ABSTRACT Quantitative imaging biomarkers (QIBs) are objective measures derived from quantitative imaging that can differentiate pathological changes from healthy biological processes. Diffusion MRI parameters derived from Diffusion Tensor Imaging (DTI) and Neurite Orientation Dispersion and Density Imaging (NODDI) could serve as potential QIBs for studying both healthy neurodevelopment and various neurological conditions. However, quantitative neuroimaging studies often require large datasets collected across multiple scanners, which introduces variability. To ensure the reliability of multi‐centre studies, the inter‐centre reproducibility of DTI and NODDI parameters must be thoroughly assessed before data collection begins. Discrepancies between results reported by previous studies can be explained by other sources of variability. The inter‐scanner reproducibility of diffusion parameters needs to be determined when the other sources of variability, such as differences in acquisition parameters, processing and ROI segmentation are controlled for. We assess the reproducibility of DTI and NODDI parameters in clinically relevant white matter (WM) tracts across three scanners of the same model, ensuring consistency in the acquisition scheme and pre‐processing pipelines. WM tract regions of interest (ROIs) are automatically segmented to standardise the analysis. Additionally, we investigate ROI and signal‐to‐noise ratio differences to better understand the sources of variability in diffusion parameters. According to the Koo and Li classification system, our results demonstrate excellent reproducibility for fractional anisotropy and mean diffusivity across scanners of the same model (ICC ≥ 0.964) when using identical acquisition schemes, pre‐processing pipelines and automated ROI segmentation. NODDI orientation dispersion index and neurite density index exhibit a similar level of reproducibility (ICC ≥ 0.942 and ICC ≥ 0.911, respectively), while free water fraction (FWF) has ICC ≥ 0.862. However, statistically significant variability was observed in the FWF, specifically within the left inferior fronto‐occipital fasciculus (CoV 9.43%) and optic radiation (CoV 9.95%), even when scanning the same cohort across sites. If there is an error in the signal fraction in one compartment in the NODDI model, the signal fractions from other compartments may likely be misestimated. The reproducibility and variability of diffusion parameters reported in this study provide guidance for future QIB research involving datasets derived from multiple scanners. These findings can help determine whether observed changes in diffusion parameters reflect meaningful biological differences or are highly influenced by measurement variability.
Sierhej et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: