Introduction

Since its inception, continuous flow-driven adiabatic inversion using pulsed RF and gradients, known as PCASL¹, has become a popular method for Arterial Spin Labeling (ASL) perfusion² because of higher efficiency compared to continuous ASL³, multi-slice approaches but also theoretically higher SNR compared to pulsed methods^4,5. Nonetheless, the off-resonance sensitivity as well as high power deposition hamper PCASL applications at ultra-high-field (e.g.≥7T). While saturation-based labeling techniques have been among the first proposed for ASL^6,7, their use has been limited because of their lower SNR compared to inversion strategies and the lack of an effective control strategy to compensate for direct MT and other effects on imaged tissue. We report here the implementation and early optimization of a pseudo-continuous saturation based labeling strategy for low SAR robust perfusion imaging.

Material and Methods

Theory
When PCASL is optimized for inversion labeling, the gradients and RF pulses are designed to create one inversion plane within the RF pulse width while aliased inversion planes are outside the pulse width. Off-resonance sensitivity arises in PCASL inversion because it induces shifts of the labeling plane away from the center of the pulse where RF power is lower. Conversely, for saturation labeling, we reduce the ratio of the RF slice selection gradient to the time averaged gradient such that multiple aliased labeling planes are within the slice profile. This introduces multiple, partially efficient “inversions” that ultimately lead to saturation of flowing spins in a manner similar to DANTE flow saturation⁸. Off-resonance sensitivity is greatly reduced relative to inversion because it merely shifts the multiple lines around within the pulse envelope (Fig.1). The unbalanced gradient control used in PCASL inversion also serves as an effective control for saturation.
Simulations
We simulated labeling efficiency using a numerical integration of Bloch equations as described previously^1,9,10. We simulated efficiency over a range of average power (B_1av) from 0.2 to 1.5μT and peak-to-average gradient ratio from R=1 to 7 by 0.5 steps with an average gradient set at 0.5mT/m (to avoid being close to background gradients levels) at a fixed flow-velocity of 50cm/s for a Hann-shaped pulse of 500μs played every 1ms or shorter when hardware compatible.
Experiments
Three volunteers were scanned at 3T (GE Discovery MR750) using 32-ch coils for brain (N=1) and kidney (N=2) imaging. A single-slice single-shot FSE was acquired, positioned axial mid-ventricles for the brain and mid-kidneys coronal. SSFSE parameters were TR/TE=6000/40ms,mtx=128²,rBW=20.83kHz, 7-mm thick slice, flip-angle=120°. We acquired two datasets in each case with either the recommended optimized inversion scheme (B_1av=1.4μT,R=7,G_av=0.5mT/m)¹⁰ and the proposed saturation scheme (B_1av=0.75μT,R=2,G_av=0.5mT/m) with a background-suppressed (BS) PCASL preparation using a w=1.5s labeling and a single PLD=1.5s (kidney) and 2s (brain), with 14 pairs for the brain and 7 for the kidneys. Kidney acquisitions were acquired with a timed-breathing strategy.
Quantification
Image reconstruction was performed offline, with a complex ASL subtraction followed by homodyne reconstruction. We quantified absolute cerebral and renal blood-flows (f) using a 1-compartment model¹¹:
$$f=[6000\cdot\lambda\cdot dM\cdot \exp(PLD/T_{1b})]/[IF\cdot \alpha\cdot T_{1b} \cdot M_{0} \cdot \exp(-w/T_{1b})]$$ With IF=1 for the saturation-based and 2 for the inversion-based labeling strategies. The labeling efficiency a was estimated at 0.6 for PCASL-I (0.8x0.75 for BS) and 0.8 for PCASL-S because of BS. We calculated a mean SNR for both labeling schemes and a ratio of PCASL-I/PCASL-S.

Results

Simulation results show that by reducing the average B₁ (0.75μT) with a low gradient ratio (2), it is possible to obtain a flat efficiency of 0 corresponding to spins saturation over a wide-range of off-resonance frequencies (Fig.2) compared to optimized PCASL inversion labeling. When looking at the spatial profile of longitudinal magnetization, we see that, in accordance with theory, multiple labeling planes appear within the Hann-pulse envelope eventually leading to spins saturation, but also that this strategy is somewhat robust across a range of flow velocities (Fig.3).
First imaging experiments in kidneys (Fig.4) and brain (Fig.5) are encouraging. As expected, we measured a lower SNR using the PCASL-S vs I (21.0vs36.9) in the kidney scan but the difference in SNR was not marked in the brain (16vs15). We observed a mild reduction in ASL signal fluctuation across repetitions from 6 to 4% in the brain (STD% of the mean perfusion signal) but more significant in the kidney case from 26 to 18%, suggesting increased labeling temporal stability. Additionally, we noted a marked reduction of SAR from 2 to 1.1W/kg for the kidney and 1.7 to 0.9W/kg for the brain. Finally, for quantification, we observed numbers consistent with healthy renal flow while markedly reducing in-ROI STD in the renal cortex in the saturation case (261±80 vs 224±40mL/100/min). In the brain, similar observations were made (80±20 vs 61±15mL/100g/min).

Discussion and conclusions

Those preliminary theoretical optimizations and experiments show that saturation-based labeling strategies based on a minor modification of PCASL can lead to an off-resonance robust labeling while greatly reducing power deposition, potentially paving the way for ultra-high-field robust ASL perfusion imaging. While somewhat reducing SNR, this saturation labeling also facilitates ASL quantification by reducing the influence of labeling efficiency on estimated blood-flow by rendering it much more insensitive to off-resonance. However, more work is required to fully characterize labeling efficiency and effects on SNR to provide optimal flow-velocity and B₀/B₁ robustness.

Acknowledgements

No acknowledgement found.