| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
|
|
|
|||||||
|
|||||||||
Editorial |
1 From the Department of Radiology, University of California-Davis Imaging Center, 4701 X St, Sacramento, CA 95817. Received December 3, 2003; accepted December 4. Address correspondence to the author (e-mail: mhbuonocore@ucdavis.edu).
Index terms: Editorials • Heart, MR, 511.121412, 511.12144 • Myocardium, MR, 511.121412, 511.12144
The article by Kim et (1) presents a significant step forward in the evolution of magnetic resonance (MR) imaging techniques to measure two-dimensional intramyocardial displacement and strain. The pulse sequence described in reference 1 is the latest in a set of sequences that have built on an innovative technique called displacement encoding with stimulated echoes (DENSE) (2). The DENSE pulse sequence directly encodes the shift in position (ie, displacement) of tissue in the phase of the transverse magnetization. The first sequence, proposed in reference 2, required a 4-minute acquisition for one end-systolic displacement-encoded MR image at one section location, while this latest publication reports a breath-hold sequence using short echo train echo-planar MR imaging (fast gradient-recalled-echo echo train) (3) with a versatile artifact-suppression technique (4) that acquires a time series of 12 displacement-encoded MR images across roughly 
of the cardiac cycle at one section location in 13 heartbeats. At this stage, a clinical evaluation of this sequence in comparison with standard tagging methods using the latest tag identification postprocessing algorithms is warranted.
Displacement encoding by phase contrast is a significant innovation, on the same level as velocity encoding by phase contrast. The mechanism whereby the magnetization phase angle becomes a function of the shift in position (ie, displacement, with units of length) of the tissue is not immediately obvious. The publication on DENSE MR imaging by Aletras et al (2) described the process of displacement encoding. Transverse magnetization is initially position encoded by an encoding gradient pulse, stored temporarily as longitudinal magnetization during the "mixing time" and then refocused just prior to data acquisition by an "unencoding" gradient pulse. The unencoding gradient pulse causes the transverse magnetization to refocus at a phase angle that is linearly proportional to the displacement that occurred during the mixing time. In later publications (1,5,6), it was made more clear to the reader that the unencoding gradient pulse refocuses only half of the transverse magnetization that had been stored and that the other half of the magnetization is not refocused and destroys the displacement encoding if it is allowed to contribute signal to the data prior to Fourier image reconstruction. The encoding and unencoding gradient pulses are each given a large zeroth moment (gradient strength x duration), so that the spatial frequencies of this "unrefocused" magnetization are shifted to outside the spatial frequency range of the acquired data. The unwanted signal from this magnetization is present as an oscillating current in the receiver coil, but most if not all of this signal is eliminated by band-pass analog or digital filtering of the signal during data acquisition. The use of the unencoding gradient pulse to shift the spatial frequencies, to effectively eliminate the unrefocused magnetization, and to preserve the refocused magnetization whose phase is linearly proportional to the displacement is a key innovation distinguishing DENSE MR imaging from amplitude-modulation methods, such as spatial modulation of magnetization (7).
Before the development of DENSE MR imaging, spatial tagging and velocity-encoded phase imaging were used to measure myocardial motion and deformation (7–10) and allowed clinicians to establish a significant diagnostic potential for these types of measurements. Scanning options to improve image quality and new postprocessing algorithms to improve automation, speed, and reliability of the motion and deformation measurements continue to be developed. Nevertheless, DENSE MR imaging provides several fundamental technical advantages over these methods. Spatial tagging relies on manipulation of the magnetization by radiofrequency and gradient pulses to create a regular grid of signal voids superimposed on the heart muscle, and these signal voids move with the muscle as the heart contracts. Spatial tagging can be implemented on most standard clinical MR imaging systems, but the method requires that the spatial tags be separated by a few pixels; consequently, the spatial resolution of the derived deformation and strain measurements is low. Phase contrast velocity encoding derives these measurements at nearly the same resolution as that of the velocity-encoded MR images. However, while spatial resolution is not a limitation, encoding of the relatively low velocities of heart contraction requires large gradient strengths, fast gradient rise times, and relatively long echo times, which result in large phase-angle errors from secondary magnetic fields produced by eddy currents and by Maxwell (concomitant) fields. Because echo times are relatively long and gradient-recalled echoes must be used instead of spin echoes, the velocity-encoded images show greater signal loss due to B0 inhomogeneity at epicardial tissue borders. The accuracy and precision of the phase angles are much less in these low-signal areas that are often most important for disease assessment.
