Accelerated two-dimensional cine DENSE cardiovascular magnetic resonance using compressed sensing and parallel imaging
© The Author(s). 2016
Received: 2 April 2016
Accepted: 20 May 2016
Published: 14 June 2016
Cine Displacement Encoding with Stimulated Echoes (DENSE) provides accurate quantitative imaging of cardiac mechanics with rapid displacement and strain analysis; however, image acquisition times are relatively long. Compressed sensing (CS) with parallel imaging (PI) can generally provide high-quality images recovered from data sampled below the Nyquist rate. The purposes of the present study were to develop CS-PI-accelerated acquisition and reconstruction methods for cine DENSE, to assess their accuracy for cardiac imaging using retrospective undersampling, and to demonstrate their feasibility for prospectively-accelerated 2D cine DENSE imaging in a single breathhold.
An accelerated cine DENSE sequence with variable-density spiral k-space sampling and golden angle rotations through time was implemented. A CS method, Block LOw-rank Sparsity with Motion-guidance (BLOSM), was combined with sensitivity encoding (SENSE) for the reconstruction of under-sampled multi-coil spiral data. Seven healthy volunteers and 7 patients underwent 2D cine DENSE imaging with fully-sampled acquisitions (14–26 heartbeats in duration) and with prospectively rate-2 and rate-4 accelerated acquisitions (14 and 8 heartbeats in duration). Retrospectively- and prospectively-accelerated data were reconstructed using BLOSM-SENSE and SENSE. Image quality of retrospectively-undersampled data was quantified using the relative root mean square error (rRMSE). Myocardial displacement and circumferential strain were computed for functional assessment, and linear correlation and Bland-Altman analyses were used to compare accelerated acquisitions to fully-sampled reference datasets.
For retrospectively-undersampled data, BLOSM-SENSE provided similar or lower rRMSE at rate-2 and lower rRMSE at rate-4 acceleration compared to SENSE (p < 0.05, ANOVA). Similarly, for retrospective undersampling, BLOSM-SENSE provided similar or better correlation with reference displacement and strain data at rate-2 and better correlation at rate-4 acceleration compared to SENSE. Bland-Altman analyses showed similar or better agreement for displacement and strain data at rate-2 and better agreement at rate-4 using BLOSM-SENSE compared to SENSE for retrospectively-undersampled data. Rate-2 and rate-4 prospectively-accelerated cine DENSE provided good image quality and expected values of displacement and strain.
BLOSM-SENSE-accelerated spiral cine DENSE imaging with 2D displacement encoding can be acquired in a single breathhold of 8–14 heartbeats with high image quality and accurate assessment of myocardial displacement and circumferential strain.
KeywordsCine DENSE Compressed sensing Parallel imaging Myocardial strain Myocardial tagging Cardiovascular magnetic resonance
Imaging myocardial strain is of growing importance for the assessment of heart disease [1–5]. For example, recent studies have shown that strain imaging is effective for quantifying mechanical dyssynchrony and predicting response to cardiac resynchronization therapy in patients with heart failure  and that strain imaging can detect subclinical systolic dysfunction in patients with diabetes . Myocardial tagging cardiovascular magnetic resonance (CMR) , a long-established method, has been considered the gold standard method for the noninvasive measurement of myocardial strain [3, 8]. However, recently cine displacement encoding with stimulated echoes (DENSE) [9, 10] has emerged as a strain imaging technique that, compared to tagging, has equivalent accuracy and better interobserver variability . Additionally, strain analysis for cine DENSE is rapid and far less time consuming than for tagging [12–14]. While cine DENSE has advantages in interobserver variability, analysis time, and spatial resolution, it has the disadvantage that data acquisition times are inherently longer than tagging. The longer scan times for cine DENSE occur because DENSE is a phase-contrast method, requiring n + 1 separate acquisitions in order to reconstruct phase images encoded for displacement in n directions . While two-dimensional (2D) grid-tagged images are typically acquired during a clinically-convenient single breathhold , common protocols for 2D cine DENSE require two separate breathholds, each acquiring 1D displacement–encoded data and phase-reference data  or using a balanced two-point encoding method . Acceleration using data undersampling has the potential to enable cine DENSE scans with 2D displacement encoding in less than 10 s without substantial compromises in spatiotemporal resolution and accuracy, which would represent a clinically-convenient single-breathhold protocol. However, acceleration using conventional parallel imaging (PI) decreases the signal-to-noise ratio (SNR)  and may compromise the accuracy of the displacement and strain measurements.
