VALIDATION OF A NEW OPTIMISATION ALGORITHM FOR REGISTRATION TASKS IN MEDICAL IMAGING

A number of problems frequently encountered in brain image analysis can be conveniently solved within a registration framework, such as alignment of mono- or multi-sequence magnetic resonance images (MRI) for single or multiple subjects, computation of the cerebral mid-sagittal plane in anatomical or diffusion-tensor MRI, correction of acquisition distortions in diffusion-weighted MRI, etc. A widely used approach for registration tasks consists of maximising a similarity criterion between the intensities of the images to be matched. In this context, efficient optimisation methods are needed to obtain good performances. In this paper, we introduce a new optimisation algorithm (called NEWUOA) to address the above registration problems, and we demonstrate its robustness and accuracy properties