The Spatial, Temporal and Interpretive Limits of Functional MRI

1
The Spatial, Temporal and Interpretive Limits of
Functional MRI

• Peter A. Bandettini, Ph.D
• Unit on Functional Imaging Methods
• Laboratory of Brain and Cognition
• National Institute of Mental Health

2
Categories of Questions Asked with fMRI
Where? When? How much? --- How to get the brain
to do what we want it to do in the context of an
fMRI experiment? (limitations time, motion,
acoustic noise.)
3
A Primary Challenge
...to make progressively more precise inferences
using fMRI without making too many assumptions
about non-neuronal physiologic factors.
4
Contrast in Functional MRI

• Blood Volume
• Contrast agent injection and time series
  collection of T2 or T2 - weighted images
• BOLD
• Time series collection of T2 or T2 - weighted
  images
• Perfusion
• T1 weighting
• Arterial spin labeling

5
Resting Active
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
(No Transcript)
11
(No Transcript)
12
Perfusion / Flow Imaging
EPISTAR
FAIR







. . .
-
-
-
-
Perfusion Time Series
. . .
13
FAIR
EPISTAR
14
-
unique information baseline information
multislice trivial
invasive low C / N for func.
Volume BOLD Perfusion
highest C / N easy to implement multislice
trivial non invasive highest temp. res.
complicated signal no baseline info.
unique information control over ves. size
baseline information non invasive
multislice non trivial lower temp. res. low
C / N
15
Neuronal
Measured
Activation
fMRI
Signal
?
?
Hemodynamics
Physiologic Factors
16
Physiologic Factors that Influence BOLD Contrast
Coupling Flow CMRO2

• Blood oxygenation
• Blood volume
• Blood pressure
• Hematocrit
• Vessel size

17
(No Transcript)
18
Where and When?
The resolution is determined by the cerebral
hemodynamics. Know the vasculature at which
you are looking. (or) Normalize to the spatial
variation in the vasculature. (or) Make
several assumptions.
19
Single Shot Imaging
T2 decay
EPI Readout Window
20 to 40 ms
20
Multishot Imaging
T2 decay
EPI Window 2
21
Partial k-space imaging
22
(No Transcript)
23
(No Transcript)
24
(No Transcript)
25
(No Transcript)
26
Pulse sequence based methods for increasing
spatial and temporal resolution

• Spin-echo
• ASL
• Diffusion weighting
• Threshold based on magnitude

27
(No Transcript)
28
Spin echo vs. Gradient echo
compartment radius lt 3 µm 3
to 15 µm gt 15 µm
29
GE TE 30 ms
SE TE 110 ms
30
?R2 /?R2
Spin-echo??
1.5
1.5 intravascular
1.5 to 3 extravascular
1.5 intravascular
3 to 8 extravascular
average ?R2 / ?R2 3 to 4
31
3T
Spin-Echo TE 105 ms TR 8
Gradient-Echo TE 50 ms
Gradient-Echo functional TE 50 ms
Spin-Echo functional TE 105 ms
32
no diffusion weighting
diffusion weighting
33
(No Transcript)
34
b 0
b 10
b 160
b 50
35
Perfusion Rest Activation
BOLD
36
Anatomy
BOLD
Perfusion
37
Simultaneous Flow and BOLD
Mz(blood)
Mz(blood)
Mz(blood)
-
blood
t
t
t
tag
control
Flow
Mz(tissue)
Mz(tissue)
Mz(tissue)
tissue
-
t
t
t
Mz(blood)
Mz(blood)
Mz(blood)

blood
t
t
t
tag
control
BOLD
Mz(tissue)
Mz(tissue)
Mz(tissue)
tissue

t
t
t
38
Simultaneous BOLD and Perfusion
BOLD
Perfusion
39
Simultaneous BOLD and Perfusion
perfusion BOLD
40
Angiogram Perfusion BOLD
41
Spatial Normalization
Hypercapnia
42
T1 - weighted
T2 weighted
T1 and T2 weighted
43
Vascular Sensitization
44
Problems with pulse sequence - based methods for
increasing resolution

