Localization of 3D anatomical point landmarks in 3D tomographic images using deformable models

Abstract

Existing differential approaches to the localization of 3D anatomical point landmarks in 3D tomographic images are relatively sensitive to noise as well as to small intensity variations, both of which result in false detections as well as affect the localization accuracy. In this paper, we introduce a new approach to 3D landmark localization based on deformable models, which takes into account more global image information in comparison to differential approaches. To model the surface at a landmark, we use quadric surfaces combined with global deformations. The models are fitted to the image data by optimizing an edge-based fitting measure that incorporates the strength as well as the direction of the intensity variations. Initial values for the model parameters are determined by a semi-automatic differential approach. We obtain accurate estimates of the 3D landmark positions directly from the fitted model parameters. Experimental results of applying our new approach to 3D tomographic images of the human head are presented. In comparison to a pure differential approach to landmark localization, the localization accuracy is significantly improved and also the number of false detections is reduced.