An Investigation into White Matter Small Vessel Disease Using Susceptibility-Weighted ... · 2015-04-24 · An Investigation into White Matter Small Vessel Disease Using Susceptibility-Weighted
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
An Investigation into White Matter Small Vessel Disease Using Susceptibility-Weighted and Perfusion Magnetic
Resonance Imaging
by
Farhang Jalilian
A thesis submitted in conformity with the requirements for the degree of Master’s of Science
Department of Medical Biophysics University of Toronto
4.2 Prominence of Veins Reflects Pathophysiological Changes in CSVD .............................39
4.3 Perfusion Findings: Evidence and Implications ................................................................41
4.4 From Structural to Physiological Imaging: A Search for New Imaging Markers of CSVD.................................................................................................................................42
4.5 Limitations and Future Work.............................................................................................43
Figure 11: A. Study-specific T1-weighted group template. B. WMHs probability map in the
group template constructed from averaging template acquired subjects’ WMHs masks. C. Group
template acquired NAWM mask is the “relative complement” of the binarized WMHs mask with
respect to the WM (i.e. voxels that are in the WM but do not belong to WMHs mask). ............. 27
Figure 12: ARWMC rating scale. A. ARWMC Score 1; Focal lesions around the horn of lateral
ventricles (arrows). B. ARWMC Score 2; Periventricular lesions are becoming confluent
(arrow), and multiple deep WMHs are present (arrowhead). C. ARWMC Score 3; diffuse
involvement; periventricular WMHs are extending into the deep WM and vice versa................ 29
Figure 13: A. The mean CBF perfusion image B. The binary mask used in the voxel-wise
analysis showing voxels with SNR>1. ......................................................................................... 30
Figure 14: Top: the tissue class ROIs (B) was used to mask the venogram (A). Bottom: Vein
fraction in different tissue ROIs.................................................................................................... 32
Figure 15: Vein fraction vs. WMHs at the level of LV (top) and vs. global WMHs (bottom). C &
D are provided as context in support of the use of WMHs at of the LV. ..................................... 34
Figure 16: Comparison of WMHs, GM CBF, WM vein fraction and vein fraction in NAWM
mask among ARWMC scores....................................................................................................... 36
Figure 17: Voxel-wise analysis results shows a region (11 voxels; shown in blue) with reduced
perfusion. The background grey-scale image is the MNI152 T1 template................................... 37
ix
Figure 18: Association between WMHs and veins. The middle diagram shows the overlay of the
WMHs and the vein schematically. thin arrows=vein; think arrows=WMHs (Adopted from Gao
et al. 2008). ................................................................................................................................... 39
x
List of Abbreviations
3D three dimentional 2D two dimensional AD Alzheimer’s disease ANCOVA analysis of covariance ANOVA analysis of variance ARWMC age-related white matter changes ASL arterial spin labelling BET Brain Extraction Tool BOLD blood oxygen level-dependent
CADASIL cerebral autosomal dominant arteriopathy with subcortical infarcts and leukoencephalopathy
FMRIB Oxford Centre for Functional Magnetic Resonance Imaging of the Brain
FSL FMRIB Software Library GLM general linear model GM gray matter HRBV high-resolution BOLD venographic LV lateral ventricle MCA middle cerebral artery MMD moyamoya disease MNI Montreal Neurological Institute MOCA Montreal cognitive assessment tool MRI magnetic resonance imaging MS multiple sclerosis NAWM normal appearing white matter OEF oxygen extraction fraction PD proton density
xi
PET positron emission tomography PLD post label delay ppm part per million QSM quantitative susceptibility mapping RF radiofrequency ROI region of interest SNR signal-to-noise ratio SPECT single-photon emission computed tomography SVD small vessel disease SWI susceptibility-weighted imaging TBI traumatic brain injury TE echo time TIA transient ischemic attack TR repetition time TRUST T2-Relaxation-Under-Spin-Tagging WM white matter WMH white matter hyperintensity
1
“He who cannot draw on three thousand years is living from hand to mouth.”
