effectively integrated with the information in the partial cerebral region to measure the difference between different cerebral activate states, so the multi-element mode and the multi-element statistical distance ensure that the arithmetic of the present invention can detect the fine cerebral action mode more completely than the traditional fMRI analyzing technology.
Description
Technical field
The invention belongs to neuroimaging data analysis technique field, be specifically related to nuclear magnetic resonance fMRI active region extraction algorithm, relate in particular to and use multi-mode recognition methods carrying out fMRI brain active region.
Background technology
(functional Magnetic Resonance Imaging, fMRI) with its high-spatial and temporal resolution, characteristics such as non-intrusion type have obtained extensive use in human brain function research with functional mri.FMRI generally is meant based on blood oxygen level and relies on (blood oxygen level-dependent, BOLD) fMRI imaging, it changes the magnetic resonance signal that causes and changes and react cerebration by measuring compositions such as the cerebral blood flow that caused by neural activity and brain blood oxygen.Along with the sharp increase of test data in recent years, it is important all the more that rational and effective fMRI data analysis technique shows.The key problem of performance data analysis be according to the fMRI data that record seek its activity can significantly distinguish the different tests condition (as, incentive condition and base line condition) the brain district.Reasonably the fMRI analytical method need be considered the general organizational principle of human brain function and some characteristics of fMRI data itself simultaneously.
It is generally acknowledged that brain function is followed two basic organizational principles: integrated and specialization of function of function.On large space scale (scale), the brain function of a complexity may be finished by interact (integrated) by the special brain district of many functions; Conversely, specificity brain district also can represent or processes, come outside different stimulated is represented by distributed cerebrations different on the scale of meticulous space many different cognitive tasks.On the other hand, along with the MRI development of technology, the fMRI image spatial resolution is improving always.Tradition fMRI data are respectively tieed up at voxel and can be obtained higher signal to noise ratio when width is about 4 millimeters.Current, each ties up width is that 2 millimeters voxel can be on standard 3T clinical magnetic resonance (MRI) machine obtains reliably.(〉=4T) the use of mr techniques, the spatial resolution of fMRI image is approached on submillimeter scale along with super high field.The raising of fMRI image spatial resolution, for studying brain function on meticulousr space scale, we provide possibility: yet, traditional fMRI data analysis is mostly based on generalized linear model (General Linear Model, GLM), GLM is a kind of monobasic statistical technique in essence, whether it decides this voxel to be activated by the time series of analyzing each voxel isolatedly, has ignored the mutual relation (especially lying in the pattern information in the local brain district) between different voxels fully.Thereby traditional univariate analysis technology can't detect the cerebration pattern on the meticulous scale that test stimulus causes, and constitute these patterns to the faint voxel of irritant reaction.In addition, in order to improve signal to noise ratio and statistics effectiveness, the univariate analysis technology can adopt gaussian kernel that the fMRI data are carried out space smoothing usually.Data smoothing is equivalent to low-pass filtering in essence, and it can blur those meticulous cerebration signals that contains the neuroscience relevant information.Thereby a large amount of meticulous information will be filtered out, and the high spatial resolution information that fMRI provides still is far from being utilized.
Summary of the invention
In order on meticulousr scale, to distinguish the cerebration pattern that causes by different stimulated, make full use of the information that comprises in the fMRI data, it is brand-new to the invention describes a class, based on pattern knowledge method detect the brain active region algorithm (local multivariate distance mapping brainactivation, LMDM).The multi-mode information that the present invention directly uses multi-voxel proton in the local brain district to constitute is represented the cerebration under the different stimulated condition, and then measures the separability of local brain district's activity under the different stimulated condition with the polynary distance between different mode as statistic.Multi-mode can reflect local cerebration state all sidedly, multivariate statistics distance as the multi-voxel proton information of statistic in then can more effectively integrated local brain district in order to distinguish different brain state of activation.Thereby with respect to the univariate analysis technology, the present invention can extract the brain active region more accurately, differentiates Different Cognitive condition hypencephalon active state.
The GLM method supposes that each voxel is separate, thereby when carrying out location, brain active region, has ignored the pattern information that is included in the local brain district fully.From the angle of cognitive neuroscience, it has ignored the probability that does not have one-to-one relationship in human brain specific function and the fMRI data between single voxel.In order to solve the shortcoming that exists in the existing fMRI data analysis technique kind, can on meticulousr scale, distinguish the cerebration pattern that causes by different stimulated, make full use of the information that comprises in the fMRI data, the purpose of this invention is to provide a class and can extract the brain active region more accurately, differentiate the brain active region detection algorithm based on pattern knowledge method of Different Cognitive condition hypencephalon active state.
In order to realize purpose of the present invention, the invention provides the brain active region detection algorithm of a class based on pattern knowledge method, comprise step:
Pre-treatment step A: the magnetic resonance image (MRI) that collects is carried out pretreatment, remove and obscure factor, be used for acquisition standardized images data;
Structure locally consistent brain district step B: to each voxel v of view data
0Use region growing algorithm with v
0As seed points, obtain comprising the locally consistent brain district N (v of K voxel
0);
Sliced time sequence step C: cut apart locally consistent brain district N (v according to the incentive condition classification of experimental design correspondence
0) in the time series of a plurality of voxels, then every class incentive condition correspondence a different set of multivariate data sample;
Structure multi-mode statistical distance step D: establish under the different stimulated condition locally consistent brain district N (v
0) in the spatial model correspondence that constitutes of K voxel different K and tie up random distribution; Use mode identification method structure multivariate statistics distance function, in order to tolerance locally consistent brain district N (v
0) statistical distance between the data sample of interior different condition correspondence; By to the above process of each voxel iteration, calculate the multivariate statistics distance, obtain multivariate statistics distance parameter figure;
Polynary distance statistics Parameter Map testing of hypothesis step e: suppose at each locally consistent brain district N (v
0) activity pattern under the different stimulated condition moving phase with, use the nonparametric sequential test that the statistical distance parameter of each voxel correspondence is tested, obtain significance nonparametric sequential test value figure;
Testing of hypothesis multiple comparisons aligning step F: utilize wrong discovery rate FDR method, be used to eliminate multiple comparisons, obtain significantly to distinguish the local brain district of different tests condition, promptly corresponding active region.
According to embodiments of the invention, the described standardized images data of pre-treatment step also comprise:
Steps A .1: the different samples of view data are carried out standardization, and making its average is 0, and standard deviation is 1;
Steps A .2: the view data different characteristic is carried out standardization, and making its average is 0, and standard deviation is 1.
