In Vivo Cardiac DTI on a 3T Clinical Scanner: An Optimized M2 Approach
Christopher Nguyen1, Zhaoyang Fan1, Yibin Xie1, Jianing Pang1, Xiaoming Bi2, Peter Speier3, Jon Kobashigawa4, and Debiao Li1,5

1Biomedical Imaging Research Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, 2MR R&D, Siemens Healthcare, Los Angeles, CA, United States, 3Siemens Healthcare GmbH, Erlangen, Germany, 4Heart Institute, Cedars-Sinai Medical Center, Los Angeles, CA, United States, 5Bioengineering, University of California Los Angeles, Los Angeles, CA, United States


Optimized second order motion compensated (M2) diffusion tensor prepared cardiac magnetic resonance (DT-CMR) was applied in healthy volunteers and heart failure patients at 3T. The pulse sequence design focused on B1 robustness at high main field. In healthy volunteers, the proposed M2 DT-CMR was compared to zero order (M0) and first order (M1) motion compensations. In addition, heart rate dependency of the proposed M2 DT-CMR was explored with contextual comparison to M0 and M1. M2 DT-CMR was the only technique capable of application in heart failure patients without bulk motion artifacts.


Currently, there are only two main methods to perform diffusion tensor CMR (DT-CMR) that either rely on the subject exhibiting stable, periodic RR cycle (stimulated echo [1]) or utilize specialized research scanners that have ultra-high gradient strengths (spin-echo [2]). Recent work has demonstrated that gradient moment nulling (GMN) of the second order is capable of yielding robust diffusion weighted images (DWI) [3]. To extend this work, we present a novel DT-CMR sequence prototype that utilizes a M2 GMN gradient scheme that is robust to imperfect B1 refocusing at high main fields (≥3T). We compare this with no GMN compensation (M0) and first order GMN compensation (M1). Patients with advanced heart failure (HF) were also scanned to test its ability in a clinical setting.


Twenty healthy subjects and five HF patients were recruited and consented under Institutional Review Board. All subjects were scanned on a 3T Siemens (MAGNETOM Verio, Siemens Healthcare GmbH, Erlangen) with the following protocol: standard morphological localizers and prototype sequences implementing 3 DTI scans (b30 + 6 directions b = 300 s/mm^2, free breathing prospective navigator gating, bSSFP readout, 2.7x2.7x8mm3, flip angle = 90°, single-shot + MoCo) utilizing M0 (TEprep = 35ms), M1 (TEprep = 46ms), and M2 (TEprep = 67ms). Acquisition was carried out during the quiescent period of diastole. Gradient amplitudes were set to 60.8 mT/m (two 43 mT/m max gradients simultaneously on).

M2 was achieved with a dual tripolar pulse that is completely balanced before and after the refocusing pulse. This allows for motion compensation to be robust against imperfect B1 refocusing since only nulling of M2 depends on robust RF refocusing (M0 and M2 are nulled via gradients alone). The single refocusing pulse was a composite-adiabatic pulse consisting of two hard pulses straddling a single BIR-4 pulse in the fashion of a MLEV configuration (90x-180y-90x). In addition, a single crusher gradient was played out preceding the refocusing pulse and then again unwound during the bSSFP readout [8]. This crusher gradient eliminates signal not dictated by the diffusion preparation, which may arise due to imperfect B1 and T1 recovery. Numerical simulation was carried out to demonstrate how much motion robustness could depend on imperfect B1 refocusing. Quiescent (velocity: 1.5cm/s, acceleration: 10mm^2/s) and peak (velocity: 15 cm/s, acceleration: 100 mm^2/s) phases were simulated between two M2 encodings (quadra-bipolar and proposed dual-tripolar).

DTI reconstruction utilized custom software developed in Python using the DIPY library [6] to generate mean diffusivity (MD), fractional anisotropy (FA), and helix angle (HA) maps. Success rates defined by >90% of the myocardium unaffected by motion was reported. Paired t-tests were utilized to statistically test for significance (p<0.05).


For mildly low heart rates (HR) (< 75 beats-per-min) in volunteers, M2 was shown to have significantly (p < 0.05) higher success rates (93%) than M1 (62%) and M0 (28%). For higher HR, M2 was still significantly (p < 0.05) higher success rates (57%) than M1 (23%) and M0 ( 7%), but much notably lower success than at lower HR. Among the scans with minimal motion artifacts, MD and FA were significantly (p<0.05) lower for M2 (1.4±0.2 μm^2/ms, 0.3±0.2) than M0 (4.8±1.3 μm^2/ms, 0.8±0.6) and M1 (1.8±0.2 μm^2/ms, 0.3±0.2) with M2 values being consistent with previous literature [1,2].

