A combined approach to 3D medical image segmentation using marker-based watersheds and active contours: the active watershed method

Active contours have been used extensively for medical image registration as they display good smoothing properties which allow to account for the discrete nature of the image and noise from the acquisition process. However, contours are generally difficult to initialise around the region of interest (ROI). The markerbased watershed segmentation can segment unique boundaries from an image or stack of images, however it has no smoothing/generalisation properties. Combining the two approaches results in a segmentation method which both solves the contour initialisation and generalisation problem. In this paper we briefly explain the marker-based watershed and active contour segmentation and the algorithm to convert watershed contours into active contours. We then illustrate the approach on the segmentation of abdominal organs from MR images and bony structures from CT images.