Kostas Haris1,2, Erik Hedström1,3, Fabian Kording4,5, Sebastian Bidhult1,6, Frederik Testud7, Katarina Steding-Ehrenborg1,8, Christian Ruprecht4,5, Einar Heiberg1,6, Håkan Arheden1, and Anthony Aletras1,2
1Lund University, Skåne University Hospital, Department of Clinical Sciences Lund, Clinical Physiology, Lund, Sweden, 2Laboratory of Computing, Medical Informatics and Biomedical-Imaging Technologies, School of Medicine, Aristotle University, Thessaloniki, Greece, 3Lund University, Skåne University Hospital, Department of Clinical Sciences Lund, Diagnostic Radiology, Lund, Sweden, 4Department of Diagnostic and Interventional Radiology, University Medical Center Hamburg-Eppendorf, Hamburg, Germany, 5North medical GmbHMartinistraße 52, 20246, Hamburg, Germany, 6Faculty of Engineering, Lund University, Department of Biomedical Engineering, Lund, Sweden, 7Siemens Healthcare AB, Malmö, Sweden, 8Lund University, Department of Health Sciences, Physiotherapy, Lund, Sweden
Synopsis
We investigated the feasibility of free-breathing
fetal cine cardiac MRI using Doppler Ultrasound retrospective binning combined
with continuous tiny golden angle radial sampling and compressed sensing via
the tyGRASP reconstruction method. Motion due to maternal respiration and fetal
movement was estimated by registration applied to reconstructed images of low temporal
resolution. K-space radial profiles containing motion were either corrected or
rejected from the reconstruction process. The reconstruction results demonstrated good image and
diagnostic quality showing that the absence of fetal ECG can be bypassed
in a computationally efficient way by Doppler Ultrasound even in the presence
of motion.
Background
Fetal cardiac
MRI is a viable tool for evaluating the fetal heart both for research and
clinically. The extraction of a reliable fetal ECG is an open problem due to
low voltages, contamination by the maternal ECG, and MRI-related interference.
Solutions based on retrospectively sorting the acquired data into the proper
cardiac phases have been proposed and applied using standard pulse sequences1,2.
The aim of this study was to demonstrate the
feasibility of free-breathing fetal cine CMR using Doppler Ultrasound-gated
retrospective binning combined with continuous tiny golden angle radial
sampling and compressed sensing tyGRASP (tiny Golden-angle
RAdial Sparse Parallel)3,4.
Motion due to maternal respiration and fetal movement is estimated and the
corresponding k-space profiles are either corrected or rejected from the
reconstruction process.Methods
Regional Review
Board approval and written informed consent from the study participants were
obtained. Exams were performed on a 1.5T MAGNETOM Aera system (Siemens
Healthcare GmbH, Erlangen, Germany).
A flowchart of the proposed reconstruction
method is shown in Figure 1. Short-axis and four-chamber view human fetal cardiac
data were acquired using a radial trajectory with a constant
azimuthal increment of tiny golden angle of order seven4 corresponding to 23.63o with a prototype balanced Steady State Free
Precession sequence with the following parameters: TE/TR=1.88/3.7ms, flip
angle 60ο, pixel size 0.7x0.7x4 mm3. A total of 13 short-axis and 6 four-chamber slices with 4,000 radial spokes-per-slice were acquired during maternal free breathing in two fetuses at
gestational week 35. The cardiac synchronization signal required for the binning
stage was produced by a prototype MRI-compatible Doppler Ultrasound (DUS)
device (sMaRT-sync, north medical Gmbh, Germany)5. The DUS synchronization
signals were determined using raw Doppler signals and processed using software
written in MATLAB (The MathWorks, Natick, MA, USA). The number of cardiac
phases depended on the number of available cardiac cycles, with typical values
ranging between 16 to 22 phases. Motion due to maternal respiration and fetal
movement was detected using translational registration applied to an image series with low temporal resolution: for each cardiac cycle, only one image was
reconstructed using the corresponding radial profiles which were available via
the DUS synchronization signal. As expected, the motion of the pulsating fetal
heart was smoothed out, and therefore, any displacements depicted in these
images were due to maternal respiration and fetal movement. When the detected
movement was approximately translational, it was corrected by the appropriate
shift operation in k-space. Otherwise, the corresponding profiles were not used
for the reconstruction. The final images were obtained by using the tyGRASP method3.
The reconstruction time was approximately 5 minutes-per-slice on a standard computer.
An observer with 17 years of cardiovascular MRI experience scored image quality and cardiac/extracardiac
diagnostic quality on 4-grade scales where intrauterine image quality 1=low
quality and/or high degree of artifact; 2=moderate quality and/or some
artifacts; 3=high quality and/or few artifacts; and 4=high quality; and
diagnostic quality 1=inadequate; 2=low; 3=moderate; 4=high. Cardiac diagnostic
quality was based on discernible epi- and endocardial borders to surrounding
tissue and blood pool, visualized papillary muscles/trabeculation, potential of
measurement of wall thickness, visualized wall thickening during contraction,
and potential detection of ventricular septal defects or other cardiac
pathology; and extracardiac diagnostic quality was based on thoracic structures
outside the heart, brain structures, contrast between brain and cerebrospinal
fluid, spinal cord visualization, abdominal structures, and placental structure
and attachment. Scores are presented as median (range).
Results
Intrauterine image quality, cardiac
and extracardiac diagnostic quality were 2 (2-3), 3 (3-4) and 3 (3-4),
respectively. In Figure 2, the sixteen reconstructed cardiac
phases of two different slices of a short-axis view are shown cropped and
magnified. In these slices, no translational motion was identified and
therefore, their reconstruction was based on the k-space profiles left after
the rejection of the ones containing movement. Figure 3 shows an indicative
translational motion correction result of a four-chamber view slice. The
reconstructed images without motion correction are blurred and the boundaries
of the anatomical structures are less clear. The improvement in sharpness after
motion correction is evident. These initial results on two subjects (four
multi-slice datasets, 36 free-breathing slices in total) verified qualitatively
that the DUS-based gating signal is accurate even in the presence of shallow
motion, and therefore, it can be used for the cardiac binning stage.Conclusions
This
study demonstrates that free-breathing Doppler Ultrasound-gated fetal cardiac
cine MRI based on tyGRASP is possible with good image and diagnostic quality.
Maternal respiratory motion and fetal movement were detected via registration
in low-temporal DUS-based reconstructions, and compensated for.
In the future, these methods can be tuned and automated thus minimizing user
interaction and further promoting the use of fetal cardiovascular imaging
clinically.Acknowledgements
Funding
was provided by the Swedish Heart and Lung Foundation, the Skåne University
Hospital and by the Medical Faculty at Lund University. The authors thank and
acknowledge Siemens Healthcare for providing support within the Master Research
Agreement with Lund University and the Region of Skåne.References
- Jansz MS et al MRM
64:1304 (2010)
- Haris K et al JMRI 46:207 (2017)
- Feng L et al MRM 72:707
(2014)
- Wundrak S et al IEEE
TMI 34:1262 (2015)
- Kording F et al MRM 74:1257 (2015)