The Basis of the BOLD Signal
Paul Forbes & Camilla Nord
Outline
• Physics (Paul)
• Physiology (Camilla)
Physics of the BOLD signal
BOLD and NMR
• To understand the basis of the BOLD signal we
first need to understand Nuclear Magnetic
Resonance (NMR)
Nuclear Magnetic Resonance
• Nuclear
• Magnetic
• Resonance
Nuclear Magnetic Resonance
• Nuclear
• Magnetic
• Resonance
Human body
Fat and water (tissue)
Hydrogen atoms
Hydrogen nucleus
Single proton
SPIN
Nuclear Magnetic Resonance
• Nuclear
• Magnetic
• Resonance
What is spin?
• A magnetic quantity that gives the proton a small magnetic field
around it
• Thus, the proton produces a nuclear magnetic resonance (NMR)
signal
• We have lots of spins so we can describe them with classical
mechanics (not quantum mechanics - see Ehrenfest theorem)
Alignment of spin
• In the absence of an external magnetic field the nuclei will point
in random directions
• Upon the application of a field the nuclei align either parallel of
anti-parallel with the fields
B0
Alignment of spin
• Some align “spin up” and some align “spin down”
• Spin up = high energy state / unstable
• Spin down = low energy state / stable
Precession
• Protons precess when aligned in a field
• The spin vector rotates around the direction of
the external field
Angular frequency at
which the nuclei
precess = Larmor
frequency,
Larmor Frequency
External magnetic
field strength
gyromagnetic ratio
The frequency of
precession (in Hz)
depends on the
magnetic field
strength
Nuclear Magnetic Resonance
• Nuclear
• Magnetic
• Resonance
Nuclear Magnetic Resonance
• Nuclei are aligned and precessing
• Pulse of electromagnetic radiation
is applied - radio frequency pulse
• The frequency of this pulse exactly
matches the frequency of precession
(Larmor frequency)
Nuclear Magnetic Resonance
• Nuclei in the low energy state absorb this
energy and flip to the high energy state (“spin
up”) – phase transition
• They also precess in phase with
each other
• Resonance effect
Nuclear Magnetic Resonance
• The nuclei precessing in phase is fundamental
to MRI signal
B0
Btranverse
as it sets up a
tranverse magnetic field
which forms the basis
of the MRI signal
Physics to physiology
- so why is this significant in terms the BOLD signal?
Relaxation
- In terms of fMRI we are interested in the
apparent relaxation time T2*
- This occurs when the spins dephase
- We are interested in the rate at which the spins
dephase (desync)
The BOLD signal
• oxyHb is diamagnetic = slow dephasing / T2*
• deoxyHb is paramagnetic = fast dephasing / T2*
1. BOLD Contrast
(Nick Todd)
Oxygenated Hb
B0
voxel
Vessel
Tissue
Deoxygenated Hb
21
21
1. BOLD Contrast
(Nick Todd)
Oxygenated Hb
B0
voxel
Vessel
Tissue
Deoxygenated Hb
22
22
1. BOLD Contrast
(Nick Todd)
Oxygenated Hb
B0
voxel
Vessel
Tissue
Deoxygenated Hb
23
23
1. BOLD Contrast
(Nick Todd)
Oxygenated Hb
B0
voxel
Vessel
Oxygenated Hb
Tissue
BOLD
Contrast
Deoxygenated Hb
Deoxygenated Hb
24
24
Physiology of the BOLD signal
Key determinants of the BOLD signal
Arthurs & Boniface, 2002
Questions
Basis of BOLD signal: action potentials or LFPs?
Implications of neurovascular coupling, capillaries vs veins, and
other physiological factors on interpretation of BOLD signal
Interpretations: How is inhibitory neuronal activity reflected?
And how does one interpret a negative BOLD signal?
What’s in a voxel?
