Giulia MC Rossi1, Nemanja Masala1, Jessica AM Bastiaansen1, Aurelien Bustin1,2, Jérôme Yerly1,3, John Heerfordt1,4, Davide Piccini1,4, Matthias Stuber1,3, and Christopher W Roy1
1Department of Radiology, Lausanne University Hospital (CHUV) and University of Lausanne (UNIL), Lausanne, Switzerland, 2LIRYC (Electrophysiology and Heart Modeling Institute), Bordeaux, France, 3CIBM Center for Biomedical Imaging, Lausanne, Switzerland, 4Advanced Clinical Imaging Technology (ACIT), Siemens Healthcare AG, Lausanne, Switzerland
Synopsis
A novel method for reconstructing motion-compensated
3D CMRA images from free-running data is proposed. We extend the use of a recent
method for non-rigid respiratory motion correction called focused navigation
(fNAV) to also encompass cardiac motion compensation that accounts for beat-to-beat
heart-rate variability. Our combined fNAV approach is compared to the
previously established cardiac and respiratory motion-resolved 5D imaging in vivo and is shown to provide overall similar image quality and comparable right
coronary artery visualizations to 5D imaging in significantly shorter
reconstruction times.
Introduction
Coronary Magnetic Resonance Angiography (CMRA) requires
robust compensation of both cardiac and respiratory motion. The free-running
framework1-2 allows for cardiac and respiratory motion-resolved 5D
imaging, where an end-expiratory mid-diastolic motion state can be
retrospectively identified to effectively “freeze” motion for coronary
visualization. However, unresolved intra-bin
motion, heart-rate variability, and undersampling artifacts may still degrade
the image quality.
Recently, an auto-focusing technique (fNAV)3 was proposed for intra-acquisition non-rigid
correction of respiratory
motion in 3D radial ECG-triggered CMRA acquisitions. Using fNAV, high quality visualizations
of the coronary arteries can be obtained by combining data from the entire
respiratory cycle. Still, prospectively triggered images can suffer from incorrect
resting phase calculation or heart-rate variability.
In this work, we developed a novel method for whole-heart
CMRA that extends and integrates the fNAV approach with a previously published free-running1-2
uninterrupted bSSFP acquisition using LIBRE water excitation pulses4-6. We test the hypothesis that, a) by
using fNAV to correct for non-rigid respiratory motion and b) by integrating a subject-specific
auto-focusing strategy for retrospective cardiac soft-gating, we can obtain high-quality
images of coronary arteries accounting for heart-rate variability without the
need for prospective calibration scans7 or long compressed sensing
(CS) reconstructions. We quantitatively compare our approach to the previously
established cardiac and respiratory motion-resolved 5D imaging in vivo.Methods
Acquisition. CMRA data were acquired in 19 healthy volunteers
(age: 23-38 y; 10 male) with written informed consent on 1.5T clinical scanners
(MAGNETOM Aera and MAGNETOM Sola, Siemens Healthcare, Erlangen, Germany) with a previously described prototype free‐running acquisition
7 while simultaneously
recording the ECG signal.
Reconstruction. For each volunteer, two reconstruction approaches were
compared. First, 5D images were reconstructed using CS as previously reported
1-2,6
and a 3D volume corresponding to a manually
identified quiescent cardiac phase at end-expiration was selected. In the
second approach, a quiescent 3D volume was obtained by correcting an
automatically selected subset of quiescent cardiac data for respiratory motion
using fNAV (
Fig.1), according to
the following steps:
- Respiratory motion amplitudes are estimated with fNAV3 and are used for rigid correction of respiratory motion in k-space.
- Quiescent cardiac data selection and soft-gating
is performed by extending the auto-focusing concept to the cardiac motion model
presented in [7]. Based on the ECG time stamp, for each heart-beat i, the trigger delay (Td) is computed as: $$Td(i)=k_0·T_{RR,AVG}(i)+(1-k_0)·k_1·log(10·(T_{RR,AVG}(i) +k_2))$$ where, to account for heart-rate variability, the R-R
interval
is computed for each heart-beat based on the
average of the five preceding heart-beats. A window of
empirically determined fixed width (w) starting from Td is considered to be
quiescent and corresponding readouts are weighted by a Gaussian
function centered at Td(i)+w/2. Here, k0 replaces
an offset value traditionally determined through calibration scans7,
while k1 and k2 are related to the
duration of the systolic and diastolic portions of the cardiac cycle and are
age and gender dependent. In our approach, the three coefficients are considered
subject-specific and are iteratively estimated by evaluating the image gradient
entropy of intermediate images obtained with respiratory motion correction from
(1) and the current cardiac soft-gating.
- Final non-rigid respiratory motion correction of the cardiac soft-gated
data from (2) is performed using fNAV3.
