MAGNETIC RESONANCE IN MEDICINE7,

79-87 (1988)

Rapid Line Scan NMR Angiography J. FFUHM,K. D. MERBOLDT,W. HANICKE,M. L. GYNGELL, AND H. BRUHN Mar-Planck-lnstitutf i r biophysikalische Chemie, Postfach 2841, 0-3400 Gottingen, Federal Republic of Germany Received October 5, 1987; revised December 28, 1987 This paper describes a new technique for NMR angiography based on rapid line scan projection imaging and presaturation of stationary spins. The resulting line Scan angiograms are free from both misregistration artifacts common to subtraction methods and motion artifacts encountered in Fourier imaging. Moreover, line scan angiograms may be recorded within seconds and offer arbitrary fields of view as well as gradient zooming without problems due to aliasing. Three-dimensional information is obtained by rotation ofthe read gradient axis generating multiple view angles. Experimental line scan angiograms of the forearms of healthy volunteers have been recorded using a Bruker 2.35-T 40-cm magnet. o 1988 Academic Press. Inc.

INTRODUCTION

Primary aims of magnetic resonance imaging of vascular structuresare the qualitative identification of flowing material, i.e., angiography, the discrimination of flow directions, and the quantitative assessment of flow velocities and related characteristics. Accordingly, a variety of means which exploit the specific properties of flowing spins has been developed. Quantitative methods generally rely on the special phase behavior of flowing spins in the presence of magnetic field gradients (I),and, in principle, allow the determination of all flow parameters. However, angiographic applications of these techniques are based on dual acquisitions and data subtraction and as a result are sensitive to all types of temporal instabilities including patient motion (2). Three-dimensional flowencoding techniques (3)such as Fourier angiography ( 4 ) tend to require long measuring times of at least several minutes. On the other hand, cross-sectional FLASH images recorded perpendicular to a major flow direction show strong signal enhancement of flowing material (5). This is due to a reflow of unsaturated spins into the selected slice resulting in high signal intensities for flowing material as compared to stationary spins being subjected to all rf pulses. This paper describes angiography experiments which take advantage of this desirable feature by means of projective line scan acquisitions (6).The basic idea is to eliminate contributions from stationary spins prior to the direct detection of flow signals in a single slice-selective acquisition without a phase-encoding gradient. A simple one-dimensional Fourier transformation then yields a projection of flow signals within the selected slice (or line) onto the axis of the read gradient. Extended projection angiograms may be recorded by multiple repetitions with different line positions. 79

0740-3 194/88 $3.00 Copyright 0 1988 by Academic Press, Inc. All rights of reproduction in any form mewed.

80

FRAHM ET AL.

METHODS

a. Line Scan Projection Imaging Although of limited efficacy for cross-sectional NMR imaging, line scan techniques provide a simple and rapid tool for projection imaging exhibiting certain unique properties for angiographic studies (7). The basic line scan experiment comprises sliceselective excitation and data acquisition in the presence of a read gradient. The NMR signal may be a gradient-recalled echo, a spin echo, or a stimulated echo depending on the selected rf excitation sequence. One-dimensional Fourier transformation of the line scan echo results in a projection of intensities within the slice or line onto the axis of the read gradient. A complete projection image may be obtained by moving the position of the line perpendicular to the slice selection gradient and arranging the projection intensities into a two-dimensional data array. Since the recording of a single line scan projection is entirely based on slice-selective excitation, the experiment may be repeated without waiting periods for magnetization recovery. Accordingly, measuring times of the order of seconds are easily accomplished depending on the number of lines desired. Typical characteristics of line scan projection images include the capability of using arbitrary fields of view given by the center to center distance of individual lines times the number of lines. Zooming of the images may be easily performed by increasing the strength of the read gradient without problems due to signal aliasing. Since no phase encoding of spatial information is employed, motion artifacts as seen in the phase-encoding dimension of Fourier images are avoided. Obviously, the individual slice-selective projections of a full line scan image are completely independent from each other, so that motion during the imaging time will not distort the overall image but result in frozen views line by line. In fact, line scan acquisitions represent true snapshot measurements with acquisition windows of the order of 2 to 10 ms and with echo times of less than 10 ms using gradient echoes. In combination with motion or flow-rephasing gradient waveforms, blurring in the frequency-encoding dimension even due to extremely rapid flow may be completely avoided, which seems to be a prerequisite for any application to cardiac angiography.

