Single-shot accelerated volumetric Arterial Spin Labeling is desirable for studies of flow and function modulation but it typically degrades image quality relative to multi-shot interleaved acquisitions. We report a Stack of Spirals RARE acquisition with rotation of spirals between slice encodes and repetitions. This acquisition supports high quality single-shot reconstruction with 3D parallel imaging/compressed sensing while retaining the ability to perform conventional interleaved reconstruction of time averaged perfusion. Application of the sequence to quantification of resting state network fluctuations is demonstrated.
A standard interleaved SOS RARE acquisition sequence was modified in two ways. First the spiral trajectory was modified to variable density for full sampling of the center of k-space in each arm and then 8 fold undersampling of the outer region. Second, the rotation strategy was modified both across shots and slice encodes. Between shots and between increments of slice encoding, the spiral gradients were rotated by an approximate golden angle (Figure 1). This rotation spreads undersampling gaps between the slice and spiral directions to improve parallel imaging while also adding an irregular sampling pattern that may assist with compressed sensing reconstruction. We used an angle of 10π/13, i.e. 138.46°, instead of the golden angle of 137.51°, because the rotation repeats after 13 rotations. The cyclic rotation enables standard slice FFT and 2D regridding reconstruction after 13 interleaves for an excellent quality mean perfusion image that can also be used for coil sensitivity calibration prior to single-shot reconstruction. The repeating rotation pattern also minimizes the memory storage needed for waveforms.
Imaging was performed in 7 healthy volunteers following an IRB approved protocol and written informed consent was obtained prior to imaging. Studies were performed on a GE 3 Tesla MR750 scanner using a 32 channel head array coil from Nova Medical. ASL studies were performed with pseudocontinuous labeling (1.8s labeling, 1.8s postlabeling delay), background suppression, and 32 centric ordered 4mm thick slices. Spiral gradient waveforms were 6.14ms in duration with echo spacings of 12ms. Label and control images were alternated and then rotations of the spiral encode patterns were performed after each pair. 39 rotations were performed corresponding to 3 averages of 13 rotations for a total scan time of 8 minutes.
Images were reconstructed and analyzed off-line. Complex k-space subtraction was performed and fully sampled data were reconstructed by slice direction FFT followed by an in-plane non-uniform FFT (nuFFT). Coil sensitivities were estimated from this fully-sampled perfusion-weighted volume using ESPIRiT(1) (σ=0.01, threshold=0.8) followed by reconstruction of the 39 single-shot volumes using an L1-wavelet Compressed-Sensing reconstruction (λ1=0.005, 100 iterations) with the BART toolbox(2) under MATLAB. Single-shot time series were analyzed for resting state fluctuations(3) using a spatial Independent Component Analysis (ICA) using FASTICA.
(1) Uecker, M. et al. ESPIRiT—an eigenvalue approach to autocalibrating parallel MRI: Where SENSE meets GRAPPA. Magn. Reson. Med. 71, 990–1001 (2014).
(2) Uecker, M. et al. Berkeley Advanced Reconstruction Toolbox. in Proc. Intl. Soc. Mag. Reson. Med 2486 (2015).
(3) Zhao L et al. J Cereb Blood Flow Metab. 2017 doi: 10.1177/0271678X17726625
(4) Li Z et al. Magn Reson Med. 2016 Jan;75(1):266-73
(5) Zhou Z et al. Magn Reson Med. 2017 Dec;78(6):2290-2298