Synopsis

4D flow MRI allows to quantify the velocity vector field non-invasively in-vivo. However, it still suffers from long acquisition times and low temporal and spatial resolution. Here, we accelerate acquisition time and increase temporal resolution without loss of spatial resolution by combining 2D selective excitation and UNFOLD. 2D selective excitation allows to limit the field-of-view in phase encoding direction and thus acquisition time, whereas UNFOLD grants to increase temporal resolution.

Introduction

4D flow MRI¹ allows to quantify the velocity vector field non-invasively in-vivo. However, long scan time and low temporal or spatial resolution are still challenges of the technique. In this work, we aim to address long scan time and low temporal resolution by two complementary methods. First, we excite 2D selectively a reduced field-of-excitation (FOX) enabling a reduced field-of-view (FOV) in phase encoding (PE) direction (FOV_PE). However, the long duration of 2D RF-pulses leads to long TR and thus reduced temporal resolution. Therefore, we additionally apply UNFOLD^2,3, an acceleration technique for dynamic imaging that has been applied successfully to 4D flow⁴, to double temporal resolution.

Methods

2D Selective Excitation with 4D Flow MRI

2D selective RF-pulses with spiral k-space trajectory (Fig.1a) and bandwidth-time-product $$$BWTP=1.2$$$ are designed in the small flip angle regime according to Pauly⁵ to excite a $$$(6\times6\times\infty)\,\text{cm}^3$$$ FOX. The trajectory starts at $$$k_\text{max}=2.51\,\frac{\text{rad}}{\text{cm}}$$$ and spirals inwards with $$$n=10$$$ turns in $$$T=3.5\,\text{ms}$$$. The actual gradient waveform was measured⁶ and integrated into the RF-pulse calculation. Fig.1b shows the resulting excitation profile.

The 2D excitation pulse was integrated into a Cartesian GRE ECG-triggered 4D flow MR sequence¹.

In-Vivo Experiment

A volunteer was scanned according to an approved ethical protocol and with informed consent at 3T (Magnetom Verio, Siemens, Germany). 4D flow imaging was performed using two protocols with varying number of PE-lines and excitation pulses to acquire two data sets: I) conventional 1D-slab-selective excitation and 96 PE-lines (full FOV_PE); II) 2D selective excitation and 48 PE-lines (FOV_PE/2). For all scans we acquired two PE-lines and four velocity encodings per cardiac phase, resulting in a temporal resolution $$$\Delta\,t=8\,T_\text{TR}$$$. Note that the different RF-pulse durations (Tab.1) result in different temporal resolutions, $$$\Delta\,t^\text{1D}=43.2\,\text{ms}$$$ and $$$\Delta\,t^\text{2D}=64.8\,\text{ms}$$$. Acquisition parameters are listed in Tab.1.

Reconstruction with UNFOLD

The sequence acquisition order (two lines per cardiac phase) allows to split odd and even lines of each cardiac phase into two separate time frames. Thereby, time-resolution is doubled, FOV_PE is halved and the set of phase-encodes is shifted by half a line from time frame to time frame (Fig.2a). The obtained complex datasets were Fourier transformed to complex image space, where they show undersampling artefacts (Fig.2b). Subsequently, a fast Fourier transform (FFT) was applied along the temporal dimension. Undersampling artefacts were then removed with a Fermi-filter $$$f(E)=\left (1+\exp{\left(\frac{|E|-E_\text{f}}{k_\text{B}T}\right)}\right)^{-1}$$$ in frequency space, where $$$E_\text{f}=0.5N_y,\,k_\text{B}T=0.022N_y$$$ and $$$N_y$$$ the Nyquist frequency (Fig.2c). UNFOLDed images are finally obtained by applying an inverse FFT to the filtered temporal frequencies (Fig2d).

UNFOLD was applied to in-vivo data set II) as explained above. Furthermore, data set I) was retrospectively four-fold undersampled in PE-direction (data set III) and reconstructed with and without UNFOLD.

Results

Fig.3a shows magnitudes and absolute velocities in the middle cerebral arteries during systole ($$$140.4$$$ and $$$145.8\,\text{ms}$$$ after ECG-trigger) for both data sets. Besides minor aliasing of residual sub-cranial fat signal outside the FOX, successful 2D selection and flow quantification was obtained. Scan times and temporal resolutions are listed in Fig.3a.

Fig.3b shows the images resulting from data set II) split into twice the number of cardiac phases and reconstructed with/without UNFOLD (columns 3&4). UNFOLD effectively removed the aliasing artifacts seen in column 3, resulting in image quality equivalent to Fig.3a (right), but with doubled temporal resolution. For comparison, applying UNFOLD to dataset III) does not fully remove all aliasing artifacts (column 1&2).

The benefit of applying UNFOLD to 2D selective excitation with reduced FOV to improve the temporal resolution is shown in Fig.4 with time-resolved flow quantification through the internal carotids and basilar artery (Fig.4b). Due to the lower temporal resolution of data set II ($$$\Delta\,t^{II}=64.8\,\text{ms}>\Delta\,t^{I}=43.2\,\text{ms}$$$), the peak volume flow is underestimated by $$$\frac{\Delta\hat{v}}{\hat{v}}=11.37\%$$$ compared to data set I. The improved temporal resolution of data set II combined with UNFOLD ($$$\Delta\,t=32.4\,\text{ms}$$$) (Fig.4a) allows to correctly capture peak volume flow ($$$\frac{\Delta\hat{v}}{\hat{v}}=3.63\%$$$) while reducing the scan time by $$$50\,\%$$$.

Discussion and Conclusion

In this work we combine 2D RF excitation with UNFOLD reconstruction to allow for reduced FOV imaging with increased temporal resolution. A main challenge of 2D selective RF excitation is the longer RF pulse duration leading to longer TR and thus to impaired temporal resolution counteracting the benefits of FOX/FOV reduction. UNFOLD was successfully applied to address this issue leading to improved quantification in 4D flow MRI. To further accelerate 2D excitation we aim to apply parallel transmission strategies, which will then also reduce the residual fat artifact that is apparent in the images.

Acknowledgements

References

1. Markl M et al. 4D flow MRI, J. Magn. Reson. Imaging 2012;36:1015-1036.

2. Madore B et al. Unaliasing by Fourier-Encoding the Overlaps Using the Temporal Dimension (UNFOLD), Applied to Cardiac Imaging and fMRI. Magn Reson Med 1999;42:813-82.

3. Zhao L et al. Reduced Field-of-View MRI With Two-Dimensional Spatially-Selective RF Excitation and UNFOLD. Magn Reson Med 2005;53:1118-112.

4. Braig M et al, Evaluation of Accelerated Preclinical 4D-Flow Imaging with UNFOLD, ISMRM Conference Proceedings 2017.

5. Pauly J et al. A k-Space Analysis of Small-Tip-Angle Excitation. J. Magn. Reson. Imaging 1989;81:43-56.

6. Duyn J H et al. Simple Correction Method for k-Space Trajectory Deviations in MRI. J. Magn. Reson. 1998;132:150-153.