According to embodiments of the invention, described structure locally consistent brain district step also comprises:
Step is B.1: the Pearson came Pearson correlation coefficient of using voxel activity time sequence is measured the activity similar quality between different voxels as criterion;
Step is B.2: use voxel v
0As seed points, the each selection of region growing algorithm joins in the locally consistent brain district with the most similar field voxel conduct of seed points activity; The region growing iteration is carried out, and stops after reaching specified regional area size; Thereby the movable all basically identicals of the voxel in the regional area that obtains.
According to embodiments of the invention, described structure multi-mode statistical distance step also comprises:
Step is D.1: voxel v
0Locally consistent brain district N (v
0) at different stimulated condition X, under the Y, the associating activity of a regional area K voxel that records, activity are from two different multiple random variable X=(X
1, X
2..., X
i..., X
K)
T, Y=(Y
1, Y
2..., Y
i..., Y
K)
T, i=1,2 ..., the pattern sample that the K sampling obtains;
Step is D.2: establish incentive condition X and Y corresponding sample set S respectively
XAnd S
YAccording to mode identification method, construct different polynary distance functions and measure sample set S
XAnd S
YBetween distance, this polynary distance is the Fisher that derives according to the Fisher linear discriminant analysis apart from 0 from function, the maximum frontier distance function of deriving according to support vector machine, and by the polynary distance function of other pattern classification algorithm to derivation;
Step is D.3: be that example describes here with FLDF, according to Fisher linear discriminant analysis, S set
XAnd S
YBetween FLDF calculate by following formula:
Wherein,
Be respectively sample S
XAnd S
YAverage vector promptly:
Σ is a sample mixing covariance:
According to embodiments of the invention, polynary distance statistics Parameter Map testing of hypothesis step:
At each local brain district N (v
0) activity pattern moving phase with promptly under the different stimulated condition: under can not remarkable isolating null hypothesis, use the nonparametric sequential test that the statistical distance parameter of each voxel correspondence is tested, obtain significance nonparametric sequential test value figure;
Resample by following magnetic resonance image data and to carry out the nonparametric sequential test: the data space pattern is constant, and magnetic resonance image data time corresponding sequence is carried out random rearrangement; Random rearrangement has destroyed the dependency of magnetic resonance image (MRI) signal and experimental condition, has kept complete space structure, thereby calculates the nonparametric sequential test value of the polynary distance statistics amount correspondence between different local cerebration patterns.
According to embodiments of the invention, testing of hypothesis multiple comparisons aligning step also comprises:
Step F .1: selecting the boundary value q (0＜q＜1) of a wrong discovery rate FDR, is the wrong discovery rate FDR of desired maximum;
Step F .2: the nonparametric sequential test p value figure that is obtained by the nonparametric permutation tests is carried out from small to large value being sorted: p (1)≤p (2)≤... ≤ p (m), if p (i) corresponding voxel statistic v (i), m is the total voxel number of being checked of fMRI data;
Step F .3: suppose that r satisfies inequality
Maximum i value, wherein c (m) is predefined constant, the distribution situation of its selection and voxel is relevant, and 2 kinds of selection: c (m)=1 are arranged under the different condition,
Step F .4: obtain the result: refusing empty assumed condition is that real activated voxel is v (1), v (2) ..., v (r) promptly is a statistic greater than the voxel of v (r) is activated voxel, claims v (r) to be the definite threshold value of multiple testing;
Step F .5: proofread and correct the multiple testing threshold value v (r) that obtains according to FDR, statistical parameter figure is carried out threshold value cut apart, obtain the brain active region that those can significantly distinguish the different tests condition.
The invention has the beneficial effects as follows:
No matter at large space scale level still on meticulous scale level, the fMRI data are all comprising abundant cerebration information.Yet, up to date,, utilize the research of the meticulous information in the fMRI data that is hidden in also fewer about how directly to extract.Tradition univariate analysis technology only relies on the time serial message of single voxel to extract the brain active region, the complete intact spatial model information that is included in the local brain district of having ignored.In addition, the common space smoothing pretreatment of adopting of univariate analysis technology also can filter out the high-frequency information on many meticulous scales that lie in the data unreasonably.This paper has proposed the algorithm that a kind of new active region is detected, represent cerebration state in the local brain district between the spatial model vector that this method at first uses a plurality of voxels to constitute, and then use polynary distance statistics amount to come integrated local brain district information, differentiate different stimulated condition hypencephalon activity variance.Multi-mode can comprehensively react local cerebration state, and multivariate statistics distance can effectively integratedly comprise information in the local brain district in order to measure the difference between different brain state of activation, and both have guaranteed that together the LMDM method can more completely detect meticulous brain enable mode than traditional univariate analysis technology.L-G simulation test and truthful data result of the test prove that all this algorithm can more completely detect the brain active region than traditional univariate analysis method, for the analysis of brain function data method provides a kind of new way.
Description of drawings
Fig. 1 is an algorithm computation schematic flow sheet of the present invention.
Fig. 2 a and Fig. 2 b extract locally consistent brain district sketch map among the present invention.
Fig. 3 is in order to measure the pattern classification conceptual schematic view of statistical distance between different enable modes among the present invention.
Fig. 4 is the contrast of algorithm of the present invention and traditional active region detection algorithm in the example.
Fig. 5 is the contrast of algorithm of the present invention and traditional active region detection algorithm in the example.
Fig. 6 is the contrast of algorithm of the present invention and traditional active region detection algorithm in the example.
Fig. 7 is the contrast of algorithm of the present invention and traditional active region detection algorithm in the example.
Fig. 8 is the contrast of algorithm of the present invention and traditional active region detection algorithm in the example.
The specific embodiment
The invention will be further described below in conjunction with accompanying drawing 1 and embodiment, is to be noted that described embodiment only is intended to be convenient to the understanding of the present invention, and it is not played any qualification effect.
The step that method of the present invention is related to describes in detail one by one below:
At first introduce, the form of specific embodiment is as described below:
Multi-mode distance between the meticulous activity pattern that comprises based on local brain district is extracted the brain active region, and the fMRI image is carried out pretreatment; Cluster obtains locally coherence brain district; Utilize the associating active tectonics multi-mode of a plurality of voxels in the locally consistent brain district; Utilize method for classifying modes to construct polynary distance function and measure the active separable character in local brain district under the different stimulated condition, judge whether the brain district activates.The multi-mode information that the present invention directly uses multi-voxel proton in the local brain district to constitute is represented the cerebration under the different stimulated condition, multi-mode can comprehensively reflect local cerebration state, and multivariate statistics distance can the effectively integrated information that comprises in the local brain district be measured the difference between different brain state of activation, and both have guaranteed that together the present invention can more completely detect meticulous brain enable mode than traditional fMRI analytical technology.