Quiescent periods in patients were significantly (p < 0.01) shorter than in healthy volunteers (56±5ms vs 120±30ms, respectively). Despite the higher heart rates (85±8 BPM) in HF patients, M2 alone was only capable of yielding motion-artifact free MD, FA, and HA maps.


The proposed B1 resistant M2 dual-tripolar pulse combined with composite adiabatic refocusing and crusher gradient dephasing scheme yielded more robust MD, FA, and HA maps in volunteers compared with M0 and M. Moreover, M0, M1, and M2 all depended on heart rate experiencing more motion robustness at lower heart rates in healthy volunteers. This is most likely due to the shorter quiescent periods and more severe motion exhibited in volunteers with higher heart rates. In patients, the high heart rate did not impede on the performance of the proposed M2 despite the short quiescent periods. However, these patients also exhibited poor ejection fraction (<30%) possibly reflecting an overall decreased and less complex bulk motion in which M2 may be sufficient in compensating. This needs to be further investigated alongside with strain, velocity, and acceleration myocardial mapping.


The proposed M2 was shown to be more motion robust than M1 and M0 compensation despite the shorter motion sensitivity periods. The proposed DT-CMR was the only method able to provide motion-free DT-CMR images in HF patients.


NIH 1F31EB018152-01A1


1. 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 vivo diffusion tensor MRI of the human heart: Reproducibility of breath-hold and navigator-based approaches. Magn Reson Med 2012, 70:454–465.

2. Gamper U, Boesiger P, Kozerke S: Diffusion imaging of the in vivo heart using spin echoes--considerations on bulk motion sensitivity. Magn Reson Med 2007, 57:331–337.

3. Nguyen C, Fan Z, Sharif B, He Y, Dharmakumar R, Berman DS, Li D: In vivo three-dimensional high resolution cardiac diffusion-weighted MRI: a motion compensated diffusion-prepared balanced steady-state free precession approach. Magn Reson Med 2014, 72:1257–1267.

4. Tannús A, Garwood M: Adiabatic pulses. NMR Biomed 1997, 10:423–434.

5. Lin AP, Bennett E, Wisk LE, Gharib M, Fraser SE, Wen H: Circumferential strain in the wall of the common carotid artery: Comparing displacement-encoded and cine MRI in volunteers. Magn Reson Med 2008, 60:8–13.

6. Garyfallidis E, Brett M, Amirbekian B, Nguyen CT, Yeh F-C, Olivetti E, Halchenko Y, Nimmo-Smith I: Dipy–a novel software library for diffusion MR and tractography. 17th Annual Meeting of the Orgnization for Human Brain Mapping 2011.

7. Tunnicliffe EM, Suttie J, Ariga R, Neubauer S, Robson MD: Myocardial diffusion tensor imaging using diffusion-prepared SSFP. Journal of Cardiovascular Magnetic Resonance 2013, 15:P1.

8. Parrish T, Hu X: A new T2 preparation technique for ultrafast gradient-echo sequence. Magn Reson Med 1994, 32:652–657.


Figure 1 – Pulse sequence diagram of the (a) novel dual tri-polar M2 diffusion preparation with crusher gradients [7] to provide additional robustness to 3T B1 inhomogeneity. (b) M0 and (c) M1 diffusion preparations used for comparison. (d) composite adiabatic refocusing pulse used.

Figure 2 – Numerical simulation demonstrating increased motion robustness in the presence of imperfect B1 refocusing between two M2 GMN quadra-bipolar and dual-tripolar. Quadra-bipolar relies on perfect refocusing to null M1 and M2, while dual-tripolar only relies on perfect refocusing to null M2. The proposed diffusion preparation’s GMN only has the B1 dependency on the M2 term while more conventional M2 diffusion encoding has a B1 dependency on M0, M1, and M2.

Figure 3 – Representative images from a healthy volunteer (HR = 65) of (A) least diffusion weighted (b30) and higher diffusion weighted image (b300) comparing M0, M1, and M2. (B) Representative MD, FA, and HA maps for HF patient (top row) and normal volunteer (bottom row).

Proc. Intl. Soc. Mag. Reson. Med. 24 (2016)