3D model-based approach to identification of laminar structures of the cerebral cortex: Application to Brodmann areas 17 and 18

The human cerebral cortex may be subdivided into architectonic fields according to variations within its laminar structure. Studies have shown correspondences between the locations of functional activation foci and architectonic regions. In order to perform accurate localization of functional activation foci to architectonic regions, a parcellation algorithm capable of segmenting architectonic regions on in vivo imaging datasets is required. This paper presents a novel 3D model-based approach to directly detect cortical layers and classify architectonic fields. The column-like structure of the cortex is modeled using a Laplace equation method which generates a collection of intensity profiles that span the cortical mantle. Bayesian evidence for intensity profile elements belonging to hyper- or hypo-intense bands, which represent cell or myelin poor or rich layers in imaging data, is gathered. A non-isotropic Markov Random Field model is used to encourage contiguous bands as well as a penalty term that completes bands across highly curved cortical regions where neighbouring evidence for banding is strong. This algorithm is validated on a 3D histological dataset of a macaque brain with visible layering at intermediate resolution between high-resolution MRI and histology. The algorithm detects the myelin-rich Stria of Gennari and uses this as the basis for finding the Brodmann Area 17/18 boundary.