compressed sensing single-breath-hold cmr for fast ... · cmr for fast quantification of lv...
TRANSCRIPT
-
J A C C : C A R D I O V A S C U L A R I M A G I N G VO L . 7 , N O . 9 , 2 0 1 4
ª 2 0 1 4 B Y T H E AM E R I C A N C O L L E G E O F C A R D I O L O G Y F O U N D A T I O N I S S N 1 9 3 6 - 8 7 8 X / $ 3 6 . 0 0
P U B L I S H E D B Y E L S E V I E R I N C . h t t p : / / d x . d o i . o r g / 1 0 . 1 0 1 6 / j . j c m g . 2 0 1 4 . 0 4 . 0 1 6
Compressed Sensing Single–Breath-HoldCMR for Fast Quantification ofLV Function, Volumes, and Mass
Gabriella Vincenti, MD,* Pierre Monney, MD,* Jérôme Chaptinel, MSC,yz Tobias Rutz, MD,*Simone Coppo, MSC,yz Michael O. Zenge, PHD,x Michaela Schmidt, RT,x Mariappan S. Nadar, PHD,kDavide Piccini, MSC,yz{ Pascal Chèvre, RT,*# Matthias Stuber, PHD,yz Juerg Schwitter, MD*
ABSTRACT
Fro
Ra
Sw
Vis
He
Sw
sto
rel
Ma
OBJECTIVES The purpose of this study was to compare a novel compressed sensing (CS)–based single–breath-hold
multislice magnetic resonance cine technique with the standard multi–breath-hold technique for the assessment of left
ventricular (LV) volumes and function.
BACKGROUND Cardiac magnetic resonance is generally accepted as the gold standard for LV volume and function
assessment. LV function is 1 of the most important cardiac parameters for diagnosis and the monitoring of treatment
effects. Recently, CS techniques have emerged as a means to accelerate data acquisition.
METHODS The prototype CS cine sequence acquires 3 long-axis and 4 short-axis cine loops in 1 single breath-hold
(temporal/spatial resolution: 30 ms/1.5 � 1.5 mm2; acceleration factor 11.0) to measure left ventricular ejection fraction(LVEFCS) as well as LV volumes and LV mass using LV model–based 4D software. For comparison, a conventional stack of
multi–breath-hold cine images was acquired (temporal/spatial resolution 40 ms/1.2 � 1.6 mm2). As a reference for theleft ventricular stroke volume (LVSV), aortic flow was measured by phase-contrast acquisition.
RESULTS In 94% of the 33 participants (12 volunteers: mean age 33 � 7 years; 21 patients: mean age 63 � 13 yearswith different LV pathologies), the image quality of the CS acquisitions was excellent. LVEFCS and LVEFstandard were similar
(48.5� 15.9% vs. 49.8� 15.8%; p¼ 0.11; r¼ 0.96; slope 0.97; p< 0.00001). Agreement of LVSVCS with aortic flow wassuperior to that of LVSVstandard (overestimation vs. aortic flow: 5.6 � 6.5 ml vs. 16.2 � 11.7 ml, respectively; p ¼ 0.012)with less variability (r ¼ 0.91; p < 0.00001 for the CS technique vs. r ¼ 0.71; p < 0.01 for the standard technique). Theintraobserver and interobserver agreement for all CS parameters was good (slopes 0.93 to 1.06; r ¼ 0.90 to 0.99).
CONCLUSIONS The results demonstrated the feasibility of applying the CS strategy to evaluate LV function
and volumes with high accuracy in patients. The single–breath-hold CS strategy has the potential to replace
the multi–breath-hold standard cardiac magnetic resonance technique. (J Am Coll Cardiol Img 2014;7:882–92)
© 2014 by the American College of Cardiology Foundation.
L eft ventricular ejection fraction (LVEF) is oneof the most important measures in cardiologyand part of nearly every cardiac imaging eval-uation because it is recognized as 1 of the strongest
m the *Division of Cardiology and Cardiac MR Center, University Hospital
diology, University Hospital and University of Lausanne, Lausanne, Swit
itzerland; xMR Product Innovation and Definition, Healthcare Sector, Siemion, Siemens Corporation, Corporate Technology, Princeton, New Jersey
althcare IM BM PI, Lausanne, Switzerland; and the #Department of Rad
itzerland. There was no financial support from Siemens for this clinical s
ckholders in Siemens. Mr. Nadar and Mr. Piccini are employees of Siemen
ationships relevant to the contents of this paper to disclose.
nuscript received December 19, 2013; revised manuscript received April 1
predictors of outcome (1). It is crucial for decisionmaking (e.g., to start or stop specific drug treatments)(2), implantation of devices (3), or assessment of theeffect of experimental procedures (4). Cardiac
of Lausanne, Lausanne, Switzerland; yDepartment ofzerland; zCenter for Biomedical Imaging, Lausanne,ens AG, Erlangen, Germany; kImaging and Computer; {Advanced Clinical Imaging Technology, Siemensiology, University Hospital of Lausanne, Lausanne,
tudy. Dr. Zenge and Ms. Schmidt are employees and
s. All other authors have reported that they have no
4, 2014, accepted April 17, 2014.
http://dx.doi.org/10.1016/j.jcmg.2014.04.016
-
AB BR E V I A T I O N S
AND ACRONYM S
CMR = cardiac magnetic
resonance
CS = compressed sensing
(technique)
LVEDV = left ventricular
end-diastolic volume
LVEF = left ventricular
J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4 Vincenti et al.S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2 Compressed Sensing CMR for Rapid LV Assessment
883
magnetic resonance (CMR) is generally accepted asthe gold standard to yield the most accurate mea-sures of LVEF and left ventricular (LV) volumes andmass. These characteristics, together with the addi-tional value of CMR to characterize pathologicalmyocardial tissue, were the basis for CMR to be rec-ommended for heart failure work-up in U.S. guide-lines (5) and to be assigned a class I indication inthe new European heart failure guidelines (6).
TABLE 1 Imaging Parameters of the Standard Cine Sequence (Standard SSFP),
Prototype CS-Based Cine Sequence, and Phase-Contrast Flow Sequence (Aortic Flow)
Standard SSFP CS Technique Aortic Flow
TR, ms 3.06 2.94 10.5
TE, ms 1.28 1.23 3.04
Field of view, mm 300 � 240 420 � 340 340 � 238Image matrix 256 � 180 272 � 272 192 � 192Spatial resolution, mm 1.2 � 1.2 1.5 � 1.5 1.8 � 1.8Temporal resolution, ms 49 30 42
Slice thickness/gap, mm 8/2 6 6
Flip angle 60� 70� 20�
Bandwidth, Hz/pixel 930 875 491
k lines/segment 16 10 4
Cardiac phases 25* 24† 20*
Breath-holds, n(duration of breathhold, s)
4–6 (12)‡ 1 (14)§ 1 (20)
VENC, cm/s — — 150
*Retrospective electrocardiogram (ECG) triggering. †Prospective ECG triggering. ‡To cover the LV in this studypopulation, 4 to 6 breath-holds (i.e., 8 to 12 short-axis slices) were required. §Seven slices were acquiredin 1 breath-hold to cover the LV.
