Declaration of Conflict of Interest or Relationship

1
Declaration of Conflict of Interest or
Relationship

• Speaker Name Thomas Liu
• I have no conflicts of interest to disclose with
  regard to the subject matter of
• this presentation.

2
Thomas LiuUniversity of California, San
DiegoApril 22, 2009
Quantitative Functional MRI
3
Topics

• Sources of Variability and Confounds
• BOLD signal model
• Calibration Methods
• Normalization Methods

4
Applications of Functional MRI
http//defiant.ssc.uwo.ca/Jody_web/fmri4dummies.ht
m
5
Interpreting BOLD

• Most fMRI studies assume that the BOLD signal is
  is proportional to brain activity.
• This is a reasonable assumption for basic studies
  of healthy young unmedicated subjects.
• However, the assumption is less valid for studies
  where disease, medication, and age may be a
  factor.

Boynton et al, 1996
6
BOLD Signal Chain
Ianetti and Wise, MRI, 2007
7
Variability in BOLD amplitude
Data courtesy of J. Liau
8
Effect of Age
DEsposito et al 2003
9
Effects of Vascular Disease
DEsposito et al 2003
Iadecola et al 2003
10
Effects of Alzheimers Disease Risk
Fleisher et al 2008
11
Carbon Dioxide
Cohen et al 2002
12
Caffeine
Liu et al 2004
13
Interpreting BOLD
Ianetti and Wise, MRI, 2007
14
Topics

• Sources of Variability and Confounds
• BOLD signal model
• Calibration Methods
• Normalization Methods

15
BOLD Contrast
Source Ogawa et al., 1992
16
BOLD Signal Change
17
Hemoglobin and Field Inhomogeneities
Oxygen binds to the iron atoms to form
oxyhemoglobin HbO2 Release of O2 to tissue
results in deoxyhemoglobin dHBO2
http//www.people.virginia.edu/rjh9u/hemoglob.htm
l
18
Signal Decay
19
BOLD Signal Equation
Simulations suggest ? ?1.5 is a reasonable value
Ogawa et al, 1993 Boxerman et al 1995, Hoge et
al. 1999
20
Blood Flow and Oxygen Metabolism
Cerebral Blood Flow (CBF) measures delivery of
blood to brain tissue (units of ml/(g-min))
Cerebral Metabolic Rate of (CMRO2) is the rate of
oxygen consumption (units of ?mol/(g-min))
CMRO2
21
Deoxyhemoglobin
CMRO2 / 4CBF
22
fMRI Spatial Temporal Dynamics
arteriole
venule
capillary bed
CBF
oxyHb
deoxyHb
CMRO2
Neural activity
Positive BOLD
Post-stimulus Response
Initial dip
23
BOLD Signal Path
Neural Activity
24
BOLD Signal Equation (Davis Model)
25
Davis Model
Hoge et al. 1999
26
Topics

• Sources of Variability and Confounds
• BOLD signal model
• Calibration Methods
• Normalization Methods

27
Motivation for CMRO2 Measures
Smith et al, PNAS 2002
28
Neural Activity
29
Calibrated fMRI (Davis et al 1998)
30
Experimental Protocol
31
Calibrated fMRI
Ances et al, NIMG 2007
32
Calibrated fMRI of HIV
Ances et al, in preparation
33
Effect of age on CBF and BOLD
Restom et al, NIMG 2007
34
Potential Issues with Calibrated fMRI

• Xu et al, ISMRM 2009, Abstract 215

Chen and Pike, ISMRM 2009, Abstract 214
35
Calibrated fMRI

• Can provide insights that BOLD measures alone
  cannot provide.
• Can be difficult to apply to cognitive tasks and
  special populations, due to low sensitivity of
  ASL CBF measures.
• The need for breathhold or hypercapnia can also
  be an issue. Hyperoxia-based methods may be an
  alternative.
• Methods that utilize measures of CBV changes
  (e.g. VASO) (e.g. Donahue et al, Abstract 14) or
  regional measures of changes in venous
  oxygenation (e.g. Bolar et al, Abstract 3659) are
  also being explored. These have the advantage of
  not requiring hypercapnia.