Compressed sensing is a newer technique that is making a major impact on accelerated CMR  and which, when combined with PI, may preserve the accuracy of cine DENSE displacement and strain measurements when acceleration is employed. In CS, high-quality images can be recovered from data sampled well below the Nyquist rate provided that the sampling pattern is incoherent, the images are sparse in a transform domain, and a sparsity-promoting iterative reconstruction is used . Cine DENSE imaging may be well-suited for acceleration using CS since the data present spatiotemporal sparsity and there are correlations between data encoded for displacement in different directions. The purposes of the present study were to develop CS-PI-accelerated acquisition and reconstruction methods for cine DENSE, to assess their accuracy for measuring myocardial displacement and strain, and to demonstrate the feasibility of these methods for acquiring high-quality prospectively-accelerated 2D cine DENSE images in a single breathhold.
Pulse sequence and image reconstruction methods
A variable-density spiral cine DENSE sequence was implemented on a 1.5 T scanner (Avanto, Siemens, Erlangen, Germany), where the center of k-space was fully sampled and the outer portion of k-space was undersampled. The undersampled variable-density sequence is a modification of a previously-described spiral cine DENSE sequence , which uses phase cycling and through-plane dephasing for artifact suppression [9, 18]. Spiral interleaves were distributed uniformly in k-space within each cardiac phase and were rotated by the golden angle through different cardiac phases to achieve randomness in time. Undersampled k-space data were reconstructed using a modified Block Low-rank Sparsity with Motion guidance (BLOSM) algorithm, a CS method that exploits low-rank spatiotemporal properties within regions or blocks . The BLOSM technique, which was recently developed and applied to accelerate Cartesian first-pass contrast-enhanced perfusion CMR , was extended to incorporate sensitivity encoding (SENSE)  for the reconstruction of multi-coil data and by using the non-uniform fast Fourier transform (NUFFT)  to transform data between the image domain and the spiral trajectories in k-space. In addition, low rank properties in space, time, and through different displacement encoding directions were exploited by applying singular value decomposition (SVD) to 2D matrices containing all these data.
Computer-generated heart phantom for evaluation of BLOSM-accelerated cine DENSE
Parameters for the variable density spiral trajectories for fully-sampled and accelerated acquisitions
Number of interleaves
Initial sampling density
Ending sampling density
Volunteer and patient scans
CMR data were collected from seven healthy volunteers (4 male; mean age 29, range 23–33 years) and seven patients (3 male; mean age 64, range 39–79 years) with known or suspected heart disease. The scans were performed in accordance with protocols approved by our institutional review board after informed consent was obtained. Short-axis 2D cine DENSE images of the left ventricle (LV) at a mid-ventricular level were collected on a 1.5 T CMR scanner (Avanto, Siemens, Erlangen, Germany) with a 5-channel body-spine combined RF coil array. For each volunteer scan, fully-sampled cine DENSE datasets with 2D in-plane displacement encoding and 6–8 spiral interleaves per image were acquired within a long breathhold (20 to 26 heartbeats). For each patient scan, two fully-sampled cine DENSE datasets with orthogonal 1D in-plane displacement encoding and 6 spiral interleaves per image were acquired within a shorter breathhold (14 heartbeats). Other sequence parameters included field of view = 280-340 × 280-340 mm2, spatial resolution = 1.8-2.4 × 1.8-2.4 mm2, slice thickness = 8 mm, ramped flip angle with the last flip angle = 15°, TR = 9.8 ms, TE = 1.3 ms, and temporal resolution = 19.6 ms. In addition, for both healthy volunteers and patients prospectively undersampled datasets at acceleration rates 2 and 4 (with 4 and 2 spiral interleaves per image, respectively) with 2D displacement encoding were acquired with shorter breathholds of 14 and 8 heartbeats, respectively. With a temporal resolution of 19.6 ms per cardiac phase, all datasets acquired 20–38 cardiac phases (35–38 phases for volunteers, and 20–30 phases for patients). These images covered on average 655 ms, and approximately 75 % of the cardiac cycle. The fully-sampled datasets provided reference images, and retrospective undersampling of these datasets was used to evaluate the new acceleration methodologies. Prospectively acquired undersampled cine DENSE datasets were used to demonstrate true acceleration. Parameters describing the fully-sampled and accelerated spiral trajectories are listed in Table 1.
