fMRI, BOLD & homodynamic response.
A Multimodal approach.
Functional magnetic resonance
During neuronal activation, metabolism is increased in the form of additional oxygen
consumption by areas of greater neuronal activity, occurring with a delay of
approximately 1-5 sec. after stimulation.
The hemodynamic response to a stimulus results in site specific concentration
differences in oxyhemoglobin and deoxyhemoglobin and changes in cerebral blood
flow (CBF) . As hemoglobin courses through the capillaries of the brain, oxygen is
consumed by means of metabolic processes.
The resulting heterogeneity can be used as an endogenous contrast agent. Blood
Oxygenation Level Dependent (BOLD) contrast.
When hemoglobin is oxygenated it is diamagnetic (non-magnetic), but paramagnetic
(magnetic) when deoxygenated. The paramagnetic species precesses at a different rate
around the local magnetic field (B0).
Because the deoxyhemoglobin does not precess at the same frequency as that of the
local magnetic field (w=γBo) it will act as a t2* contrast agents.
When a neuron, or a group of neurons is activated, cerebral blood flow (CBF) ,
cerebral blood volume (CBV) and oxygen delivery is increased. The sudden flow of
excess oxygenated hemoglobin into specified voxels, the ratio of oxygenated to
deoxygenated hemoglobin is slightly increased compared to that at rest.
As the diamagnetic ratio increases due to changes in CBF and CBV, T2* varies. Faster
dephasing, decreasing T2*
BOLD Temporal Response
The basic sampling unit for temporal resolution is
repetition time (TR)
TR range from 500ms – 4000ms.
The limitations of temporal resolution in fMRI is
based on the vasculature of the brain and not
physical limitations of the imaging methodologies.
After a stimulus is presented the BOLD signal
increases approximately 2 sec later. At around 5
sec. later signals is peaked for a short-duration
stimulus, or maintains a plateau for stimuli of a
After peak the signal undershoots to below
baseline level, remaining so for an extended
period of time.
Therefore increasing TR won’t necessarily give us
increased temporal resolution.
What color and shape was that?
This was presented to you for 50 ms.
In this time the red oval stimulated a cascade of
neural and cognitive processes. You were presented
with a stimulus and you were able to become aware
of it, eventually using descriptive language to convey
what you saw.
What if you were repeatedly presented with a stimulus?
BOLD signal changes have been detected at ~34ms in the
human visual system (Savoy et al., 1995). However as seen
before the hemodynamic response, which BOLD is dependent
is a relatively slow process, lasting up to several seconds for a
This proves to be a problem when studying cognition. If we
assume the BOLD signal to be an accurate correlate of neuronal
activity we are in trouble.
In fact there does seem to be a
refractory period where the
hemodynamic response does not
follow the principle of superposition.
-5 0 5 10 15 20 25 30
Huettel and McCarthy showed that the
BOLD signal response to a second 1.40 0
stimulus in a pair separated by 1 sec 6
Signal Change over Baseline(%)
was reduced by more then 40% and 1.00
delayed by approximately 1 sec. As the 0.80
interval between stimuli increase the 0.60
amplitude and latency of the 0.40
homodynamic response returns to 0.20
normal values. 0.00
0 1 2 3 4 5 6 7 8 9 10 11 12 13
Time since onset of second stimulus (sec)
Huettel & McCarthy, 2000
Where does that leave us…?
Using the hemodynamic response as an
inference to neural activity carries a lot of
weight, however directly measuring neuronal
activity could give us better temporal resolution
and a more accurate measure of signal
What are we measuring?
When an excitatory/inhibitory postsynaptic potential (PSP)
is initiated extracellular electric fields are generated as ion
channels open and allow for the influx of cations (Na+). A
current sink ensues as the inward flow of Na+ through
voltage-sensitive channels causes the cell to depolarize.
PSP’s travel down the axon, conducting impulses at unmyelinated portions (Nodes of Ranvier) of
the axon. At unexcited portions of the neuronal membrane there is an outward flow of K+,
establishing a current source.
The extracellular current flow, from the influx at the sink, and the efflux at the source is called the
volume current. The extracellular electric field produced amid the minute distance between these
two points is referred to as the current dipole. The signal is radially directed outward, the strength
of which decays with the inverse square of the distance.
By continuously recording the fluctuating field potentials, a result of the summation of PSP’s we
are able to more directly measure neuronal activity.
Waveform Activity Types
Delta (≤ 4 Hz). Usually associated
with deep sleep, coma and brain injury.
Theta (4-8 Hz). States of on-line
hippocampus related tasks; working
memory, quiet focus.
