Technical Note

Rapid Dark-Blood Carotid Vessel-Wall Imaging WithRandom Bipolar Gradients in a Radial SSFPAcquisition

Hung-Yu Lin, MS,1,2 Chris A. Flask, PhD,1 Brian M. Dale, PhD,1,2 andJeffrey L. Duerk, PhD1,2*

Purpose: To investigate and evaluate a new rapid dark-blood vessel-wall imaging method using random bipolargradients with a radial steady-state free precession (SSFP)acquisition in carotid applications.

Materials and Methods: The carotid artery bifurcations offour asymptomatic volunteers (28–37 years old, meanage � 31 years) were included in this study. Dark-bloodcontrast was achieved through the use of random bipolargradients applied prior to the signal acquisition of eachradial projection in a balanced SSFP acquisition. The re-sulting phase variation for moving spins established signif-icant destructive interference in the low-frequency region ofk-space. This phase variation resulted in a net nulling ofthe signal from flowing spins, while the bipolar gradientshad a minimal effect on the static spins. The net effect wasthat the regular SSFP signal amplitude (SA) in stationarytissues was preserved while dark-blood contrast wasachieved for moving spins. In this implementation, appli-cation of the random bipolar gradient pulses along all threespatial directions nulled the signal from both in-plane andthrough-plane flow in phantom and in vivo studies.

Results: In vivo imaging trials confirmed that dark-bloodcontrast can be achieved with the radial random bipolarSSFP method, thereby substantially reversing the vessel-to-lumen contrast-to-noise ratio (CNR) of a conventionalrectilinear SSFP “bright-blood” acquisition from brightblood to dark blood with only a modest increase in TR (�4msec) to accommodate the additional bipolar gradients.

Conclusion: Overall, this sequence offers a simple andeffective dark-blood contrast mechanism for high-SNRSSFP acquisitions in vessel wall imaging within a shortacquisition time.

Key Words: flow suppression; dark-blood imaging; vessel-wall imaging; black blood; SSFPJ. Magn. Reson. Imaging 2007;25:1299–1304.© 2007 Wiley-Liss, Inc.

NUMEROUS MRI techniques have been developed toproduce images that can be used to clearly distinguishthe features of the vessel wall from those of the vessellumen. This is critical for clinical applications such asgrading of stenoses and atherosclerotic plaque charac-terization. Several bright-blood methods, such as time-of-flight (TOF) (1), phase contrast (PC) (2), and externalcontrast agent enhancement (3), are currently beingused clinically for MR angiography (MRA). In addition,steady-state free precession (SSFP) acquisitions canrapidly generate images with a relatively high lumensignal and have shown promise for vessel wall imaging,despite the presence of the high signal from luminalblood (4).

To alleviate the problem of high signal from bloodwithin the lumen, several dark-blood MR techniqueswere developed in the late 1980s. These techniquesoffer opposite but superior contrast in comparison tobright-blood images (i.e., the signal amplitude (SA) offlowing blood is much lower than the vessel wall),thereby enabling visualization and further character-ization of mural thrombus and plaque. The dark-bloodcontrast is typically based on the T1 difference betweenblood and surrounding tissues (i.e., double inversionrecovery (DIR) (5)) or inflow presaturation (i.e., spatialsaturation (6)). However, each of these methods hasunique limitations that prevent their use in some rou-tine cardiovascular MRI applications. For example, DIRdark-blood preparation substantially increases theoverall acquisition time because of the long T1 of blood(�1200 msec). DIR dark-blood preparation also im-poses significant constraints on multislice acquisitions,although several variations have recently been devel-oped that better utilize the unavoidable delay time

1Department of Radiology, University Hospitals of Cleveland and CaseWestern Reserve University, Cleveland, Ohio, USA.2Department of Biomedical Engineering, Case Western Reserve Univer-sity, Cleveland, Ohio, USA.Hung-Yu Lin is now at the Department of Internal Medicine, Ohio StateUniversity, Columbus, Ohio, USA.Brian M. Dale is now at Siemens Medical Solutions, Inc., Cary, NorthCarolina, USA.*Address reprint requests to: J.L.D., Department of Radiology/MRI,B100 Bolwell Bldg., University Hospitals of Cleveland, 11100 EuclidAve., Cleveland, OH 44106. E-mail: duerk@uhrad.comReceived February 28, 2006; Accepted September 19, 2006.DOI 10.1002/jmri.20821Published online in Wiley InterScience (www.interscience.wiley.com).

