Continued Development of Infant Brain Analysis Tools Abstract: The increasing availability of infant brain MR images, such as those that will be collected through the Baby Connectome Project (BCP, on which Dr. Shen is a Co-PI, focusing on data acquisition), affords unprecedented opportunities for precise charting of dynamic early brain developmental trajectories in understanding normative and aberrant growth. However, to fully benefit from these datasets, a major barrier that needs to be overcome is the critical lacking of computational tools for accurate processing and analysis of infant MRI data, which typically exhibit poor tissue contrast, large within tissue intensity variation, and regionally-heterogeneous and dynamic changes. To fill this critical gap, in 2012 we pioneered in creating an infant-centric MRI processing software package, called infant Brain Extraction and Analysis Tool (iBEAT), and a set of infant-specific atlases, called UNC 0-1-2 Infant Atlases, and further made them freely and publicly available via NITRC. Over the last 4 years, iBEAT and UNC 0-1-2 Infant Atlases have been downloaded 2900+ and 5600+ times, respectively, and contributed to 160+ independent research papers. As indicated by 30+ support letters, iBEAT is now driving the research for MRI studies of early brain development in many labs throughout the world. Results produced by iBEAT are also highlighted in the National Institute of Mental Health (NIMH)'s 2015-2020 Strategic Plan. This project is dedicated to the continuous development, hardening, and dissemination of iBEAT, by developing innovative software modules with comprehensive user support. To achieve this goal, we propose four aims. In Aim 1, we will create an innovative learning-based multi-source information integration framework for joint skull stripping and tissue segmentation for accurate structural measurements. Our method employs random forest to adaptively learn the optimal image appearance features from multimodality images and also informative context features from tissue probability maps. In Aim 2, we will construct longitudinal infant brain atlases at multiple time points (i.e., 1, 3, 6, 9, and 12 months of age) for both T1-/T2-weighted and diffusion-weighted MR images. We propose a longitudinally-consistent sparse representation technique to construct representative atlases with significantly improved structural details by explicitly dealing with possible misalignments between images even after registration. In Aim 3, we will develop a novel learning-based approach for cortical topology correction and integrate it, along with our infant-centric analysis tools and atlases for cortical surfaces, into iBEAT for precise mapping of dynamic and complex cortical changes in infants. Unlike existing tools that perform poorly for infant brains, we will incorporate infant-dedicated tools for topology correction, surface reconstruction, registration, parcellation, and measurements. We will further integ...