Synopsis

Keywords: Arterial spin labelling, Perfusion

In this study, a framework was set up for the simultaneous acquisition of 4D MRA and perfusion ASL data while maintaining whole brain coverage. A Hadamard-8 preparation was combined with a 3D Stack of Spirals readout which resulted in 7 timepoints. By combining these two techniques we were able to obtain both angiography and perfusion from a single dataset. By improving the efficiency of the sampling scheme i.e. using variable density spirals, the total scan time could be reduced further while improving the SNR during the perfusion phase, albeit at the expense of the quality of the 4D MRA.

Introduction

The combination of 4D angiographic and perfusion information can provide essential insight into the hemodynamic condition of the brain. Recently, ASL was combined with a single slice radial readout to simultaneously obtain high temporal and high spatial resolution 2D angiography and low spatial resolution perfusion images^1,2. Compared to radial trajectories, spiral trajectories are more efficient regarding k-space coverage and are even more flexible, since the sampling density can be varied³. Especially for the simultaneous acquisition of angiography and perfusion data, a variable density spiral could be of interest, since the center of k-space can be oversampled to improve the SNR of the perfusion images. On the other hand, the outer part of k-space can easily be undersampled, as angiographic images are well-known to be sparse⁴.
The aim of this study was to simultaneously acquire 4D ASL angiographic and perfusion images from a single dataset while maintaining whole brain coverage and improve SNR during the perfusion phase. Therefore, we combined a time-encoded pseudo-continuous arterial spin labeling (te-pCASL) preparation with a (uniform or variable density) 3D Stack of Spirals (SoS) readout.

Methods

Three healthy volunteers (age 25-28 y.o., 3f) were scanned using a 32-channel head coil on a 3T-scanner (Ingenia-CX, Philips). All volunteers provided informed consent and the study was approved by the local IRB.
A Hadamard-8 matrix, which was used as the temporal resolution source, was followed by a segmented 3D SoS turbo-field echo spiral readout block having a golden angle feature, played out after a post-labeling-delay (PLD) of 280 ms (Figure 1). Two Frequency Offset Corrected Inversion (FOCI) pulses at 1815 and 3135 ms were used for background suppression. Three different spiral trajectories were used to acquire three datasets: uniform density spiral and a variable density spiral with R=2 or R=3 (Figure 2) using 19 k_z phase-encodings employing a flip-angle-sweep to maintain constant signal (TR/TE 31/1.85 ms, acquisition matrix 256).
An offline reconstruction in Python (V3.6 and V3.7), using the SigPy package, was employed. Sampling density correction was applied before the adjoint non-uniform Fast Fourier Transform⁵ and final coil combination by using a sum-of-squares approach. For the angiography an oversampling factor of 1.1 along the z-direction was used. For the perfusion phase an oversampling factor of 1.6 was used, because fold-in artefacts are more evident in the perfusion images than in the 4D MRA images. Since for perfusion imaging a low spatial resolution is preferred to enhance SNR, only the densely sampled center of k-space was included in the perfusion reconstructions (Figure 2).
To compare the SNR between the different acquired datasets, all scans were repeated once. Temporal SNR for the perfusion data was calculated as follows: tSNR=(mean signal over time)/(std over time) within the gray matter (GM).

Results

Figure 3 shows the maximum intensity projection (MIP) for the angiographic phase (A) and the perfusion phase (B) for a representative subject for all three datasets. The labeled blood is captured flowing into and through the Circle of Willis while later time points show the blood traversing the vasculature towards the tissue. The rightmost images show the MIP of the sum of all four timepoints of the angiographic phase. The second and third row show the angiographic phase for the undersampled datasets. For the fully sampled dataset the labeled blood is visible further down the vascular tree and the signal intensity is higher. The perfusion phase is reconstructed at a lower spatial resolution leading to an increased SNR as when reconstructed at the high resolution of the angiography images.
The GM tSNR showed an increase for the variable density scans compared to a uniform spiral (Table 1). The data with R=3 showed the highest tSNR, since for this dataset the center of k-space was oversampled the most, resulting in a higher percentage of k-space points being included for the perfusion reconstruction.
The two repeats of the variable density spiral with R=3 were averaged to increase SNR (Figure 4). The total scan time for this dataset was 5 min 54 sec and especially in the perfusion phase, an increase in image quality can be observed compared to the fully sampled dataset.

Discussion and Conclusion

In this study, a framework was set up for the simultaneous acquisition of 4D MRA and perfusion ASL data while maintaining whole brain coverage. By combing te-pCASL with a 3D SoS readout, which resulted in 7 timepoints, both angiography and perfusion were obtained from a single dataset. Using variable density spirals, the total scan time could be reduced further while improving the SNR during the perfusion phase, albeit at the expense of the quality of the 4D MRA.
It was checked whether a conjugate phase reconstruction^6,7 would improve the quality of the images, however, since B0 inhomogeneities were small (within 30 Hz), this was not the case. In addition, for the perfusion reconstruction only the densely sampled center of k-space was used, which resulted in a very short acquisition window of 2.3 ms per spiral. Further optimization of the exact spiral trajectories, the use of concurrent magnetic field monitoring to characterize the effective trajectories, as well as improved reconstruction approaches would be logical next steps⁸.

Acknowledgements

This work is part of the research programme Innovational Research Incentives Scheme Vici with project number 016.160.351, which is financed by the Netherlands Organisation for Scientific Research (NWO).

References

1. T.W. Okell, Combined angiography and perfusion using radial imaging and arterial spin labeling. Magn Reson Med, 2019. 81(1): p. 182-194.
2. M.C.E. van der Plas, S. Schmid, M. Versluis, T. Okell and M. van Osch, Time-encoded golden angle radial arterial spin labeling: Simultaneous acquisition of angiography and perfusion data. NMR in Biomedicine, 2021. 34: p. e4519.
3. J.G. Pipe and N.R. Zwart, Spiral trajectory design: a flexible numerical algorithm and base analytical equations. Magn Reson Med, 2014. 71(1): p. 278-85.
4. M. Lustig, D. Donoho and J.M. Pauly, Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn Reson Med, 2007. 58(6): p. 1182-95.
5. R.D. Hoge, R.K.S. Kwan and G.B. Pike, Density compensation functions for spiral MRI. Magnetic Resonance in Medicine, 1997. 38: p. 117-128.
6. Lai-Chee Man, John M. Pauly and A. Macovski, Multifrequency interpolation for fast off‐resonance correction. Magnetic Resonance in Medicine, 1997. 37: p. 785-792.
7. P. Bornert, H. Schomberg, B. Aldefeld and J. Groen, Improvements in Spiral MR Imaging. MAGMA, 199: p. 19-41.
8. B.J. Wilm, C. Barmet, S. Gross, et al., Single-shot spiral imaging enabled by an expanded encoding model: Demonstration in diffusion MRI. Magn Reson Med, 2017. 77(1): p. 83-91.