Motion Compensation Strategies - UCL Centre for Medical Image

advertisement
Motion Compensation Strategies
David Atkinson D.Atkinson@ucl.ac.uk
Imaging Strategies Course, ISMRM 2011 Montreal
Introduction
Motion during MR imaging remains a significant problem. This article outlines the some of
the acquisition strategies available for motion compensation. The article and the accompanying
lecture are intended as an educational overview and not a comprehensive review. More details
of many of the pulse sequences are described in the book by Bernstein et al [1]. Further details
about motion correction in image reconstruction are available in the Friday Sunrise Educational
Course - Image Reconstruction.
The Effect of Motion
A conventional MR image is the Fourier Transform (FT) of the acquired k-space data. The
acquisition of k-space is often slow compared to physiological motion due to timing constraints
imposed by nuclear time constants (T1 and T2), safe gradient switching speeds and power deposition. The summation operations in the discrete FT mean that each image pixel is composed
of a weighted sum of every point in k-space. Thus any alteration to k-space at any time during
the acquisition can potentially affect every pixel in the image - typical motion artefacts being
image blurring or ghosting [2]. Motion can alter k-space in three ways; movement during the
application of a gradient can cause phase errors, motion between gradients can result in phase
and amplitude inconsistencies in k-space, and the movement of tissue can affect steady state conditions or result in tissue moving out of plane. Positional information is encoded by gradients
into the phase of k-space, thus unintended alterations to the phase (or magnitude) of k-space get
translated into positional errors by the FT operation, typically leading to blurring and ghosting.
Acquisition Strategies to Reduce Motion Effects
The rest of this article concentrates on technical issues but good patient preparation can reduce
motion associated with anxiety and claustrophobia. In babies tricks such as ‘feed-wrap-scan’
can allow a baby to be scanned asleep. In other children, sedation or anesthetics may be used
where the cost and risks are outweighed by the expected diagnostic value of the scan. If scan
time is short enough, breath-holding can be used to reduce respiratory artefacts.
Many acquisition strategies have been proposed to reduce motion effects. The problem is fundamentally challenging because ideally we need to know the motion of every pixel at every time
point and change the running scan and/or reconstruction. However, even though such complete
information is hard to obtain, motion compensation strategies can be effective. The techniques
can be broadly divided into sequences with reduced motion sensitivity, and, measurements that
enable corrections.
1
Sequences With Reduced Motion Sensitivity
Averaging
Many sequences acquire multiple signal averages during free breathing. Although multiple averaging does not correct for motion, it weights coherent signal (from stationary tissue or that
with a consistent average position) greater than incoherent motion, often resulting in improved
quality scans.
Reducing Signal from Moving Tissue
Spatial saturation bands can be positioned to prevent moving tissue from providing a signal, for
example upstream of inflowing blood or over abdominal regions to reduce respiratory artefacts.
Suppression of either fat or CSF can also be used, especially with imaging sequences where fat
or CSF would appear bright and lead to prominent artefacts.
Phase Encode Re-Ordering
Motion affected k-space can be viewed as the ideal, motion-free, k-space multiplied by some
motion-dependent modulation function. In the image domain, this k-space multiplication is
equivalent to convolution of the ideal image with the Fourier Transform of the modulation
function. If the convolved function has multiple peaks, then replications (ghosts) will be seen
in the image. To reduce this ghosting, the order of k-space acquisition can be changed so that
the modulation function (and thus its FT) does not oscillate. Typically these schemes re-order
during respiration and include ROPE [3], COPE [4] and HOPE [5]. Re-ordering does not correct
for motion, but it can have a positive impact on image appearance.
Gating and Triggering
Prospectively triggering a sequence using an ECG or pulse oximeter is commonly used to capture
k-space at consistent times in the cardiac cycle over multiple heart beats. Retrospective gating
records k-space and the cardiac signal continuously and then assembles k-space into the relevant
cardiac phases after the acquisition. These techniques work well, especially if the heart rate is
regular. Respiratory triggering and gating can also be used with the respiratory phase inferred
from pneumatic bellows or the position of the diaphragm as determined from navigators (see
later section on navigation).
