Masa Bozic-Iven1,2, Stanislas Rapacchi3, Iain Pierce4, George Thornton4, Qian Tao2, Lothar Schad1, Thomas Treibel4, and Sebastian Weingaertner2
1Computer Assisted Clinical Medicine, University Heidelberg, Mannheim, Germany, 2Delft University of Technology, Delft, Netherlands, 3University Aix-Marseille, Marseille, France, 4Barts Heart Centre, London, United Kingdom
Synopsis
Despite promising results, clinical translation of myocardial arterial spin labelling (myoASL) is hampered by insufficient reproducibility and robustness. We investigated the influence of physiological and sequence parameters on FAIR-myoASL in simulations as well as phantom experiments, and developed a correction method based on separately acquired T1 maps. Our simulation and phantom results show acquisition related MBF differences, potentially undermining the reproducibility of myoASL measurements. Inaccuracies between true and reconstruction blood T1, further render the sequence susceptible to heart rate variations, particularly for larger T1 mismatch. Using accurate blood T1 times in reconstruction may improve robustness and reproducibility.
Introduction
The clinical gold standard for
detection of myocardial ischemia is first pass myocardial perfusion imaging,
where the distribution of an exogenous contrast agent is tracked using T1
weighted imaging1. However, the use of contrast agents limits repeated use and
its clinical applicability. Arterial spin labelling (ASL), on the other hand,
relies on magnetically labelled blood as an endogenous contrast and has proven
effective in quantifying neurovascular perfusion2. In cardiac applications,
however, several challenges are faced, such as a complex anatomy and high
levels of physiological noise3. Although myoASL and positron emission tomography based
myocardial blood flow (MBF) were shown to agree4, insufficient robustness and
reproducibility hinder widespread clinical translation. In this work, we sought
to investigate the influence of physiological and sequence parameters on the
reproducibility of myoASL based perfusion. Further, we sought to develop a
correction method based on precedingly acquired T1 maps to increase robustness
of myoASL.Numerical Simulations
Bloch simulations were performed to
investigate the effect of physiological and sequence parameters for bSSFP and
spGRE based myoASL. For both imaging readouts, the acquisition flip angle (FA),
matrix size, and delay between tag and control image (C-T delay) were varied.
Moreover, varying RR intervals, i.e. different effective inversion times (TI),
were simulated, with a blood volume fraction of 0.14 and in-flow rate of
0.8ml/g/min.FAIR-myoASL Sequence
Imaging was
performed across two 3T scanners (Magnetom Skyra/Prisma, Siemens Healthineers,
Erlangen Germany). For all measurements, a double ECG-gated Flow Alternating
Inversion Recovery (FAIR) ASL sequence4,5 was implemented (Fig.1). For the control and tag image, a selective and global adiabatic inversion pulse was applied, respectively, during mid-diastole. Image acquisition (bSSFP/spGRE) followed in mid-diastole of the successive heartbeat.Phantom and in vivo Measurements
In total,
one pair of baseline images (no inversion) and 5/6 (phantom/in vivo)
control-tag pairs were acquired with 8mm slice-thickness and 2x2mm/1.7x1.7mm
base resolution (phantom/in vivo). MOLLI6 was used for T1 mapping in phantom
and in vivo. Phantom experiments were performed with varying RR intervals, FAs
and C-T delays in a NiCl2-doped agarose phantom. In-vivo images of 4 healthy
subjects (4 male, 39±4.7 years) were obtained with a 6s C-T delay in 12-18s
long breath-holds per image pair depending on the heart rate.Data Analysis
After
groupwise registration7, mistriggered control-tag image pairs were excluded.
Buxton’s model8 was used for MBF reconstruction, where control and tag signal
are corrected with respective TI and blood T1 time (T1B). Previous in vivo
studies relied on a fixed literature based value for T1B (~1700ms)4,5,9. Here,
individual MOLLI-T1B times were used in a second reconstruction method to
obtain perfusion maps and septal MBF. In simulations true T1B was varied, while in
phantom the model T1 was varied. Mean MBF values were obtained by averaging MBF
over all ASL pairs and subjects.Results
Simulated MBF values were constant
(3.5/2.3 ml/g/min bSSFP/GRE) for varying heart rates when simulation and
reconstruction T1B were identical. When “true” and reconstruction T1B differ,
however, the obtained MBF values were highly heart rate dependent. This effect
was more pronounced for lower heart rates and is observed in both bSSFP and GRE
(Fig. 2a). For higher FAs and matrix sizes, bSSFP based MBF continuously
increased. In GRE, MBF decreased to zero for FAs of 30° and beyond (Fig. 2b).
