Available online at www.sciencedirect.com

ng 26 (2008) 181–187

Magnetic Resonance Imagi

Whole-body MR angiography using variable density sampling anddual-injection bolus-chase acquisition

Jiang Dua,b,⁎, Frank R. Koroseca, Yijing Wua, Thomas M. Grista, Charles A. MistrettaaaDepartment of Medical Physics and Radiology, University of Wisconsin, Madison, WI 53792, USA

bDepartment of Radiology, University of California, San Diego, CA 92103-8756, USA

Received 16 November 2006; revised 6 June 2007; accepted 11 June 2007

Abstract

Conventional bolus-chase acquisition generates peripheral runoff images using a single injection of the contrast material. Low spatialresolution, small slice coverage and venous contamination are major problems especially in the distal stations. A technique is presented hereinin which whole-body magnetic resonance angiography is performed using a dual-contrast-injection four-station acquisition protocol. Bolussharing was performed between two stations: the abdomen and calf stations share the first bolus injection, while the thorax and thigh stationsshare the second bolus injection. The combination of variable density sampling and elliptical centric acquisition order was applied to theabdomen and thorax stations. The scan time was extended to generate high spatial resolution arterial phase images with broad slice coveragefor the calf and thigh stations. The feasibility of this technique was demonstrated using phantom and in vivo human volunteer studies.© 2008 Published by Elsevier Inc.

Keywords: Variable density sampling; Bolus chase; Dual injection; Whole-body MRA; Contrast-enhanced MRA

1. Introduction

Contrast-enhanced magnetic resonance angiography (CE-MRA) of the peripheral vasculature requires extendedlongitudinal field of view (FOV) to image from the aorticbifurcation to the feet. Currently, two major approaches havebeen proposed to cover the large FOV: multiple-injection,multiple-station acquisitions [1–4] and single-injectionbolus-chase acquisitions [5–13]. In the first approach, eachinjection is used to image one station with either singlevolume acquisition [1,2] or time-resolved acquisition [3,4].In the second approach, a single contrast bolus is adminis-tered intravenously followed by imaging of multiple stations.

CE-MRA studies require accurate timing of the acquisi-tion of the central k-space data to the arterial phase of thecontrast bolus to suppress venous contamination. In themulti-injection, multistation approach, timing of the imageacquisition for each station can be resolved well through

⁎ Corresponding author. Department of Radiology, University ofCalifornia, San Diego, CA 92103-8756, USA. Tel.: +1 619 471 0519; fax+1 619 471 0503

E-mail address: jiangdu@ucsd.edu (J. Du).

0730-725X/$ – see front matter © 2008 Published by Elsevier Inc.doi:10.1016/j.mri.2007.06.005

:

bolus timing techniques [12–14] or time-resolved imaging[3,4]. However, multi-injection reduces the contrast dose foreach station and the contrast enhancement for later injectionsdue to the residual contrast material in circulation [1,2].Therefore, there is a limitation for three-station peripheralscan and significant drawback for whole-body study thatrequires four or more injections.

Bolus-chase techniques use a single injection of a largecontrast dose to image multiple stations. Timing of the imageacquisition is usually optimized for the first station.However, timing for subsequent stations depends on thedata acquisition time for each station (12–30 s) and the tabletranslation time (3–8 s) between stations. According to arecent report by Prince et al. [15], the mean bolus travel timefrom the common femoral artery to the popliteal artery isapproximately 5 s, with another 7 s to reach the ankle artery.The current status of the hardware does not allow such fast3D acquisition to chase the bolus rapidly enough. Therefore,spatial resolution and slice coverage are typically reduced toshorten the scan time [16]. The success rate at the calf stationis low for bolus-chase techniques. Wang et al. [7] reported asuccess rate of 43%. Yucel and Reid [17] reported on abolus-chase study involving 67 symptomatic limbs whereonly 58% of the studies were completely diagnostic.

