PROJECT SUMMARY We propose to develop statistical methods for the analysis of the local covariance structure between measurements in multimodal biomedical imaging studies. Despite the rise of multi-modal imaging studies and the proliferation of derived measures available from each modality, very little work has explored relationships between modalities. Notably, brain structure and function evolve dramatically in the context of adolescent brain development, but typical analyses usually evaluate each imaging phenotype separately. We propose to develop tools for measuring relationships between brain imaging phenotypes provided by disparate imaging modalities. Covariance structures at the subject level will allow identification of local relationships between measures, through weighted regressions and singular value decomposition-based techniques. These techniques will be extended to accommodate any number of imaging modalities and incorporate repeated measures. We will use extensive test-retest data provided by the Consortium of Reproducibility and Reliability (CORR) for validation, and to identify coupling measures that are particularly reliable. The rich multi-modal imaging data and associated clinical phenotypes from the Philadelphia Neurodevelopmental Cohort (PNC) will allow us to delineate how inter- modal coupling evolves in normal brain development, and also establish how such relationships are altered in association with psychosis-spectrum symptoms. This proposal builds upon established collaboration between the PIs and capitalizes on their complimentary expertise in imaging statistics, developmental neuroimaging, psychopathology, and intimate familiarity with the PNC dataset. Upon completion, this project will provide powerful new tools for the neuroscience community, as well as novel insights regarding brain development and early psychosis-spectrum symptoms.