Segmentation of Kidneys from MR-Images

by Ree, Eirik

Abstract (Summary)
Haukeland University Hospital are currently running a project for developing renal examinations using MR imaging. For many of the tasks they wish to do, i.e. visualization and volume estimation, a good segmentation of the kidneys is absolutely necessary. Most methods available today are time-consuming and labor intensive.In this paper, we have studied several possible methods for automatic or semiautomatic segmentation of the kidneys. A method has been implemented by combining watershed segmentation with active contours.The watershed algorithm is a watershed-from-markers using an Image Foresting Transform as described in [20][21]. This is simple to initialize by just setting a few marker points inside and outside of the desired region. We have simplified this initialization further, so the user only has to draw a rectangle around the kidney.The results from the watershed algorithm are used for initializing the active contours. Because watershed may result in several regions in each image, while the desired result from the active contours should be a single region, the contour is initialized as a rectangle that bounds the watershed result. This contour then shrinks until it is a smooth outline that contains the result from the watershed.The implemented algorithm has been tested with real MR images, and the results show that the method gives a good solution to the problem.
Bibliographical Information:


School:Norges teknisk-naturvitenskaplige universitet

School Location:Norway

Source Type:Master's Thesis



Date of Publication:01/01/2005

© 2009 All Rights Reserved.