Synopsis

Incorporating bipolar, flow sensitizing gradients into a magnetic resonance fingerprinting (MRF) pulse sequence permits quantification of vessel flow along with static tissue T₁ and T₂ relaxation times. This abstract demonstrates quantification of T₁, T₂ and multi-directional flow with a single, volumetric, flow-encoded MRF acquisition with whole brain coverage. Comparison of T₁, T₂, and flow measurements with inversion recovery, spin-echo, and phase contrast techniques revealed good agreement between flow-encoded MRF and gold-standard methods. The capacity to obtain volumetric relaxometry and flow measurements with a single acquisition can support efficient, comprehensive assessments of cerebral blood flow and tissue integrity.

Introduction

Magnetic Resonance Fingerprinting (MRF) techniques permit quantification of multiple MRI parameters (T₁, T₂, M₀) with a single acquisition. Recently, Flassbeck et al demonstrated the capacity to measure flow within vessels, along with T₁ and T₂ in stationary tissue by incorporating bipolar, flow sensitizing gradients into an MRF pulse sequence [1,2]. They referred to their technique as Flow MRF.

Here we demonstrate measurement of T₁, T₂, and multi-directional flow with a single volumetric, flow-encoded MRF acquisition. Furthermore, instead of randomly varying the amplitude of the bipolar gradients and using a pre-defined dictionary for velocity estimation [1,2], flow encoding was performed with a simple four-point scheme, permitting direct, phase-difference velocity quantification.

Methods

Pulse Sequence: An interleaved, four-point flow encoding scheme was incorporated into an inversion-prepared, variable flip angle, 3D (slab-selective), stack-of-spirals FISP acquisition [3]. The TR/TE was held constant at 14/4.23 ms. For each partition encode (k_z), 1000 time points were acquired. All time points were acquired for a single k_z before proceeding to the next one (after a 5 second waiting period for magnetization recovery).

Within each TR, a variable density spiral trajectory, requiring 48 interleaves for full sampling, was used to acquire k_x-k_y data. Every TR, the trajectory angle was rotated by the golden angle (111.25°) and flow sensitization was cycled between a reference acquisition (with no bipolar gradient) and bipolar gradients along the x-, y-, and z-axes. Modulation patterns for the trajectory angle and flow encoding gradients are illustrated in Figure 1.

Reconstruction: 3D volumes were reconstructed in MATLAB using the GPU-NUFFT toolbox. Separate reconstructions were used for flow and relaxometry measurements (see Figure 1). The flow reconstruction generated 4 images (one for each flow encode condition – V₀, V_x, V_y, V_z), by combining all data collected under each condition. Time-averaged flow was then calculated from the phase difference between V_x, V_y, V_z images and the V₀ image.

As suggested in [5], to reduce undersampling artifacts and improve SNR, the relaxometry reconstruction used a sliding window strategy, combining data from consecutive sets of 15 repetitions to generate 986 image frames. T₁ and T₂ were then estimated by voxel-wise matching with a dictionary generated by Bloch equation simulations followed by sliding window summation.

MRI Acquisitions: Scanning was performed on a 3T MAGNETOM Prisma system (Siemens Healthcare, Erlangen, Germany). Flow-encoded MRF data were acquired in relaxometry and flow phantoms and in the brain of 2 adult volunteers. The maximum encoded velocity (venc) was 100 cm/s. The in-plane FOV and resolution were set to 30x30 cm² and 1.2x1.2 mm². Through-plane FOV and resolution (and therefore scan duration) varied depending on the object (relaxometry phantom: 108 mm/3 mm, duration=9.7 min; flow phantom: 150 mm/5 mm, duration=11.6 min; human brain: 144 mm/2.4 mm, duration=19.2 min).

For validation, phantom T₁, T₂ and flow measurements were also obtained using conventional inversion recovery, spin echo, and phase contrast acquisitions.

Results

A comparison of T₁, T₂, and flow measurements with gold-standard methods is shown in Figure 2. In general, agreement was good, with concordance correlation coefficients of 0.99 for T₁ and T₂ and 0.98 for flow. T₁ and T₂ maps, and a time-averaged flow visualization from an in-vivo human brain acquisition are provided in Figures 3 and 4.

Discussion

The results presented demonstrate the feasibility of concurrent, volumetric measurement of T₁, T₂ and flow with a flow-encoded MRF sequence. The slight bias towards T₁ underestimation (see Figure 2b) may be resolved by modelling the adiabatic inversion process (rather than treating it as instantaneous) in the dictionary simulation. Underestimation of the longest T₂ value (Figure 2d) is likely a result of diffusion effects from z-axis spoiler gradients [6]. In general, image quality and accuracy may be improved by correcting for gradient imperfections and off-resonance blurring in the MRF reconstructions.

While a 19-minute scan duration for whole brain coverage is long, several opportunities exist for scan time reduction. As demonstrated in [3,7], a three-fold reduction can be achieved with k_z-undersampling. Advanced reconstruction techniques which exploit inherent redundancies within MRF data (e.g. compressed sensing and low-rank approaches) can support even further undersampling [8].

Correct T₁ and T₂ estimation from MRF data requires knowledge of spin history and the relaxometry reconstruction must therefore preserve the temporal acquisition order. However, more flexibility is permitted in the phase-based flow reconstruction, especially with the use of a golden angle trajectory increment. As such, in addition to exploring options for acceleration, future work will involve retrospective binning by cardiac phase in the flow reconstruction, to support cardiac-resolved flow measurements.

By extending the functionality of MRF to include flow measurement, this technique will facilitate efficient and comprehensive assessments of cerebral blood flow and tissue integrity.

Acknowledgements

References

[1] Flassbeck, S., et al. Quantification of Flow by Magnetic Resonance Fingerprinting. In Proceedings of the 25th Annual Meeting of ISMRM, Honolulu, HI, 2017. Abstract #0128.

[2] Flassbeck, S., et al. Flow-MRF: a novel way of quantifying blood velocities in combination with tissue relaxation parameters. In Proceedings of 30th Annual Meeting of the Society for Magnetic Resonance Angiography, Glasgow, Scotland, 2018. p. 47.

[3] Ma, D. et al., Fast 3D Magnetic Resonance Fingerprinting for a Whole Brain Coverage. Magnetic Resonance in Medicine, 2018, 79:2190-2197.

[4] Jiang, Y., et al. MR fingerprinting using fast imaging with steady state precession (FISP) with spiral readout. Magnetic Resonance in Medicine, 2014, 74:1621-1631.

[5] Cao, X., et al. Robust sliding‐window reconstruction for accelerating the acquisition of MR fingerprinting. Magnetic Resonance in Medicine, 2017, 78:1579-1588.

[6] Kobayashi, Y., and Yasuhiko, T. Diffusion-weighting caused by spoiler gradients in the fast imaging with steady-state precession sequence may lead to inaccurate T₂ measurements in MR fingerprinting. Magnetic Resonance in Medical Sciences, epub ahead of print, doi:10.2463/mrms.tn.2018-0027.

[7] Liao, C. et al. 3D MR fingerprinting with accelerated stack-of-spirals and hybrid sliding-window and GRAPPA reconstruction. Neuroimage, 2017, 162:13-22.

[8] Mazor, G. et al. Low-rank magnetic resonance fingerprinting. Medical Physics, 2018, 45:4066-4084.