CS ¼ compressed sensing; LV ¼ left ventricle; SSFP ¼ steady-state free precession; TE ¼ echo time;TR ¼ repetition time; VENC ¼ velocity encoded.
SEE PAGE 893
ejection fraction
LVESV = left ventricular
end-systolic volume
LVSV = left ventricular
stroke volume
The evaluation of LV volumes and LVEF is basedon well-defined and generally-accepted CMR pro-tocols (7,8), and it involves the acquisition of a stackof LV short-axis cine images that are acquired inmultiple breath-holds. Quality criteria (9) for thesefunctional images are available and have beenimplemented (e.g., within the European CMR regis-try) (10). Although standard protocols are wellestablished, it is desirable to speed up CMR acquisi-tions to reduce motion artifacts in severely ill pa-tients who cannot hold their breath for an extendedamount of time, to increase spatial and/or temporalresolution per unit time of acquisition, or simply toshorten the CMR examination. A variety of recentacceleration techniques (11), such as spatiotemporalcorrelations (12–14) and spatially selective excitations(15), could speed up acquisitions several fold. As analternative, compressed sensing (CS) was recentlyproposed as a means to considerably accelerate dataacquisition. Conventional magnetic resonance (MR)images once acquired can be compressed with littleor no perceptible loss of information (e.g., by JPEGstandards). The concept of CS aims to no longer applysuch compression algorithms to the acquired mag-netic resonance image data but to the acquisitionprocess itself and then to reconstruct the under-sampled data with novel nonlinear reconstructionalgorithms (Online Appendix) (16). This helps accel-erate image acquisition several fold and theoreticallyenables a cine acquisition of the whole heart in 1single breath-hold, whereas conventional dataacquisition that covers the whole heart by short axesoften lasts up to 10 min. The aim of the study was todemonstrate that this CS concept: 1) is applicable tothe heart in a clinical setting; 2) produces data toaccurately quantify LV function, volumes, and mass;and 3) provides data quality that allows for accept-able measurement reproducibility.
METHODS
CS TECHNIQUE. The prototype CS sequence devel-oped for this study achieves incoherent sampling by
initially distributing the readouts pseudo-randomly in the Cartesian k-space (17). Inaddition, for cine-CMR, a pseudo-randomoffset is applied from frame to frame thatresults in temporal incoherence. Finally, avariable sampling density in k-space stabi-lizes the iterative reconstruction. The recon-struction program is implemented directly onthe scanner and runs a nonlinear iterativereconstruction (80 iterations) with k-t regu-larization derived from a parallel imagingreconstruction that takes coil sensitivitymaps into account. For cine CMR, no addi-tional reference scans are needed because
the coil sensitivity maps are intrinsically calculatedfrom the temporal average of the input data in acentral region of k-space (Table 1, Online Appendix).
STUDY PROTOCOL. Healthy volunteers (n ¼ 12) andpatients with different pathologies of the LV (n ¼ 21)were included in the study. The robustness and pre-cision of the CS approach to measure LVEF andLV volumes and mass were assessed in comparisonwith a standard high-resolution steady-state freeprecession cine CMR approach. All CMR examina-tions were performed on a 1.5-T magnetic resonancescanner (MAGNETOM Aera, Siemens AG, Erlangen,Germany). The imaging protocol consisted of cardiaclocalizers followed by the acquisition of a stack ofconventional short-axis steady-state free precessioncine images covering the entire LV (Table 1). Next, totest the new CS-based technique, slice orientations
-
FIGURE 1 3D Reconstruction Derived From 7 Slices Acquired Within a Single Breath-Hold
The tracking of systolic left ventricular (LV) long-axis shortening allows for accurate LV volume measurements (yellow plane indicates mitral
valve plane) in a volunteer (A to D) and in a patient with ejection fraction of 32% (E to H). Any orientation of the 3-dimensional (3D)
representation of the LV is available for inspection of function.
TABLE 2 Characteristics of Study Population (N ¼ 33)
Volunteers(n ¼ 12)
Patients(n ¼ 21) p Value
Age, yrs 33 � 8 63 � 14
-
TABLE 3 Image Quality
Standard (12 Volunteers/21 Patients) CS (12 Volunteers/21 Patients)
Score 0 Score 1 Score 2 Score 3 Score 0 Score 1 Score 2 Score 3
LV coverage 12/21 — — — 12/21 — — —
Wrap around 12/21 — — — 9/18 —/4 — 1/2
Ghosts 12/18 —/3 — — 11/20 — — 1/—
Image blurring/mistriggering 11/18 1/2 —/1 — 12/21 — — —
Metallic artifact 12/21 — — — 12/21 — — —
Shimming artifact 12/21 — — — 12/21 — — —
Signal loss 12/21 — — — 12/21 — — —
Orientation of stack 12/21 — — — 12/21 — — —
Correct LV long axes 12/21 — — — 12/21 — — —
Slice thickness/gap 12/21 — — — not applicable
Total score 0.24 � 0.50 0.40 � 1.00
Quality score for standard versus CS was not significantly different (p ¼ 0.36 by Wilcoxon signed rank test).Abbreviations as in Table 1.
FIGURE 2 Good-Quality Compressed Sensing Acquisition in a
79-Year-Old Patient
Quality score was 0 in a 79-year-old patient with left ventricular systolic
dysfunction (left ventricular ejection fraction 41%) and severe aortic regur-
gitation (arrow) before aortic valve replacement. No flow-related or fold-over
artifacts are visible. (A and B) Three-chamber view in diastole/systole.
(C and D) Short-axis view in diastole/systole. Abbreviation as in Figure 1.