The pulse sequence described in reference 1 provides a more direct and versatile approach to reduce significant image artifacts from longitudinal magnetization generated by T1 recovery during the mixing time interval. In DENSE (2) and fast DENSE (5), the signal from this magnetization, referred to as the "T1 relaxation" signal, is eliminated by crusher gradient pulses applied in the section-select direction both before and after the mixing time interval in both displacement and reference MR images. Magnetization arising from T1 relaxation during the mixing time interval is affected by only the second crusher gradient and is significantly dephased at the echo, while the position-encoded magnetization is affected by both crusher gradients and experiences no net dephasing. This approach is not preferred because the large crusher gradients reduce desired as well as undesired signal by virtue of secondary fields caused by eddy currents and phase dispersion signal loss from tissue deformation and B0 inhomogeneity. Kim et al (1) used the complementary data acquisition technique, first described in reference 4 as complementary spatial modulation of magnetization to eliminate the T1 relaxation signal. With complementary data acquisition, an additional data set is acquired that is equivalent in timing to the original set but with negative position-encoded longitudinal magnetization. Subtraction of the complementary data set from the original data set yields a new data set with two times the normal displacement-encoded signal and near zero T1 relaxation signal. Since such subtraction is possible for all acquired cardiac phases regardless of timing, the technique can be used with multiphase acquisition. Another technique, which was proposed in meta-DENSE (6), uses a carefully timed inversion pulse in the mixing time interval to null the T1 relaxation signal. This technique does not work for multiphase acquisition, and even for one cardiac phase, the inversion pulse timing must be calibrated from a cine scout image set (to identify the time of end systole when nulling is to occur) and perhaps from another set of scout images to estimate myocardial T1. The complementary data acquisition and the inversion pulse in meta-DENSE both eliminate the need for crusher gradients used in the earlier DENSE sequences. However, because of its direct subtraction of unwanted signal and versatility with multiphase acquisition, complementary data acquisition will likely remain a fundamental part of any optimal displacement-encoding technique.
With respect to pulse sequence parameter selection in DENSE imaging, several choices are not yet settled. One of the major choices is the zeroth moment (gradient strength x duration) of the displacement-encoding gradient pulses. Low zeroth moment gradient pulses are desired to reduce the signal loss due to muscle deformation, yet displacement sensitivity is improved if the moments are increased to increase the phase angle accumulation per unit distance. Also, the moments must be large enough so that the unrefocused transverse magnetization is shifted to spatial frequencies that exceed the maximum spatial frequency acquired during data acquisition. In reference 1, half of the subjects were scanned with a spatial frequency shift (2 ke) of 0.7 cycle per pixel, which is greater than 0.5 cycle per pixel data acquisition limit, and half the subjects were scanned with a shift of 0.56 cycle per pixel, much closer to the edge of the limit. The number of k-space lines that should be acquired in the short echo train of the echo-planar imaging readout is also not settled. In reference 1, with elimination of crusher gradient pulses in the section-select direction and the use of smaller gradient strengths for displacement encoding, six echoes per excitation pulse provided acceptable image quality. However, only two echoes per excitation were used in fast DENSE (5). In addition, on some MR imaging systems, echo-planar images show Nyquist ghost artifacts, which reduce image quality and possibly clinical reliability. The severity of these artifacts may depend on the choice of echo train length, number of shots, and other sequence parameters. Strategies for minimizing the occurrence of and/or correcting these artifacts were not discussed in the publications. Finally, the question of whether signal from both halves of the transverse magnetization (the half that is refocused and the half that is not refocused by the standard unencoding gradient) should be acquired to improve overall signal-to-noise ratio is not settled. With the fast spin-echo sequence for data acquisition in meta-DENSE (6), gradients of reverse polarity are applied to alternately refocus both halves in successive echoes, and an improvement in signal-to-noise ratio associated with increased data averaging is claimed. However, it appears that there is no scheme with short echo train echo-planar MR imaging (1) whereby both halves can be acquired, nor is there clear indication that the efficiency of the signal-to-noise ratio of the sequence will be benefited. More insight regarding the optimal displacement-encoding gradient pulse moment, the optimal echo train length, the need for Nyquist ghost correction, and the true benefits of use of both halves of the stored transverse magnetization will likely be provided by additional experimentation.