Gradient Moment Nulling
A gradient means that the magnetic field strength varies as a function of position. When the
overall area (zeroth moment) of a series of gradients is zero, stationary spins experience no net
phase changes. However, spins that move in the gradient direction experience different fields and
will not always be re-phased at the time when the net area is zero. This phase accrual is used
to provide diffusion contrast in DWI and velocity measurements in phase contrast sequences. In
general, the phase accrued by spins moving with a constant velocity is determined by the first
order moment of the gradients (the integral over time of the product of the gradient and time).
When the first order moment is non-zero, unwanted motion leads to phases that can cause signal
loss and image ghosting. The technique of gradient moment nulling aims to provide a set of
gradients with a low first order moment at relevant times. For slice selection, the relevant time
is the end of the slice select gradients whilst for the readout and phase-encode, the relevant
time is at the echo peak (which occurs after the phase encode gradients and during the readout
gradients). More references, details about the design, and compensating for acceleration in
2
addition to velocity are available in Bernstein et al [1]. The subject is still a topic of on going
research, e.g. [6], and is also related to isotropic diffusion weighting e.g. [7] in moving structures.
Short TE and Non-Cartesian Sequences
For sequences such as spiral and radial where the acquisition starts near the centre of k-space,
there is relatively little gradient-on time before passing through the centre of k-space (where the
echo signal is largest) and thus the first-order moment at this point in time is low, and hence
the sensitivity to motion can be low. Often the motion artefacts associated with non-Cartesian
sequences are more benign than for regular Cartesian sampling (they can be less pronounced
and also may appear outside the region of interest) [8].
Motion Measurements: Tracking or Retrospective Compensation
Techniques such as navigators can be used to measure at least some components of motion which
can then be used either to alter a running scan (prospective correction), or to guide reacquisition
of parts of the data, or to guide a retrospective correction. Prospective correction requires rapid
data processing and feedback, but has the advantage of potentially avoiding missing data and
spin history problems. Retrospective correction is the only technique that can in principle correct
for non-affine components of motion.
Navigation
Navigation refers to inferring some components of motion from MR data, usually some additional
measurements within a sequence, rather than an external measure. One of the simplest forms of
navigation is to acquire extra non phase-encoded lines of k-space. Fourier Transformed, these give
projections through the object from which some components of motion can be inferred. This was
one of the early techniques [9] and recent applications of this central profile approach include its
use in balanced cardiac sequences [10]. When the acquisition is from multiple coils with different
local sensitivities, motion about both respiratory and cardiac cycles can be obtained from these
projections, e.g. [11].
Commonly the term ‘navigators’ refers to the signal from a cylindrical or ‘pencil beam’
shaped excitation through the diaphragm. The liver/lung border provides good contrast and
it is relatively easy to track the diaphragm here. From this position information, the scan
can be gated to acquire only during a user-determined window of diaphragm positions and the
slice/volume position may also be updated (also see [12] and Prospective Correction below).
The pencil beam profile is achieved through the use of a spiral excitation k-space. The previous
techniques are often known as 1D navigators or navigator echoes.
More complex rotational motion can be measured using orbital [13], spherical [14] and rapid
cloverleaf [15] paths through the acquisition k-space, as well as floating navigators with a small
phase encoding [16, 17].
As a means of trying to detect specifically cardiac motion, navigators from the fat signal
near the coronaries have been developed [18]. It is also possible to capture the FID following an
RF pulse and obtain motion information [19, 20, 21, 22] or make use of the butterfly or bow tie
pattern when k-space is being rewound or prepared [23, 24].
The central region of k-space can be scanned rapidly and provides a low resolution image
from which motion can be determined. Sequences such as radial and spiral can be inherently selfnavigated having a high sampling density near k-space centre (as well as some natural robustness
to motion) and can be used to infer motion, e.g. in whole heart coronary imaging using a 3D radial
3
acquisition [25]. The PROPELLER [26] method obtains low resolution images from rotating
blades of k-space that are used for motion compensation.
