New Interactive Methods for Tubular Structure Segmentation on Medical Images

In this paper we present two interactive methods for tubular structure fast extraction. The main application and motivation for this work is vessel tracking in 2D and 3D medical images. The basic tools are minimal paths solved using the fast marching algorithm. This allows interactive tools for the physician by clicking on a small number of points in order to obtain a minimal path between two points or a set of paths in the case of a tree structure. Visualization tools are very important in order to choose points and show paths and vessels in the case of 3D images. We show variants of the minimal path method that differ both on the model used (path on the image domain or centerline and surface by adding one dimension for the local radius around the centerline), the way the endpoints are chosen by the user (two points or automatic detection of key points from a single click) and the definition of the local metrics to minimize (based on gray level or also on the local orientation using a Riemannian metrics). The algorithms were made available for the physician by use of a new programming framework called CreaTools. This software suite was used to build a Graphical User Interface (GUI) and to connect the different processing methods. The GUI provides the necessary visualization and interaction for parameter tuning, source point definition and analysis of the outcome. We show results on various CT images.