J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4 Vincenti et al.S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2 Compressed Sensing CMR for Rapid LV Assessment
885
artifact involved 2 or $3 slices, 2 or 3 points weregiven, respectively. For the standard cine CMRimages, these criteria were applied to the short-axisacquisitions only because they are used exclusivelyfor volume calculations. For the CS quality assess-ment, these criteria were applied to both the short-and long-axis CS acquisitions because these sliceorientations were used for volume calculations withthis approach. For quantitative measurements, theconventional stack of short-axis cine MR images wasanalyzed by the Argus VF software (Siemens AG)applying the Simpson rule. The CS cine data wereanalyzed by the 4-dimensional (4D)–VF Argus soft-ware (Siemens AG) (18). This software is based onan LV model, and with relatively few operator in-teractions, the contours for the LV endocardiumand epicardium are generated by the analysis tool.This 4D analysis tool automatically tracks the3-dimensional (3D) motion of the mitral annulusthroughout the cardiac cycle (Figure 1) and thus al-lows for an accurate volume calculation, particularlyat the base of the heart.
Aortic flow was quantified by manually tracing aregion of interest on the ascending aorta of thephase-contrast CMR images corrected for any phaseoffset measured on the chest wall (Argus flow anal-ysis package, Siemens AG). Because aortic flow wasperformed distal to the coronary arteries, flow in thecoronary arteries was estimated as the LV massmultiplied by 0.8 ml/min/g divided by heart rate (19).This flow was added to the aortic flow before com-parison versus volumetrically-determined LVSV. Pa-tients were excluded from this analysis because thepresence of a mitral regurgitation of any degreewould cause an overestimation of the volumetricLVSV in the patient group.
To assess intraobserver and interobserver repro-ducibility of the CS technique, the CS cine imageswere analyzed by 2 experienced cardiologists (G.V.and P.M. [CMR level III experts]).
-
FIGURE
Dilated
Patient
(yellow
2-chamb
thrombu
tole. (C
Vincenti et al. J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4
Compressed Sensing CMR for Rapid LV Assessment S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2886
STATISTICAL ANALYSES. The standard multi–breath-hold and the novel single–breath-hold CStechniques were compared using Bland-Altman (20)and linear regression analyses. The same analyseswere performed to assess interobserver and intra-observer variabilities of the CS technique. Compari-sons of demographic and CMR data (Table 2) involvedthe unpaired 2-tailed Student t test. Differences inquality scores between the standard multi–breath-hold and CS technique were explored using thenonparametric Wilcoxon matched-pairs signed-ranktest. Differences between LVSV and aortic flow wereevaluated by paired Student t test. Values of p < 0.05were considered statistically significant. Stata 12software (StataCorp LP, College Station, Texas) wasused for statistical analyses.
RESULTS
STUDY POPULATION AND IMAGE QUALITY. Demo-graphic data and image quality scores are given in
3 Compressed Sensing Acquisition in a 58-Year-Old Patient With
Cardiomyopathy
had a left ventricular ejection fraction of 8%, an apical thrombus
arrow), and pleural effusion (asterisk). Mild fold-over artifact in the
er view (white arrow); quality is sufficient to detect the apical
s (quality score of 0). (A and B) Two-chamber view in diastole/sys-
and D) Short-axis view in diastole/systole. Abbreviation as in Figure 1.
Tables 2 and 3, respectively. Excellent image qualityof the single–breath-hold multislice CS acquisitionswas obtained with a mean score of 0.39 � 0.79 (notdifferent from that of standard CMR: 0.24 � 0.50;p ¼ 0.36) (Table 3, Figures 2, 3, and 4). Two CS ac-quisitions had poor image quality (score $3) due tofold-over artifacts (Figures 5A and 5B). Nevertheless,all CS data were of adequate quality for quantitative4D analysis (Figure 1). Figures 5C and 5D show ex-amples of flow-related artifacts. Although smallstructures such as trabeculations were visualizedin the CS data, smaller anatomic details such asbranches of coronary arteries were not detectable onthe CS images.
COMPARISON OF THE NEW SINGLE–BREATH-HOLD
CS APPROACH VERSUS THE STANDARD MULTI–
BREATH-HOLD CINE MR TECHNIQUE. The conven-tional cine MR acquisitions were used as the “goldstandard” for measurements of LVEF, left ventricularend-diastolic volume (LVEDV), left ventricular end-systolic volume (LVESV), LVSV, and LV mass. Theseimages were collected with 4 to 6 breath-holds,yielding 8 to 12 short-axis slices (the 3 long-axis ac-quisitions were used occasionally to decide on themost basal short-axis slice to be included for analysis;otherwise, the long-axis acquisitions were not usedfor analysis). For the acquisition of the CS cine data,2 breath-holds were performed (1 non-cine acquisi-tion to check for fold-over and 1 cine acquisition).The CS cine acquisitions yielded a similar LVEF of48.5 � 15.9% versus 49.8 � 15.8% measured by thestandard multi–breath-hold technique (p ¼ 0.11).Excellent agreement was obtained by Bland-Altmananalysis (Figure 6).
LVEDV was slightly smaller by 6.0 � 10.2% (p ¼0.0009) when measured by the CS approach, whereasno differences were found for LVESV, LVSV, and LVmass (Table 4). Bland-Altman and linear regressionanalyses yielded good agreement for all 4 parameters(Figure 7).
VALIDATION OF THE NEW SINGLE–BREATH-HOLD
MULTISLICE CS APPROACH. In the group of 12healthy volunteers, excellent correlation was foundbetween LVSVCS and aortic flow, with an over-estimation by LVSVCS of 5.6 � 6.5 ml/beat as illus-trated in Figure 8A. The standard multi–breath-holdCMR approach overestimated by 16.2 � 11.7 ml/beat(p ¼ 0.012 vs. CS) (Table 4, Figure 8C), and its vari-ability versus the reference aortic flow was larger(Figures 8B and 8D).
INTRAOBSERVER AND INTEROBSERVER VARIABILITY
OF THE NEW SINGLE–BREATH-HOLD CS APPROACH.
Intraobserver and interobserver variabilities for the
-
FIGURE 4 Good-Quality Compressed Sensing Acquisition in a
56-Year-Old Patient
Patient had a dual-chamber pacemaker (arrows) (Advisa MRI SureScan System, Medtronic
Inc., Mounds View, Minnesota) implanted for complete atrioventricular block of unknown
origin. Cardiac magnetic resonance imaging was requested for the detection of myocarditis
or infiltrative cardiomyopathy. The pacemaker leads are visible in the right atrium and right
ventricle (arrows) without affecting the image quality (quality score of 0 [good]). (A and
B) Four-chamber view in diastole/systole. (C and D) Short-axis view in diastole/systole.
FIGURE 5 Fold-Over Artifact in the Superior-Inferior Direction on Short-Axis
and 2-Chamber Views
(A) Short-axis view. (B) Two-chamber view. Quality score is 3 (fold-over artifacts in>3 slices).
