Professional Documents
Culture Documents
An example data
Xin Di, PhD & Suril Gohel New Jersey Institute of Technology
Eye open/Eye closed data from INDI 24 Subjects Eye open, 6min, 240 images Eye close, 6min, 240 images TR 2 second, Voxel size: 3.13.13.5 mm Anatomical SPGR image (MPRAGE Image)
(http://fcon_1000.projects.nitrc.org/indi/retro/BeijingEOEC.html)
Temporal processing
Noise removal Filtering Global Intensity normalization/Global Regression Specific to resting-state fMRI
Noise removal
Physiological noises Head motion
Physiological noises
o Cardiac o Respiratory
Noise removal
Physiological noises Head motion
WM/CSF Signal o High probability threshold p > 0.99 not p > 0.5 o Use eroded WM/CSF masks o Use unsmoothed fMRI data o Mean time course or principle components (Chai et al., 2012)
Filtering
Usually 0.01 0.08 Hz
Filtering
Usually 0.01 0.08 Hz
5/31/2013
Noise removal
Physiological noises Head motion
Head motion
o Six rigid-body motion parameters (translation and rotation) o First order derivatives o Autoregressive model (current and previous time points) (Friston et al., 1995, Yan et al., 2013)
Noise removal
Physiological noises Head motion
o o
Filtering
Usually 0.01 0.08 Hz
Filtering
Usually 0.01 0.08 Hz
(Murphy et al., 2009; Saad et al., 2012) (BOLD) correlations of negative connectivity (Keller et al., 2013) o o Reduce Inter-subject variance (Yan et Use with precaution about what al. 2013) negative correlation in the data represents
1055
4 1050 2
Experiment
Eye open vs. Eye close
Goal
0 50 100 150 200 250
1030
50
100
150
200
250
-8 0
50
100
150
200
250
-6
Band-pass filtering
Functional specification
Local properties
800 600
300
0.2
200
0.1
400
100
Functional integration
Connectivity
0 50 100 150 200 250
-0.2
-400
-200
-0.3
-600 -800
-300
-0.4
50
100
150
200
250
50
100
150
200
250
-0.5
First 5 PCs of WM
Motion parameters
Time-domain Properties
Standard deviation (Biswal et al., 2007)
Frequency-domain Properties
Amplitude of Low-Frequency Fluctuation (ALFF, Zang et al., 2007) Fractional Amplitude of Low-Frequency Fluctuation (fALFF, Zou et al., 2008)
Homogenous Properties
Regional Homogeneity (ReHo, Zang et al., 2004)
Network centralities
Eigen vector centrality (ECM, Lohmann et al., 2010; Wink et al., 2012)
5/31/2013
W=
( R i ) 2 n ( R )2 1 2 3 K ( n n) 12
Paired t-test Eye open mALFF Eye close mReHo Eye open Eye close
eye open > eye close
SPM mip [0, -1, 1]
mALFF
open > close
mfALFF
open - close
contrast(s)
mReHo
open - close
contrast(s)
ECM
contrast(s)
5 10 15 20 25 30 35
<
<
<
<
<
5
<
5 10
10 15 20 25 30 35 40 SPMresults: .\Group\mReHo_unsmoothed Height 45 threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 118 voxels 5 10 15 20 25 Design matrix contrast(s)
<
SPM{T23}
<
mfALFF
ECM
eye close > eye open
SPM mip [0, -1, 1]
SPM{T23}
<
SPM{T23}
15 20 25 30 35 40
SPMresults:F:\INDI_Lite_FMRI\Group\mALFF Height threshold T = 3.484964 {p<0.001 (unc.)} close Extent threshold k = 70 voxels
> open
40 SPMresults: F:\INDI_Lite_FMRI\Group\mfALFF Height 45 threshold T = 3.484964 {p<0.001 (unc.)} close Extent threshold k = 10 voxels 5 10 15 20 25 Design matrix contrast(s)
