Symmetric image normalization for mouse brain magnetic resonance microscopy

Various genetic mouse models have been used to understand aspects of the biology of the neurodegenerative disease. A rapid growth of data collection from the mouse brain has put image registration, a key prerequisite step for brain image analysis in great focus. SyN (symmetric image normalization method) is one of the most performed diffeomorphic strategies which has been widely used in human brain mapping. In this study, we optimized the SyN strategy for tiny mouse brain. The optimized protocol is able to accurately remove the anatomical variability between mouse brain MR microscopy (MRM) and offers superior performance over the SyN protocol for human. Introduction A rapid growth of brain image collection from the mice has taken place recently, strengthened by the extensive use of genetic mice for various types of brain research [1]. Inter-subject image registration technique is one of the most significant prerequisite steps that establishes a one-to-one correspondence between two brains. It is well established that linear registration is inadequate for inter-subject registration, especially between normal and lesioned or diseased mice. Various automated nonlinear algorithms have been proposed for human study, such as HAMMER [2], SPM5 [3], SPM-DARTEL [4], SPM-Shooting [5], Demons [6], FFD [7], SyN [8], et al. The performance of those methods have also been evaluated [9,10]. However, most of those methods are specially designed or optimized for human brain, and special tuning is necessary to make them work for tiny mouse brain. The animal experiment plays an irreplaceable role in medical research, because of it maneuverability, easier usability, et al. At present, the MRI has been widely used in clinical. As the technology innovation, MR microscopy (MRM) which can be used for studying mouse brain has “borned”. Because of its 3D characteristic and high resolution, it is possible to analysis the mechanism of the structures and functions of mouse brain. SyN is a kind of registration algorithms based on ANTs(Advanced Neuroimaging Tools) [11], which have showed great performance on human brain mapping(http://stnava.github.io/ANTs/). SyN has showed some advantages compared with the other algorithms mentioned above. The Exponential Mapping approach as SPM-DARTEL is achieved by controlling the velocity field time-invariant. The algorithms based on elastic or related such as HAMMER, Demons, FFD,SPM5 can’t keep topology persevering. SyN is a symmetric diffeomorphic normalization, which is more robust than SPM-Shooting that is only symmetric in theory. In this paper, we first briefly introduced the SyN registration approach; then discussed how we optimized it for mouse brain; and finally, we evaluated the performance of the optimized protocol.