Flow-related artifacts in phase-encoding direction (arrows) on systolic phases in a volunteer
propagating over the aortic valve in the 3-chamber view (C) and over the mitral valve in the
4-chamber view (D). Quality score is 1 (image blurring in 1 slice, apical region in C).
J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4 Vincenti et al.S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2 Compressed Sensing CMR for Rapid LV Assessment
887
new CS approach were excellent, as given in Table 5.The intraobserver and interobserver differencesranged from �3.8% to �0.4% and from �9.7% toþ0.7%, respectively, with slopes of regressionsranging between 0.94 and 1.06 and between 0.93 and1.06, respectively (Table 5).
DISCUSSION
PERFORMANCE OF THE SINGLE–BREATH-HOLD
CS APPROACH AND COMPARISON WITH OTHER
TECHNIQUES. Overall, the single–breath-hold CSapproach yielded a high image quality in 94% ofall participants (Table 3), and all examinations wereof sufficient quality to undergo quantitative 3DLV model–based analysis. With recently-establishedquality criteria applied (9), a score $3 (i.e., artifactsin $3 slices) was obtained in only 2 participants(6% of study participants). The standard cineCMR technique is known to yield good image qualityin patients equipped with MR-conditional pace-makers (21), and similarly, an excellent quality wasachieved with the CS approach in a pacemaker-implanted patient as illustrated in Figure 4. Thehigh image quality of the CS technique translated intohigh agreement for LVEF, with a mean differenceversus the standard multi–breath-hold technique of1.3% (95% confidence interval [CI]: �7.2% to þ9.7%).For interstudy variabilities of the standard LVEFmeasurements, 95% CIs of �4.1% to þ4.3% were re-ported (22,23), which are only slightly narrower thanthe differences found between the standard cine andCS technique in the current study. Thus, the vari-ability between the standard and CS techniqueobserved in the current study was comparable to thereproducibility of the standard technique itself.Accordingly, the novel CS-based technique, althoughmuch faster than the standard approach, may bereproducible and accurate enough to replace thestandard multi–breath-hold CMR approach.
High agreement between the standard and CStechnique was also achieved for LVESV, with anonsignificant difference of 2.0 ml (Table 4). With thepresented single–breath-hold CS technique, LV masswas also measured accurately, with a differenceversus the standard technique of 2% (95% CI: �16.8%to þ12.8%). This interval for LV mass quantificationis small considering that the intraoperator variabilityfor standard multi–breath-hold LV mass measure-ments was reported to be �12.3% to þ15.9% (22).
However, LVEDV was statistically smaller withthe CS approach by 9.9 ml (Table 4). Similarly, withother highly-accelerated single–breath-hold tech-niques that are based on spatiotemporal correlations
-
200 8040 60
Dif
fere
nce
(S
tan
dar
d -
Mu
ltis
lice
CS
)
Mean LVEF (%)
LVEF: Standard vs. Multislice CSA
40
30
20
10
0
-10
-20
-30
-40
All subjects (n = 33): y = 0.9694x + 0.2116r = 0.96 p < 0.00001
Patients (n = 21): y = 0.9334x + 0.9475r = 0.97 p < 0.00001
Vol
Pat
95% CI
Mean
200 8040 60
LV
EF
Mu
ltis
lice
CS
(%
)
LVEF Standard (%)
LVEF: Standard vs. Multislice CSB80
70
60
50
40
30
20
10
0
Vol
Pat
FIGURE 6 LVEF Quantification of CS and Standard Cardiac Magnetic Resonance
(A) Bland-Altman. (B) Linear regressions (diagonal line is line of identity). CI ¼ confidence interval; CS ¼ compressed sensing; LVEF ¼ leftventricular ejection fraction; Pat ¼ patient; Vol ¼ volunteer.
TABLE 4 Compariso
Multislice CS Techniq
LVEF, %
LVEDV, ml
LVESV, ml
LVSV, ml
LV mass, g
Validation (n ¼ 12volunteers)
Flow – LVSV, ml
Flow – LVSV, ml
Values are mean � SD.Abbreviations as in Tabl
Vincenti et al. J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4
Compressed Sensing CMR for Rapid LV Assessment S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2888
such as kt-BLAST (12), TPAT (13), or TSENSE(24), slightly smaller LVEDVs (�8.2 to �4.9 ml) andslightly larger LV mass (1.1 to 13.1 g) were foundversus the conventional multi–breath-hold tech-nique (LVESV was not significantly different).Temporal filtering of the accelerated acquisitionscould result in such an underestimation of LVEDVif, for example, the increase in LV volume duringatrial contraction was missed. In preliminary tests,we determined the CS parameters that could track
n of Standard Multi–Breath-Hold CMR Versus Single–Breath-Hold
ue and Validation
Standard CMR(n ¼ 33)
CS Technique CMR(n ¼ 33)
MeanDifference p Value
49.8 � 15.8 48.5 � 15.9 1.3 � 4.3 0.11176.4 � 56.9 166.5 � 59.0 9.9 � 10.2 0.000994.5 � 63.2 92.4 � 66.4 2.0 � 11.7 0.2482.8 � 25.6 74.1 � 22.7 8.7 � 12.8 0.11128.8 � 32.1 131.2 � 32.6 �2.5 � 9.6 0.18
Aortic flow reference CS technique CMR
86.3 � 16.1 91.9 � 14.1 �5.6 � 6.5 0.01Aortic flow reference Standard CMR
86.3 � 16.1 102.4 � 15.3 �16.2 � 11.7 0.0006
es 1 and 2.
the premature septal contraction in left bundlebranch block, a short-lived contraction occurringduring the first 30 to 60 ms after the R-wave trigger(25). Thus, the presented technique is able todetect phenomena occurring immediately after theR-wave. Alternatively, at a lower signal-to-noiseratio, the endocardial border, and consequentlyblood-filled intertrabecular spaces, might be lesswell detected, potentially resulting in a “smaller”endocardial contour (i.e., underestimation of LVEDVand overestimation of LV mass). If blood-filledintertrabecular spaces are nearly absent in the“compacted” end-systolic phase and are “under-estimated” in end-diastole by accelerated tech-niques, this may result in more precise ejectionvolumes, which was observed in this study becauseLVSVCS correlated better with aortic flow than thestandard approach.
