Detecting and localizing tree nodes in anatomic structures of branching topology

Abstract

Although several algorithms for corner detection have been proposed, limited effort has been devoted to the detection of branching points, critical points of anatomic tree structures which are located inside the tree where a branch starts splitting into two thinner branches. In this paper, we propose a novel two-step approach for identifying and inferring the location of tree nodes in medical images which visualize structures of the human body of branching topology. Our methodology begins with corner detection using adaptive thresholding in order to accommodate uneven illumination across tree levels. Furthermore, a localization procedure using the k-means algorithm finds accurate locations of branching nodes. The proposed methodology is applied to a dataset of clinical X-ray galactograms and is evaluated by means of precision, recall and localization error. Despite the challenges of this kind of medical images, the experimental results demonstrate the effectiveness of the new framework compared to state-of-the-art approaches. © 2013 Springer-Verlag.