| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
Review article |
1 Department of Neuroradiology, National Neuroscience Institute, 2 Singapore Bioimaging Consortium, A*STAR, Singapore
Correspondence: Xavier Golay, Singapore Bioimaging Consortium, Agency for Science, Technology and Research (A*STAR), Singapore, 138667. E-mail: xavier_golay@sbic.a-star.edu.sg.
 |
Abstract |
|---|
 Top Abstract IntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
|
Introduction |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
Historically, radiology has principally offered morphological imaging techniques; however, these new MRI perfusion techniques will add routine physiological imaging to its portfolio. This information is not only valuable in clinical settings, but is of great importance for basic physiological science and drug development, where, for example, perfusion changes due to a pharmacological stimulus might be monitored over time.
There exist two main MRI perfusion methods: bolus tracking after the injection of an exogenous endovascular tracer and arterial spin labelling (ASL), which uses magnetically labelled water protons as an endogenous tracer. The complete non-invasiveness of ASL makes it very suitable for perfusion studies of healthy volunteers and in patient groups requiring repetitive follow-ups. This is especially important in patients with particular conditions, such as kidney failure, or in paediatric populations where the use of radioactive tracers or exogenous contrast agents may be restricted. Recently, ASL has gained more clinical acceptance partly due to these advantages, but also as a result of the recently introduced technique for imaging the perfusion territory of individual blood vessels in the brain. In addition, the increased availability of high-field (
3 T) clinical scanners has moved ASL from the research and development stage towards the clinics.
The present article will focus on arterial spin labelling approaches. Different implementations and their advantages and disadvantages will be briefly reviewed. More attention will be given to the different ways of modelling the ASL data for perfusion quantification and the problems and pitfalls involved. Finally, research and clinical applications will be revisited.
|
Methodology |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
Quantitative perfusion
In order to measure tissue perfusion, one needs to follow the course of blood flow through the organ, and for this there exist two methods, one based on freely diffusible and the other on intravascular tracers. As their names suggest, freely diffusible tracers can leave the intravascular space without restriction and be distributed throughout the entire tissue volume, whereas intravascular tracers remain in the vasculature, which constitutes only a fraction of the full volume.
Early perfusion experiments carried out by the pioneers Kety and Schmidt [1] used inhalation of free diffusible nitrous oxide (N2O) as a tracer. This tracer distributes throughout the entire tissue volume, having a MTT in the order of minutes at normal physiological flow values. The relationship between the flow, distribution volume and the MTT is described by the central volume theorem, which states that the ratio between volume and flow is equal to the MTT. This rather long MTT allowed them to sample both arterial and venous blood, as arterial and venous tracer concentrations reached equilibrium after a few minutes. In this experiment, the time to reach this equilibrium is directly related to the cerebral blood flow using Fick's principle:
|
|
However, when a freely diffusible tracer is used, there is no access to the blood volume and an intravascular tracer would be needed in order to obtain this volume information. In this case, the distribution volume is equal to the blood volume and using the terminology for the brain, then according to the central volume theorem: MTT = CBV/CBF. This principle is used in MRI bolus tracking experiments, where an intravascular tracer such as gadolinium-DTPA allows measurements of MTT, CBF and CBV. Here, the first pass of the bolus is monitored, an arterial input function (AIF) selected and based on the indicator-dilution theory of Meier and Zierler [2] and, a model-independent perfusion estimation can be obtained [3]. Time-domain impulse functions are employed to describe and compute the tissue perfusion CBF, which is calculated using the deconvolution of the tissue concentration curve Ct(t) by the measured arterial input function Ca(t):
|
|
R(t) is the residue function describing the fraction of contrast remaining in the system after a given time, t. If, alternatively, a steady state experiment is used, i.e. letting the contrast distribute to the total body blood volume, one can measure CBV by comparing images from before and after contrast. For recent reviews on these methodologies, see Østergaard et al, Grandin, Grenier et al, and Barbier et al [4–7].
