Intracranial Dual-Venc 4D flow MRI at 7T: Effect of low versus high spatial resolution in combination with k-t GRAPPA acceleration
Susanne Schnell1, Can Wu1, Pierre-Francois Van de Moortele2, Bharathidasan Jagadeesan3, Kâmil Uğurbil2, Michael Markl1, and Sebastian Schmitter2

1Radiology, Northwestern University, Chicago, IL, United States, 2Center for Magnetic Resonance in Research, University of Minnesota, Minneapolis, MN, United States, 3Neurosurgery Department, University of Minnesota, Minneapolis, MN, United States


Dual-venc 4D flow MRI was applied in 6 healthy volunteers at a 7T MRI scanner at two different approximately isotropic spatial image resolutions, low resolution with (1.1mm)3 and high resolution with (0.8mm)3 voxel volumes. The aim of this study was to systematically investigate the potential of high-resolution k-t GRAPPA accelerated dual-venc 4D flow MRI compared to the low-resolution scan with standard GRAPPA with respect to improved image quality (vessel sharpness and depiction of small intracranial vessels) and quantification of intracranial flow parameters (net flow, peak velocity).


To investigate whether high spatial resolution dual-venc 4D flow MRI acquired with k-t GRAPPA acceleration yields improved image quality and flow quantification for evaluating cerebral hemodynamics compared to a lower resolution scan with standard acceleration (GRAPPA with R=2 along the phase encoding direction).


Dual-venc 4D flow MRI was acquired at two different approximately isotropic spatial image resolutions, low resolution with (1.1mm)3 and high resolution with (0.8mm)3 voxel volumes, on a Siemens Magnetom 7T MRI scanner (SC72 gradients;16kW RF amplifier (Stolberg); 32RX/1TX head coil (Nova Medical)) in 6 healthy volunteers (mean age=29.3±13.5yrs, age range=21-56yrs, 1 female). The standard resolution scan included GRAPPA with acceleration R=2, while k-t GRAPPA with R=5 was used for the high resolution scan with the same volume coverage. Cardiac gating was performed with standard ECG or an acoustic cardiac triggering system (MRItools GmbH, Berlin). Sequence parameters are summarized in table 1. Both acquisitions covered the same 3D volume by adjusting the amount of slices to achieve the same slab thickness.

All data sets were corrected for eddy current induced phase offset errors and Maxwell terms (1). The unaliased high-venc data was used to correct the low-venc acquisition for velocity aliasing resulting in a combined dual-venc data set with intrinsically high dynamic velocity range as well as low velocity noise (2). Phase-contrast angiograms (PC-MRA) were calculated from the velocity and magnitude data and used to identify vessel boundaries and mask blood flow velocities (figure 1). Vessel sharpness was evaluated by determining the thickness at the location of the left or right middle cerebral arteries (lMCA, rMCA) in the axial PC-MRA maximum intensity projection (MIP) images. Further analysis included regional flow quantification at the left and right internal carotid arteries (lICA, rICA), lMCA, rMCA, left and right anterior cerebral arteries (lACA, rACA), basilar artery (BA), left and right posterior cerebral arteries (lPCA, rPCA) as well as the left and right transverse sinuses (lTS, rTS), the straight sinus (SS) and superior sagittal sinus (SagSin) (figure 2) to quantify net flow and peak velocity in commercial software (EnSight, CEI).


All dual-venc 4D flow MRI data were successfully acquired in all subjects with an average scan time of 21.4min (±2.9min) for the low resolution scan. Scan time for the high resolution and k-t GRAPPA accelerated scan was 16.4min (±2.0min). Assessment of vessel thickness in the PC-MRA images yielded smaller vessel diameters of the MCA when measured with high spatial resolution than at low resolution (3.97±0.8mm vs 4.22±0.83mm, Wilcoxon signed rank test, p = 0.03). The advantage of the high resolution scan can be clearly appreciated in figures 1 and 2A and B for the depiction of small intracranial vessels. Figure 1 gives an example of a velocity MIP masked with the PC-MRA for both acquisitions. In figure 2 an axial PC-MRA MIP for the two image resolutions is shown in panels A and B with the corresponding visualization of blood flow using streamlines in panels C and D. Spearman rank correlation analysis showed that peak velocities correlated well between low and high resolution scan (peak velocitylow_res = 0.61±0.35m/s vs 0.57±0.34m/s, R=0.81, P<0.001, figure 3). Net flow was also significantly correlated but was systematically overestimated by the low resolution scan (net flowlow_res=2.94±1.74m/cycle vs net flowhigh_res=2.69±1.48m/cycle, Wilkoxon rank test: p=0.01, spearman correlation R=0.92, P<0.0001, figure 4).


The findings of this feasibility study show that high spatially resolved k-t GRAPPA accelerated dual-venc 4D flow MRI provides the opportunity to resolve flow even in smaller vessels due to its intrinsically decreased noise. k-t acceleration at 7T resulted in both reduced scan time and improved image quality and the ability to quantify venous and arterial flow across a wide range of the velocity spectrum. The low resolution data systematically overestimated net flow, which is expected due to partial volume effects.


In conclusion, dual-venc 4D flow MRI acquisition applied at 7T allows for high spatial resolution due to reduced velocity noise (high magnetic field plus lower venc value) while covering a large dynamic range. This is expected to be highly beneficial for quantification of hemodynamics in arteriovenous malformation or aneurysms with typically large velocity ranges.


Work supported in part by NIH grants P41 EB015894 and P30 NS076408.


1. Walker et al. Semiautomated method for noise reduction and background phase error correction in MR phase velocity data. J Magn Reson Imaging, 1993. 3(3): p. 521-30.

2. Schnell, C. Wu, J. Garcia, I. Murphy, M. Markl. Intracranial k-t Accelerated Dual-Venc 4D flow MRI. Proceedings ISMRM 2015, abstract 4543


Table 1: Dual-venc 4D flow sequence parameters for the low and high spatial resolution acquisitions.

Figure 1: Velocity maximum intensity projection (MIP) using the 3D PC-MRA as mask of (A) the low spatial resolution scan and (B) the high spatial resolution scan. The high resolution scan shows a clearer depiction of smaller vessels and smoother velocity profiles across the vessels when visually inspected.

Figure 2: (A) PC-MRA maximum intensity projection (MIP) of the low spatial resolution scan with standard GRAPPA, (B) PC-MRA MIP of the high spatial resolution scan. (C and D) peak systolic streamlines of low and high resolution scan, color coded by speed. The analysis plane positions at the lICA, rICA, lMCA, rMCA, lACA, rACA, BA, lPCA, rPCA, as well as the lTS, rTS, SS, and SagSin are indicated with red lines in panels B and D.

Figure 3: (A) Correlation analysis of low and high spatially resolved peak velocity at the 12 different arterial and venous locations. (B) shows the corresponding Bland-Altman plot.

Figure 4: (A) Correlation analysis of low and high spatially resolved net flow at the 12 different locations. (B) shows the corresponding Bland-Altman plot illustrating a significant underestimation the high resolution scan of 0.26 ml/cycle (10.7% higher than median net flowhigh_res, p = 0.01).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)