182 J. Du et al. / Magnetic Resonance Imaging 26 (2008) 181–187

Peripheral vascular disease reflects the systemic nature ofatherosclerotic disease and is frequently associated with theentire arterial system [8–10]. It is desirable to allow theconcomitant assessment of the whole arterial system fromthe supra-aortic arteries to the distal runoff vessels.Techniques such as AngioSURF addressed this problemeffectively by imaging four to five stations using a rollingtable platform integrated with a gliding body-array surfacecoil [8]. However, the distal station was acquired rather lateto avoid venous contamination.

In summary, the conventional bolus-chase techniquessuffer from low spatial resolution, a small number of slicesand venous contamination especially in the distal extre-mities. In this study, a technique for four-station whole-body MRA was proposed using dual-injection bolus-chaseacquisition. Variable density sampling was combined withrecessed elliptical centric acquisition to shorten scan time inthe upper stations, while extended elliptical centricacquisition was applied to the lower stations to increasespatial resolution and slice coverage. Phantom andvolunteers were scanned to investigate the feasibility ofthis technique.

2. Materials and methods

Variable density sampling reduces the sampling densityfrom low to high spatial frequencies to effectively reduce the

Fig. 1. The variable density sampling scheme is shown with (A) sampling density(dashed line) and sampling density for variable density sampling applied to both tcombination of variable density sampling and elliptical centric acquisition. The csampling along the phase-encoding direction only with strong aliasing artifact (thicmuch less aliasing artifact (thin arrow).

scan time with minimal high spatial frequency artifact[18,19]. This technique was initially implemented for MRfluoroscopy using spiral trajectory [18]. Recently, it wasapplied to Cartesian spin-warp imaging, where the densitysampling was monotonically reduced from low to highspatial frequency phase encodings [19]. Reasonable imagequality was achieved in approximately half the scan time.Elliptical centric acquisition is very efficient in depicting thearterial phase and suppressing venous contamination [14].Therefore, it is desirable to combine variable densitysampling with elliptical centric ordering for CE-MRA.Instead of variable density sampling along the phase-encoding direction only [19], a 3D fast spoiled gradient-recalled sequence was modified to permit variable densitysampling along both the phase-encoding and slice-encodingdirections. The sampling density was kept constant for thecenter 1/8 of the phase encodings and slice encodings andreduced to 1/4 of the maximum sampling density at the edgeof k-space. The center of k-space was fully sampled toreduce aliasing artifact because most of the image energy isconcentrated at the low spatial frequencies. The modifiedvariable density sampling elliptical centric acquisitionscheme is shown in Fig. 1.

A dual-injection protocol is proposed for four-stationwhole-body vasculature imaging. Fig. 2 shows the detail ofthis protocol, where the first injection of half the contrastbolus was used to image the abdomen and calf stations,followed by the second injection to image the thorax and

for variable density sampling applied to the phase-encoding direction onlyhe phase-encoding and the slice-encoding directions (solid line) and (B) theorresponding images are shown for (C) full sampling, (D) variable densityk arrow) and (E) variable density sampling elliptical centric acquisition with

Fig. 2. The dual-injection four-station whole-body angiography protocolshared a first injection of one half of the total contrast bolus between theabdomen and calf stations (solid lines) and a second injection of the otherhalf bolus between the thorax and thigh stations (dashed lines). Variabledensity recessed elliptical centric sampling with fluoro-triggering wasapplied to the two upper stations. An extended elliptical centric acquisitionup to 84 to 108 s was applied to the two lower stations.

Fig. 3. The coronal MIP whole-body 3D MR angiograms from two healthyvolunteers. The abdomen and calf stations shared the first injection of 20 mlcontrast bolus. The thorax and thigh stations shared the second injection ofanother 20 ml contrast bolus.

183J. Du et al. / Magnetic Resonance Imaging 26 (2008) 181–187

thigh stations. The thorax and abdomen stations were imagedseparately because of the requirement for breath-holding.The data acquisition of the central k-space was recessed byone quarter of the total scan time, where part of the highspatial frequency data were sampled during the rising edge ofthe contrast enhancement [16]. This procedure reducesvenous enhancement and ringing artifact and, more impor-tantly, shortens the delay time between the two bolus-chasestations. The combination of variable density sampling,recessed elliptical centric acquisition and fluoro-triggeringprovides a short delay time of 18–20 s between the upperstation and the lower station, which is significantly shorterthan that of the conventional bolus-chase techniques (33–50s). Extended elliptical centric acquisition was used for thecalf and thigh stations, where the scan time was extended to84–108 s to provide high spatial resolution, high signal-to-noise ratio (SNR) and broad slice coverage simultaneously.For comparison, dynamic imaging was performed for thetwo distal stations using a hybrid projection reconstruction(PR) TRICKS sequence [4]. Magnitude subtraction wasperformed for the lower stations to minimize the backgroundsignal. No subtraction was performed for the upper stationsto preserve SNR.

