The Magazine of MRIMAGNETOM FlashSCMR Issue 1/2013Not for distribution in the USACompressed SensingPage 4Real-Time Low-LatencyCardiac ImagingUsing Through-TimeRadial GRAPPAPage 6Myocaridal First-PassPerfusion Imagingwith High-Resolutionand Extended Coverageusing Multi-SliceCAIPIRINHAPage 10Acceleration ofVelocity EncodedPhase Contrast MR.New Techniquesand ApplicationsPage 18Free-BreathingReal-Time Flow Imagingusing EPI and SharedVelocity EncodingPage 24High AccelerationQuiescent-IntervalSingle Shot MRAPage 27Diffusion Tensor Imaging in HCM
2 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingTo aid standardization of CMR, the Societyfor Cardiovascular Magnetic Resonance (SCMR)released CMR exam protocol recommendationsfor the most frequent CMR procedures.In a collaborational effort Siemens Healthcareand the SCMR prepared clinically optimizedexam protocols in accordance to the SCMRrecommendations for 3T MAGNETOM Skyra and1.5T MAGNETOM Aera, with software versionsyngo MR D11.The following SCMR Cardiac Dot protocols areavailable:- Acute Infarct Dot- Adenosine Stress Dot- Aorta Dot- Arrhythmic RV Myopathy Dot- Chronic Ischemia Dot- Coronaries Dot- Mass & Thrombus Dot- Nonischemic Myopathy Dot- Pericardium Dot- Peripheral MRA Dot- Pulmonary Vein Dot- Valves Dot- Pediatric Teen Dot- Pediatric Child Dot- Pediatric Infant Dot- Library Cardiac Shim- Library TuneUp ShimPlease contact your SiemensApplication Specialist for the.edx files of these protocols.SCMR recommendedprotocols now availablefor Cardiac Dot EngineLocalizer Pseudo SAXLocalizer Iso ShimAdenosine Stress DotDefine LAXHaste AXLocalizerCine LAX FBCine LAX BHCine SAX FBCine SAX BHyesyesnonoBreath Hold?Breath Hold?Rest DynamicCine LAX BHFreqScout ES2.8Cine LAX FBFreqScout ES2.4DE LAX FBDE SAX FBTI Scout FBDE LAX BHDE SAX BHTI Scout BHyesyesnoyesnonoStopBreath Hold?Freq Scout?Breath Hold?STARTDefine SAXStress DynamicAcknowledgement: We would like to thank all SCMR memberswho were on the guidelines committeeChristopher M. Kramer (University of Virginia, Charlottesville, VA, USA);Jörg Barkhausen (University Hospital, Lübeck, Germany);Scott D. Flamm (Cleveland Clinic, Cleveland,OH, USA);Raymond J. Kim (Duke University Medical Center, Durham, NC, USA);Eike Nagel (King’s College, London, UK) as well asGary R. McNeal (Senior CMR Application Specialist; Siemens Healthcare, USA)for their tremendous efforts and support.
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 3ContentContentThe information presented in MAGNETOM Flash is for illustration only and is not intended to be relied upon bythe reader for instruction as to the practice of medicine.Any health care practitioner reading this information is reminded that they must use their own learning, trainingand expertise in dealing with their individual patients.This material does not substitute for that duty and is not intended by Siemens Medical Solutions to be used forany purpose in that regard. The treating physician bears the sole responsibility for the diagnosis and treatmentof patients, including drugs and doses prescribed in connection with such use. The Operating Instructions mustalways be strictly followed when operating the MR System. The source for the technical data is the correspondingdata sheets.MR scanning has not been established as safe for imaging fetuses and infants under two years of age. The respon-sible physician must evaluate the benefit of the MRI examination in comparison to other imaging procedures. 4 Compressed Sensing: A GreatOpportunity to Shift the Paradigmof Cardiac MRI to Real-TimeMichael O. Zenge 6 Real-Time Low-Latency CardiacImaging Using Through-TimeRadial GRAPPANicole Seiberlich et al. 10 Myocardial First-Pass PerfusionImaging with High Resolutionand Extended Coverage UsingMulti-Slice CAIPIRINHADaniel Stäb et al. 18 Acceleration of Velocity EncodedPhase Contrast MR. New Techniquesand ApplicationsJennifer Steeden, Vivek Muthurangu 24 Free-Breathing Real-Time FlowImaging using EPI and SharedVelocity EncodingNing Jin, Orlando “Lon” Simonetti 27 High Acceleration Quiescent-IntervalSingle Shot Magnetic ResonanceAngiography at 1.5 and 3TMaria Carr et al.References1 Nielles-Vallespin S, Mekkaoui C, Gatehouse P,Reese TG, Keegan J, Ferreira PF, Collins S,Speier P, Feiweier T, de Silva R, Jackowski MP,Pennell DJ, Sosnovik DE, Firmin D. In vivodiffusion tensor MRI of the human heart: Repro-ducibility of breath-hold and navigator-basedapproaches. Magn Reson Med 2012 Sep 21. doi:10.1002/mrm.24488. [Epub ahead of print].2 McGill LA, Ismail TF, Nielles-Vallespin S,Ferreira P, Scott AD, Roughton M, Kilner PJ, Ho SY,McCarthy KP, Gatehouse PD, de Silva R, Speier P,Feiweier T, Mekkaoui C, Sosnovik DE, Prasad SK,Firmin DN, Pennell DJ. Reproducibility of in-vivodiffusion tensor cardiovascular magnetic reso-nance in hypertrophic cardiomyopathy. J Cardio-vasc Magn Reson 2012; 14: 86.Cover imageCourtesy of Professor Dudley Pennell(Royal Brompton Hospital, London, UK)In-vivo diffusion tensor imaging in thebasal short-axis plane in a patient withhypertrophic cardiomyopathy (HCM).Acquisition used a diffusion-weightedstimulated-echo single-shot echo-planar-imaging sequence during multiplebreath-holds. The tensor is representedby superquadric glyphs, colour codedaccording to the helix angle of the maineigenvector (myocyte orientation):blue right-handed, green circumferen-tial, red left-handed.
Clinical Cardiovascular Imaging4 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldCompressed Sensing:A Great Opportunity to Shift the Paradigm of Cardiac MRI to Real-TimeMichael O. Zenge, Ph.D.Siemens AG, Healthcare Sector, Erlangen, GermanyIn the clinical practice of magnetic reso-nance imaging (MRI) today, a compre-hensive cardiac examination is consid-ered one of the most challenging tasks.In this context, the Cardiac Dot Engineoffers a useful tool to achieve reproduc-ible image quality in a standardized andtime-efficient way which can be tailoredto the specific requirements of an indi-vidual institution. However, this approachcannot completely overcome the limita-tions of ECG-gated acquisitions and thenecessity to acquire multiple images inbreath-hold. Thus, cardiac MRI is particu-larly well suited to benefit from a groupof novel image reconstruction methodsknown as compressed sensing  thatpromise to significantly speed up data1 Selected time frames of a cardiac CINE MRI, short axis view, temporal resolution 41.3 ms. Comparison of TPAT 2 segmented vs. real-timecompressed sensing with net acceleration 7.1 / 20 5 / 20 10 / 201
Cardiovascular Imaging ClinicalMAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 5ContactMichael Zenge, Ph.D.Siemens AGHealthcare SectorIM MR PI CARDPostbox 32 6091050 ErlangenGermanymichael.firstname.lastname@example.orgReferences1 Candes, E.J.; Wakin, M.B.; An Introduction ToCompressive Sampling; IEEE Signal Proc Mag.,25:21-30 (2008).2 Lustig, M.; Donoho, D.; Pauly, J.M.; Sparse MRI:The Application of Compressed Sensing for RapidMR Imaging; Magn Reson Med. 58:1182-1195(2007).3 Liang D.; Liu B.; Wang J.; Ying L.; AcceleratingSENSE using compressed sensing; Magn. Reson.Med. 62:1574-84 (2009).4 Tsao J., Boesiger, P., Pruessmann, K.P.; k-t BLASTand k-t SENSE: Dynamic MRI With High FrameRate Exploiting Spatiotemporal Correlation;Magn Reson Med. 50:1031-1042 (2003).5 Liu J. et al.; Dynamic cardiac MRI reconstructionwith weighted redundant Haar wavelets;Proc. ISMRM 2012, #178.acquisition and that could completelychange the paradigm of current cardiacimaging.Compressed sensing methods wereintroduced to MR imaging  just a fewyears ago and have since been success-fully combined with parallel imaging .Such methods try to utilize the fullpotential of image compression duringacquisition of raw input data. Thus, incase of sparse, incoherent sampling, anon-linear iterative optimization avoidssub-sampling artifacts during the pro-cess of image reconstruction. Resultingimages represent the best solution con-sistent with the input data, but have asparse representation in a specific trans-form domain, e.g. the wavelet domain.In the most favorable case, residual arti-facts are not visibly perceptible or arediagnostically irrelevant. One major ben-efit of compressed sensing compared,for example, to k-t SENSE , is thattraining data is explicitly not requiredbecause prior knowledge is formulatedvery generically.Very close attention is being paid withinthe scientific community to compressedsensing and initial experience is promis-ing. To provide broad access for clinicalevaluation, the featured solution ofcompressed sensing  has been fullyintegrated into the Siemens data acqui-sition and image reconstruction environ-ment. This has enabled real-time CINEMRI with a spatial and temporal resolu-tion that compares to segmented acqui-sitions (Fig. 1), and first clinical studieshave been initiated. In the field of car-diac MRI in particular, a generalizationof the current solution, for example, toflow imaging or perfusion, is very likelyto be achieved in the near future. Thismight shift the paradigm to real-timeacquisitions in all cases and, in the longrun, render cardiac MRI much easier. Inthis scenario, acquisitions in breath-holdas well as ECG gating will no longer berequired. This will not only increasepatient comfort and speed up patientpreparation but will also add robustnessto the entire procedure of cardiac MRIin the future.15 / 20 20 / 20
6 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingIntroductionSegmented cine imaging using balancedsteady state free precession (bSSFP) hasbeen accepted as a gold standard forassessing myocardial function. However,in order to assess the entire volume ofthe heart, this method requires multiplebreath-holds to minimize respiratorymotion, which may not be possible forvery ill or uncooperative patients. Addi-tionally, due to the signal averaging insegmented cine imaging, it is difficult toachieve reliable imaging in patients witharrhythmias, and even to assess diastolicdysfunction in patients with regularsinus rhythm.Single-shot, ungated, free-breathingreal-time Cartesian MRI is currentlyunable to match both the temporal andspatial resolution requirements ofcardiac MRI. The SCMR recommends atleast 50 ms temporal resolution for aclinically acceptable cardiac functionassessment . However, this temporalresolution is only possible by sacrificingspatial resolution, or overall imagequality, with existing Cartesian real-timeMRI methods. Radial acquisitions aremore tolerant to undersampling thanCartesian acquisitions due to the over-sampled central k-space and incoherentaliasing artifacts. However, the degreeof undersampling required to match thetemporal resolution recommended byReal-Time Low-Latency Cardiac ImagingUsing Through-Time Radial GRAPPAHaris Saybasili, Ph.D.1; Bruce Spottiswoode, Ph.D.2; Jeremy Collins, M.D.3; Sven Zuehlsdorff, Ph.D.2;Mark Griswold, Ph.D.1,4; Nicole Seiberlich, Ph.D.11Department of Biomedical Engineering, Case Western Reserve University, Cleveland, OH, USA2Siemens Healthcare, Chicago, IL, USA3Department of Radiology, Northwestern University, Chicago, IL, USA4Department of Radiology, University Hospitals of Cleveland and Case Western Reserve University, Cleveland, OH, USAthe SCMR for cardiac studies may stilllead to significant aliasing artifacts. Forthis reason, several parallel imagingreconstruction strategies for acceleratedradial imaging have been proposed toincrease the temporal resolution withoutsacrificing image quality, includingk-space domain radial GRAPPA [2, 3]and image domain iterative ConjugateGradient SENSE  methods. Bothmethods yield good image quality withvery high acceleration rates.1 Depiction of the hybrid through-time/through-k-space radial GRAPPA calibration process.(1A) Through-k-space calibration with a small segment alone does not yield the necessarynumber of kernel repetitions to calculate the GRAPPA weights (1B) Through-time calibration,using multiple calibration frames, allows the weights to be estimated for the local geometry.The GRAPPA kernel is assumed to be uniform, and the same segment is observed through-time to obtain enough data to solve the inverse problem.In k-space calibration Through-time calibration using N repetitionsTarget sample to estimateFrame 1 | Frame 2 | | Frame NSource sample in a GRAPPA block1A 1B
