Spin echo versus stimulated echo diffusion tensor imaging of the in vivo human heart

Abstract

Purpose: To compare signal-to-noise ratio (SNR) efficiency and diffusion tensor metrics of cardiac diffusion tensor mapping using acceleration-compensated spin-echo (SE) and stimulated echo acquisition mode (STEAM) imaging.

Methods: Diffusion weighted SE and STEAM sequences were implemented on a clinical 1.5 Tesla MR system. The SNR efficiency of SE and STEAM was measured (b = 50-450 s/mm(2) ) in isotropic agar, anisotropic diffusion phantoms and the in vivo human heart. Diffusion tensor analysis was performed on mean diffusivity, fractional anisotropy, helix and transverse angles.

Results: In the isotropic phantom, the ratio of SNR efficiency for SE versus STEAM, SNRt (SE/STEAM), was 2.84 ± 0.08 for all tested b-values. In the anisotropic diffusion phantom the ratio decreased from 2.75 ± 0.05 to 2.20 ± 0.13 with increasing b-value, similar to the in vivo decrease from 2.91 ± 0.43 to 2.30 ± 0.30. Diffusion tensor analysis revealed reduced deviation of helix angles from a linear transmural model and reduced transverse angle standard deviation for SE compared with STEAM. Mean diffusivity and fractional anisotropy were measured to be statistically different (P < 0.001) between SE and STEAM.

Conclusion: Cardiac DTI using motion-compensated SE yields a 2.3-2.9× increase in SNR efficiency relative to STEAM and improved accuracy of tensor metrics. The SE method hence presents an attractive alternative to STEAM based approaches. Magn Reson Med 76:862-872, 2016. © 2015 The Authors. Magnetic Resonance in Medicine published by Wiley Periodicals, Inc. on behalf of International Society for Magnetic Resonance in Medicine. This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

Keywords: STEAM; cardiac diffusion tensor imaging; myocardial fiber architecture; signal-to-noise ratio; spin-echo, motion compensation.

Figure 1

Sequence diagrams. STEAM acquisition spanning over two consecutive heartbeats (a) and SE acquisition with second order motion compensated diffusion encoding gradients including VERSE (variable rate selective excitation) echo pulse (b). Fat suppression is achieved by a binomial excitation pulse. The excitation slab is tilted with respect to the 90° or 180° pulses to allow for reduced field of view imaging. Before the first 90° excitation, rest slabs are applied orthogonal to the field of view in phase encoding direction to suppress residually excited signal. For both sequences, the trigger delay was set to mid systole and images were encoded by identical single‐shot EPI readouts.

Figure 2

Theoretical SNR efficiency. The SNR efficiency ratio of SE versus STEAM is seen to decrease with increasing heart rate. At a heart rate of 60 min^‐1, b‐value = 450 s/mm², TR/TE = 1000 ms/70 ms (SE), TR/TM/TE = 1000 ms/985/31 ms (STEAM) the relative gain in SNR efficiency of SE versus STEAM is 3.5×.

Figure 3

In vivo data. a: Example dataset acquired at b = 100 s/mm² and 450 s/mm². b: Corresponding SNR efficiency maps for b = 450 s/mm².

Figure 4

SNR efficiency ratios of isotropic agar and anisotropic diffusion phantom as well as of in vivo measurements are shown for b‐values ranging from 50 to 450 s/mm². a: While the isotropic agar phantom shows no b‐value dependency, SNR_t(SE/STEAM) is seen to decrease with increasing diffusion weighting in the anisotropic diffusion phantom. b: In vivo data reveal b‐value dependency of SNR_t(SE/STEAM) similar to anisotropic diffusion phantom.

Figure 5

Helix and transverse angle maps. The linear decrease of helix angles from endo‐ to epicardium is visible, however more coherent in the SE case. Transverse angles are close to zero degrees for both sequences except for nonnegative values at the intersection of left and right ventricular structures and near the papillary muscles.

Figure 6

Sectorwise helix and transverse angle statistics. a: Comparison of transmural helix angle distribution for SE (blue) and STEAM (red) (solid box: 50% percentile, error bars: 90% percentile of the helix angle distribution in circumferential direction). Helix angle variations are more pronounced in STEAM, particularly at the endo‐ and epicardial region. b: Histograms of transverse angles show reduced dispersion of transverse angle for SE.

Figure 7

Bland‐Altman and line plots of MD (a) and FA (b) representing intra subject reproducibility of SE versus STEAM. No statistical differences were observed between repeat measurements for MD and FA for both SE and STEAM. Significant differences were found for MD and FA between both sequences.

Figure 8

Eigenvalue analysis. Diffusion tensor eigenvalue (e1, e2, e3) histograms for SE (solid line) and STEAM (dashed line). SE eigenvalues show a distinct and dense distribution, while STEAM eigenvalue histograms are broadened.