In vivo cardiac DTI (cDTI) using Stimulated Echo Acquisition Mode (STEAM) imaging has gained significant attention^1-3. To address the sensitivity to motion and the limited signal-to-noise ratio of cDTI using STEAM, spin-echo based approaches have been pursued^4-6. Recently, second-order motion compensated diffusion gradient waveforms have been proposed for spin-echo cDTI^7-9 and the increase in SNR efficiency relative to STEAM has been demonstrated¹⁰. In addition, the single-shot method facilitates imaging during free-breathing enhancing applicability in a clinical setting. So far, however, second-order motion-compensated spin echo cDTI has only been demonstrated on MR scanners with high performance gradient systems. It is the objective of the present work to investigate minimum gradient performance requirements for second-order motion-compensated spin echo cDTI using in-vivo data.Second-order motion-compensated spin echo cDTI (Figure1) was implemented on a 1.5T Philips Achieva system (Philips Healthcare, Best, The Netherlands) equipped with a 5-channel cardiac receiver array and a gradient system delivering up to 80mT/m @ 100mT/m/ms. Data was acquired with ECG triggering and during navigator gated (7 mm gating window) free breathing. The maximum gradient strength (G_max) per physical axis was limited to 30mT/m, 40mT/m, 60mT/m and 74.5mT/m, the latter being the maximum applicable gradient strength due to peripheral nerve stimulation limits. Diffusion encoding was performed in three orthogonal directions at a b-value of 100s/mm² to suppress perfusion effects¹¹ and along nine directions at b=450 s/mm² optimized to allow for an effective G_max of 42.4mT/m, 56.6mT/m, 84.9mT/m and 105.4mT/m by utilizing the three physical gradient axes simultaneously. Corresponding TEs were 96ms, 85ms, 73ms and 66ms. Imaging parameters were: resolution 2.6×2.6mm2, slice thickness 7mm, reduced FOV¹² 230×104mm2, signal averages 12, TR=3 R-R. Three short-axis slices at basal, mid-ventricular and apical level were acquired at a trigger delay of 50% end systole. In vivo data were acquired in six healthy volunteers (1 male, age 24±2, weight 59±5kg, heart rate 70±10beats/min, max/min 55/90beats/min). Written informed consent according to institutional guidelines was obtained prior to imaging. The myocardium was manually segmented according to the AHA model with three transmural layers¹³ and helix and transverse angles were calculated¹⁴. Segment-wise comparison was performed by correlation and Bland-Altman analysis. Correlation coefficients, mean signed differences, mean absolute differences and root-mean-squared-errors (RMSE) are reported. Mean diffusivity (MD), fractional anisotropy (FA), transmural helix angle slope and range as well as mean and standard deviation of the transverse angle were computed. To model the effect of reduced SNR due to prolonged TE a subset of signal averages from data acquired with 74.5mT/m was analyzed.

Figure 2 shows helix angle maps acquired with different G_max as well as a segment-wise comparison in one volunteer. The corresponding transverse angle analysis is shown in Figure 3. Imaging at 30mT/m failed in two volunteers resulting in non-circumferential alignment of the diffusion tensors’ principal eigenvector (up to RMSE of 20.5° for helix and transverse angulation). The segment-wise comparison for different G_max at apex, mid and base across all volunteers is presented in Figure 4. Across all slices a correlation coefficient (R²) of 0.89±0.06, 0.87±0.04 and 0.77±0.11 for 60mT/m, 40mT/m and 30mT/m was found relative to data acquired at G_max=74.5mT/m. The corresponding RMSEs were 10.3±2.4°, 11.0±1.8°, 15.0±3.7° for helix angles and 8.1±2.2°, 9.0±2.0°, 11.7±3.8° for transverse angles. The (bias) signed mean difference was -3.0±2.0°, -1.9±2.9°, -4.6±4.0° for helix angles and 0.5±2.3°, -0.5±2.8°, -1.2±3.4° for transverse angles across all slices. The RMSE was lowest at mid ventricular level. The SNR analysis resulted in an RMSE of 3.9±2.8°/4.9±2.8°/7.0±4.6° and 2.6±1.4°/3.6±1.7°/4.9±2.0° for helix and transverse angles for G_max 30mT/m, 40mT/m, 60mT/m. The slice wise analysis of transmural helix angle variation, transverse angulation, MD and FA is shown in Table 1.

In this work, two confounding factors relating to the maximum available gradient amplitude for spin-echo cDTI were studied. SNR simulations showed that for helix angle estimation 43% of the RMSE and for transverse angle estimation 39% of the RMSE are attributed to the lower SNR with increased TE. This indicates that residual motion, registration and segmentation induced errors are present. To further minimize the impact of cardiac motion during long encoding times the trigger delay may be optimized for individual subjects¹⁵. Furthermore, spatial resolution may be increased reducing intra voxel dephasing at the cost of SNR. In general, G_max of 30mT/m appeared to be insufficient for robust spin-echo cardiac DTI.Second-order motion compensated diffusion gradients allow for spin-echo cardiac DTI during free breathing on clinical MR systems with standard gradient performance of 40 mT/m.