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 7Cardiovascular Imaging ClinicalThis work describes a through-timeradial GRAPPA implementation capableof reconstructing images with both hightemporal and spatial resolution whichcan be used for real-time cardiac acquisi-tions. Radial GRAPPA is similar to theCartesian GRAPPA, in that missingk-space points are synthesized by con-volving acquired points with a GRAPPAkernel. However, due to the non-uniformradial undersampling pattern in k-space,a single kernel cannot be used to esti-mate all the missing samples in theradial case. In the original formulationdescribed by Griswold et al. , k-spaceis divided into small segments andreconstructed using one GRAPPA kernelper segment, where each segment wastreated as a separate Cartesian GRAPPAproblem. When working with the highacceleration factors needed for real-timecardiac imaging, this segmentationapproach leads to errors in the GRAPPAweights as the segments no longer havethe requisite quasi-Cartesian geometry.In order to generate more accurateGRAPPA weights for non-Cartesian tra-jectories, the through-time non-Carte-sian GRAPPA can be employed . In thethrough-time approach, individual andgeometry-specific kernels are calculatedfor each missing sample for higheraccuracy. In order to collect enoughrepetitions of the kernel to calculate theGRAPPA weights for the local geometry,multiple fully sampled calibration frameseach exhibiting the same local geometryare acquired. For cardiac imaging, thesecalibration frames can be acquired as aseparate free-breathing reference scan.This hybrid through-time/through-k-space calibration scheme is depicted inFigure 1. Using a small segment (e.g.4 points in the readout direction and1 point in the projection direction, or4 × 1) combined with a long calibrationscan (~30 seconds) provides the highestimage quality. However, with only anominal loss in image quality, it is possi-ble to decrease the calibration scan timeby increasing the segment size. Figure 2shows four images reconstructed froma single undersampled dataset using dif-ferent calibration configurations. Eventhough longer calibrations provide bet-2 Radial GRAPPA reconstruction showing the effect of using a larger segment size to reducethe amount of calibration data required. A long calibration scan provides superior imagequality, but adequate image quality can be obtained in a reasonable calibration time using alarger segment. (2A) Segment size 4 × 1; calibration frames 80; calibration time 30.4 seconds.(2B) Segment size 8 × 4; calibration frames 20; calibration time 7.6 seconds. (2C) Segmentsize 16 × 4; calibration frames 10; calibration time 3.8 seconds. (2D) Segment size 24 × 8;calibration frames 2; calibration time 0.8 seconds.2A 2B2C 2D3 Example radial GRAPPA images with varying temporal/spatial resolutions. (3A) Short axisview with FOV 300 mm, image matrix 128 × 128, temporal resolution 48 ms, calibration matrixsize 128 × 256, accelerated imaging matrix size 16 × 256, radial GRAPPA acceleration rate 8.(3B) Short axis view with FOV 300 mm, image matrix 128 × 128, temporal resolution 24 ms,calibration matrix size 128 × 256, accelerated imaging matrix size 8 × 256, radial GRAPPAacceleration rate 16.3A 3B
8 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular Imagingter quality, acceptable images can beobtained using considerably shorter cali-bration times (~3 seconds). Becauseimage quality can be greatly improvedby using additional through-time infor-mation, this calibration scheme wasimplemented for real-time acquisitionand reconstruction on the scanner.Using Tim4G receiver coils, accelerationfactors of 16 are achievable, resulting ina 24 ms temporal resolution with a clini-cally acceptable spatial resolution. Bothcalibration and reconstruction steps arebased on a previous implementationdescribed in Saybasili et al. . Theonline reconstruction is multi-threadedand GPU-capable, with reconstructiontimes of less than 100 ms per frame. Thecalibration process is also multi-threaded. Figure 3 shows example shortaxis cardiac images acquired with radialGRAPPA acceleration rates of 8 and 16from a normal volunteer scanned duringfree breathing on a 3T MAGNETOMSkyra. For assessing functional parame-ters such as ejection fraction, a user maychoose to use lower acceleration ratesand benefit from the improved imagequality, whereas a higher accelerationrate with lower image quality may bepreferable to appreciate detailed myo-cardial kinetics.Radial GRAPPA is able to match thespatial resolution of the standard seg-mented bSSFP cine while providing ahigher temporal resolution. Figure 4compares short-axis cardiac images withthe same spatial resolution recon-structed using radial GRAPPA, seg-mented cine, and Cartesian real-timesequences. While the temporal resolu-tion of segmented cine images was40 ms, radial GRAPPA provided 25 mstemporal resolution for the same spatialresolution. Cartesian real-time cineimages could match neither the tempo-ral or spatial resolution of the seg-mented cine acquisition. The radialGRAPPA images show considerableimprovement compared to the standardCartesian real-time images, both inspatial and temporal resolution. The seg-mented cine images are superior inimage quality due to averaging, but as aresult also lack the temporal fidelity dur-ing late diastole, and rely on a consistentbreath-hold position.Clinical case:Left ventricular aneurysmThe patient is a 59-year-old male mara-thon runner with a past medical historyof hyperlipidemia and ST-segment eleva-tion myocardial infarction 5 months agowith symptoms starting while out fora run. The culprit lesion was in the leftanterior descending coronary and thepatient underwent urgent angiographywith bare metal stent placement with anexcellent angiographic result. Sincethen, the patient has resumed runningand is asymptomatic, although a recentechocardiogram demonstrated moder-ate systolic dysfunction with an ejectionfraction of 30%. The patient underwentcardiac MR imaging for assessment ofischemic cardiomyopathy. Due to thepatient’s intermittent arrhythmia, real-time cine imaging was employed. Thechamber is enlarged, with wall motionabnormalties in the left anteriordescending coronary artery territory,which is predominantly non-viable.There is an apical aneurysm withoutthrombus, best appreciated on hightemporal resolution cine imaging(Fig. 5).Clinical case:Diastolic dysfunctionThe patient is a 64-year-old asymptom-atic male with history of hypertensionfor which he takes two antihypertensivemedications. At a routine echocardiogra-phy, he was diagnosed with grade I dia-stolic dysfunction as well as an aorticaneurysm. The patient was referred forcardiac MRI. Real-time cine imaging wasperformed due to difficulties withbreathholding. Impaired myocardialrelaxation (grade I diastolic dysfunction)is evident by slowed myocardial relax-4 Mid ventricular short axis views. (4A) Radial GRAPPA, temporal resolution 25 ms, spatial resolution 2 × 2 × 8 mm3. (4B) Standard productsegmented bSSFP cine, temporal resolution 40 ms, spatial resolution 2 × 2 × 8 mm3. (4C) Cartesian real-time with E-PAT factor 3, temporalresolution 50 ms, spatial resolution 2.8 × 2.8 × 8 mm3.4A 4B 4C
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 9Cardiovascular Imaging Clinical5 (5A) Radial GRAPPA bSSFP cine with temporal resolution 25 ms, spatial resolution 1.7 × 1.7 × 8 mm3, radial GRAPPA acceleration rate 16. (5B)Cartesian real-time bSSFP cine with E-PAT factor 3, temporal resolution 50 ms, spatial resolution 3 × 3 × 8 mm3. (5C) Standard segmented bSSFPcine with temporal resolution 45 ms and spatial resolution 1.7 × 1.7 × 6 mm3. (5D) High temporal resolution segmented bSSFP cine acquired in a22 second breathhold, with temporal resolution 23 ms, and spatial resolution 1.7 × 1.7 × 6 mm3. The top row is end diastole and the bottom row isend systole. The bulging apical aneurism is best appreciated in the high temporal resolution cine images (5A) and (5D). The inhomogenity in(5A) is because the intensity normalization was off for the radial GRAPPA sequence and on for all other sequences.5A 5B 5C 5Dation on horizontal long axis cinesequences, better appreciated on thehighly accelerated real-time radialGRAPPA acquisition.Please scan theQR-Code® or visit us atwww.siemens.com/diastolic-dysfunctionto see the movies.ConclusionThrough-time radial GRAPPA is a robustparallel MRI method capable of produc-ing images with high spatial and tempo-ral resolution during un-gated free-breathing real-time acquisitions. Ourexperiments indicate that through-timeradial GRAPPA is capable of matchingthe spatial resolutions of segmentedbSFFP cine acquisitions, while offeringhigher temporal resolutions, withoutrequiring ECG gating or breath-holding.Additionally, this method outperformedCartesian real-time acquisitions in termof both spatial and temporal resolutions.The highly optimized multi-threadedCPU implementation (and GPU imple-mentation on supported systems)provides radial GRAPPA estimation per-formances of 55 – 210 ms per frame(12 – 56 ms on the GPU) for 20 – 30 coildata sets.This work demonstrates the quality ofimages that are achievable using highacceleration rates with Tim 4G coils andthe acquisition of separate and fullysampled reference data. The techniquehas the potential to directly improvepatient comfort, workflow, and diagnos-tic potential during a cardiac exam, anddeserves consideration for inclusion incurrent cardiac exam strategies. In addi-tion to imaging sick or uncooperativepatients, and patients with arrhythmias,the technique shows promise for assess-ing diastolic function, perfusion, andreal time flow.References1 Kramer et al. Standardized cardiovascular mag-netic resonance imaging (CMR) protocols, societyfor cardiovascular magnetic resonance: board oftrustees task force on standardized protocols.Journal of Cardiovascular Magnetic Resonance2008, 10:35.2 Griswold et al. In Proc 12th Annual Meeting ofthe ISMRM, Kyoto, Japan;2004:737.3 Seiberlich et al. Improved radial GRAPPA calibra-tion for real-time free-breathing cardiac imaging.Magn Reson Med 2011;65:492-505.4 Pruessman et al. Advances in sensitivity encodingwith arbitrary k-space trajectories. Magn ResonMed 2001; 46:638-651.5 Saybasili et al. Low-Latency Radial GRAPPAReconstruction using Multi-Core CPUs andGeneral Purpose GPU Programming. Proceedingsof the ISMRM 2012, Melbourne, Australia.ContactNicole Seiberlich, Ph.D.Assistant Professor,Department of Biomedical EngineeringCase Western Reserve UniversityRoom 309 Wickenden Building2071 Martin Luther King Jr. DriveCleveland, OH, 44106-7207USAnes30@case.eduQR-Code® is a registered trademark of DENSO WAVE Inc.
