Introduction
Contemporary neuroimaging by magnetic resonance imaging (MRI) is comprised of multi-parametric
acquisition protocols using multiple sequences that allow the radiologist to derive
information about macro- and micro-structure, function, metabolism, and/or perfusion.
Such multi-parametric approaches can facilitate initial differential diagnosis, as
well as disease and therapy monitoring of various pathologies affecting the central
nervous system (CNS). Specifically, perfusion imaging can be achieved using several
brain MRI techniques [1 ]
[2 ]
[3 ]. Most commonly in the clinical routine, information about perfusion is derived from
methods that require the intravenous application of a gadolinium-based contrast agent.
Those include dynamic contrast-enhanced MRI (DCE-MRI), making use of T1 shortening
effects of gadolinium during repeated acquisitions of T1-weighted images, and dynamic
susceptibility contrast MRI (DSC-MRI), relating to local magnetic field distortion
effects around vessels with T2* dephasing and signal loss while a bolus of gadolinium
passes, captured by a series of rapidly acquired spin or gradient echo images [1 ]
[4 ]. In contrast to these approaches, arterial spin labeling (ASL) works fundamentally
differently since it does not require the injection of a gadolinium-based contrast
agent, but instead uses blood-water as an endogenous tracer, enabling the assessment
of cerebral blood flow (CBF) [5 ]
[6 ]
[7 ]
[8 ]
[9 ].
Since the introduction of the ASL method in the early 1990s, it has shown promise
as a potential alternative to conventional perfusion imaging methods such as DCE-
or DSC-MRI [5 ]
[10 ]. With the publication of a consensus on the clinical implementation of ASL by the
Perfusion Study Group of the International Society of Magnetic Resonance in Medicine
(ISMRM) and the European Consortium for ASL in Dementia in 2015, the technique has
been further conceptualized and the transition to broader clinical application has
been facilitated [5 ]. This early consensus statement has been recently followed up by an overview of
the current state and guidance on ASL in clinical neuroimaging with a methodological
focus, published on behalf of the ISMRM Perfusion Study Group [11 ]. Nowadays, ASL can be part of imaging protocols for several diseases affecting the
CNS, ranging from cerebrovascular diseases as the most prominent clinical application
and neoplasms to concussion or migraine, which are conditions that may not even necessarily
show morphological alterations on conventional structural MRI. Furthermore, the capabilities
of ASL have been considerably expanded in recent years: while initial applications
predominantly enabled the investigation of whole-brain perfusion, recent advances
have made available vessel-selective imaging of single perfusion territories of the
brain, as well as time-resolved angiography based on such vessel-selective imaging.
Against this background, this narrative review article aims to provide an overview
of general methodological and technical characteristics of ASL, followed by a review
of ASL applications for various diseases affecting the CNS. In this context, previous
publications have provided detailed recommendations, particularly regarding the basic
methodological applications of the most common ASL techniques based on major clinical
use cases, while some specific ASL-based methods that are not yet widely applied (e. g.,
vessel-selective imaging or ASL-based angiography) or diseases that currently do not
regularly require perfusion imaging (e. g., migraine or brain injury) have not been
covered [5 ]
[11 ]. Therefore, a special focus of the present article was to 1) provide methodological
aspects and clinical examples for vessel-selective ASL and time-resolved ASL-based
angiography (i. e., two advanced methods providing insights that cannot be derived
from other pulse sequences such as DSC-MRI), and 2) present clinical examples for
diseases that do not regularly show clear morphological alterations on conventional
MRI (i. e., ASL could deliver biomarkers for diseases such as migraine or concussion
that are mostly neglected by standard neuroradiological diagnostics).
Methods and technical aspects
General overview
The high potential of ASL is the result of its methodological elegance. It is a noninvasive
MRI technique that uses the blood-water as an endogenous tracer ([Fig. 1 ]). As water can be assumed to be freely diffusible, ASL is particularly well suited
for measuring brain perfusion. The method is based on the subtraction of label and
control images, where the magnetization of blood-water is magnetically inverted during
the labeling process, while no effective labeling is performed prior to the acquisition
of the control image. In general, the ASL method makes it possible to quantify CBF
(in ml/100 g/min) [5 ]
[9 ]
[11 ].
Fig. 1 Acquisition and main concepts of arterial spin labeling (ASL). A For labeling purposes, the inflowing blood-water is magnetically labeled within the
labeling plane. The labeling plane should be placed orthogonally on a straight segment
of the brain-feeding arteries. B Prior to image acquisition, a post-label delay (PLD) is introduced to account for
the time it takes the labeled blood to arrive at the imaging volume. C By labeling single brain-feeding arteries, super-selective ASL (ssASL) allows mapping
of individual perfusion territories, for example those belonging to the right internal
carotid artery (R-ICA; red area) or left ICA (L-ICA; green area) as well as to the
vertebral arteries (VAs; blue area).
