Purpose

Vessel-selective arterial spin labeling (ASL) enables the visualization of blood flow arising from individual arteries, which is an important advantage of ASL-MRA over contrast-enhanced MRA. In vessel-encoded pseudo-continuous ASL (ve-pCASL) perfusion MRI¹, SNR-efficient territorial separation can be achieved by employing Hadamard-encodings across the feeding arteries whilst repeating measurements for averaging. For ASL-MRA, however, 3D data is usually acquired in a multi-shot fashion and the entire scan-time is used for spatial encoding, rather than averaging, to achieve high spatial resolution. Therefore, the number of vessel-encodings will increase scan-time proportionally. In a previous report, eight Hadamard-encodings of four arteries resulted in a scan-time of 18 minutes². The three-vessel encoding scheme of Günther³ only requires four Hadamard-encodings, in which each of right internal carotid artery (RICA), left ICA (LICA) and both vertebral arteries (VAs) are labeled in different combinations, along with a non-selective control image. This scheme would reduce scan-time by a factor of two. However, the pulsed-ASL approach of Günther suffers from difficult planning of the selective labeling slab to avoid contamination from non-targeted arteries, because it requires large coverage of tortuous vessels in the inferior-superior direction. Although ve-pCASL avoids this problem, it requires complicated post-processing such as the Bayesian framework5 because the inversion modulation of ve-pCASL is more gradual than pulsed-ASL. The purpose of this study is to achieve a sharper ve-pCASL spatial modulation to improve labeling efficiency and minimize partial labeling of nearby arteries, enabling simpler post-processing for 4D-MRA within a clinically feasible scan-time.

Theory

Ve-pCASL has two different approaches to achieve the spatial modulation of inversion: the bipolar approach in which the transverse gradient (G_xy) is applied alternately positive and negative, and the unipolar approach in which G_xy is not alternated. Magnetization is inverted at the positions where the phase of the labeling pulses is aligned with the phase shift produced by G_xy. Because of the pCASL mean gradient (G_mean) in z-direction, this in-phase position varies with z. With bipolar G_xy, the phase coherence away from the middle of the labeling plane is lost, but in the unipolar approach, the consistent G_xy and RF phase cycling result in phase coherence within tilted planes, visible as lines of saturation in a static phantom (Figure-1b-e). The effective inversion width (“w” in Figure-1a) depends on the maximum gradient of pCASL (G_max), whereas G_mean determines the angle of the saturation line (“θ” in Figure-1a). Accordingly, the inversion modulation of the unipolar approach is dependent on G_max and G_mean.

Methods

As Wong et.al. suggested⁴, the unipolar approach is superior to the bipolar approach, therefore all experiments were performed using unipolar gradients. By Bloch equation simulation, ve-pCASL inversion modulation was simulated with G_max of 6 mT/m and G_mean of 0.8 mT/m (default settings, as per Wong et al.^1,4) and with G_max and G_mean decreased in several steps to 2mT/m and 0.2 mT/m, respectively. Next, in-vivo studies were performed in two healthy volunteers to validate the simulation results: (1) Perfusion images were acquired whilst offsetting a left-right encoding in small increments relative to the RICA and LICA locations; averaged perfusion signal was extracted from LICA/RICA masks generated by a conventional ve-pCASL perfusion image with eight Hadamard-encodings. (2) ve-pCASL 4D-MRA was acquired using the three-vessel encoding scheme using the optimized settings of G_max and G_mean, with scan-time 7min12s. All scans were performed on a Siemens 3T Verio under a technical development protocol agreed by local ethics and institutional committees.

Results and discussion

Simulated inversion modulations showed sharper selectivity for lower G_mean (Figure-2a), which is attributed to steeper saturation line (Figure-1c/e). For a fixed G_mean, lower G_max produced broader inversion width (Figure-2b), which is explained by wider effective inversion width (Figure-1d/e) which produced concomitant broadening of the inversion width. Figure-3 shows the averaged perfusion signal curves as a function of the selective labeling offset measured in-vivo. With optimized G_mean and G_max sharper selectivity was achieved, confirming the simulation results. Comparison of the perfusion signal at offset=0 (Figure-4) did not suggest any differences in the labeling efficiency between the default and optimized settings. Figure-5 shows ve-pCASL 4D-MRA of a healthy volunteer acquired with the three-vessel encoding scheme and the optimized settings. By optimization, labeling of each RICA, LICA and VAs were achieved with high selectivity, and no obvious mislabeling was observed. Compared to the previous report using eight Hamadard-encodings, scan-time was halved.

Conclusion

In this study, we presented an approach to improve the inversion modulation in ve-pCASL, enabling faster scan-times. Controlling the inversion modulation with G_max and G_mean could be beneficial, not only for 4D-MRA, but also for perfusion imaging.

Acknowledgements

This research was supported by the EU under the Horizon2020 program (project: CDS-QUAMRI).

References

1. Wong EC. Vessel-Encoded Arterial Spin-Labeling UsingPseudocontinuous Tagging. Magn Reson Med (2007) 58:1086-91.

2. Okell TW, Schmitt P, Bic X, Chappell MA, Tijssen RHN, Sheerin F, Miller KL and Jezzard P. Optimization of 4D vessel-selective arterial spinlabeling angiography using balanced steadystatefree precession and vessel-encoding. NMR in Biomed (2016) 29:776-86

3. Guenther M. Efficient Visualization of Vascular Territories in theHuman Brain by Cycled Arterial Spin Labeling MRI. Magn Reson Med (2006) 55:671-5.

4. Wong EC, Guo J. Blind detection of vascular sources and territories using randomvessel encoded arterial spin labeling. Magn Reson Mater Phy (2012) 25:95–101

5. Chappell MA, Okell TW, Payne SJ, Jezzard P, Woolrich MW. A fast analysis method for non-invasive imaging of blood flow in individual cerebral arteries using vessel-encoded arterial spin labelling angiography. Med. Image Anal (2012) 16:831-9