10 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingBackgroundContrast-enhanced myocardial first-passperfusion MR imaging (MRI) is a power-ful clinical tool for the detection of coro-nary artery disease [1–4]. Fast gradientecho sequences are employed to visual-ize the contrast uptake in the myocar-dium with a series of saturation pre-pared images. However, the technique isstrictly limited by physiological con-straints. Within every RR-interval, only afew slices can be acquired with low spa-tial resolution, while both high resolu-tion and high coverage are required fordistinguishing subendocardial and trans-mural infarcted areas  and facilitatingtheir localization, respectively.Acceleration techniques like parallelimaging (pMRI) have recently showntheir suitability for improving the spatialresolution in myocardial first-pass perfu-sion MRI [5–7]. Within clinical settings,3 to 4 slices can be acquired every heart-beat with a spatial resolution of about2.0 × 2.0 mm2in plane . However, forincreasing anatomic coverage , stan-dard parallel imaging is rather ineffec-tive, as it entails significant reductionsof the signal-to-noise ratio (SNR):a) In order to sample more slices everyRR-interval, each single-slice mea-surement has to be shortened by acertain acceleration factor R, whichMyocardial First-Pass Perfusion Imagingwith High Resolution and ExtendedCoverage Using Multi-Slice CAIPIRINHADaniel Stäb, Dipl. Phys.1,2; Felix A. Breuer, Ph.D.3; Christian O. Ritter, M.D.1; Andreas Greiser, Ph.D.4;Dietbert Hahn, M.D.1; Herbert Köstler, Ph.D.1,21Institute of Radiology, University of Würzburg, Würzburg, Germany2Comprehensive Heart Failure Center (CHFC), Würzburg, Germany3Research Center Magnetic Resonance Bavaria (MRB), Würzburg, Germany4Siemens Healthcare, Erlangen, Germanydual-band rf pulseslice 1slice 2slice select gradientfZBWobjectrf phase slice 1rf phase slice 2kykxk-space0°0°0°0°0°0°0°0°180°0°180°0°180°0°Conventional MS-CAIPIRINHA for coverage extensionMulti-slice excitation and rf phase cycling1A MS-CAIPIRINHA with two slices simultaneously excited (NS = 2). A dual-band rf pulse is uti-lized to excite two slices at the same time (BW = excitation bandwidth). During data acquisition,each slice is provided with an individual rf phase cycle. A slice specific constant rf phase incre-ment is employed (0° in slice 1, 180° in slice 2) between succeeding excitations. Consequently,the two slices appear shifted by ½ FOV with respect to each other. In this way, the overlappingpixels originate not only from different slices, but also from different locations along the phaseencoding direction (y), facilitating a robust slice separation using pMRI reconstruction tech-niques. Using the two-slice excitation, effectively a two-fold acceleration is achieved (Reff = 2).1Asuperposition of slices NS = 2; Reff = 2slice 1 slice 2 superposition+ =xy
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 11Cardiovascular Imaging Clinicalinevitably comes along with an√R-fold SNR reduction.b) During image reconstruction, the SNRis further reduced by the so-calledgeometry (g)-factor , an inhomo-geneous noise-amplification depend-ing on the encoding capabilities ofthe receiver array.In addition, unless segmented acquisi-tion techniques are utilized , thesaturation recovery (SR) magnetizationpreparation has to be taken intoaccount:c) The preparation cannot be acceler-ated itself. Hence, the accelerationfactor R has to be higher than thefactor by which the coverage isextended. This leads to an increase ofthe SNR-reduction discussed in (a).the technique does not experienceany SNR reductions despite the g-factornoise amplification of the requiredpMRI reconstruction [11, 12]. TheMS-CAIPIRINHA concept can also beemployed with acceleration factors thatare higher than the number of slicesexcited at the same time. By utilizingthis acceleration for increasing spatialresolution, the technique facilitatesmyocardial first-pass perfusion examina-tions with extended anatomic coverageand high spatial resolution.A short overview of the concept is setout below, followed by a presentationof in-vivo studies that demonstrate thecapabilities of MS-CAIPIRINHA in con-trast-enhanced myocardial first-passperfusion MRI.Improving anatomic coverageand spatial resolution withMS-CAIPIRINHAMS-CAIPIRINHAThe MS-CAIPIRINHA concept [12, 13] isbased on a coinstantaneous excitationof multiple slices, which is accomplishedby means of multi-band radiofrequency(rf) pulses (Fig. 1A). Being subject to theidentical gradient encoding procedure,the simultaneously excited slices appearsuperimposed on each other, unless theindividual slices are provided with differ-ent rf phase cycles. In MS-CAIPIRINHA,the latter is done in a well-defined man-ner in order to control the aliasing of thesimultaneously excited slices. Making useof the Fourier shift theorem, dedicatedslice specific rf phase cycles are employedto shift the slices with respect to eachother in the field-of-view (FOV) (Fig. 1A).The slice separation is performedusing pMRI reconstruction techniques.However, the shift of the slices causessuperimposed pixels to originate fromnot only different slices, but also differ-ent locations along the phase encodingdirection. Thus, the MS-CAIPIRINHAconcept allows the pMRI reconstructionto take advantage of coil sensitivityvariations along two dimensions andto perform the slice separation withlow g-factor noise amplification .d) Subsequent to the preparation, thesignal increases almost linearly withtime. Thus, shortening the acquisitionby a factor of R is linked to an addi-tional R-fold SNR-loss.As demonstrated recently , mostof these limitations can be overcomeby employing the MS-CAIPIRINHA(Multi-Slice Controlled Aliasing InParallel Imaging Results IN HigherAcceleration) concept [12, 13] for simul-taneous 2D multi-slice imaging. Bysimultaneously scanning multiple slicesin the time conventionally requiredfor the acquisition of one single slice,this technique enables a significantincrease in anatomic coverage. Sinceimage acquisition time is preserved withrespect to the single-slice measurement,k-space extensionfor increasing thespatial resolutionacquired linek-spaceMS-CAIPIRINHA with Reff> NSfor improving coverage and resolutionMulti-slice excitation, k-space undersampling and rf phase cyclingkykxnon-acquired line0°0°0°0°0°0°0°0°180°0°180°0°180°0°1B MS-CAIPIRINHA with an effective acceleration factor higher than the number of slicesexcited simultaneously. Again, two slices are excited at the same time employing the same rfphase cycles as in (1A), but k-space is undersampled by a factor of two. The slices and theirundersampling-induced aliasing artifacts appear shifted with respect to each other by ½ FOVand can be separated using pMRI reconstruction techniques. Effectively, a four-fold accelerationis achieved (Reff = 4). The additional acceleration can be employed to extend k-space and toimprove the spatial resolution.1B+ =superposition of aliased slices NS = 2; Reff = 4xy slice 1with aliasingslice 2with aliasingsuperposition
12 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingBy conserving image acquisition timewith respect to an equivalent single-slicemeasurement, the technique is not sub-ject to any further SNR penalties. MS-CAIPIRINHA hence allows extending thecoverage in 2D multi-slice imaging in avery efficient manner. Applied to myo-cardial first-pass perfusion imaging, theconcept facilitates the acquisition of6 slices every heartbeat with an imagequality that is comparable to that ofconventional 3-slice examinations .Additional accelerationWhile extending the coverage can beaccomplished by simultaneous multi-slice excitation, improving the spatialresolution requires additional k-spacedata to be sampled during imageacquisition. Of course, as the FOV andthe image acquisition time are to beconserved, this can only be achieved bymeans of additional acceleration.However, the effective acceleration fac-tor of MS-CAIPIRINHA is not restricted tothe number of slices excited simultane-ously. By applying simultaneous multi-slice excitation to an imaging protocolwith reduced phase FOV, i.e. equidistantk-space undersampling, supplementaryin-plane acceleration can be incorpo-rated (Fig. 1B). The rf phase modulationforces the two simultaneously excitedslices and their in-plane aliasing artifactsto be shifted with respect to each otherin the FOV. As before, image reconstruc-tion and slice separation is performedutilizing pMRI methods. Employed likethis, the MS-CAIPIRINHA concept facili-tates an increase of both, anatomic cov-erage and spatial resolution with highSNR efficiency. Since image acquisitiontime does not have to be shortened,SNR is only affected by the voxel sizeand the noise amplification of the pMRIreconstruction.ImagingPerfusion datasets were obtained fromseveral volunteers and patients. Thestudy was approved by the local EthicsCommittee and written informed con-sent was obtained from all subjects.All examinations were performed on aclinical 3T MAGNETOM Trio, a Tim sys-tem (Siemens AG, Healthcare Sector,Erlangen, Germany), using a dedicated32-channel cardiac array coil (SiemensAG, Healthcare Sector, Erlangen, Ger-many) for signal reception. Myocardialperfusion was assessed using a SR FLASHsequence (FOV 320 × 300-360 mm2;matrix 160 × 150-180; TI 110-125 ms;TR 2.8 ms; TE 1.44 ms; TAcq 191-223 ms;slice thickness 8 mm; flip angle 12°).Two slices were excited at the same time(distance between simultaneouslyexcited slices: 24-32 mm) and shifted by½ FOV with respect to each other byrespectively providing the first and sec-ond slice with a 0° and 180° rf phasecycle. In order to realize a spatial resolu-tion of 2.0 × 2.0 mm2within the imagingplane, k-space was undersampled bya factor of 2.5, resulting in an overalleffective acceleration factor of 5.2 Myocardial first-pass perfusion study on a 52-year-old male patientafter ST-elevating myocardial infarction and acute revascularization.Myocardial perfusion was examined in 6 adjacent slices by performing3 consecutive MS-CAIPIRINHA acquisitions every RR-interval (FOV320 × 360 mm2; matrix 160 × 180; TI 125 ms; TAcq 223 ms; distancebetween simultaneously excited slices: 24 mm). (2A) Sections of thereconstructed images of all 6 slices showing the pass of the contrastagent through the myocardium. The hypoperfused area is depicted byarrows. (2B) Image series showing the contrast uptake in the myocar-dium of slice 3. The acquisition time point (RR-interval) is given foreach image and the perfusion defect is indicated by arrows. (2C)g-factor maps for the reconstruction shown in (2A).g-factor3,02,01,02A 2B2C
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 13Cardiovascular Imaging Clinical3 Myocardial first-pass perfusionstudy on a 51-year-old female volun-teer with whole heart coverage.6 consecutive MS-CAIPIRINHA acquisi-tions were performed every two RR-intervals (FOV 320 × 300 mm2; matrix160 x 150; TI 110 ms; TAcq 191 ms;distance between simultaneouslyexcited short/long axis slices:32/16 mm). (3A) Sections of thereconstructed short axis images. (3B)Sections of the reconstructed longaxis images. (3C) Series of image sec-tions showing the contrast uptake inthe myocardium of slice 2. The acqui-sition time point (RR-interval) is givenfor each image. (3D) g-factor mapsfor the reconstructions shown in (3A).(3E) g-factor maps for the reconstruc-tions shown in (3B). (3F) Measurementscheme. The acquisition was per-formed with a temporal resolution of1 measurement every 2 RR-intervals.g-factor3,02,01,03A 3B3C3D 3E3F1 measurement every 2 RR-intervalsSlice 1/5Slice 2/6Slice 3/7Slice 4/8Slice 9/11Slice 10/12