b. Line Scan Angiography For angiographic purposes stationary spins must be eliminated. There are at least two different approaches measuring either the inflow of previously prepared spins or the reflow after saturation. The inflow technique may be based on a stimulated-echo or spin-echo sequence. In the latter case, spins outside the desired slice are excited, e.g., by an initial 90" 1-3-1 pulse package. Flowing spins that move into the position of the projective line as determined by the selective 180" pulse become refocused, whereas stationary spins will experience only one of the two pulses and do not contribute to the spin echo. However, practical implementations of the technique may be complicated by (i) signal losses during lengthy inflow periods because of T2relaxation and the need of high-order motion rephasing, (ii) overlap of slice profiles, (iii) a dependence of the range of velocities on the bandwidth of the excitation pulse, (iv) phase cycling of the refocusing pulse to avoid contributions from spurious FID signals, and (v) waiting times needed for reflow of remote spins.

81

RAPID LINE SCAN NMR ANGIOGRAPHY

An even simpler and faster sequence for line scan angiography is depicted in Fig. 1. Its basic idea has already been proposed for velocity measurements in flowing fluids (8). A preceding slice-selective rf pulse saturates both flowing and stationary spins within a plane. After a certain reflow period short compared to T Ionly reflowing spins become excited by another slice-selective rf pulse at the same position, whereas stationary spins are still saturated. Signals from reflowing spins then are acquired in the form of a gradient-recalled echo. Of course, it is also possible to employ slice-selective spin-echo or stimulated-echo detection sequences to avoid complications from inhomogeneity or susceptibility problems encountered using gradient echoes (9). Recently, a closely related line scan angiographic sequence that attempts to suppress stationary spins by multiple repetitions of the line scan sequence was reported (10). However, in our experience the discrimination between flowing and stationary spins could be considerably improved by using a true presaturation pulse with the additional benefit of avoiding steady-state complications and reducing the measuring time per line. Flow directions may be identified by extending the regime of saturation to either side of the line. In general, signal contributions of spins reflowing from specific directions may be eliminated by inclusion of further flow suppression pulses (5). Velocity sensitivity may be achieved by changing either the time or the distance between saturating and exciting pulses. For example, velocities may be estimated from the flow signal intensity as a function of reflow time in a series of (interleaved) acquisitions. Further variants may involve ECG-synchronized data acquisitions leading to line scan movies and cardiac applications that benefit from the snapshot character of line scan measurements.

I L

m

-

-_

-__-__------J

TR

n lines

I

m

FIG. 1. Schematic rf pulse and gradient sequence for line scan angiography detecting the reflow of fresh spins after presaturation. Here a gradient echo signal has been employed. The number n of projective lines with spatially different positions determinesthe field of view. The reflow period may be adjusted as desired by changing the interval between saturation and excitation. The repetition time may be as short as the basic sequence of pulses and gradients because of its slice-selectivecharacter. Practical implementationsemploy motion/flow-rephasing gradient waveforms.

82

FRAHM ET AL.

RAPID LINE SCAN NMR ANGIOGRAPHY

83

FIG. 3. One-hundred-megahertz 'H NMR line scan angiograms of the forearm of a normal volunteer demonstrating the influence of different reflow periods: (a) 40 ms reflow, TR = 80 ms, and (b) 120 ms reflow, TR = 200 ms. Four excitations lead to measuring times of 16 and 40 s, respectively. Other parameters are as in Fig. 2.

c. Three-Dimensional Angiography

Spatial information about vascular structures in three dimensions is easily achieved using line scan angiograms at multiple view angles. Although it is, of course, possible to include a conventional phase-encoding gradient in the imaging sequence, the use FIG.2. One-hundred-megahertz 'H NMR line scan angiograms of the forearm of a normal volunteer. The angiograms have been recorded using the sequence shown in Fig. 1 using 50 lines corresponding to a FOV of about 7.5 cm in the flow direction. The reflow period and echo time were 80 and 20 ms, respectively. Using TR = 120 ms the measuring times were (a) 6 s for a single excitation, (b) 24 s for 4 excitations, and (c) 1 min for 10 excitations. With reference to the transverse 5-s FLASH image (d) the projection angiograms have been angulated by about 22.5" with respect to the horizontal axis of the image, i.e., from lower right to upper left.