One, the described standardized images data of pre-treatment step comprise:
Steps A .1 carries out standardization to the different samples of view data, and making its average is 0, and standard deviation is 1;
Steps A .2 carries out standardization to the view data different characteristic, and making its average is 0, and standard deviation is 1.
Two, structure locally consistent brain district step also comprises:
B.1, step uses the Pearson came Pearson correlation coefficient of voxel activity time sequence as criterion, measures the activity similar quality between different voxels;
B.2, step is used voxel v
0As seed points, the each selection of region growing algorithm joins in the locally consistent brain district with the most similar field voxel conduct of seed points activity; The region growing iteration is carried out, and stops after reaching specified regional area size; Thereby the movable all basically identicals of the voxel in the regional area that obtains.
Three, structure multi-mode statistical distance step also comprises:
Step is D.1: voxel v
0Locally consistent brain district N (v
0) at different stimulated condition X, under the Y, the associating activity of a regional area K voxel that records, activity are from two different multiple random variable X=(X
1, X
2..., X
i..., X
K)
T, Y=(Y
1, Y
2..., Y
i..., Y
K)
T, i=1,2 ..., the pattern sample that the K sampling obtains;
Step is D.2: establish incentive condition X and Y corresponding sample set S respectively
XAnd S
YAccording to mode identification method, construct different polynary distance functions and measure sample set S
XAnd S
YBetween distance, this polynary distance is the Fisher distance function of deriving according to the Fisher linear discriminant analysis, the maximum frontier distance function of deriving according to support vector machine, and by the polynary distance function of other pattern classification algorithm to derivation;
Step is D.3: be that example describes here with FLDF, according to Fisher linear discriminant analysis, S set
XAnd S
YBetween FLDF calculate by following formula:
Wherein,
Be respectively sample S
XAnd S
YAverage vector promptly:
Σ is a sample mixing covariance:
Four, polynary distance statistics Parameter Map testing of hypothesis step:
At each local brain district N (v
0) activity pattern moving phase with promptly under the different stimulated condition: under can not remarkable isolating null hypothesis, use the nonparametric sequential test that the statistical distance parameter of each voxel correspondence is tested, obtain significance nonparametric sequential test value figure;
Resample by following magnetic resonance image data and to carry out the nonparametric sequential test: the data space pattern is constant, and magnetic resonance image data time corresponding sequence is carried out random rearrangement; Random rearrangement has destroyed the dependency of magnetic resonance image (MRI) signal and experimental condition, has kept complete space structure, thereby calculates the nonparametric sequential test value of the polynary distance statistics amount correspondence between different local cerebration patterns.
Five, testing of hypothesis multiple comparisons aligning step also comprises:
Step F .1 selects the boundary value q (0＜q＜1) of a wrong discovery rate FDR, is the wrong discovery rate FDR of desired maximum;
Step F .2, the nonparametric sequential test p value figure that is obtained by the nonparametric permutation tests is carried out from small to large value being sorted: p (1)≤p (2)≤... ≤ p (m), if p (i) corresponding voxel statistic v (i), m is the total voxel number of being checked of fMRI data;
Step F .3 supposes that r satisfies inequality
Maximum i value, wherein c (m) is predefined constant, the distribution situation of its selection and voxel is relevant, and 2 kinds of selection: c (m)=1 are arranged under the different condition,
Step F .4 obtains the result: refusing empty assumed condition is that real activated voxel is v (1), v (2) ..., v (r) promptly is a statistic greater than the voxel of v (r) is activated voxel, claims v (r) to be the definite threshold value of multiple testing;
Step F .5 proofreaies and correct the multiple testing threshold value v (r) that obtains according to FDR, statistical parameter figure is carried out threshold value cut apart, and obtains the brain active region that those can significantly distinguish the different tests condition.
Its two, introduce emulated data experiment A
A.1 emulated data design
Because the position, active region and the shape of true nuclear magnetic resonance fMRI data generally can't accurately be known, still at first design simulation data (preestablish the position, active region, and use it that different active regions detection method is carried out performance relatively the active situation of shape and voxel).Emulated data comprises two class incentive conditions, adopts to present the incident relevant design at a slow speed.Each stimulation presents 500 milliseconds, the change between 16～20 seconds at interval of different event zero-time.Every class incentive condition comprises 30 incidents, presents order at random.(canonicalHemodynamic Response Function, cHRF) brain that causes of all kinds of stimulations of emulation activates corresponding fMRI time series by stimulating sequence of events convolution standard hemodynamics mathematic(al) function.The concrete parameter of data is as follows: the number of plies (number of slice) is 5, and voxel size (voxel sizes) is 3 * 3 * 3 cubic millimeters, and the repetition time (TR) is 2000 milliseconds, and matrix (matrix) is 64 * 64.
(Fig. 6 a) comprises the voxel number and is respectively 10,30,90,180,270 by emulation for five shapes and big or small active region all inequality.In the active region, every class stimulates the steric effect that causes to obey the white Gaussian noise distribution.Outside the active region, stimulate not cause effect.The enable mode that causes with all kinds of stimulations of white Gaussian noise emulation can make the energy even of effect signal intersperse among different space frequency ranges, thereby being activated in the local average signal (low-frequency component) in brain district and the meticulous space structure (radio-frequency component), emulation all contains the incentive condition relevant information, more near true fMRI data.Then, stimulate the cerebration signal that causes to be added on the space-time noise background.In order to simulate the local space correlation properties of noise in the true fMRI data, noise background produces as follows: at first emulation obtains the space-time white Gaussian noise, use halfwidth (Full Width at HalfMaximum then, FWHM) be that 3.5 millimeters gaussian kernel carries out smoothly obtaining having the noise background of local space correlation properties to the white Gaussian noise that produces.(five groups of data with different contrast noise ratios are produced for Contrast toNoise Ratio, the CNR) performance under, and contrast noise ratio value respectively is 0.2,0.4,0.6,0.8,1.0 for check active region extraction algorithm contrasts noise ratio in difference.The contrast noise ratio is defined as the ratio of the maximum value of space average signal amplitude in each brain active region with the background noise standard deviation.
A.2 emulated data analysis
Adopt following three kinds of modes that emulated data is analyzed:
(1) initial data-GLM analyzes;
(2) gaussian kernel smoothed data-GLM analyzes;
(3) LMDM analyzes.
