MAGNETIC RESONANCE IN MEDICINE 14,194-201 ( 1990)
Time-of-Flight MR Angiography * DWIGHT G. NISHIMURA Departmen1 of Electrical Engineering, Stanfiwd University, Stanfordz Cal(fornia 94305 Received December 19, 1989 Time-of-flighteffects depend on the displacement of blood with respect to a region of excitation. When combined with static material suppression and projection imaging, time-of-flighteffects provide a flexible means of flow sensitizationfor magnetic resonance (MR)angiography. Bolus tracking, flow enhancement by spin replacement,and selective tagging are three classes of methods being pursued for MR angiography. o 1990 Academic Press. Inc.
INTRODUCTION
Time-of-flight (or wash-in/wash-out) effects have been used to study or measure flow in a small segment of a blood vessel ( 1) . Images of transaxial planes through the vessel reveal the extent of wash-in or wash-out in the lumen; alternatively, images of longitudinal planes containing the vessel show the extent of blood displacement. In magnetic resonance (MR) angiography-where the emphasis is not on flow measurement but on anatomic depiction-vessel tortuosity, vessel conspicuity, and field of view become important considerations. The combination of time-of-flight effects with projection imaging and static material suppression addresses these considerations and results in high-contrast anatomical images of blood vessels in a longitudinal format. This paper describes the current time-of-flight-based angiographic methods being explored, first reviewing the approaches for projection imaging and static material suppression. The specific approach taken for projection imaging and static material suppression usually plays an integral role in the means of flow sensitization and is often the distinguishing characteristicbetween the various methods. PROJECTION IMAGING
Projection imaging of a volume is important because of its effectiveness and efficiency in displaying complex and often tortuous vascular anatomy. Specific approaches include 2-DFT projection imaging (full or partial projection), line scanning, and fast 3-DFT imaging with subsequent computed projection. 2-DFT projection imaging, with gradient-echo or spin-echo acquisition, employs a nonselective excitation to examine the whole volume, or a mildly selective excitation either to
* Presented at SMRM Workshop on MR Imaging of Blood Flow, Philadelphia, PA, March 13 and 14, 1989. 0740-3194/90 $3.00 Copyright 0 1990 by Academic Press,Inc. All rights of reproduction in any form reserved.
194
TIME-OF-FLIGHT MR ANGIOGRAPHY
195
limit the depth of projection or to restrict the field of view. The second approach is line scanning where a projection image is constructed in a line by line fashion by sequencing through a set of contiguous thin slices. A 1-D projection line or a 2-D image of each slice may be obtained. In the latter case, a 3-D data set becomes available, and projections can be computed at an arbitraryview angle to alleviate problems with vessel overlap and lesion identification. Typically, the thin slices are oriented perpendicular to the principal vessel direction to gain appreciable blood displacement. This method possesses some immunity to motion as motion occurringbetween different slice acquisitions results in geometric distortion as in the focal-plane shutter effect. The third approach, ungated 3-DFT imaging with small flip-angle, short-TR acquisition, has become popular recently and enables computed projection at an arbitrary view angle. Often, to reduce the scan time, a mild selective excitation is applied to restrict the field of view along one of the phase-encoding axes. STATIC MATERIAL SUPPRESSION
Vessel conspicuity is greatly enhanced by suppressing static material signals which would otherwise dominate. In certain cases such as in bolus tracking, the excitation sequence itself can be flow-selective and signal from only flowing blood is produced. However, most methods must contend with the presence of static material signals in the measurements. Because of dynamic range considerations, it is often important to minimize the static material component in the measurements by reducing its steadystate amplitude through the excitation and timing parameters. In addition, a mild phase twist (linear or nonlinear) placed along the line of projection in a direct 2-D projection imaging sequence can also improve the dynamic range by creating phase dispersion that reduces the integrated signal across larger static structures but not across the smaller vessel lumen. Given residual static material signalsin the measurements, further suppression is possible via subtraction of two images with different blood signals as induced by the appropriate change in the pulse sequence. Recently, nonlinear postprocessing has become a popular suppression technique for methods acquiring a flow-enhanced 3-D data set. A specificexample of nonlinear postprocessing is taking the maximum value along the ray path during the computed projection process. This technique, which obviates subtraction, relies on the a priori knowledge that blood often represents the highest signal in the data set. FLOW SENSITIZATION BY TIME-OF-FLIGHT EFFECTS
As opposed to flow-dependent phase shifts which depend on gradient modulations, time-of-flight effects in MR imaging depend on the spatial selectivity, flip angle, and timing of radiofrequency (RF) excitations. A simplified sequence with which to describe the basic time-of-flight effects used for angiography is given in Fig. la. In general, two RF pulses separated by time Texcite the appropriate regions with flip angles O1 and 02. It is the displacement of blood relative to the regions of excitation that give rise to the time-of-flight effects. The three main classes of methods for time-of-flightbased angiography are bolus tracking, flow-related enhancement, and selective tagging.
