This document provides an overview of functional magnetic resonance imaging (fMRI) and the blood oxygen level dependent (BOLD) contrast mechanism. It discusses how BOLD fMRI works by measuring changes in oxygenated blood flow and volume in the brain during neural activation. The history and key discoveries are summarized, including how deoxyhemoglobin causes local magnetic field distortions that reduce the MRI signal. Experimental paradigms and considerations for optimizing BOLD pulse sequences and analysis are also briefly outlined.

1. BASIC OF FUNCTIONAL
MAGNETIC RESONANCE
IMAGING-FMRI
REMIX
MAHARJAN
Bsc.MIT 4th year

2. Outline
 Introduction.
 History of fMRI
 How does fMRI work.
 BOLD Contrast Mechanism.
 BOLD Signal.
 BOLD and Brain Activity.
 fMRI Paradigm Designs.
 On-off Paradigm
 BOLD pulse Sequences and parameter choice.
 fMRI of Visual and Sensorimotor cortex
 Phonologic and Semantic Paradigms and language
Testing.

3. Introduction
 The Oxford physiologist Charles Sherrington made the
observation that when a small area of exposed cat brain
was stimulated electrically there is a flush of red blood.
 This increases the blood supply to the brain locally,
ensuring an adequate supply of oxygen to regions
working harder in thinking.
 So, if we could measure the amount of oxygenated
blood in a specific area we could have a measure of the
underlined neural activity at this area.
 Blood has a lot of water molecules and water molecules
have a lot of hydrogen atoms that we can measure
using MRI.
 But what about oxygenation ? Oxygenation can be

4. History.
 Belliveau et al (1991) used susceptibility changes from
gadolinium bolus to show increased blood volume in the
human visual cortex during photic stimulation.
 This method is known as Dynamic susceptibility contrast
(DSC) MRI.
John (Jack) Belliveau 1959-2014

5. Seiji Ogawa, pioneer of BOLD fMRI, c. 1990
 Ogawa et al (1992) demonstrated the same phenomenon using
BOLD (Blood Oxygen Level Dependent) MRI.
 Based on physical principles of nuclear magnetic resonance
(NMR) and the intrinsic effects of blood oxygenation on the MR
signal due to the magnetic properties of Oxyhemoglobin and
Deoxyhemoglobin.
 Even a brief stimulus elicits a strong blood flow change that

6. How does fMRI works
 Actually BOLD does not measure neuronal activity, but
measures the amount of oxygenated blood in a specific
area.(Changes in regional blood flow).
 So fMRI demonstrates brain activation by recording T2(*)
signal changes due to increased regional blood flow.
 BOLD contrast depends on local ratio of oxy- and deoxy-
hemoglobin, thus affecting T2(*) signal intensity.
 EPI with GRE sequences obtained while patient performs task
(e.g. Finger tapping) and at rest, the processed to generate
activation maps.

9. Magnetic susceptibility:
 Susceptibility (χ): degree of magnetization of a
material in response to an applied magnetic field
 Ferromagnetism: – Large, positive magnetic
susceptibility, – Attracted to external magnetic fields –
Exhibit magnetism even when field is removed
 Diamagnetism: – Negative susceptibility – All
electrons are paired – Slightly repelled by magnetic field
– Very weak
 Paramagnetism: – Small, positive susceptibility to
magnetic fields – Unpaired electrons – Slightly attracted
to external field – Slightly stronger magnetic properties
than diamagnetic materials in presence of external field

10. BOLD Contrast Mechanism
Deoxy-hemoglobin is
strongly paramagnetic due to
4 unpaired electrons at each
iron center.

11.  Presence of Deoxy Hb creates
local magnetic field distortions
in and around the vessels.
 These causes nearby stationary
and slowly moving spins to
have different resonance
frequencies and phase shift
resulting intra-voxel dephasing-
classic T2* shortening effect,
most prominent near larger vein
and emphasized by use of GRE
sequence with TE(s) close to
T2*.
 This effect is the dominant
mechanism for BOLD contrast
Paramagnetic
deoxyhemoglobin (D)
confined to red blood cells
causes a local field distortion
in and around the vessel.

12.  Local magnetic field distortions produced by intravascular
deoxy Hb also affect protons in water molecules diffusing in
and around these vessels.
 Such protons experience randomly changing frequency
offsets and undergo unrecoverable dephasing.
 This diffusion-related T2-signal loss is best appreciated
using spin echo techniques and is more prominent adjacent
to capillaries (than near larger vessels).
 This consitute the dominant mechanism for BOLD contrast
at 4.0T and higher. (At 3.0T, where most clinical fMRI
studies take place, the T2 and T2* effects make comparable
contributions to BOLD contrast.)