During gaussian kernel smoothed data-GLM was analyzed, halfwidth was that the gaussian kernel wave filter of 6 millimeters and 9 millimeters is used for smoothed data respectively.Why adopt these two gaussian kernel wave filter to be because they are generally adopted in fMRI data analysis and bibliographical information.LMDM analyzes the regional area size K that adopts and is respectively 10 and 30 voxels, this guaranteed LMDM analyze the local voxel number that uses be approximately equal to respectively gaussian kernel smoothed data-GLM analyze in two local voxel numbers that gaussian kernel can contain, make us can compare the LMDM analysis objectively and gaussian kernel smoothed data-GLM analyzes the integrated efficient of local spacing wave.
(Receiver Operating Characteristic, ROC) curve is used to quantitatively contrast the performance of different analytical methods to operating curve ROC.To the statistical parameter figure that the particular analysis method calculates, set different threshold values and cut apart, just obtain different activation graphs.The ROC curve has been portrayed when threshold value when minima is to maximum value from Parameter Map, and true activity ratio of the detected activation graph correspondence of analytical method (true activation rate) and false activation rate (false activation rate) are with the situation of threshold variation.True activity ratio be defined as correctly be identified as activate voxel number of voxels than last true activation number of voxels; Be defined as the false activation rate and be identified as the number of voxels that activates voxel by mistake than last true non-activation number of voxels.According to signal detection theory, the ROC curve has embodied the restricting relation between signal detecting method sensitivity (sensitivity) and specificity (specificity), can be used for the quantitative contrast of unlike signal detection method.Here, true activity ratio correspondence the sensitivity of active region extraction algorithm, the false activation rate then corresponding the specificity of active region extraction algorithm.ROC curve encirclement area (ROC area) can be used as an aggregative indicator and comes the gauge signal detection method to what extent can obtain hypersensitivity and specificity simultaneously.In other words, this index can be weighed the statistical parameter that calculated by ad hoc approach and can to what extent correctly activate voxel to emulation and make a distinction with background noise.
Fig. 2 extracts locally consistent brain district sketch map among the present invention, Fig. 2 a has marked the seed voxels of using when carrying out a region growing, and Fig. 2 b has shown that this sub-region increases the locally consistent zone that obtains.
Fig. 3, among the present invention in order to measure the pattern classification conceptual schematic view of statistical distance between different enable modes, (this sentences linear discriminant analysis is example).Among Fig. 3, empty circles is illustrated in the sample point that records under the incentive condition X, and solid circles is illustrated in the sample point that records under the incentive condition Y.For ease of on X-Y scheme, showing, suppose that here the local space pattern is by two voxels (X-axis and Y-axis).Linear discriminant analysis makes two class samples have best separability on this, and can measure the separation property of two class samples on this by the Fisher linear discriminant function by two class sample points being projected on the optimum differentiation axle.
The contrast of algorithm of the present invention and traditional active region detection algorithm in Fig. 4 example, curve on average obtains by 30 emulation experiments, and the ROC curve that has provided three kinds of analytical method correspondences surrounds the situation of area along with the change of contrast noise ratio.Can clearly observe, under the contrast noise ratio in office, data smoothing has all reduced the detection performance that GLM analyzes.Data smoothing serious more (smoothing windows is big more), performance reduce many more, this is because during the gaussian kernel smoothed data, and the high-frequency information on a lot of meticulous scales is by filtering.The performance that LMDM analyzes all is better than GLM in all cases and analyzes, because the spatial model by directly using a plurality of voxels to constitute, LMDM intactly describes and utilized the cerebration information that is included in the regional area.
Fig. 5, the contrast of algorithm of the present invention and traditional active region detection algorithm in the example.Shown that three kinds of analytical methods are the ROC curve under 0.6 the emulated data in an example contrast noise ratio.Can see that LMDM analyzes corresponding ROC curve and all be in upper left side under all scenario, be higher than all GLM and analyze corresponding ROC curve.Initial data-GLM analyzes the ROC curve that obtains and analyzes between the ROC curve that obtains in LMDM analysis and gaussian kernel smoothed data-GLM, shows that the performance of initial data-GLM analysis is poorer than the LMDM analysis, analyzes but be better than gaussian kernel smoothed data-GLM.Emulated data has the energy rank that is equal in the frequency range that all are lower than Nyquist frequency (by the voxel size decision), gaussian kernel smoothed data-GLM has been analyzed filtering high-frequency information, subdued signal energy, thereby can't utilize these high-frequency informations, reduced the detection performance, so performance is the poorest in the three; Initial data-GLM analyzes and does not carry out filtering, has kept this part high-frequency information, but owing to only utilized the information of single voxel, does not utilize the information in the local brain district, analyzes so performance will be lower than LMDM; And the LMDM analysis has effectively utilized the information that is distributed in all frequency ranges in the local space pattern to differentiate different cerebration states just, so its best performance.
In order to understand Fig. 4 intuitively, the quantitative result that provides among Fig. 5, the contrast of algorithm of the present invention and traditional active region detection algorithm in Fig. 6 (a) and (b), (c), (d) example.Three kinds of methods have been provided at the Parameter Map of cutting apart without threshold value (all taking from the emulated data intermediate layer) that on an example contrast noise is 0.6 used emulated data, obtains among Fig. 5.Fig. 6 (a) emulation active region profile diagram.Fig. 6 (b) initial data, the GLM analysis result.Fig. 6 (c) data are after 9 millimeters the gaussian kernel filtering through halfwidth, the GLM analysis result.Fig. 6 (d) regional area size is made as 30 voxels, LMDM analysis result.Among the figure, true border, brain active region is marked by contour line.Can see that initial data-GLM analyzes the absolute t value figure that obtains and do not detect the more weak voxel of some activities in the emulation active region well, the whole signal to noise ratio of Parameter Map is also lower simultaneously.After to adopt halfwidth be 9 millimeters gaussian kernel smoothed data, GLM analyzed the Parameter Map signal to noise ratio of extracting and increases, but has also occurred some significantly pseudo-at random voxels that activate outside the real simulation active region.Analyze with respect to two kinds of GLM, the signal to noise ratio that LMDM analyzes the polynary distance statistics Parameter Map that obtains is higher, and has accurately detected true brain active region, also simultaneously the pseudo-voxel that activates not occur outside true active region.These undivided Parameter Map have illustrated that qualitatively LMDM analyzes by using the complete information in the next integrated local brain district of multivariate statistics distance, can detect the local brain district that can distinguish different cerebration states more accurately.