The theory behind the two methods described above provides the basis for all applied perfusion measurement techniques used today. They range from MRI to CT and nuclear medicine imaging, all capable of measuring perfusion and/or blood volume in different parts of the body, depending on the selected tracer and imaging technique.
Perfusion imaging encompasses physiological mass transport and exchange mechanisms, where the system is required to be stationary, linear and time invariant in order to satisfy the underlying flow quantification theory. This means that no major physiological alterations are allowed during the acquisition scheme in order to obtain quantitative perfusion. This should be kept in mind when planning an experiment, especially when dealing with functional studies for instance, where the perfusion rate is manipulated over time.
Basic arterial spin labelling
The overall goal of all existing ASL techniques is to produce a flow-sensitized image or "labelled" image and a "control" image in which the static tissue signals are identical, but where the magnetization of the inflowing blood differs. The subtraction control-label yields a signal difference 
M that directly reflects local perfusion because the signal from stationary tissue is completely eliminated (Figure 1
). The label is usually performed by inverting or saturating the water molecules of the blood supplying the imaged region. By adding a delay between labelling and image acquisition, called inversion delay (TI), the labelled blood spins are allowed to reach the capillaries where they exchange with tissue water and thereby give rise to the perfusion signal. The signal difference, which is only 0.5–1.5% of the full signal, depends on many parameters such as the flow, T1 of blood and tissue, as well as the time it takes blood to travel from the labelling to the imaging region. Multiple repetitions are needed for ensuring sufficient signal-to-noise, and a model of the perfusion signal is usually used in order to quantify the perfusion.
|
). The PASL approach labels a thick slab of arterial blood at a single instance in time, and the imaging is performed after a time long enough for that spatially labelled blood to reach the tissue and exchange at the region of interest (Figure 2b). Both methods need a control experiment in order to visualize and quantify the perfusion.
|
, which is important for further quantification, depends on factors like the mean velocity of the blood, angulations of the vessels to the plane and the selected RF amplitude and gradient strength. Typical labelling efficiency is in the range of 80–95% [8, 11–19]. Among the confounding factors of these long lasting inversion pulses are magnetization transfer (MT) effects [20]. When using a single coil for labelling and imaging, the off-resonance labelling pulse (with respect to imaging slice) will act as a powerful MT pulse in a way similar to an MT-weighted technique. The resulting saturation effect of the macromolecular pool will result in a reduced signal of the free water pool from the tissue of interest [21]. This is a very important issue, as the perfusion weighted images are calculated by subtracting a labelled from a control acquisition, and if this MT effect is present only during the labelling scheme, it will lead to overestimated perfusion.
In the first implementation, these MT-effects were compensated for by applying a distal labelling during the control experiment. This produces identical saturation effects but, due to the applied gradients during labelling, this is unfortunately valid only for a single slice. For multislice acquisition, Alsop et al [22] proposed the use of two closely spaced inversion planes, also called double adiabatic inversion (DAI). In the control experiment, the magnetization gets inverted while traversing the first plane and returns theoretically to its original state during the passage through the second plane (the CASL experiment is shown in Figure 2a
). Double inversion is achieved by applying a sinusoidal modulation of the RF waveform. Global control of the MT-effects is obtained by matching the RF power and the location of the planes. Figure 3 shows a perfusion map obtained using this method. Another method was also proposed, called simultaneously proximal and distal RF irradiation (SPDI) for multislice acquisition [23]. In this later scheme, the RF power on the control scan is distributed evenly on both sides of the acquisition volume. However, a big limitation of these approaches is the doubled RF deposition, resulting in higher specific absorption rates (SAR). This is particularly important at higher field strength, and must be carefully considered for human studies at 3 Tesla and above.