14 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingAll first-pass perfusion measurementswere conducted in rest over a total of 40heartbeats. All subjects were asked tohold their breaths during the acquisitionas long as possible. Every RR-interval, 3to 4 consecutive MS-CAIPIRINHA acquisi-tions were performed in order to samplethe contrast uptake of the myocardium.For contrast enhancement, a contrastagent bolus (4 ml, Gadobutrol, BayerHealthCare, Berlin, Germany) followedby a 20 ml saline flush was administeredat the beginning of each perfusion scan.Image reconstruction was performedusing an offline GRAPPA  reconstruc-tion. The according weights were deter-mined from a separate full FOV calibra-tion scan. To evaluate the GRAPPAreconstruction, an additional noise scanwas obtained and the g-factor noiseenhancement was quantified . Allcalculations were performed on a stand-alone PC using Matlab (The MathWorks,Natick, MA, USA).ResultsFigure 2 shows the results of a myocardialfirst-pass perfusion examination with6-slices on a 52-year-old male patient(91 kg, 185 cm) on the fourth day afterSTEMI (ST-elevating myocardial infarc-tion) and acute revascularization. Sec-tions of the reconstructed images of allexamined slices are depicted, showingthe first pass of contrast agent throughthe myocardium (Fig. 2A). Also dis-played is a series of image sectionsdemonstrating contrast uptake in slice 3(Fig. 2B). The GRAPPA reconstructionseparated the simultaneously excitedslices without visible artifacts. The g-fac-tor maps (Fig. 2C) indicate generallylow noise amplification. Thus, in com-parison to the high effective accelera-tion factor of 5, the images provideexcellent image quality. Both contrastand SNR allow for a clear delineation ofthe subendocardial hypoperfused areawithin the anterior and septal wall of themyocardium (arrows).The results of a first-pass perfusionstudy with whole heart coverage are dis-played in Fig. 3. In a 51-year-old femalevolunteer, first pass perfusion was exam-ined in eight short (Fig. 3A) and fourlong axis slices (Fig. 3B). In order toaccomplish the 12-slice examination andto achieve whole heart coverage, tem-poral resolution was reduced by a factorof two with respect to the examinationdisplayed in Fig. 2. The contrast uptakeof the myocardium was sampled with1 measurement every two RR-intervalsby performing 3 out of 6 consecutivedouble-slice MS-CAIPIRINHA acquisitionsevery heartbeat (Fig. 3F). Image recon-struction could be performed withoutvisible artifacts and generally low noiseamplification (Figs. 3D and E). Only ina few regions, the g-factor maps showmoderate noise enhancement. In theimages, the myocardium is homoge-neously contrasted and the contrastagent uptake is clearly visible (Fig. 3C).The findings of a first-pass perfusionstudy in a 48-year-old male patient(80 kg, 183 cm) on the eighth day afterSTEMI and acute revascularization arepresented in Fig. 4. An overall of 8 sliceswere acquired with a temporal resolutionof 1 measurement every RR-interval byperforming 4 consecutive MS-CAIPIRINHAacquisitions after each ECG trigger pulse.Sections of the reconstructed images,showing the first pass of contrast agentthrough the myocardium in all 8 slices(Fig. 4A) are depicted together withsections demonstrating the process ofcontrast uptake in slice 5 (Fig. 4B).Despite the breathing motion (Fig. 4D),the GRAPPA reconstruction performedrobustly and separated the slices with-out significant artifacts. The g-factornoise amplification is generally low andmoderate within a few areas (Fig. 4C).In the images, the hypoperfused suben-docardial region in the anterior wall canbe clearly identified (arrows). As can beseen from the enlarged section of slice 5(Fig. 4F), the technique provides suffi-cient spatial resolution to distinguishbetween subendocardial and transmuralperfusion defects. These findings cor-respond well to the results of a subse-quently performed Late Enhancementstudy (Fig. 4E) delineating a transmuralinfarction zone of the anterior wall(midventricular to apical).DiscussionContrast-enhanced myocardial first-passperfusion MRI with significantlyextended anatomic coverage and highspatial resolution can be successfullyperformed by employing the MS-CAIPIR-INHA concept for simultaneous multi-slice imaging. Basically, two differentacceleration approaches are combined:the simultaneous excitation of two sliceson the one hand and k-space undersam-pling on the other. While the firstdirectly doubles the number of slicesacquired, the second provides sufficientacceleration for improving the spatialresolution. The proposed imaging proto-cols provide an effective accelerationfactor of 5, which is sufficient for theacquisition of 6 to 8 slices every RR-interval with a high spatial resolution of2.0 × 2.0 × 8 mm3. Correspondingly,whole heart coverage can be achievedby sampling 12 slices with a temporalresolution of 1 measurement every 2 RR-intervals. Since the slices can be plannedwith individual thickness and pairwisespecific orientation, the concept therebyprovides high flexibility. With imageacquisition times of 191 ms, it also sup-ports stress examinations in 6 slices upto a peak heart rate of 104 bpm.Employing a dedicated 32-channel car-diac array coil, the image reconstruc-tions could be performed without signi-ficant reconstruction artifacts and onlylow to moderate g-factor noise amplifi-cation. Also in presence of breathingmotion, the GRAPPA reconstruction per-formed robustly. The images providedsufficient SNR and contrast betweenblood, myocardium and lung tissue todelineate small perfusion defects and todifferentiate between subendocardialand transmural hypoperfused areas.Compared to conventional parallel MRI,the MS-CAIPIRINHA concept benefitsfrom the high SNR efficiency discussedearlier. Simultaneous multi-slice excita-tion allows increasing the coveragewithout supplementary k-space under-
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 15Cardiovascular Imaging Clinicalsampling. At the same time the g-factornoise amplification is minimizedby exploiting coil sensitivity variationsin both, slice and phase encodingdirection. Thus, despite the doubledanatomic coverage, the image qualityobtained is comparable to that of anaccelerated measurement with stan-dard coverage and high spatial resolu-tion . An important feature of theMS-CAIPIRINHA concept in myocardialfirst-pass perfusion MRI is the frame-by-frame reconstruction, which preventsthe reconstructed images to be affectedby temporal blurring. Moreover, arrhyth-mia and breathing motion only impactthe underlying time frame and not thewhole image series, as it is likely forreconstruction techniques incorporatingthe temporal domain [16–19].4 Myocardial first-pass perfusion study on a 48-year-old male patient after ST-elevating myocardial infarction and acute revascularization. 8slices were acquired every RR-interval by performing 4 consecutive MS-CAIPIRINHA acquisitions (FOV 320 × 300 mm2; matrix 160 × 150; TI 110 ms;TAcq 191 ms; flip angle 10°; distance between simultaneously excited slices: 32 mm). (4A) Sections of the reconstructed images of all 8 slicesshowing the first pass of the contrast agent through the myocardium. The hypoperfused area is depicted by arrows. (4B) Image series showingthe contrast uptake in the myocardium of slice 5. The acquisition time point (RR-interval) is given for each image and the perfusion defect is indi-cated by arrows. (4C) g-factor maps for the reconstruction shown in (4A). (4D) Displacement of the heart due to breathing motion, example forslice 4. (4E) Late gadolinium enhancement. (4F) Enlarged section of slice 5. The technique allows distinguishing between subendocardial andtransmural perfusion defects.g-factor3,02,01,04A 4B4C 4D4E4F
16 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingReferences1 Atkinson DJ, Burnstein D, Edelman RR. First-passcardiac perfusion: evaluation with ultrafast MRimaging. Radiology 1990; 174:757–762.2 Wilke N, Jerosch-Herold M, Wang Y, Yimei H,Christensen BV, Stillman E, Ugurbil K, McDonaldK, Wilson RF. Myocardial Perfusion Reserve:Assessment with Multisection, Quantitative,First-Pass MR Imaging. Radiology 1997;204:373–384.3 Rieber J, Huber A, Erhard I, Mueller S, SchweyerM, Koenig A, Schiele TM, Theisen K, Siebert U,Schoenberg SO, Reiser M, Klauss V. Cardiac mag-netic resonance perfusion imaging for the func-tional assessment of coronary artery disease:a comparison with coronary angiography andfractional flow reserve. Eur Heart J 2006;27:1465–1471.4 Schwitter J, Nanz D, Kneifel S, Bertschinger K,Büchi M, Knüsel PR, Marincek B, Lüscher TF,Schulthess GK. Assessment of Myocardial Perfu-sion in Coronary Artery Disease by MagneticResonance. Circulation 2001; 103:2230–2235.5 Ritter CO, del Savio K, Brackertz A, Beer M, HahnD, Köstler H. High-resolution MRI for the quanti-tative evaluation of subendocardial and subepi-cardial perfusion under pharmacological stressand at rest. RoFo 2007; 179:945–952.6 Strach K, Meyer C, Thomas D, Naehle CP,Schmitz C, Litt H, Bernstein A, Cheng B, Schild H,Sommer T. High-resolution myocardial perfusionimaging at 3 T: comparison to 1.5 T in healthyvolunteers. Eur Radiol 2007; 17:1829–1835.ContactDaniel StäbInstitute of RadiologyUniversity of WürzburgOberdürrbacher Str. 697080 WürzburgGermanystaeb@roentgen.uni-wuerzburg.de7 Jung B, Honal M, Hennig J, Markl M. k-t-Spaceaccelerated myocardial perfusion. J Magn ResonImag 2008; 28:1080–1085.8 Köstler H, Sandstede JJW, Lipke C, Landschütz W,Beer M, Hahn D. Auto-SENSE perfusion imagingof the whole human heart. J Magn Reson Imag2003; 18:702–708.9 Pruessmann KP, Weiger M, Scheidegger MB,Boesiger P. SENSE: sensitivity encoding for fastMRI. Magn Reson Med 1999; 42:952–962.10 Kellman P, Derbyshire JA, Agyeman KO, McVeighER, Arai AE. Extended coverage first-pass perfu-sion imaging using slice-interleaved TSENSE.Magn Reson Med 2004; 51:200–204.11 Stäb D, Ritter CO, Breuer FA, Weng AM, Hahn D,Köstler H. CAIPIRINHA accelerated SSFP imaging.Magn Reson Med 2011; 65:157–164.12 Breuer FA, Blaimer M, Heidemann RM, MuellerMF, Griswold MA, Jakob PM. Controlled aliasingin parallel imaging results in higher acceleration(CAIPIRINHA) for multi-slice imaging. MagnReson Med 2005; 53:684–691.13 Breuer F, Blaimer M, Griswold M, Jakob P. Con-trolled Aliasing in Parallel Imaging Results inHigher Acceleration (CAIPIRINHA). MagnetomFlash 2012; 49:135–142.14 Griswold MA, Jakob PM, Heidemann RM,Nittka M, Jellus V, Wang J, Kiefer B, Haase A.Generalized autocalibrating partially parallelacquisitions (GRAPPA). Magn Reson Med 2002;47:1202–1210.15 Breuer FA, Kannengiesser SAR, Blaimer M,Seiberlich N, Jakob PM, Griswold MA. Generalformulation for quantitative G-factor calculationin GRAPPA reconstructions. Magn Reson Med2009; 62:739–746.16 Adluru G, Awate SP, Tasdizen T, Whitaker RT,Dibella EVR. Temporally constrained reconstruc-tion of dynamic cardiac perfusion MRI. MagnReson Med 2007; 57: 1027–1036.17 Otazo R, Kim D, Axel L, Sodickson DK. Combina-tion of compressed sensing and parallel imagingfor highly accelerated first-pass cardiac perfu-sion MRI. Magn Reson Med 2010; 64:767–776.18 Ge L, Kino A, Griswold M, Mistretta C, Carr JC,Li D. Myocardial perfusion MRI with sliding-win-dow conjugate-gradient HYPR. Magn Reson Med2009; 62:835–839.19 Plein S, Kozerke S, Suerder D, Luescher TF,Greenwood JP, Boesiger P, Schwitter J. Highspatial resolution myocardial perfusion cardiacmagnetic resonance for the detection ofcoronary artery disease. Eur Heart J 2008;29:2148–2155.Simultaneous multi-slice excitation is,of course, linked to an increase in theamount