The method we have developed to extract and label from 3-D MRI the cortical topography is divided in three steps : a) extraction of the exterior trace of the sulci and detection of the connections between these entities. At this level, we extract the superficial structural component of the cortical topography ; b) modelling of the deeper cortical infolding by tracking the median surface of each cortical fold ( fig 1.a). As a result, we obtain a parametric surface representing the 3-D shape of the buried part of each sulcus [2]. Finally, a set of parameters are associated with each extracted fold and junction, and the result is encoded in a graph structure. c) labelling of these graphs were then performed by an anatomist through an iteractive display interface for 51 subjects in order to study the spatial distributions of the cortical sulci. Data expressed with respect to a standardized brain coordinate system (i.e Talairach-like coordinate system) permits the creation of SP_AMs (Statistical Probability Anatomy Map)[1] of the main sulci (fig 1.b). These SP_AMs, representing the normal spatial variability of the neuroanatomy are then used to perform the recognition of these anatomical structures on a new subject (fig 1.c)
 |