Longitudinal Reproducibility of Neurite Orientation Dispersion and Density Imaging (NODDI) Derived Metrics in the White Matter

Diffusion-weighted magnetic resonance imaging (DWI) is undergoing constant evolution with the ambitious goal of developing in-vivo histology brain. A recent methodological advancement Neurite Orientation Dispersion and Density Imaging (NODDI), a histologically validated multi-compartment model to yield microstructural features brain tissue such as geometric complexity neurite packing density, which are especially useful in white matter. Since NODDI increasingly popular clinical research fields developmental neuroscience neuroplasticity, it vast importance characterize its reproducibility (or reliability). We acquired multi-shell DWI data 29 healthy young subjects twice over rescan interval 4 weeks assess within-subject coefficient variation (CVWS), between-subject (CVBS) intraclass correlation (ICC), respectively. Using these metrics, we compared regional voxel-by-voxel most common image analysis approaches (tract-based spatial statistics [TBSS], voxel-based different extents smoothing [“VBM-style”], ROI-based analysis). observed high test–retest for orientation dispersion index (ODI) slightly worse results density (NDI). Our findings also suggest that choice approach might have significant consequences study. Collectively, Gaussian kernels ≥4 mm FWHM ROI-averaging yielded highest across NDI ODI maps (CVWS mostly ≤3%, ICC ≥0.8), respectively, whilst smaller TBSS performed consistently worse. Furthermore, demonstrate quality (signal-to-noise ratio [SNR]) an important determinant metric reproducibility. discuss implications longitudinal cross-sectional designs commonly employed neuroimaging field.