of energy that is deployed tothe subject under investigation. Thus,limitations have to be expected at higherfield strengths or when using sequenceswith larger flip angles, such as TrueFISP.At 1.5 Tesla, the application of the MS-CAIPIRINHA concept to TrueFISP hasbeen successfully demonstrated utilizingadvanced rf phase cycling .In all in-vivo examinations, the distancebetween the two slices excited simulta-neously was maximized in order to makeuse of the highest possible coil sensitiv-ity variations in slice direction and tominimize the g-factor penalty. Thus, thespatial distance between consecutivelyacquired cardiac phases is large whichmight be a possible drawback for corre-lating hypoperfused regions of the myo-cardium. While the latter was feasiblefor all in-vivo studies, slice distancenaturally can be reduced at the expenseof slightly more noise enhancement.ConclusionUtilizing the MS-CAIPIRINHA concept forsimultaneous multi-slice imaging, con-trast-enhanced myocardial first-pass per-fusion MRI can be performed with ananatomic coverage of 6 to 8 slices everyheart beat and a high spatial resolutionof 2.0 × 2.0 × 8 mm3. Based on the simul-taneous excitation of multiple slices, theconcept provides significantly higherSNR than conventional in-plane acceler-ation techniques with identical accelera-tion factor and facilitates an accurateimage reconstruction with only low tomoderate g-factor noise amplification.Taking into account the high flexibility,simple applicability and short recon-struction times in addition to the highrobustness in presence of breathingmotion or arrhythmia, the concept canbe considered a promising candidatefor clinical perfusion studies.AcknowledgementsThe authors would like to thank the DeutscheForschungsgemeinschaft (DFG) and the FederalMinistry of Education and Research (BMBF),Germany for grant support.
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 17Evaluating cardiac flow data is nowautomated thanks to syngo.via.1Patients with compromised cardiac function require a more comprehensiveevaluation to determine areas of significant pathology. A key step in this processincludes cardiac flow quantification, which typically involves many labor-intensive steps. syngo®.via eliminates the need for most cumbersome tasks,providing a more streamlined workflow.• Time saving automation provides results quicker.• syngo.via compensates for up to 200 % inaccuracy in velocity encoding values,helping to avoid repeat scans2• Integrated reporting enables flow results to be accessible everywhere.1syngo.via can be used as a standalone device or together with a variety of syngo.via-based software options,which are medical devices in their own rights.2May not work in all situations. Please ensure proper VENC values by running a VENC-scout, and then setting theVENC for the diagnostic scan based on the outcome of the VENC-scout. In situations where the VENC was stillset too low, despite the results of the scout, the anti-aliasing tool (part of syngo.MR Cardiac Flow) can be usedto correct for inaccuracy in the VENC values and alleviate the resulting aliasing artifacts. Do not misuse this tooland intentionally set VENC values at a lower than optimal rate as the VENC anti-aliasing function will not workin 100% of the cases.
18 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingAcceleration of Velocity EncodedPhase Contrast MR. New Techniquesand ApplicationsJennifer A Steeden; Vivek MuthuranguVascular Imaging and Physics Group, UCL Centre for Cardiovascular Imaging, London, UKMagnitudePhaseFlowCurveMPAAorta RPA LPA11 Rapid Cardiac-Gated PCMR achieved using undersampled spiral trajectories, in a child, age 2 years. Flow curves show good agreement with a con-ventional Cartesian acquisition throughout the entire cardiac cycle (blue: retrospectively-gated Cartesian PCMR acquisition, red: prospectively-gatedundersampled spiral acquisition).
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 19Cardiovascular Imaging ClinicalIntroductionThe cardiovascular system is character-ized by motion, whether that be of theblood, the myocardium or the valves.For this reason, cardiovascular disease isoften associated with abnormalities inflow and movement. Thus, assessmentof motion is a vital part of the cardiovas-cular work-up. One of the most impor-tant methods of assessing motion is tomeasure the velocity of a moving struc-ture and this is an area in which cardio-vascular magnetic resonance (CMR) hasan important role to play. In fact, CMRhas become the gold standard methodof assessing volume blood flow in condi-tions such as congenital heart disease.However, velocity encoded CMR tech-niques are slow and this has limitedtheir uptake in some clinical situations.In this review we will address novelacceleration techniques that are open-ing up new areas for CMR velocity quan-tification.Basic CMR velocity encodingThe CMR technique that is most com-monly used to measure velocity is phasecontrast MR (PCMR). PCMR is achievedthrough the addition of a bipolar velo-city-encoding gradient to a standardspoiled gradient echo sequence. Thisbipolar gradient induces an additionalphase in moving objects, which isdirectly proportional to the velocity.Thus, the velocity of an object (i.e.blood) can be calculated using the phaseimage. Once the velocities are known,it is trivial to estimate volume flow in aregion of interest. This technique hasbeen widely validated both in-vivo andin-vitro and is considered the referencestandard method of measuring volumeflow. Thus, it has become heavily usedin the assessment of cardiac output,cardiovascular shunts and valvar regur-gitation. However, PCMR is intrinsicallyslow because each line in k-space mustbe acquired twice (with different velo-city-encodings) in order to performbackground phase subtraction. Thisprolongs the acquisition time and canbecome extremely problematic inpatients in whom several flow mapsmust be acquired (i.e. congenital heartdisease). Consequently, there has beena significant push to accelerate PCMRsequences.Acceleration techniquesTo accelerate CMR, it is necessary to fillk-space more quickly. As we havealready reached the physical limits of MRgradient systems, this can only beaccomplished through undersampling ofk-space or more efficient filling. Undersampling in k-space leads to fold-overartifacts, which makes the data unus-able if the artifact overlays the region-of-interest. Parallel imaging techniques(e.g. SENSE or GRAPPA) can be used toreconstruct artifact free images fromundersampled data, through the use ofindependent coil sensitivities. These par-allel imaging techniques can signifi-cantly speed up 2D PCMR (typically up to2 times for Cartesian acquisitions. Dataundersampling is currently used by mostCMR units to speed up the acquisition offlow data. However, the amount ofachievable acceleration is limited, lead-ing to studies into alternative techniquesfor reconstructing artifact free images.The most obvious examples are tempo-ral encoding techniques (e.g. k-t BLAST)that can be used alone or in conjunctionwith parallel imaging techniques (e.g.UNFOLD-SENSE). These techniques offerhigher acceleration factors, althoughoften at the cost of some temporal blur-ring. Currently, these techniques havenot become clinically mainstream,however they should further improveworkflow when they do.An additional method of acceleratingPCMR is more efficient filling of k-space,using a non-Cartesian trajectory. Exam-ples are EPI and spiral k-space filling,both of which can significantly speed upacquisition. Furthermore, they can becombined with the previously men-tioned techniques to achieve high levelsof acceleration. In the next section wewill discuss specific uses of accelerationin PCMR.Rapid cardiac-gated PCMRPCMR is heavily used in the assessmentof conditions such as congenital heartdisease, where between 4 to 8 separateflow measurements are often per-formed. However, particularly in chil-dren1, the accuracy of PCMR is depen-dent on spatial and temporal resolution.Thus, in this population it is desirable toperform high spatio-temporal resolutionflow imaging. Unfortunately becausePCMR is intrinsically slow it is difficult toacquire this sort of data in a breath-hold.Hence, cardiac-gated PCMR often relieson multiple signal averages to compen-sate for respiratory motion, resulting inscan times of approximately 2 minutes.Therefore, complete flow assessmentcan take around 10 minutes. Thus, inthis population there is a need for a highspatio-temporal resolution gated PCMRsequence that can be performed withina short breath-hold.Previous studies have investigatedspeeding up Cartesian gated PCMR withthe use of SENSE parallel imaging. Beer-baum, et al. [1, 2] showed that undersampling by factors of 2 or 3 did nothamper their ability to measure strokevolumes and pulmonary-to-systemicflow ratios. Lew, et al.  were able toshow that breath-hold PCMR (acquisitiontime: 7-10 seconds) was possible ifaccelerated by SENSE (x3). Thus, parallelimaging can be combined with PCMR tosignificantly accelerate acquisition.Even greater acceleration is possible ifparallel imaging is combined with effi-cient k-space trajectories. Our group hasshown that combining efficient spiralk-space filling with SENSE (x3) allowshigh temporal-spatial resolution PCMRdata (Fig. 1) to be acquired in a shortbreath-hold (6 RR-intervals) . Further-
20 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular Imagingmore, we have shown in a large groupof children1and adults with congenitalheart disease that there were no signifi-cant differences compared to a free-breathing Cartesian sequence.Such sequences allow a reduction in thetotal duration of flow imaging in con-genital heart disease from ~10 minutes,to less than 1 minute. This would leadto a marked reduction in total scan timeand has implications for patient through-put and compliance for congenitalcardiac MR scanning.Another approach that has recentlybecome popular is temporal encoding.Baltes, et al.  have shown that Carte-sian PCMR accelerated with k-t BLAST ork-t SENSE can lead to significant reduc-tions in scan time. In the future, thesetemporal encoding techniques could becombined with efficient k-space fillingstrategies to produce very high accelera-tion factors.Real-time PCMRGated PCMR is the mainstay of clinicalflow imaging. However, there are sev-eral situations where gated imaging isnot suitable, for example, during exer-cise. Exercise is a powerful stimulator ofthe cardiovascular system and can beused to unmask subtle disease. How-ever, real-time PCMR is necessary tomeasure flow during exercise. This canalso be achieved through data undersampling and efficient k-space filling.For example, Hjortdal, et al.  havedemonstrated the use of efficient EPItrajectories, with partial-Fourier acquisi-tion, to achieve real-time imaging forthe investigation of flow during exercise.This sequence was used to successfullyassess exercise hemodynamics in com-plex congenital heart disease.Our group has investigated the use ofefficient spiral trajectories combinedwith SENSE (x4) to measure flow at restand during continuous exercise . Thesequence had a high temporal resolu-tion and was validated against a stan-dard gated Cartesian PCMR sequence at2B2C2A2 Real-time PCMR used for quantification of flow during exercise. (2A) MR-compatible ergometer used within the scanner (MR cardiac ergometer Up/Down,Lode, Groningen, Netherlands). (2B) Example of image quality achieved usingundersampled spiral trajectories in the ascending aorta, left: magnitude, right:phase. (2C) Real-time flow curves achieved using this technique.