Johann Wolfgang von Goethe
1 Background
1.1 Cerebral Small Vessel Disease (CSVD)
Small Vessel Disease (SVD) is a multi-organ vascular disease phenomenon that is associated
with human aging and affects the brain, eyes, and kidneys, for example. Cerebral Small Vessel
Disease (CSVD) is associated with structural and functional changes in the brain, principally in
the connective white matter (WM), in the form of ischemic lesions. White matter
hyperintensities (WMHs) are one of the most common manifestations of CSVD. The WMH
appearance on diagnostic imaging is in contrast to subcortical infarcts, lacunes, perivascular
spaces or cerebral micro bleeds 1. Sometimes referred to as “covert strokes”, WMHs are named
“hyperintensity” because of their bright signal intensity appearance as seen on some radiological
images. CSVD is considered to be the most common neurological disorder. Its high prevalence is
predominantly due to the fact that age is the most important risk factor for developing the
disease. The prevalence of CSVD increases dramatically with age; from approximately 6% to
7% at age 60 to 28% at age 80 2. WMHs are often incidental findings and in up to 89% of
individuals no history of stroke or transient ischemic attack exists 2.
1.1.1 White Matter Hyperintensities
The term WMHs to denote white matter CSVD pathology stems from the fact that WMHs are
readily apparent on T2-weighted magnetic resonance (MR) images, like the FLuid Attenuated
Inversion Recovery (FLAIR) sequence. WMHs can also be detected with reduced sensitivity on
CT, as areas of low attenuation and on proton density (PD) and T1-weighted MRI where they
appear hyper-intense and hypo-intense respectively. The prevalence of WMHs ranges from 11-
12% in adults at the age of 64, and increases with age to about 94% at the age of 84 3. Clinically
WMHs are associated with 1) cognitive decline (lower attention and information processing
2
speed and impaired executive function 4–6 2) increased risk of depression 7,8, gait problem 9, and
most notably increased chance of dementia and stroke 3.
There are several terms that are often used interchangeably to describe WMHs; this is an
unfortunate consequence of the development of the literature in this field. A systematic review of
940 studies by Wardlaw and colleagues in 2013 revealed that there are as many as 50 different
terms that have been used to describe WMHs, including leukoaraiosis, white matter lesions,
white matter changes, and white matter disease 1. This position paper recommends WMHs be
defined as white matter hyperintensities of presumed vascular origin.
Figure 1: WMHs in two 80 year old patients. WMHs are best visualized on T2 weighted fluid
attenuated inversion recovery (FLAIR) images. Left: minor amounts of WMHs voxels. Right:
more pronounced WMHs in both periventricular and deep WM regions (Adopted from Debette
et al. 2010).
1.1.2 Etiology and Consequences of WMHs Leukoaraiosis is derived from the Greek word leuko, “white” and araios, “rarefaction”. WMHs
(leukoaraiosis) were first described as radiological findings in 1986 10 and was later followed by
attempts to provide a pathological description 1. The etiology of WMHs is not completely
understood, however there are histopathological, epidemiological, and physiological studies that
provide important information.
3
WMHs tend to accumulate in the deep white matter and regions around the lateral ventricle
(periventricular WM). Deep medullary arteries provide the main blood supply to these regions.
The limited number of arteries that supply blood to the deep white matter with little collateral
supply along with their susceptibility to vascular pathology make the deep WM more vulnerable
to ischemia. Also, these vessels are among the longest in the brain and often become tortuous
and compromises cerebral blood flow.