close - open
45
5 10 15 20 25 30 35
mip
<
<
<
<
<
5
<
5 10
10 15 20 25
<
SPM{T23}
SPM{T23}
<
SPM{T23}
15 20 25 30 35 40 45 5 10 15 20 Design matrix 25
<
30 35 40 SPMresults: .\Group\mReHo_unsmoothed
SPMresults:F:\INDI_Lite_FMRI\Group\mALFF
Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 40 voxels
40 SPMresults: F:\INDI_Lite_FMRI\Group\mfALFF
Height 45 threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 30 voxels 5 10 15 20 25 Design matrix
Height 45 threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 118 voxels 5 10 15 20 25 Design matrix
cluster-level
pFWE-corr
0.012 0.015 0.034
peak-level
qFDR-corr
0.048 0.048 0.066
T
7.05 6.94 6.59
(Z)
5.09 5.05 4.89
puncorr
0.000 0.000 0.000
mm mm mm
30 -19 39 -31 -6 -7 19 10 40
250
cluster-level
qFDR-corr
0.000
peak-level
T
6.26 6.22 5.69 5.78 5.71 5.43 4.91 4.73 4.73
pFWE-corr
0.000
kE
puncorr
pFWE-corr
0.045 0.048 0.128 0.108 0.124 0.201 0.454 0.574 0.574
qFDR-corr
0.096 0.096 0.096 0.096 0.096 0.096 0.199 0.239 0.239
(Z)
puncorr
mm mm mm
27 -27 -27 36 33 42 -30 -51 -30 -31 -40 -19 -31 -22 -31 -25 -37 -13
1895
0.000
200
200
0.000
0.000
561
0.000
0.000
0.000
249
0.000
150
150
67 58 61 13 22 19 10 16 10
100
100
50
50
table shows 3 local maxima more than 8.0mm apart Height threshold: p = 0.001 (1.000) 0 0.1 T = 3.48,0.2 0.3 Extent threshold: k = 40 voxels, p = 0.010 (0.130) Expected voxels per cluster, <k> = 5.391 Expected number of clusters, <c> = 0.14 FWEp: 6.430, FDRp: 6.908, FWEc: 4035, FDRc: 4035
0.4
FWHM = 12.1 12.0 11.8 mm mm mm; 4.0 4.0 3.9 {voxels} Volume: 1844532 = 68316 voxels = 982.0 resels Voxel size: 3.0 3.0 3.0 mm mm mm; (resel = 63.12 voxels)
0.8
0.9
table shows 3 local maxima more than 8.0mm apart Height threshold: p = 0.001 (1.000) 0 0.1 T = 3.48,0.2 0.3 Extent threshold: k = 118 voxels, p = 0.002 (0.014) Expected voxels per cluster, <k> = 9.857 Expected number of clusters, <c> = 0.01 FWEp: 6.201, FDRp: Inf, FWEc: 249, FDRc: 249
0.4
FWHM = 15.0 13.6 15.3 mm mm mm; 5.0 4.5 5.1 {voxels} Volume: 1844532 = 68316 voxels = 537.1 resels Voxel size: 3.0 3.0 3.0 mm mm mm; (resel = 115.40 voxels)
0.8
0.9
Functional connectivity
Correlation Mutual information
A B
ROI-based analysis
Simple Need strong hypothesis Selectivity problem
Effective connectivity
Granger causality Dynamic causal model
A B
5/31/2013
Seed-based correlation
Less strong hypothesis Exploratory Specificity Correlation maps might be affected by seed selection
Sn(1) WM PCA1
Sn(1) WM PCA2
Sn(1) WM PCA3
Sn(1) WM PCA4
Sn(1) WM PCA5
Sn(1) LV 2
Sn(1) R1
Sn(1) R2
Sn(1) R3
Sn(1) R4
Sn(1) R5
Local differences (ALFF differences) Structural (AAL, Brodmanns area) Previous study (Coordinates) Meta-analysis (Activation likelihood estimation )
Motion parameters (translation, rotation, derivatives, autoregressive model...) White matter, CSF (mean, PCA princomp) Global signal High-pass filter (1/100 Hz) Series correlation (AR(1)?)
Design matrix
Sn(1) R6
Defining ROIs
Covariates
Sn(1) constant
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
images
Seed
WM
parameters CSF
Motion Constant
(gray not uniquely specified)
parameter estimability
Design description...