Historically, two different categories of labeling methods were proposed: continuous
ASL (CASL) and pulsed ASL (PASL) [10 ]
[12 ]
[13 ]. While CASL offers a comparably good signal-to-noise ratio (SNR), it is limited
by magnetization transfer-induced artifacts and the high amount of energy deposition
in the subject. Additionally, the availability of continuous pulses is limited on
most clinical MRI systems [14 ]. In contrast, short-pulsed schemes are widely available and result in way lower
energy deposition. However, single pulses as implemented in PASL result in a comparably
low SNR. Both CASL and PASL have shown promising initial applications. However, they
have been used quite rarely for clinical imaging, which is due to their inherent limitations.
Pseudo-continuous ASL
A major advancement was a third approach called pseudo-continuous ASL (pCASL; [Fig. 1 ]), which was proposed by Dai et al. in 2008 as a hybrid of CASL and PASL [15 ]
[16 ]. Conceptually being a CASL technique, the long labeling period is split into a series
of short pulses, which has several advantages. First, short pulses are available on
most clinical MRI systems, providing wide applicability and availability. Second,
the technique results in lower energy deposition in the subject, allowing longer labeling
periods. The labeling is hereby performed within a thin slice, which is often referred
to as the labeling plane ([Fig. 1 ]).
In pCASL, a post-label delay (PLD) is introduced between the labeling and image acquisition
([Fig. 1 ]). This is set to account for the arterial transit time (ATT), which is the time
it takes the labeled blood to travel from the labeling region to the tissue of interest,
which is then captured by the imaging volume. The PLD is a critical parameter when
setting up a pCASL sequence: short PLDs can result in underestimation of CBF, while
the signal intensity will be reduced for long PLDs due to venous draining and the
inherent T1 decay of the labeled spins [5 ]. Therefore, the consensus statement of the Perfusion Study Group of the ISMRM and
the European Consortium for ASL in Dementia included consistent recommendations for
the implementation of ASL sequences and the choice of parameters, such as different
PLDs for different populations [5 ]. Furthermore, updated recommendations for the implementation of clinical perfusion
imaging sequences using ASL have been recently provided by the ISMRM Perfusion Study
Group [9 ]
[11 ].
Time-encoded ASL
Instead of adapting the PLD to the respective study cohorts (e. g., in terms of age),
multiple PLDs can be acquired, which can be used to fit a kinetic model to estimate
the CBF and ATT (e. g., Buxton general kinetic model) [17 ]. This can greatly improve quantitative CBF estimates, as it accounts for intra-
and inter-subject ATT variations. In general, multi-PLD data can be obtained by acquiring
multiple single-PLD datasets with variations of the PLD between each scan. However,
this results in rather long scan times, which may not be feasible for clinical imaging
protocols. Therefore, several approaches have been proposed to acquire multi-PLD data
[18 ]
[19 ]
[20 ]
[21 ]
[22 ].
One readout-based approach uses Look-Locker sampling, which basically employs a series
of low flip angle excitations during the relaxation of the longitudinal magnetization
to the thermal equilibrium, which allows image acquisition with a considerably high
number of different delays [18 ]
[19 ]. However, the readout pulse train will effectively reduce the amount of signal available
for later PLDs, resulting in a reduced SNR [20 ]. As an alternative, Günther et al. proposed a time-encoded labeling scheme based
on CASL, which was later also adapted for pCASL [21 ]
[22 ]. By alternating label and control conditions according to a Hadamard matrix, decoded
images can be acquired in a highly time-efficient manner. Those images can be decoded
during postprocessing, resulting in CBF and ATT maps within clinically feasible scan
times.
ASL-based perfusion territory mapping
All whole-brain ASL implementations including pCASL perform non-selective labeling
of the blood-water flowing through all the brain-feeding arteries. However, especially
in steno-occlusive diseases, variations in blood supply between different perfusion
territories are a typical scenario. Therefore, vessel-selective imaging could be of
high clinical impact, allowing the noninvasive determination of individual perfusion
territories. In this regard, different approaches have been proposed. Based on PASL,
rotated labeling slabs have been used that have been manually placed onto the major
brain-feeding arteries [23 ]
[24 ]. However, this comes with the inherent drawbacks of PASL and does not provide high
spatial selectivity. Another approach is vessel-encoded ASL, where off-resonance effects
introduced by gradients are used to generate effective labeling regions [25 ]
[26 ]. Although it is time-efficient, the method is based on population-averaged distances
between the internal carotid arteries (ICAs) and the vertebral arteries (VAs) and
is less sensitive for collateral blood supply.
However, specific vessel-selective labeling can be facilitated by super-selective
ASL (ssASL), which was proposed by Helle et al. based on pCASL [27 ]. Time-varying gradients are applied perpendicular to the selected area and result
in effective labeling spots. Thus, ssASL provides a high degree of freedom in placing
the labeling spot together with high spatial selectivity. Clinical applicability was
further enhanced by options for automated planning of the labeling spots [28 ]. Recently, a combination of ssASL labeling with the contrast-enhanced timing-robust
angiography (CENTRA) keyhole technique and view-sharing was proposed (4D-sPACK), leading
to noninvasive and time-resolved ASL angiography with clinically reasonable scan times
of less than 5 minutes per labeled vessel [29 ]
[30 ].