Its three, be to introduce truthful data experiment B
B.1 true fMRI data are obtained
6 tested has participated in experiment, and the men and women half and half.Chunkization (block) design with two class incentive conditions is adopted in experiment, and a class incentive condition is people's face picture, and another kind of incentive condition is the house picture.In the incentive condition chunk, present stimulation (people's face or the house) picture that the center indicates " ten " number; In baseline chunk (baseline), rendering content only is the picture of " ten " number.Require tested when watching picture, point of fixation being placed on " ten " number.Each stimulates chunk to continue 30 seconds, during present 20 similar pictures altogether, each picture presents 500 milliseconds, is stimulating picture to present " ten " number picture at interval.Baseline chunk length is about 10 seconds, allows baseline chunk variable-length, more fully gathering the hemodynamics response of human brain to stimulating, thereby helps the detection of brain active region.Adopt T2* weighting gtadient echo planar imaging (Echo-Planar Imaging, EPI) sequence is obtained the BOLDfMRI data, detail parameters is as follows: bed thickness (slice thickness) is 4 millimeters, form (FOV) is 240 * 240 square millimeters, repetition time (TR) is 2000 milliseconds, and matrix (matrix) is 64 * 64.
Adopt SPM2 (http://www.fil.ion.ucl.ac.uk/spm/) that data are carried out pretreatment, comprising: time unifying, reset in the space, goes baseline drift, removes non-brain voxel.When LMDM analyzed in addition, the fMRI time series was moved forward 4 seconds to offset the hemodynamics response delay; The data of chunk conversion place correspondence are dropped, and only keep the data that record when hematodinamics is stablized.
B.2 true fMRI data analysis
Adopt the mode identical with analyzing emulated data that truthful data is analyzed: (1) initial data-GLM analyzes, and (2) gaussian kernel smoothed data-GLM analyzes, and (3) LMDM analyzes.Can see that by emulation experiment when the regional area size was made as 30 voxels, LMDM analyzing and testing performance was best, it is the local voxel number that 9 millimeters gaussian kernel contains that its local voxel number of using is approximately equal to halfwidth.So only use these two parameters that data are analyzed below.The result as shown in Figure 7, three kinds of analytical methods can detect all that corresponding people's face picture stimulates and the main brain district (FFA and PPA) of house picture stimulation difference.Among the figure, all significance p value figure resample by 1000 independent arrangements to obtain.Different is that the significance p value figure that gaussian kernel smoothed data-GLM analyzes and LMDM analyzes has all used FDR to proofread and correct, sets mean F DR and is no more than q=0.05; And the significance p value figure that initial data-GLM analysis provides does not carry out the FDR correction, if use and other both same corrections, will can reach remarkable activation without any voxel.GLM analyzes the activation graph that obtains and represents that this people from position face picture stimulates the voxel activity that causes significantly to be better than the house picture and stimulates the voxel activity that causes, or the house picture stimulates this position voxel activity that causes significantly to be better than the activity that the stimulation of people's face picture causes.LMDM analyzes the activation graph obtain and illustrates that voxel activity in corresponding voxel and the regional area around it unites the spatial model of formation and have good separability under two class pictures stimulate, and the pattern that presents of local brain district's activity of causing of two class incentive conditions is remarkable different in other words.Shown in the contrast of algorithm of the present invention and traditional active region detection algorithm in Fig. 7 a, Fig. 7 b, Fig. 7 c example, initial data-GLM analyzes can roughly obtain position, brain active region, but exists serious salt-pepper noise.The contrast noise ratio of single voxel signal is lower in the fMRI data, Fig. 7 a initial data-GLM analysis result only utilizes single voxel action message to decide the state of voxel, and the information of not utilizing local brain district to contain, thereby the Parameter Map signal to noise ratio that obtains is lower.Employing half-breadth height is after 9 millimeters gaussian kernel is carried out filtering to data, it is clean and concentrated that the activation graph that uses the GLM analysis result to obtain is analyzed the activation graph that obtains than initial data-GLM, is detected (Fig. 7 b) but have only those that stronger core of two class irritant reaction differences is activated voxel.Yet use regional area size is that the LMDM analysis result of 30 voxels arrives the more voxel that activates than gaussian kernel smoothed data-GLM analyzing and testing, and these distribution are analyzed around the active region, center that obtains at gaussian kernel smoothed data-GLM.This shows the core active region of arriving except gaussian kernel smoothed data-GLM analyzing and testing, around it, still exist and much comprise meticulous information, to the fainter voxel (Fig. 7 c) of experiment irritant reaction, when these voxels are carried out conjoint analysis, can distinguish the different experiments incentive condition well equally.The GLM analysis can't detect these edges, the voxel that contains faint information, mainly come from two reasons: the one, it is univariate analysis in essence that GLM analyzes, it is when carrying out the active region detection, discretely each voxel is handled, can't be fully utilized the spatial model information that is included in the regional area; The 2nd, render a service although gaussian kernel filtering can be strengthened on the one hand the statistics that GLM analyzes, improve the signal to noise ratio of voxel signal in the core active region, simultaneously unreasonably filtering be included in the interior most of meticulous cerebration information of regional area.By analyzing all tested data, we further find, the LMDM analyzing and testing to active region with respect to gaussian kernel smoothed data-GLM analyzing and testing to active region more extend, present distributed form, this is consistent with former research results.
L represents the axially left side of position of the corresponding human brain of this side among Fig. 7, and R represents the axially right side of position of the corresponding human brain of this side.Face represents people's face picture incentive condition, and house represents house picture incentive condition.Among the figure people's face picture and house picture being had the voxel that activates difference all checks through permutation.B and c use FDR to proofread and correct, and the mean F DR that guarantees all activated voxel is less than 0.05.
The Parameter Map that provides for quantitative different analytical methods of contrast is tested repeatability (similarity or concordance), tested Parameter Map at first is registered on the standard MNI template, and the Pearson's correlation coefficient (r) that calculates every pair of tested Parameter Map then is in order to measure the similarity between them.The contrast of algorithm of the present invention and traditional active region detection algorithm in Fig. 8 example, provided the concordance of three kinds of analytical method corresponding parameters figure among the figure tested of difference, GLM on the transverse axis, sm-GLM, on behalf of initial data-GLM, LMDM analyze respectively, gaussian kernel smoothed data-GLM analyzes, and LMDM analyzes; Longitudinal axis scale is represented Pearson's correlation coefficient.With the mean value standard units of the error of the mean (r ± form SEM) to every kind of analytical method following 15 to (C
6 2) Pearson's correlation coefficient of Parameter Map adds up.In three kinds of analysis results, it is the poorest at tested similarity (r=0.25 ± 0.04) that initial data-GLM analyzes the statistical parameter figure that obtains, and this causes by the Parameter Map signal to noise ratio is too low.The Parameter Map that LMDM analyzes is more less better than the Parameter Map (r=0.57 ± 0.02) that gaussian kernel smoothed data-GLM analyzes at tested similarity (r=0.49 ± 0.03).This is that different tested diencephalon enable modes itself differ greatly because on the scale of meticulous space on the one hand; Also might be that space smoothing is not adopted in the LMDM analysis in addition, responsive to the error of introducing in the space criteria process, cause the concordance of tested Parameter Map to descend.Therefore the present invention is specially adapted to single tested fMRI brain active region, and with respect to traditional fMRI active region extracting method, this method can be extracted single tested meticulousr brain active information.