In multi-shot diffusion weighted imaging, 2D navigators have been used to measure the phase
changes caused by motion during diffusion sensitization [27] for retrospective correction (e.g. [28])
or as a guide to re-acquisition of the worst affected shots [29].
External Sensors
External sensors can provide additional information with no compromise to the scan sequence
or imaging. Optically tracked markers or active markers cannot directly measure the motion
of internal tissues but have been the subject of recent demonstrations for neurological imaging
[30, 31, 32]. Recently an ultrasound probe has been used to monitor internal motion for cardiac
imaging [33].
Prospective Correction
The position of an imaged slice or volume can be adjusted during acquisition by altering the
gradients and/or RF pulses with the intention of tracking a region of interest. Successful tracking
can reduce k-space inconsistencies due to both position changes and problems associated with inconsistent spin histories. A 1D navigator can be used for slice tracking [34], but for more complex
motion compensation, a more complete description of motion is required. This can come from
rapid low resolution images with motion determined by image registration [35, 36, 37], or using
orthogonal spiral navigators [38, 39]. Note that using prospective alterations to gradients, it is
possible to compensate for 3D affine motion. For non-rigid motion components that are more
complex than affine, currently the only general method for compensation is to use retrospective
techniques in image reconstruction (see Sunrise Educational Course on Motion Correction in Image Reconstruction). Although prospective correction aims to effectively fix the imaging volume
of interest, this can lead to a relative motion of coil sensitivities, distortions associated with B0
variations [40] and other tissue moving non-rigidly. Many prospective techniques also include
some postprocessing to reduce these effects and overcome any latency issues. Spectroscopy as
well as imaging can also benefit from these methods e.g. [41].
Outlook
It seems likely that in the near future, most diagnostic quality scans will not be faster than
physiological motion. This means there will continue to be a role for developments in motion
correction. In the future we might expect to see scan time being fully exploited for motion
measures and a more coherent integration of prospective scanner control, motion measures,
learned models of motion and reconstructions that include not just inter-shot motion but effects
such as motion during a shot, T2 signal decay and the relative motion of field inhomogeneities
and coil sensitivities. The area of parallel RF transmit offers the possibility of new ways to
monitor and perhaps compensate prospectively for motion during acquisition.
References
[1] M.A. Bernstein, K.F. King, and X.J. Zhou. Handbook of MRI Pulse Sequences. Elsevier,
2004.
[2] M. L. Wood and R. M. Henkelman. MR image artifacts from periodic motion. Med Phys,
12(2):143–151, 1985.
4
[3] D. R. Bailes, D. J. Gilderdale, G. M. Bydder, A. G. Collins, and D. N. Firmin. Respiratory
ordered phase encoding (ROPE): a method for reducing respiratory motion artefacts in MR
imaging. J Comput Assist Tomogr, 9(4):835–838, 1985.
[4] E. M. Haacke and J. L. Patrick. Reducing motion artifacts in two-dimensional fourier
transform imaging. Magn Reson Imaging, 4(4):359–376, 1986.
[5] P. Jhooti, F. Wiesmann, A. M. Taylor, P. D. Gatehouse, G. Z. Yang, J. Keegan, D. J.
Pennell, and D. N. Firmin. Hybrid ordered phase encoding (HOPE): an improved approach
for respiratory artifact reduction. J Magn Reson Imaging, 8(4):968–980, 1998.
[6] Kurt Majewski, Oliver Heid, and Thomas Kluge. MRI pulse sequence design with first-order
gradient moment nulling in arbitrary directions by solving a polynomial program. IEEE
Trans Med Imaging, 29(6):1252–1259, Jun 2010.
[7] E. C. Wong, R. W. Cox, and A. W. Song. Optimized isotropic diffusion weighting. Magn
Reson Med, 34(2):139–143, Aug 1995.
[8] G. H. Glover and J. M. Pauly. Projection reconstruction techniques for reduction of motion
effects in MRI. Magn Reson Med, 28(2):275–289, Dec 1992.
[9] R. L. Ehman and J. P. Felmlee. Adaptive technique for high-definition MR imaging of
moving structures. Radiology, 173(1):255–263, Oct 1989.
