A surface-constrained volumetric alignment method for image atlasing

In this paper, we propose a prototype system capable of incorporating 3D shape information with conventional TPS-based (thin-plate-spline) volumetric registration method for image atlasing. Our method consists of two phases. The former phase registers and warps the 3D mesh surface models describing the tissue shape boundary of the input image volumes, and the latter aims to align the input image volumes with the aid of the boundary constraints suggested by the former. The proposed volumetric registration method is driven and constrained by the pre-registered 3D mesh surface model. Experiments show that using our framework for volumetric image registration and warping obtains a performance comparable to or better than a well-known benchmark method.