A model-based vertebral segmentation method using GVF and ASM

In this paper, we describe a new image segmentation technique applied to vertebral segmentation in medical X-ray images. We propose a combination of two kinds of deformable models based approaches, which are Gradient Vector Flow (GVF) and Active Shape Models (ASM). GVF is a well known external force field increasing the capture range and concavities detection of active contour models. However, the definition of this external force field is not always sufficiently accurate to determine vertebrae contours. In some cases, it leads to erroneous contours which means further human post-processing. To solve this problem, we propose to use active shape models in order to initialize the deformable GVF model and to control the iterative deformable process. The principal idea is to start from a vertebral model and to deform it using the GVF force field while the model fit to certain knowledge of the vertebrae. This knowledge will be brought through statistical shape models to form a GVF controlled by Active Shape Model.