The main problem with this protocol is the contrastreduction due to dual injection of the contrast material. Theresidual contrast in circulation from the first injectionshortens blood T1, thus reducing contrast enhancement forthe second injection. This effect was investigated by imagingthe calf station twice using exactly the same protocol in fourhealthy volunteers.

Phantom and eight volunteer studies were performed on astandard 1.5-T MR scanner (Signa LX, GE Medical System,Waukesha, WI). A phantom was scanned using full samplingand the conventional and modified variable density sam-pling, respectively. Volunteers were scanned using body coilfor the abdomen/thorax stations and peripheral phased arraycoil (MRI Devices, Milwaukee, WI) for the calf/thighstations. The acquisition parameters for the abdomen/thoraxstations were as follows: FOV=40×32 cm, TR=5.0 ms,TE=1.3 ms, flip angle=30°, BW=41.67 kHz, acquisitionmatrix size=320×156, number of slices=36, slice

184 J. Du et al. / Magnetic Resonance Imaging 26 (2008) 181–187

thickness=2.5 mm. The extended elliptical centric acquisi-tion parameters for the calf/thigh stations were the following:FOV=40×32 cm, TR=4.8/4.5 ms, TE=1.4/1.3 ms, flipangle=30°, BW=62.5/41.67 kHz, acquisition matrixsize=512×312/384×312, number of slices=72/60, slicethickness=1.2/2.0 mm. The dynamic imaging parametersfor the calf/thigh stations were as follows: FOV=40×40 cm2,TR=7.5 ms, TE=2.7 ms, flip angle=30°, BW=62.5 kHz,acquisition readout=384, number of projections=72, numberof slices=72, slice thickness=1.5/2.0 mm. An injection of20 ml Gd-DTPA (Omniscan, Nycomed Amersham, Prince-ton, NJ) followed by 20 ml saline flush was applied to eachbolus-chase acquisition. A computer-controlled powerinjector (Spectris, Medrad, Indianola, PA) was used toensure a precise injection rate of 2.0 ml/s for the first 10 mlcontrast agent and 0.5 ml/s for the next 10 ml contrast agent.Written consent in accordance with institutional reviewboard rules was obtained before imaging procedures.

3. Results

The variable density sampling scheme and its combina-tion with elliptical centric acquisition are shown in Fig. 1A

Fig. 4. Isotropic spatial resolution and broad slice coverage are demonstratedthrough the sagittal (A) and coronal (B) MIP of the lower leg. The highspatial resolution and broad slice coverage offer excellent depiction of thesmall tibial and tibioperoneal arteries in the calf. Limited venous edgecontamination is seen in the MIP images.

and B. The phantom images demonstrated that the modifiedvariable density sampling scheme reduced the scan time by43% with limited high-frequency aliasing artifact (Fig. 1E),while the variable density sampling along the phase-encoding direction only produced much stronger high-frequency aliasing artifact (Fig. 1D).

SNR was measured for mask-subtracted calf images withthe first and second injection of 20 ml contrast dose underexactly the same imaging protocol for four volunteers. Themean SNR was 23.6 for the first injection and 18.8 for thesecond injection, corresponding to an SNR reduction of 20%for the second injection. Therefore, SNR for the thoraxstation will be reduced, but it can be compensated for thethigh station through extended acquisition. Further, there isno SNR degradation for the abdomen and calf stations sincethe first injection was used.