Alpha (8-12 Hz). Relaxed, alert state of
Beta (> 12 Hz). Active, busy or
anxious thinking and active
Gamma (26-100 Hz). Higher mental
activity; perception, problem solving.
Through the use of electrodes placed directly on the scalp
(2,3,8,19 or even up to 100) graded post synaptic potentials
(PSP) of the cell body, large dendrites and vertically oriented
pyramidal cells in layers III-V are measurable. (Lopes del Silva,
EEG electrode placement
Each electrode is connected to a differential amplifier, used to amplify
voltages changes over time at each electrode.
The electrode-amplifier relationship is arranged in one of three montages.
1) Common reference derivation:
Each amplifier records the difference between a scalp electrode and a reference electrode. A
single reference electrode is placed somewhere along the midline or attached to both earlobe
2) Average reference derivation:
Activity from all the electrodes are recorded and averaged before being passed through a
high value resistor. The resulting signal is then used as a reference point for each amplifier.
3) Bipolar derivation:
Electrodes are connected in a straight line from the front to the back of the head or
transversely across the head. An amplifier sequentially measures the difference between two
The conductivity of the skull is relatively low,
(1/80 to 1/100) to that of the brains
conductivity (Toga. TW, et al., 2002).
The electrical field is best detected from
electrodes placed directly on the scalps
surface. The magnetic field can be measured
outside of the head because the signal is not
impeded to the same degree as the electrical
The goal is to record EMF’s at varying
locations outside the head and to then
calculate the most probable source currents
within the brain.
The magnetic signals produced by evoked
magnetic fields (similar to ERP’s) are weak,
on the order of 100 femtoTesla (an fT is 10-15
To detect these very weak fields
Superconducting Quantum Interference
Devices (SQUID’s) are used. Current
neurmagnetometers contain more then 300 http://biomagnetism.kriss.re.kr/English/meg_application-e.htm
SQUID’s, operating in liquid helium at 4 K
The SQUID is a magnetic flux-to-voltage or
Find voltages at electrodes given source neuronal activity
The calculation of scalp potentials (EEG) and magnetic
fields near the scalp (MEG) given neuronal currents in the
In estimating the neural source of electromagnetic fields
we employ a forward model that maps a source of known
locations, strength and orientation to an array of sensors.
The Inverse problem.
Find a source of neuronal activity given voltages at electrodes.
Locating the source of field potentials is difficult to determine with certainty. The decay
of electric field strength has varying affects on nearby electrodes, however dipoles will
contribute similarly to more distantly placed electrodes.
A priori source modeling is required.
MEG’s temporal resolution can detect changes occurring in <1ms, while
spatial resolution ranges from a few millimeters up to a couple of centimeters.
MEG recordings in a normal control MEG recordings in Alzheimers Disease
Fig. 1. A, MEG recording (left hemisphere channels) with a coronal magnetic resonance image
imposed on the waveforms shows normal baseline activity in a control subject. B, Similar MEG
recording (left hemisphere channels) with a coronal magnetic resonance image imposed on the
waveforms shows pronounced slow wave activity (5–6 Hz) in a patient with Alzheimer’s disease. White
dots indicate dipole localization of increased slow wave activity in the left temporal lobe. The dipole
localization is consistent with the neuronal destruction patterns associated with Alzheimer’s disease.
The spatiotemporal resolution differs depending on your imaging technique. MEEG
both have good temporal resolution in the order of sub-millisecond, however there
spatial resolution is poor due to physiological impedance and complex source
fMRI has good spatial resolution (2-3 mm), but poor temporal resolution due to
physiological limitations of the homodynamic response, not physical limitations of
A multimodal analysis of neuronal activity may provide us with superior
spatiotemporal resolution. Providing a more direct and causal interpretation of
Recorded EEG activity of interest triggers MRI acquisition. Due to delayed hemodynamic
response (2-4 sec) the EEG can be measured soon after the end of the previous MRI block.
This way there is limited distortion from the RF pulses or gradient fields from the MRI
acquisition block (Goldman et al., 2000).
The attenuation of EEG electrodes is limited, thereby limiting our window of measurement.
Leads are typically constructed of carbon fiber (non-magnetic).
EEG is continuously recorded while fMRI data is acquired in blocks
The EEG is continues.
fMRI data is acquired 1 sec after the delay.
Data is obtained simultaneously.
Artifacts due to fMRI acquisition is unavoidable. Artifact removing methods are
complex and only result in an estimated EEG signal.
Good for researching neuronal activity of cognitively varying tasks.
EEG with no artifact