Wavelet-driven knowledge-based MRI calf muscle segmentation

Abstract

We propose a novel representation of shape variation using diffusion wavelets, and a search paradigm based on local features. The representation can reflect arbitrary and continuous interdependencies in the training data. In contrast to state-of-the-art methods our approach during the learning stage optimizes the coefficients as well as the number and the position of landmarks using geometric constraints. During the learning stage the approach obtains a landmark shape model, based on diffusion maps. For the model search we apply an approach related to active feature models; the location of landmarks is updated iteratively, using local features, and the canonical correlation analysis. The resulting search is independent from the topology of the anatomical structure, and can represent complex geometric and photometric dependencies of the structure of interest. We report promising results on challenging medical data sets of T1 MRI full calf muscles.