JOURNAL OF MAGNETIC RESONANCE IMAGING 25:1299–1304 (2007)

© 2007 Wiley-Liss, Inc. 1299

(�600 msec) necessary for dark-blood magnetizationpreparation (7,8). Spatial slab-selective RF pulses canbe applied to saturate the magnetization of incomingspins prior to entering the imaging slice. However, spa-tial saturation methods are ineffective for suppressingin-plane flow, can often lead to incomplete flow sup-pression at vessel bifurcations, and can cause a signif-icant increase in the specific absorption rate (SAR),particularly with high-field systems.

In this study we developed a radial SSFP sequencewith randomized bipolar gradients that are appliedalong each imaging axis prior to MR data acquisition.The k-space oversampling inherent in some non-Carte-sian trajectories (e.g., radial) has been shown to resultin signal suppression caused by destructive interfer-ence in the low-frequency regions of k-space for off-resonance spins associated with certain compounds,such as fat (9). In this work we propose a similar con-cept that uses the velocity sensitivity of bipolar gradi-ents to introduce phase variation over each repetitiontime (TR) for moving spins in a radial SSFP acquisition.As a result, the signal from moving spins is expected todestructively interfere while the static tissue maintainsthe typical bright SSFP signal acquisition if there is

sufficient phase variation due to flow during the bipolargradients.

MATERIALS AND METHODS

Pulse Sequence Development

The true fast imaging with steady-state precession(True-FISP) sequence acquires relatively high SSFP SAby fully refocusing the gradients along all three imagingaxes over each short TR (10). However, this sequencehas strong bright-blood contrast due to the inflow ef-fect, in which unsaturated spins have experienced noprior nutation, dephasing, or relaxation effects prior toentering the imaging slice, and therefore contributehigh SA (11). To achieve the goal of dark-blood imagingand a high contrast-to-noise ratio (CNR) between thevessel wall and the lumen, random amplitude bipolargradients are applied prior to the radial SSFP data ac-quisition in each TR (Fig. 1). The bipolar gradient (Fig.2a) introduces an additional velocity-dependent phaseaccumulation (��) for moving spins (Fig. 2b), which issimilar to that typically used for PC angiography (12).The velocity-induced phase shift, ��, created by such abipolar gradient pulse can be expressed as

Figure 1. Pulse sequence diagram for radial SSFP with random bipolar gradients. During each TR the random amplitude ofbipolar gradients (shaded regions) is applied to balance the zeroth gradient moment and randomize the first gradient moment.Combined with the radial acquisition, this causes destructive interference in the low-frequency region of k-space for movingspins. This results in the suppression of flowing spins while the high SSFP signal is maintained for stationary spins.

Figure 2. Velocity-dependentphase accumulation (�v) froma bipolar gradient. a: The bipo-lar gradient pulse is a combi-nation of two gradient lobeswith equal and opposite zerothgradient moment but oppositepolarities. b: The bipolar gradi-ent introduces a difference inthe accumulated phase be-tween moving and static spins.

1300 Lin et al.

�� � �m1� (1)

where � is the gyromagnetic ratio of the moving spin, m1

is the first gradient moment of the bipolar gradient, andv is the component of the velocity vector in the directionof the gradient.

