Knowledge-based segmentation of spine and ribs from bone scintigraphy

Abstract

We propose a novel method for the segmentation of spine and ribs from posterior whole-body bone scintigraphy images. The knowledge-based method is first applied to determine the thoracic region. An adaptive thresholding method is then used to extract the thoracic spine from it. The rib segmentation algorithm is carried out in two steps. First, the rib skeleton is extracted based on standard template and image information. The skeleton is then used to locate the accurate boundary of the respective ribs. The introduction of standard template can deal with significant variations among different patients well, while the skeleton-based method is robust against the low contrast between the ribs and the adjacent intervals. The experiments show that our method is robust and accurate compared to existing methods. © 2011 Springer-Verlag.