Cerebral blood flow measurement with arterial spin labeling MRI: An underutilized technique
Correspondence Address: Source of Support: None, Conflict of Interest: None DOI: 10.4103/0028-3886.263224
Source of Support: None, Conflict of Interest: None
Measurement of cerebral perfusion is integral for understanding brain pathophysiology. With recent advances in neuroradiology, several techniques are available for the measurement of cerebral perfusion of which ASL (arterial spin labeling) MRI is unique. ASL is a non-invasive technique which can measure cerebral blood flow without any contrast injection or radiation exposure. The technique was developed in the 90s, but did not gain popularity due to the low resolution of images or low signal-to-noise ratio (low SNR). Now, with tremendous improvements in MRI scanners and advances in the ASL technique, clinically relevant research in ASL is gaining popularity.
ASL aims to assess the tissue perfusion rate, which is different from the macrovascular blood flow. It provides quantitative blood flow maps by following magnetically labeled blood water molecules from the arterial compartment to the capillary bed. Labeling is done by magnetic inversion of the water protons in the feeding arteries, e.g., the neck region. The labeling plane is kept below the tissue of interest. Following the labeling, time is allowed for the labeled blood protons to reach the tissue of interest. This is called the post-labeling delay (PLD), after which image acquisition is started, to obtain signals from the labeled protons inflowing from the feeding arteries [Figure 1].
Three labeling methods are available: continuous ASL, pseudo-continuous ASL (PCASL) and pulsed ASL. Three dimensional PCASL has a higher labeling efficiency, a better SNR, lesser artifacts and a higher reliability over other methods.
Several physical and physiological factors influence the quality of ASL images, of which labeling efficiency and post labeling delay (PLD) are the most relevant. The choice of the optimum PLD remains a challenge, as it is highly dependent on the presumed blood velocity 
Most clinically available ASL sequences use a single time delay between the labeling and image acquisition, based on values calculated from healthy young adults. However, since optimal PLD depends on blood velocity, it will vary between individuals and in disease conditions. For instance, in cases of proximal vessel occlusion, there will be delayed arrival of blood in the brain parenchyma, hence a longer PLD would be needed to accurately measure the CBF. Reduced cardiac output in elderly individuals would also have a similar effect.
In this article Hu et al., have well emphasized that when PLD equals the arterial transit time (ATT), the measured perfusion value would be accurate. PLD is a scanning parameter referring to the time interval between blood labeling and image acquisition, whereas ATT is a parameter referring to the time taken for arterial blood to flow from the labeling level to the tissue of interest. If PLD is less than the ATT, falsely low CBF values will be obtained, because all the labeled blood would not have not reached the tissue of interest.
Grade et al., recommended PLD of 2000 ms in neonates, 1500 ms in children, 1800 ms in healthy adults <70 years age, and 2000 ms in healthy adults >70 years age. In this study, Hu et al., explored the optimal post labeling delay (PLD) of ASL in different age groups and the correlation between CBF and age in adults. Based on the hypothesis that blood flows faster in younger individuals, they used shorter PLDs (1025 ms, 1525 ms and 2525 ms) in youth (20-44 years) and found that most brain lobes had a higher mean CBF with shorter PLDs; 1525 ms being the best. For this age, a PLD of 2525 ms was probably too long, so excessive arterial relaxation occurred resulting in a falsely low CBF value.
In the middle aged (45-59 years), and elderly subjects (60-80 years) they used PLDs of 1525 ms, 2525 ms and 3025 ms and found that the mean CBF was higher with longer PLDs; 3025 ms being the best. In this age group, since the blood flow is relatively slow, a shorter delay time of 2525 ms was probably too early to measure CBF, as all the labeled blood would not have reached the acquisition level, resulting in underestimation of CBF.
Hu et al., also found that CBF of global grey matter and all brain lobes of the elderly was lower than that of the youth. All brain lobes showed a statistically significant difference between the youth and middle aged groups; and no statistical difference between the middle-aged and elderly groups. This could be explained by the fact that CBF decreases with age and gradually reaches a plateau after middle age (>44 yrs). Several factors have been implicated for the decrease in CBF with increasing age; loss of brain substance resulting in decreased brain metabolism; and, a progressive decrease in the number of cortical neurons with age.
The age-related changes found in the study by Hu et al., were highly consistent with the study by Soni et al. Biagi et al., observed that CBF continued to increase during 4-10 years, decreased rapidly during adolescence and reached a plateau at 25-30 yrs. The possible explanation for a high CBF during the development phase could be due to the excess production of neurons, synapses and dendritic spines, achieving a maximum value by 12 yrs of age after which there is a plateau.
In all prior studies, it has been consistently observed that CBF decreases with increasing age; nevertheless it is possible that the measured CBF could be lower than the actual one because of the use of single PLD (which may be inappropriate for age) resulting in underestimation of CBF.
To overcome this potential pitfall, multi-delay ASL imaging can be used to improve the accuracy of CBF measurement and additionally permit the imaging of multiple hemodynamic parameters. But these sequences require a longer imaging time.
In this study, Hu et al., performed multiple acquisitions at multiple delays in each subject to find out which PLD yielded the highest CBF for that age group. The acquisition times of each ASL sequence was 4.08 to 5.16 min (varying according to the PLD employed) resulting in the total scan duration of approximately 14 min per subject.