From a pathological perspective, endothelial damage caused by various factors such as
hypertension may lead to vessel wall thickening and narrowing of the vessel lumen in arteries
supplying the WM. Endothelial damage may also result in the disruption of the blood brain
barrier. Consequently, blood plasma and other substances that cannot pass the blood brain barrier
may penetrate to the brain tissue and damage the cells. Histopathological studies of WMHs have
found evidence of demyelination, loss of glial cells axon damage and spongiosis 11. It has been
suggested that the incomplete infarcts (i.e. demyelination, axonal and oligodendrocytes damage)
are the consequences of impaired haemodynamics and subsequent ischemia 12,13. A number of
vascular pathologies are also shown to be associated with WMHs including tortuous arterioles,
and periventricular venous collagenosis. Overall, there is ample evidence that WMHs is driven,
at least in part, by chronic ischemia.
Aside from arterial dysfunction, there is a form of venopathy that has also been observed in the
periventricular regions of patients with WMHs. First described by Moody and Brown, the
periventricular venous collagenosis is manifested by deposition of collagen in the venous walls
resulting in intramural thickening and stenosis of veins 14. It has been shown that venous
collagenosis increases with age and is positively correlated with the severity of periventricular
WMHs 14,15. Venous collagenosis can result in ischemia via two mechanisms 1) increasing
vascular resistance and 2) leakage of fluid resulting in vasogenic edema (vessel leakage) 16. It has
also been suggested that venous collagenosis may cause the veins to dilate and result in “venous
insufficiency”, a condition characterized by impaired venous flow and damaged vessel function 16.
Global changes in cerebral physiology support the notion that WMHs are caused by underlying
ischemia. Investigation of alterations of cerebral blood flow (CBF) in patients with WMHs using
neuroimaging started in early 1990s 17. More recently Marstand et al. and O’Sullivan et al. have
4
used contrast-enhanced MRI to study CBF in WMHs and NAWM of elderly cohorts 18–20. The
common findings of the studies that have investigated CBF in relation to WMHs are that 1)
relative to age matched controls, the WMH cohorts have lower CBF both globally and in their
NAWM and 2) in individuals with WMHs, CBF is lower in the lesions compared to NAWM.
Apart from CBF measurements, the hypothesis that WMHs are ischemic in nature have been
tested more directly via measurements of oxygen extraction fraction (OEF), defined as the
fraction of oxygen taken out from the blood at the level of capillaries, and cerebral metabolic rate
of oxygen (CMRO2) using positron emission tomography (PET). Meuguro and colleagues have
shown that when a group of 21 adults with periventricular WMHs are divided into two groups
with mild and more severe lesion loads, the group with more severe WMHs has lower gray
matter CBF, and higher OEF compared to the first group 21. Hatazawa et al. have studied
asymptomatic patients with WMHs and normal controls and have found reduced CBF and higher
OEF in the WM of the patient group compared to the control. WMHs and its relationship with
hemodynamic measures in symptomatic patients have also been investigated. Patients with a
dementia as well as WMHs have higher cerebral cortex OEF and lower CBF compared to
hypertensive controls suggesting that hemodynamic impairment in the cortex could be related in
parts to the “disconnection of neural fibers in deep WM” 22. Yamaji and others have reported that
Alzheimer’s disease (AD) patients with WMHs have lower regional WM and GM CBF and
higher OEF compared to patients with AD alone 23. Similarly, in a group of patients with lacunar
stroke Nezu et al. demonstrated a reduction in CBF and CMRO2 and an increased OEF in
patients with severe WMHs compared to those with mild WMHs 24. Lastly, it has repeatedly
been reported that among patients with carotid artery occlusive disease, those with WMHs tend
to have lower CBF and higher OEF compared to WMHs-negative patients, which suggests that
hemodynamic impairment plays a role in pathogenesis of WMHs 25–27. The increased OEF can
be viewed as an attempt to maintain oxygen delivery to tissues due to impaired haemodynamic.
The observation that WMHs are associated with problems in grey matter (GM) is notable. WM
consists of glial cells and axons and is responsible for the transmission of neural signals between
different brain regions. Therefore, structural or functional impairment of WM can negatively
affect the information processing speed and tasks that require complex communication across the
brain. The association between WMHs and GM hypoperfusion is less clear, but it may be related
to the fact that both WMHs and GM hypoperfusion are consequences of a systemic brain injury.