Approaches to ASL data processing and analysis
Post-processing of acquired ASL data regularly includes control-label subtraction
and averaging of the subtraction series, which could be extended by additional steps
for motion correction or outlier scrubbing [9 ]
[11 ]. Depending on the particular clinical use case, further post-processing steps might
be helpful for data interpretation, such as partial volume correction (e. g., to remove
pseudo-hypoperfusion effects secondary to cerebral atrophy in neurodegenerative diseases
[ND]) or normalization to normal-appearing brain parenchyma (e. g., to potentially
improve accuracy for brain tumor grading) [9 ]
[11 ]
[31 ]
[32 ]. For pCASL as today’s most common ASL-based method, generation of perfusion maps
and quantification of CBF can be obtained. Typically, pCASL is used to derive information
on whole-brain perfusion, but segmentation (e. g., by co-registration with anatomical
T1-weighted or T2-weighted sequences) can be added to derive more localized information
from a lesion or a specific brain structure. In contrast, ssASL provides territorial
perfusion information from a preselected brain-supplying artery, thus is more dynamic
and may provide the most comprehensive information when more than a single vessel
is labeled successively (e. g., to derive information about perfusion territory shifts
and individual cerebrovascular architecture).
In recent years, considerable advancements have been made regarding both standardized
ASL data storage as well as standardized perfusion analyses. One example is the extension
of the Brain Imaging Data Structure (BIDS) for ASL data, in order to provide data
storage standards that meet the need for structured image data organization, including
also metadata beyond the image files (e. g., acquisition characteristics such as voxel
sizes) [33 ]
[34 ]. Furthermore, software packages such as ExploreASL (written in MATLAB and based
on Statistical Parametric Mapping [SPM]) have been developed recently, which may facilitate
standardized analysis of ASL data across centers and scanners [35 ]. Another software application is ASLPrep, aiming to provide a generalizable and
robust workflow targeting reproducible processing of heterogeneous ASL data [36 ]. As such, it also provides advanced analysis approaches beyond the commonly used
kinetic model for CBF quantification, including two different Bayesian models incorporating
information regarding brain structure with the Bayesian Inference for ASL (BASIL)
and Structural Correlation with Robust Bayesian (SCRUB) methods [36 ].
Clinical use cases
Cerebrovascular diseases
Ischemic stroke
Globally, stroke is ranked as the second leading cause of disability and mortality,
with approximately 13.7 million incidents of stroke in 2016 [37 ]. Of those, more than 80 % are categorized as ischemic strokes, which commonly occur
on the basis of cardioembolism, large artery atherosclerosis, or vessel occlusions
[37 ]
[38 ]. Ischemic stroke is one of the most frequent indications for perfusion imaging by
computed tomography (CT) or MRI. Perfusion imaging-based selection of patients for
endovascular therapy and/or intravenous thrombolysis in ischemic stroke has found
entrance into the clinical routine based on a large body of evidence both for CT and
MRI [39 ]
[40 ]
[41 ]
[42 ]
[43 ]
[44 ]
[45 ]
[46 ]
[47 ]
[48 ]. In recent years, ASL has been studied many times as a potential alternative to
DSC-MRI in stroke imaging ([Fig. 2 ]).
Fig. 2 Post-stroke perfusion imaging using pseudo-continuous arterial spin labeling (pCASL)
with corresponding axial diffusion-weighted imaging (DWI), which was acquired three
days after ischemic stroke due to intracranial vessel occlusions. A In a 55-year-old female patient, DWI showed a large infarct with diffusion restriction
after thrombotic left-sided occlusion of the M1 segment, and cerebral blood flow (CBF)
maps from pCASL showed marked hyperperfusion of the infarct territory that occurred
after successful revascularization therapy. B In a 70-year-old male patient, DWI showed a large temporo-occipital infarct after
right-sided thrombotic occlusion of the P1 segment, and the infarct core was markedly
hypoperfused on pCASL-derived CBF maps, while a small region in the lateral aspect
of the temporal lobe showed marked hyperperfusion where no diffusion restriction was
observed.
Studies comparing DSC-MRI and ASL-based perfusion imaging have found overall good
agreement in ischemic stroke [49 ]
[50 ]. Bokkers et al. found that ASL-based perfusion may be used alternatively to DSC-MRI
in penumbra imaging for acute stroke, especially in patients with contraindications
to gadolinium-based contrast media [51 ]. Furthermore, ASL has been shown to be more sensitive than DSC-MRI for detecting
post-stroke hyperperfusion, which often occurs after successful revascularization
therapy [50 ]
[52 ]. It should be noted that a previous study indicated that hyperperfusion may be associated
with a good outcome after stroke, likely as a surrogate of successful reperfusion
and reactive infarct hyperemia [52 ]
[53 ]
[54 ]. On the other hand, ASL-based hyperperfusion following stroke has also been suggested
as a risk factor for the development of intracranial bleeding [53 ]
[54 ]
[55 ].
