ASL: Data Acquisition

Jun Hua, Ph.D.

F.M. Kirby Research Center for Functional Brain Imaging, Kennedy Krieger Institute

Department of Radiology, Johns Hopkins University School of Medicine

A. BASIC PRINCIPLES

Arterial spin labeling (ASL) uses the magnetization of water protons in the arterial blood stream as an endogenous tracer for perfusion measurements (Detre et al., 1992; Williams et al., 1992). Any perturbation to the magnetization of the arterial blood that feeds the tissue can serve as a magnetic tracer. The perturbation is typically introduced by an RF pulse at a location proximal (i.e., upstream) to the tissue of interest. After a time delay that allows the magnetically labeled arterial blood to reach the tissue capillary bed, the labeled water molecules exchange with the water molecules in the tissue, causing a change in the MR signal. At this point, a pulse sequence (readout) is played to acquire an image at a location distal (i.e., downstream) to the labeling position. The spatially resolved signal change is measured by subtracting this labeled image with a second image acquired without labeling (the control image). The difference signal between the labeled and control images is fitted to a quantitative model, from which maps of cerebral blood flow (CBF) and vascular transit times can be obtained.

B. LABELING SCHEMES

Various approaches have been developed for labeling arterial blood spins. In general, they can be categorized into the following groups:

1. PULSED

In pulsed ASL (PASL), an RF inversion pulse is applied to produce a bolus of labeled magnetization at a labeling location. The bolus travels from the artery to the capillary bed and exchanges the labeled magnetization with the unlabeled magnetization of the tissue water. Pulsed ASL is a family of pulse sequences, all sharing the same basic principles, but differing from one another in the strategies by which the labeled and control images are acquired.

a) STAR

The echo planar imaging and signal targeting with alternating radiofrequency (EPISTAR) approach (Edelman et al., 1994) is one of the first ASL sequences developed in this category. The labeling pulse sequence begins with a 90° slice-selective saturation pulse applied at the location where the perfusion measurement is sought. This pulse saturates the spins in the imaging slice, providing some immunity to any perturbation that can be caused by the subsequent labeling pulse. Following the saturation pulse, a spoiler gradient is typically used to dephase the magnetization. Next, a spatially selective inversion pulse (the labeling pulse) inverts spins within a thick slab proximal to the imaging slice. At the end of the labeling pulse, a long delay time is introduced to allow the inverted arterial spins to travel from the labeled slab to the imaging plane and to perfuse into the tissue. Because the labeled arterial blood carries inverted magnetization, it causes a signal reduction in tissue. The reduced magnetization is imaged using an EPI readout. The control image without arterial labeling can be acquired by simply repeating the sequence without the labeling pulse. However, the labeling pulse is an off-resonance irradiation with respect to the imaging slice, resulting in a magnetization transfer (MT) effect in the labeled image that masks the subtle signal change due to perfusion. To balance the MT effect, an inversion pulse identical to that in the labeling sequence is also played prior to the acquisition of the control image, except that its carrier frequency is chosen to place the labeling slab distal to the imaging plane by an equal distance.

Several variants of this approach have also been developed. For instance, the STAR-HASTE method (Chen et al., 1997) uses a RARE readout instead of EPI. Proximal inversion with a control for off-resonance effects (PICORE) (Wong et al., 1997) is another variation of EPISTAR. The labeling sequence is the same as that in EPISTAR, but an off-resonance inversion pulse in the control sequence is played without an accompanying slab-selection gradient. The carrier frequency of the inversion pulse is the same between the control and the labeling pulse sequences, so the MT effect can be subtracted out. The magnetization of the imaging slice is virtually unperturbed due to the large resonance offset. Compared to EPISTAR, the difference signal in PICORE rejects the inflow from the distal side of the imaging slice, whereas the inflow causes a reduction of the difference signal in EPISTAR. Another advantage is that asymmetry in MT effects is compensated in PICORE, but not in EPISTAR. A disadvantage of PICORE is, however, that it is less robust against eddy-current effects than EPISTAR because the control sequence uses a different gradient waveform from the labeling sequence. Other variants in this group include QUIPSS (Wong et al., 1998), TILT (Golay et al., 1999), DIPLOMA (Jahng et al., 2003), PULSAR (Golay et al., 2005), QUASAR (Petersen et al., 2006) and others.

b) FAIR

The flow-sensitive alternating inversion recovery (FAIR) (Kim, 1995; Kwong et al., 1995) ASL sequence employs a frequency-selective inversion pulse with and without an accompanying slice-selection gradient to produce the labeled and the control images, respectively. Unlike EPISTAR, the inversion pulses for the control and labeled images have the same carrier frequency. When the pulse is played with the slice-selection gradient, it inverts the spins within the imaging slice while leaving spins elsewhere virtually unaffected. After the pulse, an optional spoiler gradient can be applied, followed by a delay time, just as in EPISTAR. Finally, a single-shot EPI readout is played to produce the labeled image. To acquire the control image, the slice-selection gradient can be either played with zero amplitude or played at a different time away from the slice-selection gradient. The latter approach uses the slice-selection gradient as a spoiler (Kwong et al., 1995). By keeping the same gradient waveform, it also allows better cancelation of eddy-current effects between the labeled and the control images. In the absence of an accompanying slice-selection gradient, the inversion pulse inverts spins in the entire volume of the transmitting RF coil. Because both the arterial blood and the tissue experience similar inversion recovery, and the T1 of arterial blood is only slightly longer than that of gray matter, there is virtually no sensitivity to arterial inflow in the control image. FAIR is more robust against the MT effect because no off-resonance irradiation is applied with respect to the imaging slice. This facilitates multi-slice imaging. Another advantage of FAIR over EPISTAR is that arterial blood feeding the tissue from both proximal and distal sides of the imaging slice is labeled. When the flow direction is unknown or when the feeding arteries have tortuous paths, FAIR reduces underestimation of perfusion, because inflows from both directions are registered and contribute to the difference signal.

Several variations of the FAIR pulse sequence have been developed. In a technique called uninverted flow-sensitive alternating inversion recovery (UNFAIR) (Helpern et al., 1997; Tanabe et al., 1999) or extraslice spin tagging (EST) (Berr and Mai, 1999), two consecutive inversion pulses with the same carrier frequency are used in the pulse sequence. To acquire the labeled image, one of the inversion pulses is played with a slice-selection gradient to selectively invert the magnetization of the imaging slice while the other inversion pulse is nonselective (or very slightly selective with a much broader spatial profile than that of the selective inversion pulse) to invert all spins. As a consequence, the magnetization of the imaging slice experiences a 360° rotation and ends at a positive z axis. The inflow magnetization is affected by only one inversion pulse and thus is inverted. To acquire the control image, both inversion pulses are nonselective (or very slightly selective with the same spatial profiles), resulting in an uninverted image. Compared to FAIR, UNFAIR does not invert the control image. Therefore, it is insensitive to the inversion time or the difference between blood and tissue T1s. This property also makes UNFAIR less sensitive to errors due to radiation damping (Zhou et al., 1998). BASE is another variation of FAIR (Schwarzbauer and Heinke, 1998). This technique acquires a basis image (i.e., a control) without any spin preparation and a labeled image with a selective inversion pulse applied at the imaging slice location. Because the nonselective inversion pulse is not used in the control image, BASE is more robust against a mismatch between the inversion profile and the imaging slice profile than FAIR. Another variation of FAIR is known as flow-sensitive alternating inversion recovery with an extra radiofrequency pulse (FAIRER) (Mai et al., 1999). As the name implies, FAIRER employs a slice-selective saturation pulse delivered to the imaging location immediately after the inversion pulse of a FAIR sequence. This technique was developed to reduce the TI sensitivity of the subtracted image and improve the robustness against TI values that are close to the nulling point of specific tissues. Another technique, also called FAIRER (FAIR excluding radiation damping), addresses the problem of radiation damping (Zhou et al., 1998). In the presence of radiation damping, FAIR is subject to errors in perfusion quantification. In this FAIRER technique, the effect of radiation damping is suppressed by employing a very weak gradient during the delays in the pulse sequence.

2. CONTINUOUS

Continuous ASL (CASL) was developed before pulsed ASL (Detre et al., 1992; Williams et al., 1992). Unlike pulsed ASL, adiabatic inversion in continuous ASL relies on the flow-induced (also known as flow-driven or velocity-driven) fast adiabatic passage principle (Dixon et al., 1986) to provide a continuous supply of inverted arterial spins to the imaging location. Similar to pulsed ASL, a continuous ASL pulse sequence also acquires at least two images, a labeled and a control image. The pulse sequence to acquire the labeled image begins with a flow-induced adiabatic inversion pulse and an accompanying gradient G along the direction of arterial flow. Although a rectangular RF pulse can be used, the extremely long pulse duration (e.g., several seconds) needed to achieve a continuous arterial spin inversion often exceeds the RF amplifier capability or regulatory limits on power deposition in human subjects. Therefore, in practice, a rectangular pulse is approximated by a series of shorter hard pulses separated by a time delay. Following the long labeling pulse, a spoiler gradient is often played out to dephase any transverse magnetization that might be produced by imperfections of the inversion pulse. After this inversion module and a time delay, an imaging pulse sequence (readout) is executed to acquire a labeled image at the imaging slice location. During the next TR, this pulse sequence repeats itself with the frequency of the labeling pulse changed to the opposite sign, or with a negative labeling gradient –G to obtain a control image. The former approach ensures that the control and labeled images have the same eddy-current effects induced by the labeling gradient, whereas the latter method compensates for the MT effect more effectively. For multiple averages, the labeled and control images are almost always interleaved to improve robustness against patient bulk motion and system instability.