The presented CS approach acquires 1 slice per 2heart beats, whereas kt-BLAST acquires 1 slice per 1.6heart beats (12) and TPAT acquires 1 slice per 2.5 heartbeats (13). If the acquisition time is corrected forspatial resolution, the CS approach is approximately 2and 3 times faster than kt-BLAST and TPAT, respec-tively, and it is 4 to 5 times faster than TPAT if cor-rected for temporal resolution. The elegant real-timeCS approach of Feng et al. (26) acquires 1 slice per heartbeat, but voxels are 3 times larger and temporal reso-lution is 1.5 times lower versus that in the current
-
15050 250100 200 350300
Dif
fere
nce
(S
tan
dar
d -
Mu
ltis
lice
CS
)
Mean LVEDV (ml)
LVEDV: Standard vs. Multislice CSA80
60
40
20
0
-20
-40
-60
-80150100 200 35050 250 300 400
LVE
DV
Mu
ltis
lice
CS
(m
l)
LVEDV Standard (ml)
LVEDV: Standard vs. Multislice CSB400
350
300
250
200
150
100
50
10050 1500 200 300250 350
Dif
fere
nce
(S
tan
dar
d -
Mu
ltis
lice
CS
)
Mean LVESV (ml)
LVESV: Standard vs. Multislice CSC60
40
20
0
-20
-40
-6015010050 200 3000 250 350
LVE
SV
Mu
ltis
lice
CS
(m
l)
LVESV Standard (ml)
LVESV: Standard vs. Multislice CSD350
300
250
200
150
100
50
0
Vol
Pat
95% CI
Mean
Vol
Pat
95% CI
Mean
All subjects (n = 33): y = 0.9979x – 9.4924r = 0.96 p < 0.00001
Patients (n = 21): y = 1.0143x – 8.4616r = 0.97 p < 0.00001
All subjects (n = 33): y = 1.0352x – 5.3778r = 0.98 p < 0.00001
Patients (n = 21): y = 1.0318x – 3.0236r = 0.98 p < 0.00001
Vol
Pat
Vol
Pat
500 100 150
Dif
fere
nce
(S
tan
dar
d -
Mu
ltis
lice
CS
)
Mean LVSV (ml)
LVSV: Standard vs. Multislice CSE80
60
40
20
0
-20
-40
-60
-800 5025 10075 150125
LVS
V M
ult
islic
e C
S (
ml)
LVSV Standard (ml)
LVSV: Standard vs. Multislice CSF150
125
100
75
50
25
0
25 22575 125 175
Dif
fere
nce
(S
tan
dar
d -
Mu
ltis
lice
CS
)
Mean LV Mass (g)
LV Mass: Standard vs. Multislice CSG80
60
40
20
0
-20
-40
-60
-8015010050 2000 250
LV M
ass
Mu
ltis
lice
CS
(g
)
LV Mass Standard (g)
LV Mass: Standard vs. Multislice CSH250
200
150
100
50
0
Vol
Pat
95% CI
Mean
Vol
Pat
95% CI
Mean
All subjects (n = 33): y = 0.9616x + 0.6212r = 0.86 p < 0.00001
Patients (n = 21): y = 0.931x + 1.1282r = 0.82 p < 0.00001
All subjects (n = 33): y = 0.9718x + 6.0989r = 0.95 p < 0.00001
Patients (n = 21):y = 1.0037x – 0.7116r = 0.95 p < 0.00001
Vol
Pat
Vol
Pat
FIGURE 7 Quantification of LVEDV, LVESV, LVSV, and LV Mass of CS and Standard Cardiac Magnetic Resonance
(A, C, E, and G) Bland-Altman. (B, D, F, and H) Linear regressions (diagonal line is line of identity). LVEDV ¼ left ventricular end-diastolicvolume; LVESV ¼ left ventricular end-systolic volume; LVSV ¼ left ventricular stroke volume; other abbreviations as in Figure 6.
J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4 Vincenti et al.S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2 Compressed Sensing CMR for Rapid LV Assessment
889
-
500 100 150
Dif
fere
nce
(A
ort
ic F
low
- M
ult
islic
e C
S)
Mean Aortic Flow - LVSV (ml)
Validation: Multislice CS LVSV vs. Aortic FlowA60
40
20
0
-20
-40
-60
Vol
95% CI
Mean
5025 75 1250 100 150
y = 0.8028x + 22.634
r = 0.91 p < 0.00001
LVS
V M
ult
islic
e C
S (
ml)
Aortic Flow (ml)
Validation: Multislice CS LVSV vs. Aortic FlowB150
125
100
75
50
25
0
Vol
500 100 150
Dif
fere
nce
(A
ort
ic F
low
- S
tan
dar
d L
VS
V)
Mean Aortic Flow - LVSV (ml)
Validation: Standard LVSV vs. Aortic FlowC60
40
20
0
-20
-40
-605025 75 1250 100 150
y = 0.6855x + 43.282
r = 0.72 p < 0.01
Sta
nd
ard
LV
SV
(m
l)
Aortic Flow (ml)
Validation: Standard LVSV vs. Aortic FlowD150
125
100
75
50
25
0
VolVol
95% CI
Mean
FIGURE 8 Validation of CS and Standard Cardiac Magnetic Resonance Imaging for Quantification of LVSV Versus Aortic Outflow
(A and C) Bland-Altman. (B and D) Linear regressions (diagonal line is line of identity). The novel CS approach was more accurate than the
conventional technique (A and C: underestimation of 5.6 � 6.5 ml vs. 16.2 � 11.7 ml; p ¼ 0.012), and its variability versus aortic flow was lower(B and D). Abbreviations as in Figures 6 and 7.
Vincenti et al. J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4
Compressed Sensing CMR for Rapid LV Assessment S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2890
study. Future studies are needed to determinewhether this higher speed achieved by the presentedCS approach is translated into higher accuracy andreproducibility versus other highly-acceleratedtechniques.
