ASL is a powerful technique to image perfusion without the injection of tracers. A challenge in ASL is the effect of the bolus arrival time variability on the observed signals. Velocity selective ASL (VSASL) has been shown to dramatically reduce the sensitivity to bolus arrival time [1]. However, saturation labeling gives up half of the contrast because magnetization of the arterial water is saturated, rather than inverted.

Velocity selective inversion (VSI) pulses, on the other hand, could produce nearly twice as much contrast in the magnetization. Recent work has demonstrated the feasibility of VSI pulses, but these are very sensitive to B0 and B1 homogeneity [2]. In this article, we develop, optimize and test a new approach to VSI that inverts the magnetization of adiabatically, as long as they are moving within a specific velocity band. A major advantage of adiabatic pulses is that they are known to be significantly more robust to B1 inhomogeneity.

Consider a hyperbolic secant adiabatic pulse that is broken up into multiple small segments. Velocity selectivity can be achieved by introducing a bipolar pair of gradients (Gv) between the RF pulse segments, whose phase is incremented linearly from pulse to pulse. The phase increment is calculated to match the quadratic phase gain that is experienced by the spins traveling at a specific velocity in the presence of the bipolar pair. The magnetization “tracks” the effective field, provided that the sweep rate maintains the adiabatic condition [3]. Spins that do not match that velocity (and therefore that phase gain) cannot “lock on” to the effective field and do not achieve adiabatic inversion. Thus, one can achieve velocity selective adiabatic inversion (VSAI) pulses that invert only those spins moving at the target velocity while having negligible effect on the magnetization of the spins outside the velocity range.

In order to characterize the properties of the VSAI pulses, we simulated the magnetization vector of an isochromat moving at a constant velocity under multiple VSAI pulse design parameters. Unless otherwise specified, we used Ns = 32 segments, max. B1 amplitude = 200 mG, B1 sweep width = 500 Hz, bipolar gradient duration = 2 ms (two 1ms lobes), gradient amplitude = 3.5 G/cm, target velocity = 0 cm/s (i.e., =0). In all cases, we obtained velocity selectivity profiles for each VSI pulse by repeating the simulation at different velocities of the isochromat from -50 to 50 cm/s.

We investigated the effects of (1) the number of segments (Ns) from a single pulse to 32 segments (2) RF pulses’ maximum amplitude (0 – 300 mG) and sweep widths (250 Hz – 1000 Hz). (3) the bipolar gradient’s amplitude by from 0 to 4 G/cm.

We then tested a 32 segment VSAI pulse in an ASL experiment on two human participants with a 3T scanner (GE MR750). The Labeled image was preceded by a VSAI pulse targeting velocity=0, and the Control image was non-velocity selective (no velocity selective gradients). The difference between images results from the uninverted moving spins in the arteries as they move into the tissue during a post labeling delay. We collected a label uptake curve by incrementing the post labeling delay from 50 to 2000 sec. Image acquisition consisted of a single shot spiral sequence with FOV=24 cm, matrix = 64 x 64, BW = 85 kHz, sl. th. = 6 mm.

Figure 1 shows the velocity profile of two such pulse trains targeting 0 and 20 cm/s. The simulations in figures 2-4 indicate that the velocity bandwidth is largely affected by the number of segments, bipolar gradient amplitude and duration, and the sweep width of the original adiabatic pulse. The simulations confirm that VSAI pulses are very stable over a wide range of B1 amplitudes (figure 5) .

The human data (figure 6) show a uptake curve typical of a pulsed ASL experiment, and consistent with tracer kinetic models. The early points in the uptake, however, are significantly affected by eddy currents induced by the velocity selective bipolar gradients. These artefacts are largely absent beyond 500 ms.

This work demonstrates that VSAI pulses are potentially very useful for ASL. They are very stable over a wide range of B1 amplitudes, thus achieving spatially uniform labeling efficiency. Further research is currently under way to characterize sensitivity to off-resonance, optimize the pulses in terms of their velocity selectivity , power deposition, and to reduce eddy currents.