Statistical modeling of pairs of sulci in the context of neuroimaging probabilistic atlas
Résumé
In the context of neuroimaging probabilistic atlases, we propose a statistical framework to model the inter-individual variability of pairs of sulci with respect to their relative position and orientation. The approach extends previous work [3], and relies on the statistical analysis of a training set. We first define an appropriate data representation, through an observation vector, in order to build a consistent training population, on which we then apply a normed principal components analysis (normed-PCA). Experiments have been performed on pairs of major sulci extracted from 18 MR images.