Advantages of Multi-shell Diffusion Models for Studies of Brain Development in Youth

Diffusion-weighted imaging (DWI) has advanced our understanding of how brain microstructure evolves during neurodevelopment. Most existing studies of brain development use a simple scalar measure - fractional anisotropy (FA) - to characterize brain tissue microstructure. However, advances in the modeling of diffusion images, such as Neurite Orientation Dispersion and Density Imaging (NODDI) and Laplacian-regularized Mean Apparent Propagator MRI (MAPL), allow for greater richness in how microstructure is characterized. Nonetheless, the differential utility of these new diffusion models for understanding brain development remains sparsely investigated. Additionally, despite ample evidence that motion artifact is a major confound for studies of brain development, the relative vulnerability of these models to in-scanner motion has not been described. Accordingly, in a sample of 123 youth (ages 12-30) we evaluated FA in comparison to two recently introduced diffusion metrics: intra-cellular volume fraction (ICVF) from NODDI and return-to-origin probability (RTOP) from MAPL. Associations with age were evaluated for each of these measures at multiple scales, including mean white matter scalar values, voxelwise analyses, and tractography-based networks. Additionally, we examined the association between in-scanner head motion and each of these measures. Our results reveal that ICVF and RTOP are more associated with development than FA, and are less impacted by motion. Collectively, our findings suggest that multi-shell diffusion models confer notable advantages over FA for studies of the developing brain.