Pseudo continuous arterial spin labeling (pCASL) can provide valuable flow and function information by efficiently labeling flow in the upstream arteries. Recently consensus recommended parameters for pCASL labeling have been published based on prior optimizations¹. These optimizations have largely focused on maximizing efficiency for an average vessel velocity on resonance². In practice, the flow in the major feeding vessels is highly pulsatile and the pulsatile velocities can vary across time or subject. Additionally, off-resonance efficiency loss can be a significant factor. These factors have been considered in simulations for balanced pCASL³, but since this pCASL version is highly off-resonance sensitive⁴, correction of off-resonance was emphasized. Here we present simulations and initial in-vivo data suggesting that robustness and accuracy of unbalanced pCASL can be substantially improved by changing labeling parameters and/or pulses from those recommended in the consensus document. Adjustment of these parameters may reduce not just clear failures, but also variable efficiency across flow territories, scans, and subjects.

Optimization of continuous labeling requires trading off two different sources of inefficiency: incomplete rotation and T2 related signal loss. These are expressed by the adiabatic condition.

In pCASL, the B1 amplitude experienced by flowing spins can be reduced if the labeling plane is shifted relative to the RF pulse envelope by off-resonance (Figure 1 dash line). This leads to several insights regarding B0 robustness: (1) Higher velocity spins will suffer disproportionately from off-resonance. (2) The RF pulse excitation profile should be widened (by decreasing the ratio of the gradient during the RF pulse to the time averaged gradient) until aliased labeling planes begin to contribute to inefficiency (Figure 1 orange line). (3) A flatter RF pulse profile may help to reduce B0 sensitivity (Figure 1 yellow line).

Flow driven inversion was simulated with the hard pulse approximation of the Bloch equations⁵ using a step size of 4us. Labeling RF pulses were 500us duration (with several shapes as described below), followed by a 500us gap. Blood spins originated from 30mm below and stopped at 80mm above the tagging plane.

Pulsatile flow was approximated as a maximum peak velocity of 80cm/s at systole and 20cm/s at diastole, which occupied 80% of the cardiac cycle. Blood flow was assumed to follow a laminar distribution. Labeling efficiency $$$\alpha$$$ was weighted averaged according to flow contribution. To quantify the performance, the averaged inversion efficiency ($$$Q=-\sum w(v,B_{off})\beta$$$, $$$\beta$$$ is inversion efficiency with -1 for perfect inversion ) was calculated across off-resonance 0-300Hz and velocity distribution.

A minimum phase (MP) RF pulse was optimized for flatter excitation profile than the Hann RF pulse typically employed for pCASL. Simulations were performed both with the Hann pulse and the MP pulse. Using the pulsatile velocity distribution, the optimal performance of Hann was searched with fixed SAR, Gave 0.1-2mT/m and the ratio between slice selective gradient of tagging pulse (Gmax) and averaged gradient (Gave) 1-14.

Volunteers were scanned on a 3T GE HDxt scanner. pCASL was performed with Hann labeling RF and 3D stack-of-spiral FSE readout. Scans were performed with parameters near those suggested in the consensus document and with parameters suggested by our optimization. A range of off-resonance was simulated by including a variable phase increment to the phase shift between labeling pulses.

Averaged inversion efficiency (Figure 2) suggested different optimal Hann parameters (red square) than consensus recommendation parameters (red circle), when considering velocity and off-resonance.

Simulations of labeling with the consensus recommended parameters confirms that off-resonance sensitivity is much higher for high velocity spins (Figure 3a). By reducing the gradient pulse during the RF pulse to widen the response and lowering the average gradient to increase efficiency at higher velocities, major improvements in high velocity and off-resonance robustness can be achieved (Figure 3b). Simulations using a MP pulse for flatter excitation profile also confirm the theoretical benefit (Figure 3c). Compared to Hann, MP resulted in better off-resonance performance but worse high velocity coverage. The simulated performance of consensus recommended Hann, optimal Hann and Minimum phase RF are shown in Table 1.

Figure 4 the consensus recommended parameters resulted in higher labeling of the posterior circulation and efficiency variability with frequency offset. The newly optimized parameters showed more homogenous perfusion signal across flow territories and frequencies across a wider range of frequencies.

In this work, we improved the labeling efficiency of unbalanced pCASL at off-resonance and high velocity. Simulations and preliminary data suggest the proposed parameters are more robust. Further experimental validation in realistic clinical cohorts is required prior to changing recommendations.