13. Why T2* and not T1?
 MRI scanner is tuned to resonate and image hydrogen
atoms as in conventional MRI; however, T2*-weighted
images are performed which take advantage of the fact that
deoxy-hemoglobin is paramagnetic whereas oxy-
hemoglobin dimagnetic.
 Because of the magnetic properties of the unflipped
magnetic deoxy-hemoglobin molecule which causes rapid
dephasing, T2* signal is retained longer in a region when it
has more oxygenated blood. Thus, an area with more
oxygenated blood will show up more intense on T2*-
weighted images compared to when there is less
oxygenated blood around.

14. BOLD signal
 Cerebral activation causes increased regional
blood flow exceeding the brain’s immediate
metabolic demands (hemodynamic
“uncoupling”).
 “Overshoot” of oxygenated bloods (increase in
ratio of Oxy: Deoxy hemoglobin ), resulting in
increased BOLD signal.

16. BOLD and Brain Activity
 Increased neuronal activity induces changes in regional CBF,
CBV, and oxygen extraction(CMRO2) through a process
known as neurovascular coupling.
 BOLD signal best correlates with extracellular local field
potentials (LFPs) which reflects the total activity of regional
neural networks including neuronal discharges and sum of
+ve and –ve post-synaptic potentials at multiple dendritic
connections.
 The change in the MR signal (regional BOLD response
generated) from neuronal activity (stimuli such as finger
tapping) is called the hemodynamic response Function
(HRF).
 HRF typically demonstrates a small initial dip, followed by tall

17.  Initial dip:
 Occurs due to Increased
early metabolic extraction
of blood oxygen
 Increased local cerebral
blood volume.
 Fast and transient – Need
high temporal resolution
and good SNR – observed
at very High field(≥7.0T)
BOLD Hemodynamic Response Function (HRF)
following a single brief stimulus

18.  Positive dominant
peak: constitutes the bulk
of the BOLD response
 With Multiple repeated
stimuli the dominant peak
becomes a broad plateau,
not dropping off until the
stimulation ends.
 Post-stimulus
undershoot:
 Occurs due to slow recovery
of arterial blood volume and
decrease in regional CBF.
 Also continuous metabolic

20.  neural system provides feedback to the vascular system of
its need for more glucose is partly the release
of glutamate as part of neuron firing
 glutamate affects nearby supporting cells, astrocytes,
causing a change in calcium ion concentration. This, in
turn, releases nitric oxide at the contact point of astrocytes
and arterioles.
 Nitric oxide is a vasodilator causing arterioles to expand
and draw in more blood.

21. fMRI paradigm Design
Paradigm Design of BOLD/fMRI experiments

22.  Block design (“Boxcar”):
 Task periods alternated with periods of rest.
 Low BOLD signal are digitally “subtracted” from higher
BOLD signal to reveal focal areas of cortical activation.
 Best for simple experiments and pre-surgical mapping.
 Highest SNR, statistical power and time efficiency.
 Limitation: blocks are measured over relatively long periods
(10-20 sec), info about hemodynamic response and fMRI
signal timing are difficult to measure.

23. Finger tapping fMRI study using simple block design comparing
signals during activity and at rest. Because the "on-off" pattern of
activation resembles the passing of a train, the paradigm is often
referred to as a "boxcar" design.

24.  Physiological noise:
 Contributions from scanner instability, heating, etc.
 Contributions from respiratory rate, cardiac rate, coherent
synchronous hemodynamic activity
 To eliminate the noise, fMRI studies repeat a stimulus
presentation multiple times

25.  Event-related Designs:
 Allows single or multiple tasks and stimuli to take place at
short and variable time intervals.
 High degree of flexibility for neuropsychological experiments.
 randomized and different types of events can be mixed.
 Allows better temporal resolution and estimation of HRF time
course.
 Limitation:
 Low SNR, statistical power, with longer imaging times and
more trials per subject required.
 Data analysis is more complex and dependent on accurate
modeling of HRF.

29.  Mixed Design:
 Combined features of blocked and event-related
designs.
 Semi-randomized events take place during the task
blocks, with rest periods in between.
 Preserve favorable SNR with flexibility of event-related
ones.

