Resting-state fMRI and DTI
David C. Zhu, Ph.D.
Associate Professor of Radiology and Psychology
Cognitive Imaging Research Center
Michigan State University, East Lansing, Michigan, USA
Facts
• The brain is only about 2% of total body
mass.
• But it consumes about 20% of the body’s
total energy at “rest”.
• Engaging in active tasks increases neuronal
metabolism less than 5%.
What is this energy at “rest” consumed by?
The on-going spontaneous neuronal activity
Huettel SA, Song AW, McCarthy G. Functional Magnetic Resonance Imaging.
Oxygenated
and
deoxygenated
hemoglobin
( )
(Hb)
Two Main Paths of BOLD (Blood Oxygen LevelLevel-Dependent) fMRI
First 1-2 seconds
Later seconds
Stimulation
Neuronal Activity
CMRglucose
CMRO2
Blood Oxygen Level
Deoxygenated hemoglobin: paramagnetic
Blood Magnetic
g
Susceptibility Effects
Cerebral Blood
Flow (CBF)
Blood Oxygen Level
Oxygenated hemoglobin: diamagnetic
Blood Magnetic
g
Susceptibility Effects
T2* decay
T2* decay
fMRI Image Signal Intensity
fMRI Image Signal Intensity
Resting-state fMRI based functional connectivity analyses
Assumptions
Strong BOLD fMRI temporal correlation
Strong neuronal synchronous activity
Strong functional connectivity
Same/intact network
B
A
D
C
E
Two popular methods of processing
for resting-state fMRI
1 Correlation
1.
C
l i analysis
l i
p
component
p
analysis
y (ICA)
(
)
2. Independent
Seed-based
Seedcorrelation analysis
seed
EPI
Smeasure = Sintrinsic + Srandom
Time courses of correlated and uncorrelated regions
Seed
pC/rsp
Left
et
MeFG
Right
MTG
Left
IPL
R ti
Resting-state
t t fMRI pre-processing
i steps
t
1.
2.
3
3.
4.
Slice-timing and motion correction.
Remove baseline, linear and quadratic system trends.
Spatial blurring
blurring.
Remove “Nuisance” signals of
(a) Global mean
(b) CSF
(c) White matter
5. Band-pass
p filtering:
g 0.009 Hz – 0.08 Hz
Seed Region #1: pC/rsp (posterior cingulate/retrosplenial cortex)
Group Integration
(17 ssubjects):
bjects):
pC/rsp
Whole-brain
corrected P < 0.0325.
Mean structural
M
t t l
normalized
connectivity
distribution > 10-4.
MeFC/ACC
(Medial Frontal Cortex/
Anterior Cingulate Cortex)
Zhu DC, Majumdar S. Integration of resting-state fMRI and diffusion-weighted MRI
connectivity analyses of the human brain: limitations and improvement. J Neuroimaging.
2014 Mar-Apr;24(2):176-86.
Seed Region
eg o #2
# : Right
g SPL
S (supe
(superior
o parietal
pa e a lobule)
obu e)
Gro p Integration
Group
(17 subjects):
Whole-brain
corrected
t d P < 0.0325.
0 0325
Mean structural
normalized
connectivity
distribution > 10-4.
Right SPL
Left SPL
Seed Region #3: Left Cuneus
Group
p Integration
g
(17 subjects):
Whole-brain
corrected P < 0.0325.
Mean structural
normalized
connectivity
distribution > 10-4.
Left cuneus
Right cuneus
Independent Component Analysis
http://www.fmrib.ox.ac.uk/fsl/melodic/index.html
http://www.fmrib.ox.ac.uk/fsl/melodic/index.html
ICA example results with MELODIC in FSL
IC 1: Default Mode Network
IC 25: Visual Network
Default--Mode Network
Default
Medial temporal lobe (MeTL)
Medial prefrontal cortex (MePFC)
(hippocampus, parahippocampal gyrus,
entorhinal gyrus)
((memory
y pprocessing)
g)
(Medial frontal cortex, anterior cingulate
cortex, superior frontal cortex)
((facilitation))
Structural connectivity
(diffusion MRI fiber tracking)
Posterior cingulate cortex/retrosplenial cortex (PCC/RSC)
(Integration)
temporoparietal junction cortices
(TPJC):
(angular gyrus, superior/middle temporal gyrus,
parietal lobule,
lobule supramarginal gyrus)
Buckner RL, Andrews-Hanna JR, Schacter DL (2008) The brain's default network: anatomy,
function, and relevance to disease. Ann N Y Acad Sci 1124, 1-38.