|
Pulsed arterial spin labelling
In 1994, Edelman et al [31] proposed the first pulsed ASL scheme. Contrary to CASL, the labelling is performed once in a 10–15 cm slab proximal to the image slices. For the PASL sequences, MT-effects have to be considered as well, although these are much smaller compared with CASL. In this first version of the "Echo-Planar MR Imaging and Signal Targeting with Alternating Radio frequency" (EPISTAR) sequence, inversion was performed distal to the image slice during the control experiment to induce identical MT effects in both cases. Again, this truly compensates for a single slice only and, therefore, the sequence was modified for multislice acquisition using a single 180° adiabatic pulse for the label experiment and two 180° + 180° pulses of half the power for the control experiment at the same proximal location (Figure 2b) [31, 32]. Figure 4 shows a multislice experiment using this technique.
|
PASL allows inversion of the arterial spins closer to the image slices and the inversion efficiency is improved; however, due to imperfect inversion profiles, a gap between the labelling region and the image slices of typically 1–2 cm is needed, depending on the type of RF pulse used. This increases transit time 
a from the labelling slab to the imaging slices leading to decreased efficiency. In addition, T1 relaxation of all the inverted spins will also result in a theoretically lower signal difference. Nevertheless, the ease of implementation and reduced practical problems, as compared with CASL, have made PASL a popular choice for perfusion imaging. This is reflected in the wide range of sequences available today, the common ones being listed in Table 1. For further in depth explanation of these sequences, see recent ASL reviews [4, 35].
|
). The relationship between the M signal and the actual CBF depends mainly on proton density and T1 relaxation rates of tissue and inflowing blood, the and their respective differences. In addition, the label transit time from the inversion slab to the observed region in the images is also an important factor. Traditionally, quantitative CBF estimation is carried out using the tracer clearance theory originally proposed by Kety and Schmidt [1], which was first adapted to ASL experiments by Detre and Williams et al [8, 9]. In the original model, it is assumed that the labelled arterial blood water is a free diffusible tracer, implying that the exchange of blood water with tissue water happens instantaneously upon its arrival to the parenchyma. Therefore, this model corresponds to a single compartment tracer kinetic, described by a mono-exponential tissue response function. The modified Bloch equation, including the flow dependent exchange term, becomes:
|
|
is the blood–brain partition coefficient, and T1t is the longitudinal relaxation rate of the tissue. In the original quantification model, further assumptions about uniform plug flow and equal T1 relaxation of both tissue and arterial blood were made [8, 9].
Many derivatives and improved versions of this first solution exist for both CASL and PASL experiments. Calamante et al [36] took the difference in R1 relaxation rate for tissue and arterial blood into account as well as the trailing edge d of the bolus. In the work of Kwong et al [33], the transit time a was also considered.
Buxton et al [37] proposed a general kinetic model where all the above mentioned parameters can be taken into account. Here, the magnetization difference between labelled and control measurements is described using the convolution integral in a way similar to Equation (2):
|
|
) describes the washout of labelled spins from a voxel, and m(t–) includes the longitudinal magnetization relaxation effects. The possibility to choose a particular arterial input function and to consider a certain exchange mechanism (single or multicompartment) allows greater flexibility for data analysis of both CASL and PASL experiments.
However, analytical solutions are only possible with simple assumptions, and the most widely used "Standard Model" can be summarized using the following terms:
|
|
Again, the assumptions are: uniform plug flow and fast exchange, which is equivalent to single compartment kinetics. Solving Equation (5) using Equation (4) in the case of a PASL experiment gives a stepwise defined equation:
|
|
R = R1a – R1app and R1app = R1t + CBF/, also called the apparent tissue relaxation rate. A similar set of equations can be obtained for the CASL experiments, with the only difference being that c(t) in Equation (5) is constant for CASL, whereas it is subject to T1a decay for PASL, since the latter is not a steady-state method. As can be seen, various parameters like the transit time a, blood–tissue partition coefficient , M,0, R1a and R1t need to be estimated or measured in order to obtain quantitative CBF values. The difference between the many ASL sequences is mainly in the measurement (or not) of these parameters. Different issues relating to the actual perfusion quantification, using a single or multiple inversion time points, and the possible errors introduced using the standard model vs more advanced methods, will be discussed next.