196
DWIGHT G. NISHIMURA
FIG. 1. Comparison of time-of-flight methods: (a) Generic sequence-two RF pulses separated by time T excite the appropriate regions with flip angles 0, and 02. (b) Bolus tracking-the 180" excitation tracks the bolus of blood excited by the 90" pulse, resulting in an echo from only the bolus. (c) Flow-related enhancement-an enhanced blood signal due to spin replacement occurs as the region of interest is repetitively excited by readout pulses. Larger flip angles are used for smaller excited volumes. ( d ) Selective tagging-a 180" excitation tags upstream blood for a first image. A second image with the tagging pulse turned off allowsstatic material subtraction, leaving a differencesignal from the differentiallytagged blood.
Bolus trucking. In bolus tracking, signal generation depends on receiving two or more excitations. An example is presented by Fig. 1b which shows a bolus of blood excited by a 90" excitation and later ( T = TE/2) by a 180" excitation downstream. Static material suppression is inherent with this method as only blood, receiving both excitations, produces a spin echo (2, 3). Alternatively, stimulated-echo generation can be used using a trio of 90" excitations (4). Whereas the interpulse interval in the spin-echo case is limited by blood's T2,the interval is limited by TIin the stimulatedecho case. Flow-reluted enhancement. As illustrated by Fig. 1c, fully magnetized blood flowing into the imaged region creates an enhanced flow signal relative to the steady-state signal of surrounding static tissue. In this case, the interpulse interval T corresponds to the repetition time TR of the readout excitations. One acquisition mode exploiting this effect is sequential thin-slice imaging, a linescan approach that uses repetitive (or a burst of) larger flip-angle excitations to create a near-saturated steady-state signal from static material and a high signal from inflowing blood ( 5 - 1 0 ) . This method is sensitive to slow flow owing to the thinness of the slice being refreshed by blood; the minimum-velocity sensitivity is approximately Az/TR, where Az is the slice thickness. A 2-D projection image can be formed either
TIME-OF-FLIGHT MR ANGIOGRAPHY
197
FIG.2. Line-scan carotid angiogram of a normal volunteer: 3-D acquisition consisting of a set of 2-D images of thin ( 1.5 mm) slices is followed by maximum value ray tracing to produce this projectionimage. Scan time was 6.5 min to acquire 55 slices (image provided courtesy of Dr. P. J. Keller).
directly, applying a readout gradient after each selective excitation, or indirectly, acquiring a 3-D data set by assembling 2-D images of each thin slice and then computing projections at any desired view angle. Keller et al. ( 10) have implemented such a 3-D imaging sequence, acquiring a set of 1.5-mm-thick 2-D images using a short TR and relatively large (>90") flip angles. Maximum value ray tracing during the computed projection step suppresses stationary material in the resultant projection image. Figure 2 is an example of a carotid angiogram of a normal volunteer obtained with this method in about 6.5 min (55 slices). Another acquisition mode, 3-DFT with short TR, also results in a 3-D data set with a large signal from inflowing blood and a relatively small signal from stationary material ( 11,12).Whereas the line-scan method uses larger flip angles because blood refreshes the thin slice within a single TR, 3-DFT imaging of a volume requires smaller flip angles to avoid saturation as blood experiences more than a single excitation as it passes through. Again nonlinear (maximum value) ray tracing as described earlier suppresses static material and can produce a projection image at any view
198
DWIGHT G. NISHIMURA
FIG.3 . Carotid angiogram using the inversion-taggingmethod Subject with 80%stenosis in the internal branch (a) X-ray DSA and (b) correspondingMR angiogram-4 min scan time for this 5-cm (vertical) by 12-cm (horizontal) field-of-viewimage.
