M. Takahashi and M. Nakano (Japan)
Contour extraction, Segmentation, Snakes, Nucleus, Histopathology, Liver
A novel segmentation method, which we call ‘Radial Snakes’, for identifying nuclei in histopathologic images is proposed. The basic idea is based on the fact that the shapes of most nuclei are convex, i.e., circular or elliptical. Using this idea, the method becomes one of solving a combinatorial optimization problem to select one contour point from a limited number of candidates for about 15 to 20 groups. As a result, the search space of Snakes can be drastically reduced. The simulation results using hepatocellular images were promising. Most of the unclear contours that were difficult to extract using the conventional Snakes could be successfully extracted with Radial Snakes, giving a 90% correct ratio compared with the 40% of conventional Snakes. The average iteration cycles for convergence was only 17, which corresponds to a computation time of 10 ms on a Pentium III 933 MHz. The method can be applied to other time-consuming segmentation problems involving convex-shaped objects, such as 3D segmentation of organs.
Important Links:
Go Back