|
Quantification errors |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
a, which even in healthy subjects differs across the brain, being longest in distal branches, especially in the regions between perfusion territories also known as border-zone areas. In most PASL approaches, information on perfusion is assessed at a single inversion time point (Figure 5), and therefore without information about the transit time. Quantification is then based on the second step in Equation (6), either where a is assumed equal all over, or simply set to zero. The problem in doing this is that not only is the quantification of CBF impossible, but relative perfusion values between regions are not valid either. Sequences like QUIPSS II and Q2-TIPS [38, 39] were developed to render ASL more transit-time insensitive. Constant bolus duration is assured by saturating the part of the label that remains within the labelling slab at a time delay short enough to be able to intersect the trailing edge of the fastest blood. Appropriate timing of this saturation and the following image acquisition makes these sequences insensitive to the transit time. This is really an advantage in volunteers and patients without vascular diseases where the difference in transit time is small (<1.5 s) [40], as well as for quantitative functional studies where it is known that the transit time changes are small (0.1–0.2 s) between baseline and activation conditions [41, 42].
|
M curve (Figure 6). The standard sequences, in which a single time point is acquired at a time (2–4 min per time point) are generally too lengthy to be suitable for clinical examinations [37, 41]. Günther et al [43] introduced an elegant solution to this problem using a Look-Locker-like readout to measure the ASL signal at multiple inversion times in a single scan (Figure 6). Because of multiple low flip angle readouts, the general model needs to be modified by substituting R1app in Equation (6) with R1app,eff = R1t + CBF/ - ln(cos
)/TI, where is the flip angle and TI is the interval between the excitation pulses. A similar scheme was recently implemented using the transfer insensitive labelling technique (TILT) [44] by Hendrikse et al [45].
|
Vascular artefacts
Vascular artefacts, associated with the inflow of labelled arterial blood into the arteries, can introduce important errors in CBF quantification. Strictly speaking, for Equation (5) to be valid, none of the M signal should originate from within the arterial vasculature. This assumption is often violated in voxels containing feeding vessels or traversing arteries, resulting in a substantial overestimation of perfusion values. Ye et al [47] proposed the use of bipolar crusher gradients to eliminate the signal from the large feeding arteries in CASL experiments. This was also adapted for PASL sequences. Another solution is to choose a sufficiently long inversion time, allowing the feeding vessels to empty before acquisition [40, 46–48].
Inversion pulse shape and efficiency
In PASL sequences, where a spatially defined label is used, the shape of the inversion pulse is of great importance. Ideally, the profile should be truly rectangular, allowing zero spacing between the labelling slab and the imaging region; due to finite duration of the RF pulse, this is not realizable in practice. Imperfect profile can reduce labelling efficiency, but, more importantly, there can be contamination in the imaging region from the labelling slab. To avoid this, a gap between the inversion and image plane is often introduced, with increased transit times as a result. In order to minimize this gap, longer adiabatic pulses like the hyperbolic secant (HS) [49], frequency offset corrected inversion (FOCI) [50] or bandwidth-modulated adiabatic selective saturation and inversion (BASSI) [51] pulses are often used. Another approach, as implemented in the TILT sequence [44], is the self-refocusing concatenated 90° Shinar-Leroux pulse combination that maintains the profile efficiency of the 90° pulses [52].
For CASL labelling, where labelling is performed in general at a distance from the imaging slices, the "profile" is less of a concern. It is the fulfilment of the adiabatic condition ensuring proper inversion that can be problematic [9, 15, 16, 22, 53].
Signal to noise issues
At normal perfusion rates of 40–100 ml 100 g–1 min–1 the signal change M is in the order 0.5–1.5% of the full signal. Therefore, an average from typically 30 to 40 pairs of subtracted control and labelled images are required to get the desired signal to noise ratio (SNR) in the perfusion-weighted maps. A total scan time of 3–4 min is needed in order to acquire these data, which makes the technique very sensitive to motion artefacts. Proper head fixation and collaboration of the subject are necessary in order to obtain good results. Fast imaging techniques like single shot EPI and spiral sequences are often used to reduce the scan time between successive control-label pairs as well as overall scan time. Also, prior saturation of the image plane reduces the sensitivity to motion. In case of limited motion, realignment of the images can improve quantification, but in general it is compromised because of low resolution, small number of slices, and small M signal.