angle. For carotid imaging, the typical parameters of the 3-DFT sequence are TR = 80 ms, 6 = 40",and 16 (32) X 256 X 256 matrix, for a scan time of 5.5-1 1 min. Examples and analyses of results produced by this method can be found in articles by Masaryk et al., Laub, and Haacke et al. in this issue. Selective tugging. In this method, blood upstream to the region of interest gets tagged by an RF excitation (Fig. Id). Two images, one with the tagging on and the other with the tagging off, are subtracted, leaving a difference signal from only the washed-in blood. Flow-related enhancement from fresh unexcited spins occurs for one of the two images; a second image set based on tagged blood allows subtraction of static material. The various implementations use either a saturation-tagging excitation of upstream blood (13, 1 4 ) or an inversion-tagging excitation via adiabatic fast passage
TIME-OF-FLIGHT MR ANGIOGRAPHY
199
FIG.3-Continued
( 15, 16) or a selective 180” pulse ( 1 7). For saturation tagging, excitations can be
applied repeatedly during the acquisition of one of the two images to create a magnitude difference by “darkening” the blood signal; in this case, however, the two images must be acquired sequentially. In the case of inversion tagging, which typically involves complex subtraction of two interleaved data sets, a long transit time, T = TI, limited by the TI of blood, is used to allow tagged blood to flow into the imaged region. If gated, application of the tagging pulse prior to systole and the readout pulse at diastole maximizes wash-in of blood and minimizes flow during readout. At Stanford, we have been developing an inversion-taggingmethod using a selective 180” pulse and typical values of TI of 350-550 ms. The method is designed to produce high resolution (0.5 mm) angiograms with a smaller field of view (5-7 cm), as limited by scan time and extent of inflow. A comparison of a conventional X-ray angiogram with an MR carotid angiogram acquired in about 4 min is given in Fig. 3. DISCUSSION
Optimization of time-of-flight methods involves mainly the RF excitation parameters: slice profiles, flip angles, and timing. Compared to using flow-dependent phase
200
DWIGHT G. NISHIMURA
shifts, some of the advantages of using time-of-flight effects for angiography are ( 1 ) better immunity to gradient eddy current problems, (2) flexibility in the gradient waveforms for flow compensation, and ( 3 ) inherent sensitivity to all flow directions given sufficient in/outflow. Also because flow is not required during readout, the chance of signal dropout may be reduced by timing the signal acquisition to a quiescent period. Its principal disadvantage relates to problems with insufficient in/ outflow which can lead to a reduced signal or inadequate vessel filling. Recent clinical results with the spin replacement and selective tagging methods have clearly demonstrated their potential clinical utility. Among these methods, evaluation and comparison of their relative strengths and weaknesses from a technical perspective at this stage are complicated and involve many considerations. In addition to non-flow-related concerns such as SNR, spatial resolution, scan time, and field of view, specific flow-imaging considerations include dynamic range, motion (vessel and tissue), flow rate sensitivity, vessel overlap, flow-induced dephasing, and vessel orientation. With respect to some of these considerations, trade-offs exist between 3-D and 2D projection imaging, between 3-DlT imaging and line scanning, between gating and not gating, and between subtraction and nonsubtraction. In general, 3-D imaging offers flexibility in displaying the vasculature