Describing above is to be used to realize embodiments of the invention, it should be appreciated by those skilled in the art, in any modification or partial replacement that does not depart from the scope of the present invention, all belongs to claim of the present invention and comes restricted portion.
Claims (6)
1, a kind of method based on pattern recognition classifier extraction of magnetic resonance imaging brain active region is characterized in that: comprise step:
Pre-treatment step A: the magnetic resonance image (MRI) that collects is carried out pretreatment, remove and obscure factor, be used for acquisition standardized images data;
Structure locally consistent brain district step B: to each voxel v of view data
0Use region growing algorithm with v
0As seed points, obtain comprising the locally consistent brain district N (v of K voxel
0);
Sliced time sequence step C: cut apart locally consistent brain district N (v according to the incentive condition classification of experimental design correspondence
0) in the time series of a plurality of voxels, then every class incentive condition correspondence a different set of multivariate data sample;
Structure multi-mode statistical distance step D: establish under the different stimulated condition locally consistent brain district N (v
0) in the spatial model correspondence that constitutes of K voxel different K and tie up random distribution; Use mode identification method structure multivariate statistics distance function, in order to tolerance locally consistent brain district N (v
0) statistical distance between the data sample of interior different condition correspondence; By to the above process of each voxel iteration, calculate the multivariate statistics distance, obtain multivariate statistics distance parameter figure;
Polynary distance statistics Parameter Map testing of hypothesis step e: suppose at each locally consistent brain district N (v
0) activity pattern under the different stimulated condition moving phase with, use the nonparametric sequential test that the statistical distance parameter of each voxel correspondence is tested, obtain significance nonparametric sequential test value figure;
Testing of hypothesis multiple comparisons aligning step F: utilize wrong discovery rate FDR method, be used to eliminate multiple comparisons, obtain significantly to distinguish the local brain district of different tests condition, promptly corresponding active region.
2, the method for extraction of magnetic resonance imaging brain active region according to claim 1 is characterized in that the described standardized images data of pre-treatment step also comprise:
Steps A .1 carries out standardization to the different samples of view data, and making its average is 0, and standard deviation is 1;
Steps A .2 carries out standardization to the view data different characteristic, and making its average is 0, and standard deviation is 1.
3, the method for extraction of magnetic resonance imaging brain active region according to claim 1 is characterized in that described structure locally consistent brain district step also comprises:
B.1, step uses the Pearson came Pearson correlation coefficient of voxel activity time sequence as criterion, measures the activity similar quality between different voxels;
B.2, step is used voxel v
0As seed points, the each selection of region growing algorithm joins in the locally consistent brain district with the most similar field voxel conduct of seed points activity; The region growing iteration is carried out, and stops after reaching specified regional area size; Thereby the movable all basically identicals of the voxel in the regional area that obtains.
4, the method for extraction of magnetic resonance imaging brain active region according to claim 1 is characterized in that described structure multi-mode statistical distance step also comprises:
Step is D.1: voxel v
0Locally consistent brain district N (v
0) at different stimulated condition X, under the Y, the associating activity of a regional area K voxel that records, activity are from two different multiple random variable X=(X
1, X
2..., X
i..., X
K)
T, Y=(Y
1, Y
2..., Y
i..., Y
K)
T, i=1,2 ..., the pattern sample that the K sampling obtains;
Step is D.2: establish incentive condition X and Y corresponding sample set S respectively
XAnd S
YAccording to mode identification method, construct different polynary distance functions and measure sample set S
XAnd S
YBetween distance, this polynary distance is the Fisher distance function of deriving according to the Fisher linear discriminant analysis, the maximum frontier distance function of deriving according to support vector machine, and by the polynary distance function of other pattern classification algorithm to derivation;
Step is D.3: be that example describes here with FLDF, according to Fisher linear discriminant analysis, S set
XAnd S
YBetween FLDF calculate by following formula:
Wherein,
Be respectively sample S
XAnd S
YAverage vector promptly:
∑ is a sample mixing covariance:
5, the method for extraction of magnetic resonance imaging brain active region according to claim 1 is characterized in that polynary distance statistics Parameter Map testing of hypothesis step:
At each local brain district N (v
0) activity pattern moving phase with promptly under the different stimulated condition: under can not remarkable isolating null hypothesis, use the nonparametric sequential test that the statistical distance parameter of each voxel correspondence is tested, obtain significance nonparametric sequential test value figure;
Resample by following magnetic resonance image data and to carry out the nonparametric sequential test: the data space pattern is constant, and magnetic resonance image data time corresponding sequence is carried out random rearrangement; Random rearrangement has destroyed the dependency of magnetic resonance image (MRI) signal and experimental condition, has kept complete space structure, thereby calculates the nonparametric sequential test value of the polynary distance statistics amount correspondence between different local cerebration patterns.