Basis functions : Number of sessions : Trials per session : Interscan interval : High pass Filter : Global calculation : Grand mean scaling : Global normalisation : hrf 1 0 2.50 {s} Cutoff: 100 {s} mean voxel value session specific None
Eye open
Eye close
R. visual seed
open - close
contrast(s)
PCC seed
open - close
contrast(s)
contrast(s)
open - close
5 10
5 10
<
<
<
<
<
<
5 10
<
SPM{T23}
15 20 25 30 35
<
SPM{T23}
15 20 25 30 35
<
SPM{T23}
15 20 25 30 35
40
close - open
45
SPMresults:.\Group\Seeding_RV Height threshold T = 3.484964 {p<0.001 (unc.)} close Extent threshold k = 40 voxels 5 10 15 20 25 Design matrix contrast(s) Statistics: p-values adjusted for search volume
set-level
p
0.000
40
- open
45
SPMresults:.\Group\Seeding_PCC Height threshold T = 3.484964 {p<0.001 (unc.)} open Extent threshold k = 76 voxels 5 10 15 20 25 Design matrix contrast(s) Statistics: p-values adjusted for search volume
40
- close
c
5
pFWE-corr
0.000
qFDR-corr
0.000
kE
420
puncorr
0.000
pFWE-corr
0.031 0.239 0.245 0.091 0.202 0.234 0.250 0.601 0.851 0.999 0.806 0.818
kE
puncorr
0.002
pFWE-corr
0.240 0.340 0.544 0.243 0.482 0.997
qFDR-corr
0.482 0.482 0.482 0.482 0.482 0.997
<
<
250
cluster-level <
<
<
<
peak-level
T (Z)
4.30 4.18 3.99 4.29 4.04 3.28
puncorr
0.000 0.000 0.000 0.000 0.000 0.001
mm mm mm
-42 -22 -14 -39 -25 -5 -51 -37 -11 -12 38 -5 12 35 -5 9 50 -14
5 200 10
0.000
0.000
486
0.000
<
SPM{T23}
15 20 150 25 30 100 35
0.041 0.005
0.010 0.002
64 104
0.005 0.001
<
0.134 0.027 43
SPM{T23}
0.017
0.017 -18 0.023 0.000 -76 34 76 0.000 -12 -82 43 0.000 -30 -79 10 0.001 18 0.003 0.000 -79 46 131 0.000 30 -46 49 0.000 33 -73 22 0.000 -27 -55 55 0.000 -27 -55 -14 0.000 -39 -67 -5 0.001 -24 -61 -8 0.000 6 -73 -20 0.000 -9 -76 -17
0.000
<
SPM{T23}
15 20 25 30 35
100 35
SPMresults:.\Group\Seeding_LV
40 45
50
Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 57 voxels 5 10 15 20 25 Design matrix
Height threshold: p = 0.001 (1.000) 0 0.1 T = 3.48,0.2 0.3
SPMresults:.\Group\Seeding_RV
40 45
50
Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 50 voxels 5 10 15 20 25 Design matrix 0.4
SPMresults:.\Group\Seeding_LV
40 45
25
Extent threshold: k = 40 voxels, p = 0.020 (0.160) Expected voxels per cluster, <k> = 6.903 Expected number of clusters, <c> = 0.17 FWEp: 6.245, FDRp: Inf, FWEc: 64, FDRc: 43
0.4
0 Height threshold: p = 0.001 (1.000) Degrees of freedom 23.0] 0.5 0.6 = [1.0,0.7 0.8 0.9 10.3 0 0.1 T = 3.48, 0.2
table shows 3 local maxima more than 8.0mm apart Degrees of freedom 23.0] 0.5 0.6 = [1.0,0.7 FWHM = 12.5 12.8 12.7 mm mm mm; 4.2 4.3 4.2 {voxels} Volume: 1422630 = 52690 voxels = 622.4 resels Voxel size: 3.0 3.0 3.0 mm mm mm; (resel = 75.44 voxels)
