No abstract

Abstract. This paper presents a prototype segmentation system for three-dimensional ultrasound data. 3D ultrasound is cheap and noninvasive but the data has a low signal-to-noise ratio and contains artifacts. To overcome these difficulties we have developed a system which uses a prior model, initialised by a clinician, to provide the starting point for a data-driven segmentation algorithm based on active contours. Results are presented showing how the technique can facilitate the segmentation of a gall-bladder.

No abstract