Alzheimer’s disease and amnestic mild
cognitive impairment weaken connections
within the default-mode network
Zhu DC, Majumdar S, Korolev IO, Berger KL, Bozoki AC. Alzheimer’s
di
disease
andd amnestic
i mild
ild cognitive
i i impairment
i
i
weaken
k connections
i
within the default-mode network: a multi-modal imaging study. J
Alzheimer's Dis. 2013;34(4):969-84.
7 minute resting-state fMRI EPI scan
(relax and eyes open in dim-light
dim light
condition):
38 contiguous 3-mm axial slices,
22 cm × 22 cm FOV,
64 × 64 matrix size
size,
27.7 ms TE, 2500 ms TR,
80° flip angle,
164 time points.
Seed location
GM seed
Amnestic mild
cognitive impairment
Normal
R cingulum
L TPJC
Alzheimer’s disease
R MeFG
(a)
WM seed
R TPJC
R isthmus of cingulate
R TPJC
L cingulum
L MeFG
GM seed
(b)
WM seed
L TPJC
L isthmus
i th
off cingulate
i
l t
Thresholds: whole-brain corrected p ≤ 0.033 for functional connectivity and mean
connectivity distribution of > 1000 for structural connectivity. Green: functional
connection only. Orange: structural connection only. Red: coexistence of both
connections. GM: gray matter. WM: white matter. R = right. L = left. MeFG =
medial frontal gyrus. TPJC = temporoparietal junction cortices.
Functional connectivity with seed region at
the right isthmus of cingulate cortex (Normal > AD)
R STG/MTG
R precuneus
PET
L precuneus
L AG
L MTG
L CG
R AG/IPL
(a) R TPJC
(b) PCC/RSC
(c) L TPJC
(d) PET: NC > AD
Structural connectivity with the seed region in the associated
white matter of the right isthmus of cingulate cortex
NC > AD
NC > aMCI
A potential biomarker in sportssports-related
concussion: brain functional connectivity
alteration of the defaultdefault-mode network measured
with longitudinal resting
resting--state fMRI over 30 days.
days
Zhu DC, Covassin T, Nogle S, Doyle S, Russell D, Pearson RL,
Monroe J, Liszewski CM, DeMarco JK, Kaufman DI. J Neurotrauma.
