There have been several methods developed for outer volume suppression (OVS) and T₂-Prep for coronary magnetic resonance angiography^[1],[2]. This work presents an alternative OVS T₂-Prep sequence that additionally provides fat saturation and B₁-insensitivity.

The proposed preparation sequence is shown in Figure 1. A nonselective 90° B₁-insensitive rotation-4 (BIR-4) pulse tips down all longitudinal magnetization into the transverse plane. The transverse magnetization then experiences T₂ decay during this TE period. A spectral-spatial 180° pulse with an HSn adiabatic spectral weighting and sinc spatial weightingis applied at TE/4 with spatial selectivity in x. At 3TE/4, an identical spectral-spatial pulse is applied but with spatial selectivity in y. At TE, a -90° BIR-4 pulse tips up the T₂-weighted magnetization back into the longitudinal axis. The minimum TE of this preparatory pulse is 39ms and including the spoiler gradients at the end, the entire sequence is 55ms.

The pulses were designed for robustness to B₀ and B₁ inhomogeneities at 1.5T. The BIR-4 pulses were designed with a hyperbolic tangent amplitude of β=10, a tangent frequency modulation of tan(λ)=10, max B₁=0.15G, and pulse width=6ms^[3]. The -90° tip up BIR-4 pulse is a time reversal of the 90° tip down BIR-4 pulse so that the phases of the two pulses cancel.

The true null spectral-spatial HSn pulse envelope was designed with n=2, β=3.2, bandwidth=260Hz, pulse width=15.12ms, and max B₁=0.10G^[4],[5]. The sinc subpulses have time-bandwidth (TBW)=4 and frequency FOV=1kHz. The HSn parameters were chosen for more robustness in the passband so that in Bloch simulations at ±1ppm there is 90% of the on-resonance signal (Figure 2b). The tradeoff is that fat is more sensitive to off-resonance for positive ΔB₀. The sinc TBW and frequency FOV were chosen so that the HSn envelope is sampled often enough to maintain adiabatic spin-lock while also keeping gradient slew rates small enough to accommodate various FOVs. A higher TBW can reduce the transition band pin-cushion shape so that it is more rectangular (Figure 3a).

Outer volume suppression is achieved by spatial selectivity of the HSn pulses. The first HSn pulse is selective in x while the second HSn pulse is selective in y. Thus only spins in the intersection of these two regions are refocused while spins outside the region are spoiled. After tip up, only longitudinal magnetization from the inner volume remains. Although each HSn pulse has quadratic spectral phase and linear spatial phase, the spectral phase is cancelled out by the double refocusing so each voxel has linear phase after the two refocusing pulses.

We achieve fat saturation by the spectral selectivity of the HSn pulses. Fat is not affected by the refocusing pulses so it is consecutively spoiled by the four G_z pulses before any remaining transverse magnetization is tipped up. Bloch simulations show a longitudinal magnetization of 0.75%M₀ at the end of the preparation sequence.

Bloch simulations of the proposed sequence in Figure 3a show M_z less than 0.2%M₀ in the outer volume and an inner volume resembling the desired rectangle. Figures 3b-d show that at x=0cm, at an off resonance of ±1ppm, the worst-case M_z is 97%M₀ and at B₁/B_1norm=0.8, the worst-case M_z is 92%M₀.

A 1.5T Signa scanner (GE Healthcare, Waukesha, WI) and a single channel head coil were used to acquire phantom data with a single-slice Cartesian gradient-echo readout, FOV (24cm)², RF excitation angle 30°, TE/TR 6.7/200ms, and T₂-Prep TE 41ms. Figure 4 demonstrates T₂-Prep and OVS; the suppression outside the passband is well appreciated when comparing the images with and without preparation. Figure 4c verifies the T₂-Prep and by quantitative ROI analysis the stopband signal is below 7%M₀. Figure 5 demonstrates OVS and fat saturation. Using quantitative ROI analysis, the image with preparation has a 13.1%M₀ fat signal relative to the image without preparation.

From the results, we see that the proposed sequence has effective OVS, fat saturation, and T₂-Prep while its B₁-insensitivity comes from its adiabatic design. The key difference between the proposed sequence and existing sequences is the achievement of OVS and fat saturation by the spectral-spatial refocusing pulses. This alleviates the constraints on the 90° tip down and -90° tip up pulses so that we can choose adiabatic BIR-4 pulses for both, leading to an entirely adiabatic sequence. This method of OVS also provides a more uniform FOV than 2D spiral excitation, which produces a jinc-weighted FOV. Furthermore, the FOV can be designed as any volume that can be constructed by the intersection of two spatially selective pulses. This provides versatility of the sequence to adapt to various anatomical structures.