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 21Cardiovascular Imaging Clinicalrest, with good agreement (Fig. 2).Using this sequence it was possible tocomprehensively assess the responseto exercise both in volunteers andpatients .Of course, one major problem with thisapproach is the long reconstructiontimes associated with non-Cartesian par-allel imaging. For example, our grouphas investigated continuous flow assess-ment over ~10 minutes during exercise. Using the scanners CPU-basedreconstruction, images would only beavailable ~1 hour after the end of dataacquisition. This is too long to be used ina clinical environment; therefore it wasnecessary to develop a faster reconstruc-tion. One method to achieve this was touse graphical processing unit (GPU)-based reconstruction. This methodologyhas previously been shown to signifi-cantly speed up complex MR reconstruc-tions . In our implementation, aseparate GPU-based reconstructor waslinked into the scanners reconstructionpipeline, allowing online reconstructionthat is hidden from the end-user. Usingthis technique, images were available~10 seconds after the end of data acqui-sition. This type of highly acceleratedacquisition and real-time reconstructionmay open up many novel areas incardiovascular MR that are currentlyimpeded by long reconstruction times.Fourier velocity encodingReal-time PCMR requires significantacceleration because each frame mustbe acquired quickly. However, accelera-tion is also required if large data sets areto be acquired in a short period of time.One good example is Fourier velocityyxyxyxMaxvelocity3 Fourier Velocity Encoding using an undersampled spiral acquisition, in one patient. (3A) Acquired data in x-y, MIP’d through v. (3B) Velocityspectrum along x, at time of peak flow. (3C) Velocity profile through time as measured from the spiral MR acquisition. (3D) Velocity profile in thesame patient as measured using Doppler Ultrasound.3A 3B3D3C
22 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular Imaging4 4D PCMR data acquired in-vivo using an undersampled spiral acquisition.encoding (FVE), which is a 3D k-spaceacquisition with spatial encoding in twodimensions (x and y) and velocity encod-ing in the z direction (denoted as kv).Multiple kv positions are acquired usingbipolar flow-encoding gradients withdifferent first-order moments. InverseFourier transformation of k-space pro-duces a 3D image with each point in x-yspace associated with a spectrum ofvelocities in v. This spectrum of velo-cities allows accurate assessment ofstenotic flow, which is often underesti-mated by traditional PCMR. The mainproblem with FVE is that it is timeconsuming to acquire and thus rarelyused in the clinical setting.Time-efficient spiral trajectories withpartial-Fourier acquisition along thevelocity-encoding dimension have beenused to speed up FVE . This formof acceleration does allow breath-holdFVE to be achieved (of 7 RR-intervals),although with a low spatial resolution(of 7 mm). Other groups have investi-gated the use of k-t SENSE for CartesianFVE  with higher acceleration fac-tors (x8 and x16) allowing better spatialresolution (of 1.3-2.8 mm, within abreath-hold of 15-20 seconds).Our solution to achieve high resolutionFVE was to combine spiral trajectorieswith SENSE in kx-ky (x4), in addition topartial-Fourier acquisition in kv (67%)and velocity unwrap in the v dimension(Fig. 3). The result of all these differentacceleration techniques is that it is pos-sible to achieve high spatial (~2.3 mm),temporal (~41 ms) and velocity resolu-tion (14-25 cm/s) FVE data in a relativelyshort breath-hold (of 15 RR-inter-vals) .This FVE technique was shown to pro-vide more accurate peak velocity mea-sures than PCMR in-vitro and in-vivo,compared to Doppler US. Thus, by usingacceleration techniques it was possibleto bring a technique that has beenavailable for some time, into the clinicalenvironment.4D FlowAnother technique that has been avail-able for some time, but is rarely usedclinically, is time-resolved 3-dimensionalPCMR imaging. This technique acquiresflow in three encoding directions(known as 4D PCMR) and allows quanti-fication of flow in any imaging plane, inaddition to the visualization of complexflow patterns. However, 4D PCMR israrely used in the clinical setting due tolong acquisition times, often in theorder of 10-20 minutes.Several acceleration approaches havebeen investigated in 4D PCMR. Forinstance, SENSE and k-t BLAST can accel-erate the acquisition of Cartesian 4DPCMR at 1.5T and 3T . Using suchtechniques, it is possible to acquire high-resolution data in ~10 minutes, givingreasonable agreement in terms of stroke4
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 23Cardiovascular Imaging ClinicalReferences1 Beerbaum P, Korperich H, Gieseke J, Barth P,Peuster M, Meyer H. Rapid Left-to-Right ShuntQuantification in Children by Phase-ContrastMagnetic Resonance Imaging Combined WithSensitivity Encoding (SENSE). Circ2003;108(11):1355-1361.2 Beerbaum P, Körperich H, Gieseke J, Barth P,Peuster M, Meyer H. Blood flow quantification inadults by phase-contrast MRI combined withSENSE--a validation study. Journal of cardiovas-cular magnetic resonance 2005;7(2):361-369.3 Lew CD, Alley MT, Bammer R, Spielman DM,Chan FP. Peak Velocity and Flow QuantificationValidation for Sensitivity-Encoded Phase-Contrast MR Imaging. Academic Radiology2007;14(3):258-269.4 Steeden JA, Atkinson D, Hansen MS, Taylor AM,Muthurangu V. Rapid Flow Assessment ofCongenital Heart Disease with High-Spatiotem-poral-Resolution Gated Spiral Phase-Contrast MRImaging. Radiology 2011;260(1):79-87.5 Baltes C, Kozerke S, Hansen MS, Pruessmann KP,Tsao J, Boesiger P. Accelerating cine phase-contrast flow measurements using k-t BLASTand k-t SENSE. MRM 2005;54(6):1430-1438.6 Hjortdal VE, Emmertsen K, Stenbog E, Frund T,Schmidt MR, Kromann O, Sorensen K, PedersenEM. Effects of Exercise and Respiration onBlood Flow in Total Cavopulmonary Connection:A Real-Time Magnetic Resonance Flow Study.Circ 2003;108(10):1227-1231.13 Steeden JA, Jones A, Pandya B, Atkinson D,Taylor AM, Muthurangu V. High-resolution slice-selective Fourier velocity encoding in congenitalheart disease using spiral SENSE with velocityunwrap. MRM 2012;67(6):1538-1546.14 Carlsson M, Toger J, Kanski M, Bloch K, StahlbergF, Heiberg E, Arheden H. Quantification andvisualization of cardiovascular 4D velocity map-ping accelerated with parallel imaging or k-tBLAST: head to head comparison and validationat 1.5 T and 3 T. Journal of Cardiovascular Mag-netic Resonance 2011;13(1):55.15 Tariq U, Hsiao A, Alley M, Zhang T, Lustig M,Vasanawala SS. Venous and arterial flow quanti-fication are equally accurate and precise withparallel imaging compressed sensing 4D phasecontrast MRI. Journal of Magnetic ResonanceImaging 2012:n/a-n/a.7 Steeden JA, Atkinson D, Taylor AM, MuthuranguV. Assessing vascular response to exercise usinga combination of real-time spiral phase contrastMR and noninvasive blood pressure measure-ments. Journal of Magnetic Resonance Imaging2010;31(4):997-1003.8 McKee A, Davies J, Bilton D, Taylor A, GatzoulisM, Aurora P, Muthurangu V, Balfour-Lynn I.A Novel Exercise Test Demonstrates ReducedStroke Volume Augmentation But No EvidenceOf Pulmonary Hypertension In Patients WithAdvanced Cystic Fibrosis. Pediatr Pulm 2011:354.9 Kowalik GT, Steeden JA, Pandya B, Odille F,Atkinson D, Taylor A, Muthurangu V. Real-timeflow with fast GPU reconstruction for continuousassessment of cardiac output. Journal ofMagnetic Resonance Imaging 2012;36(6):1477-1482.10 Hansen MS, Atkinson D, Sorensen TS. CartesianSENSE and k-t SENSE reconstruction usingcommodity graphics hardware. MRM2008;59(3):463-468.11 Carvalho JLA, Nayak KS. Rapid quantitation ofcardiovascular flow using slice-selective fouriervelocity encoding with spiral readouts. MRM2007;57(4):639-646.12 Baltes C, Hansen MS, Tsao J, Kozerke S, Rezavi R,Pedersen EM, Boesiger P. Determination of PeakVelocity in Stenotic Areas: Echocardiographyversus kt SENSE Accelerated MR Fourier VelocityEncoding. Radiology 2007;246(1):249.ContactJennifer SteedenVascular imaging and physics groupUniversity College London (UCL)Centre for Cardiovascular ImagingLondonUnited Kingdomj.email@example.com with gated 2D PCMR. Anotherpossibility is the use of combined paral-lel imaging and compressed-sensing toaccelerate the acquisition of 4D PCMR. K-space data can be acquiredwith variable-density Poisson disc under-sampling with a total acceleration of4–5, giving an acquisition time of ~11minutes.We have taken a similar approach to FVEand used a stack-of-spirals acquisitionin addition to data undersampling, andreconstructing with SENSE to greatlyreduce the acquisition time for 4D PCMR.For example, a developed spiral SENSEsequence with 8 uniform density spiralinterleaves in kx-ky, and 24 slices, canbe acquired with a SENSE undersam-pling factor of 4 in kx-ky, and SENSEundersampling factor of 2 in kz, giving atotal undersampling factor of 8. Thisallows a temporal resolution of ~49 msand a spatial resolution of 2.6 × 2.6 × 2.5 mm, to be achieved within a scantime 2 minutes 40 seconds (Fig. 4).In this case, accelerated imaging canoffer huge reductions in scan time.Conclusion and futureAccelerated techniques offer the possi-bility of significantly speeding up PCMRacquisitions. This should allow betterreal-time imaging, as well as the acquisi-tion of larger data sets in much shortertimes (e.g. 4D PCMR). In the future newacceleration techniques such as com-pressed sensing may further improveour ability to acquire this sort of data.However, as the acceleration techniquesbecome more complex, faster recon-struction will become necessary. Thus,the advent of GPU based MR reconstruc-tion has a pivotal role to play in thedevelopment of new accelerated PCMRsequences.1MR scanning has not been established as safe forimaging fetuses and infants under two years of age.The responsible physician must evaluate the benefit ofthe MRI examination in comparison to other imagingprocedures.