Image quality assessments and functional analysis
Myocardial displacement and strain were computed using previously published semi-automatic cine DENSE analysis algorithms [12, 14]. To calculate myocardial displacement and strain, endocardial and epicardial contours were manually delineated at one cardiac phase and propagated automatically to all other cardiac phases . After this segmentation process, the phase-reconstructed images were phase unwrapped and displacement and strain were computed within the segmented myocardium . Circumferential strain (Ecc), the most commonly-reported and accurately-measured strain element , was selected in this study to evaluate the reconstruction algorithms. Short-axis images were segmented into 6 regions according to the standard American Heart Association model of the left ventricle  and segmental Ecc values at all cardiac phases were recorded. Linear correlation and Bland-Altman analyses were performed to assess the relationship between the accelerated reconstructions and the reference data at each spatiotemporal Ecc data point (1488 points in total for 7 volunteers, 1062 points in total for 7 patients). When comparing Ecc from retrospectively-undersampled images to fully-sampled images, only the volunteer data were used because these fully-sampled DENSE images with 2D displacement encoding were acquired in a single breathhold. All of the rRMSE and Bland-Altman plots included all of the data from all acquired cardiac phases.
To show the behavior of BLOSM-SENSE compared to SENSE under the condition of low SNR, for one dataset various levels of noise were added to the fully-sampled data to simulate different SNR values (-10 dB to -2 dB), and the data were then rate-4 undersampled and reconstructed using both reconstruction methods.
Computer-generated heart phantom for evaluation of BLOSM-accelerated cine DENSE
Retrospective undersampling of fully-sampled volunteer and patient scans
Evaluation of prospectively-accelerated volunteer and patient scans
In this study we developed an accelerated 2D cine DENSE method that can provide accurate 2D displacement and strain maps within a single breathhold of 8-14 heartbeats utilizing variable-density undersampled spiral k-space trajectories for data sampling and BLOSM-SENSE for image reconstruction. Studies using fully-sampled computer-generated phantom data and in vivo data with retrospective undersampling demonstrated low rRMSE of CS-PI-reconstructed images and accurate displacement and strain values compared to fully-sampled data. Prospectively-accelerated in vivo scans of healthy volunteers and heart disease patients at acceleration rates of 2 and 4 demonstrated good image quality, typical values of displacement and strain for healthy volunteers, and impaired strain for patients. These results demonstrate that undersampled variable-density spiral imaging with BLOSM-SENSE reconstruction can substantially reduce the scan time for cine DENSE CMR compared to conventional protocols while maintaining accurate measurements of displacement and strain.
For CS-PI reconstruction we used a method that exploits the low-rank property in space, time, and among displacement encodings in sub-regions of the images. The BLOSM method has been extended in this study to incorporate PI and exploit data sparsity among RF coils. BLOSM was also extended to handle non-Cartesian k-space data. The properties of BLOSM and its successful application to cine DENSE imaging of the heart suggest that it can be readily extended to tagging and phase-contrast methods that also probe myocardial motion. Regional low-rank methods have recently been used successfully for other applications such as dynamic contrast-enhanced CMR [19, 26], myocardial parameter mapping  and imaging of cardiac morphology and function . Other sparsity promoting reconstructions such as x-f methods  or others  may also be applicable to cine DENSE. A comparison of all the various possible sparsifying transforms that may be applicable to accelerating cine DENSE is beyond the scope of the present study.
Using rate-4 acceleration, a common clinical 2D cine DENSE protocol can be reduced from 28 heartbeats (two breathholds of 14 heartbeats each) to 8 heartbeats. For clinical scanning, this is expected to provide a significant improvement in efficiency, better toleration and compliance by patients, and fewer misregistration errors that may occur due to inconsistencies between different breathholds. Compared to the single-breathhold SENSE-accelerated 2D cine DENSE protocol previously developed by Kim et al , the present methods provide three-fold better spatial resolution with a similar or shorter scan time.
