Effect of number of diffusion encoding directions in Neonatal Diffusion Tensor Imaging using Tract-Based Spatial Statistical analysis

Abstract

Diffusion Tensor Imaging (DTI) has been used to study the developing brain in early childhood, infants and in utero studies. In infants, number of used diffusion encoding directions has traditionally been smaller in earlier studies down to the minimum of 6 orthogonal directions. While the more recent studies often involve more directions, number of used directions remains an issue when acquisition time is optimized without compromising on data quality and in retrospective studies. Variability in the number of used directions may introduce bias and uncertainties to the DTI scalar estimates that affect cross-sectional and longitudinal study of the brain. We analyzed DTI images of 133 neonates, each data having 54 directions after quality control, to evaluate the effect of number of diffusion weighting directions from 6 to 54 with interval of 6 to the DTI scalars with Tract-based spatial statistics (TBSS) analysis. The TBSS analysis was applied to DTI scalar maps, and the mean Region of Interest (ROI) values were extracted using JHU atlas. We found significant bias in ROI mean values when only 6 directions were used (positive in FA, negative in MD, AD, RD), while when using 24 directions and above, the difference to scalar values calculated from 54 direction DTI was negligible. Using DTI measurements from data with at least 24 directions may be used in comparisons with DTI measurements from data with higher numbers of directions.