Vascular basis of BOLD signal
➢ = Neural activity-dependent changes in
relative concentration of oxygenated and
deoxygenated blood
○ Deoxyhaemoglobin is paramagnetic
(as decreases, causes increase in
BOLD signal), unlike oxygenated
haemoglobin
○ Paradox? → CBF
○ Decoupling more O2 provided
than neccessary
➢ Varies widely across cortical areas; between
species. What if it varied between the groups you
were testing?
Additional physiological factors
○
○
○
○
blood volume
vascular geometry
haematocrit
basal oxygenation levels
○ ‘brain vs vein debate: capillary vs
large draining veins downstream
from neuronally active regions
■ could displace signal
changes Spin-echo fMRI
technique minimizes, but
decreased StN ratio
Relative contribution to T2* weighted GRE fMRI from
capillaries vs. larger blood vessels (veins) as a function of the
frequency on the vessel wall for a cylinder mimicking blood
vessels. (Ugurbil, 2011)
Does the BOLD signal reflect energy
demands?
• Originally thought that increase in blood flow due to energy
requirements
• But most of energy used in post-synaptic signalling;
maintaining APs
• So…energy use does not directly increase blood flow
• Increase in glucose and O2 demands causes release of
vasodilators
• Adenosine
• Arachidonic Acid metabolites
• Nitric Oxide (NO)
• Global blood flow changes also associated with DA, NA, 5-HT
(Attwell & Iadecola, 2002)
Does the BOLD signal reflect firing rate?
fMRI responses in human V1 are proportional to average
firing rates in monkey V1.
(Rees et al, 2000; Heeger at al, 2000)
➢ V5 neurons- directionspecific
➢ fMRI responses in MT+ also
increase linearly with
motion coherence (field of
moving dots)
➢ Proportional to the slopes of
single-neuron firing rates in
monkeys
Or does it reflect another neural property?
➢i.e., APs or LFPs?
➢ Logothetis et al 2001:
Simultaneous intracortical
recordings of neural signals
& fMRI responses
➢ Largest magnitude changes
in LFP correlated with
haemodynamic response
BOLD signal reflects input and intracortical
processing.
BOLD signal/AP correlation
depends on location!
• Rees/Heeger findings resulted from recordings in
areas where inputs, intracortical activity, and
outputs (recordings) are all highly correlated, so all
increase with stimulus contrast
•
Would not have worked in other areas, such as…
Cerebellar APs and the BOLD signal
➢ GABA-A receptor blockade in
rats
➢ Increase in spiking activity
(Purkinje cells)
○ caused no change in
neurovascular coupling that
forms basis of BOLD signal
○ || fibre stimulation causes
increase in BOLD, inhibit
Purkinje cells
Effect of picrotoxin on evoked cerebellar blood flow responses,
local field potentials, synaptic activity, and neurovascular
coupling during inferior olive stimulation (Thomsen et al, 2004)
One possible explanation: Ca2+
a. Excitation increases Ca2+
b. Inhibition decreases Ca2+
c. …but not always, depending on
inhibitory interneuron and its
complement of enzymes &
vasoactive substances, e.g.
cerebellum (INs increase blood
flow because of high NO content)
Why does this matter?
• Implications for group
differences in patient studies (as
well as differences between
neural regions, development
ages, etc.- anything you can
think of)
• Concerns over interpretation of
BOLD response in autism
• Alterations in signalling
pathways (eg, NO),
neurotransmitter differences
• potential for nonlinear shifts in
BOLD response
• Signalling changes- changes in
BOLD w/o neuronal changes
Reynell & Harris 2013
Interim summary
➢ fMRI BOLD signals are dependent on variability and interrelationships of several factors.
➢ This includes multiple cellular mechanisms that could affect vasodilation
other than direct neuronal activity, and is relevent for studies examining
group differences
➢ Debate currently favours relatively direct correlation
between fMRI signals and population synaptic activity
(including inhibitory and excitatory activity) with a secondary
and potentially more variable correlation with cellular action
potentials.