• Spin-echo (sensitivity, specificity)
• Arterial spin-labeling (sensitivity, time, range)
• Diffusion weighting (sensitivity, specificity)
• Threshold based on magnitude (sensitivity,
  specificity)

45
(No Transcript)
46
,
Finger Movement
Anatomical
5 CO2
12 O2
47
(No Transcript)
48


49
Hoge et al
50
Hoge et al
51
Mapping CMRO2 using CO2 Calibration
Hoge et al
52
(No Transcript)
53
Types of Temporal Resolution
1. Maximum on-off switching rate. 2. Minimum
detectable activation duration. 3. Minimum
detectable difference in activation duration or
onset in same region. 4. Minimum detectable
activation interval across separate brain
regions. 5. Maximum image acquisition rate.
54
(No Transcript)
55
S k t 8.6 e - t / 0.547
Cohen, Neuroimage 6, 93-103 (1997)
56
(No Transcript)
57
2 sec
Latency
- 2 sec
Magnitude
58
Temporal Normalization
Relative Timing
59
Regions of Interest Used for Hemi-Field
Experiment
Left Hemisphere
Right Hemisphere
60
(No Transcript)
61
Hemi-field with 500 msec asynchrony
Average of 6 runs Standard Deviations Shown
Percent
MR
Signal
Strength
Time (seconds)
62
Average of 6 runs Smoothed Data
Percent
MR
Signal
Strength
Time (seconds)
63
500 ms
500 ms
Right Hemifield
Left Hemifield
2.5 s

-
0 s
- 2.5 s
64
250 ms
250 ms
Right Hemifield
Left Hemifield
2.5 s

-
0 s
- 2.5 s
65
How Much?
Central Issue Spatial and temporal neuronal
firing integration to create an fMRI signal
change. - is the hemodynamic response a linear
system? -what is the dynamic range?
66
Motor Cortex
Auditory Cortex
67
DeYoe et al.
68
(No Transcript)
69
Time (sec)
70
(No Transcript)
71
Stimululs - Duration Dependent Deviation from
Linearity of the fMRI Response (Hemodynamic or
Neuronal?)
gtlinear
linear
72
Spatial Distribution of the Hemodynamic Response
Linearity
73
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
74
Ultimate Limits?
Spatial 0.5 mm Temporal 100 ms Interpretability
too early to tell, but hopeful
75
Neuronal Input Strategies

• Peter A. Bandettini, Ph.D
• Unit on Functional Imaging Methods
• Laboratory of Brain and Cognition
• National Institute of Mental Health