[10] Sergio Uribe, Vivek Muthurangu, Redha Boubertakh, Tobias Schaeffter, Reza Razavi, Derek
L G Hill, and Michael S Hansen. Whole-heart cine MRI using real-time respiratory selfgating. Magn Reson Med, 57(3):606–613, Mar 2007.
[11] Freddy Odille, Sergio Uribe, Philip G Batchelor, Claudia Prieto, Tobias Schaeffter, and
David Atkinson. Model-based reconstruction for cardiac cine MRI without ECG or breath
holding. Magn Reson Med, 63(5):1247–1257, May 2010.
[12] Y. Wang, S. J. Riederer, and R. L. Ehman. Respiratory motion of the heart: kinematics
and the implications for the spatial resolution in coronary imaging. Magn Reson Med,
33(5):713–719, May 1995.
[13] Z. W. Fu, Y. Wang, R. C. Grimm, P. J. Rossman, J. P. Felmlee, S. J. Riederer, and R. L.
Ehman. Orbital navigator echoes for motion measurements in magnetic resonance imaging.
Magn Reson Med, 34(5):746–753, Nov 1995.
[14] Edward Brian Welch, Armando Manduca, Roger C Grimm, Heidi A Ward, and Clifford R
Jack. Spherical navigator echoes for full 3D rigid body motion measurement in MRI. Magn
Reson Med, 47(1):32–41, Jan 2002.
[15] Andr J W van der Kouwe, Thomas Benner, and Anders M Dale. Real-time rigid body
motion correction and shimming using cloverleaf navigators. Magn Reson Med, 56(5):1019–
1032, Nov 2006.
[16] Yasser M Kadah, Ayman A Abaza, Ahmed S Fahmy, Abou-Bakr M Youssef, Keith Heberlein, and Xiaoping P Hu. Floating navigator echo (FNAV) for in-plane 2D translational
motion estimation. Magn Reson Med, 51(2):403–407, Feb 2004.
[17] Wei Lin, Feng Huang, Peter Brnert, Yu Li, and Arne Reykowski. Motion correction using
an enhanced floating navigator and GRAPPA operations. Magn Reson Med, 63(2):339–348,
Feb 2010.
5
[18] Thanh D Nguyen, Anthony Nuval, Suresh Mulukutla, and Yi Wang. Direct monitoring of
coronary artery motion with cardiac fat navigator echoes. Magn Reson Med, 50(2):235–241,
Aug 2003.
[19] Xiaoping Hu and Seong-Gi Kim. Reduction of signal fluctuation in functional MRI using
navigator echoes. Magn Reson Imaging, 31:495–503, 1994.
[20] N. Gai and L. Axel. Correction of motion artifacts in linogram and projection reconstruction
MRI using geometry and consistency constraints. Med Phys, 23(2):251–262, Feb 1996.
[21] Anja C S Brau and Jean H Brittain. Generalized self-navigated motion detection technique:
Preliminary investigation in abdominal imaging. Magn Reson Med, 55(2):263–270, Feb 2006.
[22] Martin Buehrer, Jelena Curcic, Peter Boesiger, and Sebastian Kozerke. Prospective selfgating for simultaneous compensation of cardiac and respiratory motion. Magn Reson Med,
60(3):683–690, Sep 2008.
[23] M. Lustig, C. H. Cunningham, E. Daniyalzade, and J. M. Pauly. Butterfly: A self navigating
cartesian trajectory. In ISMRM, page 865, 2007.
[24] Candice A Bookwalter, Mark A Griswold, and Jeffrey L Duerk. Multiple overlapping kspace junctions for investigating translating objects (MOJITO). IEEE Trans Med Imaging,
29(2):339–349, Feb 2010.
[25] C. Stehning, P. Brnert, K. Nehrke, H. Eggers, and M. Stuber. Free-breathing whole-heart
coronary MRA with 3D radial SSFP and self-navigated image reconstruction. Magn Reson
Med, 54(2):476–480, Aug 2005.
[26] J. G. Pipe. Motion correction with PROPELLER MRI: application to head motion and
free-breathing cardiac imaging. Magn Reson Med, 42(5):963–969, Nov 1999.