30. On-off paradigms
 Raw MR BOLD signal during activation is a relative, not an
absolute quantity/units which is affected by numerous technical
and patient-specific factors.
 Technical factors: field strength, amplifier gain, no. and
location of head coil receiver elements, pulse sequence( SE or
GRE), slice acquisition order, sequence timing parameters (
TR/TE/α), and voxel size.
 Patient-related factors: Hematocrit, respiratory rate, head
size, age, gender, hormonal status, and medications (including
caffeine).
 BOLD signal expressed in arbitrary units (A.U.s) or percent
change from baseline.

31.  fMRI-BOLD is best used for studying processes that can
be rapidly turned on and off like language, vision,
movement, hearing, and memory.
 The study of emotion is hampered by its slow and
variable onset and its inability to be quickly reversed.

32. BOLD pulse sequences
 A BOLD pulse sequence should have the following
characteristics:
1) sensitivity to T2 and/or T2* changes
2) ability to detect the intrinsically low BOLD signal, often just a
few percent different than baseline
3) sufficient spatial and temporal resolution to cover the entire
brain at multiple closely spaced time points.
 At 3.0T and below: T2*- weighted GRE sequence are commonly
used.
 At 7.0T and higher: T2-weighted spin echo (SE) techniques are
generally preferred.
 Most common technique is echo-planar imaging(EPI), 2D-
multislice, T2*-weighted GRE sequence
 The hemodynamic response to neural activation occurs over
fractions of seconds – EPI can acquire whole-brain volumes in a

34.  Ultra Fast Methods for BOLD fMRI:
• Half-Fourier
• SENSE (Sensitive Encoding)
• SMASH (simultaneous acquisition of spatial
harmonics)
• UNFOLD

35. Parameter choice
 Plane of imaging: parallel to the anterior commissure-
posterior commissure (AC-PC) line with whole-brain
coverage.
 Slice thickness: trade of between low SNR(thin) Vs Partial
volume averaging(too thick), usually 2-4mm
 Echo time (TE): TE ≈ 30−35 ms. ( At 3.0T) ,At longer TE’s
more susceptibility artifacts and signal dropout on GRE-EPI
images.
 Repetition time (TR). Should be less than HRF time course.
(1−​4 sec), Short TR's (≤ 1.5 s) provide better estimation of
the HRF and more statistical power.
 Matrix size: Typically 64x64 − 128x128 (or 2x2 − 3x3 mm in-
plane resolution)

36.  Slice order. Interleaved acquisition (1,3,5,...2,4,6..)
generally selected to reduce slice cross-talk artifacts.
 Total imaging time. For maximal subject compliance
overall imaging time should not exceed the 45-60 minute
range, with no more than 10-12 minutes per individual
experiment.
 Parallel imaging. Generally advised to decrease
acquisition time, increase temporal resolution, and reduce
susceptibility artifacts

37. Advance techniques and
options:
 Distortion Correction. Pre-acquisition field-mapping to
minimize susceptibility-induced spatial distortions in fMRI, A
quick shimming procedure to improve homogeneity, especially
used in non-research centers.
 3D Methods: Reduction in image dropout and contiguous
acquisition without gaps or need to perform slice-time
correction. Due to prolonged TR and readout time it is difficult
to perform 3D, so segmented 3D methods with parallel
imaging in 2-D.
 Multiplexed Acquisition: Use of "multi-band" techniques to
excite several (2-3) slices simultaneously without SNR penalty.
 Other Non-GRE Pulse Sequences. Techniques using fast
spin echo (FSE/TSE) offers high SNR and reduced geometric
distortion, for ultra-high fields(≥7.0T), SE and SSFP methods
are used.

39. fMRI of Sensorimotor Cortex
 Block design is typically used.
 Up to three different anatomic areas are typically interrogated as
part of a complete motor fMRI study: hand, foot, and mouth.
 Hand functions are evaluated using fist clenching or ball
squeezing(partially dysfunctional hand) or tapping of individual
fingers
 Lower extremity motor functions are commonly tested by
toe/ankle/foot dorsiflexion
 face/mouth functions are evaluated using lip puckering and/or
tongue wiggling maneuvers
 For patients with complete paralysis pure sensory stimulation of
the hand (by stroking or brushing) may elicit fMRI activation of
both the somatosensory and primary motor cortices.

41. BOLD-fMRI maps obtained during performance of right-sided finger tapping.
The contralateral (left-side) primary sensorimotor cortex is most strongly
activated. Also note bilateral activation of the supplementary motor area
(green arrow) and ipsilateral (right-side) superior cerebellum (right image).