While BLOSM makes use of and derives benefit from motion estimation and motion compensation for free-breathing single-shot first-pass perfusion CMR , motion estimation is not needed for breathhold cine DENSE imaging and the implementation of BLOSM with motion compensation for segmented acquisitions would be complex. We have reconstructed all of the images in the present study using BLOSM-SENSE both without and with motion tracking, and the results are nearly identical in terms of image appearance and quantitative metrics of image quality such as rRMSE for both rate-2 and rate-4 acceleration. Since the results are equivalent and the method is simpler and less computationally expensive without motion tracking, we chose to use the method without motion tracking for this application.
It is well established that the magnitude of the displacement-encoded stimulated echo signal decreases with time across the cardiac cycle due to T1 relaxation. We used a ramped flip angle method to partially compensate for this effect, however we did observe that SNR decreased and that rRMSE increased for cardiac phases later in the cardiac cycle. Using the present methods, we imaged approximately 75 % of the cardiac cycle, but did not reliably capture late diastole. Further optimization of the ramped flip angle method and potentially using a magnetic field strength of 3 T as recently shown  may enable BLOSM-SENSE accelerated cine DENSE imaging of the entire cardiac cycle.
To make fair comparisons in the present study, we held parameters such as readout duration and TR constant. However, acceleration can enable shorter readout durations and shorter TRs, which in turn can reduce spiral blurring artifacts and provide increased temporal resolution. Reduced readout duration also has important implications regarding better spiral cine DENSE image quality at higher field strengths such as 3 T, where off-resonance effects are greater. Thus, acceleration of cine DENSE may have broad implications for achieving shorter scan times, better temporal resolution, use at 3 T, and subsequently higher SNR and/or spatial resolution. Thus these technical innovations will likely lead to better assessment of myocardial strain for patients with heart disease.
CS-PI-accelerated spiral cine DENSE imaging with 2D in-plane displacement encoding can be acquired in a single breathhold, as short as 8 heartbeats, with high image quality and accurate functional assessments. These methods promise to improve clinical CMR of myocardial strain.
BLOSM, block low-rank sparsity with motion guidance; CS, compressed sensing; DENSE, displacement encoding with stimulated echoes; Ecc, circumferential strain; HARP, harmonic phase; NUFFT, non-uniform fast Fourier transform; PI, parallel imaging; ROI, region of interest; rRMSE, relative root mean square error; SENSE, sensitivity encoding; SOS, sum-of-squares; SVD, singular value decomposition.
This work was supported by NIH grant R01 EB 001763, American Heart Association Predoctoral Award 12PRE12040059 and Siemens Healthcare.
Availability of supporting data
All authors were responsible for algorithm development and sequence development. X.Chen was responsible for data acquisition and data analysis. X. Cai and D. A. Auger were responsible for patient data acquisition. X.Chen and F.H.Epstein were responsible for data interpretation and study design. X.Chen and F.H.Epstein drafted the manuscript. Y.Yang, M.Salerno and C.H.Meyer critically revised the manuscript. All authors read and approved the final manuscript.
X.Chen performed most of the work during his PhD study at University of Virginia.
X.Chen is a Siemens Medical Solutions USA Inc. employee. F. Epstein received research support from Siemens Healthcare.
Consent for publication
Written consent for publication was obtained from all participants.
Ethical approval and consent to participate
The study was performed in accordance with protocols approved by the University of Virginia institutional review board and informed consent was obtained from all study participants.
Open AccessThis article is distributed under the terms of the Creative Commons Attribution 4.0 International License (http://creativecommons.org/licenses/by/4.0/), which permits unrestricted use, distribution, and reproduction in any medium, provided you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons license, and indicate if changes were made. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated.
