Andrew James Patterson1, Martin John Graves1, and David John Lomas1
1Radiology, University of Cambridge & Addenbrooke’s Hospital, Cambridge, United Kingdom
Synopsis
This work describes the development of a novel approach for prescribing
cardiac scan plane locations which will benefit cardiac examinations where the
anatomy is abnormal (for example, in congenital anomalies). We have developed a
desktop sized constrained surface controller which mimics an ultrasound transducer.
This enables retrospective reformatting of a 3D localizer image to determine the
desired cardiac view. The cardiac scan plane is then prospectively transferred
to a 2D CINE bSSFP sequence to enable interactive assessment of cardiac function.
Initial testing has established that our approach allows us to replicate
standard cardiac views.
Purpose
Standard cardiac protocols involve multiple
localization scans which require consecutive breath-holds (the de facto
standard), alternatively system vendors have developed automated prescription techniques
to improve time efficiency and repeatability.[1] Both of these to
some extent assume normal cardiac chamber morphology. The application of both
of these techniques is challenging when the anatomy is abnormal, for example,
in congenital anomalies. The aim of this work is to develop a novel approach to
prescribing cardiac scan plane locations using an initial 3D "localizer"
dataset of the heart which is then retrospectively reformatted to define the
desired acquisition planes using a constrained surface controller (CSC).[2]
The geometry for these planes is then sent to the MRI system to facilitate high-quality CINE acquisitions at the desired
locations.
Methods
A desktop sized planar constrained surface controller (CSC)
was utilized (figure 1). To prevent disorientation the motion of the CSC is analogous
to an ultrasound transducer where the operator is constrained to a reference
planar surface mimicking the skin surface whilst limiting the total number of
degrees of freedom to four (two translations and two rotations). Purpose built
software was developed to enable 2D planar reformatting of the 3D localizer
image and to transfer scan plane geometry. The software was developed in C++
incorporating the following open-source libraries: ITK (Insight Toolkit, www.itk.org), VTK (Visualization Toolkit, www.vtk.org) and QT (www.qt.io)
for image processing, image visualization and the graphical user interface
respectively on a 64-bit Linux operating system. A 3D sagittal, ECG triggered,
navigated, balanced steady-state-free-precession (bSSFP) acquisition was
acquired for localization (TE/TR= 1.4/3.2ms, FA=15°, FOV=34, matrix size=256×256×60,
ASSET factor=2 (phase)). T
2 preparation was included to improve contrast
between the blood and myocardium. The volumetric localizer images were imported
into the CSC software. Homogenous transfer matrices (HTM) were evaluated to
reformat the localizer image and to prospectively define the scan plane
positions for subsequent CINE acquisitions. The HTM were derived separately for
the image reformats and the scanner acquisitions to match geometrical prescriptions
after accounting for the fact that both systems have different origins and thus
centers of rotation. A 2D ECG-triggered CINE bSSFP sequence was acquired
(TE/TR= 1.1/3.2, FA=35°, matrix size=192×128, phases=20, scan time=20s) after
modifying it to enable interactive scan plane prescription updates from the CSC.
The sequence was modified to support data transfer using the User Datagram Protocol (UDP) which enables the
sequence to pause and listen for incoming geometry packets from the CSC software.
Upon navigating the CSC to the desired cardiac view (figure 2), a key-press
event transfers the geometry to the scanner. Following the breath-held
acquisition, the scanner pauses again until the next geometry packet is
transmitted. To assist with the scan plan prescription relative sagittal and
coronal transforms, with superimposed guidelines, are defined to represent the
current imaging plane (figure 3). The completed system was initially tested on
standard geometric phantoms using a 3T whole body scanner (MR750, GE Healthcare,
Waukesha, WI) and an 8-channel receiver coil.
Results
The 2D planar images reconstructed at 4Hz (using
a 6-core Intel i7 3.2GHz). Testing on geometrical phantoms established the
required systematic correction to reformat the image plane and scan plane at
matched geometrical locations.
Discussion
The results suggest a reformat rate
correspondent with requirements to achieve interactive feedback which has been achieved
by utilizing C++ (i.e. a compiled program language). We found it was possible
to reformat standard cardiac views using the proposed paradigm.
Conclusion
This work demonstrates the development of a
novel interactive approach to facilitate cardiac scan plane prescription based
on using a CSC. This is potentially useful in
complex cardiac anatomy cases where the conventional scan plane prescription approach
may be either inappropriate or excessively time consuming.. This interactive
approach allows for the combination of high resolution reformats of 3D bSSFP and
targeted 2D CINE acquisitions to assess anatomy and function in complex cardiac
cases. Future work will compare this CSC
based method against the current techniques initially in normal volunteers and
then in patients with complex cardiac anatomy.
Acknowledgements
We would like to thank Addenbrooke’s Charitable
Trust, the NIHR Cambridge Biomedical Research Centre, and the MRIS
Radiographers.References
[1] Lu X, Jolly MP, Georgescu
B, et al. Automatic view planning for cardiac MRI acquisition. Med Image Comput
Comput Assist Interv. 2011. 14(3): 479-86
[2] Graves MJ, Black RT, Lomas
DJ. Constrained surface controllers for three-dimensional image data
reformatting. Radiology. 2009. 252(1): 218-24