5
Overall, it is reasonable to conclude that WMHs are associated with reduced CBF and increased
OEF. In the hypoperfused state the OEF should increase as a compensatory mechanism to meet
metabolic demand, hence preserving CMRO2. As the CBF falls further, uncoupling between
oxygen delivery and metabolic rate occurs, resulting in a decrease of CMRO2, which highlights
the condition when increased OEF cannot compensate the effects of reduced CBF. A decrease in
CMRO2 is often associated with loss of neuronal functionality and cognitive decline.
1.1.3 Geometrical Location of White Matter Hyperintensities
WMHs are typically formed around the horns of the lateral ventricles (See Figure 1). These
periventricular lesions tend to grow larger and spread towards the lateral wall of the ventricles
and then into the deep white matter (i.e. closer to the grey matter). Other than these
periventricular WMHs, some lesion are formed in the deep white matter far away from the
ventricles. In general, the bigger and the more confluent the lesions the more severe they are.
Age related WMHs have a specific pattern. The deep white matter in the temporal lobe almost
never has any WMHs, except in more rare cerebrovascular conditions 28. A hereditary form of
covert stokes known as cerebral autosomal dominant arteriopathy with subcortial infarcts and
leukoencephalopathy (CADASIL) is one example where WMHs are diffuse in many white
matter regions, including the temporal lobe.
1.2 Overview of Magnetic Resonance Imaging: Principles and Relevant Techniques
This section briefly discusses the general principles of magnetic resonance imaging (MRI) and
the imaging techniques that are used in this thesis, namely susceptibility-weighted imaging
(SWI), and arterial spin labeling (ASL).
6
1.2.1 Basic Principles of MRI
While a complete and accurate description of MR physics requires quantum mechanics, a
classical description can adequately explain the majority of macroscopic events. This section
uses this classical treatment to briefly describe the basics of MRI.
Atoms with an odd number of protons and/or an odd number of neutrons have a property known
as nuclear spin angular momentum. This spin angular momentum is a vector quantity and is
often called simply the spin. It can be visualized as a charged particle spinning around an axis
producing a magnetic field similar to that of a bar magnet. In the human body, the hydrogen (H)
atom, with a single proton, is highly abundant and the most studied atom that possess a spin that
is used to generate an MRI signal. In the absence of an external magnetic field, the spins of
individual atoms are oriented in random directions resulting in zero magnetization (the vector
sum of spins is zero). When an external magnetic field B0 is applied, the spins exhibit two
important behaviours. First, the spins that were randomly oriented now tend to align in the
direction of the external magnetic field (the z direction by convention) producing a net
magnetization M. Second, the individual spins also precess around the axis of the of external
magnetic field with a resonance frequency known as the Larmor frequency, ω, which is
proportional to the applied magnetic field and a unique atom specific constant known as the
gyromagnetic ratio γ. The following equation represents the relationship between the Larmor
frequency, the external magnetic field and the gyromagnetic ratio:
7
Figure 2: A) The radiofrequency field B1 applied in the transverse direction and tuned to the
Larmor frequency so as to rotate the magnetization. B) Behaviour as seen from a rotating frame
of reference. (Adopted from Nishimura 1996)
In the presence of an only constant B0, the spins continue to precess at the Larmor frequency and
no change in the magnetization vector is observed. In order to obtain an MR signal, the
equilibrium state of spins should be disturbed. This is done by applying a radiofrequency
magnetic field B1 at the resonant frequency of the spins in the x-y (transverse) plane. The
application of this radiofrequency pulse moves the spins out of equilibrium, and rotates the net
magnetization towards the x-y plane (see Figure 2). The amount of rotation depends on the
strength and duration of B1. When the pulse is turned off, the magnetization continues to precess
at the Larmor frequency and also starts to return back to its equilibrium state along the z-axis
characterized by the longitudinal time constant T1. While the longitudinal magnetization is
recovering along the z axis, the transverse magnetization decay due to spin-spin interactions
resulting in the dephasing of the spins. The rate at which this dephasing takes place is given by
T2 transverse relaxation time. These processes are independent of each other and are briefly
explained in the following subsections.
