Several neuroimaging modalities such as magnetic resonance imaging (MRI) are becoming increasingly important for diagnostic classification, prognostic evaluation and tracking treatment response in patients with brain disorders. They provide detailed, quantitative information about brain structure and function, and aspects of ongoing disease processes. Measures derived from these neuroimaging modalities have been studied through automated algorithms as potential predictors of disease outcomes. This proposal aims to introduce and evaluate new computerized algorithms, based on the field of machine learning that would incorporate not only brain imaging measures, but also biochemical and genetic information to create more powerful predictions of diagnostic and prognostic outcomes. Such novel, automated, multimodal predictors, we propose, may have important applications in future clinical decision making and clinical trial design. Brain imaging offers new quantitative measures that may be closer than cognitive assessments to the underlying biological mechanisms that lead to disease. By studying the associations of genetic factors with phenotypes based on cutting edge imaging techniques such as diffusion tensor imaging (DTI), we plan to examine mechanistically meaningful genetic contributions to brain disorders. The rapidly expanding field of neuroimaging genetics will provide the nexus for the applicant's intensive training in the world-class imaging and genetics programs at the UCLA School of Medicine. This proposal will introduce new automated algorithms for gene discovery and risk prediction into the field of neuroimaging genetics. Algorithms that consider multiple genetic variants jointly, we propose, are likely to (1) more powerfully detect new gene effects on brain images, and (2) identify profiles of candidate genetic variants to assist prediction of an individual's brain integrity and risk for disease.
Neuropsychiatric disorders are the leading causes of disability across the world. By developing new computerized models for the prediction of clinical outcomes based on disease-specific neuroimaging, biochemical and genetic biomarkers as well as the early prediction of white matter integrity based on genetic profiles, we hope to pave the way to personalized prevention and management of brain disorders.
Showing the most recent 10 out of 15 publications
