A semi-automated technique for vertebrae detection and segmentation from CT images of spine

Abstract

Spine or backbone forms a supportive structure for all vertebrates, which is composed of complex bones known as vertebrae. Spine related pathologies are common and they are analyzed with help of various medical imaging techniques. Thus the detection and segmentation of vertebrae is gaining prime importance, as it can be used for the clinical analysis of spine. There are many semi-automated and automated approaches which are used to detect and segment individual vertebra from computed tomography (CT) images. But most of these approaches do not provide satisfactory results due to the anatomical similarities between the adjacent vertebrae and the effect of underlying artifacts. So to overcome these problems, a framework for individual vertebrae segmentation from CT images of spine is put forward. First of all, the vertebral body is identified with the help of iterative Normalized-cut algorithm which uses eigenvalue decomposition for the detection procedure and the segmentation of individual vertebra is done using region based active contour method.