[27] K. Butts, A. de Crespigny, J. M. Pauly, and M. Moseley. Diffusion-weighted interleaved echoplanar imaging with a pair of orthogonal navigator echoes. Magn Reson Med, 35(5):763–770,
May 1996.
[28] David Atkinson, Serena Counsell, Joseph V Hajnal, Philip G Batchelor, Derek L G Hill,
and David J Larkman. Nonlinear phase correction of navigated multi-coil diffusion images.
Magn Reson Med, 56(5):1135–1139, Nov 2006.
[29] David A Porter and Robin M Heidemann. High resolution diffusion-weighted imaging using
readout-segmented echo-planar imaging, parallel imaging and a two-dimensional navigatorbased reacquisition. Magn Reson Med, 62(2):468–475, Aug 2009.
[30] M. Zaitsev, C. Dold, G. Sakas, J. Hennig, and O. Speck. Magnetic resonance imaging of
freely moving objects: prospective real-time motion correction using an external optical
motion tracking system. Neuroimage, 31(3):1038–1050, Jul 2006.
[31] Lei Qin, Peter van Gelderen, John Andrew Derbyshire, Fenghua Jin, Jongho Lee, Jacco A
de Zwart, Yang Tao, and Jeff H Duyn. Prospective head-movement correction for highresolution MRI using an in-bore optical tracking system. Magn Reson Med, 62(4):924–934,
Oct 2009.
[32] Melvyn B Ooi, Sascha Krueger, William J Thomas, Srirama V Swaminathan, and Truman R
Brown. Prospective real-time correction for arbitrary head motion using active markers.
Magn Reson Med, 62(4):943–954, Oct 2009.
6
[33] David A Feinberg, Daniel Giese, D. Andre Bongers, Sudhir Ramanna, Maxim Zaitsev,
Michael Markl, and Matthias Gnther. Hybrid ultrasound MRI for improved cardiac imaging
and real-time respiration control. Magn Reson Med, 63(2):290–296, Feb 2010.
[34] P. G. Danias, M. V. McConnell, V. C. Khasgiwala, M. L. Chuang, R. R. Edelman, and W. J.
Manning. Prospective navigator correction of image position for coronary MR angiography.
Radiology, 203(3):733–736, Jun 1997.
[35] Dirk Manke, Kay Nehrke, and Peter Bòˆrnert. Novel prospective respiratory motion correction approach for free-breathing coronary MR angiography using a patient-adapted affine
motion model. Magn Reson Med, 50(1):122–131, Jul 2003.
[36] Kay Nehrke and Peter Bòˆrnert. Prospective correction of affine motion for arbitrary MR
sequences on a clinical scanner. Magn Reson Med, 54(5):1130–1138, Nov 2005.
[37] S. Thesen, O. Heid, E. Mueller, and L. R. Schad. Prospective acquisition correction for head
motion with image-based tracking for real-time fmri. Magn Reson Med, 44(3):457–465, Sep
2000.
[38] Nathan White, Cooper Roddey, Ajit Shankaranarayanan, Eric Han, Dan Rettmann, Juan
Santos, Josh Kuperman, and Anders Dale. PROMO: Real-time prospective motion correction in MRI using image-based tracking. Magn Reson Med, 63(1):91–105, Jan 2010.
[39] Timothy T Brown, Joshua M Kuperman, Matthew Erhart, Nathan S White, J. Cooper
Roddey, Ajit Shankaranarayanan, Eric T Han, Dan Rettmann, and Anders M Dale. Prospective motion correction of high-resolution magnetic resonance imaging data in children. Neuroimage, 53(1):139–145, Oct 2010.
[40] P. Jezzard and R. S. Balaban. Correction for geometric distortion in echo planar images
from B0 field variations. Magn Reson Med, 34(1):65–73, Jul 1995.
[41] Brian Keating, Weiran Deng, J. Cooper Roddey, Nathan White, Anders Dale, V. Andrew
Stenger, and Thomas Ernst. Prospective motion correction for single-voxel 1H MR spectroscopy. Magn Reson Med, 64(3):672–679, Sep 2010.
7
Download