Post-stroke (hyper-)perfusion could become a clinically relevant imaging marker, but
thus far it is unclear at what threshold the risk of intracranial bleeding overtakes
potential physiological benefits. Due to its sensitivity to post-stroke hyperperfusion,
ASL seems especially suitable for studying this phenomenon and may be used in the
future to screen for patients that are at high risk for the development of infarct
bleeding.
Arteriovenous malformations/Moyamoya disease
Pre-treatment imaging of brain arteriovenous malformations (AVMs) is usually performed
with digital subtraction angiography (DSA) as the reference standard [56 ]. A noninvasive alternative for vessel-selective angiography is time-resolved ASL-based
4D-sPACK ([Fig. 3 ]). Specifically, it has been shown that this method could reliably identify arterial
feeders, nidus size, and venous drainage in comparison to DSA in a series of 15 AVMs
[30 ]. Furthermore, a case report showed high visual concordance between DSA and ASL-based
MRA and demonstrated the feasibility of the segmentation of vascular territories and
border zones using perfusion maps [57 ]. Furthermore, ASL has been used to monitor treatment success after radiosurgery
for AVMs by enabling confirmation of the obliteration or detection of residual manifestations
following treatment [58 ]
[59 ]. As a potential advantage compared to DSA, ASL-based MRA is not affected by intravascular
pressure changes resulting from contrast medium application through a syringe, which
may be beneficial when trying to understand the hemodynamics of an AVM.
Fig. 3 50-year-old male patient with a right-hemispheric arteriovenous malformation (AVM)
as indicated by irregular vessels according to susceptibility artifacts on axial T2*-weighted
imaging, flow voids on axial time-of-flight magnetic resonance angiography (TOF-MRA),
and hypointense tubular structures on axial non-contrast-enhanced T1-weighted imaging.
According to perfusion maps from super-selective arterial spin labeling (ssASL), the
AVM showed marked hyperperfusion, while parts of the right superior frontal gyrus
appeared without clear perfusion for labeling of the right internal carotid artery
(R-ICA). For labeling of the left ICA (L-ICA), hyperperfusion in projection on the
location of the AVM can still be observed, together with perfusion of parts of the
right-hemispheric superior frontal gyrus from the contralateral side. The cerebral
blood flow (CBF) maps from whole-brain pseudo-continuous ASL (pCASL) resemble hyperperfusion
in projection on the location of the AVM. ASL-based MRA (4D-sPACK) derived from labeling
of the L-ICA showed that the right-hemispheric AVM was to large extents supplied by
the left anterior cerebral artery with early venous drainage according to the third
image of the time series, in agreement with the results from digital subtraction angiography
(DSA) for the L-ICA.
Moyamoya disease, often characterized by progressive stenoses of the distal ICA and
its proximal branches with associated characteristic micro-collateralization, is also
highly accessible to ASL-based imaging including ASL-based MRA ([Fig. 4 ]). Due to alterations of the vascular architecture during the disease course of Moyamoya,
severe and rapid changes in brain hemodynamics can be observed [60 ]
[61 ]. Previous studies have indicated good agreement between DSC-MRI and ASL-based perfusion
imaging for the monitoring of cerebral hemodynamic changes before and after bypass
surgery [60 ]
[61 ].
Fig. 4 26-year-old female patient with Moyamoya disease showing multiple changes of arterial
vessel calibers and reduced flow signals on time-of-flight magnetic resonance angiography
(TOF-MRA), particularly for the left internal carotid artery (L-ICA) and left middle
cerebral artery. According to perfusion maps derived from super-selective arterial
spin labeling (ssASL), large areas of the left anterior frontal lobe were supplied
by the right ICA (R-ICA), as indicated by increased perfusion within this location
when the R-ICA was labeled and lacking perfusion when the L-ICA was labeled. Correspondingly,
ASL-based MRA (4D-sPACK) with labeling of the R-ICA depicted multiple arterial branches
extending to the contralateral hemisphere, which was in agreement with results from
digital subtraction angiography (DSA).
Internal carotid artery stenosis
Usually caused by atherosclerotic plaques at the inner arterial wall, the prevalence
of ICA stenosis (ICAS) ≥ 50 % in patients with acute ischemic stroke ranges between
approximately 15 % and 20 % [62 ]
[63 ]
[64 ]. Besides ischemia, the persistently reduced blood supply of the brain can manifest
as severe chronic perfusion deficits and may result in cognitive decline [65 ]. Both symptomatic patients with previous signs of permanent cerebral ischemia and
transient ischemic attacks, as well as asymptomatic patients with no obvious symptoms
coexist in the case of ICAS.