24 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingIntroductionPhase-contrast magnetic resonanceimaging (PC-MRI) is used routinely as animportant diagnostic tool to measureblood flow and peak velocities in majorblood vessels and across heart valves.Conventional PC-MRI commonly usesECG synchronization and segmentedk-space acquisition strategies [1, 2]; thisapproach requires breath-holding andregular cardiac rhythm, rendering itimpractical in a large number of patients.Alternatively, signal averaging and respi-ratory gating approaches can be usedto reduce respiratory motion artifacts inpatients unable to breath-hold, butthese methods still require a reliable ECGsignal and regular heart beats. Further-more, the velocity information resultingfrom an ECG-synchronized segmentedPC sequence represents a weighted tem-poral average of the velocity waveformover the acquisition time; short-livedhemodynamic effects, such as responseto pharmacological stress, physical exer-cise or respiration and interventionalmaneuvers cannot be assessed. Hence,free-breathing real-time (RT) PC-MRIwith sufficient spatial/temporal resolu-tion would be desirable to provide beat-to-beat hemodynamic information tocharacterize transient blood flow phe-nomena and as an alternative approachfor patients with irregular cardiacrhythm or inability to hold their breath.Free-Breathing Real-Time Flow Imagingusing EPI and Shared Velocity EncodingNing Jin1; Orlando Simonetti, Ph.D.21Siemens Healthcare, Malvern, PA, USA2The Ohio State University, Columbus, OH, USAMethodsRT PC-MRI sequence was implementedusing a gradient-echo echo-planar imag-ing (GRE-EPI) sampling trajectory withan echo train length (ETL) of 7 to 15echoes. A rapid phase modulated bino-mial spatial and spectral selective waterexcitation pulse was used to eliminateghosting artifacts from fat tissue. Tem-poral parallel acquisition technique(TPAT) with an acceleration factor of 3was used by temporally interleavingundersampled k-space lines [3, 4]. Fourshots per image were acquired resultingin an acquisition time about 40 to 50 ms(depending on ETL and matrix size) foreach full k-space dataset per flow encod-ing step. To further improve the effec-tive temporal resolution of RT-PC,a shared velocity encoding (SVE) scheme was implemented.In PC-MRI, a critical step is the elimina-tion of background phase errors. This isachieved by subtracting two measure-ments with different flow encodings. Itis typically performed using one of twomethods: one-sided velocity encoding inwhich velocity compensated (k0) andvelocity encoded (ke) data are acquiredfor each cardiac phase, or two-sidedvelocity encoding in which equal andopposite velocity encodings (k+/k-) areacquired. Both cases result in a reduc-1 (1A) The conventional PC with 1-sided velocity encoding. Velocity information wasextracted using the flow encoded (ke) and flow compensated (k0) data within the same cardiacphase. (1B) SVE with 2-sided velocity encoding. Additional velocity frames (V2 and V4) arereconstructed by sharing the flow encoded data across cardiac phases to double the effectiveframe rate by a factor of 2.CardiacPhase1 2 31A1 2 31Bke k+ k- k+ k- k+ k-k0V1 V1 V3 V5V2 V2 V4V5ke k0 ke k0
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 25Cardiovascular Imaging Clinicaltion in temporal resolution or an increasein scan time by a factor of two. The SVEconcept is to share flow encoded databetween two adjacent frames in a two-sided velocity encoded PC-MRI. Figure 1illustrates the difference between con-ventional PC reconstruction withone-sided velocity encoding and SVEreconstruction with two-sided velocityencoding. The odd-number velocityframes (V1, V3 and V5) are identical tothe frames generated by the conven-tional PC reconstruction, while SVEreconstruction results in the intermediateeven frames 2 and 4. While each ofthese additional frames shares velocitydata with the adjacent frames, eachcontains a unique set of data that repre-sents a velocity measurement centeredat the time between the original frames.As the result, SVE increases the effectivetemporal resolution by a factor of two.While the SVE method doubles the effec-tive frame rate; the temporal window ofeach frame is unchanged. In the RT-PCworks-in-progress (WIP) sequence, theeffective, reconstructed temporal resolu-tion of each frame is 40 to 50 ms whilethe temporal window to acquire all flowencoding steps is 80 to 90 ms.ResultsFigure 2 shows the representative mag-nitude (Fig. 2A) and phase (Fig. 2B)images and the velocity waveforms ofthe aortic valve in one normal volunteeracquired using ECG-triggered segmentedPC-MRI in breath-hold (Figs. 2C and 2E)and RT PC-MRI in free-breathing (Figs.2D and 2F). The results from ECG-trig-gered segmented PC-MRI are used as thereference standard. The mean velocity2 Example of magnitude (2A)and phase (2B) images acquiredusing RT PC-MRI of the aorticvalve in a normal volunteershowing typical image quality.Mean velocity (2C, 2D) and peakvelocity (2E, 2F) waveforms fromthe same volunteer acquiredusing ECG-triggered segmentedPC-MRI as gold standard andfree-breathing RT PC-MRIMeanVelocityVelocity(cm/s)120100806040200-20 2000 400 600 800 1000Time (ms)100140120100806040200-20806040200-202C2A 2BPeakVelocitymax = 117.83 cm/sVelocity(cm/s)1401201008060402002000 400 600 800 1000Time (ms)2EVelocity(cm/s)Velocity(cm/s)1000 2000 3000 4000 5000 60000Time (ms)1000 2000 3000 4000 5000 6000Time (ms)max = 116.59 ± 1.15 cm/s2D2F
26 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular Imaging26 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldContactOrlando “Lon” Simonetti, Ph.D.Professor of Internal Medicine andRadiologyThe Ohio State University460 W. 12thAve., BRT316Columbus, OH 43210Phone: +1 (614) 293-0739Fax: +1 (614) 247-8277Orlando.Simonetti@osumc.eduReferences1 Chatzimavroudis GP et al. Clinical blood flow quantification with segmented k-space magneticresonance phase velocity mapping. J Magn Reson Imaging 2003;17:65–71.2 Edelman RR et al. Flow velocity quantification in human coronary arteries with fast, breath-holdMR angiography. J Magn Reson Imaging 1993;3:699–703.3 Kellman P et al. Adaptive Sensitivity Encoding Incorporating Temporal Filtering (TSENSE).Magn Reson Med. 45: 846-852 (2001).4 Breuer FA et al. Dynamic autocalibrated parallel imaging using temporal GRAPPA (TGRAPPA).Magn Reson Med. 53(4): 981-985 (2005).5 Lin HY et al. Shared velocity encoding: a method to improve the temporal resolution ofphase-contrast velocity measurements, Magn Reson Med 2012;68:703–7106 Thavendiranathan et al., Simultaneous right and left heart real-time, free-breathing CMR flowquantification identifies constrictive physiology. JACC Cardiovasc Imaging. 2012 Jan;5(1):15-24.curve (Fig. 2D) and the peak velocitycurve (Fig. 2F) acquired using RT PC-MRImatch well with those of the referencestandard (Figs. 2C and 2E). The maxi-mum peak velocity from RT-MRI is116.59 ± 1.15 cm/s averaged over sixheart beats, which agrees well with thereference standard (117.83 cm/s).Figure 3 shows the magnitude and phaseimages of the aortic valve acquired withRT PC-MRI in one patient with bicuspidaortic valve and moderate aortic steno-sis. RT PC-MRI measured a peak velocityof 330 cm/s and planimetered valvearea of 1.2 cm2.ConclusionWe have described a free-breathingRT PC-MRI technique that employs aGRE-EPI sampling trajectory, TPAT accel-eration and SVE velocity encodingscheme. It is a promising approach forrapid and RT flow imaging and measure-ments. The technique has been shownto provide accurate blood flow measure-ments with sufficient SNR and spatial/temporal resolution . With RT EPIacquisition, the adverse effects of car-diac and respiratory motion on flowmeasurements are minimized. Thus, RTPC-MRI can be used on patients suffer-ing from arrhythmia, and in pediatric orother patients incapable of breath-hold-ing. Furthermore, RT PC-MRI providestransient hemodynamic information,which could be used to evaluate theeffects of respiration, exercise, or otherphysical maneuvers on blood flow .3 Magnitude (3A) and phase (3B) images acquired using RT PC-MRI of the aortic valve in one patient with bicuspid aortic valve andmoderate aortic stenosis.3A 3B
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 27Cardiovascular Imaging ClinicalHigh Acceleration Quiescent-IntervalSingle Shot Magnetic ResonanceAngiography at 1.5 and 3TMaria Carr R.T.R (CT)(MR)1; Christopher Glielmi, Ph.D.2; Robert R. Edelman, M.D.3; Michael Markl, Ph.D.1;James Carr, M.D.1; Jeremy Collins, M.D.11Northwestern Memorial Hospital and Northwestern University Feinberg School of Medicine, Chicago, IL, USA2Cardiovascular MR RD, Siemens Healthcare, Chicago, IL, USA3NorthShore University HealthSystem, Evanston, IL, USA1 QISS maintains excellent image quality with PAT 4 acquisition. 3T imaging demonstrates higher SNR and potentially better results at high PATfactors. All protocols have high correlation to CE-MRA (right).Reprinted from MAGNETOM Flash 1/2012.Healthy volunteer: Comparison across field strength and PAT factor1.5T PAT 2 1.5T PAT 4 3T PAT 2 3T PAT 4 CE-MRA 3T1
28 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular ImagingBackgroundContrast-enhanced MR angiography (CE-MRA) is routinely used for the evaluationof peripheral arterial disease (PAD). How-ever, due to the increased concern ofnephrogenic systemic fibrosis (NSF) asso-ciated with gadolinium administration inpatients with impaired renal function,there has been increased interest in thedevelopment of non-contrast MRA (NC-MRA) techniques. Quiescent-intervalsingle-shot (QISS) imaging* has recentlybeen described as an NC-MRA techniquefor assessment of the lower extremityvasculature  with proven clinical util-ity at 1.5T [2, 3]. These studies havedemonstrated that QISS is easy to use,does not require patient-specific imagingparameters, has minimal flow dependenceand is less sensitive to motion than sub-tractive techniques. Increased SNR at 3Tpotentially provides improved imagequality and enable higher parallel imag-ing acceleration (PAT) factors. Increasingthe PAT factor can be critical to maintain-ing an acquisition rate of one slice perheartbeat in patients with fast heart rates.In order to evaluate these aspects ofQISS NC-MRA, the purpose of our studywas to determine1. the effect of field strength on QISSimage quality,2. the potential to accelerate imaging at3T using higher PAT factors and3. to compare NC-MRA QISS at 3T to aconventional CE-MRA protocol.In addition, the diagnostic quality ofQISS at 3T in patients with PAD wasevaluated.2 3T QISS in a patient shows segmental occlusion in the adductor canal with recon-stitution in the profunda femoral collateral for patient with low GFR (no CE-MRAreference). Standard PAT factor 2 (left), as well as higher PAT factors 3 (center) and4 (right) clearly depict disease, enabling shorter single shot acquisition capabilitiesfor patients with fast heart rates.2Reprinted from MAGNETOM Flash 1/2012.