6, the method for extraction of magnetic resonance imaging brain active region according to claim 1 is characterized in that testing of hypothesis multiple comparisons aligning step also comprises:
Step F .1 selects the boundary value q (0＜q＜1) of a wrong discovery rate FDR, is the wrong discovery rate FDR of desired maximum;
Step F .2, the nonparametric sequential test p value figure that is obtained by the nonparametric permutation tests is carried out from small to large value being sorted: p (1)≤p (2)≤... ≤ p (m), if p (i) corresponding voxel statistic v (i), m is the total voxel number of being checked of fMRI data;
Step F .3 supposes that r satisfies inequality
Maximum i value, wherein c (m) is predefined constant, the distribution situation of its selection and voxel is relevant, and 2 kinds of selection: c (m)=1 are arranged under the different condition,
Step F .4 obtains the result: refusing empty assumed condition is that real activated voxel is v (1), v (2) ..., v (r) promptly is a statistic greater than the voxel of v (r) is activated voxel, claims v (r) to be the definite threshold value of multiple testing;
Step F .5 proofreaies and correct the multiple testing threshold value v (r) that obtains according to FDR, statistical parameter figure is carried out threshold value cut apart, and obtains the brain active region that those can significantly distinguish the different tests condition.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200710098691A CN101292871B (en) | 2007-04-25 | 2007-04-25 | Method for specification extraction of magnetic resonance imaging brain active region based on pattern recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN200710098691A CN101292871B (en) | 2007-04-25 | 2007-04-25 | Method for specification extraction of magnetic resonance imaging brain active region based on pattern recognition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN101292871A true CN101292871A (en) | 2008-10-29 |
| CN101292871B CN101292871B (en) | 2010-05-26 |
Family
ID=40063606
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200710098691A Expired - Fee Related CN101292871B (en) | 2007-04-25 | 2007-04-25 | Method for specification extraction of magnetic resonance imaging brain active region based on pattern recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN101292871B (en) |
Cited By (18)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101847261A (en) * | 2010-05-10 | 2010-09-29 | 南京航空航天大学 | Obtaining method of rat head magnetic resonance image Monte Carlo simulation model |
| CN102508184A (en) * | 2011-10-26 | 2012-06-20 | 中国科学院自动化研究所 | Brain function active region detection method based on moving average time series models |
| CN101706561B (en) * | 2009-11-20 | 2012-07-04 | 电子科技大学 | Clustering method for functional magnetic resonance images |
| CN101666865B (en) * | 2009-09-09 | 2012-12-19 | 西北工业大学 | Method for registering diffusion tensor nuclear magnetic resonance image in local quick traveling mode |
| CN103020653A (en) * | 2012-12-07 | 2013-04-03 | 中国科学院自动化研究所 | Structure and function magnetic resonance image united classification method based on network analysis |
| WO2013097118A1 (en) * | 2011-12-28 | 2013-07-04 | 中国科学院自动化研究所 | Method for processing brain function magnetic resonance data |
| CN101744612B (en) * | 2008-11-28 | 2014-07-23 | Ge医疗系统环球技术有限公司 | Blood flow dynamic analysis apparatus and magnetic resonance imaging system |
| CN105513058A (en) * | 2015-12-01 | 2016-04-20 | 中国科学院苏州生物医学工程技术研究所 | Brain active region detection method and device |
| CN105653861A (en) * | 2015-12-31 | 2016-06-08 | 首都医科大学宣武医院 | Functional magnetic resonance imaging (fMRI) based category concepts similarity calculation method |
| CN106548206A (en) * | 2016-10-27 | 2017-03-29 | 太原理工大学 | Multi-modal nuclear magnetic resonance image data classification method based on minimum spanning tree |
| CN108197661A (en) * | 2018-01-19 | 2018-06-22 | 上海海事大学 | Cognitive activities state classification system and method based on voxel level brain function information |
| CN108596228A (en) * | 2018-04-13 | 2018-09-28 | 江南大学 | A kind of brain functional magnetic resonance image classification based on unsupervised fuzzy system |
| CN109830286A (en) * | 2019-02-13 | 2019-05-31 | 四川大学 | Brain function magnetic resonance code energy imaging method based on nonparametric statistics |
| CN110087546A (en) * | 2016-12-21 | 2019-08-02 | 弗门尼舍有限公司 | For determining the fMRI method of the brain activation pattern of the feeling excited in response to smell |
| CN111329446A (en) * | 2020-02-26 | 2020-06-26 | 四川大学华西医院 | Visual stimulation system and method for processing spatial frequency of facial pores through brain visual pathway |
| CN112529918A (en) * | 2020-12-23 | 2021-03-19 | 沈阳东软智能医疗科技研究院有限公司 | Method, device and equipment for ventricular region segmentation in brain CT image |
| CN112686902A (en) * | 2019-10-17 | 2021-04-20 | 西安邮电大学 | Two-stage calculation method for brain glioma identification and segmentation in nuclear magnetic resonance image |
| CN116583221A (en) * | 2021-09-17 | 2023-08-11 | 皇家飞利浦有限公司 | Determination of subject-specific hemodynamic response functions |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11826133B2 (en) * | 2017-02-28 | 2023-11-28 | Board Of Trustees Of Michigan State University | Method and system for determining brain-state dependent functional areas of unitary pooled activity and associated dynamic networks with functional magnetic resonance imaging |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6292683B1 (en) * | 1999-05-18 | 2001-09-18 | General Electric Company | Method and apparatus for tracking motion in MR images |
| EP1219243A1 (en) * | 2000-12-28 | 2002-07-03 | Matsushita Electric Works, Ltd. | Non-invasive brain function examination |
| DE10162927A1 (en) * | 2001-12-20 | 2003-07-17 | Siemens Ag | Evaluation of images of the brain obtained using functional magnetic resonance tomography |
| CN1294875C (en) * | 2003-12-12 | 2007-01-17 | 中国科学院自动化研究所 | Time series analysis method of nuclear magnetic resonance for brain functions based on constrained optimization |
| CN1626031A (en) * | 2003-12-12 | 2005-06-15 | 中国科学院自动化研究所 | Method for detecting functional connection between brain regions based on graph theory |
| CN1939211A (en) * | 2005-09-30 | 2007-04-04 | 中国科学院自动化研究所 | Analysis of sea-horse quantitative shape in magnetic resonance image |
-
2007
- 2007-04-25 CN CN200710098691A patent/CN101292871B/en not_active Expired - Fee Related
Cited By (31)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101744612B (en) * | 2008-11-28 | 2014-07-23 | Ge医疗系统环球技术有限公司 | Blood flow dynamic analysis apparatus and magnetic resonance imaging system |
| CN101666865B (en) * | 2009-09-09 | 2012-12-19 | 西北工业大学 | Method for registering diffusion tensor nuclear magnetic resonance image in local quick traveling mode |
| CN101706561B (en) * | 2009-11-20 | 2012-07-04 | 电子科技大学 | Clustering method for functional magnetic resonance images |
| CN101847261B (en) * | 2010-05-10 | 2012-02-15 | 南京航空航天大学 | Obtaining method of rat head magnetic resonance image Monte Carlo simulation model |
| CN101847261A (en) * | 2010-05-10 | 2010-09-29 | 南京航空航天大学 | Obtaining method of rat head magnetic resonance image Monte Carlo simulation model |