Current diagnostic procedures usually rely on estimations of the degree of stenosis
by extracranial Doppler ultrasound [66 ]
[67 ]. Even with additional CT angiography (CTA) or MRA, information on the complex local
effects of ICAS on brain tissue (e. g., collateral blood flow) is limited. Furthermore,
while contrast agent-based perfusion imaging methods such as DSC-MRI are promising
especially with regards to the imaging of regional perfusion delay, collateral pathways
cannot be detected even though they are known to severely alter stroke risk patterns
in ICAS [65 ]
[68 ]. Thus, ASL has high potential as a noninvasive imaging tool to quantify regional
CBF by pCASL as well as for collateral flow mapping by ssASL to support delicate treatment
decisions ([Fig. 5 ]).
Fig. 5 71-year-old female patient with right-sided high-grade asymptomatic internal carotid
artery stenosis (ICAS). A Cerebral blood flow (CBF) derived from imaging with pseudo-continuous arterial spin
labeling (pCASL) showed ipsilateral decreases (blue arrow), while at the same time,
contralateral frontal CBF was decreased (red arrow). B These findings were in agreement with delayed perfusion from dynamic susceptibility
contrast magnetic resonance imaging (DSC-MRI)-based time-to-peak (TTP) assessments
regarding the same location. C Super-selective ASL (ssASL) provided additional insight, showing that the right ICA
(R-ICA) perfusion territory was shifted towards the contralateral hemisphere. Contralateral
frontal perfusion impairments were located in the border zone between shifted territories
of the R-ICA and left ICA (L-ICA).
A recent study by Göttler et al. evaluated cerebral perfusion in asymptomatic patients
with unilateral high-grade ICAS using pCASL [69 ]. Even in those asymptomatic patients, significant hypoperfusion was found [69 ]. Correlations of lateralized perfusion deficits with ipsilateral attention bias
were shown as well [69 ]. Moreover, CBF was ipsilaterally decreased by around –18 % compared to the contralateral
hemisphere in such patients [70 ]. This is in good agreement with previous positron emission tomography (PET), DSC-MRI,
and ASL studies [71 ]
[72 ]
[73 ]. Detailed comparisons of six hemodynamic parameters in the same study cohort by
Kaczmarz et al. revealed the most severe pathophysiologic effects for CBF, as measured
by pCASL [70 ]. While the absolute contralateral CBF values were comparable to age-matched healthy
controls, their variability was increased by around +22 % in ICAS [70 ]. This can be explained by collateral flow via the circle of Willis, which could
be evaluated by ssASL.
Comparisons of pCASL, ssASL, and DSC-MRI in ICAS highlight the plausibility of ASL-based
measurements and provide additional insights by mapping perfusion territory shifts.
Moreover, vessel-selective ASL enables the delineation of individual border zones
between perfusion territories [74 ]
[75 ]. Spatial variability of those individual watershed areas (iWSAs) can be increased
in patients with ICAS. Moreover, hemodynamic impairments were enhanced by up to +117 %
within iWSAs compared to brain regions outside of iWSAs [70 ]. As an alternative to vessel-selective ASL, perfusion territory border zones can
be also assessed by time-encoded ASL to map ATT, based on known perfusion delays within
iWSAs [76 ]
[77 ]. In this regard, Di Napoli et al. showed that the presence of arterial transit artifacts
on standard pCASL perfusion maps predicted the presence of symptoms in patients with
ICAS and was highly correlated to a poor collateral status within the circle of Willis
[78 ]. Thus, the technique might be regarded as a candidate to determine indications for
surgical or interventional ICAS therapy. After revascularization, hemodynamic improvement
was demonstrated in asymptomatic as well as in symptomatic patients using ASL [79 ]
[80 ].
Overall, ASL may be particularly suitable for periodic application in ICAS as a noninvasive
technique, including preventive screening, disease progression monitoring of patients
receiving best medical treatment, or treatment efficacy testing after revascularization.
Gliomas
Gliomas are heterogeneous neuroepithelial tumors that stem from the glial cells and
show an age-adjusted average rate of 6.03 per 100 000 of the population [81 ]. Those tumors are commonly categorized into low-grade glioma (LGG, grades 1 and
2) and high-grade glioma (HGG, grades 3 and 4) in relation to the World Health Organization
(WHO) Classification of Tumors of the CNS [82 ]. Typically, first-line treatment includes maximum neurosurgical tumor resection
for cytoreduction and to avoid complications, with the aim of prolonging survival
and improving quality of life [83 ]
[84 ]. In this context, ASL-based perfusion imaging can be applied for several purposes,
including differential diagnosis, preoperative tumor characterization and phenotyping,
as well as monitoring of treatment response after surgery ([Fig. 6 ]).
Fig. 6 Perfusion imaging based on pseudo-continuous arterial spin labeling (pCASL) in the
presence of space-occupying intracranial lesions. A In an 18-year-old male patient, a T2-hyperintense lesion with discrete contrast enhancement
on axial contrast-enhanced T1-weighted imaging and corresponding mild hyperperfusion
on cerebral blood flow (CBF) maps was shown in the neighborhood of the right precentral
gyrus, which was histopathologically confirmed as a high-grade glioma (HGG). B In a 39-year-old female patient, an extensive T2-hyperintense edematous lesion was
detected within the left temporal lobe, with a circumscribed oval lesion in the left
lateral inferior and middle temporal gyrus. In comparison to the contralateral side,
the left temporal lobe showed marked hypoperfusion according to CBF maps, and the
patient was diagnosed with a brain abscess after stereotactic biopsy.
Regarding initial tumor phenotyping, differential diagnosis, and monitoring of therapy,
multi-parametric advanced MRI has an emerging role [85 ]
[86 ]
[87 ]. Specifically, ASL-based perfusion imaging can delineate heightened CBF in glioma,
and it has been demonstrated that it can differentiate glioma from other intracranial
neoplasms such as lymphoma, metastases, or brain abscess ([Fig. 6 ]) [88 ]
[89 ]. It has also been shown that ASL can facilitate distinguishing between LGG and HGG,
with CBF being typically significantly higher in HGG than in LGG [90 ]. This is due to higher perfusion and vascularity in HGG as compared to LGG, which
is related to higher tumor tissue metabolism and neovascularization profiles [91 ]
[92 ]. Two meta-analyses provided cumulative evidence for the role of ASL in differentiating
between LGG and HGG, indicating that absolute and relative tumor blood flow values
could support grading, with a pooled sensitivity of 86 %, specificity of 84 %, and
an area under the curve (AUC) of 91 % for differentiating LGG from HGG [32 ]
[93 ]. Besides categorization into LGG and HGG, ASL-derived tumor perfusion has been associated
with multiple markers that can impact treatment decision making and survival, including
isocitrate dehydrogenase status, methylguanine-DNA methyltransferase promoter methylation,
p53 status, as well as vascular endothelial growth factor expression and tumor microvascular
density [94 ]
[95 ]
[96 ]
[97 ]. Furthermore, it has also been suggested that malignant progression within 12 months
could be predicted with ASL-based perfusion imaging in patients with LGG, with a sensitivity
of 73 %, specificity of 82 %, and odds ratio of 12 [98 ].
Regarding monitoring of treatment response, the main purpose of perfusion imaging
in glioma is to differentiate true tumor relapse or progression from treatment-induced
alterations. Specifically, in HGG, pseudoprogression can typically appear as early
changes within a few months after treatment (particularly after radiotherapy and/or
temozolomide), while pseudoresponse can take effect after the administration of anti-angiogenic
agents (such as bevacizumab) [99 ]
[100 ]
[101 ]
[102 ]. In this context, it has been demonstrated that ASL-based perfusion imaging can
distinguish predominant recurrent HGG from radiation necrosis with a sensitivity of
more than 80 %, which was comparable to findings from DSC-MRI and fluorodeoxyglucose
PET (FDG-PET) [103 ]. Moreover, in patients with HGG who developed progressively enhancing lesions within
the radiation field after resection and chemoradiation, ASL-derived CBF demonstrated
the highest AUC of 0.95 and misclassified the fewest cases regarding true progression
versus pseudoprogression [104 ].
Overall, ASL for glioma imaging has the potential of restricting administration of
contrast agents, which might be relevant in light of gadolinium deposition particularly
for frequent follow-up examinations as regularly scheduled in patients with glioma
[105 ]. Furthermore, compared to DSC-MRI, ASL-based perfusion measurements should not be
biased by blood-brain barrier (BBB) permeability effects, given that water is a freely
diffusible tracer. In contrast, DSC-MRI relies on gadolinium-based contrast agents,
which are intravascular tracers and therefore, even after leakage corrections, DSC-MRI-based
perfusion estimates may be deteriorated by BBB leakage [106 ].
Neurodegenerative diseases
Typically characterized by a progressive loss of specific neuron populations, NDs
become increasingly prevalent with aging and can be classified according to primary
clinical features including Alzheimer’s disease (AD), other dementia syndromes, and
Parkinson’s disease (PD) [107 ]
[108 ]. The most common NDs are proteinopathies, which can lead to abnormal conformational
properties [109 ]. Because of the increasing deposition of the proteins in the cerebral parenchyma,
the final pathway is increasing permeability of the BBB, a decreasing expression of
different receptors, and a disrupted structural and functional connectivity of nervous
fibers [110 ].
In this context, CBF is thought to act as a proxy for synaptic activity throughout
the parenchymal changes [111 ]. Specifically, a study comparing ASL-based perfusion and FDG-PET in patients with
mild-to-moderate AD reported a considerable overlap between the hypometabolic areas
from PET and the hypoperfusion areas from ASL imaging [112 ]. Typically, AD has been characterized by hypoperfusion/hypometabolism in the predilection
sites of the posterior cingulate, precuneus, and/or posterior temporal and parietal
cortices [113 ]
[114 ]
[115 ]
[116 ]. Thus, a certain pattern of alterations in CBF may exist for AD ([Fig. 7 ]). Interestingly, several studies suggested a correlation with decreased perfusion
depending on the tau and amyloid burden [117 ]
[118 ]. Focusing on the alterations of perfusion patterns, ASL might prospectively play
a role in screening for AD, especially with regard to the US Food and Drug Administration
(FDA)-approved AD medication Lecanemab, with early detection becoming even more important
in light of patient selection for a certain therapy [119 ].
Fig. 7 Perfusion imaging based on pseudo-continuous arterial spin labeling (pCASL) in neurodegenerative
diseases (ND). Comparison of perfusion analysis results from A a 58-year-old female patient diagnosed with Alzheimer’s disease (AD; Mini-Mental
State Examination score 19) and B a healthy 60-year-old female control in axial, coronal, and sagittal representations.
The pCASL data were aligned to the anterior commissure-posterior commissure (AC-PC)
line prior to analysis. Blue arrows indicate regions of reduced cerebral blood flow
(CBF), including frontal and parietal areas in the patient. Images were thresholded
at 30 ml/100 g/min.
Furthermore, hypoperfusion with hypometabolism has been revealed primarily in the
frontal brain in patients with frontotemporal dementia, with overall good spatial
agreement between both methods, but slightly lower sensitivity, specificity, and AUC
in discriminating patients from controls for ASL-based perfusion imaging compared
to PET (0.75 versus 0.87) [120 ]
[121 ]. For the semantic variant of primary progressive aphasia, hypoperfusion with hypometabolism
has been identified in the left anterior temporal lobe [122 ]. With respect to PD, the characteristic propagation of alpha-synuclein pathology
can disrupt normal brain function [123 ].
Overall, the physiological basis of the ASL technique, with its ease of repeatability,
offers a great opportunity to derive measures potentially representative of metabolic
information. Thus, in the future, ND may be ideally continuously classified and monitored
during the course of disease by this technique. However, the inherently low SNR of
ASL compared to PET may hamper detection of early changes related to ND, but future
large-scale studies are needed to further explore the role of ASL in this regard.
Traumatic brain injury
Worldwide, it is estimated that over 60 million incidences of traumatic brain injury
(TBI) occur every year, of which approximately 80 % are considered mild TBI (mTBI)
based on initial symptom presentation [124 ]
[125 ]
[126 ]. Over the last decade, ASL has emerged as a promising imaging technique for the
study of cerebral perfusion changes following moderate-to-severe TBI as well as mTBI
[127 ]
[128 ]
[129 ]
[130 ]
[131 ]
[132 ]
[133 ]
[134 ]
[135 ]
[136 ]
[137 ]
[138 ]
[139 ]
[140 ].
Studies in patients with moderate-to-severe TBI revealed decreased regional or global
CBF years after the injury [137 ]
[138 ]
[139 ]
[140 ]. Moreover, a study reported reduced CBF in several cortical and subcortical regions
to be correlated with injury severity, defined as the duration of post-traumatic amnesia
[138 ]. Another study focused on chronic vascular abnormalities in areas of tissue loss
and in normal-appearing brain tissue [139 ]. Specifically, areas of encephalomalacia appeared to have both reduced perfusion
and cerebrovascular reactivity (CVR), with the latter representing the vascular response
after hypercapnic conditions [141 ]. Normal-appearing tissue, on the other hand, revealed only changes in CVR suggesting
global vascular alterations post-TBI [139 ].
Findings from studies on mTBI are heterogeneous, with several studies reporting an
increase in global or regional CBF in the acute and subacute phase after mTBI [127 ]
[128 ]
[129 ]
[130 ]
[131 ]
[132 ]. In contrast, another study found a regional decrease in CBF [131 ]. A study focusing on patients who required hospitalization found that global increase
in CBF acutely after trauma was associated with a better clinical outcome, suggesting
that increases in perfusion might represent a compensatory mechanism (i. e., metabolic
or inflammatory response) [128 ]. In addition, a higher mean global and gray matter CBF was also found in the chronic
phase post-injury [128 ]. However, months after mTBI, studies mostly reported decreased perfusion in various
brain regions, such as the thalamus or in parts of the frontal and temporal lobes
[133 ]
[134 ]. It should be noted that there is limited evidence from longitudinal investigations
suggesting increased CBF acutely and decreased CBF in the chronic phase [129 ]. Furthermore, it remains to be elucidated whether age at the time of injury or sex
are associated with specific alterations in CBF. In this regard, previous studies
have reported an increase in CBF or a decrease in CBF in adolescents [134 ]
[135 ].
Overall, ASL is a promising technique for the evaluation of CBF following both moderate-to-severe
TBI as well as mTBI. However, further research is needed to better characterize the
underlying pathophysiology as well as the effects of important demographic variables
(e. g., age and sex). Moreover, comprehensive study designs are needed to appreciate
the association of CBF with other imaging measures, fluid biomarkers (e. g., neurofilaments),
and outcome measures (e. g., neuropsychological function) to pave the way for CBF
to serve as a marker for diagnosis and prognosis following TBI.
Migraine
Migraine belongs to the entity of primary headaches and has an estimated global prevalence
of about 14 % [142 ]. Multiple factors contributing to the pathophysiology of migraine have been discussed,
with the trigemino-vascular system playing a major role in migraine with and without
aura [143 ]. In this context, central (including the hypothalamus, thalamus, brainstem, prefrontal
dorsolateral cortex, M1, and S1) and peripheral mechanisms (including the trigeminal
nerve and trigemino-cervical complex) may promote neurogenic inflammation: retrograde
trigeminal delivery of vasoactive mediators including calcitonin gene-related peptide
could trigger dilation of arteries and plasma exudation, which may perpetuate nociceptive
excitation of the trigeminal nerve endings surrounding vessels [143 ]
[144 ]
[145 ]. Pain triggering and processing mechanisms of the respective brain structures and
modulation of their in-between networks are likely to alter patterns of brain perfusion.
In addition, in migraine with aura, alterations in CBF can be observed in the context
of acute aura symptoms with cortical spreading depression [143 ]. Thus, ASL-based techniques have been used in various designs to examine cerebral
perfusion in migraine.
Some studies employed longitudinal designs to investigate changes in brain perfusion
over the migraine cycle [146 ]
[147 ]
[148 ]
[149 ]. Herein, one study demonstrated decreased CBF in the right hypothalamus, retrosplenial
cortex, and left visual cortex compared to healthy controls only pre-ictally, thus
potentially emphasizing changing perfusion patterns over the migraine cycle [147 ]. Another study demonstrated cyclical perfusion changes within the right nucleus
accumbens, right insular cortex, and right precentral gyrus, with perfusion increasing
leading up to the attack, while a superior parietal lobule cluster demonstrated perfusion
at its lowest during the attack and increasing afterwards [146 ]. Additionally, another study induced pharmaceutically triggered attacks in patients
with migraine and aura and observed regional CBF increases in the ipsilateral dorsolateral
pons (with respect to the most painful side) compared to baseline [148 ]. However, there are potentially conflicting results given that one study performed
scans in migraine patients without aura both during spontaneous attacks and in the
interictal state, revealing no difference in global or regional CBF between the two
conditions [149 ].
Other studies employed cross-sectional designs to investigate cerebral perfusion in
migraine [150 ]
[151 ]
[152 ]
[153 ]
[154 ]
[155 ]. Specifically, reduced CBF was described in the left nucleus accumbens of patients
with interictal chronic migraine and in the cerebellar vermis of patients suffering
from interictal migraine without aura [150 ]
[151 ]. Furthermore, elevated CBF was found in the right V5 and superior temporal gyrus
of patients with interictal migraine with aura, and in the right orbitofrontal gyrus
and middle frontal gyrus, as well as for the bilateral somatosensory cortex and left
primary motor cortex among patients with interictal migraine without aura [150 ]
[152 ]
[153 ]
[154 ]. Regional CBF differences between migraine patients with and without aura as well
as healthy controls (in the superior frontal gyrus, postcentral gyrus, cerebellum,
middle frontal gyrus, thalamus, and occipital cortex) have additionally been used
to establish a support vector machine classifier, which achieved an AUC of 0.86 for
differentiating migraine with and without aura [155 ].
Mostly due to its noninvasive nature, ASL-based perfusion imaging has also been used
in cases of pediatric migraine [156 ]
[157 ]
[158 ]
[159 ]. A series investigating 12 pediatric cases demonstrated changes in cerebral perfusion
corresponding to aura symptoms (mostly transient paresis), with a relationship between
the time to symptom onset and perfusion changes [156 ]. Specifically, the authors observed early hypoperfusion and later hyperperfusion
[156 ]. A similar pattern of initial hypoperfusion followed by hyperperfusion in aura-corresponding
regions was observed in a case-control study of 10 pediatric patients and matched
controls, albeit with differences regarding the timing of phase transition [159 ]. An analysis conducted in a larger cohort of 49 pediatric patients demonstrated
localized hypoperfusion in all cases scanned within 24 hours of symptom onset, while
patients scanned after a longer interval for the most part demonstrated normal perfusion
[157 ]. Another study in pediatric migraine patients with a median time-to-scan interval
of almost two hours after symptom onset demonstrated hypoperfusion matching neurological
symptoms in 14/15 cases [158 ].
Overall, findings currently show little overlap between studies for adult patients,
likely due to the inherent heterogeneity of migraine as a disease, as well as the
rather infrequent use of ASLbased techniques in migraine to date. In pediatric migraine,
however, multiple reports converge on the finding of cerebral hypoperfusion matching
aura symptoms in the early phase after symptom onset. Furthermore, ASL-based perfusion
imaging may also show promise with respect to investigating the specific pathological
mechanisms of migraine and their association with vascular and/or global or regional
perfusion alterations.
