Diffeomorphic demons: Efficient non-parametric image registration. Tom Vercauteren, Xavier Pennec, Aymeric Perchant, Nicholas Ayache. NeuroImage, 2009, 45 (1), Supplement 1, Pages S61-S72. 10.1016/j.neuroimage.2008.10.040. This paper presents an efficient non-parametric diffeomorphic image registration algorithm based on Thirion's demons algorithm. The authors show that Thirion's demons algorithm can be viewed as an optimization procedure on the entire space of displacement fields. They provide strong theoretical justification for the different variants of the demons algorithm, predicting a theoretical advantage for the symmetric forces variant. Experiments confirm that this variant converges faster in practice. The authors adapt the demons algorithm to a space of diffeomorphic transformations, which are computationally efficient as they only replace an addition of displacement fields by a few compositions. The algorithm is tested on both simulated and realistic registration setups, showing that it provides results similar to the demons algorithm but with smoother transformations and closer to the gold standard in terms of Jacobians. The algorithm is also compared to other approaches, showing that it is more efficient and provides better results. The authors also show that the algorithm can be extended to handle different types of images and to be symmetric with respect to the order of the input images. The algorithm is computationally efficient and can register typical 3D MR images in less than five minutes on a desktop computer. The paper also discusses the theoretical and practical properties of the algorithm and shows that it is a good working framework for image registration when no additional information about the spatial transformation is available. The algorithm is based on the Lie group structure on diffeomorphic transformations and uses a fast vector field exponential algorithm to compute the exponential map. The algorithm is tested on a variety of experiments, showing that it provides better results than other approaches in terms of convergence, smoothness, and accuracy. The authors conclude that the diffeomorphic demons algorithm is a powerful and efficient method for image registration.Diffeomorphic demons: Efficient non-parametric image registration. Tom Vercauteren, Xavier Pennec, Aymeric Perchant, Nicholas Ayache. NeuroImage, 2009, 45 (1), Supplement 1, Pages S61-S72. 10.1016/j.neuroimage.2008.10.040. This paper presents an efficient non-parametric diffeomorphic image registration algorithm based on Thirion's demons algorithm. The authors show that Thirion's demons algorithm can be viewed as an optimization procedure on the entire space of displacement fields. They provide strong theoretical justification for the different variants of the demons algorithm, predicting a theoretical advantage for the symmetric forces variant. Experiments confirm that this variant converges faster in practice. The authors adapt the demons algorithm to a space of diffeomorphic transformations, which are computationally efficient as they only replace an addition of displacement fields by a few compositions. The algorithm is tested on both simulated and realistic registration setups, showing that it provides results similar to the demons algorithm but with smoother transformations and closer to the gold standard in terms of Jacobians. The algorithm is also compared to other approaches, showing that it is more efficient and provides better results. The authors also show that the algorithm can be extended to handle different types of images and to be symmetric with respect to the order of the input images. The algorithm is computationally efficient and can register typical 3D MR images in less than five minutes on a desktop computer. The paper also discusses the theoretical and practical properties of the algorithm and shows that it is a good working framework for image registration when no additional information about the spatial transformation is available. The algorithm is based on the Lie group structure on diffeomorphic transformations and uses a fast vector field exponential algorithm to compute the exponential map. The algorithm is tested on a variety of experiments, showing that it provides better results than other approaches in terms of convergence, smoothness, and accuracy. The authors conclude that the diffeomorphic demons algorithm is a powerful and efficient method for image registration.