General Advantages of Parallel Imaging 16

General Advantages of Parallel Imaging

173

O. Dietrich, PhD

Department of Clinical Radiology, University Hospitals – Gross hadern, Ludwig Maximilian University of Munich, Marchioni nistr. 15, 81377 Munich, Germany

C O N T E N T S

16.1 Introduction 173

16.2 Spatial and Temporal Resolution 173 16.3 Image Quality 174

16.4 High-Field MRI 175 16.5 Conclusion 176 References 176

General Advantages of Parallel Imaging 16

Olaf Dietrich

16.2

Spatial and Temporal Resolution

When applying parallel imaging with a reduction (or acceleration) factor of, for example, R=3, only a third of the k-space lines of a conventional, non-accelerated MR acquisition are required to reconstruct an image data set of the same spatial resolution and geometry as without acceleration. The two most important applica- tions of this reduction of raw data are decreasing the total scan time (e.g. to a third of the original scan time for R=3) or increasing the spatial resolution.

The fi rst of these approaches can be benefi cial in acquisitions with comparatively long scan times (such as conventional T2-weighted high-resolution brain MRI or time-of-fl ight MR angiography of brain vessels that can take several minutes; cf. Chap. 26); however, reduc- ing the acquisition time is even more important in dynamic MRI in order to increase the temporal resolu- tion (e.g. for MRI of the cardiac function (cf. Chap. 35), or for multi-slice perfusion MRI; cf. Chaps. 33, 36, 37, and 39). Frequently, only a certain part of the time reduc- tion obtained by parallel imaging is used to increase the temporal resolution, while at the same time the number of acquired slices or the spatial resolution is increased.

In many important MR applications, the total scan time is restricted by physiological constraints such as the maximal breath-hold duration of 2025 s. These applications particularly profi t from parallel-imag- ing techniques that increase the spatial resolution without exceeding the maximum possible scan time.

Typical examples are abdominal MRI as described in Chaps. 2123 or breath-hold MR angiography (cf.

Chaps. 28 and 29). In both applications substantially higher spatial resolutions can be acquired with parallel imaging than with non-accelerated MRI. Often a com- promise between increased in-plane resolution and reduced scan time per slice is chosen in order to increase the total number of acquired slices within one breath- hold; thus, either the covered volume can be extended or a better resolution (i.e. a smaller slice thickness) can be obtained in through-plane direction.

16.1

Introduction

As discussed in the previous chapters, the concept of parallel imaging is based on the reduction of the number of phase-encoding steps without compro- mising the spatial resolution or the fi eld of view of an MR acquisition. Several, very different advantages arise from this idea that are summarized for various pulse-sequence types in Fig. 16.1. The most obvious of these advantages is the possibility of increasing the spatial resolution or of accelerating data acquisition.

A supplementary benefi cial consequence of a shorter scan time is a reduced susceptibility to motion arte- facts. In addition, sequences with long echo trains and particularly single-shot pulse sequences benefi t from improved image quality due to reduced geomet- ric distortions and blurring. These advantages of par- allel imaging are discussed in the following sections.

Subsequently, the specifi c synergistic effects of paral-

lel imaging at high fi eld strengths, i.e. at 3 Tesla and

above, will be elucidated at the end of this chapter.

174

16.3

Image Quality

An important positive side effect of shorter scan times is reduced susceptibility to motion artefacts.

There are different kinds of involuntary motion aris- ing, for example, from diaphragmatic contractions (cf. Chap. 29) or peristalsis (Froehlich et al. 2005) with typical time periods of 530 s; thus, MRI with relatively long scan times of 20 s or more will fre- quently be affected by motion artefacts, which can be reduced by applying parallel imaging to shorten the acquisition time. In order to maintain a certain signal-to-noise ratio (SNR) it can be advantageous to combine parallel imaging with data averaging of repeated acquisitions; thus, k-space data are more consistent before averaging because of the shorter acquisition duration (i.e. less motion during a single acquisition), and remaining motion artefacts will be attenuated due to the averaging process. More sophisticated data-correction schemes and advanced techniques for motion correction based on parallel imaging are described in Chap. 5.

