A New Deformable Model Based on Level Sets for Medical Image Segmentation
Devappa Jayadevappa, S. Srinivas Kumar, D. S. Murty · 2009
This paper presents a new deformable model based on level sets for medical image segmentation which plays a pivotal role in medical diagnosis. The current popu lar Image segmentation deformable models such as Snakes, Geometric Active Contours, Gradient Vector Flow, Level sets and Variational Level sets have a limitation that the co nvergence of the contour towards the object boundary is slow and hence not suitable for real time medical diagnosis. To counte r this limitation we present an improved image segmentation algorithm which is computationally efficient and al so the proximity of the contour towards the object is high er compared to existing algorithms. A new speed term is introd uced in the evolution step of variational level set in order to speed up the convergence process. The variational level sets in images with intensity inhomogeneity, tend to be slower and prone to leakage of contour outside the object boundary. This is due to the selection of gradient information for the terminati on of convergence process. However, this limitation is ov ercome in the proposed algorithm by modifying the edge indicator function embedded with the speed term that optimizes the effective distance of the attractive force. Experimental resu lts are provided using real time medical images. Comparative tables and graphs highlighting the performance of various deformable models are also presented.