The presented CS technique also offers the advan-tage that functional information in at least a singleplane can be obtained in patients unable to hold theirbreath for several heart beats or in patients witharrhythmias.VALIDATION OF THE NOVEL SINGLE–BREATH-HOLD
CS TECHNIQUE. To accurately account for thelong-axis shortening of the LV, a set of radial long-axis acquisitions was proposed, which achievedgood reproducibility in humans and excellent agree-ment for volumes in porcine hearts (27). In thecurrent approach, we combined long-axis CS acqui-sitions to benefit from their high in-plane spatialresolution to accurately assess the position of themitral valve annulus, and short-axis CS acquisitionswere added to minimize partial volume artifacts thatcan occur near the papillary muscles on long-axis
orientations. In the volunteer group without mitralinsufficiency, the flow in the ascending aorta wasused as a reference for LVSV. When this multisliceCS acquisition scheme was combined with the 3DLV model–based analysis, the LVSV results indicatedthat the single–breath-hold CS approach was moreaccurate than the conventional multi–breath-holdapproach when volumetric LVSV was compared withaortic flow (difference 5.6 ml vs. 16.2 ml, respec-tively; p ¼ 0.012) (Figure 8A). More importantly,LVSVCS was not only more precise but also lessvariable than that found with the conventionalapproach (i.e., smaller SD of 6.5 ml/beat vs. 11.7 ml/beat of the standard multi–breath-hold approach incomparison with aortic flow) (Table 4, Figure 8B).One might infer from these LVSV results that LVEDVand LVESV were also measured more accurately bythe CS approach as compared with the conventionalmulti–breath-hold approach. Most likely the higheraccuracy of the CS approach was due to the fact thatthis technique allows correct tracking of the 3D mo-tion of the base of the heart during the cardiac cycle;
-
TABLE 5 Intraobserver and Interobserver Variabilities of the Single–Breath-Hold
Multislice CS Technique
Bland-Altman (n ¼ 33) Correlation (n ¼ 33)
Mean SD % Mean % SD r slope p Value
Intraobserver
LVEF, % �0.4 �3.4 �2.5 �13.3 0.97 0.99
-
Vincenti et al. J A C C : C A R D I O V A S C U L A R I M A G I N G , V O L . 7 , N O . 9 , 2 0 1 4
Compressed Sensing CMR for Rapid LV Assessment S E P T E M B E R 2 0 1 4 : 8 8 2 – 9 2892
RE F E RENCE S
1. Curtis JP, Sokol SI, Wang Y, et al. The associa-tion of left ventricular ejection fraction, mortality,and cause of death in stable outpatients with heartfailure. J Am Coll Cardiol 2003;42:736–42.
2. Gharib MI, Burnett AK. Chemotherapy-inducedcardiotoxicity: current practice and prospects ofprophylaxis. Eur J Heart Fail 2002;4:235–42.
3. Bardy GH, Lee KL, Mark DB, et al. Amiodaroneor an implantable cardioverter-defibrillator forcongestive heart failure. N Engl J Med 2005;352:225–37.
4. Sürder D, Manka R, Lo Cicero V, et al. Intra-coronary injection of bone marrow-derivedmononuclear cells early or late after acute myo-cardial infarction: effects on global left ventricularfunction. Circulation 2013;127:1968–79.
5. Hunt SA, Abraham WT, Chin MH, et al. 2009focused update incorporated into the ACC/AHA2005 Guidelines for the Diagnosis and Manage-ment of Heart Failure in Adults. J Am Coll Cardiol2009;53:e1–90.
6. McMurray JJV, Adamopoulos S, Anker SD,et al. ESC guidelines for the diagnosis and treat-ment of acute and chronic heart failure 2012. EurHeart J 2012;33:1787–847.
7. Kramer C, Barkhausen J, Flamm S, et al. Stan-dardized cardiovascular magnetic resonance(CMR) protocols 2013 update. J Cardiovasc MagnReson 2013;15:91.
8. Kilner PJ, Geva T, Kaemmerer H, et al. Recom-mendations for cardiovascular magnetic resonancein adults with congenital heart disease from therespective working groups of the European Soci-ety of Cardiology. Eur Heart J 2010;31:794–805.
9. Klinke V, Muzzarelli S, Lauriers N, et al. Qualityassessment of cardiovascular magnetic resonancein the setting of the European CMR Registry:description and validation of standardized criteria.J Cardiovasc Magn Reson 2013;15:55.
10. Bruder O, Wagner A, Lombardi M, et al. Eu-ropean Cardiovascular Magnetic Resonance(EuroCMR) Registry—multi national results from57 centers in 15 countries. J Cardiovasc MagnReson 2013;15:1–9.
11. Kozerke S, Plein S. Accelerated CMR usingzonal, parallel and prior knowledge driven imagingmethods. J Cardiovasc Magn Reson 2008;10:29.
12. Jahnke C, Nagel E, Gebker R, et al. Four-dimensional single breathhold magnetic resonanceimaging using kt-BLAST enables reliable assess-ment of left- and right-ventricular volumes andmass. J Magn Reson Imaging 2007;25:737–42.
13. Eberle H, Nassenstein K, Jensen C, et al.Rapid MR assessment of left ventricular systolicfunction after acute myocardial infarction usingsingle breath-hold cine imaging with the tem-poral parallel acquisition technique (TPAT) and4D guide-point modelling analysis of left ven-tricular function. Eur Radiol 2010;20:73–80.
14. Plein S, Schwitter J, Suerder D, et al. k-tSENSE-accelerated myocardial perfusion MR im-aging at 3.0 Tesla—comparison with 1.5 Tesla.Radiology 2008;249:493–500.
15. Schwitter J, Oelhafen M, Wyss BM, et al. 2D-spatially-selective real-time magnetic resonanceimaging for the assessment of microvascularfunction and its relation to the cardiovascular riskprofile. J Cardiovasc Magn Reson 2006;8:759–69.
16. Lustig M, Donoho D, Pauly JM. Sparse MRI: theapplication of compressed sensing for rapid MRimaging. Magn Reson Med 2007;58:1182–95.
17. Liu J. Dynamic cardiac MRI reconstructionwith weighted redundant Haar wavelets (abstr).Proc Int Soc Magn Reson Med 2012;67:4249.
18. Young AA, Cowan BR, Thrupp SF, Hedley WJ,Dell’Italia LJ. Left ventricular mass and volume:fast calculation with guide-point modeling on MRimages. Radiology 2000;216:597–602.
19. Schwitter J, DeMarco T, Kneifel S, et al. Mag-netic resonance-based assessment of global cor-onary flow and flow reserve and its relation to leftventricular functional parameters: a comparisonwith positron emission tomography. Circulation2000;101:2696–702.
20. Bland J, Altman D. Statistical methods forassessing agreement between two methods ofclinical measurement. Lancet 1986;1:307–10.
21. Schwitter J, Kanal E, Schmitt M, et al. Impactof the Advisa MRI pacing system on the diagnosticquality of cardiac MR images and contractionpatterns of cardiac muscle during scans: AdvisaMRI randomized clinical multicenter study results.Heart Rhythm 2013;10:864–72.
22. Danilouchkine M, Westenberg J, De Roos A,Reiber J, Lelieveldt P. Operator induced variabilityin cardiovascular MR: left ventricular measure-ments and their reproducibility. J Cardiovasc MagnReson 2005;7:447–57.
23. Grothues F, Smith G, Moon J, et al. Compari-son of interstudy reproducibility of cardiovascularmagnetic resonance with 2-dimensional echocar-diography in normal subjects and in patients withheart failure or left ventricular hypertrophy. Am JCardiol 2002;90:29–34.
24. Young A, Cowan B, Schoenberg S,Wintersperger B. Feasibility of single breath-holdleft ventricular function with 3 Tesla TSENSEacquisition and 3D modeling analysis. J CardiovascMagn Reson 2008;10:24.
