An intelligent tool for anatomical object segmentation using deformable surfaces

Image segmentation is a very active area of research in machine vision. In this work, an innovative methodology is presented that allows the segmentation of objects in three-dimensional images with initial user intervention. The paper describes the adopted approach for implementing the algorithm of deformable / active surfaces (AS), using the explicit scheme for numerical evaluation of the partial derivative equation of the AS evolution. Both the Vector Field Convolution (VFC) and the Gradient Vector Flow (GVF) image dynamic field are investigated for 3D segmentation using the AS. The proposed methodology is implemented as software tool, which allows the initialization of AS using cylinder-like surfaces with user intervention. Initial results are provided for the case of three-dimensional synthetic data and clinical Computed Tomography (CT) images, in terms of segmentation accuracy and speed of convergence.