42. fMRI of Visual System/cortex
Pathways from the retina to the primary visual cortex showing crossed and uncrossed
projections

43.  most complex integration of visual information occurs in the
extrastriate cortex, visual association areas denoted V2−V6
located in the adjacent occipital, temporal, and parietal lobes.
 Ventral stream is responsible for object recognition and the
storage of long-term visual memories.
 Dorsal stream associated with the spatial location of objects
and perception of motion.

44. Primary visual cortex at occipital pole (gray) is connected to posterior parietal
(green) and inferior temporal (purple) association regions via the dorsal and
ventral streams respectively

45. BOLD/fMRI of
visual cortex by
checkerboard
pattern. Also note
activation of lateral
geniculate nucleus
(green arrow)

46. Language Testing
 A minimum of one semantic and one phonological
paradigm should be employed, and the best centers
typically employ 2-3 of each.

47. "Classic" and "Dual
Stream" models of
language processing.
Classic model is centered
on Wernicke's (involved in
the reception and
comprehension of
language,) and Broca's
(involved in expression/
articulation of speech,)
areas. "Dual stream" model
incorporates these areas
into a ventral stream (red)
serving comprehension and
a dorsal stream (blue) for
articulation.

48. Semantic Paradigms
 designed to elicit activation of Wernicke's area as well
as other portions of the ventral stream involved in
decoding language and establishing meaning.
 Typical semantic tasks used in fMRI studies
include sentence completion ("I drive to work in my
____"), true/false statements ("Is the displayed
statement true or false?"), reading
comprehension (whole paragraphs with questions),
and listening comprehension (spoken language with
questions vs garbled speech)

49. Sentence completion paradigm. Note strong activation of Wernicke's
area (A), but also other portions of the superior and anterior left
temporal lobe (B,C).
(A) (B) (C)

50. Phonologic Paradigms
 to elicit activation of Broca's area and other parts of the
dorsal stream involved in the production of sounds and
articulation of language.
 Involved silent word generation. The subject is shown a letter
(e.g., "B") and over the next 5-10 seconds is asked to silently
think of as many words as possible starting with that letter
("baby", "bed", etc.) During the control/rest periods the
subject is shown nonsense symbols (e.g., "✜​", "⌘").

51. Silent word generation task producing activation of left prefrontal
cortex (A);
bilateral Broca's areas, left greater than right (B); and right cerebellum
(C)
(A) (B) (C)

52. Non-BOLD Methods
 Perfusion fMRI
• measures regional cerebral blood flow –Intravenous
Bolus-Tracking fMRI(T2*)
• Injection of a magnetic compound (gadolinium-DTPA) .
Areas perfused with the magnetic compound show less
signal intensity as the compound creates a magnetic
inhomogeneity that decreases the T2* signal.
•Belliveau et al. (1991) first functional magnetic resonance
maps of human task activation using a visual stimulation
paradigm.

53.  Arterial Spin-Labelling(T1)
• Magnetic tagging of hydrogen atoms as they course through
the blood and imaging them as they course through the slice
of interest.
• Uses RF-pulse to “tag” spins in a slice
• Used to generate quantified resting blood flow maps and
perform functional experiments
•Acquire each slice twice: one tagged and one untagged
• Subtract tagged from untagged to get slice with perfusion
• Using models of flow it is possible to quantify perfusion to
obtain regional cerebral blood flow (rCBF)

55. Radio Frequency Pulse to Tag Protons

56. • More suited to measuring state differences between
groups (i.e. bipolar euthymic Versus bipolar depressed)
• It is non-invasive.
 ASL
 Provides better spatial specificity
 Not affected by “draining veins”
 Less susceptible to scanner signal drift (useful for
studies of changes that occur slowly over a long time
scale)
 Takes several minutes to image a single slice.
 BOLD
 Better temporal resolution
 Better spatial resolution

57. fMRI Using ASL vs. BOLD

58. Other non- BOLD methods
 MRI spectroscopy (MRSI): which can measure
certain cerebral metabolites non-invasivelly and study
tissue biochemistry.
 Diffusion-weighted fMRI: which measures random
movement of water molecules.

59. 1) Who invented functional MR imaging (fMRI)?
2) How does fMRI work?
3) How is image contrast produced by BOLD fMRI?
4) How is image contrast produced by BOLD fMRI?
5) Why does the BOLD signal increase during activation?
6) Does the BOLD response result from the firing of nerve cells?
7) How do you design a BOLD/fMRI study?
8) Why do you have to do an "on-off" comparison? Why not just
measure the absolute BOLD signal instead?
9) What is the best pulse sequence to use for BOLD fMRI?
10) What is the best way to identify the sensorimotor and visual
cortex using fMRI?
11) What paradigms do you use to test language function prior to
surgery?