One problem with complementary data acquisition is residual signal after subtraction. This residual signal may be due to errors in the flip angles of the radiofrequency pulses in the position-encoding module of the sequence. The acquisition requires that the flip angles be exactly 90°, since two such pulses must give exactly 180° to ensure that the complementary data set has inverted longitudinal magnetization equal in magnitude to that created by the normal 90°, -90° flip angle combination. It is possible that the residual T1 relaxation signal in the image (eg, 8% of total signal energy at end systole) reported in reference 1 is due to the nominal +90°, +90° radiofrequency pulse combination not equaling exactly 180°. Regardless of the cause, the residual T1 relaxation signal can be suppressed by k-space "filtering," which in reference 1 is simply setting to 0 all values in a circular region centered on the T1 relaxation signal in k space. This ad hoc correction of unwanted signal improves image appearance but eliminates valid signal from displacement-encoded magnetization and reduces the effective spatial resolution of the displacement-encoded image. Additional experiments are needed to determine the optimal distribution of weighting factors for this k-space filter.
There appear to be a few lingering fundamental limitations with the proposed DENSE cardiac imaging techniques. For example, loss of image intensity and displacement measurement errors occur because the spatial distribution of B0 inhomogeneity changes with contraction of the heart; hence, magnetization is not completely refocused at the echo with respect to the effect of B0 inhomogeneity. Image intensity loss and displacement measurement errors are most severe at air-epicardial interfaces, where the underlying B0 field changes most rapidly over a short distance. Also, beat-to-beat variability of heart motion and electrocardiographic triggering in each RR interval can contribute to these effects. Quantitative methods to reduce the loss of accuracy and precision of displacement and myocardial strain measurement due to these influences (eg, use of B0 field maps to correct the displacement measurements) have not yet been proposed, but these influences have been taken into account qualitatively in the design of pulse sequences and in the empirical selection of pulse sequence parameters. Sequence design and parameter choices have served to minimize the signal loss and systematic errors in the measurements as much as possible.
The benefits of multiphase myocardial strain measurements with high temporal and spatial resolution as provided by the pulse sequence in reference 1 now require assessment in the clinical setting. The clinical use of tagged and nontagged cine-based MR imaging for cardiac diagnosis has repeatedly demonstrated that time series of MR images across the cardiac cycle provide significant diagnostic advantages whether they are used for visualization of the overall contraction of the heart or for analysis of myocardial deformation and strain. It remains to be seen whether other signal-to-noise ratio and image artifact problems, which could reduce the reliability and reproducibility of imaging results, will impede the clinical use and diagnostic impact of this latest DENSE pulse sequence.
Are the technical innovations presented in reference 1 the last that are likely to appear for myocardial strain measurement? Probably not! In the literature, there are several points made regarding the struggle to increase signal-to-noise ratio by the process of limiting the amount of phase dispersion signal loss, by using optimal flip angles to increase signal at each echo, and by acquiring signal from both halves of the stored transverse magnetization. The meta-DENSE sequence implemented a fast spin-echo acquisition with 90° excitation pulses, and the level of signal-to-noise ratio was better than that with previous short echo train gradient-recalled-echo sequences with small flip angle pulses. However, as stated in reference 1, fast spin-echo acquisition appears to be incompatible with multiphase acquisition of displacement-encoded images because after the 90° pulse, new longitudinal magnetization with new displacement encoding is not available to contribute to the signal at the beginning of each cardiac phase. Kim et al (1) recognize that ultrashort repetition time sequences—such as TrueFISP, Fiesta, or BalancedFFE—which fully refocus transverse magnetization at the end of each repetition time interval—would provide high signal levels as well as allow new longitudinal magnetization with new displacement encoding to contribute to the signal in each successive repetition time interval. Also, these ultrashort repetition time sequences may allow acquisition of both halves of the stored transverse magnetization. So, expect to see additional improvements in myocardial strain measurement by displacement encoding with these sequences. Regarding future advances of postprocessing algorithms, harmonic phase analysis (11) may provide a means to eliminate residual image components that arise from the unrefocused part of the stored transverse magnetization. This use of harmonic phase analysis with DENSE MR imaging may provide higher image quality and allow smaller gradient moments for displacement encoding, resulting in reduced signal loss from phase dispersion.
FOOTNOTES
See also the article by Kim et al in this issue.
Abbreviation: DENSE = displacement encoding with stimulated echoes
REFERENCES
Related Article
This article has been cited by other articles:
|
|
 |
|
  H. Wen, K. A. Marsolo, E. E. Bennett, K. S. Kutten, R. P. Lewis, D. B. Lipps, N. D. Epstein, J. F. Plehn, and P. Croisille Adaptive Postprocessing Techniques for Myocardial Tissue Tracking with Displacement-encoded MR Imaging Radiology, January 1, 2008; 246(1): 229 - 240. [Abstract] [Full Text] [PDF]
|
|
| ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
| HOME | HELP | FEEDBACK | SUBSCRIPTIONS | ARCHIVE | SEARCH | TABLE OF CONTENTS |
| RADIOLOGY | RADIOGRAPHICS | RSNA JOURNALS ONLINE |