To achieve the desired flow suppression (dark-bloodcontrast), m1 is varied by randomly changing the mag-nitude of the bipolar gradient lobes. For moving spinsthis leads to a varying phase accumulation (i.e., ��1, ��2

. . . ��N) for each projection in the radial k-space trajec-tory. Due to the inherent oversampling of the centralk-space region of radial acquisitions, the view-depen-dent phase variation results in 1) destructive interfer-ence of signal from moving spins without disturbing thesignal from stationary spins, and 2) incomplete refocus-ing over each TR, which is effectively “gradient spoiling”for moving spins (13). The net effect of the pulse se-quence shown in Fig. 1 is that the stationary spins willfully refocus over each TR, while moving spins will ex-perience a form of velocity spoiling and therefore gen-erate an incoherent SSFP signal. In this study the ran-dom bipolar gradients were applied over all threegradient axes to suppress both in-plane and through-plane flow. In addition, we optimized flow suppressionon a Siemens Sonata 1.5 T scanner by utilizing thehardware’s entire gradient amplitude range of 40 mT/mto –40 mT/m, which resulted in the greatest possiblevariation in the velocity-dependent phase. All radialimages were reconstructed by means of a table-basedconvolution gridding algorithm that incorporated pre-viously measured radial trajectories (14,15).

Numerical Simulations

All simulations were performed with blood-like relax-ation time constants (T1 � 1200 msec, T2 � 223 msec).Analytic expressions for the resulting rotation matricesand magnetization distributions were generated usingMathematica (Wolfram Research, Inc., Champaign, IL,USA) to solve the Bloch equations for a radial SSFPacquisition with specified bipolar gradients and a givenamount of flow. The numerical simulations considered128 repetitions, each with a different random ampli-

tude of the bipolar gradients, and calculated the result-ing magnitude and phase for on-resonance spins (slicethickness � 5 mm, number of radial projection lines �128, flip angle � 65°, TR � 12 msec, TE � 7 msec, andbipolar gradients amplitude variation Gb � 40 mT/m to–40 mT/m). The SA was simulated for blood velocitiesfrom zero to 1.5 m/second. The simulation was re-peated 500 times with different series of random bipolargradient amplitudes in order to determine the averageeffect of the signal suppression and its variability(Fig. 3).

Phantom Preparation

Phantom experiments were performed using a constantvelocity flow phantom with an additional stationary wa-ter component to simulate the signal properties of flow-ing blood and static tissues. The flow phantom wasconstructed from a flexible plastic tube with a 5.5-mminternal diameter and 1.5-mm wall thickness. It wasfilled with doped fluid to obtain blood-like relaxationproperties. The stationary saline phantom was a 1.0-kguniform cylinder (diameter � 12 cm) doped with 1.25 gNiSO4�6H2O and 5 g NaCl. The phantom flow studieswere performed under two experimental conditions:primarily through-plane flow or primarily in-plane flow.The comparison of flow signal suppression betweenconventional radial SSFP and radial SSFP with randombipolar gradients was performed in two different imag-ing planes: transverse (Fig. 4a and b) and coronal (Fig.4c and d).

Human Volunteer Studies

Axial slices through the carotid artery bifurcations offour asymptomatic volunteers (28–37 years old, meanage � 31 years) were imaged using a conventional turbospin-echo (TSE) sequence with DIR preparation, a con-ventional radial SSFP sequence, and the new radialrandom bipolar dark-blood SSFP sequence. All studieswere performed on a 1.5 T Siemens Sonata MRI scanner(Siemens Medical Solutions, Erlangen, Germany) with-out cardiac or respiratory gating. To obtain comparativeimages, the following sequence parameters were used

Figure 3. SA vs. through-plane blood flow velocity withthe radial random bipolar gra-dient technique. The circlepoints/error bar stand for themean value/SD of 500 permu-tations of the randomizationseries for the bipolar gradientamplitude. Note that the bloodsignal is significantly reducedby this method when blood ve-locity is �0.1 m/second (flipangle � 65°, TR � 12 msec,TE � 7 msec, slice thickness �5 mm, number of radial projec-tion lines � 128).

