Modeling anatomical heterogeneity in populations

Our goal is to model anatomical variability across individuals, which presents substantial challenges in clinical population studies and in building atlases for segmentation. Based on a mixture model for a population, we derive an efficient algorithm that clusters a set of images while co-registering them into a common coordinate frame. The output of the algorithm is a small number of template images that represent different modes of a population. This is in contrast to traditional computational anatomy methods that assume a single template for population modeling. The experimental results demonstrate the promise of our approach for statistical analysis in clinical studies of anatomy.