Project Summary Alzheimer’s disease (AD) is a heterogeneous neurodegenerative disorder, not only in pathophysiology, but also at different disease progression stages. Despite numerous studies that have investigated the clinical utility of magnetic resonance imaging (MRI) based biomarkers in characterizing AD stages from asymptomatic to mildly symptomatic to dementia, making a personalized precision prediction and early diagnosis of AD is still challenging. Existing imaging biomarkers are limited in representing significant heterogeneity across different individuals and at different clinical stages. This challenge originates from the lack of reliable brain landmarks that can simultaneously characterize and represent robust population correspondences and individual variation during normal aging and AD progression. In response, this project aims to: 1) Identify a set of brain anchor- nodes as population landmarks based on both group-wise consistent patterns and individualized anatomical and connectivity properties during normal aging and AD progression among massive, publicly available neuroimaging data sources; 2) Develop an efficient individualized shape transformation approach based on deep learning to map population anchor-nodes to individual brains by flexibly leveraging multimodal individual features; and 3) Construct a progression tree using anchor-nodes derived brain measures to unveil and represent the wide spectrum of AD development. Individual subjects can thus be projected to the tree structure to effectively and conveniently access their clinical status and predict the trend of AD progression. We will test our new frameworks on four large independent aging/AD cohorts including HCP-Aging, UK Biobank, ADNI and the latest stage of Open Access Series of Imaging Studies (OASIS-3), and freely release our computational tools and processed data to the public.