1345589 - Evaluation of Distortion Correction for EPI-based Cardiac Diffusion Tensor Imaging

Cardiac diffusion tensor imaging (cDTI) has demonstrated the ability to estimate the underlying microstructural organization of the myocardium [1-3]. Single-shot echo-planar imaging (EPI) is the preferred acquisition strategy due to its fast and efficient coverage of k-space. However, EPI suffers from geometric distortion caused by B₀ field inhomogeneities due to tissue susceptibility differences [4]. Image distortions can impact the quantitative accuracy of cDTI. FSL TOPUP [5] and TORTOISE DR BUDDI [6] are two techniques used to mitigate geometric distortion by acquiring two datasets with opposing phase encoding polarities. Both techniques have been evaluated for brain DTI, but only TOPUP has been evaluated in cDTI [7]. Herein, we analyze the impact of distortion correction on qualitative and quantitative measures of cDTI.

Methods:

10 healthy volunteers (25.5±2.0 years) were imaged (IRB approved, consented) at 3T (Skyra, Siemens) for three (basal, mid-ventricular, and apical) slices of the heart. Balanced steady-state free precession images were acquired at a mid-systolic cardiac phase to provide a distortion-less reference image (resolution=1×1×8mm³). A free-breathing, ECG-gated, M₁+M₂-compensated spin echo EPI cDTI sequence [2] was used to acquire blip-up (BU) and blip-down (BD) datasets (TE=71ms, TR=3×R-R intervals, BW=1776Hz/px, resolution=2×2×8mm³, b-values=[0,350]s/mm², six diffusion directions, and 10 averages). Five BU averages and five BD averages were used for distortion correction to ensure equivalency of SNR between all datasets. Segmentations were conducted using the mean diffusivity (MD) and primary eigenvector to exclude papillary muscles. Distortion correction results were compared using the Dice similarity coefficient (DSC), average Hausdorff distance (AHD), and maps of MD, fractional anisotropy (FA), and helix angle (HA).

Results:

Fig. 1A shows a qualitative assessment of DSC for two different volunteers relative to the distortion-less reference image. Fig. 1B presents an analysis of DSC for BU, BD, TOPUP (TU), and DR BUDDI (DB) compared to the distortion-less reference image. TU and DB generally outperform traditional BU or BD acquisition schemes. All statistics for DSC are reported in Fig 1B. Fig. 2A shows a qualitative assessment of AHD for two different volunteers relative to the distortion-less reference image. Fig. 2B presents an analysis of AHD for both epicardial and endocardial contours for BU, BD, TU, and DB compared to the distortion-less reference image. TU and DB generally outperform traditional BU or BD acquisition schemes. All statistics for contours are reported in Fig 2B. Fig. 3 shows MD, FA, and HA maps for one volunteer. Histograms for MD and FA are plotted with a median line.

Conclusion:

TOPUP and DR BUDDI provided similar distortion-free cDTI images. A more comprehensive quantitative investigation of MD, FA, and HA maps is required to delineate differences in non-shape characterizing metrics.