from any view angle through computed projection, provides cross-sectional information if desired, and reduces signal dephasing owing to the smaller voxel size. Direct 2-D projection imaging yields fewer view angles but requires less measurements, enabling shorter imaging times (or better spatial resolution) and use of cardiac gating if desired. Line-scan acquisition appears to be better suited for slower flow compared to a 3-DFT acquisition which may saturate slower flowing blood. However, the spatial resolution in line scanning is limited by the slice thickness of the selective excitation while the resolution in 3-DFT imaging can be isotropic. Cardiac gating suffers from problems of implementation and variable heart rate but is helpful in regions of pulsatile flow and where vessel motion is significant. For ungated sequences, the effective spatial resolution will be limited by vessel motion. Use of subtraction produces images that are readily interpretable (reflecting the conventional line integral) and potentially free of surrounding structures, but incurs a doubling of the minimum scan time and potential problems with incomplete subtraction due to motion or a variable heart rate (if gated). The nonlinear processing methods such as maximum value ray tracing have demonstrated effectiveness in suppressing stationary material and do not require a second data set; however, these methods are less predictable and can produce artifacts given an eccentric lesion or large signals from nonblood structures. Overall, because of these tradeoffs, all of the methods possess limitations, and it may be appropriate to view the different methods as complementary. ACKNOWLEDGMENTS The author gratefully acknowledges the contributions and support of Dr. Albert Macovski and colleagues in the Magnetic Resonance Systems Research Lab at Stanford University. The author also thanks Dr. Paul Keller and his colleaguesat the Barrow Neurological Institute in Phoenix for providing the linescan angiogram.
TIME-OF-FLIGHT MR ANGIOGRAPHY
20 1
REFERENCES I. L. AXECL, Amer. J. Roentgenol. 143, 1157 ( 1984). 2. A. MACOVSKI, IEEE Trans. Med. Imaging 1,42 ( 1982). 3. D. NORRIS, in “Proceedings, Fifth SMRM-WIP,” p. 19, August 1986. 4. T. K. F. Foo, W. H. PERMAN,AND J. T. CUSMA,in “Proceedings, Sixth SMRM,” p. 424, August 1987. 5 . J. M. PAULY,D. G. NISHIMURA, AND A. MACOVSKI, in “SMRI, Program ofthe 5th Annual Meeting,” February 1987. 6. J. FRAHM,K. D. MERBOLDT, W. HANICKE,M. L. GYNGELL, AND H. BRUHN,Magn. Reson. Med. 7, 79 (1988). 7. G. T. GULLBERG, F. W. WEHRLI, A. SHIMAKAWA, AND M. A. SIMONS, Radiology 165,241 ( 1987). 8. J. HENNIG, M. MUERI,H. FRIEDBURG, AND P. BRUNNER, J. Catal. 11( 5 ) , 872 ( 1987). 9. J. P. GROEN,R. G. DE GRAAF,AND P. VAN DIJK, in “Proceedings, Seventh SMRM,” p. 906, August 1988. 10. P. J. KELLER,B. P. DRAYER, E. K. FRAM,C. L. DUMOULIN, AND S. P. SOUZA,Mag. Reson. Imaging 7(I), 186(1989). I I . G. LAUB,E. MULLER,W. LOEFFLER,B. KIEFER, AND P. RUGGIERI,in “Proceedings, Seventh SMRM,” p. 876, August 1988. 12. T. J. MASARYK, M. T. MODIC, J. S. ROSS, W. SELMAN, S. HARIK,P. RUGGIERI, G. LAUB,AND E. M. HAACKE,in “Proceedings, Seventh SMRM,” pp. 177-178, August 1988. 13. S. P. SOUZA,C. L. DUMOULIN, H. E. CLINE,AND W. WAGLE,in “Proceedings, Seventh SMRM,” p. 895, August 1988. 14. M. H. CHO,Y. H. KIM,J. B. RA, AND 2. H. CHO,in “Proceedings, Seventh SMRM,” p. 919, August 1988. IS. W. T. DIXON,L. N. Du, D. D. FAUL,M. GADO,AND S. ROSSNICK, Magn. Reson. Med. 3,454 ( 1986). 16. H. K. LEE,0. NALCIOGLU, AND P. R. MORAN,in “Proceedings, Seventh SMRM,” p. 722, August 1988. 17. D. G. NISHIMURA, A. MACOVSKI,J. M. PAULY,AND S. M. CONOLLY,Magn. Reson. Med. 4, 193 (1987).