Fig. 3 shows the whole-body 3D MR angiograms fromtwo healthy volunteers. The arterial system from the supra-aortic arteries to the distal arteries was well depicted withhigh spatial resolution and SNR. Fig. 4 shows the coronaland sagittal projection of the mask-subtracted 3D calfimages. Excellent details of the tiny tibial and peronealarteries were depicted. The broad slice coverage of 86.4 cmdepicted the whole distal vessels and almost the whole foot.Fig. 5 shows the comparison between the coronal maximalintensity projection (MIP) of the calf using extendedelliptical centric acquisition and hybrid PR TRICKSacquisition, respectively. The dynamic images demonstrateda fast contrast transit time of 16 s from the abdomen stationto the calf station and a short arterial window of 12 s for thisvolunteer. Even under such short arteriovenous transit timeof the contrast agent, high-quality arterial phase image wasstill achieved using extended elliptical centric acquisition.

Fig. 6 shows the projection of the mask-subtracted 3Ddata sets for the thigh station. Dynamic images (not shownhere) demonstrated a short arterial window and quick venousreturn for this volunteer. However, there was little venouscontamination because of the short delay time between dataacquisitions in the thorax and thigh stations. There was somebackground tissue enhancement, which was due to thecontrast leakage from the first injection.

4. Discussion

MRA of the whole-body arterial system is challenging dueto several reasons: (a) broad coverage from the supra-aorticartery to the distal runoff vessels, (b) matching the contrasttransit time to the peak arterial enhancement for each stationand (c) high spatial resolution and high SNR arterial phaseimages for accurate diagnosis. Conventional bolus-chasetechniques address adequately the broad coverage through athree- to five-station acquisition with a single injection of thecontrast bolus. However, the calf station is typically acquired33 to 60 s after the contrast arrival at the abdomen station,which is too long for patients with quick venous

Fig. 5. The comparison between dual-injection extended elliptical centric acquisition (left) and time-resolved acquisition (right) demonstrates the excellendepiction of the tibial and tibioperoneal arteries using the new protocol. This volunteer has a very fast contrast arrival at the abdomen station (about 14 s), a shorcontrast transit time from the abdomen station to the calf station (about 16 s) and a short arterial window (about 12 s). In spite of this, little venous contaminationwas observed in the calf station using the two-station bolus-chase acquisition protocol.

185J. Du et al. / Magnetic Resonance Imaging 26 (2008) 181–187

enhancement. Furthermore, it is difficult to achieve highspatial resolution and broad slice coverage simultaneouslyfor the thigh and calf stations due to the short scan timeper station.

The dual-injection four-station protocol has a much lesslikelihood of venous contamination due to the bolus sharingbetween only two stations. The combination of variabledensity sampling and recessed elliptical centric acquisitionsignificantly improved the acquisition efficiency in the upperstations. As a result, data acquisition in the calf and thighstations could be initiated 18–20 s after contrast arrival at theupper station, which is a significant improvement over the33- to 60-s delay time in the conventional bolus-chasetechniques. The fact that there is no requirement for breath-holding, with less likelihood of motion, and minimizedvenous contamination in the thigh and calf stations allowsextended acquisition. Extended acquisition in the distalstations provides many advantages over the short acquisitiontime in the conventional bolus-chase acquisition. First,higher SNR can be achieved since SNR is proportional tothe square root of scan time. Second, higher spatial resolutionand larger slice coverage can be achieved. Broad slicecoverage is important so that the tibial artery, the peronealartery and the whole foot can be imaged in a single scan.Third, extended acquisition makes use of the residualcontrast in circulation, which typically maintains a steadysignal for several minutes [12]. High-quality MR angiograms

tt

of the distal stations are especially important since the lowerlimbs are the most frequently affected vascular territory with90% of identified atherosclerotic lesions found below theaortic bifurcation [20].

Dual-injection peripheral MRA could also be performedthrough dynamic imaging of the distal extremity followed bybolus chasing between the upper stations [21,22]. A recentreport by Schmitt et al. [22] showed a significant increase inanatomic coverage and temporal resolution of the cruropedalarteries using a hybrid dual-bolus three-station approachover the standard single bolus-chase acquisition. Although aquantitative comparison between these protocols and thenew four-station dual-injection protocol was not performedin this study, one advantage of the new approach is theextended coverage of the whole-body vasculature and highspatial resolution and broad coverage for both the calf andthigh stations.

