PROJECT SUMMARY Modern multimodal biomedical imaging data have the potential to advance our ability to diagnose medical conditions and to understand the biological mechanisms underlying disease progression. However, these data typically display non-linear geometric structure, i.e., manifold structure, which limits the applicability of classical statistical methods to gain further knowledge from the analysis of the contemporary biomedical datasets. This project will focus on the development of novel statistical methods for the analysis of biomedical images with manifold structure that are subject- specific anatomical objects coupled with ‘signals' that are structural or functional images. Examples of such data are fMRI signals, seed-based connectivity maps, or cortical thickness measurements located on the highly convoluted subject-specific cortical surfaces. The proposed methods will model these data as ‘functional data', i.e., without relying on oversimplified representations that could lead to the loss of relevant biological information. In practice, the proposed framework will allow researchers to relate anatomical, structural, and functional imaging features to other variables typically collected, such as disease status, treatment type, or genetic information, with the aim of validating scientific hypotheses or discovering novel imaging biomarkers. The models developed will be made available as free and open-source tools that can easily interface with the most popular data analysis software for them to become part of the larger imaging software ecosystem.