| CN102508184A (en) * | 2011-10-26 | 2012-06-20 | 中国科学院自动化研究所 | Brain function active region detection method based on moving average time series models |
| CN102508184B (en) * | 2011-10-26 | 2015-04-08 | 中国科学院自动化研究所 | Brain function active region detection method based on moving average time series models |
| WO2013097118A1 (en) * | 2011-12-28 | 2013-07-04 | 中国科学院自动化研究所 | Method for processing brain function magnetic resonance data |
| US9633430B2 (en) | 2011-12-28 | 2017-04-25 | Institute Of Automation, Chinese Academy Of Sciences | Method for analyzing functional MRI brain images |
| CN103020653B (en) * | 2012-12-07 | 2015-07-22 | 中国科学院自动化研究所 | Structure and function magnetic resonance image united classification method based on network analysis |
| CN103020653A (en) * | 2012-12-07 | 2013-04-03 | 中国科学院自动化研究所 | Structure and function magnetic resonance image united classification method based on network analysis |
| CN105513058A (en) * | 2015-12-01 | 2016-04-20 | 中国科学院苏州生物医学工程技术研究所 | Brain active region detection method and device |
| CN105513058B (en) * | 2015-12-01 | 2019-04-23 | 中国科学院苏州生物医学工程技术研究所 | A kind of brain active region detection method and device |
| CN105653861B (en) * | 2015-12-31 | 2018-03-23 | 首都医科大学宣武医院 | Similarity calculation method between a kind of class concepts based on fMRI |
| CN105653861A (en) * | 2015-12-31 | 2016-06-08 | 首都医科大学宣武医院 | Functional magnetic resonance imaging (fMRI) based category concepts similarity calculation method |
| CN106548206A (en) * | 2016-10-27 | 2017-03-29 | 太原理工大学 | Multi-modal nuclear magnetic resonance image data classification method based on minimum spanning tree |
| CN106548206B (en) * | 2016-10-27 | 2019-08-02 | 太原理工大学 | Multi-modal nuclear magnetic resonance image data classification method based on minimum spanning tree |
| CN110087546A (en) * | 2016-12-21 | 2019-08-02 | 弗门尼舍有限公司 | For determining the fMRI method of the brain activation pattern of the feeling excited in response to smell |
| CN108197661B (en) * | 2018-01-19 | 2022-03-01 | 上海海事大学 | Cognitive activity state classification system and method based on voxel horizontal brain function information |
| CN108197661A (en) * | 2018-01-19 | 2018-06-22 | 上海海事大学 | Cognitive activities state classification system and method based on voxel level brain function information |
| CN108596228A (en) * | 2018-04-13 | 2018-09-28 | 江南大学 | A kind of brain functional magnetic resonance image classification based on unsupervised fuzzy system |
| CN108596228B (en) * | 2018-04-13 | 2022-08-23 | 江南大学 | Brain function magnetic resonance image classification method based on unsupervised fuzzy system |
| CN109830286A (en) * | 2019-02-13 | 2019-05-31 | 四川大学 | Brain function magnetic resonance code energy imaging method based on nonparametric statistics |
| CN109830286B (en) * | 2019-02-13 | 2022-09-30 | 四川大学 | Brain function magnetic resonance encoding energy imaging method based on nonparametric statistics |
| CN112686902A (en) * | 2019-10-17 | 2021-04-20 | 西安邮电大学 | Two-stage calculation method for brain glioma identification and segmentation in nuclear magnetic resonance image |
| CN112686902B (en) * | 2019-10-17 | 2023-02-03 | 西安邮电大学 | Two-stage calculation method for brain glioma identification and segmentation in nuclear magnetic resonance image |
| CN111329446B (en) * | 2020-02-26 | 2022-06-10 | 四川大学华西医院 | Visual stimulation system and method for processing spatial frequency of facial pores through brain visual pathway |
| CN111329446A (en) * | 2020-02-26 | 2020-06-26 | 四川大学华西医院 | Visual stimulation system and method for processing spatial frequency of facial pores through brain visual pathway |
| CN112529918A (en) * | 2020-12-23 | 2021-03-19 | 沈阳东软智能医疗科技研究院有限公司 | Method, device and equipment for ventricular region segmentation in brain CT image |
| CN112529918B (en) * | 2020-12-23 | 2024-02-27 | 沈阳东软智能医疗科技研究院有限公司 | Method, device and equipment for segmenting brain room area in brain CT image |
| CN116583221A (en) * | 2021-09-17 | 2023-08-11 | 皇家飞利浦有限公司 | Determination of subject-specific hemodynamic response functions |
Also Published As
| Publication number | Publication date |
|---|---|
| CN101292871B (en) | 2010-05-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN101292871B (en) | Method for specification extraction of magnetic resonance imaging brain active region based on pattern recognition | |
| Liu et al. | A correlation-matrix-based hierarchical clustering method for functional connectivity analysis | |
| Sui et al. | An ICA-based method for the identification of optimal FMRI features and components using combined group-discriminative techniques | |
| Mensen et al. | Advanced EEG analysis using threshold-free cluster-enhancement and non-parametric statistics | |
| CN104887224B (en) | Feature extraction and automatic identifying method towards epileptic EEG Signal | |
| CN101287410B (en) | Brain function analysis method and brain function analysis program | |
| Clare et al. | Detecting activations in event‐related fMRI using analysis of variance | |
| CN104715261A (en) | FMRI dynamic brain function sub-network construction and parallel connection SVM weighted recognition method | |
| CN110575164B (en) | Method for removing artifacts of electroencephalogram signal and computer-readable storage medium | |
| CN101706561B (en) | Clustering method for functional magnetic resonance images | |
| CN109640814A (en) | Check the method and magnetic imaging equipment of human brain cortex hormone function | |
| Ylipaavalniemi et al. | Analyzing consistency of independent components: An fMRI illustration | |
| CN105842642A (en) | Fractional anisotropy microstructure characteristic extraction method based on kurtosis tensor and apparatus thereof | |
| Liu et al. | Towards tunable consensus clustering for studying functional brain connectivity during affective processing | |
| CN115359305B (en) | Accurate positioning system for abnormal area of cerebral fiber bundle | |
| DE102019206827A1 (en) | Magnetic resonance fingerprinting method | |
| CN102508184A (en) | Brain function active region detection method based on moving average time series models | |
| Zou et al. | Brain anomaly networks uncover heterogeneous functional reorganization patterns after stroke | |
| CN106264460A (en) | The coding/decoding method of cerebration multidimensional time-series signal based on self study and device | |
| CN113796873B (en) | Wearable dynamic electrocardiosignal classification method and system | |
| CN105931281A (en) | Method for quantitatively describing cerebral function network based on network characteristic entropy | |
| CN105395198A (en) | Method for acquiring brand new diffusion magnetic resonance imaging contrast ratio and application of method | |
| Gao et al. | Flexing the principal gradient of the cerebral cortex to suit changing semantic task demands | |
| CN103006215A (en) | Brain functional region positioning method based on local smoothing regressions | |
| Khalid et al. | Adaptive 2DCCA based approach for improving spatial specificity of activation detection in functional MRI |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20100526 Termination date: 20210425 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |