Digital Poster 5: Ex-vivo cardiac DTI: measure what matters - effects of varying diffusion time with spin echo and stimulated echo sequences

Recent advances in cardiac MR diffusion and improvements in gradient performance in MR scanners have enabled sub-millimeter spatial resolution ex vivo cardiac diffusion tensor imaging (cDTI) of large animal and human hearts, improving our understanding of cardiac microstructure [1,2].

These studies have used spin echo (SE) sequences that benefit from high signal to noise ratio (SNR) and can accurately depict the primary helical arrangement of cardiomyocytes in the left ventricular myocardium. However, due to its intrinsic short diffusion time (TD~20ms), SE-cDTI has been shown to be less suited to probing the secondary laminar arrangement of cardiomyocytes (sheetlets) in vivo [3].

cDTI data can also be acquired with stimulated echo (STEAM) sequences. In vivo, STEAM results in a longer diffusion time of one cardiac cycle (~1000ms) and has been shown to be superior to SE in the depiction of sheetlets and sheetlets dynamics in healthy volunteers [4]. However, STEAM suffers from lower SNR, limiting its adoption in high resolution ex vivo cDTI. In this work, we aim to study the effect of varying diffusion time on the diffusion tensor ex vivo and the ability of short diffusion time sequences to precisely characterize sheetlet orientation.



Methods:

A fixed healthy porcine heart (systolic arrest, 4 weeks immersed in 10% formalin) was scanned at 3T (MAGNETOM Cima.X, Siemens Healthineers AG, Forchheim, Germany) with the following protocol: b=150, 750 s/mm², 6 directions, 20 averages, TE/TR=40/6000ms, single slice, voxel size=3x3x8mm³ (to minimize the effects of noise). Scans were performed with a monopolar SE sequence (TD~20ms) and STEAM (TD 30-1000ms). Subsequently, high resolution scans were performed for SE and STEAM (TD=300ms) with 1.5mm isotropic voxels, 40 averages.

A diffusion tensor was fitted voxel-wise to each dataset using an in house developed tool, and maps of the components of the diffusion tensor (eigenvalues 𝜆1, 𝜆2, 𝜆3), fractional anisotropy (FA), mean diffusivity (MD), tensor mode [5], helix angle (HA) and absolute secondary eigenvector angle (|E2A|) were obtained for each dataset.



Results:

Eigenvalues decrease with increasing TD, resulting in a median MD decrease of 20% and median FA increase of 35% when increasing TD from ~20ms to 300ms, after which point these metrics plateau (Fig 1a). Tensor mode distribution remains similar across TDs, with increased variability at higher TDs likely due to reduced SNR. Fig 1b shows that 𝜆2/𝜆3 ratio increases with TD, suggesting that higher diffusion times might differentiate secondary diffusion directions better. Correspondingly, a shift toward higher values and broader distribution can be seen in the example histograms of 𝜆2/𝜆3 from SE to STEAM.

Example cDTI maps for high resolution SE and STEAM show apparent changes in MD and FA, whereas HA and |E2A| maps do not show evident visual differences (Fig 2). Finally, superquadric glyphs from both SE and STEAM data show that the higher MD and reduced FA of SE results in cube-like shaped glyphs, whereas STEAM produces more elongated glyphs as expected for tensors with a higher 𝜆2/𝜆3 (Fig 3).



Conclusion:

Despite having a lower 𝜆2/𝜆3 ratio, cDTI-SE provides a similar depiction of the primary and secondary cardiomyocyte structures to cDTI-STEAM, and can therefore provide and accurate characterization of sheetlet organization ex vivo at moderate spatial resolution. Future work will explore if this holds at the higher resolutions delivered by small-bore preclinical systems and for hearts arrested in diastole.