What about inhibitory activity?
Majority of cortex = pyramidal cells
Hypothesis: inhibitory cells do not
produce change in BOLD because lower
metabolic demand from increased
efficiency
Figure: Left column, primary motor cortex, ‘no go
condition’ show no change from baseline, comp to preSMA. (Waldfogel et al, 2000).
• Recent model: inhibition increases
BOLD if low level of excitation, but reduces
when excitation high- can explain
contradictory results (Tagamets & Horowitz, 2001)
•
More complex than this?
Is inhibition just reflected as a
negative BOLD signal?
• Unknown.
• Motor and visual cortices show NBR as
a result of neuronal inhibition
• Many claim that NBR = inhibitory
activity
• Monkey V1 show NBR
corresponds with decreased
population spiking (comb
fMRI/electrophys)
• Cannot draw conclusions without
understanding intrinsic correlation
• BUT is there another
explanation? Example of
seizures.
• Cortex vs subcortical
dissocation?
Schridde et al, 2008
Intrinsic connectivity
What is the majority of the brain’s energy
devoted to?
NOISE (spontaneous fluctuations in the
BOLD signal)
 Synchrony/coherence with
anatomically and functionally
connected regions
 Potential for clinical biomarkers
 Techniques: seed-based (AD
hippocampal connectivity); ICA
(MDD subgenual cingulate).
•
Zhang & Raichle, 2010
The ‘Default Mode Network’
•
Seminal meta-analysis suggested a baseline state of the
human brain using PET (Raichle et al 2001)
• Consistent reductions from this baseline observed in
goal-directed behaviour (PET; fMRI). Temporary
suspension?
• Shown as a negative bold response (NBR)
• Has spurred an entire field of research into rs-fMRI
• A priori studies vs. model-based/data-driven
Designing an fMRI experiment
• Because fMRI does not measure neural
activity directly, we must quantify relative
changes in activity
• A typical fMRI experiment employs a task to
examine differences in the BOLD signal (eg
event-related) vs baseline (rest)
– Timing of BOLD signal: intertime interval &
jitter
• What if there is no task?
Interpreting DMN results
•Grecius et al, 2007
•Increased DMN functional
connectivity with subgenual
cingulate, thalamus,
precuneus in MDD
•Interpretation: Treatment
refractoriness, ‘biomarker’
Carhart-Harris et al, 2012
Cowdrey et al, 2012
Increased DMN-task
positive network coupling
(normally orthogonal)
after psilocybin
Increased DMN functional
connectivity in anorexia
nervosa
Blue= recovered AN group
average Green= healthy
control group
Interpretation: ‘disturbed
ego boundaries’, with
implications for
schizophrenia
Interpretation: rumination
(precuneus) and inhibition
(dlPFC)
Using the BOLD signal
advantageously
•
87% studies used
conventional gradientecho echoplanar imaging
(rest, spin-echo)
• can lead to localization
errors without
optimised protocol
(e.g. pulse sequences)
• Spatial specificity
improves with
increasing magnetic
field strength
•
•
Lawson et al, 2012
Higher-resolution sequences
available
• Requires extensive
collaboration with MR
physicists
Do we need this for all studies?
• “reading a newspaper with a
microscope”
Logothetis, 2008
Next time you read an fMRI study
•
•
•
•
Are there other physiological effects that
could have produced their result?
• Vasodilation differences in region;
disease state
What was their fMRI protocol? Was it
high-res? If no, should it have been?
Do they report a negative BOLD signal?
What is their interpretation? Does it rely
on APs or LFPs?
• Depends on their region of interest!
•
Example critique: Conner et al (2013): Social
proximity to a caregiver during a potentially
stressful situation attenuated activity in the
hypothalamus, VMPFC, and VLPFC (left) in
clinically anxious youths.
Thank you
Many thanks to Nikolaus Weiskopf for advice,
and to you for your attention!
Any questions?