A dynamic tree-based registration could handle possible large deformations among MR brain images

• A dynamic tree-based strategy for effective large-deformation registration in multi-atlas segmentation.
• The node links of the tree may change to reduce local minima and thus registration error.
• Registration is vital to multi-atlas segmentation and can greatly improve segmentation accuracy when performed accurately.

Multi-atlas segmentation is a powerful approach to automated anatomy delineation via fusing label information from a set of spatially normalized atlases. For simplicity, many existing methods perform pairwise image registration, leading to inaccurate segmentation especially when shape variation is large. In this paper, we propose a dynamic tree-based strategy for effective large-deformation registration and multi-atlas segmentation. To deal with local minima caused by large shape variation, coarse estimates of deformations are first obtained via alignment of automatically localized landmark points. The dynamic tree capturing the structural relationships between images is then employed to further reduce misalignment errors. Evaluation based on two real human brain datasets, ADNI and LPBA40, shows that our method significantly improves registration and segmentation accuracy.