84

FRAHM ET AL.

RAPID LINE SCAN NMR ANGIOGRAPHY

85

of rotated read gradients to create multiply angulated angiograms turns out to be advantageous because it does not sacrifice the advantages of the line scan technique as outlined above. A rapid rotation of multiview angiograms using appropriate display systems results in an impressive three-dimensional visualization of a vascular tree. However, in such cases two-dimensional filtered back projection will provide improved signal-to-noise ratios of projections taken from a reconstructed data set as compared to the individually acquired angiograms. RESULTS AND DISCUSSION

Experimental line scan angiograms of the forearms of normal volunteers have been recorded using a Bruker Medspec system with a 2.35-T 40-cm magnet. The angiograms in Figs. 2a-2c have been recorded using a repetition time of 120 ms, a reflow period of 80 ms, and an echo time of 20 ms including flow-rephasing gradients in the slice direction. For comparison the 5-s FLASH image shown in Fig. 2d depicts a transverse view of the same volunteer delineating major vessels in the human forearm. The projective angiograms have been obtained at an angle of about 22.5" with respect to the horizontal image axis. Their spatial resolution of 50 X 256 pixels has been limited to 50 lines per angiogram due to the actual system software. Slice-selective Gaussianshaped rf pulses with a duration of 6 ms lead to slice profiles without distortions as observed using other sequences with short repetition times ( 1 1 , 12).The line thickness was 1.5 mm using 7.5 mT m-l slice selection gradients. For excitation the flip angle was 90", while the flip angle of the saturation pulse was experimentally optimized, and, in general, turned out to be higher than 90". Of course, more sophisticated selective rf pulses that generate rectangular slice profiles may considerably improve both the suppression of stationary spins and the strength of the detected signal. Frequency advancement of the rf pulses in repeated applications was 300 Hz corresponding to a spatial advancement of the line by about 1.5 mm per repetition time. Thus, for TR = 120 ms, the speed of the line was of the order of 1.2 cm s-I, i.e., less than the velocities expected in human limbs ( 4 ) . Figure 2 clearly demonstrates that line scan angiograms may be obtained within seconds with sufficient signal-to-noise and spatial resolution. In particular, the image in Fig. 2a has been recorded in 6 s using 50 lines with a repetition time of TR = 120 ms. The loss of signal observed within the radial artery in every eighth line of the angiogram, i.e., every 0.96 s, directly reflects the heartbeat of the volunteer which in this case corresponded to a pulse rate of about 60 per minute. The signal pattern is probably due to the occurrence of accelerated flow in systole not compensated for in the present implementation of the sequence. Obviously, the effects from pulsatile flow during the cardiac cycle may be studied in more detail by means of dynamic applications using ECG-synchronized line scan angiograms. On the other hand, averaging

FIG. 4. One-hundred-megahertz 'H NMR line scan angiograms of the forearm of a normal volunteer demonstrating nine angulated views covering a full 180" rotation in 22.5" steps from (a) to (i). The reflow period was 80 ms. The measuring times of the individual angiograms were 24 s using TR = 120 ms and four excitations.

86

FRAHM ET AL.