36
Topics

• Sources of Variability and Confounds
• BOLD signal model
• Calibration Methods
• Normalization Methods

37
Variability in BOLD amplitude
Data courtesy of J. Liau
38
Neural Activity
39
Arterial spin labeling (ASL)
40

• Whole brain CBF Images from 1 subject scanned at
  each of the 4 sites are shown below. Grayscale
  bar indicates units of ml/(100g-min).

41
ASL Time Series
Perfusion Images
42
BOLD variability and Baseline CBF
Liau and Liu, NIMG 2009
43
Measures of Venous Oxygenation (O2,v)

• T2-Relaxation-Under-Spin-Tagging (TRUST) MRI (Lu
  and Ge, MRM 2008)

1
Apply T2 Prep
2
Apply T2 Prep
44
?T2 weighting
eTE160ms
eTE40ms
eTE80ms
eTE0ms
Control
Tag
Difference
O2,v68.2ms
T268.2ms
45
Venous Oxygenation (TRUST MRI)
Liau et al, ISMRM 2009, Poster 1635
46
TRUST MRI
Lu et al, Abstract 855, ISMRM 2008
47
Breathold/Hypercapnic Normalization
48
Breathold/Hypercapnic Normalization
Liau and Liu, NIMG 2009
49
Breathhold Calibration
Breathhold Signal
Thomason et al, 2007
50
Breathhold Calibration
Thomason et al, 2007
51
Scaling with Resting-State Fluctuations
Wise et al 2004
52
Scaling with Resting-State Fluctuations
Birn et al 2006
Kannapurti and Biswal 2008
53
Summary of Relations
Lu et al, ISMRM 2009, Abstract 218
54
BOLD Dependence on Baseline State
BOLD
55
Davis Model
Davis et al, 1998 Hoge et al. 1999
56
Factors Driving BOLD Variability

• Decrease in CMRO2 with CBF0 would tend to lead to
  a larger BOLD response.
• M appears to be independent of CBF -- effects of
  increased CBF and decreased dHb appear to
  cancel out
• Variability in the BOLD response across subjects
  appears to be driven by inter-subject differences
  in the CBF response.

57
Dependence of CBF changes on Baseline CBF

• Hypercapnic CBF response also inversely
  proportional to CBF
• Vasculature may be constructed to conserve
  absolute change in CBF

58
BOLD Normalization
Is all the variability really vascular?
59
BOLD Dependence on Resting-State Alpha Power
Koch et al 2008
60
BOLD Dependence on Resting-State Alpha Power
IAF/Hz
Koch et al 2008
61
Stimulus-related CBF and BOLD
Neural Response
Neurovascular Coupling
Stimulus
Resting-state Fluctuations
CMRO2 Estimates
Baseline Vascular State
Resting-state Neural Activity
Baseline CBF Venous O2 Hypercapnic Reponses
?

• Across Subjects Higher Alpha Power --gt Lower
  CBF
• Instead of a vascular mechanism, higher CBF
  changes observed at lower baseline CBF may be
  driven by true increases in neural activity.
• Suggests that normalization methods need to be
  treated with caution.
• Relationship may not be so simple with medication
  and disease.

62
Summary of Normalization Approaches

• Additional measures can account for BOLD
  variability due to variations in baseline blood
  flow, volume, oxygenation, field strength, etc.
• Measures of venous oxygenation, cerebral blood
  flow, and resting fluctuations have the advantage
  of not requiring the subject to perform
  additional tasks.
• These approaches offer some insight into the
  factors that can affect the BOLD response between
  subjects, groups, and conditions.
• However, it is possible that these approaches may
  also remove variability related to true
  differences in neural activity.

63
Conclusions

• The BOLD signal is a complex function of the
  baseline state and changes in blood flow, volume,
  and metabolism.
• Differences in the BOLD signal do NOT always
  reflect differences in neural activity.
• Instead they make reflect differences in the
  baseline vascular or metabolic state.
• Calibrated fMRI can provide additional insights
  into differences in brain activity, especially in
  the presence of disease, medication, and age.
  However, it is a technically challenging method
  and may be difficult to apply in certain
  populations.
• Normalization methods can help to reduce
  inter-subject variability, but need to be treated
  with caution.

64
Acknowledgements

• Beau Ances
• Yashar Behzadi
• Rick Buxton
• Joy Liau
• Oleg Leontiev
• Joanna Perthen
• Khaled Restom
• Eric Wong