The main problem with the dual-injection protocol is thatthe residual contrast bolus from the first injection enhancessignal from the vascular systems and background tissues and,thus, reduces contrast enhancement for the second injection.Rofsky et al. [1] investigated the application of two separateinjections for two different peripheral stations. Their resultsshowed no significant differences in diagnostic usefulness oroverall image quality between the first and second MRangiograms. Westenberg et al. [2] also performed similarstudies and concluded that there was little or no difference

Fig. 6. High spatial resolution and broad slice coverage are demonstrated for the thigh station in (A) the sagittal MIP of the right leg, (B) the coronal MIP of bothlegs and (C) the sagittal MIP of the left leg. High spatial resolution and broad slice coverage were demonstrated with little venous contamination. Theenhancement of background tissue (arrows) was due to the contrast leakage from the first injection.

186 J. Du et al. / Magnetic Resonance Imaging 26 (2008) 181–187

between the angiograms from the two injections. The calfimages from four volunteers in this study scanned underexactly the same protocol demonstrated a contrast enhance-ment reduction of about 20% for the second injection. Theextended acquisition not only can compensate for the SNRreduction but also can provide high spatial resolution andbroad slice coverage for both the thigh and calf stations.

It is critical to shorten the scan time in the thorax/abdomenstation to ensure an optimal coordination of central k-spaceacquisition with the contrast bolus arrival at the thigh/calfstation. Variable density sampling along both the phase- andslice-encoding directions reduces the scan time by 43% withminimal undersampling artifact (Fig. 1E). In our study, theperipheral phased array coil could only cover the thigh andcalf stations; thus, body coil was used for the thorax andabdomen stations. Phased array coils for the upper station willimprove image SNR and permit parallel imaging [11]. Thecombination of variable density sampling and parallel imagereconstruction will further reduce undersampling artifact andscan time, thus allowing faster transition from the upperstation to the lower station. Furthermore, the calf and thighstations can be acquired using either extended ellipticalcentric acquisition or dynamic imaging sequences such asCartesian or hybrid PR TRICKS [3,4]. The latter techniqueprovides single-volume images in the thorax and abdomen

stations and contrast dynamics in the distal extremities (thighand calf), a solution for patients with pathology-relateddelayed filling, which is problematic with any other bolus-chase techniques [5–13].

5. Conclusions

The feasibility of whole-body four-station MRA usingdual-injection bolus-chase acquisition was demonstratedthrough limited volunteer studies. Variable density sam-pling along both the phase-encoding and slice-encodingdirections significantly reduced the scan time with minimalaliasing artifact. Bolus sharing between two stationseffectively reduced the venous contamination in thesubsequent station and allowed extended acquisition inthe distal stations to significantly increase image SNR,spatial resolution and slice coverage.

References

[1] Rofsky N, Johnson G, Adelman M, Rosen RJ, Krinsky GA, WeinrebJC. Peripheral vascular disease evaluated with reduced-dose gadoli-nium-enhanced MR angiography. Radiology 1997;205:163–9.

[2] Westenberg JJ, Wasser MN, van der Geest RJ, Pattynama PM, de RoosA, Vanderschoot J, et al. Gadolinium contrast-enhanced three-

187J. Du et al. / Magnetic Resonance Imaging 26 (2008) 181–187

dimensional MRA of peripheral arteries with multiple bolus injection:scan optimization in vitro and in vivo. Int J Card Imaging 1999;15:161–73.

[3] Hany TF, Carroll TJ, Omary RA, Esparza-Coss E, Korosec FR,Mistretta CA, et al. Aorta and runoff vessels: single-injection MRangiography with automated table movement compared with multi-injection time-resolved MR angiography-initial results. Radiology2001;221:266–72.

[4] Du J, Carroll TJ, Wagner HJ, Vigen KK, Fain SB, Block WF, et al.Time-resolved, undersampled projection reconstruction imaging forhigh resolution CEMRA of the distal runoff vessels. Magn Reson Med2002;48:516–22.