Dark-Blood Random Bipolar Radial SSFP 1301

for both radial SSFP sequences: slice thickness � 5mm, acquisition matrix � 256 � 256, flip angle � 65°,TR � 12 msec, and TE � 7 msec. Note that extensivevariation of the first gradient moment of bipolar gradi-ents is critical for flow suppression in the current dark-blood method. Hence, we utilized a series of bipolargradients with a maximally randomizing scheme (i.e.,bipolar gradient amplitude has maximal range basedon the limitation of MRI scanner) for gradient amplitudeand appropriate gradient duration (total duration � 4msec for one pair of bipolar gradients). This was essen-tial in order to induce sufficient phase accumulation ineach projection despite the wide velocity range of hu-man blood flow in different applications. All of the invivo imaging experiments were conducted using an in-stitutional review board-approved protocol.

Quantitative Measurements

A circular region of interest (ROI) of 32 mm2 total areawas placed in the lumen of the right carotid artery in theimages to measure the mean SA of the flowing bloodthrough the entire vessel. The SAs of all of the pixelswithin the lumen were averaged. A ring-shaped ROIwas placed in the right carotid artery wall to measurethe SA of the vessel wall. For consistency, the sameROIs were chosen in the standard TSE, conventional,and radial random bipolar SSFP images. The lumen-to-vessel CNR value was calculated via the following equa-tion:

CNRVesselLumen � SAVessel � SALumen�/�Noise (2)

RESULTS

Numerical Simulations

Figure 3 shows a simulation of the steady-state SA vs.through-plane blood flow velocity with the random bi-polar gradient SSFP sequence. The circles represent themean signal magnitude from the 500 different randomseries, and the error bars represent the standard devi-ation (SD). Note that the blood signal is significantlyreduced by this method when blood velocity is over 0.1m/second. Note also that for flow velocities less than 3cm/second, there is a very slight enhancement relativeto stationary blood.

Phantom Imaging Studies

Figure 4 demonstrates the ability of the random bipolargradients to suppress both through-plane and in-planeflow. In the absence of random bipolar gradients (Fig.4a and c), the vessel phantoms appears as bright as orbrighter than the stationary water phantom. The addi-tion of the bipolar gradients produced a strong reduc-tion in the SA of the vessel phantom. It is important tonote that the SA of the spins in the static phantom werelargely unaffected by the application of the bipolar gra-dients.

Human Volunteer Studies

Figure 5 shows representative human volunteer carotidimages from the conventional (Fig 5b) and dark-blood(Fig. 5c) radial SSFP acquisitions along with the corre-sponding DIR-TSE image (Fig. 5a). Note that the con-ventional radial SSFP acquisition results in the familiarbright-blood signal in the conventional SSFP acquisi-tion (arrows in Fig. 5b). The carotid arteries are stillclearly visible in the conventional radial SSFP acquisi-tion with no flow suppression (Fig. 5b). The images fromthe radial random bipolar gradient SSFP acquisition(Fig. 5c) show significant signal loss near the bifurca-tion of the carotid artery. Also note that the vessel wallis clearly delineated from the lumen, while the contrastproperties of the other tissues are preserved.

The radial random bipolar SSFP sequence preservesthe major benefits of gradient-echo sequences (i.e.,rapid acquisition, multislice capability, and relativelylow SAR even at high field) and adds additional imagecontrast (dark-blood contrast) within a short acquisi-tion time (21 seconds/slice) relative to the current stan-dard protocol DIR-TSE (48 seconds/slice).

Quantitative Measurements

The in vivo imaging trials demonstrate substantiallyincreased vessel-to-lumen CNR from 10.23 for the con-ventional radial SSFP acquisition to 22.26 for the newrandom bipolar acquisition, with a modest increase inacquisition time to accommodate the additional bipolargradients (Table 1). Note the excellent suppression ofinflowing blood in the lumen of the carotids yet consis-tent SSFP contrast for the stationary tissues.

DISCUSSION

This study suggests that the use of random bipolargradients within a radial SSFP sequence is a time-effi-

Figure 4. Conventional radial SSFP (a and c) and radial SSFPwith random bipolar gradient (b and d) images of a large staticcylindrical water phantom flanked by two tubes of flowingwater. Application of the randomized bipolar gradients effec-tively suppressed both through-plane and in-plane flow.

