Group analyses of fMRI data
description
Transcript of Group analyses of fMRI data
![Page 1: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/1.jpg)
Group analyses of fMRI data
Methods & models for fMRI data analysisNovember 2012
With many thanks for slides & images to:FIL Methods group, particularly Will Penny & Tom Nichols
Klaas Enno Stephan
Translational Neuromodeling Unit (TNU)Institute for Biomedical Engineering, University of Zurich & ETH Zurich
Laboratory for Social & Neural Systems Research (SNS), University of Zurich
Wellcome Trust Centre for Neuroimaging, University College London
![Page 2: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/2.jpg)
Overview of SPM
Realignment Smoothing
Normalisation
General linear model
Statistical parametric map (SPM)Image time-series
Parameter estimates
Design matrix
Template
Kernel
Gaussian field theory
p <0.05
Statisticalinference
![Page 3: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/3.jpg)
Time
BOLD signalTim
esingle voxel
time series
Reminder: voxel-wise time series analysis!
modelspecificati
onparameterestimationhypothesis
statistic
SPM
![Page 4: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/4.jpg)
The model: voxel-wise GLM
=
e+y
X
N
1
N N
1 1p
pModel is specified by1. Design matrix X2. Assumptions about
eN: number of scansp: number of regressors
eXy
The design matrix embodies all available knowledge about experimentally controlled factors and potential confounds.
),0(~ 2INe
![Page 5: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/5.jpg)
GLM assumes Gaussian “spherical” (i.i.d.) errors
sphericity = iid:error covariance is scalar multiple of identity matrix:Cov(e) = 2I
1001
)(eCov
1004
)(eCov
2112
)(eCov
Examples for non-sphericity:
non-identity
non-independence
![Page 6: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/6.jpg)
Multiple covariance components at 1st level
),0(~ 2VNe
iiQV
eCovV
)(
= 1 + 2
Q1 Q2
Estimation of hyperparameters with ReML (restricted maximum likelihood).
V
enhanced noise model error covariance components Qand hyperparameters
![Page 7: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/7.jpg)
WeWXWy
c = 1 0 0 0 0 0 0 0 0 0 0
)ˆ(ˆˆ
T
T
cdtsct
cWXWXc
cdtsTT
T
)()(ˆ
)ˆ(ˆ
2
)(
ˆˆ
2
2
RtrWXWy
ReML-estimates
WyWX )(̂
)(2
2/1
eCovV
VW
)(WXWXIRX
t-statistic based on ML estimates
iiQ
V
TT XWXXWX 1)()( For brevity:
![Page 8: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/8.jpg)
Distribution of population effect
8
Subj. 1
Subj. 2
Subj. 3
Subj. 4
Subj. 5
Subj. 6
0
Fixed vs.random effectsanalysis
• Fixed Effects– Intra-subject variation
suggests most subjects different from zero
• Random Effects– Inter-subject variation
suggests population is not very different from zero
Distribution of each subject’s estimated effect 2
FFX
2RFX
![Page 9: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/9.jpg)
Fixed Effects
• Assumption: variation (over subjects) is only due to measurement error
• parameters are fixed properties of the population (i.e., they are the same in each subject)
![Page 10: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/10.jpg)
• two sources of variation (over subjects)– measurement error– Response magnitude: parameters are probabilistically
distributed in the population• effect (response magnitude) in each subject is randomly
distributed
Random/Mixed Effects
![Page 11: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/11.jpg)
Random/Mixed Effects
• two sources of variation (over subjects)– measurement error– Response magnitude: parameters are probabilistically
distributed in the population• effect (response magnitude) in each subject is randomly
distributed• variation around population mean
![Page 12: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/12.jpg)
Group level inference: fixed effects (FFX)
• assumes that parameters are “fixed properties of the population”
• all variability is only intra-subject variability, e.g. due to measurement errors
• Laird & Ware (1982): the probability distribution of the data has the same form for each individual and the same parameters
• In SPM: simply concatenate the data and the design matrices lots of power (proportional to number of scans),
but results are only valid for the group studied and cannot be generalized to the population
![Page 13: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/13.jpg)
Group level inference: random effects (RFX)
• assumes that model parameters are probabilistically distributed in the population
• variance is due to inter-subject variability • Laird & Ware (1982): the probability distribution of
the data has the same form for each individual, but the parameters vary across individuals
• hierarchical model much less power (proportional to number of subjects), but results can be generalized to the population
![Page 14: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/14.jpg)
FFX vs. RFX
• FFX is not "wrong", it makes different assumptions and addresses a different question than RFX
• For some questions, FFX may be appropriate (e.g., low-level physiological processes).
• For other questions, RFX is much more plausible (e.g., cognitive tasks, disease processes in heterogeneous populations).
![Page 15: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/15.jpg)
Hierachical models
fMRI, single subject
fMRI, multi-subject ERP/ERF, multi-subject
EEG/MEG, single subject
Hierarchical models for all imaging data!
time
![Page 16: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/16.jpg)
Linear hierarchical model
)()()()1(
)2()2()2()1(
)1()1()1(
nnnn X
X
Xy
)()()( i
k
i
kk
i QC
Hierarchical model Multiple variance components at each level
At each level, distribution of parameters is given by level above.
What we don’t know: distribution of parameters and variance parameters (hyperparameters).