Wang et al., used a 4-delay PCASL protocol in their study within a single acquisition in a scan time of 4 min 30 sec. The advantage of their protocol was measurement of multiple perfusion parameters (ATT, CBF and arterial cerebral blood volume [CBV]) simultaneously within a time equivalent to a single delay ASL scan. Assessment of collateral perfusion information was also possible with this protocol. Another interesting observation was that CBF calculated by their protocol showed only a slight improvement compared to CBF acquired at a single PLD of 2 seconds. The PLD of 2 s is used as a benchmark based on ASL studies in cerebrovascular disorders and stroke by Qui et al. A single delay PCASL scan is more advantageous in acute stroke where CBF is the primary parameter of interest and the scan time needs to be short (can be acquired in 1 min). A 4-delay PCASL would be more useful when additional hemodynamic parameters need to be measured including ATT and CBV, which may be more sensitive in characterizing chronic stroke (Mackintosh et al.) and brain tumors (Law et al.)
Other approaches directed towards overcoming the problem of inappropriate PLDs have also been suggested. According to Monet et al., a dynamic magnetic resonance angiography (MRA) sequence based on ASL obtained early, when the labeled protons are still intravascular, could help in determining the time delay of blood proton arrival for each vessel.
ASL measurements are likely to be affected by other factors also such as end-tidal CO2, heart rate, respiratory rate, wakefulness, and the use of caffeine. A short pre-scan to determine the optimal PLD before the actual ASL acquisition in individual subjects would be an ideal approach. This has been tried by using a dynamic sequence or an MRA sequence; however, such sequences are not widely available in routine MR scanners.
Current and potential clinical applications of ASL
Cerebrovascular diseases: ASL has an important role in the assessment of cerebral perfusion in both acute as well as chronic cerebrovascular disease. It can be used even in patients with contraindications to the administration of an intravenous contrast. Moreover, the quantification of CBF by ASL is simpler compared to the contrast-based perfusion techniques. Vessel selective ASL has the additional ability to identify perfusion changes induced by the collateral flow. ASL can be applied to assess patients with carotid stenosis and Moyamoya disease.
Arteriovenous malformation (AVM), and fistula (AVF): Normally, ASL signal is not seen in venous structures. However if the capillary bed is absent, as in AVM and AVF, ASL signal can be seen in venous structures. ASL can be used to diagnose small lesions or recurrence of lesions after embolization. Even in the setting of intracerebral haemorrhage, ASL can show the AVM when conventional angiograms are negative. The conspicuity of these lesions, however, heavily depends on the precise timing of imaging.
Neoplasms: Haemodynamic changes are present in many neoplasms, and in general, cerebral blood volume (CBV) increases with tumor grade. ASL cannot measure CBV but it can measure CBF. Glioblastomas are the most common high grade tumours and are associated with a high metabolism and CBF. ASL is superior to dynamic susceptibility contrast based (DSC) perfusion and fluorodeoxyglucose (FDG)-positron emission tomography (PET) scan in differentiating radiation necrosis from tumour recurrence in patients on follow up showing a new contrast enhancing region or an increase in the size of lesion. Radiation necrosis typically shows a reduced CBF while recurrent tumours show an increased CBF [Figure 2].
Neurodegenerative disorders: Due to the close link between brain metabolism and perfusion, patterns of FDG PET closely resemble ASL; hence, it is implied that known patterns of hypometabolism in FDG PET in dementia can be transferred to patterns of hypoperfusion on ASL. Brain metabolic alterations precede structural abnormalities during cognitive decline in Alzheimer's dementia. Since structural MR imaging is done for these patients, addition of ASL sequence in the MR protocol can potentially avoid the need for FDG PET.
Epilepsy: ASL can locate a potential epileptogenic focus. During the acute ictal phase, CBF is typically increased due to the pathological neuronal activity, while in the inter-ictal period, CBF is reduced as the epileptogenic focus is less functional and active.
Depression and psychosis: Specific patterns of CBF alterations have been observed in different subtypes of depression and in patients with schizophrenia and this is a topic of ongoing research.
Its potential utility in post-traumatic stress disorders (PTSD), mild traumatic brain injury, pain and migraine is also being explored.
Extensions of ASL: Super-selective ASL mapping of vascular territories may be used to assess the impact of vascular variants of circle of Willis on brain perfusion.
ASL can be used to assess the cerebrovascular reserve, similar to a stress test using an acetazolamide challenge.
In this study, the authors have done a good job of re-generating interest in the ASL technique, and of highlighting the importance of adjusting PLDs according to the age of subjects.
Another interesting observation of this study is that a PLD of 2525 ms seems to be inappropriate for all age groups. According to this study, the appropriate value of PLD is 1525 ms up to the age of 44 years and 3025 ms at 45 years and above. These PLD values may be used as a rough guide for acquisitions in some clinical situations. Multi-delay ASL holds promise, provided the problem of reduced spatial resolution is solved without an increase in scan time.
However, even in the absence of known cerebrovascular disease, hemodynamics of subjects cannot be predicted with accuracy. Hence, the need for estimating and optimizing PLD on an individual basis remains important; probably taking into account dynamic variables such as heart rate, blood pressure, and respiratory rates. Perhaps MR sequences which can optimize PLDs in individual subjects would give the best results.
[Figure 1], [Figure 2]