of incoherent data will eliminate the beat pattern as shown in Fig. 2b using four excitations and a total measuring time of 24 s. A further improvement of the signalto-noise as shown in Fig. 2c using 10 excitations might not be necessary for most applications. Residual signals from stationary material may be ascribed to the insufficient suppression capability of the Gaussian pulses as well as to severe eddy currents in our current system generating slightly different slice (line) profiles for the two rf pulses involved. Another reason for residual background noise is contributions from tissues such as fat having rather short TI relaxation times. These signals may be effectively suppressed by incorporation of a chemical-shift-selective (CHESS) pulse applied to the methylene frequency range prior to excitation of the line scan echo. The influence of different reflow times on line scan angiograms is demonstrated in Fig. 3 using a reflow period of 40 ms and TR = 80 ms (a) and a reflow period of 120 ms and TR = 200 ms (b). The measuring times were 16 and 40 s, respectively, using four excitations. The appearance of a further vein with slow flowing blood is quite obvious in Fig. 3b where a threefold reflow period has been employed. In principle, the accessible range of flow velocities in line scan angiograms may be limited by different physical parameters. A lower value may be determined either by the ratio of line thickness and reflow period (here 1-4 cm s-') or by the speed of a unidirectional line advancement (here 0.8-2 cm sC1). However, the latter restriction may be circumvented by manipulating the order of the individual line recordings. For biological applications an upper limit of flow velocities is not observed because it is approximately given by the ratio of one-half the length of the probehead and the echo time (here 3 m s-I) which refers to the fact that the spins should not have left the coil between excitation and detection. Figure 4 shows a series of nine angulated line scan angiograms of the forearm of a normal volunteer obtained by changing the effective direction of the read gradient as in back projection imaging. The angiograms have been obtained using a reflow period of 80 ms and TR = 120 ms leading to measuring times of 24 s for four excitations. The total measuring time for all angiograms representing a full 180" rotation of the view angle was less than 4 min. Interleaved acquisitions of angulated line projections may improve the suppression of stationary spins within a selected line. Although rotation of the set of angiograms results in a three-dimensional impression, future applications will benefit from two-dimensional back projection. In general, image quality of NMR angiograms may also be improved using pattern recognition principles that exploit a priori knowledge about the simple data structure of vascular images. CONCLUSIONS

New methods for rapid NMR angiography using projective line scan imaging techniques have been described. The feasibility of gradient echo line scan angiography detecting the reflow of spins after slice-selectivepresaturation has been experimentally demonstrated on humans. Analogous sequences employing spin echoes or stimulated echoes are also possible. In comparison with phase-sensitive flow-imagingtechniques, line scan angiography bears several important advantages. It is rapid with a basic measuring time of the order of seconds. Implementation and application of the sequence is simple. Signals

RAPID LINE SCAN NMR ANGIOGRAPHY

87

from flowing spins are detected directly without the need of subtraction techniques. Furthermore, projective line scan techniques are robust with respect to both motion artifacts because of the use of independent snapshot acquisitions and eddy currents because of the recording of flow amplitudes. In addition, line scan angiography offers an arbitrary choice of the field of view as well as true gradient zooming capabilities without aliasing as in the phase-encoding dimension of Fourier images. These features allow a purpose-oriented optimization of the spatial resolution and measuring time. Angulated views are easily obtainable by rotation of the read gradient and lend themselves to back projection algorithms. Work is in progress for whole-body applications. ACKNOWLEDGMENT Financial support by the Bundesminister fur Forschung und Technologie ( B M R ) of the Federal Republic of Germany (Grant 01 VF 8606/6) is gratefully acknowledged. REFERENCES 1. P. R. MORAN,Magn. Reson. Imaging I, 197 (1982). 2. C. L. DUMOULIN AND H. R. HART,Radiology 161, 7 17 ( 1986). 3. J. HENNIG,H. FRIEDBURG, AND M. MUERI,Radiology P 161, 159 (1986). 4. K. D. MERBOLDT, J. FRAHM,AND W. HANICKE,“Society of Magnetic Resonance in Medicine, Sixth Annual Meeting, New York, August 17-2 1,1987,” Book of Abstracts, p. 414; submitted for publication. 5. J. FRAHM, K. D. MERBOLDT,W. HANICKE,AND A. HAASE,Mugn. Reson. Med. 4,372 (1987). 6. P. MANSFIELD AND A. A. MAUDSLEY, Phys. Med. Biol. 21,847 (1976). 7. D. G. NISHIMURA, A. MACOVSKI, AND J. M. PAULY,IEEE Truns. Med. Imaging MI 5, 140 (1986). 8. A. N. GARROWAY, J. Phys. D I, L159 (1974). 9. J. FRAHM, K. D. MERBOLDT,AND W. HANICKE,Mugn. Reson. Med. 6,474 (1988). 10. J. M. PAULY,D. G. NISHIMURA, AND A. MACOVSKI, “Society for Magnetic Resonance Imaging, Fifth Annual Meeting, San Antonio, February 28-March 4, 1987,” oral presentation. 11. I. R. YOUNGAND J. A. PAYNE,Mugn. Reson. Med. 5, 177 (1987). 12. W. HANICKE,K. D. MERBOLDT, AND J. FRAHM, J. Magn. Reson., in press.