Geometric Strategies for Neuroanatomic Analysis from MRI
James Duncan
Yale University
Electrical Engineering
Quantitative analysis of brain structure is important in the study of many neurological and neuropsychiatric disorders. Furthermore, accurate delineation of structure can provide important baseline information for quantifying brain function. This talk presents a body of work grounded in the use of geometrical constraints and mathematical optimization to analyze the neuroanatomical structure of the human brain from Magnetic Resonance Images (MRI).
More specifically, we describe our applied mathematical approaches to the segmentation of cortical and subcortical structures, the analysis of white matter fiber tracts using diffusion tensor imaging (DTI), and the intersubject registration of neuroanatomical (aMRI) datasets. Many of our methods center on the use of geometric constraints and level set evolution strategies. In addition, we are currently integrating these methods with functional MRI (fMRI) analysis strategies with the goal of developing a more fully integrated approach to functional and structural brain image analysis.
The analysis of gray matter structures and connecting white matter pathways, combined with the ability to bring all information into a common space via intersubject registration, provides a rich set of data for investigating structure and variation in the normal, abnormal, and developing human brain. This framework also serves as a basis for ongoing work in the development of integrated brain function and structure analysis.
An accompanying example illustrates our efforts in segmenting the amygdala and hippocampus from T1-weighted MRI data. These subcortical structures are important in the study of autism and other disorders. The approach is based on a three-dimensional level set parametrization of object surfaces and uses inter-object statistical priors to constrain the solution. The sequence demonstrates three stages of the simultaneous evolution of four zero level sets (left and right hippocampus, left and right amygdala), shown both overlaid on the image data and in three-dimensional visualization.
Joint work with Xenios Papademetris, Jing Yang, Marcel Jackowski, Xiaolan Zeng, and Lawrence Staib (Yale University, Biomedical Engineering, Diagnostic Radiology, and Electrical Engineering).