An improved implementation of CASL, known as pseudocontinuous ASL (PCASL), has been recently developed (Dai et al., 2008). In PCASL, the continuous RF is replaced by a long train of slice-selective RF pulses applied at the labeling location, along with a train of gradient pulses that have a small but non-zero mean value. The mean value of both RF and gradient pulses over time are similar to those used in CASL, and the mechanism of inversion is the same. PCASL provides superior labeling efficiency and is compatible with modern body coil RF transmission hardware that is now ubiquitous on clinical MRI scanners. Therefore, it is recommended as the workhorse labeling approach for ASL in most cases (Alsop et al., 2014).

3. VELOCITY SELECTIVE

In this type of labeling method, a velocity selective pulse train is first applied to saturate blood flowing above a chosen cutoff velocity. After a post-labeling delay, a second velocity selective pulse module with the same cutoff velocity is applied either within or prior to the imaging pulse sequence (readout). Thus, only signals from spins that have decelerated from above the cutoff velocity to below the cutoff velocity are measured in the final image, which provides selectivity for arterial delivery, as blood on the venous side of the circulation generally accelerates with time (Wong, 2007; Wong et al., 2006). One advantage of this method is that the velocity selective saturation is usually spatially non-selective, and can even be within the imaging volume. Therefore, velocity selective labeling is not sensitive to vascular transit time variations in the brain.

C. BACKGROUND SUPPRESSION

One main challenge in ASL MRI is its relatively low SNR, as the difference between label and control images is typically <1% of the overall MR signal measured. On the other hand, subject motion, which is typically the dominant noise source in ASL, produces signal fluctuations that are proportional to the signal intensity in the unsubtracted images. Therefore, the signal-to-noise ratio (SNR) in ASL can be improved substantially if the signal intensity of the unsubtracted images can be reduced without affecting the ASL difference signal. This can be achieved by a combination of spatially selective saturation and inversion pulses, usually referred to as background suppression (BS) (Ye et al., 2000). A typical BS module in ASL consists of an initial saturation pulse selective to the imaging region, followed by carefully timed inversion pulses, which results in the longitudinal magnetization of static tissue passing near or through zero at the time of image acquisition. The blood that is to be labeled by the labeling pulses does not experience the initial saturation, but does experience the inversion pulses. For perfect inversion pulses, each inversion changes the sign of the ASL label/control magnetization difference, but nominally does not affect the magnitude of this difference. Thus, the ASL difference signal is preserved, while the static tissue signal is nearly eliminated. Details about the implementation and optimization of BS for ASL can be found in (Dai et al., 2008; Garcia et al., 2005; Maleki et al., 2012).

D. READOUT

A number of pulse sequences can be used as the imaging pulse sequence (readout) for ASL following the labeling module. Single-shot EPI is a common choice in the early days due to its superior acquisition efficiency. Currently, segmented 3D sequences are the preferred methodology. Commonly used 3D segmented methods include 3D multiecho (RARE) stack of spirals (Vidorreta et al., 2013; Ye et al., 2000) and 3D GRASE (Fernandez-Seara et al., 2005; Gunther et al., 2005). These methods provide nearly optimal SNR for measurement of the magnetization prepared by the labeling pulses, and they are relatively insensitive to field inhomogeneity. They strike a balance between the T2* insensitivity of pure RARE methods, and the time efficiency of pure EPI or spiral acquisitions, enjoying most of the benefits of both. Compared with 2D multi-slice readouts, these methods allow for significantly better BS. BS is only optimal at one time point, and because segmented 3D readouts only require one excitation per TR period, the excitation can be timed to provide a very high degree of BS. Single-shot 3D readout may be a promising choice in the future. Multi-slice single-shot 2D (EPI) or spiral readout are also widely used, mainly because of their availability on all modern MRI scanners and the immunity to motion artifacts from the inconsistency between excitations that can affect multi-shot methods. However, for 2D imaging, BS will only be optimal for one or a few slices, and the acquisition time is usually longer in 2D imaging.

E. ADVANCED METHODS TO COMBINE ASL WITH OTHER MEASUREMENTS

ASL MRI can be combined with other approaches to measure CBF, CBV, and BOLD signals, from which the oxygen extraction fraction (OEF), and cerebral metabolic rate of oxygen (CMRO2) can be determined. Such measures are typically performed consecutively, which is not time efficient for clinical use. More importantly, the physiology may be slightly different between scans. By incorporating FAIR ASL and the vascular-space-occupancy (VASO) approach, CBF and CBV changes can be measured in one single scan (Cheng et al., 2014). Furthermore, CBF, CBV and BOLD signal changes can be detected simultaneously by hybrid sequences combining ASL, VASO, and BOLD MRI (Yang et al., 2004). This was first implemented in single-slice version (Gu et al., 2005; Krieger et al., 2015), and more recently expanded with 3D whole-brain acquisition (Cheng et al., 2015).