Interactive medical image sequences segmentation based on non-rigid registration
Hongbo Chen · Computer Engineering and Applications Journal · 2007
This paper presents a new algorithm for medical image sequences segmentation,which converts image segmentation into image registration by combining non-rigid registration technology with prior knowledge about anatomy.Firstly,we select Demons algorithm to register image,adopt optical flow method to compute instantaneous displacement,design a new stop criterion for Demons algorithm and implement it in a pyramid multi-scale framework.Then,the deformation fields obtained from image registration are used to wrap the prior image that has been segmented.So the segmented result is obtained.Repeating this process,the segmented results of the whole image sequence can be achieved.Experimental results demonstrate the new algorithm can rapidly get accurate results with few interventions.