The rotating magnetization induces an electromotive force (emf) in a receiver coil based on the
Faraday’s law of induction. This MR signal is then used to reconstruct images. However the MR
signal is not recorded immediately after flipping the magnetization into the transverse plane. The
signal is measured after a short period of time known as the echo time or TE (time to echo). The
choice of TE is of critical importance since it contributes to the type of contrast that we see in the
final image.
Figure 3: Precession of the magnetization vector in the xy-plane and generating an emf, which is
detected by the receiver coils. (Adopted from Nishimura 1996)
8
The RF receiver coil(s) in MRI record the total signal generated by all the spins in an excited
region. Since the excitation done by B1 is not spatially selective, spatial localization is required
and is obtained by using linear gradients. This process starts with the selective excitation of a
slice in one direction (z direction for example) via the application of slice selective gradient Gz.
Then the application of an RF pulse excites the magnetization and rotates it toward the xy plane.
During the decay of this magnetization and while a signal is being detected a second gradient in
the x direction is applied which is known as the frequency-encoding gradient. The strength of
this gradient is linearly changing as a function of x. As a result, different spins at different x
positions now precess at different frequencies. This frequency encoding implicates that the
precession frequency can be used to distinguish the spatial location of the spins in the x
direction. Lastly, to localize the spins in the y direction a third gradient known as the phase
encoding gradient is applied. This results in a phase offset among spins in the y direction. This
gradient is turned on with a constant strength prior to the detection of the MR signal. Then the
frequency-encoding gradient is applied and the signal is detected simultaneously. The process is
then repeated with different Gy strengths. It should be noted that the method described above is
only one of the common ways of applying the gradient in MRI.
1.2.2 Contrast in MRI
Image contrast in MRI is due to the fact that biological tissues have unique relaxation time
constants, which results in differences in the evolution of their magnetization and hence the MR
signal. The relaxation time is a property of a material and its surrounding tissues and therefore is
fixed. In this section the T1 and T2 relaxation constants and their significance in MRI is briefly
discussed.
1.2.2.1 T1 Relaxation Time
The T1 relaxation time (also known as longitudinal or spin-lattice relaxation time) is related to
the rate at which spins transfer their energy to the surrounding after the administration of the RF
pulse. The faster the spins transfer their energy the faster they return to their lowest energy state
along the z direction. Immidiately after the RF excitation pulse the length of the transverse
9
magnetization decays and consequently the longitudinal component of the magnetization (Mz)
recovers to its pre-excitation state as given by the following equation.
After a while another RF pulse is applied and the longitudinal magnetization is flipped into the
transverse plane again and the same recovery follows. The time interval between two
consecutive excitations is known as the repetition time or TR. When the frequency of the motion
(translational, vibrational and rotational) of protons matches the Larmor frequency the maximum
energy transfer takes place. In the body the motional frequencies of hydrogen protons in fat is
almost equal to the Larmor frequency of hydrogen. Therefore fat has the greatest ability to
transfer energy to its surrounding hence the shortest T1.
1.2.2.2 T2/T2* Relaxation Times
After the adminesteration of an RF pulse and while the longitudenal magnetization is being
recovered along the z direction, the tranverse magnetization undergoes an independent form of
decay. Immidiately after the RF pulse, all the spins that constitute the transverse magnetization
are in phase (i.e. rotating with the same angular velocity). However, as time goes by the spins
dephase which and therefore their vector sum decreases. The dephasing process is a result of two
phenomena namely the spin-spin interaction and magnetic field inhomogeneities. The former is
explained by the T2 time constant while the latter denoted by the T2’ relaxation time. The
overall rate of dephasing can be written as:
and the magnetization in the x-y plane decys base on the following:
where t is the time after excitation.