Extent threshold: k = {voxels} 76 voxels, p = 0.002 (0.017) FWHM = 12.9 13.0 13.0 mm mm mm; 4.3 4.3 4.3 Volume: 1422630 = 52690 voxels = 581.0 resels Expected voxels per cluster, <k> = 6.443 Expected number of clusters, Voxel size: 3.0 3.0 3.0 mm mm mm; (resel = 80.82 voxels)<c> = 0.02 FWEp: 6.280, FDRp: Inf, FWEc: 76, FDRc: 76
5/31/2013
DMN
1st Visual
Right Executive
Higher Visual
Left Executive
Motor
1st Visual IC
open - close
Higher Visual IC
open - close
contrast(s) contrast(s)
DMN IC
open - close
contrast(s)
5 10
5 10
<
<
<
<
<
<
5 10
<
SPM{T23}
15 20 25 30 35
<
SPM{T23}
15 20 25 30 35
<
SPM{T23}
15 20 25 30 35
SPMresults:.\ICA_preprocessed_IC8_1stVis Height threshold T = 3.484964 {p<0.001 (unc.)} open Extent threshold k = 50 voxels
40
- close
45
SPMresults:.\ICA_preprocessed_IC9_2ndVis Height threshold T = 3.484964 {p<0.001 (unc.)} open Extent threshold k = 55 voxels 5 10 15 20 25 Design matrix contrast(s)
40
- close
45
SPMresults:.\Group\Seeding_LV Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 50 voxels 5 10 15 20 25 Design matrix contrast(s)
40
open - close
5 10
5 10
DMN IC
<
<
<
<
<
<
5 10
<
SPM{T23}
15 20 25 30 35
<
SPM{T23}
15 20 25 30 35
<
SPM{T23}
15 20 25 30 35
SPMresults:.\Group\Seeding_LV
40 45
SPMresults:.\Group\Seeding_LV Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 50 voxels 5 10 15 20 25 Design matrix
40 45
SPMresults:.\Group\Seeding_LV Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 50 voxels 5 10 15 20 25 Design matrix
40 45
Whether the effects are reliable across subject? Whether the effects are different between two groups? Whether the effects are different between conditions? More than two groups/conditions
Condition 1 Condition 2 Subject 1 Subject 2 Subject 3 Subject 4 Subject 5 Subject 6 Subject 7 Subject 8 Subject 9 Subject 10 Subject 11 Subject 12 Subject 13 Subject 14 Subject 15 Subject 16 Subject 17 Subject 18 Subject 19 Subject 20 Subject 21 Subject 22 Subject 23 Subject 24
Within-subject effects Inter-individual differences are less likely to affect the results
images
Design matrix
Paired t-test
. . . . . . . . . . . . . . . . . . . . . . . .
Conditions
parameters Subjects
(gray not uniquely specified)
parameter estimability
Design description...
Design : Global calculation : Grand mean scaling : Global normalisation : Parameters : Paired t-test omit <no grand Mean scaling> <no global normalisation> 2 condition, +0 covariate, +24 block, +0 nuisance 26 total, having 25 degrees of freedom leaving 23 degrees of freedom from 48 images
5/31/2013
Groups
Groups
Groups
Design matrix
gender
Group differences Dealing with potential confounding variables Covariates Age Gender Motion Brain structure ALFF
age
. . . . . . . . . . . . . . . . . . . . . . . . .
images
Groups
parameters
Effects of no interest
(gray not uniquely specified)
parameter estimability
Design description...
