Salient feature volume and its application in brain MRI image registration

Brain MRI image registration is crucial for the diagnoses and treatments of various brain disease. Moreover, it is a fundamental procedure during the functional analysis of brain. Currently, brain MRI image registration based on volumetric iterative search of the best transformation parameters is still a little time-consuming. In this paper, we propose a new method based on the salient feature volume (SFV) to register the brain MRI images. First, an innovative saliency definition of brain MRI image is given to extract the SFVs. For extracted SFVs, we compute corresponding feature descriptor, which consists of both local gradient field distribution and geometric information. The global rigid transformation parameters are then estimated by matching these extracted SFVs. Finally, a local to global registration strategy is adopted. The proposed method is evaluated by detailed experiments, and the experimental results prove that our method could accelerate the brain MRI image registration efficiently.