Computational Diffusion MRI for Studying Early Human Brain Development Abstract In the first years of life, the human brain develops dynamically in both structure and function. Many neurodevel- opmental disorders are associated with aberrations from normative growth during this critical period of early brain development. The increasing availability of longitudinal baby MRI data, such as those acquired through the Baby Connectome Project (BCP), affords unprecedented opportunity for precise charting of early brain developmental trajectories in order to understand normative and aberrant growth. Dedicated computational tools are needed for accurate processing and analysis of baby MR images, which typically exhibit dynamic heterogeneous changes across time. The goal of this project is to equip brain researchers with computational tools effective for studying the early developing human brain in terms of tissue microstructure and white matter pathways using diffusion MRI. We propose three aims. In Aim 1, we will develop computational tools for effective estimation of white matter pathways in the baby brain via diffusion tractography. We will tackle the challenge of tracking through regions with low diffusion anisotropy owing to ongoing myelination in the developing brain. Our tools will allow proper characterization of complex white matter pathway patterns such as fanning and bending. This will allow solving the gyral bias problem ubiquitous in existing tractography algorithms with fiber streamlines terminating predomi- nantly at gyral crowns but not sulcal banks. Our tools will allow tracing of cortico-cortical and cortico-subcortical pathways with more uniform coverage of the cortex. In Aim 2, we will develop microstructural analysis meth- ods that are unconfounded by complex fiber configurations, such as crossing, bending, branching, kissing, and fanning, allowing more accurate and specific characterization of changes in tissue microarchitecture during early brain development. In Aim 3, we will develop techniques that will allow diffusion MRI data collected at multiple sites, which are very common in the era of big data, to be harmonized to mitigate the negative effects of inter-site variability. Unlike existing methods that harmonize derived quantities such as fractional anisotropy, our method can be applied directly to the diffusion-weighted images, allowing measurements based on microstructure and connectivity to be subsequently computed for consistent analysis. We will also develop deep learning tools for multi-shell data prediction so that diffusion MRI data collected with different numbers of shells can be harmonized. Successful completion of this project will empower the neuroscience community with computational tools to better chart the normative early development of the human brain using diffusion MRI. The developed tools will also enable quantitative brain examinations of children who are affected by neurological developmental disorders.