Finally, the running state of the scanner is important. Having a perfusion signal M at only a fraction of the full signal, a slight drop in the scanner performance can preclude acquisition of clinically relevant ASL images, even if no effects can be detected in the standard clinical images.
Fast vs intermediate water exchange
The complex nature of the brain vasculature might not act like a single compartment to water based tracers. In particular, the blood–brain barrier seals the vasculature from the extravascular space, allowing a dynamic interface that protects the brain from toxic substances while allowing nutrients and other essential compounds to pass and thereby maintain its homeostasis. The "free diffusible" water is known to be limited to "free" diffusion through the lipid membrane of the endothelial cells as well as through dedicated water channels [54, 55], whereas diffusion in between adjacent endothelial cells is impossible due to tight junctional complexes [56].
The speed of the exchange in comparison with the 2–5 s duration of the "control" or "label" experiment still remains controversial. At higher flow rates, the exchange appears restricted [54], as though the water channels have been saturated. In order to deal with these effects, a few groups have suggested more elaborate multicompartmental approaches [57–59]. In general, the conclusion is that these effects are negligible at normal human perfusion rates at 1.5 Tesla, whereas the effect might be more pronounced when going toward higher field strengths. However, the intrinsic low SNR of the ASL techniques combined with the problems related to fitting all these additional exchange parameters will keep these models within the animal experimental world for at least another few years.
Bolus dispersion
Even when using vascular crushing and modelling assuming multicompartmental behaviour, it still leaves another possible error source, namely the dispersion of the labelled bolus. In a PASL experiment the label is close to being rectangular at the time of labelling and if the blood does not experience any resistance, this shape would persist all the way to the image slices. In reality, the blood friction with the vessel wall, resistance in bifurcations and the pulsatile behaviour of the flow make the flow profile look like something between a parabolic and ideal plug flow profile [60]. This results in dispersion of the bolus while it travels from the labelling slab to the imaging region, which will lead to underestimation of the perfusion.
Various attempts to incorporate these effects in the modelling have been made, and Hrabe et al [61] recently provided two analytical solutions that might reflect the physiology better. While the dispersion might be fairly homogeneous in healthy volunteers, patients with vascular diseases can be expected to have completely different dispersion in affected areas, and the flow estimation using these advanced models would still be non-quantitative.
Blood–brain partition coefficient
When a tracer consists of water or is dissolved in water, it is necessary to know the ratio of water in the tissue of interest and the feeding blood water in order to correct the distribution volume and thereby obtain the correct perfusion values. In the ASL technique, this is more related to the proton density, which tells us about how much signal we can expect from different tissues.
This ratio is called the blood–brain partition coefficient , and it was first defined by Kety [62] and corrected by Herscovitch et al [63] to the commonly used values today: whole brain = 0.9 ml g–1, grey matter g = 0.98 ml g–1 and white matter w = 0.82 ml g–1.
The relevance of this coefficient is controversial; first of all because the valid use of it requires instantaneous exchange of the tracer, which, as earlier mentioned, can be questionable, but also due to the fact that Roberts et al [64] demonstrated that this coefficient not only varies between different tissues, but also from region to region. Nevertheless, a whole brain value of 0.9 ml g–1 is commonly used for the calculations. Notice the unit of ml g–1, which means that the unit of CBF/ in Equation (5) becomes s–1 when CBF is expressed as ml g–1 s–1.
Blood equilibrium magnetization
The inflowing blood has a proton density different from the tissue. For absolute quantification, the equilibrium magnetization of the arterial blood Ma,0 is needed, i.e. the available longitudinal magnetization from a fully relaxed blood filled voxel. This could be obtained from a partial volume-free arterial voxel, but with a standard in-plane resolution of 3–4 mm this is not possible in practice. Alternatively, this information can be taken from the much larger sagittal sinus. However, the T2* of the deoxygenated venous blood, especially at higher field strength, is shorter than that of arterial blood. This results in a underestimated Ma,0. A different approach is to measure the equilibrium magnetization in a grey or white matter region and estimate Ma,0 using the blood–brain partition coefficient. For single inversion time experiments, this is often done on a voxel by voxel basis using the control experiment as Mt,0-map after correction for TR and T1.
Note that Ma,0 is a direct scaling factor of the perfusion (Equation (6)) and therefore an error in the product ·Ma,0 will directly change the calculated CBF value.
Functional studies and non-steady state perfusion
Finally, a requirement for using the various models and formulae for perfusion quantification is that tissue perfusion is a stationary, linear and time invariant system. This would most often be satisfied in standard perfusion scans where the subject is resting. However, ASL has gained more and more popularity in functional studies where perfusion is monitored during various stimulation paradigms. Here, a typical experiment would have a repetition time of 2–3 s, while 4–6 s would be needed in order to acquire both control and label. Now, in functional experiments, the haemodynamic response time is often in the same range, which results in a violation of the steady state requirement [65, 66]. Not only will the flow change in between the control and label experiment, but also in between labelling and acquisition.
Two main problems arise from this; first, the models do not assume increasing or decreasing perfusion during the experiment and calculation errors will occur, and second, and probably the most important, is that T2* will change due to the BOLD effect, which will reduce the signal intensity changes when acquiring at typical echo times of 15–25 ms.
In the case of a standard block paradigm, where the rising edges are discarded in the statistical analysis, this is not a problem. With basic research moving towards the characterization of the haemodynamic response on the other hand, the information in the rising and the falling edges are especially important. Recently, Lu et al [67] proposed a new subtraction method for reducing the T2* effect, which is very suitable for this type of experiments, however, without addressing the problem of non-steady-state regimen.
Comparison with other imaging modalities
At this point, it should be kept in mind that related issues such as the ones highlighted in the previous paragraphs apply to other perfusion modalities like DSC-MRI, CT-perfusion and positron emission tomography (PET) as well. Dynamic susceptibility contrast using tracers such as gadolinium-DTPA for instance, suffers from the fact that the relationship between measured signal and contrast is non-linear and depends on parameters such as the field strength, shim of the magnet and the constitution of the vessels [68]. In addition, correct scaling of the AIF is influenced by partial volume effects in the voxels from where it is measured. Altogether, this makes quantification troublesome, and in the general clinical practice only a relative perfusion measure is possible. Similar problems exist in CT-perfusion, mainly related to the extraction of the global AIF [69]. PET can be said to be a more "pure" method for measuring perfusion due to the use of free diffusible tracers and experimental durations that ensure a steady state. However, a relatively low resolution of typically 6–10 mm introduces partial volume effects, i.e. a mixture of grey-matter, white-matter and CSF will be present in almost all voxels, making direct comparisons to the abovementioned methods questionable.
|
Applications |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
Although most publications on ASL to date have focused on brain perfusion, other organs like the lungs, kidneys and the heart have recently gained attention with improved techniques and hardware available. In particular, the move toward 3 T high field scanners in the standard clinical environments seems to push this method from the research and development stage towards clinical applications.
The various applications of ASL have recently been extensively covered elsewhere [4, 35] and only a short summary of the active fields and most recent applications will be listed here. Two main categories exist – the basic science, mainly neuroscience, and the general clinical use.
|
Neuroscience |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
For pharmaceutical validation, the suitability of ASL for prolonged functional studies is advantageous for studying perfusion-altering drugs. Alternative techniques often require the injection of a tracer, which cannot be repeated more than a few times and would possibly interfere with the drug itself. Among investigated pharmaceutical agents are the vasodilators acetazolamide [80, 81] and 2-chloradonosine [82, 83], as well as the vasoconstrictor indomethacin [84, 85] and finally isoflurane, which is used for anaesthesia [86, 87]. This field can be expected to expand with broader awareness of ASL and availability of the sequences used.