In addition to reduced susceptibility to motion artefacts, particularly the image quality of single-shot pulse sequences as well as of multi-shot techniques

with long echo trains benefi ts from parallel imag- ing as described in detail in Chaps. 9 and 10. A major source for image artefacts and especially for image blurring is transversal relaxation during the echo- train readout. By shortening the echo train of single- shot or conventional turbo-spin-echo sequences or of echo-planar sequences, the signal decay during the echo-train readout is reduced and the point-spread function of the sequence, i.e. the image quality, is improved. This effect has been proven to be espe- cially useful for applications such as lung MRI with half-Fourier-acquisition single-shot turbo-spin-echo (HASTE) sequences (cf. Chap. 20) or MR cholangio- pancreaticography (cf. Chap. 22).

Even more dramatic improvements due to paral- lel acquisition techniques are found in echo-planar imaging, which generally suffers from (often severe) geometric-distortion artefacts due to B

inhomo- geneities and associated off-resonance effects. The extent of these artefacts depends the line distance in k-space (i.e. on the acquired fi eld of view prior to parallel-imaging reconstruction) and on the echo spacing; hence, geometric distortions can be reduced by acquiring only every other or every third line in k- space with parallel-imaging techniques (cf. Chap. 10).

Therefore, virtually all echo-planar-imaging applica- tions benefi t from parallel imaging, such as diffusion

Fig. 16.1. Advantages of parallel-imag- ing techniques for some frequently used pulse-sequence types:

echo-planar-imaging (EPI) sequences; single- shot turbo-spin-echo sequences (HASTE, RARE); turbo-spin-echo sequences (TSE); and spin-echo and gradi- ent-echo sequences (SE, GRE).

General Advantages of Parallel Imaging

175

and perfusion brain MRI (Chaps. 18, 33, and 34) or black-blood liver MRI (Chap. 21).

Finally, parallel imaging allows for reduced echo times in single-shot sequences with linear k-space sampling in phase-encoding direction (cf. Chap. 10);

thus, higher signal intensities, i.e. improved SNRs, can be obtained, for example, in diffusion-weighted or diffusion-tensor MRI or in HASTE lung MRI (Chaps. 20 and 38).

16.4

High-Field MRI

In principle, the application of parallel-imaging tech- niques is completely independent of the static fi eld strength, B

, of an MRI system, although it may be limited by signal-to-noise-ratio constraints at very low fi eld strengths (cf. Chap. 17); however, it turns out that parallel imaging provides several specifi c advantages when used at high fi eld strengths, e.g. at 3 T or above. These specifi c advantages are associated with both the main motivation to proceed to higher fi eld strengths, i.e. the increased signal-to-noise ratio, on the one hand, and with one of the main limitations

of MRI at high fi eld strengths, namely the substantial increase of required RF energy, on the other hand.

The SNR (cf. Chap. 1) of an MR acquisition increases approximately linearly with the static fi eld strength, i.e. it is about twice as high at 3 than at 1.5 T.

This effect can compensate an intrinsic disadvantage of parallel imaging, which is that the SNR decreases with the square root of the acceleration factor, R, if the geometry (voxel size and fi eld of view) is held constant because of the shortened acquisition dura- tion as illustrated in Fig. 16.2. Thus, the same SNR can be obtained at 3 T with an acceleration factor of R=4 as at 1.5 T without parallel-imaging acceleration, i.e. images of comparable quality can be acquired in only 25% of the original acquisition time. Compar- ing 1.5 and 7 T, substantially higher acceleration fac- tors become theoretically feasible. The SNR increase by a factor of about 4.7 corresponds to an accelera- tion factor of approximately R=22, i.e. a reduction of scan time to <5% of its original value. In practice, smaller acceleration factors will be chosen because of limitations with respect to coil geometry and the g-factor; however, high two-dimensional acceleration of three-dimensional MR imaging has already been demonstrated, e.g. with R=3×4=12 (cf. Chap. 44), and appears particularly feasible at very high fi eld strengths.

Fig. 16.2. Relative signal- to-noise ratios (SNR) for different parallel-imaging acceleration factors, R, and different static fi eld strengths, B₀, of 1.5 T, 3 T, and 7 T. The resulting reduced acquisition time (ratio of acquisition time of accelerated acquisition, t_R, and non-accelerated acquisition, t₀) is shown below the acceleration factor. The SNR increases proportional to B₀ and decreases proportional to

R

1 ; effects of regional noise amplifi cation due to the g-factor are neglected. For all three fi eld strengths, accelera- tion factors resulting in the same relative SNR of 100 % are marked.