10
1.2.3 Susceptibility-weighted Imaging (SWI)
Susceptibility-weighted magnetic resonance imaging (SWI), as we know it today, is a recently
new imaging technique, first described in 1997 by Reichenbach and Haacke, and has gained
clinical use in the past decade 29,30. Historically, it was known as the high-resolution BOLD
venographic (HRBV) technique where BOLD stands for the blood oxygenation level-dependent.
Later on, when it was used in clinical applications, the term AVID BOLD was coined to the
technique where AVID stands for the application of venographic imaging to diagnose disease.
The term susceptibility-weighted imaging or SWI was suggested in 2004 to highlight the broader
application of the technique, which goes beyond the visualization of the cerebral veins 31. It is
entirely appropriate to view the development of SWI in parallel with the BOLD fMRI technique
for functional brain activation. In the former case, acquisition time is dedicated towards spatial
resolution, whereas in the latter case BOLD volumes are acquired repeatedly, i.e. every
2seconds, at the expense of spatial resolution.
Magnetic susceptibility χ is an intrinsic property of materials that describes the degree of
magnetization of an object when placed in an external magnetic field and is given by
where M is the induced magnetization, and H is the external magnetic field.
The relationship between induced magnetic field inside an object and the external magnetic field
can be written as
11
where is permeability [N.A-2], is permeability in vacuum and is equal to 4π×10-7 N.A-2 and
is the relative permeability. The above equation shows that the induced magnetization M is
proportional to both susceptibility and the magnetic field.
Materials with susceptibility of greater than zero are called paramagnetic and substances with
negative susceptibility are called diamagnetic. Changes in magnetic susceptibility result in
magnetic field inhomogeneity which in turn affects the T2* relaxation of the nuclei. It has been
shown that the blood T2* is a quadratic function of oxygen saturation (Y), but can be estimated
as a linear function over a wide range of Y 32. For blood with oxygen saturation of Y the signal as
a function of echo time can be written as
The change in magnetic susceptibility not only changes the signal intensity on the T2 or T2*
weighted image, but also results in the changes in the phase of the MR signal. The measured
phase is proportional to the local change in the magnetic field and also to the echo time (TE).
Starting with the Larmor equation
The phase can be written as
and the phase difference between two tissues after a time of TE would be
€
Δϕ = ΔωTEΔϕ = (γΔB)TEΔϕ = γ (ΔχB0)TE
Different tissue types and structures in the brain have different magnetic susceptibilities, which
can result in local a magnetic field inhomogeneity. For example, hemoglobin has two
oxygenation states depending on whether it is bounded to oxygen or not. When bounded to
oxygen, the oxyhemoglobin is slightly diamagnetic compared to most brain tissues. The state
where the oxygen is not bounded to the hemoglobin is deoxyhemoglobin and is paramagnetic
12
compared to surrounding brain tissue. The deoxygenated blood, which is more abundant in the
cerebral veins (with oxygenation levels of approximately 60%) than in cerebral arteries (with
oxygenation levels that are close to 100%), has Δχ ≈ 0.27 ppm 33 compared to surrounding
tissue. This susceptibility difference of the venous blood is the basis of SWI venography. When
vessels are modelled as infinitely long cylinder (as an approximation of long cerebral veins with
diameter much smaller than their length), the magnetic field between the vein and the
surrounding can be written as
€
ΔB = 2π⋅ Δχ⋅ B0 ⋅ cos2θ −
13
⎛
⎝ ⎜
⎞
⎠ ⎟ ⋅ (1−Y )⋅ Hct
where Δχ is the susceptibility difference between deoxygenated and oxygenated blood, B0 is the
strength of the main magnetic field, θ is the angle the angle between B0 and the blood vessel, Y is
the oxygen saturation of the blood inside the vessel, and Hct is the hematocrit34,35. For a cerebral
vein, which is running parallel to the magnetic field, the TE that results to the maximum phase
difference can be calculated. Setting Y = 0.7, Hct = 0.40, and Δχ=0.27×10-6, as suggested by
Koopmans and colleagues, the TE that corresponds to a phase difference of π (i.e. maximum
signal cancelation on the phase image) is equal to 28 ms 35. When θ ≠ 0, the situation becomes
more complicated due to the contribution of extravascular magnetic field components in addition
to the above intravascular field difference and the analytical calculation becomes nontrivial 34,35.
