Landmark-based Elastic Matching of Tomographic Images

We describe a novel approach for elastic matching of tomographic images of the human brain. The approach is based on a set of corresponding anatomical point landmarks and uses approximating thin-plate splines. Previous work on this topic has concentrated on using interpolation schemes which means that corresponding landmarks are exactly matched to each other. Consequently, the underlying assumption is that the landmark positions are known exactly. However, in real applications the localization of landmarks is always prone to error. Therefore, to cope with these errors, we have investigated the application of an approximation scheme which is based on a minimizing functional. This approach renders possible to individually weight the landmarks according to their localization uncertainty. Our approach has been applied to 2D as well as 3D tomographic data.