- Gorcsan 3rd J, Tanaka H. Echocardiographic assessment of myocardial strain. J Am Coll Cardiol. 2011;58(14):1401–13.View ArticlePubMedGoogle Scholar
- Geyer H, Caracciolo G, Abe H, Wilansky S, Carerj S, Gentile F, Nesser HJ, Khandheria B, Narula J, Sengupta PP. Assessment of myocardial mechanics using speckle tracking echocardiography: fundamentals and clinical applications. J Am Soc Echocardiogr. 2010;23(4):351–69. quiz 453-5.View ArticlePubMedGoogle Scholar
- Shehata ML, Cheng S, Osman NF, Bluemke DA, Lima JA. Myocardial tissue tagging with cardiovascular magnetic resonance. J Cardiovasc Magn Reson. 2009;11:55.View ArticlePubMedPubMed CentralGoogle Scholar
- Epstein FH. MRI of left ventricular function. J Nucl Cardiol. 2007;14(5):729–44.View ArticlePubMedGoogle Scholar
- Stanton T, Leano R, Marwick TH. Prediction of all-cause mortality from global longitudinal speckle strain: comparison with ejection fraction and wall motion scoring. Circ Cardiovasc Imaging. 2009;2(5):356–64.View ArticlePubMedGoogle Scholar
- Bilchick KC, Kuruvilla S, Hamirani YS, Ramachandran R, Clarke SA, Parker KM, Stukenborg GJ, Mason P, Ferguson JD, Moorman JR, Malhotra R, Mangrum JM, Darby AE, Dimarco J, Holmes JW, Salerno M, Kramer CM, Epstein FH. Impact of mechanical activation, scar, and electrical timing on cardiac resynchronization therapy response and clinical outcomes. J Am Coll Cardiol. 2014;63(16):1657–66.View ArticlePubMedPubMed CentralGoogle Scholar
- Emande L, Thibault H, Bergerot C, Mouline P, Wen H, Derumeaux G, Croisille P. Systolic myocardial dysfunction in patients with type 2 diabetes mellitus: identification at MR imaging with cine dispalcement encoding with stimulated echoes. Radiology. 2012;265(2):402–9.View ArticleGoogle Scholar
- Simpson RM, Keegan J, Firmin DN. MR assessment of regional myocardial mechanics. J Magn Reson Imaging. 2013;37(3):576–99.View ArticlePubMedGoogle Scholar
- Kim D, Gilson WD, Kramer CM, Epstein FH. Myocardial tissue tracking with two-dimensional cine displacement-encoded MR imaging: development and initial evaluation. Radiology. 2004;230(3):862–71.View ArticlePubMedGoogle Scholar
- Zhong XD, Spottiswoode BS, Meyer CH, Kramer CM, Epstein FH. Imaging three-dimensional myocardial mechanics using navigator-gated volumetric spiral cine DENSE MRI. Magn Reson Med. 2010;64(4):1089–97.View ArticlePubMedPubMed CentralGoogle Scholar
- Young AA, Li B, Kirton RS, Cowan BR. Generalized spatiotemporal myocardial strain analysis for DENSE and SPAMM imaging. Magn Reson Med. 2012;67(6):1590–9.View ArticlePubMedGoogle Scholar
- Spottiswoode BS, Zhong X, Lorenz CH, Mayosi BM, Meintjes EM, Epstein FH. Motion-guided segmentation for cine DENSE MRI. Med Image Anal. 2009;13(1):105–15.View ArticlePubMedGoogle Scholar
- Gilliam AD, Epstein FH. Automated motion estimation for 2-D cine DENSE MRI. IEEE Trans Med Imaging. 2012;31(9):1669–81.View ArticlePubMedPubMed CentralGoogle Scholar
- Spottiswoode BS, Zhong X, Hess AT, Kramer CM, Meintjes EM, Mayosi BM, Epstein FH. Tracking myocardial motion from cine DENSE images using spatiotemporal phase unwrapping and temporal fitting. IEEE Trans Med Imaging. 2007;26(1):15–30.View ArticlePubMedGoogle Scholar
- Zhong X, Helm PA, Epstein FH. Balanced multipoint displacement encoding for DENSE MRI. Magn Reson Med. 2009;61(4):981–8.View ArticlePubMedPubMed CentralGoogle Scholar
- Kim D, Kellman P. Improved cine displacement-encoded MRI using balanced steady-state free precession and time-adaptive sensitivity encoding parallel imaging at 3 T. NMR Biomed. 2007;20(6):591–601.View ArticlePubMedGoogle Scholar
