Abstract
Four-Dimensional (4D) image has broadened 3D image information by embedding temporal information, which makes it possible to investigate the dynamic anatomy and pathology of patients. This paper aims to provide an overview of the transformation models for 4D image registration. We classify 4D image transformations into the spatial and spatio-temporal categories according to whether temporal information is included in the transformation. A spatial transformation estimates the full displacement field for each image in the 3D image sequence using only isolated 3D volumes, which is applicable to estimate motion model of organs using one sequence for a patient. Spatio-temporal transformations take advantage of the temporal information contained in image sequences, map the corresponding objects and exploit the temporal coherence of the two image sequences. Spatio-temporal transformation models align not only the spatial but also the temporal displacement in the motion cycle between subjects. This paper focuses on how temporal information is utilized in the 4D transformation models. We classify existing 4D transformation models into following categories: free-form deformation (FFD) based on the B-spline, the radial basis functions (RBFs) expansions, optical flow-based methods, and continuum biomechanics. This paper compares these transformation models and illustrates their advantages and disadvantages for 4D image registration. Finally, evaluation methods for 4D image registration are also described.
Keywords: 4D image, free-form deformation (FFD), image registration, spatial transformation, spatio-temporal transformation, transformation models.
30