SWI is a gradient echo imaging technique. A conventional SWI is performed at high spatial
resolution (i.e. in plane voxel dimensions are less than 1 mm), with 3D imaging (i.e. phase
encoding in two directions), and with flow compensation in all three directions (i.e. x, y and z).
SWI is fairly unique in MRI because both magnitude and phase signals are used in the
reconstruction to form the susceptibility-weighted composite image. The reason for this is based
on the following: changes in magnetic susceptibility lead to changes in local magnetic field.
These field inhomogeneity result in two notable effects. First, a reduction of T2* relaxation time
which corresponds to areas of lower signal intensity on a magnitude image. Second, a phase shift
is observed compared to the surrounding regions as reflected by the phase image. SWI is
designed primarily to produce an impressive magnitude image and the phase image is used to
accentuate the magnitude tissue differences. Therefore, the acquisition parameters are chosen
such that the final image would be of adequate quality.
13
SWI post-processing is based on the following. First the phase images are unwrapped, where
multiples of 2π are added or subtracted such that the 2π discontinuities in the phase images are
removed. Then macroscopic magnetic field variations, which arise from sources such as
background field inhomogeneity, are removed by applying a high pass filter. The unwrapped
filtered phase image now has high frequency phase variations corresponding to local areas of
changes in magnetic susceptibility.
The filtered phase mask, which has values between -π and π, is now used to create a so-called
“phase filter”, which is designed to enhance the contrast of the magnitude image. The phase
mask f(x), suppresses voxels with certain phase and is usually designed in the following manner:
The phase mask defined above has values between zero and one. If a voxel has a phase shift of –
π it will be completely suppressed, whereas those with values between –π and 0 are partially
suppressed. This phase mask is then multiplied by the magnitude image, ρ(x), a number of times
to provide a unique form of contrast. It has been shown that four times multiplication produces
the optimal contrast for detection of cerebral veins 36. The resultant composite image (i.e. f 4(x)
ρ(x)) is called the susceptibility-weighted image.
SWI produces images that are T2* weighted, hence it requires an image acquisition with a long
TE. In conventional gradient echo imaging a long TE results in longer TR hence increasing the
acquisition time. When long TE in combination with fast imaging is desired, echo-shifted (ES)
pulse sequences could be employed where TE is longer than TR. In a simple ES gradient echo
pulse sequence, modified gradient lobes are applied such that the gradient echo will be formed
with a minimum delay of one TR hence making the TR shorter than TE and decreasing the scan
duration (See Figure 4).
14
Figure 4: A general example of an ES pulse sequence where gradients lobes with areas of –A
and +2A are added prior and after each readout to shift the echo by one TR interval (Adopted
from Bernstein et al. 2004 37).
1.2.4 Clinical Applications of SWI
MR venography with deoxygenated haemoglobin as an intrinsic contrast agent and with the use
of both phase and magnitude images began with the work of Reichenbach and Haacke in 1997 38.
Subsequently, the feasibility of MR venography at 3T was shown by Reichenbach et al. in 2000 34. Figure 5 shows a typical SWI image at 3T, where deep medullary veins are clearly
distinguishable. Since 2004 there has been a unified understanding of SWI processing techniques
and major MRI vendors have adopted the SWI sequence 31. SWI has established its position in
clinical settings as well. Nowadays, SWI has clinical application in both adult and paediatric
neuroimaging. This section focuses mostly on the clinical applications of SWI in adults. An in
depth review of the clinical application of SWI in children has been written by Tong 39.
15
Figure 5: A typical SWI from a healthy participant. The hypointense lines are the cerebral veins.
To date SWI has been used in following settings to provide additional or complementary