The Volume Radius Function: A new descriptor for the segmentation of volumetric medical images
Ghassen Selmi, Zouhour Ben Azouz, Dhafer Malouche · 2015
In this paper, we consider the problem of segmenting volumetric medical images. Our non-supervised approach uses a tetrahedral mesh representation of volumetric data. The segmentation is formulated as a non-linear optimization of an energy function defined over a graph. The main contribution of the proposed approach is the introduction of a novel geometric descriptor, called the Volume Radius Function (VRF). The new descriptor is a scalar function representing the position of a given point within the tetrahedral mesh relatively to the image boundaries. The VRF is used to compute the energy function during the segmentation step. To demonstrate the efficiency of our approach, we applied our approach on different medical images. The obtained results show the discriminative power of our descriptor.