Quality assessment. For both methods, reconstruction times were recorded
for comparison. Curved reformats of the right coronary artery (RCA) were obtained
from the resulting 3D images and visually compared. Quantitative comparison of
image quality was performed via evaluation of RCA percentage vessel sharpness (%VS)
and visible length using Soap-Bubble
8 and the blood-to-myocardium
contrast ratio (CR) measurements using 3DSlicer
9. Statistical
significance was evaluated using paired t-tests. RCA detection rates were compared for the two
methods.
Results
Overall, fNAV reconstructions were significantly faster
than 5D reconstructions (39±6 minutes vs 479±88 minutes, p<10-13),
while achieving comparable image quality (Fig.2). Reformats of
the RCA were also similar; however, the fNAV approach provided improved visual vessel conspicuity at specific locations (Fig.3).
Quantitative comparisons revealed improved CR with fNAV (p<10-4),
comparable visible RCA length, equal RCA detection rates and no statistically
significant difference in %VS (Table1).Discussion and Conclusion
Using our fNAV approach, we were able to correct
free-running whole-heart CMRA data for non-rigid respiratory motion and by
extending the auto-focusing concept to retrospective cardiac soft-gating, we
were able to provide subject-specific compensation of cardiac motion that
accounts for heart-rate variability. This combined approach produced high-quality
CMRA images that were comparable to previously established 5D image
reconstructions of the same data, but with a much shorter reconstruction time. In the
current work, we used fNAV to produce static 3D volumes in contrast to the
dynamic 5D image reconstructions. In principle, dynamic data could be generated
by the fNAV method by shifting the cardiac soft-gating window, and provided
enough data remain in our temporal footprint during peak motion. Furthermore,
in patient populations where un-resolved intra-bin respiratory motion and
heart-rate irregularities may be further amplified
and degrade image quality of 5D
imaging, our approach may be particularly beneficial, but this remains to be
studied. Finally, if sufficient SNR gain could be achieved whilst preserving
%VS, the increased amount of signal could be traded for higher resolution or
abbreviated scan times, potentially improving the current clinical utility of
3D radial CMRA.Acknowledgements
No acknowledgement found.References
- Di Sopra, L., Piccini, D., Coppo, S., Stuber, M., & Yerly, J. (2019). An automated approach to fully self‐gated free‐running cardiac and respiratory motion‐resolved 5D whole‐heart MRI. Magnetic
resonance in medicine, 82(6), 2118-2132.
- Feng, L. I., Coppo,
S., Piccini, D., Yerly, J., Lim, R. P., Masci, P. G., ... & Otazo, R.
(2018). 5D whole‐heart sparse MRI. Magnetic
resonance in medicine, 79(2), 826-838.
- Roy, C. W., Heerfordt J., Piccini, D., Schwitter, J.,
& Stuber, M. Motion Compensated Coronary MRA using Focused Navigation
(fNAV). Proc. Intl. Soc. Mag.
Reson. Med. 28 1321 (2020).
- Bastiaansen, J. A.,
& Stuber, M. (2018). Flexible water excitation for fat‐free MRI at 3T using lipid insensitive binomial off‐resonant RF excitation (LIBRE) pulses. Magnetic resonance in
medicine, 79(6), 3007-3017.
- Bastiaansen, J. A., van Heeswijk, R.
B., Stuber, M., & Piccini, D. (2019). Noncontrast free-breathing
respiratory self-navigated coronary artery cardiovascular magnetic resonance
angiography at 3 T using lipid insensitive binomial off-resonant excitation (LIBRE). Journal
of Cardiovascular Magnetic Resonance, 21(1), 38.
- Masala, N., Bastiaansen, J. A., Di
Sopra, L., Roy, C. W., Piccini, D., Yerly, J., ... & Stuber, M. (2020).
Free‐running 5D coronary MR
angiography at 1.5 T using LIBRE water excitation pulses. Magnetic
resonance in medicine, 84(3), 1470-1485.
- Roes, S. D., Korosoglou, G., Schär, M., Westenberg, J. J., Van Osch, M. J., De Roos, A., & Stuber, M. (2008). Correction for heart rate variability during 3D whole heart MR coronary angiography. Journal of Magnetic Resonance Imaging: An Official Journal of the International Society for Magnetic Resonance in Medicine, 27(5), 1046-1053.
- Etienne,
A., Botnar, R. M., Van Muiswinkel, A. M., Boesiger, P., Manning, W. J., &
Stuber, M. (2002). “Soap‐Bubble” visualization and quantitative analysis of 3D coronary magnetic
resonance angiograms. Magnetic Resonance in Medicine: An Official Journal
of the International Society for Magnetic Resonance in Medicine, 48(4),
658-666.
- Fedorov, A., Beichel, R., Kalpathy-Cramer, J., Finet, J., Fillion-Robin, J. C.,
Pujol, S., ... & Buatti, J. (2012). 3D Slicer as an image computing
platform for the Quantitative Imaging Network. Magnetic resonance imaging, 30(9),
1323-1341.