Digital Poster 1: Diffusion Tensor Imaging of the Human Heart in Vivo: Impact of Spatial Resolution on Helix Angle Range and Slope

Background: Diffusion tensor imaging (DTI) of the heart has been performed in vivo at a range of in-plane spatial resolutions (0.85mm [1] to 2.5mm [2] for M2 spin echo) and 2.8 mm (STEAM [3]). Helix angle (HA) range and slope are frequently reported in these studies, but the impact of spatial resolution on these metrics has not been systematically studied. We recently reported the ability to acquire DTI images of the human heart in vivo with an acquired (non-interpolated) in-plane spatial resolution of 0.85mm x 0.85mm [1]. We aimed here to systematically downsample these data to evaluate HA range and slope as a function of spatial resolution.

Methods: DTI images were acquired in a cohort of healthy subjects (n=19) on a 3T scanner with 80mT/m gradients (Prisma, Siemens) using a free-breathing M2-compensated spin echo sequence and a custom 64-channel coil. Parameters included: acquired resolution 8mm x 0.85mm x 0.85mm, b-value 500, 12 directions, 12 averages, TE 76ms [1]. Analysis of the diffusion tensor produced HA maps with an in-plane resolution of 0.85mm x 0.85mm, which were systematically downsampled via bi-cubic interpolation to produce HA maps at the following resolutions: 1.2mm, 1.5mm, 1.8mm, 2.2mm, 2.5mm, and 2.8mm. In each subject, 3-5 mid-ventricular images were selected for analysis. The endocardial and epicardial borders were determined using the 0.85mm resolution scans and the resulting masks were scaled down to the various resolutions. HA range and HA slope were calculated for each image, averaged on a per-subject basis, and compared to the 0.85mm resolution images. Statistical significance was determined by a 1-way ANOVA with Dunnett’s post-test.

Results: HA maps at 0.85mm and 1.5mm resolution allowed fine detail to be detected (Fig. 1-2). In contrast, at the resolution range frequently used in vivo (2.2-2.8mm), the maps appeared highly pixelated and under-resolved (Fig. 1-2). The number of voxels from endocardium to epicardium decreased from 12.8±1.8 voxels at 0.85mm resolution to 3.9±0.5 voxels at 2.8mm resolution (Fig. 3A). HA range decreased linearly with resolution (105±7°at 0.85mm to 80±10° at 2.8mm, Fig. 3B). HA slope showed no significant changes between 0.85mm and 1.5mm, but thereafter decreased (1.10±0.07°/% at 0.85mm to 1.00±0.09°/% at 2.8mm, Fig. 3C-D).

Conclusion: Our data show that HA range and spatial resolution have a linear relationship, with even small changes in spatial resolution impacting HA range. Of note, downsampling from 0.85mm to 2.8mm resulted in a reduction of mean HA range from 105° to 80°. The impact of spatial resolution on HA slope was not significant below 1.5mm, but thereafter showed a consistent reduction with decreased resolution. Further study will be needed on the effect of scanner-based interpolation, however, the values we obtain here are similar to those previously reported with scanner-based interpolation [3]. Our results suggest that in vivo DTI should ideally be performed with an acquired in-plane resolution of 1.5 mm or less.