|Stochastic Systems Group|
Prof. Bruce Fischl
Computational Core Director
MGH/MIT/HMS Athinoula A. Martinos Center for Biomedical Imaging
Assistant Professor, Harvard Medical School
Visiting Scientist, MIT AI Lab
Computational neuronanatomy can either refer the manner in which the computational architecture of the brain helps it to carry out computations, or the application of computational techniques to build models of neuroanatomical structures. While the former definition is the one that is of ultimate interest, it most likely requires the models indicated in the latter definition. In this talk I will discuss research at MGH with the goal of building such models. The neuronanatomical structures of interest can be broadly subdivided into two categories - cortical and non-cortical.
Cortical structures (particularly the cerbral cortex) are typically highly folded, thin sheets of gray matter. Functionally, the cerebral cortex has been shown to have a "columnar" architecture. For this reason, we construct surface-based models for analysis of cortical properties. The construction of such models is a difficult task due to the high degree of folding of the cortical manifold in conjunction with the limited (~ 1 mm) resolution of current neuroimaging technologies. While the geometry of the cortex is highly variable across subjects, the topology is fixed except in cases of pathology. I will demonstrate a technique for correcting the topology of the cortical models, while limiting the modifications to the small portion of the manifold in which the "defect" occurs. I will then discuss various surface deformations we apply to the cortical models for visualization and computational purposes (e.g. flattening, parcellation, inter-subject registration).
A different set of techniques have been developed for the construction of models of subcortical structures. Here, we model the segmentation as an anisotropic nonstationary Markov Random Field. The anisotropy lets us model the local spatial relationships that exist between neuroanatomical structures (e.g. hippocampus is posterior and inferior to amygdala), while the nonstationarity facilitates the encoding of inhomogeneous properties of the tissue within a structure. This approach is based on extracting the relevant model parameters from a manually labeled training set, and has been shown to be comparable in accuracy to the manual labeling. Finally, I will discuss the incorporation of the physics of the imaging process in order to aid in the construction of accurate models that are independent of the hardware and software used in the acquisition of the underlying imaging data.
Problems with this site should be emailed to email@example.com