Time-of-Flight MR Angiography * DWIGHT G. NISHIMURA Departmen1 of Electrical Engineering, Stanfiwd University, Stanfordz Cal(fornia 94305 Received December 19, 1989 Time-of-flighteffects depend on the displacement of blood with respect to a region of excitation. When combined with static material suppression and projection imaging, time-of-flighteffects provide a flexible means of flow sensitizationfor magnetic resonance (MR)angiography. Bolus tracking, flow enhancement by spin replacement,and selective tagging are three classes of methods being pursued for MR angiography. o 1990 Academic Press. Inc.
INTRODUCTION
Time-of-flight (or wash-in/wash-out) effects have been used to study or measure flow in a small segment of a blood vessel ( 1) . Images of transaxial planes through the vessel reveal the extent of wash-in or wash-out in the lumen; alternatively, images of longitudinal planes containing the vessel show the extent of blood displacement. In magnetic resonance (MR) angiography-where the emphasis is not on flow measurement but on anatomic depiction-vessel tortuosity, vessel conspicuity, and field of view become important considerations. The combination of time-of-flight effects with projection imaging and static material suppression addresses these considerations and results in high-contrast anatomical images of blood vessels in a longitudinal format. This paper describes the current time-of-flight-based angiographic methods being explored, first reviewing the approaches for projection imaging and static material suppression. The specific approach taken for projection imaging and static material suppression usually plays an integral role in the means of flow sensitization and is often the distinguishing characteristicbetween the various methods. PROJECTION IMAGING
Projection imaging of a volume is important because of its effectiveness and efficiency in displaying complex and often tortuous vascular anatomy. Specific approaches include 2-DFT projection imaging (full or partial projection), line scanning, and fast 3-DFT imaging with subsequent computed projection. 2-DFT projection imaging, with gradient-echo or spin-echo acquisition, employs a nonselective excitation to examine the whole volume, or a mildly selective excitation either to
* Presented at SMRM Workshop on MR Imaging of Blood Flow, Philadelphia, PA, March 13 and 14, 1989. 0740-3194/90 $3.00 Copyright 0 1990 by Academic Press,Inc. All rights of reproduction in any form reserved.
194
TIME-OF-FLIGHT MR ANGIOGRAPHY
195
limit the depth of projection or to restrict the field of view. The second approach is line scanning where a projection image is constructed in a line by line fashion by sequencing through a set of contiguous thin slices. A 1-D projection line or a 2-D image of each slice may be obtained. In the latter case, a 3-D data set becomes available, and projections can be computed at an arbitraryview angle to alleviate problems with vessel overlap and lesion identification. Typically, the thin slices are oriented perpendicular to the principal vessel direction to gain appreciable blood displacement. This method possesses some immunity to motion as motion occurringbetween different slice acquisitions results in geometric distortion as in the focal-plane shutter effect. The third approach, ungated 3-DFT imaging with small flip-angle, short-TR acquisition, has become popular recently and enables computed projection at an arbitrary view angle. Often, to reduce the scan time, a mild selective excitation is applied to restrict the field of view along one of the phase-encoding axes. STATIC MATERIAL SUPPRESSION
Vessel conspicuity is greatly enhanced by suppressing static material signals which would otherwise dominate. In certain cases such as in bolus tracking, the excitation sequence itself can be flow-selective and signal from only flowing blood is produced. However, most methods must contend with the presence of static material signals in the measurements. Because of dynamic range considerations, it is often important to minimize the static material component in the measurements by reducing its steadystate amplitude through the excitation and timing parameters. In addition, a mild phase twist (linear or nonlinear) placed along the line of projection in a direct 2-D projection imaging sequence can also improve the dynamic range by creating phase dispersion that reduces the integrated signal across larger static structures but not across the smaller vessel lumen. Given residual static material signalsin the measurements, further suppression is possible via subtraction of two images with different blood signals as induced by the appropriate change in the pulse sequence. Recently, nonlinear postprocessing has become a popular suppression technique for methods acquiring a flow-enhanced 3-D data set. A specificexample of nonlinear postprocessing is taking the maximum value along the ray path during the computed projection process. This technique, which obviates subtraction, relies on the a priori knowledge that blood often represents the highest signal in the data set. FLOW SENSITIZATION BY TIME-OF-FLIGHT EFFECTS
As opposed to flow-dependent phase shifts which depend on gradient modulations, time-of-flight effects in MR imaging depend on the spatial selectivity, flip angle, and timing of radiofrequency (RF) excitations. A simplified sequence with which to describe the basic time-of-flight effects used for angiography is given in Fig. la. In general, two RF pulses separated by time Texcite the appropriate regions with flip angles O1 and 02. It is the displacement of blood relative to the regions of excitation that give rise to the time-of-flight effects. The three main classes of methods for time-of-flightbased angiography are bolus tracking, flow-related enhancement, and selective tagging.