2015 Mar 1;32(5):327-41.
The DMN Network
R IPL/AG
L PCC
R PCC
L IPL/AG
L SFG R SFG
L Hippo
R ACC/MeFC
L ACC/MeFC
Ovverall DMN Connnectivitty (R)
Mean default
default--mode network connectivity
(7 concussed, 11 control)
0.5
0.45
0.4
0.35
0.3
0.25
02
0.2
0.15
01
0.1
0.05
0
Concussed
Control
Day 1
Day 7
Day 30
DMN after concussion (n = 7)
(Node # 1 = left PCC, 2 = left ACC/MeFC, 3 = left SFG, 4 = left IPL/AG, 5 = left hippocampus,
6 = right PCC, 7 = right ACC/MeFC, 8 = right SFG, 9 = right IPL/AG, 10 = right hippocampus)
3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 R value
1
1
0.8
2
2
3
3
0.6
4
4
5
5
0.4
6
6
7
7
0.2
8
8
0
9
9
-0.2
10
10
Mean correlation R on Day 1
Mean correlation R on Day 7
Mean correlation R on Day 30
Node 1 2
1
2
3
4
5
6
7
8
9
10
3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 p value
0.05
1
1
2
2
0.04
3
3
4
4
0.03
5
5
6
6
0.02
7
7
8
8
0.01
9
9
10
10
0
t test of Day 1 vs. Day 7
t test of Day 7 vs. Day 30
t test of Day 1 vs. Day 30
Node 1 2
1
2
3
4
5
6
7
8
9
10
DMN of control subjects (n=11)
(Node # 1 = left PCC, 2 = left ACC/MeFC, 3 = left SFG, 4 = left IPL/AG, 5 = left hippocampus,
6 = right PCC, 7 = right ACC/MeFC, 8 = right SFG, 9 = right IPL/AG, 10 = right hippocampus)
Node 1 2
1
2
3
4
5
6
7
8
9
10
3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 R value
1
1
2
2
0.8
3
3
06
0.6
4
4
5
5
0.4
6
6
7
7
0.2
8
8
0
9
9
-0.2
10
10
Mean correlation R on Day 1
Mean correlation R on Day 7
Mean correlation R on Day 30
3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 Node 1 2 3 4 5 6 7 8 9 10 p value
0.05
1
1
2
2
0.04
3
3
4
4
0.03
5
5
6
6
0.02
7
7
8
8
0.01
9
9
10
10
0
t test of Day 1 vs. Day 7
t test of Day 7 vs. Day 30
t test of Day 1 vs. Day 30
Node 1 2
1
2
3
4
5
6
7
8
9
10
The mean functional connectivity to left isthmus
of cingulate cortex (ICC) of the concussed group
((correlation R > 0.25,, n = 7))
Day 1
(b) ANOVA: Day 1 vs. Day 7 (n = 8)
Day
y7
Day 30
(c) ANOVA: Day 1 vs. Day 30 (n = 7)
Case study
The functional and structural connectivityy to left isthmus of
cingulate cortex of a concussed subject over one month
(correlation R > 0.4 and connectivity distribution > 1000).
Seed regions
Day 1
structural
functional
Green: functional seed
Orange: structural seed
Day 7
Day 30
Red: overlap regions.
The above
Th
b
functional
f ti l connectivity
ti it reduction
d ti from
f
Day
D 1 to
t Day
D 7 was
seen in 8 of our 9 concussed cases.
Resting-state fMRI based functional connectivity analyses
Assumptions
?
Strong BOLD fMRI temporal correlation
Vascular confound
St
Strong
neuronall synchronous
h
activity
ti it
Why?
Strongg functional connectivityy
The underlying cellular
and molecular bases
Same/intact network
B
A
D
C
E
Introduction to Diffusion Tensor Imaging (DTI)
Fick’s Law describes particle movement
Net flux (mole
mm2/s):
ΔC
J = −D
Δx
Diffusion coefficient D is in mm2/sec
ΔC
= concentration gradient in mole/mm4
Δx
High
concentration
J
Low
concentration
Figure 5
5.20
20 Diffusion
Stejskal-Tanner Diffusion-weighted Sequence
G
Isotropic Diffusion
Signal attenuation due to diffusion coefficient D
S
= e −bD
A=
S0
Where b = commonly called “b factor”,
which characterizes the gradient pulses
(timing, amplitude, shape) = clinical
practice, 1000 s/mm2
Stejskal-Tanner Diffusion-weighted Sequence
G
δ
b = γ G δ (Δ − )
3
2
2
2
Anisotropic Case (for example, axons)
A=e
⎡bxx
⎢
b = ⎢b yx
⎢bzx
⎣
bxy
b yy
bzy
bxz ⎤
⎥
b yz ⎥
bzz ⎥⎦
A=e
− bD
⎡ D xx
⎢
D = diffusion tensor = ⎢ D yx
⎢ D zx
⎣
D xy
D yy
D zy
D xz ⎤
⎥
D yz ⎥
D zz ⎥⎦
− ( bxx Dxx + b yy D yy + bzz Dzz + 2 bxy Dxy + 2 bxz Dxz + 2 b yz D yz )
Diagnolization
DE = EΛ
E = eigen vector (unit vector)
⎡λ1 0
⎢0 λ
2
⎢
⎢⎣ 0 0
Λ = eigen value =
0⎤
0 ⎥⎥
λ3 ⎥⎦
z
x’
y’
y
x
Laboratory frame
λ2
λ1
λ3
z’
Diffusion ellipsoid
Mean diffusivity = λmean = (λ1+ λ2 +λ3 )/3
The level of tissue constraint
F ti l Anisotropy
Fractional
A i t
= FA =
3[(λ1 − λmean ) 2 + (λ2 − λmean ) 2 + (λ3 − λmean ) 2 ]
2(λ12 + λ22 + λ32 )
the directionalityy of diffusion
Axon fiber integrity
Mean diffusivity map
FA map
Pierpaoli C, Jezzard P, Basser PJ, Barnett A, Di Chiro G. Diffusion tensor MR
imaging of the human brain. Radiology. 1996 Dec;201(3):637-48.
