Reliability on multiband diffusion NODDI models: A test retest study on children and adults

High-frequency longitudinal white matter diffusion- & myelin-based MRI database: reliability and variability

“…Indeed, despite the additional thresholding applied on ISOvf, this measure systematically presents the largest CV in all the bundles, which is consistent with studies that suggest that ISOvf is a poorly reliable parameter (Tariq, 2013, 2018). Finally, our findings of overall greater CVb for most measures compared with their corresponding CVw are in accordance with other DTI- and NODDI-based reliability studies (Andica et al, 2020; Chung et al, 2016; Lucignani et al, 2021; Tariq, 2013; Veenith et al, 2013). This result is expected, as there will be greater microstructural heterogeneity in a localized region across a population compared with multiple observations within the same subject.…”

“…Together, this suggests that there is a need for further validation of this measure before adopting it in longitudinal studies. On the other hand, even though NODDI measurements are inherently noisier than DTI measurements for white matter modelling - likely due to a more complex model and requiring high b-value data - within- and between-variability for ICvf and ECvf were lower or comparable to the FA variability in most bundles (Andica et al, 2020; Chung et al, 2016; Lucignani et al, 2021). However, this sensitivity to noise inherent to NODDI could explain the higher within-variability of ISOvf and OD.…”

“…Since ISOvf accounts for the isotropic volume fraction, this parameter has generally very low values in the WM (Tariq, 2018), and many studies emphasized the poor reliability of this NODDI parameter (Andica et al, 2020; Chung et al, 2016; Lehmann et al, 2021; Lucignani et al, 2021). To improve the consistency of this parameter, we proposed an evaluation of different thresholds to remove values close to zero.…”

“…Among DTI-derived measures, Fractional anisotropy (FA) and Mean Diffusivity (MD) generally show the highest reliability across different WM regions (Acheson et al, 2017; Hakulinen et al, 2021; Luque Laguna et al, 2020; Palacios et al, 2017; Shahim et al, 2017; Thieleking et al, 2021; Zhou et al, 2018). For NODDI-derived measures, studies reported similar (intracellular volume fraction, ICvf) or higher (orientation dispersion, OD) reliability compared to DTI measures, while isotropic volume fraction (ISOvf) showed the poorest reliability (ICC<0.6) (Andica et al, 2020; Chung et al, 2016; Granberg et al, 2017; Lucignani et al, 2021; Tariq, 2013). In contrast, to the best of our knowledge, the reliability of HARDI-derived measurements such as apparent fiber density (AFD, D. Raffelt et al, 2012) and the number of fiber orientations (NuFO, Dell’Acqua et al, 2013a) has not been yet examined in healthy subjects or clinical population.…”

Machine learning classification of chronic traumatic brain injury using diffusion tensor imaging and NODDI: A replication and extension study

“…The ODI and Viso compartment data were warped to the same standard FMRIB FA template as the voxel-based FA using FSL’s applywarp routine, and then masked with a WM segmentation map, to only include WM voxels, with age regressed out of the remaining voxels, and residual maps used for subsequent classification analyses. Previously published studies suggest that NODDI metrics, including ODI and Viso, are reliable using single-shell, low b-value data ( Cottaar et al, 2019 ; Lucignani et al, 2021 ; Timmers et al, 2016 ; Zhang et al, 2012 ) and through the use of MRI scanners with low field strength ( Chung et al, 2016 ).…”