196
DWIGHT G. NISHIMURA
FIG. 1. Comparison of time-of-flight methods: (a) Generic sequence-two RF pulses separated by time T excite the appropriate regions with flip angles 0, and 02. (b) Bolus tracking-the 180" excitation tracks the bolus of blood excited by the 90" pulse, resulting in an echo from only the bolus. (c) Flow-related enhancement-an enhanced blood signal due to spin replacement occurs as the region of interest is repetitively excited by readout pulses. Larger flip angles are used for smaller excited volumes. ( d ) Selective tagging-a 180" excitation tags upstream blood for a first image. A second image with the tagging pulse turned off allowsstatic material subtraction, leaving a differencesignal from the differentiallytagged blood.
Bolus trucking. In bolus tracking, signal generation depends on receiving two or more excitations. An example is presented by Fig. 1b which shows a bolus of blood excited by a 90" excitation and later ( T = TE/2) by a 180" excitation downstream. Static material suppression is inherent with this method as only blood, receiving both excitations, produces a spin echo (2, 3). Alternatively, stimulated-echo generation can be used using a trio of 90" excitations (4). Whereas the interpulse interval in the spin-echo case is limited by blood's T2,the interval is limited by TIin the stimulatedecho case. Flow-reluted enhancement. As illustrated by Fig. 1c, fully magnetized blood flowing into the imaged region creates an enhanced flow signal relative to the steady-state signal of surrounding static tissue. In this case, the interpulse interval T corresponds to the repetition time TR of the readout excitations. One acquisition mode exploiting this effect is sequential thin-slice imaging, a linescan approach that uses repetitive (or a burst of) larger flip-angle excitations to create a near-saturated steady-state signal from static material and a high signal from inflowing blood ( 5 - 1 0 ) . This method is sensitive to slow flow owing to the thinness of the slice being refreshed by blood; the minimum-velocity sensitivity is approximately Az/TR, where Az is the slice thickness. A 2-D projection image can be formed either
TIME-OF-FLIGHT MR ANGIOGRAPHY
197
FIG.2. Line-scan carotid angiogram of a normal volunteer: 3-D acquisition consisting of a set of 2-D images of thin ( 1.5 mm) slices is followed by maximum value ray tracing to produce this projectionimage. Scan time was 6.5 min to acquire 55 slices (image provided courtesy of Dr. P. J. Keller).
directly, applying a readout gradient after each selective excitation, or indirectly, acquiring a 3-D data set by assembling 2-D images of each thin slice and then computing projections at any desired view angle. Keller et al. ( 10) have implemented such a 3-D imaging sequence, acquiring a set of 1.5-mm-thick 2-D images using a short TR and relatively large (>90") flip angles. Maximum value ray tracing during the computed projection step suppresses stationary material in the resultant projection image. Figure 2 is an example of a carotid angiogram of a normal volunteer obtained with this method in about 6.5 min (55 slices). Another acquisition mode, 3-DFT with short TR, also results in a 3-D data set with a large signal from inflowing blood and a relatively small signal from stationary material ( 11,12).Whereas the line-scan method uses larger flip angles because blood refreshes the thin slice within a single TR, 3-DFT imaging of a volume requires smaller flip angles to avoid saturation as blood experiences more than a single excitation as it passes through. Again nonlinear (maximum value) ray tracing as described earlier suppresses static material and can produce a projection image at any view
198
DWIGHT G. NISHIMURA
FIG.3 . Carotid angiogram using the inversion-taggingmethod Subject with 80%stenosis in the internal branch (a) X-ray DSA and (b) correspondingMR angiogram-4 min scan time for this 5-cm (vertical) by 12-cm (horizontal) field-of-viewimage.
angle. For carotid imaging, the typical parameters of the 3-DFT sequence are TR = 80 ms, 6 = 40",and 16 (32) X 256 X 256 matrix, for a scan time of 5.5-1 1 min. Examples and analyses of results produced by this method can be found in articles by Masaryk et al., Laub, and Haacke et al. in this issue. Selective tugging. In this method, blood upstream to the region of interest gets tagged by an RF excitation (Fig. Id). Two images, one with the tagging on and the other with the tagging off, are subtracted, leaving a difference signal from only the washed-in blood. Flow-related enhancement from fresh unexcited spins occurs for one of the two images; a second image set based on tagged blood allows subtraction of static material. The various implementations use either a saturation-tagging excitation of upstream blood (13, 1 4 ) or an inversion-tagging excitation via adiabatic fast passage
TIME-OF-FLIGHT MR ANGIOGRAPHY
199
FIG.3-Continued
( 15, 16) or a selective 180” pulse ( 1 7). For saturation tagging, excitations can be
applied repeatedly during the acquisition of one of the two images to create a magnitude difference by “darkening” the blood signal; in this case, however, the two images must be acquired sequentially. In the case of inversion tagging, which typically involves complex subtraction of two interleaved data sets, a long transit time, T = TI, limited by the TI of blood, is used to allow tagged blood to flow into the imaged region. If gated, application of the tagging pulse prior to systole and the readout pulse at diastole maximizes wash-in of blood and minimizes flow during readout. At Stanford, we have been developing an inversion-taggingmethod using a selective 180” pulse and typical values of TI of 350-550 ms. The method is designed to produce high resolution (0.5 mm) angiograms with a smaller field of view (5-7 cm), as limited by scan time and extent of inflow. A comparison of a conventional X-ray angiogram with an MR carotid angiogram acquired in about 4 min is given in Fig. 3. DISCUSSION
Optimization of time-of-flight methods involves mainly the RF excitation parameters: slice profiles, flip angles, and timing. Compared to using flow-dependent phase
200
DWIGHT G. NISHIMURA
shifts, some of the advantages of using time-of-flight effects for angiography are ( 1 ) better immunity to gradient eddy current problems, (2) flexibility in the gradient waveforms for flow compensation, and ( 3 ) inherent sensitivity to all flow directions given sufficient in/outflow. Also because flow is not required during readout, the chance of signal dropout may be reduced by timing the signal acquisition to a quiescent period. Its principal disadvantage relates to problems with insufficient in/ outflow which can lead to a reduced signal or inadequate vessel filling. Recent clinical results with the spin replacement and selective tagging methods have clearly demonstrated their potential clinical utility. Among these methods, evaluation and comparison of their relative strengths and weaknesses from a technical perspective at this stage are complicated and involve many considerations. In addition to non-flow-related concerns such as SNR, spatial resolution, scan time, and field of view, specific flow-imaging considerations include dynamic range, motion (vessel and tissue), flow rate sensitivity, vessel overlap, flow-induced dephasing, and vessel orientation. With respect to some of these considerations, trade-offs exist between 3-D and 2D projection imaging, between 3-DlT imaging and line scanning, between gating and not gating, and between subtraction and nonsubtraction. In general, 3-D imaging offers flexibility in displaying the vasculature from any view angle through computed projection, provides cross-sectional information if desired, and reduces signal dephasing owing to the smaller voxel size. Direct 2-D projection imaging yields fewer view angles but requires less measurements, enabling shorter imaging times (or better spatial resolution) and use of cardiac gating if desired. Line-scan acquisition appears to be better suited for slower flow compared to a 3-DFT acquisition which may saturate slower flowing blood. However, the spatial resolution in line scanning is limited by the slice thickness of the selective excitation while the resolution in 3-DFT imaging can be isotropic. Cardiac gating suffers from problems of implementation and variable heart rate but is helpful in regions of pulsatile flow and where vessel motion is significant. For ungated sequences, the effective spatial resolution will be limited by vessel motion. Use of subtraction produces images that are readily interpretable (reflecting the conventional line integral) and potentially free of surrounding structures, but incurs a doubling of the minimum scan time and potential problems with incomplete subtraction due to motion or a variable heart rate (if gated). The nonlinear processing methods such as maximum value ray tracing have demonstrated effectiveness in suppressing stationary material and do not require a second data set; however, these methods are less predictable and can produce artifacts given an eccentric lesion or large signals from nonblood structures. Overall, because of these tradeoffs, all of the methods possess limitations, and it may be appropriate to view the different methods as complementary. ACKNOWLEDGMENTS The author gratefully acknowledges the contributions and support of Dr. Albert Macovski and colleagues in the Magnetic Resonance Systems Research Lab at Stanford University. The author also thanks Dr. Paul Keller and his colleaguesat the Barrow Neurological Institute in Phoenix for providing the linescan angiogram.