25. Rutz AK, Manka R, Kozerke S, et al. Left ven-tricular dyssynchrony in patients with left bundlebranch block and patients after myocardialinfarction: integration of mechanics and viabilityby cardiac magnetic resonance. Eur Heart J 2009;30:2117–27.
26. Feng L, Srichai MB, Lim RP, et al. Highlyaccelerated real-time cardiac cine MRI using k-tSPARSE-SENSE. Magn Reson Med 2013;70:64–74.
27. Bloomer T, Plein S, Radjenovic A, et al. CineMRI using steady-state free precession in theradial long-axis orientation is a fast accuratemethod for obtaining volumetric data of theleft ventricle. J Magn Reson Imaging 2001;14:685–92.
28. Marcus J, Goette M, DeWaal L, et al. The in-fluence of through-plane motion on left ventric-ular volumes measured by magnetic resonanceimaging: implications for image acquisition andanalysis. J Cardiovasc Magn Reson 1999;1:1–6.
29. Muzzarelli S, Monney P, O’Brien K, et al.Quantification of aortic flow by phase-contrastmagnetic resonance in patients with bicuspidaortic valve. Eur Heart J Cardiovasc Imaging 2013;15:77–84.
30. Captur G, Muthurangu V, Cook C, et al.Quantification of left ventricular trabeculae usingfractal analysis. J Cardiovasc Magn Reson 2013;15:36.
31. Papavassiliu T, Kühl HP, Schröder M, et al.Effect of endocardial trabeculae on left ventricularmeasurements and measurement reproducibilityat cardiovascular MR imaging. Radiology 2005;236:57–64.
32. Moon J, Lorenz C, Francis J, Smith G,Pennell D. Breath-hold FLASH and FISP cardio-vascular MR imaging: left ventricular volume dif-ferences and reproducibility. Radiology 2002;223:789–97.
KEY WORDS cardiac function, fastimaging, fast LV assessment
APPENDIX For supplemental information,please see the online version of this article.
http://refhub.elsevier.com/S1936-878X(14)00483-5/sref1http://refhub.elsevier.com/S1936-878X(14)00483-5/sref1http://refhub.elsevier.com/S1936-878X(14)00483-5/sref1http://refhub.elsevier.com/S1936-878X(14)00483-5/sref1http://refhub.elsevier.com/S1936-878X(14)00483-5/sref2http://refhub.elsevier.com/S1936-878X(14)00483-5/sref2http://refhub.elsevier.com/S1936-878X(14)00483-5/sref2http://refhub.elsevier.com/S1936-878X(14)00483-5/sref3http://refhub.elsevier.com/S1936-878X(14)00483-5/sref3http://refhub.elsevier.com/S1936-878X(14)00483-5/sref3http://refhub.elsevier.com/S1936-878X(14)00483-5/sref3http://refhub.elsevier.com/S1936-878X(14)00483-5/sref4http://refhub.elsevier.com/S1936-878X(14)00483-5/sref4http://refhub.elsevier.com/S1936-878X(14)00483-5/sref4http://refhub.elsevier.com/S1936-878X(14)00483-5/sref4http://refhub.elsevier.com/S1936-878X(14)00483-5/sref4http://refhub.elsevier.com/S1936-878X(14)00483-5/sref5http://refhub.elsevier.com/S1936-878X(14)00483-5/sref5http://refhub.elsevier.com/S1936-878X(14)00483-5/sref5http://refhub.elsevier.com/S1936-878X(14)00483-5/sref5http://refhub.elsevier.com/S1936-878X(14)00483-5/sref5http://refhub.elsevier.com/S1936-878X(14)00483-5/sref6http://refhub.elsevier.com/S1936-878X(14)00483-5/sref6http://refhub.elsevier.com/S1936-878X(14)00483-5/sref6http://refhub.elsevier.com/S1936-878X(14)00483-5/sref6http://refhub.elsevier.com/S1936-878X(14)00483-5/sref7http://refhub.elsevier.com/S1936-878X(14)00483-5/sref7http://refhub.elsevier.com/S1936-878X(14)00483-5/sref7http://refhub.elsevier.com/S1936-878X(14)00483-5/sref7http://refhub.elsevier.com/S1936-878X(14)00483-5/sref8http://refhub.elsevier.com/S1936-878X(14)00483-5/sref8http://refhub.elsevier.com/S1936-878X(14)00483-5/sref8http://refhub.elsevier.com/S1936-878X(14)00483-5/sref8http://refhub.elsevier.com/S1936-878X(14)00483-5/sref8http://refhub.elsevier.com/S1936-878X(14)00483-5/sref9http://refhub.elsevier.com/S1936-878X(14)00483-5/sref9http://refhub.elsevier.com/S1936-878X(14)00483-5/sref9http://refhub.elsevier.com/S1936-878X(14)00483-5/sref9http://refhub.elsevier.com/S1936-878X(14)00483-5/sref9http://refhub.elsevier.com/S1936-878X(14)00483-5/sref10http://refhub.elsevier.com/S1936-878X(14)00483-5/sref10http://refhub.elsevier.com/S1936-878X(14)00483-5/sref10http://refhub.elsevier.com/S1936-878X(14)00483-5/sref10http://refhub.elsevier.com/S1936-878X(14)00483-5/sref10http://refhub.elsevier.com/S1936-878X(14)00483-5/sref11http://refhub.elsevier.com/S1936-878X(14)00483-5/sref11http://refhub.elsevier.com/S1936-878X(14)00483-5/sref11http://refhub.elsevier.com/S1936-878X(14)00483-5/sref12http://refhub.elsevier.com/S1936-878X(14)00483-5/sref12http://refhub.elsevier.com/S1936-878X(14)00483-5/sref12http://refhub.elsevier.com/S1936-878X(14)00483-5/sref12http://refhub.elsevier.com/S1936-878X(14)00483-5/sref12http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref13http://refhub.elsevier.com/S1936-878X(14)00483-5/sref14http://refhub.elsevier.com/S1936-878X(14)00483-5/sref14http://refhub.elsevier.com/S1936-878X(14)00483-5/sref14http://refhub.elsevier.com/S1936-878X(14)00483-5/sref14http://refhub.elsevier.com/S1936-878X(14)00483-5/sref15http://refhub.elsevier.com/S1936-878X(14)00483-5/sref15http://refhub.elsevier.com/S1936-878X(14)00483-5/sref15http://refhub.elsevier.com/S1936-878X(14)00483-5/sref15http://refhub.elsevier.com/S1936-878X(14)00483-5/sref15http