Design : Global calculation : Grand mean scaling : Global normalisation : Parameters : ANOVA omit <no grand Mean scaling> <no global normalisation> 3 condition, +2 covariate, +0 block, +0 nuisance 5 total, having 5 degrees of freedom leaving 70 degrees of freedom from 75 images
Including ALFF maps as a voxel-wise covariate revealed significant increased connectivity in 11 RSNs. Interestingly, the number of voxels with significant increased connectivity generally decreased. Abou Elseoud et al., (2012) FSL SPM/MATLAB: REST, WFU Biological Parametric Mapping (BPM)
Lower ECM
contrast(s)
<
<
<
<
10 15 20 25 30 35
10
<
SPM{T23}
<
Be careful when using global signal regression Take care of multiple comparison problem
Not only multiple voxels, but also multiple ICs, and multiple seeds
SPM{T23}
15 20 25 30 35 40 45 5
40 SPMresults: .\Group\mReHo_unsmoothed 45 Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 118 voxels 5 10 15 20 25 Design matrix
<
<
10 15 20 25 30 35
10
<
SPM{T23}
<
SPM{T23}
15 20 25 30 35 40 45 5 10 15 20 Design matrix 25
40 SPMresults: .\ICA_preprocessed_IC8_1stVis 45 Height threshold T = 3.484964 {p<0.001 (unc.)} Extent threshold k = 50 voxels 5 10 15 20 25 Design matrix
cluster-level
c
5
peak-level
kE
420
pFWE-corr
0.000
qFDR-corr
0.000
puncorr
0.000
pFWE-corr
0.031 0.239 0.245 0.091 0.202 0.234 0.250 0.601 0.851 0.999 0.806 0.818
qFDR-corr
0.090 0.119 0.119 0.119 0.119 0.119 0.119 0.222 0.345 0.802 0.330 0.330
T
6.51 5.37 5.36 5.92 5.47 5.39 5.35 4.74 4.35 3.64 4.43 4.41
(Z)
4.85 4.28 4.27 4.57 4.34 4.29 4.27 3.92 3.68 3.20 3.73 3.71
puncorr
0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.000 0.001 0.000 0.000
mm mm mm
-18 -12 -30 18 30 33 -27 -27 -39 -24 6 -9 -76 34 -82 43 -79 10 -79 46 -46 49 -73 22 -55 55 -55 -14 -67 -5 -61 -8 -73 -20 -76 -17
200
0.000
0.000
486
0.000
150
0.041 0.005
0.010 0.002
64 104
0.005 0.001
0.134
0.027
43
0.017
100
50
table shows 3 local maxima more than 8.0mm apart Height threshold: p = 0.001 (1.000) 0 0.1 T = 3.48,0.2 0.3 Extent threshold: k = 40 voxels, p = 0.020 (0.160) Expected voxels per cluster, <k> = 6.903 Expected number of clusters, <c> = 0.17 FWEp: 6.245, FDRp: Inf, FWEc: 64, FDRc: 43
0.4
FWHM = 12.9 13.0 13.0 mm mm mm; 4.3 4.3 4.3 {voxels} Volume: 1422630 = 52690 voxels = 581.0 resels Voxel size: 3.0 3.0 3.0 mm mm mm; (resel = 80.82 voxels)
0.8
0.9
5/31/2013
ALFF: Zang YF, He Y, Zhu CZ, Cao QJ, Sui MQ, Liang M, Tian LX, Jiang TZ, Wang YF, (2007). Altered baseline brain activity in children with ADHD revealed by resting-state functional MRI. Brain Dev 29(2):83-91. fALFF: Zou QH, Zhu CZ, Yang Y, Zuo XN, Long XY, Cao QJ, Wang YF, Zang YF, (2008). An improved approach to detection of amplitude of low-frequency fluctuation (ALFF) for restingstate fMRI: fractional ALFF. J Neurosci Methods172(1):137-41. Reho: Zang Y, Jiang T, Lu Y, He Y, Tian L, (2004). Regional homogeneity approach to fMRI data analysis. Neuroimage22(1):394-400. ECM: Lohmann G, Margulies DS, Horstmann A, Pleger B, Lepsien J, Goldhahn D, Schloegl H, Stumvoll M, Villringer A, Turner R, (2010). Eigenvector centrality mapping for analyzing connectivity patterns in fMRI data of the human brain. PLoS One 5(4):e10232. ECM: Wink AM, de Munck JC, van der Werf YD, van den Heuvel OA, Barkhof F, (2012). Fast eigenvector centrality mapping of voxel-wise connectivity in functional magnetic resonance imaging: implementation, validation, and interpretation. Brain Connect 2(5):265-74. Seed-based correlation: Biswal B, Yetkin FZ, Haughton VM, Hyde JS, (1995). Functional connectivity in the motor cortex of resting human brain using echo-planar MRI. Magn Reson Med34(4):537-41. Seed-based correlation: Cordes D, Haughton VM, Arfanakis K, Wendt GJ, Turski PA, Moritz CH, Quigley MA, Meyerand ME, (2000). Mapping functionally related regions of brain with functional connectivity MR imaging. AJNR Am J Neuroradiol 21(9):1636-44. ICA: Beckmann CF, DeLuca M, Devlin JT, Smith SM, (2005). Investigations into resting-state connectivity using independent component analysis. Philos Trans R Soc Lond B Biol Sci 360(1457):1001-13.