176

As mentioned above, the required RF energy increases substantially at high fi eld strengths, since the energy of an RF pulse is proportional to the square of B

; hence, the specifi c absorption rate (SAR) of two identical pulse sequences is about four times higher at 3 than at 1.5 T. This effect turns out to be a serious limitation for many protocols at 3 T, particularly if high-SAR sequences, such as turbo-spin-echo (TSE), HASTE, or steady-state free-precession (e.g. True- FISP) sequences, are applied.

Fortunately, parallel imaging can help to reduce the average SAR of MR protocols under certain con- ditions. Consider, for example, a T2-weighted breath- hold multi-slice coronal TSE acquisition of the liver without parallel-imaging acceleration. While the SAR may be acceptable at 1.5 T, it is increased by a factor of 4 at 3 T and, thus, may exceed the allowed maximum. If parallel imaging with an acceleration factor of R=4 is applied, the number of refocusing RF pulse is decreased by 75%, and the fi eld-strength- related SAR increase is completely compensated, i.e.

the total RF energy of this sequence is reduced to the level of the non-accelerated sequence at 1.5 T. It is noteworthy, however, that the (long-term) averaged SAR of this protocol is only reduced, if no further RF pulses are applied during the total scan time of the non-accelerated sequence, i.e. the examination must be paused for the time difference between the non- accelerated and accelerated sequence. Similarly, the maximum possible fl ip angle of TrueFISP sequences may be higher with parallel-imaging acceleration than without (cf. Chap. 35). In all these cases, how- ever, only the long-term average of the SAR can be decreased; thus, parallel-imaging reduction of SAR is particularly helpful for relatively short meas- urements such as breath-hold examinations with alternating intervals of RF-pulse application (meas- urement during breath-hold) and breaks between measurements.

Apart from the discussed compensatory effects with respect to SNR and SAR, there are several other synergistic implications of parallel-imaging tech- niques and high magnetic fi eld strengths. Geometric distortions of echo-planar-imaging sequences, for example, increase at higher fi eld strengths (Wang et al. 2002; Xu et al. 2004; Ardekani and Sinha 2005).

This effect can ideally be compensated by parallel imaging as demonstrated in Chap. 10. Furthermore, it has been shown by Wiesinger et al. (2004) that lower g-factors at high linear acceleration factors can be obtained at high fi eld strengths of more than about 4.5 T (cf. Chaps. 14 and 44). Finally, parallel-

imaging techniques can be transferred to RF excita- tion; so-called parallel-excitation techniques can be used to compensate inhomogeneous signal distribu- tion frequently observed at high and ultra-high fi eld strengths. This approach is described in detail in Chap. 45.

16.5 Conclusion

Parallel imaging provides several essential advan- tages for a multitude of MRI applications. The pri- mary motivation for parallel imaging was faster imaging and MRI with higher spatial resolution;

however, it soon turned out that parallel-acquisition techniques offer several other advantages as well such as improved image quality of single-shot MRI due to decreased blurring and reduced geometric distor- tions. In addition, MRI at high and ultra-high fi eld strengths specifi cally benefi ts from parallel imaging, which ideally complements the increased SNR and compensates the growing specifi c absorption rate and increasing geometric distortions. Consequently, parallel imaging has been introduced in an ever- growing number of MRI protocols for an immense variety of clinical and research applications.

References

Ardekani S, Sinha U (2005) Geometric distortion correction of high-resolution 3 T diffusion tensor brain images. Magn Reson Med 54:1163–1171

Froehlich JM, Patak MA, Weymarn C von, Juli CF, Zollikofer CL, Wentz KU (2005) Small bowel motility assessment with magnetic resonance imaging. J Magn Reson Imaging 21:370–375

Wang J, Alsop DC, Li L, Listerud J, Gonzalez-At JB, Schnall MD, Detre JA (2002) Comparison of quantitative perfusion imaging using arterial spin labeling at 1.5 and 4.0 Tesla.

Magn Reson Med 48:242–254

Wiesinger F, Van de Moortele PF, Adriany G, De Zanche N, Ugurbil K, Pruessmann KP (2004) Parallel imaging per- formance as a function of fi eld strength: an experimental investigation using electrodynamic scaling. Magn Reson Med 52:953–964

Xu D, Henry RG, Mukherjee P, Carvajal L, Miller SP, Barkovich AJ, Vigneron DB (2004) Single-shot fast spin-echo diffu- sion tensor imaging of the brain and spine with head and phased array coils at 1.5 T and 3.0 T. Magn Reson Imaging 22:751–759