://refhub.elsevier.com/S1936-878X(14)00483-5/sref16http://refhub.elsevier.com/S1936-878X(14)00483-5/sref16http://refhub.elsevier.com/S1936-878X(14)00483-5/sref16http://refhub.elsevier.com/S1936-878X(14)00483-5/sref17http://refhub.elsevier.com/S1936-878X(14)00483-5/sref17http://refhub.elsevier.com/S1936-878X(14)00483-5/sref17http://refhub.elsevier.com/S1936-878X(14)00483-5/sref18http://refhub.elsevier.com/S1936-878X(14)00483-5/sref18http://refhub.elsevier.com/S1936-878X(14)00483-5/sref18http://refhub.elsevier.com/S1936-878X(14)00483-5/sref18http://refhub.elsevier.com/S1936-878X(14)00483-5/sref19http://refhub.elsevier.com/S1936-878X(14)00483-5/sref19http://refhub.elsevier.com/S1936-878X(14)00483-5/sref19http://refhub.elsevier.com/S1936-878X(14)00483-5/sref19http://refhub.elsevier.com/S1936-878X(14)00483-5/sref19http://refhub.elsevier.com/S1936-878X(14)00483-5/sref19http://refhub.elsevier.com/S1936-878X(14)00483-5/sref20http://refhub.elsevier.com/S1936-878X(14)00483-5/sref20http://refhub.elsevier.com/S1936-878X(14)00483-5/sref20http://refhub.elsevier.com/S1936-878X(14)00483-5/sref21http://refhub.elsevier.com/S1936-878X(14)00483-5/sref21http://refhub.elsevier.com/S1936-878X(14)00483-5/sref21http://refhub.elsevier.com/S1936-878X(14)00483-5/sref21http://refhub.elsevier.com/S1936-878X(14)00483-5/sref21http://refhub.elsevier.com/S1936-878X(14)00483-5/sref21http://refhub.elsevier.com/S1936-878X(14)00483-5/sref22http://refhub.elsevier.com/S1936-878X(14)00483-5/sref22http://refhub.elsevier.com/S1936-878X(14)00483-5/sref22http://refhub.elsevier.com/S1936-878X(14)00483-5/sref22http://refhub.elsevier.com/S1936-878X(14)00483-5/sref22http://refhub.elsevier.com/S1936-878X(14)00483-5/sref23http://refhub.elsevier.com/S1936-878X(14)00483-5/sref23http://refhub.elsevier.com/S1936-878X(14)00483-5/sref23http://refhub.elsevier.com/S1936-878X(14)00483-5/sref23http://refhub.elsevier.com/S1936-878X(14)00483-5/sref23http://refhub.elsevier.com/S1936-878X(14)00483-5/sref23http://refhub.elsevier.com/S1936-878X(14)00483-5/sref24http://refhub.elsevier.com/S1936-878X(14)00483-5/sref24http://refhub.elsevier.com/S1936-878X(14)00483-5/sref24http://refhub.elsevier.com/S1936-878X(14)00483-5/sref24http://refhub.elsevier.com/S1936-878X(14)00483-5/sref24http://refhub.elsevier.com/S1936-878X(14)00483-5/sref25http://refhub.elsevier.com/S1936-878X(14)00483-5/sref25http://refhub.elsevier.com/S1936-878X(14)00483-5/sref25http://refhub.elsevier.com/S1936-878X(14)00483-5/sref25http://refhub.elsevier.com/S1936-878X(14)00483-5/sref25http://refhub.elsevier.com/S1936-878X(14)00483-5/sref25http://refhub.elsevier.com/S1936-878X(14)00483-5/sref26http://refhub.elsevier.com/S1936-878X(14)00483-5/sref26http://refhub.elsevier.com/S1936-878X(14)00483-5/sref26http://refhub.elsevier.com/S1936-878X(14)00483-5/sref27http://refhub.elsevier.com/S1936-878X(14)00483-5/sref27http://refhub.elsevier.com/S1936-878X(14)00483-5/sref27http://refhub.elsevier.com/S1936-878X(14)00483-5/sref27http://refhub.elsevier.com/S1936-878X(14)00483-5/sref27http://refhub.elsevier.com/S1936-878X(14)00483-5/sref27http://refhub.elsevier.com/S1936-878X(14)00483-5/sref28http://refhub.elsevier.com/S1936-878X(14)00483-5/sref28http://refhub.elsevier.com/S1936-878X(14)00483-5/sref28http://refhub.elsevier.com/S1936-878X(14)00483-5/sref28http://refhub.elsevier.com/S1936-878X(14)00483-5/sref28http://refhub.elsevier.com/S1936-878X(14)00483-5/sref29http://refhub.elsevier.com/S1936-878X(14)00483-5/sref29http://refhub.elsevier.com/S1936-878X(14)00483-5/sref29http://refhub.elsevier.com/S1936-878X(14)00483-5/sref29http://refhub.elsevier.com/S1936-878X(14)00483-5/sref29http://refhub.elsevier.com/S1936-878X(14)00483-5/sref30http://refhub.elsevier.com/S1936-878X(14)00483-5/sref30http://refhub.elsevier.com/S1936-878X(14)00483-5/sref30http://refhub.elsevier.com/S1936-878X(14)00483-5/sref30http://refhub.elsevier.com/S1936-878X(14)00483-5/sref31http://refhub.elsevier.com/S1936-878X(14)00483-5/sref31http://refhub.elsevier.com/S1936-878X(14)00483-5/sref31http://refhub.elsevier.com/S1936-878X(14)00483-5/sref31http://refhub.elsevier.com/S1936-878X(14)00483-5/sref31http://refhub.elsevier.com/S1936-878X(14)00483-5/sref32http://refhub.elsevier.com/S1936-878X(14)00483-5/sref32http://refhub.elsevier.com/S1936-878X(14)00483-5/sref32http://refhub.elsevier.com/S1936-878X(14)00483-5/sref32http://refhub.elsevier.com/S1936-878X(14)00483-5/sref32
Compressed Sensing Single–Breath-Hold CMR for Fast Quantification of LV Function, Volumes, and MassMethodsCS techniqueStudy protocolImage analysesStatistical analyses
ResultsStudy population and image qualityComparison of the new single–breath-hold CS approach versus the standard multi–breath-hold cine MR techniqueValidation of the new single–breath-hold multislice CS approachIntraobserver and interobserver variability of the new single–breath-hold CS approach
DiscussionPerformance of the single–breath-hold CS approach and comparison with other techniquesValidation of the novel single–breath-hold CS techniqueRobustness of the new single–breath-hold multislice CS approachStudy limitations
ConclusionsReferences