1302 Lin et al.

cient method for achieving dark-blood contrast images.This technique utilizes the oversampling properties ofradial trajectories to effectively eliminate the typicalbright-blood signal in SSFP acquisitions for flowingspins. At the same time, the bipolar gradients maintainthe phase properties of static spins, resulting in thedesired high-SNR images for other tissues. The appli-cation of random bipolar gradients in all spatial direc-tions can achieve flow suppression regardless of theflow direction, e.g., close to the carotid bifurcation,where nearly in-plane flow exists, as well as through-plane flow. Furthermore, increasing the first gradientmoment of the bipolar gradients can achieve a higherlevel of phase dispersion for flowing spins at the ex-pense of slightly increased TR.

The main advantages of the radial random bipolarSSFP sequence, compared to the standard DIR prepa-ration, are that one can readily achieve dark-blood con-trast in a radial SSFP sequence by merely introducingthe bipolar gradients for dephasing moving spins, andflow artifact can be completely eliminated within theshort acquisition time of typical gradient-echo se-quences. Unlike most other dark-blood approaches,pulsatility effects do not interfere with this form of flowsuppression since the temporal variation simply addsadditional randomization of the view-to-view phasevariation. This new technique offers dramatically im-proved temporal resolution, multislice imaging capabil-

ities, and reduced acquisition time compared to DIR,the existing standard clinical dark-blood technique.Relative to normal SSFP, this acquisition technique canachieve dark-blood contrast with a short increase in theTR and TE (�4 msec) while maintaining the high tissuesignal-to-noise ratio (SNR) of static spins. Furthermore,this method places no additional restrictions on multi-slice acquisitions with dark-blood contrast. It allows astraightforward implementation of any radial steady-state acquisition, which is difficult to achieve using thecurrent standard dark-blood contrast (e.g., DIR withTSE acquisition). In addition, this simple flow cancella-tion technique can be applied to any other k-spacetrajectory that oversamples the central region ofk-space (i.e., rosette or multi-interleaf spiral). Thismethod can also be combined with the numerous fat-suppression methods in non-Cartesian k-space trajec-tories to obtain both blood- and fat-suppression imagesfor potential use in cardiac imaging plaque character-ization applications.

One important limitation of this dark-blood method isthat the additional large bipolar gradient lobes increasethe possibility of gradient-induced peripheral nervestimulation (PNS). However, PNS limits were notreached in any of the human sessions in this study.

Another limitation is that the imaged vessel wallthickness is a little greater than is typical in histologicalanalysis. This is likely due to the low blood flow veloc-

Figure 5. In vivo axial carotid vessel images from an asymptomatic volunteer are compared with images obtained by DIR-TSE(a), standard radial SSFP (b), and radial SSFP with the random bipolar gradient method (c) in the same section. The blood signalis significantly reduced and vessel wall anatomy is clearly delineated by the radial random bipolar gradient method andDIR-TSE. Acquisition parameters for a: DIR-TSE (TR � 2000 msec, TE � 12 msec, slice thickness � 5 mm, matrix � 256 � 256,NEX � 2, acquisition time � 48 seconds); b: standard radial SSFP (flip angle � 65°, TR � 12 msec, TE � 7 msec, slice thickness �5 mm, matrix � 256 � 256, NEX � 6, acquisition time � 21 seconds); and c: radial SSFP with random bipolar gradients sequence(flip angle � 65°, TR � 12 msec, TE � 7 msec, slice thickness � 5 mm, matrix � 256 � 256, NEX � 6, acquisition time � 21seconds).

Table 1Signal Amplitude and Contrast-to-Noise Measures in the Asymptomatic Human Studies

Sequences SA in vessel wall SA in lumen SA of noise CNR

DIR-turbo spin echo 128.6 12.20 30.1 2.87 3.38 29.14Radial SSFP 130.1 8.82 89.8 9.89 3.94 10.23Random bipolar radial SSFP 131.2 5.23 42.6 9.91 3.98 22.26

SA � signal amplitude, CNR � (SAVessel SALumen)/�Noise.