For larger matrix sizes, MBF increased in bSSFP and continuously decreased in
GRE (Fig. 2c). With varying C-T delay (Fig. 2d), MBF values showed almost no
deviation from values simulated with the global inversion pulse acting on fully
recovered magnetization (infinite delay).
Phantom data
depicts similar trends, where MBF difference due to T1B mismatch was stronger
for lower heart rates (Fig. 3a). MBF values in bSSFP increased with increasing
FA, while in GRE the values increased until 18°, then decreased and stayed
constant after 35°. bSSFP based MBF remained constant over the entire range of
matrix sizes, whereas GRE based MBF continuously decreased with increasing
matrix size. For varying C-T delay, MBF remained largely constant in bSSFP and
GRE.
In vivo perfusion maps are shown in Fig. 4a
for bSSFP and GRE acquisition. Septal MBF was 1.53±1.04ml/min/g(bSSFP)
/1.9±1.55ml/min/g(GRE) with fixed T1B, whereas with adaptive T1B MBF was
1.46±0.97ml/min/g (bSSFP) /1.81±1.44ml/min/g (GRE).Discussion
Our simulation and phantom results
show that sequence parameters such as FA and matrix size lead to MBF
differences, potentially eroding the reproducibility of myoASL. Due to the impact of the imaging readout, effects were particularly pronounced for a large mismatch between reconstruction and true T1B. Moreover, MBF values show a heart rate dependence when true and model T1B differ. Without correction, this may lead to measuring spurious MBF
changes in stress and rest conditions.
The effect
of varying C-T delay on MBF values is negligible in comparison, indicating that
6s may suffice despite long T1B. Both in bSSFP and GRE, simulated and phantom
MBF values as well as MBF maps match previously reported perfusion values4,5.
The use of adaptive instead of fixed T1B yields slightly more robust
reconstruction of myoASL based perfusion.Conclusions
MyoASL shows residual sensitivity
to sequence parameters and potentially HR dependence. This may reduce the
reproducibility in clinical use and corrupt stress/rest performance.
Simulations show that the use of accurate blood T1 time may mitigate this
effect for improved robustness.Acknowledgements
No acknowledgement found.References
- Qayyum,
A. A., & Kastrup, J. (2015). Measuring
myocardial perfusion: the role of PET, MRI and CT. Clinical radiology, 70(6),
576-584.
- Chen,
Y., Wang, D. J., & Detre, J. A. (2011). Test–retest
reliability of arterial spin labeling with common labeling strategies. Journal
of Magnetic Resonance Imaging, 33(4), 940-949.
-
Jao,
T., & Nayak, K. (2019). Analysis of physiological noise in quantitative
cardiac magnetic resonance. Plos one, 14(8),
e0214566
-
Zun,
Z., Wong, E. C., & Nayak, K. S. (2009). Assessment of myocardial blood
flow (MBF) in humans using arterial spin labeling (ASL): feasibility and
noise analysis. Magnetic Resonance in
Medicine: An Official Journal of the International Society for Magnetic
Resonance in Medicine, 62(4), 975-983.
-
Do,
H. P., Jao, T. R., & Nayak, K. S. (2014). Myocardial arterial spin labeling
perfusion imaging with improved sensitivity. Journal of Cardiovascular Magnetic Resonance, 16(1),
1-6.
- Messroghli,
D. R., Radjenovic, A., Kozerke, S., Higgins, D. M., Sivananthan, M. U., &
Ridgway, J. P. (2004). Modified Look‐Locker inversion recovery (MOLLI) for
high‐resolution T1 mapping of the heart. Magnetic Resonance in Medicine: An Official
Journal of the International Society for Magnetic Resonance in Medicine, 52(1),
141-146.
- Tao, Q., van der Tol, P., Berendsen, F. F., Paiman, E.
H., Lamb, H. J., & Van Der Geest, R. J. (2018). Robust motion
correction for myocardial T1 and extracellular volume mapping by principle
component analysis‐based groupwise image registration. Journal of Magnetic Resonance Imaging, 47(5),
1397-1405.
- Buxton,
R. B., Frank, L. R., Wong, E. C., Siewert, B., Warach, S., & Edelman, R. R.
(1998). A general kinetic model for quantitative
perfusion imaging with arterial spin labeling. Magnetic resonance in medicine, 40(3),
383-396.
- Lu,
H., Clingman, C., Golay, X., & Van Zijl, P. C. (2004). Determining the longitudinal relaxation time
(T1) of blood at 3.0 Tesla. Magnetic
Resonance in Medicine: an Official Journal of the International Society for
Magnetic Resonance in Medicine, 52(3),
679-682.