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 29Cardiovascular Imaging ClinicalMethodsQISS is a two-dimensional electrocardio-graphically gated single-shot balancedsteady-state free precession acquisitionof one axial slice per heartbeat. Imagecontrast is generated using in-plane sat-uration to suppress background tissueand a tracking saturation pulse to sup-press venous signal prior to a quiescentinflow period. ECG gating ensures thatthe quiescent inflow period coincides withrapid systolic flow to maximize inflow ofunsaturated spins into the imaging slice;since only a 2D slice is saturated, onlyminimal inflow is required. Images areacquired with a series of imaging stations,each with 70 slices around the magnetisocenter. Stations in the pelvis andabdomen typically use a series of breath-hold concatenations to minimize respira-tory motion.To validate QISS at 3T, twelve healthyvolunteers (9 males age 22–49, 3females age 26–43) were scanned at 1.5and 3T within one week using a dedi-cated 36-channel peripheral vascular coiland an 18-channel body array coil (1.5TMAGNETOM Aera and 3T MAGNETOMSkyra, Siemens Healthcare, Erlangen,Germany). In addition, CE-MRA was per-formed at 3T (8.4–10 ml at 2 ml/sec,Ablavar, Lantheus Medical Imaging, N.Billerica, MA, USA). Additionally, 3patients with lower extremity PAD werescanned using 3T QISS NC-MRA and CE-MRA as the reference standard if glomer-ular ﬁltration rate (GFR) was sufficient.Imaging parameters are provided inTable 1. For each scan session, threeQISS run-off scans were acquired in ran-domized order with GRAPPA PAT factors2, 3 and 4. The total acquisition andshim time for each QISS NC-MRA run-offwas approximately 10 minutes depend-ing on the heart rate. Two blinded radi-ologists scored image quality for venouscontamination, arterial conspicuity andarterial artifacts using a 4-point scale (0= poor, 1 = fair, 2 = good, 3 = excellent)and inter-observer reliability was measuredusing the kappa statistic.Validation resultsQISS image quality was consistent forboth field strengths. 3T results were par-ticularly robust, even with high PAT fac-tors and showed higher signal-to-noiseratio (SNR) relative to 1.5T (representa-tive volunteer in Fig. 1). Arterial conspi-cuity and artifact scores were compara-ble to CE-MRA, and slightly higher for 3T(2.8 ± 0.1) relative to 1.5T (2.5 ± 0.2).Venous suppression was superior at1.5T; venous signal did not impact arte-rial assessment at 3T, however. QISS at3T also had higher inter-observer agree-ment (κ = 0.759) than at 1.5T (κ = 0.426)for image quality scoring. Overall, QISSperformed comparably to CE-MRA atboth field strengths, even with a highPAT factor of 4 . Despite the presenceof respiratory or bowel motion in somevolunteers, QISS with concatenatedbreath-holding performed consistentlyin the abdomen and pelvis. Moreover,one always has the option to repeat allor part of the data acquisition in case oftechnical difficulty or macroscopicpatient motion - an option that is notafforded by CE-MRA.Clinical resultsBased on impressive results in volunteers,several patients were scanned at 3T todemonstrate clinical utility. Initial resultssuggest that QISS is effective at 3T andhas high correlation with CE-MRA.Clinical case 142-year-old male with longstanding his-tory of type I diabetes complicated byrenal failure presents with left diabeticfoot requiring transmetatarsal amputa-tion. He subsequently received a kidneytransplant and now presents to vascularsurgery clinic with non-healing woundsof the right 3rdand 4thtoes. The patient’srenal function is poor, with a Creatinineof 2.35 mg/dl and an eGFR of 31 ml/min/1.73 m2. The patient was referredfor QISS NC-MRA at 3T to assess optionsfor right lower extremity revascularization(Fig. 2). NC-MRA demonstrated segmen-tal occlusion of the distal left superficialfemoral artery at the adductor canal overa distance of approximately 3 cm withreconstitution via branches of the pro-funda femoris artery (Fig. 2). Small col-lateral vessels are clearly depicted forPAT factors 2, 3, and 4 (suitable for 738,660 and 619 ms R-R intervals, respec-tively) suggesting that high PAT factorscan be used to maintain acquisition ofone slice per heartbeat for patients withhigh heart rates. The QISS NC-MRA imagesof the calves clearly depict the arterialrun-off beyond the level of occlusion. Nosignificant arterial stenoses were appar-ent in the right lower extremity and thepatient was diagnosed with microvascu-lar disease. The patient continued woundcare, with plans for transmetatarsalamputation if conservative therapy wasunsuccessful.Clinical case 264-year-old male with a history of hyper-tension, chronic back pain, and tobaccoabuse presents with diminished exercisetolerance. The patient reports that thesesymptoms have been slowly progressiveover two years such that he is only ableto walk half a block, limited by fatigueand left leg weakness of the hip and but-tock. Lower extremity arterial flow stud-ies demonstrated an ankle brachial indexof 0.69 with duplex findings suggestiveof inflow disease. The patient was referredfor CE-MRA at 1.5T and underwent QISSNC-MRA at 3T for research purposes(Fig. 3). CE-MRA with a single dose ofgadobenate dimeglumine (Multihance,Bracco Diagnostics Inc., Princeton, NJ,USA) demonstrated long segment occlu-sion of the left external iliac artery, withreconstitution of the common femoralartery. There was no significant left lowerextremity outflow disease; three-vesselrunoff was noted to the left foot. QISSNC-MRA demonstrated similar findings,with segmental occlusion of the externaliliac artery, with the common femoralartery reconstituted by the inferior epi-Reprinted from MAGNETOM Flash 1/2012.
30 MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-worldClinical Cardiovascular Imaginggastric and deep circumflex iliac arteries.The patient will undergo revasculariza-tion with left iliac stent placement at alater date.Clinical case 356-year-old female with a history ofhypertension, type II diabetes, and lum-ber radiculopathy presents with inter-mittent left lower extremity claudicationgradually increasing in severity over thecourse of a year, now limiting her walk-ing to a half block. The patient describesincreased sensitivity to her left foot andpain in the anterior left thigh, which keepsher awake at night. She notes improvedpain with Vicodin. The patient was referredfor CE-MRA and underwent QISS NC-MRA at 3T for research purposes (Fig. 4).CE-MRA was performed at 1.5T followinga double-dose of gadopentetate dimeglumine (Magnevist, Bayer HealthcarePharmaceuticals, Inc, Wayne, NJ, USA)demonstrating long segment occlusionof the left mid to distal superficial femo-ral artery, with reconstitution of thepopliteal artery via branches of thehypertrophied profunda femoris artery.These findings are well seen using QISSNC-MRA, which delineates the lengthof the occlusion and the run-off vesselswithout contrast media. Mild venoussignal contamination is noted on QISS3 Both QISS and CE-MRA depict long segment occlusion of the leftexternal iliac artery, with reconstitution of the common femoral artery.QISS PAT 3 acquisition (left) maximizes SNR while still maintainingsingle shot acquisition every heartbeat.34 QISS and CE-MRA demonstrate long segment occlusion of theleft mid to distal superficial femoral artery, with reconstitution ofthe popliteal artery via branches of the hypertrophied profundafemoris artery. For this patient, QISS PAT 4 acquisition (left) main-tained single shot acquisition every heart beat.4Reprinted from MAGNETOM Flash 1/2012.
MAGNETOM Flash · SCMR 2013 · www.siemens.com/magnetom-world 31Cardiovascular Imaging ClinicalNC-MRA without impacting diagnosticyield of the maximum intensity projec-tion (MIP) images. The patient under-went successful endovascular recanali-zation of the left superficial femoralartery with stent placement with resolu-tion of her claudication symptoms.ConclusionQISS NC-MRA is promising at 3T anddemonstrates comparable image qualityto CE-MRA in depicting the peripheralarteries in both volunteers and patientswith PAD. Improved SNR at 3T enableshigher PAT factors to reduce image acqui-sition time while maintaining imagequality in patients with faster heart rates.While QISS at 1.5T has shown excellentclinical utility, our results suggest thatimaging at 3T may provide further clinicalbenefit. QISS NC-MRA at 3T demonstratedarterial vessels with slightly greater con-spicuity and lower arterial artifacts com-pared to QISS NC-MRA at 1.5T. Althoughvenous signal contamination is greaterat 3T, the increased arterial signal enablesready differentiation on MIP imaging anddoes not impact diagnostic utility. Volun-teer and patient data demonstrate thatthe QISS acquisition can be acceleratedup to four fold without significant imagequality degradation, suitable for heartrates approaching 100 beats per minute.At both field strengths, the QISS techniqueis a valuable alternative to CE-MRA espe-cially in patients with renal insufficiency.Table 1: Quiescent-interval single-shot (QISS) sequence details.QISS sequence detailsTR (ms) / TE (ms) / flip angle (deg.) 3.2/1.7/90In-plane resolution 1 x 1 mmPartial Fourier 5/8Slice thickness 3 mmOrientation axialNo. of slices per group 70No. of slice groups 7Acquisition time / slice group ~55 s (heart rate dependent)Parallel imaging GRAPPA x 2–4ECG gating YesInversion time 345 msContactMaria Carr, RT (CT)(MR)CV Research TechnologistDepartment of RadiologyNorthwestern UniversityFeinberg School of Medicine737 N. Michigan Ave. Suite 1600Chicago, IL 60611USAPhone: +1 firstname.lastname@example.org References1 Edelman, R.R., et al., Quiescent-interval single-shot unenhanced magnetic resonance angiographyof peripheral vascular disease: Technical consid-erations and clinical feasibility. Magn Reson Med,2010. 63(4): p. 951–8.2 Hodnett, P.A., et al., Evaluation of peripheral arterialdisease with nonenhanced quiescent-interval sin-gle-shot MR angiography. Radiology, 2011. 260(1):p. 282–93.3 Hodnett, PA, et al., Peripheral Arterial Disease ina Symptomatic Diabetic Population: ProspectiveComparison of Rapid Unenhanced MR Angiogra-phy (MRA) With Contrast-Enhanced MRA. AJR December 2011 197(6): p. 1466–73.4 Glielmi, C., et al. High Acceleration Quiescent-In-terval Single Shot Magnetic Resonance Angiographyat 1.5 and 3T. ISMRM 2012, 3876.Reprinted from MAGNETOM Flash 1/2012.