Talk by Vedrana Andersen Dahl, Technical University of Denmark

Computing Segmentation Directly from Projection Data using Parametric Deformable Curve

Abstract

Processing of projection data usually starts by computing a reconstructed image, often followed by a segmentation step. Combining reconstruction and segmentation leads to methods that compute a segmented reconstruction directly from the projection data. This is useful when the object under study consists of a number of domains with approximately homogeneous absorption coefficients. In the presented work, we use a parametric deformable curve to obtain such segmentation. Unlike similar approaches based on level sets, we never consider a pixel representation of the scanned object. Our current implementation uses a simple closed curve and is capable of separating one object from the background. However, our basic algorithm can be applied to an arbitrary topology and multiple domains of different absorption coefficients. We demonstrate excellent performance under high noise and a small number of projections.