Project: TR&D 1 (Data Science)

University Of Southern California, Los Angeles CA

Investigators

Linked publications & trials

Abstract

This Project is focused on methodological developments for the analysis of brain imagery in which we coregister brains to a common coordinate system so as to precisely align cortical surface anatomy as well as subcortical structures. While automated methods that perform segmentation, labeling, and coregistration of brain imaging data are now in routine use, they frequently fail when applied to imperfect data or to images with pathology. This limits the scope of their application, increases manual labor requirements, and may reduce the accuracy of scientific findings. To address these issues, we will develop methods to identify errors as they occur during the processing chain, and when possible, automatically adapt parameters to resolve these errors. We will also adapt our existing methods to be more robust to the imaging artifacts that can occur during routine acquisition of imaging data, and we will enhance our software to facilitate analysis of clinical data with pathology. Finally, we will extend our work on segmentation and cortically-constrained registration of structural brain MRI to include diffusion images, thus enabling inter-subject registration of high dimensional diffusion data. This Technology Research and Development (TR&D) Project will produce and support open-source, publicly available software that will have applications in the methods described in TR&D Projects 2 and 3 and facilitate the scientific research described in the Driving Biological Projects (DBPs) and Collaboration and Service (C&S) Projects. In particular we will work closely with Dr. Damasio at USC in DBP6 on the use of the methods described here for analysis of normal and lesioned brains for investigation of the anatomic substrates of complex human behavior. Similarly we will collaborate with Dr. Machado at the Cleveland Clinic (CSS) on the application of our methods in his work on deep brain stimulation. Dr. Pantazis (CS7) at MIT on the application of our methods to structural image analysis in conjunction with multivariate analysis of cognitive studies using EEG/MEG, and Dr. Klein at Columbia (086) on new approaches to validation of image registration and segmentation algorithms.