![Page 17: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/17.jpg)
Example: Two-level model
=
2221
111
X
Xy
1
1+ 1 = 2X
2
+ 2y
)1(1X
)1(2X
)1(3X
Second levelFirst level
![Page 18: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/18.jpg)
Two-level model
(1) (1) (1)
(1) (2) (2) (2)
y X
X
(1) (2) (2) (2) (1)
(1) (2) (2) (1) (2) (1)
y X X
X X X
Friston et al. 2002, NeuroImage
fixed effects random effects
![Page 19: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/19.jpg)
Mixed effects analysis
(1) (2) (2) (1) (2) (1)y X X X
(1) (1)
(2) (2) (2) (1) (1)
(2) (2) (2)
ˆ X y
X X
X
(2) (2) (1) (1) (1) T
Cov C X C X
Non-hierarchical model
Variance components at 2nd level
Estimating 2nd level effects
between-level non-sphericity
( ) ( )( ) i iik k
kQC
Within-level non-sphericity at both levels: multiple
covariance componentsFriston et al. 2005, NeuroImage
within-level non-sphericity
![Page 20: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/20.jpg)
Algorithmic equivalence
)()()()1(
)2()2()2()1(
)1()1()1(
nnnn X
X
Xy
Hierarchicalmodel
ParametricEmpirical
Bayes (PEB)
EM = PEB = ReML
RestrictedMaximumLikelihood
(ReML)
Single-levelmodel
)()()1(
)()1()1(
)2()1()1(
...
nn
nn
XXXX
Xy
![Page 21: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/21.jpg)
Estimation by Expectation Maximisation (EM)
EM-algorithm
gJd
LdJ
ddLg
1
2
2
E-step
M-step
kk
kQC
maximise L ln ( )p y | λ
111
NppNN
Xy
Friston et al. 2002, NeuroImage
Gauss-Newtongradient ascent
• E-step: finds the (conditional) expectation of the parameters, holding the hyperparameters fixed
• M-step: updates the maximum likelihood estimate of the hyperparameters, keeping the parameters fixed
1 1 1
|
1| |
Ty
Ty y
C X C X C
η C X C y C η
![Page 22: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/22.jpg)
Practical problems
• Full MFX inference using REML or EM for a whole-brain 2-level model has enormous computational costs
• for many subjects and scans, covariance matrices become extremely large
• nonlinear optimisation problem for each voxel
• Moreover, sometimes we are only interested in one specific effect and do not want to model all the data.
• Is there a fast approximation?
![Page 23: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/23.jpg)
Summary statistics approach: Holmes & Friston 1998
Data Design Matrix Contrast Images )ˆ(ˆ
ˆ
T
T
craV
ct
SPM(t)1̂
2̂
11̂
12̂
21̂
22̂
211̂
212̂
Second levelFirst level
One-samplet-test @ 2nd level
![Page 24: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/24.jpg)
Validity of the summary statistics approach
The summary stats approach is exact if for each session/subject:
But: Summary stats approach is fairly robust against violations of these
conditions.
Within-session covariance the same
First-level design the same
One contrast per session
![Page 25: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/25.jpg)
Mixed effects analysis: spm_mfx
Summarystatistics
EMapproach
(2)̂
jjj
i
Tii QXQXV
XXY
)2()2()1()1()1()1(
)2(
)1(ˆ
yVXXVX TT 111)2( )(ˆ
yVXXVX TT 111)1( )(ˆ
},,{ QXnyyREML T
Step 1
Step 2
},,,{
][)1()2(
1)1()1(
1
)2()1()0(
TXQXQQ
XXXX
IVXXX
][ )1()0(
datay
Friston et al. 2005, NeuroImage
non-hierarchical model
1st level non-sphericity
2nd level non-sphericity
pooling over voxels
![Page 26: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/26.jpg)
2nd level non-sphericity modeling in SPM8
• 1 effect per subject→use summary statistics approach
• >1 effect per subject→model sphericity at 2nd level using variance basis
functions
![Page 27: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/27.jpg)
Reminder: sphericity
„sphericity“ means:
ICov 2)(
Xy )()( TECovC
Scans
Scan
s
i.e.2)( iVar
1001
)(Cov
![Page 28: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/28.jpg)
2nd level: non-sphericity
Errors are independent but not identical:
e.g. different groups (patients, controls)
Errors are not independent and not identical:
e.g. repeated measures for each subject (multiple basis functions,
multiple conditions etc.)
Errorcovariance
![Page 29: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/29.jpg)
Example of 2nd level non-sphericity
Qk:
Error Covariance
...
...
![Page 30: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/30.jpg)
Example of 2nd level non-sphericity
30
y = X + eN 1 N p p 1 N 1
N
N
error covariance
• 12 subjects, 4 conditions
• Measurements between subjects uncorrelated
• Measurements within subjects correlated
• Errors can have different variances across subjects
Cor(ε) =Σk λkQk
![Page 31: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/31.jpg)
2nd level non-sphericity modeling in SPM8:assumptions and limitations
• Cor() assumed to be globally homogeneous
• k’s only estimated from voxels with large F
• intrasubject variance assumed homogeneous
![Page 32: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/32.jpg)
Practical conclusions• Linear hierarchical models are used for group analyses of multi-
subject imaging data.
• The main challenge is to model non-sphericity (i.e. non-identity and non-independence of errors) within and between levels of the hierarchy.
• This is done by estimating hyperparameters using EM or ReML (which are equivalent for linear models).
• The summary statistics approach is robust approximation to a full mixed-effects analysis.– Use mixed-effects model only, if seriously in doubt about validity of
summary statistics approach.
![Page 33: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/33.jpg)
Recommended reading
Linear hierarchical models
Mixed effect models
![Page 34: Group analyses of fMRI data](https://reader036.fdocuments.us/reader036/viewer/2022062520/56816071550346895dcf9981/html5/thumbnails/34.jpg)
Thank you