76
How to get the brain to do what we want it to do
in the context of an fMRI experiment?
Noise Artifact
"Interesting" Aspects of MR Signal
77
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
78
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
79
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
80
(No Transcript)
81
0.08 Hz 0.05 Hz
spectral density
c.c. gt 0.5 with spectra
82
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
83
(No Transcript)
84
(No Transcript)
85
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
86
(No Transcript)
87
(No Transcript)
88
(No Transcript)
89
(No Transcript)
90
Overt Word Production
2
3
4
5
6
7
8
9
10
11
12
13
91
Tongue Movement
Jaw Clenching
92
Event-Related fMRI Questions 1. What is the
optimal ISI? 2. How does functional contrast
compare with blocked timing? (Is the
hemodynamic response a linear system?)
93
Contrast in Event Related fMRI
Dependency on Inter-stimulus Interval
(ISI) Stimulus Duration (SD) Comparison
with Blocked strategies Synthesized responses
created using convolution
94
Issues 1. ISI Issue Shorter ISI provides
more trials per unit time. Shorter ISI causes
overlap in hemodynamic response, reducing
dynamic range. 2. Contrast Issue Does signal
behave like a linear system with brief SD?
95
Experimental Methods
Two imaging planes containing motor and
visual cortex. EPI, 3.75 x 3.75 x 7 mm, TE
40 ms, TR 1 sec. Time series duration 360
images (6 minutes). 10 series compared Single
Trial SD 2, ISI 24, 20, 16, 12, 10, 8, 6,
4, 2. Blocked SD 20, ISI 20. Subjects
instructed to tap fingers when GRASS goggles
were on.
96
Visual Cortex
ISI, SD
ISI, SD
8, 2
20, 20
6, 2
12, 2
4, 2
10, 2
2, 2
97
Motor Cortex
ISI, SD
ISI, SD
8, 2
20, 20
6, 2
12, 2
4, 2
10, 2
2, 2
98
ISI
Motor Cortex
ISI
Visual Cortex
99
ISI
Motor Cortex
ISI
Visual Cortex
100
Contrast to Noise Images
( ISI, SD )
S1
S2
101
Motor
Visual
( ISI, SD )
20, 20
2, 2
Relative differences in activation
intensities may reflect spatial differences in
hemodynamic responsivity. (draining veins vs.
capillaries).
102
Functional Contrast
( Block design 1 )
103
Response Synthesis
S k t 8.6 e - t / 0.547
Cohen, Neuroimage 6, 93-103 (1997)

104
Synthesized Responses
ISI, Dur
ISI, Dur
8, 2
20, 20
6, 2
12, 2
4, 2
10, 2
2, 2
105
Convolution
ISI
106
Functional Contrast
ISI (sec)
1
32
1
SD (sec)
32
107
Functional Contrast
( Block design 1 )
108
(No Transcript)
109
Possible implications for interpretation of
event-related data using short, randomized ISI w/
deconvolution.
Dale AM, Buckner RL (1997), Human Brain Mapping,
5, 329-340.
110
BOLD response - constant ISI
ISI 2 s
ISI 4s
ISI 8 s
Tasks can be performed faster by varying the ISI
111
(No Transcript)
112
(No Transcript)
113
(No Transcript)
114
You can go even faster with the assumption of
linearity...
115
If ISI is randomized, and if ON / OFF
distribution is 50, the optimal average ISI is
as short as you can make it.
116
BOLD response - varying ISI
BOLD response
Stimulus
Blocked trial
Event-related random ISI
Event-related constant ISI
117
fCNR vs. Average ISI
118
Conclusions
The fMRI signal is able to be calibrated.
Physiologic, neuronal, and pulse sequence
calibration techniques are just starting to
develop to complement pulse sequence
advances. -spatial resolution lt 0.5 mm -temporal
resolution lt 100 ms -information content
quantitative flow, CMRO2... A large amount of
additional information exists in the fMRI signal
(i.e. fluctuations..). To aid the development of
calibration, more work needs to be done using
extremely well understood neuronal activation
(across several temporal, spatial, and intensity
scales) to better characterize of the fMRI signal.
119
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design.
120
Neuronal Activation Input Strategies
1. Block Design 2. Frequency Encoding 3. Phase
Encoding 4. Single Event 5. Orthogonal Block
Design 6. Free behavior Design
121

• Free behavior Design
• Use the following as reference functions
• Skin Conductance
• EEG
• Eye tracking
• Task performance
• Heart rate
• Respiration rate

122
Pulse sequences
Paradigms
Processing
Basic
Parametric manipulation
Shimming
Phase and freq. encoding
Contrast comparisons
Orthogonal multi-task encoding
RF coil arrays
Physiologic fluctuations
Physiologic manipulations
Embedded contrast
Event - related fMRI
Motion correction
Distortion / dropout correction
Real time fMRI
Perfusion quantitation
Effective connectivity mapping
lt- Multi - modal integration -gt
lt- Sub - second resolution -gt
lt- Sub - millimeter resolution -gt
lt- CMRO2 mapping -gt
Advanced
123
1991-1992
1992-1999
124
(No Transcript)