Fiber Tracking
Deterministic approach:
pp
Answer Yes/no.
Software: (1) DTI Studio
(2) MedINRIA
Probabilistic
P
b bili i approach:
h
How probable two voxels/regions
connected together
g
Software: FSL
Probabilistic tracking at Right Fornix
Data Acquisition
GE 3T Signa® HDx MR scanner with an 88-channel
channel head coil.
coil
12 minute and 6 second DTI scan (full-brain coverage):
Dual spin echo EPI sequence,
48 2.4-mm axial slices,
22 cm × 22 cm FOV,
FOV 128 × 128,
128 2 NEX,
NEX
75 ms TE, 13.7 s TR,
parallel imaging acceleration factor = 2,
b = 1000 s/mm2, 25 directions
DTI Analysis with the Diffusion Toolbox (FDT v2.0)
in FSL software package
Eddy-current distortion and motion correction.
Applied Bayesian Estimation of Diffusion Parameters Obtained
using Sampling Techniques with the crossing fibers (n = 2)
modeled
d l d (BEDPOSTX).
(BEDPOSTX)
Applied probabilistic tractography (PROBTRACKX) to each
seed region to calculate the corresponding connectivity
distributions.
The connectivity distributions were then normalized by the total
number of generated tracts from the seed region.
Behrens TE, Berg HJ, Jbabdi S, Rushworth MF, Woolrich MW. NeuroImage 2007, 34:144-155.
Integration
g
• Resting-state fMRI allows the examination
of brain function connectivity (1).
(1)
• Diffusion tensor imaging (DTI) fiber
tracking
ki allows
ll
the
h evaluation
l i off structurall
connection between cortical regions (2).
11. Fox
F MD
MD, R
Raichle
i hl ME
ME. N
Natt R
Rev N
Neurosci.
i 2007;
2007 8:700-711.
8 700 711
2. Le Bihan D, Mangin JF, Poupon C, Clark CA, Pappata S, Molko N,
Chabriat H. J Magn Reson Imaging. 2001;13:534-546.
Left PCC
Left ACC
Test both functional and structural connectivity.
Case study
The functional and structural connectivityy to left isthmus of
cingulate cortex of a concussed subject over one month
(correlation R > 0.4 and connectivity distribution > 1000).
Seed regions
Day 1
structural
functional
Green: functional seed
Orange: structural seed
Day 7
Day 30
Red: overlap regions.
The above
Th
b
functional
f ti l connectivity
ti it reduction
d ti from
f
Day
D 1 to
t Day
D 7 was
seen in 8 of our 9 concussed cases.
Volumetric Analysis and Segmentation
with
ith F
FreeSurfer
S f
Left-Putamen
Left-Pallidum
Left
Pallidum
Time point 0 Time point 1
6708
4869
2090
1618
% change
-27.4
-22.6
22.6
Region
lh_postcentral_volume
lh_precentral_volume
lh_p
precuneus_volume
Time 0 (cc) Time 1 (cc)
10724
7408
14308
11312
12278
9660
% change
-30.92
-20.94
-21.32