Clinical
In the clinics, an important issue is how well ASL performs when compared with established "gold-standard" methods. Validation studies have been carried out in animals [57, 88, 89] using radioactive microspheres and 14C-iodoantipyrine autoradiography, as well as in humans [90] comparing ASL and PET. In general, good correlation is observed for grey matter, whereas white-matter often shows an underestimation of CBF. This is mainly attributed the prolonged arrival times and low perfusion values of white matter, which is in the lower range of measurable flow values using ASL.
The benefits of ASL particularly suit the requirements of the paediatric population, where there is a need to avoid ionizing radiation as procedures are likely to be repeated for disease monitoring and venous cannulation can be highly traumatic. In addition, the general higher blood flow seen in children increases the difference signal M and thereby the perfusion SNR [91], making ASL a very promising tool for paediatric perfusion studies [92].
Another particularly appropriate clinical arena is that of cerebrovascular diseases. In fact, stroke is the third cause of death behind cardiovascular diseases and cancer. Consequently, a lot of resources are put into the research of this disease – how to prevent it, the immediate therapy following stroke and post-stroke rehabilitation procedures. In recent years, ASL techniques have become an alternative tool to CT and bolus tracking MRI for studying the mechanisms of stroke and the underlying processes of brain damage resulting from ischaemia. Studies of the evolution of acute stroke and validation of outcome predictors for the identification of potentially salvageable tissue have been performed in animal models [93–96] and patients [97, 98] using both PASL and CASL. For the assessment of the different perfusion territories and eventually collateralized flow in steno-occlusive patients, a range of ASL sequences has been made capable of labelling individual perfusion territories of the major feeding vessels one at a time [24, 27, 28, 32, 99–102]. An example of regional perfusion imaging of a healthy volunteer is shown in Figure 7. These techniques are very promising in the evaluation of the successful recruitment of subsidiary blood vessels, which are believed to play an important role in the clinical outcome of patients with cerebral artery occlusion. Until recently, diagnostic strategies to evaluate the collateral circulation could be divided into those that directly visualized collateral blood vessels, such as conventional X–ray angiography and indirect methods that assess tissue perfusion, including acetazolamide-challenge tests of cerebrovascular reserve. This new class of ASL techniques makes the combination of both methods possible for the first time and quantitative information of the flow territories supplied by each major brain vessel is obtainable. As such, these methods could become alternatives to conventional X–ray based subtraction angiography, which is so far the only modality giving temporal as well as spatial regional blood flow information. Using these techniques, Van Laar et al [103] showed a wide interindividual variability in the perfusion territories caused mainly by anatomical variations of the circle of Willis in a population study of 115 healthy ageing volunteers.
|
Finally, ASL has also been applied to organs other than the brain, although the majority of research and applications have focused on the latter. Applications range from combined lung perfusion using ASL and ventilation scans using hyperpolarized helium-3 for evaluation of pulmonary diseases [108–111], to cardiac [112–114] and renal [6, 115–119] perfusion imaging. However, these techniques are very challenging, mainly due to non-rigidity and increased movement of these organs as compared with the brain.
|
Conclusion |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
Despite the remarkable progress, ASL has still not overtaken traditional invasive methods in the clinics. One reason for this is the intrinsically low SNR, making averaging necessary, and therefore good cooperation from the subject is essential. This is not always the case with, for instance, stroke patients or children. Nevertheless, the great improvements in hardware and the increasing availability of high field clinical scanners pre-installed with ASL sequences seem to put more focus on these techniques. Robust sequences for regional perfusion imaging and functional studies of either cognitive or pharmacological nature are also gaining more and more interest, and the complete non-invasiveness of these techniques will always be an attractive asset.
|
Acknowledgments |
|---|
Received for publication December 15, 2005. Accepted for publication March 29, 2006.
|
References |
|---|
|
TopAbstractIntroductionMethodologyQuantification errorsApplicationsNeuroscienceConclusionReferences |
|---|
