A New Algorithm for Inverse Consistent Image Registration

This paper presents a novel variational model for inverse consistent deformable image registration. This model deforms the source and target image simultaneously, and aligns the deformed source and deformed target images in the way that the both transformations are inverse consistent. The model does not computes the inverse transforms explicitly, alternatively it finds two more deformation fields satisfying the invertibility constraints. Moreover, to improve the robustness of the model to noises and the choice of parameters, the dissimilarity measure is derived from the likelihood estimation of the residue image. The proposed model is formulated as an energy minimization problem, which involves the regularization for four deformation fields, the dissimilarity measure of the deformed source and deformed target images, and the invertibility constraints. The experimental results on clinical data indicate the efficiency of the proposed method, and improvements in robustness, accuracy and inverse consistency.