An exploration of diffusion tensor eigenvector variability within human calf muscles

Purpose: To explore the effect of diffusion tensor imaging (DTI) acquisition parameters on principal and minor eigenvector stability within human lower leg skeletal muscles. Materials and Methods: Lower leg muscles were evaluated in seven healthy subjects at 3T using an 8-channel transmit/receive coil. Diffusion-encoding was performed with nine signal averages (NSA) using 6, 15, and 25 directions (NDD). Individual DTI volumes were combined into aggregate volumes of 3, 2, and 1 NSA according to number of directions. Tensor eigenvalues ((1), (2), (3)), eigenvectors (epsilon(1), epsilon(2), epsilon(3)), and DTI metrics (fractional anisotropy [FA] and mean diffusivity [MD]) were calculated for each combination of NSA and NDD. Spatial maps of signal-to-noise ratio (SNR), (3):(2) ratio, and zenith angle were also calculated for region of interest (ROI) analysis of vector orientation consistency. Results: epsilon(1) variability was only moderately related to epsilon(2) variability (r = 0.4045). Variation of epsilon(1) was affected by NDD, not NSA (P < 0.0002), while variation of epsilon(2) was affected by NSA, not NDD (P < 0.0003). In terms of tensor shape, vector variability was weakly related to FA (epsilon(1):r = -0.1854, epsilon(2): ns), but had a stronger relation to the (3):(2) ratio (epsilon(1):r = -0.5221, epsilon(2):r = -0.1771). Vector variability was also weakly related to SNR (epsilon(1):r = -0.2873, epsilon(2):r = -0.3483). Zenith angle was found to be strongly associated with variability of epsilon(1) (r = 0.8048) but only weakly with that of epsilon(2) (r = 0.2135). Conclusion: The second eigenvector (epsilon(2)) displayed higher directional variability relative to epsilon(1), and was only marginally affected by experimental conditions that impacted epsilon(1) variability.