Computer assisted medical systems often require segmentation of 3D medical image data for efficient visualization or data analysis. So various 3D-segmantation methods have been developed recently, but it is very difficult to automatically extract exact boundaries because of the property of medical images, that is the boundaries of organ are indistinct and complex. In this paper, we propose an efficient semi-automatic algorithm to segment a 3D object from the given segmentation result of a single slice, based on the assumption that the region to be segmented is homogeneous and has discernable boundaries. The proposed algorithm is composed of 3 steps. First, we estimate a parametric motion model of an organ from the previous slice to the current slice, and find an estimated boundary of the organ by projecting the previous result. After that, 3 kinds of seeds are extracted by using the projection result and the pixel values of current slice. Finally, all seeds are grown to produce true boundaries of the organ. The proposed algorithm provides satisfactory results in segmenting kidneys from 82 slices of X-ray CT images.
