Systematic Assessment of the Impact of DTI Methodology on Fractional Anisotropy Measures in Alzheimer’s Disease

Abstract

White matter microstructural changes in Alzheimer’s disease (AD) are often assessed using fractional anisotropy (FA) obtained from diffusion tensor imaging (DTI). FA depends on the acquisition and analysis methods, including the fitting algorithm. In this study, we compared FA maps from different acquisitions and fitting algorithms in AD, mild cognitive impairment (MCI), and healthy controls (HCs) using the Alzheimer’s Disease Neuroimaging Initiative (ADNI) database. Three acquisitions from two vendors were compared (Siemens 30, GE 48, and Siemens 54 directions). DTI data were fit using nine fitting algorithms (four linear least squares (LLS), two weighted LLS (WLLS), and three non-linear LLS (NLLS) from four software tools (FSL, DSI-Studio, CAMINO, and AFNI). Different cluster volumes and effect-sizes were observed across acquisitions and fits, but higher consistency was observed as the number of diffusion directions increased. Significant differences were observed between HC and AD groups for all acquisitions, while significant differences between HC and MCI groups were only observed for GE48 and SI54. Using the intraclass correlation coefficient, AFNI–LLS and CAMINO–RESTORE were the least consistent with the other algorithms. By combining data across all three acquisitions and nine fits, differences between AD and HC/MCI groups were observed in the fornix and corpus callosum, indicating FA differences in these regions may be robust DTI-based biomarkers. This study demonstrates that comparisons of FA across aging populations could be confounded by variability in acquisitions and fit methodologies and that identifying the most robust DTI methodology is critical to provide more reliable DTI-based neuroimaging biomarkers for assessing microstructural changes in AD.