Abstract The diagnosis of neuropsychiatric disorders such as autism spectrum disorder (ASD) is based on behavioral assessments that cannot be confirmed until 2-3 years of age. ASD affects 1 in 59 children in the U.S., and infants from high-risk families have 20 times of risk to develop ASD compared to the general population. Current paradigm misses the precious time window for potential early intervention from birth to 2-3 years of age. Therefore, the focus in ASD research is shifting toward developing early biomarkers which can predict the risk of an infant developing future behavioral abnormalities, while the infant is still in pre- symptomatic stage. Imaging markers play major roles in understanding of both ASD and typical developing (TD) brains. Neural MRI at early infancy may improve the current paradigm for diagnosis for ASD in high-risk infants. Macrostructural measurements such as cortical thickness and surface area from conventional T1 weighted MRI (T1w) have been the primary measurements for characterizing the maturation of infant cortex. However, T1w MRI cannot reveal information about the complex microstructural changes inside the cortical mantle. Cortical microstructure, associated with the underlying cellular and molecular processes, plays a vital role in neuronal circuit formation and emergence of brain functions. We have recently developed a novel protocol to quantify the cortical microstructure using advanced diffusion MRI (dMRI), including diffusion tensor imaging (DTI) and diffusion kurtosis imaging (DKI). These advanced dMRI-based cortical microstructural measurements are sensitive imaging markers for the microstructural differentiation that characterizes the infant cortex maturation. We hypothesize that dMRI-based cortical microstructure measurements in early infancy could be potential biomarkers for early detection of ASD in high-risk infants. The goal of this study is to develop and test a novel protocol, using dMRI-based cortical microstructure feature, to reliably predict clinical score of infants at high risk for ASD and other brain disorders in general. We aim: 1) to develop a technique that incorporates dMRI-based cortical microstructural measures at early infancy with cutting-edge multi-kernel machine learning algorithms to reliably predict infants’ future neurodevelopmental outcomes at 2 years of age; 2) to demonstrate the initial clinical utility of the technique in infants at high-risk for ASD. To accomplish these goals, we leverage a large cohort of infant study that consists 100 TD infants who longitudinally undergo multi- modal MRI scans at early infancy and neuropsychological testing at 2 years for developing the prediction techniques, and a high-risk ASD study that allows us to longitudinally image the infants during infancy for testing our developed techniques. Finally, it should be emphasized that, although the present project focuses on its clinical applications in infants at high risk for ASD, the method ...