TIME-OF-FLIGHT MR ANGIOGRAPHY
20 1
REFERENCES I. L. AXECL, Amer. J. Roentgenol. 143, 1157 ( 1984). 2. A. MACOVSKI, IEEE Trans. Med. Imaging 1,42 ( 1982). 3. D. NORRIS, in “Proceedings, Fifth SMRM-WIP,” p. 19, August 1986. 4. T. K. F. Foo, W. H. PERMAN,AND J. T. CUSMA,in “Proceedings, Sixth SMRM,” p. 424, August 1987. 5 . J. M. PAULY,D. G. NISHIMURA, AND A. MACOVSKI, in “SMRI, Program ofthe 5th Annual Meeting,” February 1987. 6. J. FRAHM,K. D. MERBOLDT, W. HANICKE,M. L. GYNGELL, AND H. BRUHN,Magn. Reson. Med. 7, 79 (1988). 7. G. T. GULLBERG, F. W. WEHRLI, A. SHIMAKAWA, AND M. A. SIMONS, Radiology 165,241 ( 1987). 8. J. HENNIG, M. MUERI,H. FRIEDBURG, AND P. BRUNNER, J. Catal. 11( 5 ) , 872 ( 1987). 9. J. P. GROEN,R. G. DE GRAAF,AND P. VAN DIJK, in “Proceedings, Seventh SMRM,” p. 906, August 1988. 10. P. J. KELLER,B. P. DRAYER, E. K. FRAM,C. L. DUMOULIN, AND S. P. SOUZA,Mag. Reson. Imaging 7(I), 186(1989). I I . G. LAUB,E. MULLER,W. LOEFFLER,B. KIEFER, AND P. RUGGIERI,in “Proceedings, Seventh SMRM,” p. 876, August 1988. 12. T. J. MASARYK, M. T. MODIC, J. S. ROSS, W. SELMAN, S. HARIK,P. RUGGIERI, G. LAUB,AND E. M. HAACKE,in “Proceedings, Seventh SMRM,” pp. 177-178, August 1988. 13. S. P. SOUZA,C. L. DUMOULIN, H. E. CLINE,AND W. WAGLE,in “Proceedings, Seventh SMRM,” p. 895, August 1988. 14. M. H. CHO,Y. H. KIM,J. B. RA, AND 2. H. CHO,in “Proceedings, Seventh SMRM,” p. 919, August 1988. IS. W. T. DIXON,L. N. Du, D. D. FAUL,M. GADO,AND S. ROSSNICK, Magn. Reson. Med. 3,454 ( 1986). 16. H. K. LEE,0. NALCIOGLU, AND P. R. MORAN,in “Proceedings, Seventh SMRM,” p. 722, August 1988. 17. D. G. NISHIMURA, A. MACOVSKI,J. M. PAULY,AND S. M. CONOLLY,Magn. Reson. Med. 4, 193 (1987).