Dark-Blood Random Bipolar Radial SSFP 1303

ities near the vessel wall. The partial-volume effect in-creases the blood signal near the vessel wall, whichappears as an increased wall thickness. However, thecurrent standard vessel-wall imaging protocol (DIR)suffers from a similar artificial increase in vessel-wallthickness for all clinical applications.

Another potential problem with this method is addi-tional eddy-current effects resulting from the rapidlyvarying gradient amplitude of the additional bipolargradients. However, one can reduce eddy-current ef-fects by increasing the time interval between the bipolargradients and the data acquisition in order to allow timefor the decay of the eddy currents prior to data acqui-sition.

The modest increase of TR and TE that results fromadding the bipolar gradients also increases the sensi-tivity to field inhomogeneity and banding artifacts inthe SSFP signal. Therefore, optimization may be partic-ularly important for clinical use, since it will be impor-tant to mitigate some of the additional banding artifactand signal loss that may appear in the SSFP acquisitiondue to the longer TR.

In conclusion, we have shown that it is possible toachieve dark-blood contrast with high vessel-to-lumenCNR in a rapid acquisition by adding random bipolargradients to a conventional radial SSFP sequence. Thisapproach offers great promise for the clinical evaluationof atherosclerosis and vessel stenosis using a rapidacquisition.

REFERENCES1. Keller PJ, Drayer BP, Fram EK, Williams KD, Dumoulin CL, Souza

SP. MR angiography with two-dimensional acquisition and three-

dimensional display. Work in progress. Radiology 1989;173:527–532.

2. Dumoulin CL, Souza SP, Walker MF, Wagle W. Three-dimensionalphase contrast angiography. Magn Reson Med 1989;9:139–149.

3. Koenig SH. From the relaxivity of Gd(DTPA)2- to everything else.Magn Reson Med 1991;22:183–190.

4. Hillenbrand CM, Jesberger JA, Wong EY, et al. Toward rapid highresolution in vivo intravascular MRI: evaluation of vessel wall con-spicuity in a porcine model using multiple imaging protocols. JMagn Reson Imaging 2006;23:135–144.

5. Edelman RR, Chien D, Kim D. Fast selective black blood MR imag-ing. Radiology 1991;181:655–660.

6. Felmlee JP, Ehman RL. Spatial presaturation: a method for sup-pressing flow artifacts and improving depiction of vascular anatomyin MR imaging. Radiology 1987;164:559–564.

7. Song HK, Wright AC, Wolf RL, Wehrli FW. Multislice double inver-sion pulse sequence for efficient black-blood MRI. Magn Reson Med2002;47:616–620.

8. Yarnykh VL, Yuan C. Multislice double inversion-recovery black-blood imaging with simultaneous slice reinversion. J Magn ResonImaging 2003;17:478–483.

9. Flask CA, Dale B, Lewin JS, Duerk JL. Radial alternating TE se-quence for faster fat suppression. Magn Reson Med 2003;50:1095–1099.

10. Oppelt A, Graumann R, Barfuss H, Fischer H, Hartl W, Schajor W.FISP: a new fast MRI sequence. Electromedica (Engl Ed) 1986;54:15–18.

11. Markl M, Alley MT, Elkins CJ, Pelc NJ. Flow effects in balancedsteady state free precession imaging. Magn Reson Med 2003;50:892–903.

12. Moran PR. A flow velocity zeugmatographic interlace for NMR im-aging in humans. Magn Reson Imaging 1982;1:197–203.

13. Crawley AP, Henkelman RM. Errors in T2 estimation using multi-slice multiple-echo imaging. Magn Reson Med 1987;4:34–47.

14. Dale BM, Wendt M, Duerk JL. A rapid look-up table method forreconstructing MR images from arbitrary k-space trajectories.IEEE Trans Med Imaging 2001;20:207–217.

15. Matej S, Bajla I. A high-speed reconstruction from projections us-ing direct Fourier method with optimized parameters—an experi-mental analysis. IEEE Trans Med Imaging 1990;9:421–429.

1304 Lin et al.