[5] Meaney JF, Ridgway JP, Chakraverty S, Robertson I, Kessel D,Radjenovic A, et al. Stepping-table gadolinium-enhanced digitalsubtraction MR angiography of the aorta and lower extremity arteries:preliminary experience. Radiology 1999;211:59–67.

[6] Earls JP, DeSena S, Bluemke DA. Gadolinium-enhanced three-dimensional MR angioraphy of the entire aorta and iliac arteries withdynamic manual table translation. Radiology 1998;209:844–9.

[7] Wang Y, Winchester PA, Khilnani NM, Lee HM, Watts R, TrostDW, et al. Contrast-enhanced peripheral MR angiography from theabdominal aorta to the pedal arteries. Invest Radiol 2001;36:170–7.

[8] Goyen M, Quick HH, Debatin JF, Ladd ME, Barkhausen J, HerbornCU, et al. Whole-body three-dimensional MR angiography with arolling table platform: initial clinical experience. Radiology 2002;24:270–7.

[9] Leiner T, Ho K, Nelemans P, de Haan M, van Engelshoven J. Three-dimensional contrast-enhanced moving-bed infusion-tracking (MoBI-Track) peripheral MR angiography with flexible choice of imagingparameters for each field of view. J Magn Reson Imaging 2000;11:368–77.

[10] Shetty AN, Bis KG, Narra VR. Lower extremity MR angiography:universal retrofitting of high-field-strength systems with steppingkinematic imaging platforms — initial experience. Radiology2002;222:284–91.

[11] Maki JF, Wilson GJ, Eubank EB, Hoogeveen RM. Utilizing SENSE toachieve lower station sub-millimeter isotropic resolution and minimalvenous enhancement in peripheral MR angiography. J Magn ResonImaging 2002;15:484–91.

[12] Foo TKF, Ho VB, Hood MN, Marcos HB, Hess SL, Choyke PL. High-spatial-resolution multistation MR imaging of lower-extremity periph-eral vasculature with segmented volume acquisition: feasibility study.Radiology 2001;219:835–41.

[13] Earls JP, Rofsky NM, DeCorato DA, Krinsky GK,Weinreb JC. Breath-hold single-dose gadolinium-enhanced three-dimensional MR aorto-graphy: usefulness of a timing examination and MR power injector.Radiology 1996;201:705–10.

[14] Wilman AH, Riederer SJ, King BF, Debbins JP, Rossman PJ, EhmanRL. Fluoroscopically triggered contrast-enhanced three-dimensionalMR angiography with elliptical centric view order: application to therenal arteries. Radiology 1997;205:137–46.

[15] Prince MR, Chabra SG, Watts R, Chen CZ, Winchester PA, KhilnaniNM, et al. Contrast material travel times in patients undergoingperipheral MR angiography. Radiology 2002;224:55–61.

[16] Watts R, Wang Y, Prince MR, Winchester PA, Khilnani NM, Kent KC.Anatomically tailored k-space sampling for bolus-chase three-dimen-sional MR digital subtraction angiography. Radiology 2001;218:899–904.

[17] Yucel K, Reid SK. Pitfalls of moving-table peripheral MRA. XIIInternational Workshop on Magnetic Resonance Angiography, Lyon,France; 2000. p. 26.

[18] Spielman DM, Pauly JM, Meyer CH. Magnetic-resonance fluoroscopyusing spirals with variable sampling densities. Magn Reson Med1995;34:388–94.

[19] Tsai CM, Nishimura DG. Reduced aliasing artifacts using variable-density k-space sampling trajectories. Magn Reson Med 2000;43:452–8.

[20] Rutkow IM, Ernst B. An analysis of vascular surgical manpowerrequirements and vascular surgical rates in the United States. J VascSurg 1986;3:74–83.

[21] Du J, Korosec FR, Thornton FJ, Grist TM, Mistretta CA. High-resolution multistation peripheral MR angiography using under-sampled projection reconstruction imaging. Magn Reson Med2004;52:204–8.

[22] Schmitt R, Coblenz G, Cherevatyy O, Brunner H, Frohner S, Wedell E,et al. Comprehensive MR angiography of the lower limbs: a hybriddual-bolus approach including the pedal arteries. Eur Radiol 2005;15:2513–24.