- Lustig M, Donoho D, Pauly JM. Sparse MRI: The application of compressed sensing for rapid MR imaging. Magn Reson Med. 2007;58(6):1182–95.View ArticlePubMedGoogle Scholar
- Zhong X, Spottiswoode BS, Cowart EA, Gilson WD, Epstein FH. Selective suppression of artifact-generating echoes in cine DENSE using through-plane dephasing. Magn Reson Med. 2006;56:1126–31.View ArticlePubMedGoogle Scholar
- Chen X, Salerno M, Yang Y, Epstein FH. Motion-compensated compressed sensing for dynamic contrast-enhanced MRI using regional spatiotemporal sparsity and region tracking: Block low-rank sparsity with motion-guidance (BLOSM). Magn Reson Med. 2014;72(4):1028–38.View ArticlePubMedGoogle Scholar
- Pruessmann KP, Weiger M, Scheidegger MB, Boesiger P. SENSE: Sensitivity encoding for fast MRI. Magn Reson Med. 1999;42(5):952–62.View ArticlePubMedGoogle Scholar
- Fessler JA, Sutton BP. Nonuniform fast Fourier transforms using min-max interpolation. IEEE T Signal Proces. 2003;51(2):560–74.View ArticleGoogle Scholar
- Walsh DO, Gmitro AF, Marcellin MW. Adaptive reconstruction of phased array MR imagery. Magn Reson Med. 2000;43(5):682–90.View ArticlePubMedGoogle Scholar
- Feng L, Donnino R, Babb J, Axel L, Kim D. Numerical and in vivo validation of fast cine displacement-encoded with stimulated echoes (DENSE) MRI for quantification of regional cardiac function. Magn Reson Med. 2009;62(3):682–90.View ArticlePubMedPubMed CentralGoogle Scholar
- Epstein FH, Gilson WD. Displacement-encoded cardiac MRI using cosine and sine modulation to eliminate (CANSEL) artifact-generating echoes. Magn Reson Med. 2004;52(4):774–81.View ArticlePubMedGoogle Scholar
- Cerqueira MD, Weissman NJ, Dilsizian V, Jacobs AK, Kaul S, Laskey WK, Pennell DJ, Rumberger JA, Ryan T, Verani MS. Standardized myocardial segmentation and nomenclature for tomographic imaging of the heart. A statement for healthcare professionals from the Cardiac Imaging Committee of the Council on Clinical Cardiology of the American Heart Association. Circulation. 2002;105(4):539–42.View ArticlePubMedGoogle Scholar
- Zhang T, Cheng JY, Potnick AG, Barth RA, Alley MT, Uecker M, Lustig M, Pauly JM, Vasanawala SS. Fast pediatric 3D free-breathing abdominal dynamic contrast enhanced MRI with high spatiotemporal resolution. J Magn Reson Imaging. 2015;41(2):460–73.View ArticlePubMedGoogle Scholar
- Mehta BB, Chen X, Bilchick KC, Salerno M, Epstein FH. Accelerated and navigator-gated look-locker imaging for cardiac T1 Estimation (ANGIE): Development and application to T1 mapping of the right ventricle. Magn Reson Med. 2015;73(1):150–60.View ArticlePubMedGoogle Scholar
- Yoon H, Kim KS, Kim D, Bresler Y, Ye JC. Motion adaptive patch-based low-rank approach for compressed sensing cardiac cine MRI. IEEE Trans Med Imaging. 2014;33(11):2069–85.View ArticlePubMedGoogle Scholar
- Jung H, Sung K, Nayak KS, Kim EY, Ye JC. k-t FOCUSS: A General Compressed Sensing Framework for High Resolution Dynamic MRI. Magn Reson Med. 2009;61(1):103–16.View ArticlePubMedGoogle Scholar
- Kim D, Dyvorne HA, Otazo R, Feng L, Sodickson DK, Lee VS. Accelerated phase-contrast cine MRI using k-t SPARSE-SENSE. Magn Reson Med. 2012;67(4):1054–64.View ArticlePubMedGoogle Scholar
- Wehner GJ, Suever JD, Haggerty CM, Jing L, Powell DK, Hamlet SM, Grabau, J.D., Mojsejenko, W.D., Zhong, X., Epstein, F.H. and Fornwalt, B.K. Validation of in vivo 2D displacements from spiral cine DENSE at 3 T. J Cardiovasc Magn Reson. 2015;17(1):5.View ArticlePubMedPubMed CentralGoogle Scholar